Electromagnetic Compatibility Between Power Networks and Telecommunications Networks, Knut J Ramleth

Contents

Feature Special

Guest editorial, Egil Hauger ...... 3 Travelling wave antennas, Knut Stokke ...... 63 Electromagnetic interference XDR/RPC without TCP/IP: client-server – physical mechanisms, Geir Løvnes ...... 5 communication with file exchange as transport mechanism, Per Sigmond ...... 67 Radio interference in the band 30 – 100 MHz – the international limits of interference seen Isolation levels in relational database from a radio-engineer’s point of view, management systems, Ole J Anfindsen Egil Hauger ...... 16 and Mark Hornick ...... 71 Design of electronic equipment with Mathematical methods and algorithms EMC in mind, Agnar Grødal ...... 22 in the network utilization planning tool RUGINETT, Ralph Lorentzen ...... 73 Electromagnetic compatibility between power networks and telecommunications networks, Knut J Ramleth ...... 30

EMP in telecommunications networks, Bjørn Fossum ...... 37 Telecom overhead cables as antennas for Status long wave radio signals, Arne Thomassen ...... 41 International research and standardization EMC-norms and regulations – the situation activities in telecommunication: Introduction, in Norway, Sverre Tannum ...... 43 Tom Handegård ...... 85 EMC measurements, Sverre Tannum ...... 46 Eurescom work on ATM Network Studies and the European ATM Pilot Network, Electrical medical equipment control Inge Svinnset ...... 86 and vigilance in Norway, Erik Fønstelien ...... 50 Terminal equipment and user aspects, Company EMC organization, Agnar Grødal .... 52 Trond Ulseth ...... 89 The Radio Interference Service in Norway Telecommunications Management Network, – a historic review, Sverre Tannum ...... 56 Ståle Wolland ...... 93

Editorial office: Telektronikk Telektronikk Telenor AS Telenor Research Volume 90 No. 4 - 1994 P.O. Box 83 ISSN 0085-7130 N-2007 Kjeller, Norway Editor: Editorial board: Ola Espvik Ole P Håkonsen, Senior Executive Vice President Tel. + 47 63 80 98 83 Karl Klingsheim, Vice President, Research Bjørn Løken, Vice President, Market and Product Strategies Status section editor: Tom Handegård Graphic design: Tel. + 47 63 80 98 00 Design Consult AS Editorial assistant: Layout and illustrations: Gunhild Luke Gunhild Luke, Britt Kjus, Åse Aardal Tel. + 47 63 80 91 52 Telenor Research

1 Guest editorial

BY EGIL HAUGER SENIOR ENGINEER, TELENOR RESEARCH

In 1895, when Marconi started sound of the first signal from the wireless communication in Italy, mobile phone in his pocket. he may have observed what we These dramatic situations have today define as EMC – electro- all been taken care of by the magnetic compatibility problems. manufacturers of safety equipment, so there is no need to Electromagnetic interference worry. (EMI) a hundred years ago came mainly from natural sources such As mentioned above, the use of as lightning. But the electrifi- personal radio equipment consti- cation of the railway in the 1920s tutes a more significant EMC and the introduction of electric problem. If you are working at tramways in the big cities caused your PC, the reason you do not electromagnetic compatibility receive the beep-beep from the problems to the new communi- pager carried in your belt may cation medium – wireless broad- be EMC problems. The same casting. thing may apply when your colleague in the next office uses So, as long as man has been his GSM telephone and it dis- using the electromagnetic turbs your conversation in the spectrum as a medium for com- old fashion fixed line telephone. munication, there have been The regulations authorities actu- EMC problems. What then high- ally assume a protection distance lights EMC problems today? The of 10 m from an electromagnetic answer may be found in the stat- interferer to a radio receiver. istics of the number of PCs, the The 10/1 distance ratio increases number of mobile telephones, the the interference by 20 dB! number of pagers, and the amount of equipment using the radio spectrum for communi- What about the legal aspect of EMC and regulations authori- cation. The IT revolution has had a tremendous impact on the ties? In Europe, the EU has been the driving force since the total pollution of electromagnetic energy in the radio communi- 89/336/EU EMC directive “Council Directive on the appro- cation part of the electromagnetic spectrum. ximation of the laws of the Member States relating to electromagnetic compatibility”. This directive was to be implemented The IT revolution together with the digitization of the telecom within the EU/EFTA area (18 countries) by January 1, 1992, network will bring the source of radio noise, the interferer, to with a transition period until December 31, 1992. That was the every home in the industrialized countries in the years to come. situation in 1989. In the period from 1989 to 1992 there were Looking back 10 years when the limits of interference were laid heavy protests from the European IT industry claiming that the down, the electromagnetic environment was quite different from time schedule for fulfilling the objectives was unrealistic and today. The protection distance of 10 m was realistic in those pointing at the lack of measurements facilities and admin- days, but what is the reality 10 years later, and what can be seen istrative precautions taken to ensure the implementation of the on the horizon from a radio engineer’s point of view. directive. The result was, as we all know, that the date of implementation was put off until December 31, 1995. From that date The scenario for the year 2000 and beyond, and which for many no electronic equipment whatsoever is allowed to be placed on people is a reality today, is the personal radio equipment. the European market without fulfilling the EMC directive. This Technology has now enabled us to communicate with anyone, includes restrictions on electromagnetic radiation as well as anywhere, anytime. But what about the communication link immunity. The discussion on the limits of both interference and which we must all share? Is there enough space for everyone to immunity level will continue, but from a radio engineer’s point travel along the same link at the same time with no restrictions of view, a great step has been taken to ensure proper operation on required bandwidth and with no restrictions on use wherever of all kinds of electronic equipment. we want to use our radio communication link? Looking ahead, we see a lot of work still to be done in the area Today, there are areas where the use of mobile telephones or of EMC. The use of the electromagnetic spectrum for communi- electronic equipment with active transmitters is prohibited, such cations purposes will increase dramatically in the years to come, as in hospitals and aircraft. These restrictions are for safety and the potential for conflicts between different users will in- reasons and are not caused by the use of the radio spectrum as crease accordingly. We must all be aware of the fact that the such. electromagnetic spectrum is a common resource for us all; its use cannot be based upon the rights of the strongest, but must be If we go into the actual EMC situation today, where do we have regulated in a manner which benefits everybody. the most pressing conflict between users of different radio equipment and other electronic devices? One of the areas which have been focused on in the media, is the possibility of dramatic situations arising if for example the airbag of a car is inflated because of incoming calls to the mobile phone when driving along the motorway, or a person’s heart implant stops on the

3 4 Electromagnetic interference - physical mechanisms

BY GEIR LØVNES

1 Introduction and on the current flowing through them. 2 Electromagnetic radia- Besides, the casing around the equipment All unwanted currents or voltages that may work as a resonant cavity. tion and induction may arise in a communication system are called electrical noise. All kinds of elec- Inductive or capacitive coupling may 2.1 Radiation from wires and tronic equipment produce such noise. cause the current on one wire to affect loops other wires. Signals from one loop may The term Electromagnetic Interference – In section 2.1.1 we will see that any wire affect other loops via common EMI – describes the situation that electri- carrying a time variable current may impedance couplings. Power and signal cal noise destroys the functionality of radiate. Dimensions and geometry will cables may in this way intercept systems. This noise may come from other decide whether the wires have good or unwanted signals from other wires inside parts of the system, from other systems poor antenna properties. In section 2.1.2 the equipment. Such couplings contribute or from natural phenomena (like light- we will give some examples of wiring to both radiated and conducted interfer- ning). which have “good” – but unwanted – ence. antenna properties. Depending upon the noise source, there A transmitter antenna also works as a are two main kinds of EMI: receiving antenna. Therefore, all wires 2.1.1 Infinitesimal radiation - Intrasystem – The noise sources are and PCB traces will also intercept inter- sources within the disturbed system ference. And noise may penetrate as con- We will start by setting up the complete ducted interference in power and signal - Intersystem – The noise sources are equations for the electrical and magnetic cables. other systems. field vectors around an infinitesimal In this paper we will look into the mech- dipole and current loop. Three basic elements must be present to anisms that turn electronic equipment We assume the dipole’s length is l, and create an EMI situation: into emitters and susceptors. The same that it carries a sinusoidal current I. As mechanisms largely apply to analogue - Emitter (noise source) shown in Figure 2.1, r is the distance equipment as to digital. The most impor- from the dipole to our observation point. - Susceptor (noise receiver, victim) tant difference in receiving noise is the The requirement to infinitesimality is met circuit’s susceptibility to interference, - Transfer or coupling medium (radiated when l « r, l « λ and I does not vary while an important difference in emission or conducted). along the length l. is the frequency spectrum which the circuits produce. High speed digital circuits Ignoring the transmission delay, we have When noise is transferred through cables, produce wideband noise and can give a for an infinitesimal dipole: it may be common mode or differential substantial contribution to interference 3 mode. Common mode noise is currents I · l · k into other systems. Eθ = − · flowing in the same direction in all wires. 4 · π · ω · ε 0 Differential mode noise is current flow- Most of the formulas are deduced in the 1 1 1 + + · sin θ ing in opposite directions in two wires. text. This may impair the readability, but (j · k · r) (j · k · r)2 (j · k · r)3 we have chosen to do it in this way in The opposite situation to Electromag- [V / m] (2.1) order to give the reader an understanding netic Interference is Electromagnetic of the prerequisites behind the formulas – Compatibility – EMC. An EMC situation that is to say under which conditions the occurs when all systems work as intend- formulas apply. ed, both isolated and in their natural environments. It must be noted that there will always be some interference between systems and internally within each system. How- ever, the term EMI denotes the situation that the interference is so severe that the functionality of one or several systems is affected in an unwanted manner, while the term EMC denotes the ~ situation that the interference is so low that all systems work properly. Hence, EMI and EMC may be regarded as two contrary situations. The source of interference from electronic equipment is the circuits generating the signals. These signals may be emitted because cables and printed circuit board (PCB) traces work as antennas. The power of this emission depends on geometry and dimensions of the cables and PCB traces, Figure 1.1 Electronic equipment as emitter and susceptor of interference. Interference may be radiated or conducted

5 H Z Er Z r Hφ Eφ I · l · k3 Er = − · P P 2 · π · ω · ε0 Eθ Hθ 1 1 + · cos θ (j · k · r)2 (j · k · r)3 θ θ r r I I [V / m] (2.2) YY l I · l · k2 φ φ H = − · A φ 4 · π 1 1 + · sin θ (j · k · r) (j · k · r)2 XX [A / m] (2.3) Figure 2.1 Infinitesimal dipole and current loop in spherical coordinate system k = 2π / λ

We will now look at a current loop with I2 area A carrying a sinusoidal current I. The current does not vary around the loop. When the diameter of the loop is A2 very small in relation to r and λ, and we r2 I3 ignore the transmission delay, we have I A A3 3 I1 1 the following equations for the field I2 components: r 3 I1 1 r j · I · A · k A5 3 H = − · φ 4 · π A r4 4 r3 r 1 1 1 5 + + · sin θ r2 (j · k · r) (j · k · r)2 (j · k · r)3 r I 1 I5 4 [A / m] (2.4)

j · I · A · k3 P H = − · r 2 · π P 1 1 + · cos θ Figure 2.2 Total field from conductors is found by superposition of conductor segments (j · k · r)2 (j · k · r)3 or loop segments [A / m] (2.5)

η[Ω] j · I · A · k4 1200 E = · φ 4 · π · ω · ε 1000 dipole 0 1 1 800 + · sin θ (j · k · r) (j · k · r)2 600 [V / m] (2.6) 377 k = 2π/λ 300 When r is very small, the r-3 terms dominate. These terms describe what is called 200 the static fields. The r-2 terms are induction fields, while the r-1 terms, which loop dominate when r is large, are called radi- Figure 2.3 Wave impedance of 100 ation terms. When r = λ/2π, the three infinitesimal dipole and current terms r-1, r-2 and r-3 are of equal magni- loop. (A small current loop is also tude. called a magnetic dipole) 0,4 0,5 0,7 1 2 3 4 6 8 10 2πr λ

6 We can draw some important conclusions: Firstly: Emission from any conductor or loop may be calculated on the basis of these equations by dividing the conductor or loop into many infinitesimal segments and find the total field by superposition (Figure 2.2). Besides, I we may find approximate expres- sions for the fields from conductors I and loops which are “quite” small (b) by using the equations for the (a) infinitesimal instances directly. This means that any wire carrying a time variable current will work as a transmitting antenna. However, the antenna properties may become poor because field contribution from various segments may Equipment A Equipment B be added in counter phase, so the far field may be strongly reduced. In the same way any wire will work as a receiving antenna. The principle of reciprocity states that the radiation pattern of an antenna is the same whether I the antenna works as transmitter or receiver. (The Radiation pattern is a three-dimensional graph describing the relative far-zone field strength versus (c) direction at a fixed distance from the Figure 2.4 Example of current loops working as antennas antenna.) We observe that field strength from the 4d2 r> infinitesimal sources is proportional to An antenna consisting of one or more λ for the far field the current I. current loops is usually called a magnetic antenna, while an antenna consisting of When r → ∞, E and H will approach d2 r r dipoles is an electrical antenna. r< zero, while the ratio between E and H λ for the near field components perpendicular to the propa- Figure 2.3 shows the wave impedance of gation direction, will approach the infinitesimal radiation elements. 2.1.2 Examples of cabling acting µ/ε = 120 π = 377 [Ω] as antennas 0 An important issue is where to draw the The ratio between E and H is called wave border line between near and far field. Any cabling1 constituting a current loop impedance, η. The value 120π of the As already mentioned, the three links in will work as a magnetic antenna. Exam- wave impedance is called free space (2.1)–(2.3) and (2.4)–(2.6) are of equal ples of loops which may create radiation impedance and is denoted by η . So, in 0 λ are given in Figure 2.4. the far field, wave impedance is equal to size when r = . For a point source free space impedance and the field from 2π Electrical antennas are usually created by a dipole and a current loop looks the we thus say that we are in the far field power and signal cables. Unintended same; we are unable to determine whether λ voltage drops in a circuit may, cause r> it was generated by a dipole or a current when 2 π . common mode currents to be set up in loop. connecting cables. Such voltage drops When the largest extent d of the source is are usually caused by potential differ- Very close to the dipole or the current of the same size as the wavelength, this ences in the ground plane (Figure 2.5). loop, terms which are proportional to r-3 far field condition is used [1]: Even if a cable is terminated to a socket in equations (2.1) – (2.6) will dominate. 2d2 r> or other equipment, from the circuit it This means that close to a dipole, λ may look like a monopole. This is E/H > 377 Ω, while close to a current For a half-wave dipole we then have: because the cable must be regarded as a loop, E/H < 377 Ω. When the field’s E λ transmission line when the cable length component dominates over its H compo- r> nent, or vice versa, we say that we are in 2 the near field of the radiation sources. A For very directive antennas [2]: 1 By cable is meant any pliable structure current loop is then seen as a source of containing one or more isolated wires magnetic fields; a dipole is seen as a surrounded by a rubber or plastic cov- source of electrical fields. This is import- ering. One single insulated wire will ant when choosing shielding materials. thus also be a cable.

7 I · A H = [A/m] (2.7) 2 · π · r3 This expression then gives the maximum magnetic near field from a small current Signal loop when the frequency is so low that Ι cable Common mode r « λ/2π.

Field close to a straight wire Ground At the distance r from a straight wire, the Vj plane Ι Common mode magnetic field is: I V H = [A/m] (2.8) Grounding j 2 · π · r Equivalent circuit when r is small in relation to the length of the conductor. Figure 2.5 Radiation caused by potential differences in the ground plane The formula comes from Maxwell’s equation: ∂ H · ds = I + D · n · dA ∂t (2.9) Near field Far field I is the conductor current and we pre- µ E(dB V/m) sume harmonic current variation. We 100 may with good approximation ignore the Z=10Ω (Ι=100mA) second part of the equation. The effect of Z=30Ω (I=30mA) the approximation is that the current I 80 cannot vary along the conductor. Z=10Ω (I=10mA) When integrating along a circle around 60 ≥ Ω Z 377 the conductor at a distance of r, we have: I H = [A/m] (2.10) 40 φ 2 · π · r 2 1v ~ 1 cm Z Based on another of Maxwell’s 20 equations, B · n · dA =0 0 (2.11) it can be shown that the other -20 components of the H field is zero. 1 3 10 30 100 300 1000 Far field from wire loop Frequency (MHz) The far field from a wire loop with area Figure 2.6 The field 3 metres from a 1 cm2 current loop excited by a 1 V sinusoidal [4]. For A is found from formula (2.6). The field other values corrections must be made by the formula: E = Ecurves + 20 · lg(V) + 20 · lg(A) (A = component is area in cm2) I · A · k3 1 Eφ = · · sin θ [V/m](2.12) 4 · π · ω · ε0 k · r is of the same order of magnitude as the wavelength. We will look closer at trans- We are not interested in the direction of E[dB] mission lines in section 5. the E field, and when rearranging the expression, we have this formula for the A major part of the radiation from equip- far field from a wire loop: ment with small dimensions compared to the wavelength is caused by common I · A 20 dB/decade E = π · η · · sin θ [V/m] mode currents on cables between the 0 λ2 · r (2.13) equipment units [3]. Such radiation is often denoted common mode radiation. We observe maximum radiation in the 40 dB/decade loop plane. 2.1.3 Practical equations The E field increases by 40 dB/decade 1 ∝ 2 f = 1 lg f (Ef). This applies to sinusoidal sig- 1 πτ f = πτ Near field around current loop 2 rf nals. By combining the typical spectrum From the equations (2.4) and (2.5) we from digital circuits (section 6), we have Figure 2.7 The far field spectrum from a have that the static field around a small current loop with periodic square pulses current loop will never be larger than τ τ with pulse length . rf is explained in Fig- ures 7.3 and 7.4

8 the resulting frequency spectrum for the compared to the wavelength. If the This is the maximum voltage to be far field from a current loop carrying dig- monopole is of the same order of magni- induced. If the polarisation of the antenna ital signals in Figure 2.7. tude as λ, we may apply the effective is different from that of the field, or the length of the monopole instead of l – see propagation direction of the field is not It is often more interesting to observe the next section – and use (2.16) as an app- perpendicular to the antenna rods, the field as function of voltage rather than roximate expression for the E field. voltage will be lower. current. If I in equation (2.13) is replaced by V/Z, we have The resulting spectrum with digital sig- From this we can determine the effective nals is shown in Figure 2.9. length, l , as the length the dipole would V · A eff E = π · η · · sin θ [V/m] have had if the current distribution were 0 λ2 · Z · r (2.14) 2.2 Induction in wires and rectangular. leff will therefore comply loops with the equation Z is the load, and if this is greater than 377 Ω, the field is largely related to the As mentioned above, a transmitting voltage V; the circuit becomes similar to antenna will also work as a receiving a voltage driven dipole [4]. When antenna. We will here consider the size E[dB] Z > 377 Ω, the generated field will be the of voltages which may be induced in a same size as at Z = 377 Ω. dipole/monopole and in a wire loop when 20 dB/decade these are placed in an electromagnetic Formulas (2.13) and (2.14) are derived field. from (2.6). Even if we in (2.6) assumed the diameter of the loop to be very small 40 dB/decade 2.2.1 Voltage induced in a in relation to λ, we will get a reasonable dipole/monopole result even when using (2.13) and (2.14) for larger loops. But none of the values The relation between electrical field 60 dB/dekade we use in A (length, width, or diameter) strength and voltage between two points must exceed λ/4 at any frequency. At a and b in space is: 1 1 πτ πτ lg f these frequencies the values used must be b rf reduced to λ/4 [4]. Above this limit the V = E · dl [V ] (2.17) Figure 2.8 Same situation as in Figure 2.7, but the spec- load impedance Z must also be replaced a by characteristic impedance for the PCB trum is assessed on the basis of voltage rather than cur- where the lines d l represent the integra- traces. rent, and we have assumed a purely capacitive load tion path between a and b. In many cases, particularly when MOS If a dipole has rectangular current distri- circuits are used, the load impedance will bution, i.e. the current is equal over the E[dB] be very capacitive. Assuming the load whole of the dipole length, the induced impedance to be purely capacitive, we voltage will according to (2.17) be can replace Z by 1/ωC in (2.14) and have V = E ⋅ l , where l is the dipole length. 20 dB/dekade -20 dB/dekade a spectrum as shown in Figure 2.8, that is d d a 20 dB/decade increase compared to In a real dipole the current decreases to Figure 2.7. zero at each end. Then the antenna does not receive as efficiently at the ends as in Far field from monopole the centre. Assuming cosinus shaped current distribution over the dipole and the E The far field from a short monopole over field being parallel to the dipole element, a ground plane is given by: 1 1 lg f we have for a half-wave dipole: πτ πτ sf f · I · l E = µ · · sin θ [V/m] λ/4 0 (2.15) 2πl λ r V = E · cos · dl = E · [V ] (2.18) Figure 2.9 The far field spectrum from a monopole over −λ/4 λ π l = length of the monopole a ground plane. The monopole is excited by square pulses as in Figure 2.7 The basis for this formula is (2.1). When l is the length of the monopole, and we only observe the radiation term, (2.1) gives: E µ I · l · f E = − 0 · · sin θ [V/m] 0 2 j · r (2.16) w l The monopole is placed over an ideally 2 l conducting ground plane, therefore we 1 must double the length of the monopole. w We are interested in the absolute value of the E field, and have formula (2.15). Direction of propagating E-field The formula presupposes uniform current distribution, which is a valid assumption λ/2 when the length of the monopole is small Figure 2.10 Situation when a plane wave moves through a frame antenna

9 Direction of propagating E-field length because here the current distribu- If the magnetic field is sinusoidal, we tion is defined as rectangular. have: α Likewise, a short monopole over a V = 2 · π · f · A · B · cosφ [V] (2.24) ground plane will have effective length w cos α equal to half the physical length, while a In the far field E/H = 377 Ω, and the quarter wave monopole will have effect- equations (2.22) and (2.24) give the same λ π w ive length /2 . But here we also have a result. Equations (2.21), (2.22) and (2.24) field contribution by reflection in the presuppose a uniform current distribution Figure 2.11 Angle of incidence in relation to the orienta- ground plane, so the maximum voltage on the loop. tion of the frame antenna measured over the monopole will be If the dimensions of the antenna are so ⋅ Vmonopole = 2E leff [V] (2.20) big compared to the wavelength that the current around the loop is no longer in 2.2.2 Voltage induced in a loop phase, the radiation pattern will be greatly affected. For example, a square For the sake of simplicity we assume that loop antenna with sides λ/4 will have the the loop is a square frame and that the same radiation pattern as a half-wave dimensions are much smaller than the dipole, see Figure 2.12. wavelength. When a plane wave enters as shown in Figure 2.10, a voltage will When the dimensions are such that we do be induced in the sides l1 and not have uniform current distribution, we α θ φ l2. The induction is the may put = = = 0 into formulas same in l1 as in l2, (2.22) and (2.24) and use the formulas as w b but because there an estimate of maximum induced volt- is a certain dis- age. The induced voltage increases as l tance between long as the largest dimension in the loop the sides, there is smaller than λ/2. Above this value the l will be a phase voltage will not increase further, but will difference oscillate with peak values for each half between the volt- wavelength. As an estimate for maxi- ages. If the enter- mum induced voltage we can use the ing direction is as value at λ/2 [4]. shown in Figure 2.11, l, w << λ l=w=λ/4 the phase difference will 2.2.3 Transfer of power be Figure 2.12 Radiation pattern for a small loop antenna An equivalence diagram for an antenna 2 · π with receiver is shown in Figure 2.13. and a square loop antenna with sides λ/4. For the small ∆k = · w · cos α, and we have: loop antenna the shape does not affect the radiation pat- λ The antenna’s impedance, Zant, consists of three parts: R which is the antenna’s tern as long as the loop is plane r ∆k radiation resistance, Rι is ohmic loss in V =2· E · l · sin · cos θ [V ] 2 (2.21) the antenna, while X is the reactive part of the input impedance. If the antenna is where we have multiplied by cosθ to in resonance, X = 0. Radiation resistance Zant make allowances for the E field not nec- is an imagined resistance consuming the essarily being parallel with l and l . 1 2 same power as the antenna emits when R R X r l Because the loop dimension is small used as a transmitting antenna. Zm is the compared to the wavelength and input impedance of the receiver. sin(x) ≈ x when x is a small angle, we V is the voltage induced when the ~ V Z may use sin(·) = (·) in (2.21). Letting l · w ind ind m antenna is situated in an electromagnetic = A, we have: 2 · π field. The voltage is distributed on both V = E · A · · cos α · cos θ [V ] (2.22) the antenna and receiver impedance. In λ Figure 2.13 Equivalence diagram for antenna with re- order to obtain maximum power transfer, ceiver The formula applies independent of the Zant = Zm* must apply. shape of the loop, as long as it is plane. The radiation resistance of a loop Induced voltage from a time varying antenna is [5]: magnetic field is: V = E · l [V] (2.19) 2 dipole eff 4 A dB Rr = 320 · π · [Ω] (2.25) V (t)=−A · · cos φ [V ] λ2 Thus, a half-wave dipole has effective dt (2.23) length λ/π. For a short dipole the effect- Small current loops have very low radia- A = area of the loop ive length will be half the dipole length tion resistance. If for example A = l·l because we may here assume triangular B = magnetic flux density and l = 0.1 λ, radiation resistance is 3 Ω. current distribution. An infinitesimal Transferred power from interference φ = angle between normal in the loop dipole has the same effective and real received with this current loop is small plane and the magnetic field. when the load resistance » 3 Ω.

10 wire E

R R Radiation resistance of a short dipole is LE LS also low – for example 8 Ω at dipole M length 0.1 λ. Large electrical antennas I E have higher values. A quarter wave I Ω E V monopole has the value 36.6 . A wire wire S S acting as a dipole antenna, not having V R E LS symmetric feed, may have radiation R Ω a GS a R R resistances in the k area. E VE S LE GS d On a printed circuit board most of the ES ground current loops are small compared to the plane wavelength of the interfering fields. (a) (b) Therefore, the current loops act as low impedance magnetic antennas. On print- Figure 3.1 Inductive transfer between two parallel wires ed circuit boards low impedant circuits will therefore be most susceptible to magnetic fields because this gives the best impedance match to the interference. circuit 1 Similarly, high impedant circuits will be most susceptible to electrical fields l because the interference is received by the input cables which are often the same size as the wavelength, and therefore have higher radiation impedances. circuit 2 However, we should be aware of the fact that the consequences of the interference b b are not necessarily determined by power 2 1 transfer to the interfering element. For example, a current loop connected to a a a 1 2 MOS circuit may interfere greatly even if the impedance match is poor, because MOS circuits are interfered by interference voltages.

3 Crosstalk Figure 3.2 Transfer between two conductor pairs In section 2 we observed how cabling works as transmitting or receiving antenna independently. This section will conductor S. The inductive transfer may a1 · a2 deal with capacitive and inductive trans- be expressed as: M =4.61 · 10−7 · l · lg [H](3.5) 1 · 2 fer between (at least) two conductors in ∂IE b b V = M · [V ] (3.2) the electrical and magnetic near field. S ∂t Returning to Figure 3.1 we observe that there may also be capacitive transfer be- Unwanted inductive or capacitive trans- where V is shown in Figure 3.1 and M is E tween the wires. We redraw the circuit ferred power from a wire or pair of wires mutual inductance between wires. and get the equivalence circuit as shown to another wire or pair of wires is called With harmonic time variation we have: in Figure 3.3. crosstalk. The term wire may here mean anything from the traces inside an IC to a V The voltage across RGS and RLS given by V = M · ω · I = M · ω · E [V ] pair of wires in an Atlantic Sea cable. S S R (3.3) the capacitive transfer is, with harmonic LE time variation: Crosstalk is defined as: If we have a configuration as shown in Vsusceptor Crosstalk =20· lg [dB] (3.1) Figure 3.1.a, we have the following R · VE Vemitter VS = 1 [V ] (3.6) expression for the mutual inductance M R + jωC In Figure 3.1 are shown two parallel [2]: where wires with earth as return. The current in µ (a + a )2 + d2 E produces a m,agnetic field affecting M = 0 · ln E S ES [H] R · R 2 2 (3.4) GS LS 4π (aE − aS) + dES R = [Ω] (3.7) RGS + RLS where all lengths are in metres and given C is the mutual capacitance between the by Figure 3.1. wires. Ignoring the effect of the ground If we have two wire pairs as shown in plane, C is given by [2]: Figure 3.2, we have the following expression for M [2]:

11 wirer E I1 R R LE LS I2 l LS C V1 ~ V2 ~ Z2 Z1 wire S R R R V VE LE GS LS S Z R c a GS a E VE S (b) Figure 4.1 Common impedance coupling dES across Z ground c plane (a) Figure 3.3 Capacitive transfer between two parallel wires Figure 5.1. We see that if the length l in the transmission line model in Figure 5.1.a becomes very small (l « λ), we have 3.5 · 10−10 · l a situation as shown in Figure 5.1.b, C = ES [F ] 4 Common impedance 2 (3.8) which is pure circuit technique. 2π · ln dES coupling gE gS In Figure 5.1.a there will be a voltage gE = radius of wire E This type of coupling is made by two wave VGL on the transmission line from loops in a circuit sharing an impedance. generator to load. If there is impedance gS = radius of wire S When the current I1 in one loop produces mismatch between load and line, a volt- lES is mutual length of E and S. a drop in voltage over the common age wave VLG is reflected from load to impedance, the current in the other loop generator. The ratio between these two All lengths are in metres and given by will be interfered by this (see Figure 4.1). voltage waves is called reflection coeffi- Figure 3.3. cient, and at load this is given by: As for inductive and capacitive coupling, ω If 1/ C » R we have: the common impedance crosstalk will VLG Z1 − Z0 also increase with frequency. When the ρ = = (5.1) V = R · V · jωC [V] (3.9) V Z + Z S E common impedance is a ground plane, GL 1 0 the increase is 10 dB/decade, when the The ratio between the currents is equal to From (3.3) we observe that inductive common impedance is a conductor, the (voltage) reflection coefficient with the crosstalk increases when impedance in increase is 20 dB/decade [1]. sign reversed: the interfering circuit decreases. (3.9) shows that capacitive crosstalk increases I 5 Impedance mismatch LG = −ρ when impedance in the interfered circuit IGL (5.2) increases. In case of poor impedance match we may have mutliple reflections on the transmis- The line’s characteristic impedance is Inductive crosstalk is strongest sion line between generator and load. - in transformers This may cause high frequency oscilla- R + j · ω · L Z0 = [Ω] (5.3) tions to arise. G + j · ω · C - when wires are parallel If the wire length from generator to load R, G, C, and L are distributed ohmic - usually at lower frequencies. is considerably less than λ, where λ is the resistance in the wires, distributed leak- shortest wavelength of the transferred age conductance between the wires, dis- Capacitive crosstalk is strongest signal, we find currents and voltages eas- tributed capacitance between the wires - in multiconductor twisted cables ily by using Ohm’s law. If this is not the and distributed serial inductance in the case we must consider the wire as a wires, respectively – all per length unit of - at high frequencies transmission line and use transmission the wire. - at high impedance levels. line theory in order to calculate voltages Assuming R = G = 0, the line is loss free. and currents. We note that Z is real in the case of loss From (3.3) and (3.9) we observe that 0 A transmission line model of transmis- free lines. Furthermore, we note from both capacitive and inductive crosstalk sion from generator to load is shown in (5.1) that we have a positive voltage increases by 20 dB/decade when the frequency increases. We get a rough estimate of crosstalk by l using the formulas for capacitive and Z Z inductive coupling and choosing the one g g giving the highest result. If wire lengths are greater than λ/4, we ~ Z ~ Z cannot use the formulas in this section. In V Z0 l V l this case we must consider the conductors as transmission lines and find mutual inductance and capacitance per length unit. (a) (b) Figure 5.1 (a) Transmission line model, (b) circuit technique model

12 Z >Z Z =Z Z Z0, i.e. the voltage 2,0 2,0 2,0 drop across the load is above the excita- 1,4 1,4 1,4 tion level. When Z1 < Z0 we have negative voltage reflection and the voltage 1,2 1,2 1,2 drop across the load is below the excita- 0,8 0,8 0,8 tion level. 0,4 0,4 0,4 Supposing, in Figure 5.1.a, there is impedance match either by generator 0 0 0 (Z = Z ) or by load (Z = Z ), the net- 01020 01020 01020 g 0 l 0 t t t Z =5Ω Z =5Ω Z =5Ω work will enter a stable end state as soon g g g as the pulse front on the line hits the Z =500Ω Z =50Ω Z =5Ω matched load. However, transferred l l l Z =Z Z Z Z

13 Time: Frequency: X (f) aτ 1 1 quency plane we draw the frequency logarithmic and the amplitude in dB, we will x1(t) a see that the envelope of the maximum amplitudes decreases by 20 dB/ decade. If we also consider rise time and fall time of the square pulses, and define parame- −τ /2 −τ /2 t f ters as in Figure 7.3, we will have a fre- 1 1 quency plane approximation as shown in −1/τ τ Figure 7.4. Note that if the rise time is 1 1/ 1 different from the fall time the lowest of the two is used. Figure 7.5 shows how the lines in Figure 7.4 are upper bars of x (t) X (f) 2 τ 2 the harmonics. a a 2 In Figures 7.6 and 7.7 are shown corresponding envelopes for an individual square pulse and periodic triangular pulses. −τ −τ 2/2 2/2 t f A bit sequence consisting of “ones” and “zeros” coded as +A volts and 0 volts −1/τ2 1/τ2 may in the frequency plane be drawn in the same way as periodic square pulses. The points f1 and f2 are determined by x3(t) a X (f) the pulse width and the rise time or fall aτ 3 time of an individual “one”, while the 3 distance between the harmonics is determined by the bit rate. In addition, we will here have frequency components lower −τ −τ 3/2 3/2 t − τ 1/τ f 1/ 3 3 than f0, because we may have several consecutive “ones” or “zeros”. For the Figure 7.1 The Fourier transformed of square pulses of various lengths same reason we will also have components between n · f0 and (n + 1) · f0.

The Fourier transformed of a square component at 1/T is called the first har- 8 Conclusions pulse is a sinc function, monic, the component at 2/T is called the In this paper we have looked at physical sin(π · f) second harmonic, etc. mechanisms behind electromagnetic sin c(f)= – see Figure 7.1. interference. π · f When the length of each square pulse is A square pulse in time domain will there- half the distance between the pulses, i.e. The term Electromagnetic Interference – fore give a sinc shaped frequency spec- at 50 % duty cycle, the even harmonics EMI – describes the situation that electri- trum. will correspond to the zero points of the cal noise destroys the functionality of sys- envelope. At 50 % duty cycle we will tems. The opposite situation is called Elec- Supposing that in time domain we have therefore only have uneven harmonics tromagnetic Compatibility – EMC. An periodic square pulses, one pulse every T (1., 3., 5., etc.). (Duty cycle: δ = τ/T.) EMC situation is when all systems work as seconds, we will in the frequency domain intended in their natural environments. have discrete frequency components sep- There is no need to carry out Fourier arated by 1/T and a sinc shaped envelope. transformations in order to quantitatively To create an EMI situation, a noise When the length of each square pulse is determine the envelopes of the frequency source and a receiver or victim must be τ, the first zero point at the envelope will spectra. We will first look at periodic present, and the noise must be transferred be 1/τ (see Figure 7.2) The frequency square pulses. Supposing that in the fre- from the source to the victim. This may be done either as radiated or conducted noise.

A 0,9A 1/T τ 0,5A

0,1A -2/τ -1/τ 1/τ 2/τ τ τ f r f t T Figure 7.2 The Fourier transformed of periodic square pulses. Each square pulse is τ long, and the distance between them is 1/T Figure 7.3 Periodic square pulses

14 V VdB dB f =1/T fundamental frequency 0 A(0) f = 1 -20 dB/decade 1 πτ A(f0) f = 1 2 πτ A(2f0) rf A(nf0) τ τ τ -40 dB/decade rf=min{ r, f } δ ≈ τ lg f f0 f f 2f f nf lg f Figure 7.5 Example of correct frequency 1 0 0 2 0 A(0)= 20 lg (2Aδ) spectrum of the periodic square pulses in Fig- A(f )= A(0)-20 lg (f /f ) ure 7.3 0 0 1 Figure 7.4 Envelope of the frequency spectrum for the periodic square pulses in Figure 7.3

9 References

1 Krebs, P, Nissen, H G. HMC-hånd- bog del 3: EMC-hensyn ved print- udlægning. Hørsholm, Elektron- VdB ikCentralen, 1989. A(0) -20 dB/decade 2 Keiser, B. Principles of electromag- A netic compatibility. Norwood, MA, 0,9A τ Artech House, 1987. ISBN 0-89006- 0,5A 206-4. -40 dB/decade 0,1A 3 Hubing, T H, Kaufman, J F. Model- τ τ t r f f f lg f ing the electromagnetic radiation 1 2 from electrically small table-top 1 1 f = f = πτ products. IEEE transactions on Elec- 1 πτ 2 rf tromagnetic Compatibility, 31, 74–84, 1989. A(0)=20lg(2Aτ) Figure 7.6 The envelope of the frequency spectrum of an individual square pulse. We 4 Mardiguian, M. Interference control will here have a continuous spectrum below the envelope in computers and microprocessor based equipment. Gainesville, VA, Don White Consultants, 1984. ISBN 0-932263-23-2.

5 Andreassen, S O. Antenner. Oslo, Teledirektoratet, 1987.

V A dB

A(0)

-40dB/decade

τ τ r f T f0 f1 lg f

1 1 f0 = f = πτ T 1 rf A(0)=20lg(2Aδ) Figure 7.7 Periodic triangular pulses in the time and frequency planes

15 Radio interference in the band 30 – 1000 MHz – the international limits of interference seen from a radio-engineer’s point of view

BY EGIL HAUGER

The radio interference limits stated by We have a report in Norwegian which may be subject to restrictions on its EN55022 (European Norm) have its gives some background material. Unfor- sale and/or use. background in CISPR 22 (Comité Inter- tunately, we do not have the title of the Note: The limits for Class A equip- national Spécial des Perturbations report, but it is probably written by a del- ment are derived for typical commer- Radioéléctriques). egate working in the area of radio inter- cial establishments for which a 30 m ference, Radiostøykontrollen (the Radio Before discussing the actual limits in this protection distance is used. The Class Interference Service). frequency range, let us have a look into A limits may be too liberal for domes- the background for the chosen figure and Section 5.5 in this document gives the tic establishments and some residential what type of information the limits are background for limits and methods of areas. based upon. measurements, Bakgrunn for gren- Class B equipment is data processing severdier og målemetoder. In a document from CISPR/B (Secre- and electronic office equipment which tariat) September 30, 1983 the interfer- In para 5.5.1 General, we have: satisfies the Class B interference lim- ence limits in the frequency range 30 – its. Such equipment should not be sub- “CISPR/B has concluded as follows: 1000 MHz are presented. This is a sec- ject to restrictions on its sale and is ond draft recommendation for data and - The potential of interference from generally not subject to restrictions on electronic office equipment (DP/OE) and DP/OE is different for professional use its use. takes account of the detailed considera- than for the common use at home. Note: The limits for Class B equip- tion of the comments on Document - DP/OE commonly for professional ment are derived for typical domestic CISPR/B (Central Office) 9. It was pre- use, e.g. within production, medicine, establishments for which a 10 m pro- pared by CISPR/B/WG2. education, official business etc., may tection distance is used.” In section 2.1 we have the definitions of be used both in the domestic area and Data Processing and Electronic Office in the commercial and industrial area. The most interesting point here is to see Equipment (DP/OE): that also for Class B equipment the - In professional use we may have dif- CISPR committee recommended a pro- “Electrical/electronic units or systems ferent factors of propagation compared tection distance of 10 m from the source which predominantly generate a multi- to domestic use. of interference to the victim. In 1983 this plicity of periodic binary pulsed elec- - These differences in propagation fac- may have been a reasonable distance be- trical/electronic wave forms and are tors require different recommendations tween electronic equipment, but the designed to perform data processing for interference limits. question is: what is the reality today and functions such as electronic word pro- what about the electromagnetic environ- cessing, electronic computation, data - Electromagnetic energy may switch ment for the 1990s and beyond 2000? transformation, recording, filing, sort- from the source to nearby radio system ing, storage, retrieval and transfer, and by conducting, inductive or capacitive When going into the background material reproduction of data/or images but coupling and radiation. we see that the maximum radiated field excluding equipment designated solely strength by a receiver placed at least - In the frequency band below 30 MHz for telecommunications purposes”. 30 m (E30) from a class A DP/OE is the dominating factor of coupling is by The emphasizing done by myself. given by: the conducting networks and the inductive and capacitive coupling from E30 = ES - S/N + AB In CISPR Publication 22 First edition the same. 1985, Limits and methods of measure- where ments of radio interference characteris- - In the frequency band above 30 MHz tics of information technology equipment, radiation is the dominating factor, but E30 = radiated limits of interference we have mainly the same definition of in some installations the inductive and from professional DP/OE at the source of interference but under the capacitive coupling from cables con- least 30 m from a receiver in abbreviation ITE – Information Technol- necting host units to peripherals may the chosen broadcasting band ogy Equipment. The interesting differ- be of importance. S/N = signal/noise ratio for the chosen ence is that the exclusion of equipment - In the frequency band from 30 to 300 broadcasting service designated solely for telecommunications MHz the source of radiation from purposes is removed. AB = attenuation factor for the build- smaller equipment giving interference ing What had happened from September will be the mains or other cables con- 1983 to the official publication in 1985? nected, mainly the cables nearest the ES = usable field strength for the The limits of interference are the same, equipment”. receiver. “equipment designated solely for telecommunications purposes” seems to In the document from September 1983 The calculated field strength for the have no influence upon the chosen figures we have the definition of Class A and B chosen band is given in Table 1. for recommended limits of interference. equipment: Investigations published by CBEMA The only change is that in the 1985 Publi- “Class A equipment is data processing (Computer and Business Equipment cation there is a section 4.2, Limits of and electronic office equipment which Manufacturers Association) shows that telecommunications line interference satisfies the Class A interference limits 89 % of the receiver antennas within 100 voltage, which is “under consideration”. but does not satisfy the Class B limits. m from a class A DP/OE are located at The interesting point is now what the In some countries, such equipment least 30 m away from the installation. interference limits are based upon?

16 Table 1 Calculated field strength for the chosen band Frequency range MHz Band I III IV V Primary service 54–88 80–108 108–136 470–1000 (1) (1) TV FM Aero TV dB(µV/m) 48 55 65 70 (1) The values for Bands IV and V ES dB (µV/m) 68 60 15 74 should be increased by 2 dB for the S/N dB 45 30 0 45 625-line(OIRT) system”. AB dB 8 8 21 8 CCIR also assumes that, in the absence E30 dB (µV/m) 31 38 36 37 of interference from other television transmissions and man-made noise, the minimum field strength at the receiver Remarks: 2 More than 8 dB attenuation in the antenna that will give a satisfactory grade interjacent walls (also more walls) - ES in TV band is minimum grade A of picture, taking into account receiver “Statistical expected” usable signal 3 The frequencies with the highest noise, cosmic noise, antenna gain and defined by FCC. radiation do not fall within the re- feeder loss is 47 dB(µV/m) in Band I, 53 ceiver selectivity. dB(µV/m) in Band III, 62 dB(µV/m) in - ES in Aeronautic band is chosen to Band IV and 67 dB(µV/m) in Band V. 118 MHz where air–ground service The measuring distance was chosen to be starts because the interference po- For the FM sound broadcasting at VHF, the protection distance of 10 m. Since all tential by ground stations is higher. CCIR “Unanimously recommends” that conditions except protection and measur- the following planning standards should - Experiments and literature studies ing distances are the same, the limits be used for frequency-modulation sound show that buildings nearly almost give evaluated for class A equipment in the broadcasting in band 8 (VHF): attenuation of signals. A factor of 8 dB 30 m distance will give the same protec- attenuation for a single wall was tion for the receiver also for class B 1 Minimum usable field strength chosen except in the aeronautic band. equipment at a distance of 10 m. In the presence of interference - Empirical results published by Table 2 gives the interference limits in from industrial and domestic CBEMA show a building attenuation the frequency band 30 – 1000 MHz for equipment a satisfactory service from 21 to 39 dB in 27 American air- class A and B equipment. requires a median field strength ports. A factor of 21 dB is used for all (measured 10 m above ground The aerial should preferably be located at aeronautic band. level) of at least: the horizontal distance from the test unit - The S/N ratio is found empirically by as specified above. If the field strength 1.1 for the monophonic service: use of a CISPR quasi-peak receiver as measurement at 30/10 m cannot be re- 48 dB(µV/m) in rural areas reported by CBEMA. corded because of high ambient noise 60 dB(µV/m) in urban areas levels or for other reasons, measurements 70 dB(µV/m) in large cities Furthermore, is was stated that the quasi- may be made at a closer distance. If so, peak receiver specified by CISPR publi- 1.2 for the stereophonic service an inverse proportionality factor of cation 16 part 1 will give the same pro- 54 dB(µV/m) in rural areas 20 dB/decade shall be used to normalize tection for the TV receiver both against 66 dB(µV/m) in urban areas the measured data to the specified dis- wideband interference as well as narrow- 74 dB(µV/m) in large cities. tance for determining compliance. band signals. Therefore, the limits will be the same for both type of interference in Comparing the E30 values in Table 1 to When comparing these values with the range 30 – 1000 MHz by use of a the interference limits in Table 2 we see CISPR emission levels, the different ser- CISPR quasi-peak receiver. a high degree of conformity. We can vices require different radio-frequent conclude that the figures in Table 1 are protection ratios. CCIR has, in recom- For the Class B DP/OE a protection dis- the basis for the EN55022 interference mendations 412-4 and 655 (1986), given tance of 10 m was chosen. This distance limits. protection ratios for FM broadcasting and should give sufficient protection when television services. The given figures are installing DP/OE in domestic areas. only valid for interference from FM or The protection distance was based upon Radio communication and TV transmitters and other types of radio the following supposition: interference limits interference are not considered. In [1] - The receiver should be protected from The ES values in Table 1 give the “Sta- the neighbour DP/OE tistical expected” usable signal defined Table 2 Interference limits in the frequency band 30 – by FCC. - There will be relatively few receivers 1000 MHz for class A and B equipment closer than 10 m In CCIR Rec. 412-4 and 417-3 1986, CCIR Class Frequency Quasi-peak Test - In case of less than 10 m distance, range limits distance other factors of contributions may “Unanimously recommends: MHz dB (µV/m) m reduce the possibility of interference. 1 that when planning a television ser- These may include, but are not restrict- vice in Bands I, III, IV or V, the A 30 – 230 30 30 ed to: median field strength for which A 230 – 1000 37 30 protection against interference is 1 Orientation of source and receiver B 30 – 230 30 10 planned should never be lower than: B 230 – 1000 37 10

17 some investigations have been done con- ing a high ignition interference and a In both these example, the FM and TV cerning interference from different types higher utilization of the radio spectrum broadcasting has the potential problem of of radio sources onto FM sound broad- for communication purposes. But when interference from IT equipment which is casting and television services. The typi- we come to the use of IT equipment, our in line with CISPR 22B. cal interference spectrum from a modern private, in-house environment is the Other types of radio communication may IT-equipment or tele-terminal (ISDN) is same regardless of it being a flat in a have different requirements to interfer- multiples of clock frequencies up to large city or a villa in the rural areas. So, ence immunity. We will go through some some hundreds of MHz and with a white if we are talking of interference from IT examples illustrating the great variations noise spectrum 10 – 20 dB below the equipment, there should be no difference of external interference in which the harmonics. This white spectrum is gener- depending on the customers place of resi- radio system is intended to operate. ated in the processor chips with its data dence. The figures from CCIR seem to lines and seen from the outside this acts take no consideration of the wide use of An essential factor in all calculations as a random pulse generator with white IT equipment even in non-commercial involving field strength is the antenna noise spectrum. areas. Going back ten years, this is not so factor. In the following we will use the surprising, but we have to draw the con- equation: If we consider the FM stereophonic sequences of the IT revolution. broadcasting the NTR report measures AF(dB) = 20 log(f,MHz) the protection ratio to 42 dB when the The same exercise as for the FM broad- - 29.8 - G(gain, dBi) interference is white Gaussian noise. The casting can be done for television ser- In some installation we may have some quality criteria are in accordance with vices. As shown above we have the fig- dB gain, but for interference calculations CCIR rec. 641 with a S/N-ratio of 50 dB. ure 48/55 dBµV/m in band I/III “for where no information is given regarding For comparison CCIR recommends a which protection against interference is antenna position and interference source protection ratio of 45 – 51 dB for stereo- planned”. In the same CCIR recommen- we use isotropic antenna with G = 0 dB. phonic broadcasting when a standard FM dation we have the figure of radio-fre- signal is used as interferer. From the fig- quency protection ratio. This is based When referring to interference field ure of field strength and protection ratio, upon TV-signals as interferer and figures strength we have to be aware of the differ- we can now calculate the maximum for CW signal or white noise are not ence in bandwidth for the different radio interference level which can be accepted given. As mentioned earlier, some com- systems. In the examples given below we in the vicinity of the receiver antenna for parisons between white noise and FM will use the equivalent bandwidth of 120 an acceptable reception quality. and TV as interferers have been done in kHz which is in line with CISPR 22. From [1]. When using a 120 kHz BW and QP this follows that we are talking about In the case of FM stereophonic service detector the actual protection ratio is broadband interference and not single fre- we have the following figure for maxi- measured at 42.3 dB. This has to be com- quencies generated by harmonics. mum interference: pared with the continuous interference Large cities: from 33 – 52 dB where TV signal is - Radiotelephony, 80 MHz, analogue: 74 dB(µV/m) used. The variation of protection ratios Sensitivity 0.0 dB(µV) - 42 dB (protection ratio) depends on frequency offset in multiples AF + 8.3 dB = 32 dB(µV/m) of 1/12 line-frequency. C/I - 8.0 dB Imax 0.3 dB(µV/m)/10 kHz Urban areas: The maximum interference can now be Imax 66 dB(µV/m) calculated for band I/III (120 kHz) = 11.3 dB(µV/m) - 42 dB (protection ratio) 48/55 dB(µV/m) = 23 dB(µV/m) - 42 dB (protection ratio) - FM car-radio mono, 100 MHz: Rural areas: = 6/13 dB(µV/m) Sensitivity 48 dB(µV/m) 54 dB(µV/m) C/I - 36 dB Comparing with the interference of - 42 dB (protection ratio) Imax = 12 dB(µV/m)/165 kHz 30 dB(µV/m) 10 m away from IT equip- = 12 dB(µV/m). Imax ment we obviously have a problem. The (120 kHz) = 10.6 dB(µV/m) These figures can now be compared with CCIR recommendation refers to field CISPR 22 Class B as mentioned above: strength at a height of 10 m above - Paging, 170 MHz 30 dB(µV/m) in 10 m distance from ground level but no figure of antenna Sensitivity 25.0 dB(µV/m) interferer. The figure for large cities is gain is given. Using an outdoor antenna C/I - 10.0 dB quite compatible with the calculated one, we may have a front/back ratio of 10 dB, Imax = 15.0 dB(µV/m)/10 kHz but when we go outside the city we may so the interference level can be increased Imax have problems. Especially in the rural to 16/23 dB(µV/m). The missing 14/7 dB (120 kHz) = 25.8 dB(µV/m) areas where the wanted signal is weak, for band I/III can now be achieved by the interference from IT equipment may increasing the distance from IT equip- - Mobile telephony analogue, 460 MHz have severe consequences. One may ask ment to antenna to 50/22 m, turning the Sensitivity 4.0 dB(µV) why we have different figures from equipment in a direction where the radia- AF + 23.4 dB CCIR for the broadcasting services in tion lobe has a minimum or by adding C/I - 18 dB cities and rural areas. The obvious expla- 14/7 dB to the wall attenuation. In prac- Imax = 9.4 dB(µV/m)/10 kHz nation is the higher interference level. tice, a combination of all these adjust- Imax This interference has a lot of sources; ments will be the best solution. (120 kHz) = 20.2 dB(µV/m) heavy industrial plants with high powered machinery, high density of cars caus-

18 - Mobile telephony analogue, 900 MHz The other interesting figure which can be Table 3 Maximum interference to radio systems Sensitivity 4.0 dB(µV) seen from Table 3 is the comparison be- Radio Max. interference CISPR22B AF + 29.3 dB tween analogue (NMT) and digital system dB(µV/m) dB(µV/m) C/I - 18.0 dB (GSM) when talking about interference 120 kHz 10 m Imax = 15.3 dB(µV/m)/10 kHz immunity. The lower immunity to GSM, Imax in spite of the 9 dB lower C/I ratio, is of TV band I 6 30 (120 kHz) = 26.1 dB(µV/m) course due to the fact that GSM requires a bandwidth of 230 kHz while NMT only Radiotelephony - Mobile telephony digital, 900 MHz 80 MHz 11.3 30 occupies 10 kHz. Sensitivity 6.0dB(µV) FM car radio, AF + 29.3 dB From a practical point of view, we see mono 10.6 30 C/I - 9.0 dB that the most exposed radio system is the Imax = 26.3 dB(µV/m)/230 analogue system in the lowest frequency FM stereo 12 – 32 30 kHz band. In Norway we use the band I for TV band III 13 30 Imax TV transmission, and in rural areas Paging 170 MHz 25.8 30 (120 kHz) = 23.5 dB(µV/m) where the TV-signal is on the margin, the introduction of IT equipment for com- Mobile telephony The figures for sensitivity are of vital mon use may well cause problems for analogue 460 MHz 20.2 37 importance for these calculations. For TV broadcasting. This frequency band is Mobile telephony radio telephony in the 80 MHz band the also the band with the highest interfer- analogue 900 MHz 26.1 37 sensitivity is decreased by 6 dB from the ence level. When measuring interference Mobile telephony receiver’s performance under static con- from IT equipment we generally have a digital 900 MHz 23.5 37 ditions due to fading under normal use, falling interference level above some and we accept a C/I ratio of 8 dB. This will give a reasonably good speech quality, but not up to telephone standard. For mobile telephony in 460 and 900 MHz band (NMT), we have used the receiver sensitivity under fading condition. For digital systems in 900 MHz (GSM) we have used the reference sensitivity + 3 dB due to the higher BER with fading. The paging performance data are taken from the ETSI recommendation. For FM and TV broadcasting the CCIR recommendation is used. For the mobile system the equivalent bandwidth is the 3 dB bandwidth with standard signalling measured on the running system. The maximum interference seen from a radio engineer’s point of view can be summarised in Table 3. One obvious conclusion can be drawn Figure 1 Interference ISDN terminal 30 – 200 MHz from Table 3: the higher the immunity, the higher the frequency. This is because the antenna factor is increasing by 6 dB/ octave, or in other words, the receiver sensitivity is decreasing by 6 dB/octave when we refer to field strength with equal antenna gain. So, when we keep the interference field strength constant in a given frequency band, the interference power delivered to the receiver is dropping by the same 6 dB/octave when we use an antenna with a constant gain. This frequency dependency may be turned the other way round; the higher the radio frequency, the higher the wanted signal must be in terms of field strength. This is a general problem with all radio communication and is one of the reasons for the reduction of radio link distances with higher radio frequency. Figure 2 Interference ISDN terminal 200 – 1000 MHz

19 hundreds of MHz. This is of course relat- spectrum and are usually located around The mobile radio trans- ed to the clock frequency and the proces- or above 1 GHz. In this band the man- sor speed, but with today’s IT equipment made noise is negligible in most environ- mitter as an interference with clock frequencies up to 50 MHz the ments, but here we have the drawback source to other electronic highest interference level is in the fre- with radio link distance. quency band 30 – 200 MHz. Above equipment So if you do not receive the message on 200 MHz we may have high levels from We have been talking about the IT equip- your pager when staying in a modern harmonics of internal oscillators, but the ment as a potential source of electromag- office environment, the reason may well “white” processor noise is normally re- netic pollution, but this is only one side be the interference from all the IT equip- duced by more than 20 dB. of the compatibility problem in the elec- ment and not the missing radio planning tromagnetic world. To illustrate a typical radio interference from the telecom operator. CISPR never spectrum, we have in Figures 1 and 2 considered today’s penetration of “Data All radio transmitters, like the mobile given the frequency plot of an ISDN ter- Processing and Electronic Office Equip- telephone, are intended to radiate electro- minal from 30 to 1000 MHz. According ment, DP/EO” when the limits of inter- magnetic energy. This radiation is what to CISPR the resolution bandwidth is ference were stated. we can call “wanted” in contrast to the 120 kHz and the green line is the CISPR electromagnetic energy radiated from What about ETSI and their contribution 22B limit. For the quasi peak measure- non-radio IT equipment whose commu- to the compatibility between IT and radio ments the test receiving antenna and the nication is on copper cable or optic fibre. equipment? terminal are orientated for maximizing If we use Radio-LAN for communication the field strength. we again have an ordinary radio trans- In prETS 300 339 June 1993 “Radio mitter included in the IT equipment. In the range below 70 MHz we see the Equipment and Systems (Res); Generic “white” processor noise. Above 70 MHz Electro-Magnetic Compatibility (EMC) The radio transmission is regulated in the spectrum is dominating by the harm- for radio equipment”, we have criteria CCIR and co-ordination between differ- onics of the internal oscillators with a for emission and immunity. The “ETS is ent radio systems are part of the radio fre- level around 10 dB higher than the pro- based upon the Generic Standards EN quency management work. If this co-ordi- cessor noise. This is very typical for this 50081-1 [3], EN 50082-1 [4] and other nation is done as intended and the guide- type of equipment. When we pass standards, where appropriate, to meet lines for transmitter power, antenna gain, 230 MHz we see that the interference is the essential requirements of the Council antenna position etc., are respected, the dropping by 20 dB. Directives 89/336/EEC [1] and 92/31 radio systems should not give any inter- [2]. The ETS applies to domestic, com- ference to each other. The problem is the The probability of disturbances from IT mercial light industry and vehicular envi- potential interference to the non-radio equipment onto TV transmissions is of ronments only and therefore may not electronic equipment which can be used course influenced of many factors. One give sufficient protection in more hostile in the vicinity of the radio transmitter. import factor is the fact that the radio environments”, and “This ETS incorpo- This problem is not a new one. In the interference must be within the receiver rates normative annexes which specify neighbourhood of high powered AM band, we must have “in-band” interfer- test methods relevant to a number of tests transmitters we have for several years had ence. The broadband processor noise in this ETS. These test methods have problems with interference to electronic may be 10 – 20 dB below the harmonics been taken from the latest available equipment, particularly the telephone sets peak values. Taking 15 dB as a normal information on the development of EMC including audio amplifiers. peak/broadband factor, we still have a Basic Standards in the International too high interference level for the radio In the 27 MHz band, the private radio Electrotechnical Commission (IEC) and systems in the lowest frequency band. band, the amplitude modulation was pro- Comité Européen de Normalisation Éléc- For the paging system we are just on the hibited in the early 1980s due to the tronique (Cenelec)”. limit when the distance from source to interference to other electronic equip- victim is 3 m. ment. In all the international recommendations In many of the radio systems mentioned and standards the emission limits are The problem with AM is the way in here, the 10 m distance source – victim is based upon CISPR 22 which, as seen which the demodulation can take place. unrealistic. For FM and TV this can be above, has its background material from All kinds of amplifiers have some non- achieved by using outdoor antennas, but a world totally different from today’s linearity and the non-linearity itself for the mobile terminal in car installation reality when talking about electromag- demodulates the carrier giving the inter- or body worn equipment like pager or netic environment. The sentence “The ference as a result. It is easily shown that portable mobile telephone the distance in limits for class B equipment are derived when you express the in/out relation in an ordinary office environment is in the for typical domestic establishments for an amplifier as seen in Figure 3 in terms range of less than 3 m. That means a which a 10 m protection distance is of power series 10 dB higher interference level from the used” is among the key points. Today’s U = a + a U + a U 2 disturbing IT equipment. tendency of increasing use of the radio out 0 1 in 2 in + a U 3 + ... + a U n (1) spectrum for personal communication 3 in n in As seen from Table 3, the most exposed needs highlights the electromagnetic pol- radio system is in the lower frequency and insert the AM equation lution, and an increased effort of “clean- band. As mentioned earlier, this band is ing up” the radio spectrum is needed. U = U cos(ω t)(1 + mcos(ω t)) (2) also very attractive from a radio trans- in o o m mission point of view. New systems, where ω represents the carrier fre- especially the digital ones, need more o quency, m the modulation index and ωm

20 the modulation frequency, one term in References Uout the output signal is 1 Hauger, E, Eriksen, J. ISDN-termi- a mU 2cos(ω t) (3) 2 o m naler som potensielle støykilder for kringkasting. Kjeller, Norwegian This is the demodulated signal where we Telecom Research, 1991. (Report TF have the quadratic function from input to R 50/91) output. In [2] we have done interference mea- 2 Hauger, E. Interference from the surements on 12 analogue telephone sets. TDMA structure in digital mobile The reason for these measurements was communication to PSTN. Kjeller, the audible 217 Hz interference from the Norwegian Telecom Research, 1992. GSM telephone. As shown in Figure 4 (Report TF R 40/92) the TDMA structure in GSM generates most of its energy in the lowest audio Uin band where an ordinary telephone is most sensitive. In Figure 5 we see the Figure 3 Amplifier in/out characteristics 217 Hz interference and its harmonics together with the reference 1000 Hz tone with level -10 dBm. The measurements dB are done with 0.8 W GSM peak power 3 0 m away from the PSTN telephone. In this instrument set-up with perfect reflecting -4 ground the electric field exposure to the test object is 3.7/3.9 V/m for vertical and -8 horizontal field. -12 When measuring the interference as a function of the GSM RF effect, we see -16 the quadratic function mentioned above: 1 dB increase in RF level gives 2 dB -20 increase in audio interference. This compatibility problem is due to the -24 missing immunity standard for PSTN terminals. The GSM system operates as -28 intended, the TDMA structure gives the 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 amplitude pulse modulation and from an kHz EMC point of view this problem should Figure 4 GSM TDMA frequency spectrum have been foreseen. Hearing aids have the same problem as telephone sets. A person wearing the hearing aid cannot be a customer of the GSM telecom operator, he may use the car-mounted mobile telephone set, but a handheld GSM telephone may give a field strength exceeding 30 V/m to the hearing aid. The coming immunity recommendation for teleterminals will be 3 V/m. To achieve 30 V/m a completely new design for hearing aids is required. Here we see again the compatibility problem with the penetration of personal radios and electronic equipment where the designer has paid insufficient attention to the RF layout or filtering. The problem can only be solved by a total EMC concept where all the aspects of the electromagnetic environment are taken care of before launching new radio systems.

Figure 5 GSM TDMA interference in analogue telephone set

21 Design of electronic equipment with EMC in mind

BY AGNAR GRØDAL

Introduction Specification - Shielding The design process of electronic equip- The starting point of a design process - Screened connectors and screened ment has traditionally concentrated on should always be to write the product cables making a reliable product according to its specification. The product shall meet - Extra “real estate” on the printed cir- functional specification, and at a compet- mainly two specifications: cuit board. itive price. - The functional specification, and - The EMC specification. A typical EMC specification will, as an Tests have shown that design along these example, dictate EN 55022 A or B for traditional guidelines often fails the EMC It is important to incorporate the EMC emission, and the IEC-801 (EN 61000-4) Directive. specification at this early stage, because series for immunity. It is important that both specifications will influence the log- the relevant immunity severity level is The multitude and complexity of elec- ical design, the choice of components, indicated. tronic equipment in modern society have the number of layers in the printed circuit also led to numerous reliability problems boards, and the enclosure. and to accidents caused by unintentional Components interactions between electronic equip- Early knowledge of the EMC specifica- The EMC challenge has come almost ment of different types. tion will also guard against a possible simultaneously with a radical shift in costly overdesign. component technology. The reason for The EMC Directive has almost simul- this is mainly new requirements from the The EMC specification will be given in taneously come with a change in compo- surface mount (SMD) technology. The terms of norms and limits relevant to the nent and production technology, namely integrated circuit chip technology has product type. It is important to realize the surface mount technology. This tech- not, however, altered as much as their that these norms and limits reflect mini- nology gives important technical bene- encapsulation. mum performances for the equipment to fits, but it may also give rise to problems pass legal requirements according to the Passive components, mainly non-elec- with respect to EMC. EMC Directive. trolytic capacitors, have undergone the most radical change. The reason for this Design for EMC thus imposes new Practice indicates that the EMC specifi- is the massive change from hole-mounted restrictions to the design process, and cation for some product types has to be plastic film capacitors to ceramic capaci- more and better competence is required more stringent than what is required to tors. Some ceramic capacitor types have in order to succeed. pass the legal minimum limits. The in- parasitic properties that make their creasing use of mobile radio telephone capacitance change radically with tem- During the last few years we have been systems has clearly shown that the perature and applied voltage. Wet elec- through a period where the design app- immunity level of 10 V/m, which is the trolytic capacitors for surface mounting roach has been to concentrate mainly on level most systems are designed for, may represent another problem, as these may meeting the functional specification, then be far too low for certain types of elec- be harmed by the soldering process. to test the prototype and hope for the best tronic equipment. This fact will impose with regard to EMC. This trial and error new and even more stringent restrictions The change in component technology has method inevitably leads to costly time to the design process. unfortunately led to a decline in the cir- consuming testing, fixing, new testing, cuit designers’ understanding of which The specification shall be tailored to CE- and so on. component types should be used where marking of the product. CE-marking in the circuit design. In many companies demands that all relevant requirements Lack of EMC experience is the main rea- there seems to be a shortage of qualified shall be met, not only the EMC require- son why this design method has been component experts. The component work ments according to the EMC Directive. used. Experience over the last few years has in some cases been reduced to mere A common additional requirement is that has shown that we today have a better registering and purchase. Using primarily the safety norm EN 60950 shall be met. understanding of the factors that will CECC approved components is a good For equipment connected to the public influence EMC. With this knowledge it is help in such cases. telecom network, the Directive on now possible to design economical com- Telecommunications Terminal Equip- The following list exemplifies new types petitive electronic equipment that has a ment 91/263/EEC applies. This directive of components used in modern logic fair chance of passing the relevant EMC poses requirements on EMC and safety, designs: test criteria, even at first attempt. but also demands that the equipment - SMD components shall be produced according to a quality It may seem that too much time and standard, normally ISO 9001. - Components with small pitch (distance money is spent on discussing EMC stan- between terminals) dards and levels of emissions, relative to The specification should preferably be the time spent on discussing how to meet written in co-operation with the client. - Components with low output these standards and levels of emissions Alterations to the specification later in impedance, producing strong switch- by using adequate design principles. It is the design process can have severely ing current transients true that there is more broad knowledge costly and time consuming conse- - Components with high edge rates available on standards and limits than quences, and may require: there is in general, effective EMC design - Components with high output level, - Extra filter components principles. Both disciplines, however, switching rail to rail must be dealt with. - Components for overvoltage protection

22 - Components using high clock rate, - Avoid using Advanced CMOS compo- limited to pass the relevant tests; hope- typically up to 50 MHz nents, especially byte-wide types fully without the use of costly mechanical screening. - Microprocessors with wide buses - BICMOS types are preferable when high speed and high drive capability is - Microprocessors with high bus activity. The latter method is by far the cheapest needed and the most effective, even if some Some of these components have qualities - Low voltage types are favourable additional screening and filtering have to that are advantageous for realizing logi- be used. - Use special components for clock cal circuits with advanced system specifi- phase distribution The use of extensive screening may, cations. But the same qualities may also however, work. This may even be cost be disadvantageous with respect to EMC. - Use special components for driving effective when trying to prolong the sale CMOS integrated circuits from the backplanes. of an older design, but the method has its “Advanced” series is a typical example. limitations. The enclosure will normally These circuits are capable of driving high Be aware of the prices, components with be more expensive, it may introduce capacitive loads at very high clock rates, favourable EMC attributes may be cooling problems, and worst of all; the and are often used in fast microprocessor expensive! strong internal emissions may sneak out bus circuits, but they are very noisy. along the cables. The change some years ago from Low General sources of EMC Designing electronics with little radiation Power TTL circuits to High Speed problems requires wide competence within elec- CMOS circuits was a great success. Prac- The causes of the EMC problems can tronic design and manufacturing, and it is tice has shown, however, that these are simplified be derived from Maxwell’s of utmost importance that involved noisier than the corresponding LSTTL equations as: departments and individuals co-operate. circuits, but the EMC threat was not an issue at the time of shifting from LSTTL Emission The component price for a low-emitting 2 to HCMOS. = K1 x Area x Current x Frequency design will in most cases not be signifi- cantly higher than for a comparable high- The widespread use of the newer Susceptibility er emitting design. The difference in Advanced CMOS circuits has caused so = K x Area x Field strength x Frequency 2 design time for the circuit diagram will many and serious EMC problems that normally also be insignificant. many designers would rather use older See Figure 1. TTL circuits, were it not for the TTL The printed circuit layout on a low-emit- Observe that the term “Frequency” power dissipation problem. ting board will normally take longer to appears squared in the equation for the design, and the layout designer must be Help has come in the form of the BI- emission. EMC competent. The board will nor- CMOS circuit families, circuits with low The term “Frequency” in the equations mally also require more layers than a power CMOS “kernels” and with TTL includes both fundamental frequency and conventional two-layer board. Low-emit- low noise and high drive capability. The its harmonics. A 1 MHz square-wave sig- ting boards will thus be more expensive. shift to high speed 3 V families will also nal may thus emit noise up to tens or be a success with respect to EMC. Some years ago, before the “discovery” even hundreds of MHz. The task of of EMC, the electronics designer and the Some newer component types have been reducing the emissions part of the EMC circuit board designer would perform designed specially to achieve better EMC problem can be simplified as: their tasks without even knowing one qualities. Examples are: - Reducing the Frequency another. Today, they have to co-operate - Components with short terminals, giv- - Reducing the Current closely to succeed with EMC. That also ing lower parasitic inductivity - Reducing the Area. applies for the departments or persons responsible for the EMC-, component-, - Components with controlled output It is in fact, at least theoretically, that and design review activities. pulse edge rates (no sharp pulse cor- simple. ners) All relevant expertise needed to succeed is seldom found within one company. - Components with better V and CC Solutions Help must often be hired from external ground pinout (centre pinout instead of Whereas the theory for reducing the EMC consulting firms specializing on EMC corner pinout) problems may seem simple, practice may issues. - Components with low VCC voltage be more complicated. The problems may (3.3 V) in principle be solved by either:

- Microprocessors with separate VCC - Designing the electronics without pay- pinout for kernel and bus drivers. ing attention to EMC design rules, but Frequency Current Emission instead to encapsulate the emissions Some general advice for selecting com- within the products’ enclosure by ponent types: using screens, tight metal boxes, filters Area on the cable outlets, and by using - Use fast, low output impedance com- Field strength screened connectors and cables, or by ponents only where necessary for the (Susceptibility) function of the circuit - Designing the electronics so that the generation of emissions is sufficiently Figure 1 Simplified illustration of Maxwell’s law

23 The circuit board designer may have equipment shall and will be tested emitted noise. As a rule, liquid crystal automatic routing programs at his dis- according to levels specified by regula- displays are less noisy. posal, which pay attention to EMC to tions. Real life conditions may expose - Do not use a switch-mode power supp- some degree. Practice shows that most the equipment to even higher levels. ly if a linear power supply can be cost autorouting programs so far are of lim- It is also important not to overdesign the effective. Switch mode power supplies ited value, because they very often gener- protection, especially on boards connect- are clock controlled, always noisy, and ate layouts with poor EMC performance. ed to a multitude of external signal lines. must in most cases have extra filtering Practice has shown that a moderate pro- components. This is an example where Self compatibility tection circuitry may be sufficient, even a clock controlled design may be Newer component technology with lower when calculations point to the use of replaced by a non clock controlled output impedance, lower rise and fall more costly and larger components. design. times, higher signal levels, higher tran- - The use of optical fibre cables are sient currents, and smaller distance be- Emission, principles of always favourable with respect to tween signal tracks on the printed circuit EMC. They do not carry currents and boards, have given rise to numerous design will thus not emit noise. They will also problems with internal noise on the Alternative equivalent functioning circuit be immune to overvoltages and to PCBs. As a consequence, circuits may boards, but using different design princi- external electromagnetic fields. disturb one another within the same PCB. ples or using different components, may A board that disturbs its own function is have a different noise emission spectrum. said to have low self compatibility. The difference in noise level may be con- Emission, design details siderable. Price and power dissipation Such problems are often marginal, and Many corrective circuit details will affect may differ a lot, too. may come or go with a slight shift in the the EMC performance of an electronic power feed voltage, with a shift in ambi- It is of prime importance to choose design. Some of them may be very influ- ent temperature, with ageing, or with the design principles that have the lowest ential, and some only marginal. Some of use of “identical” components from a possible price, the lowest possible power them may be helpful over a broad band “second source” vendor. dissipation, sufficiently low emissions, of noise frequencies, and some only at and sufficiently high immunity. One discrete frequencies. Some examples are: Such circuit malfunctions may not be should not aim for a better EMC perfor- detected by simulation or testing. The - The capacitive circuit loads should be mance than what is needed, as that will faults may be detected or avoided by minimized. Noise from signal tracks normally have consequences for price on the printed circuit board is generat- - Applying good EMC oriented design and power dissipation. Some alternative ed mainly because of the charging and rules design principles with different EMC discharging of the parasitic capacitive qualities will be mentioned: - Employing an EMC competent design loads as the signal is pulsing. reviewer - Do not use a higher microprocessor - Signal tracks having high capacitive clock frequency than what is neces- - Using more and better simulation tools loads should be as short as possible to sary. This is perhaps the most common make the noise area as small as possi- - Employing an EMC competent circuit design error. Designers tend to use the ble. board designer, paying particular atten- same clock frequency as they did in tion to component placement and how their previous designs, without check- - Capacitive loading should normally the tracks are routed. ing what is really needed. Because the not be higher than what the compo- noise depends on the square of the fre- nent data specify. Component delay quency, the noise will be reduced by data are given as a function of their Immunity/susceptibility 75 % – at least theoretically (and simp- capacitive load. If the load is higher, The words immunity and susceptibility, lified) – if the frequency is halved. the delay will increase, and the delay despite having opposite meanings, are data will no longer be valid. This may - If possible, use a microcontroller used to describe an electronic system’s result in circuit malfunction. instead of a microprocessor. The behaviour when exposed to overvoltages microcontroller has the bus system - Noise generating currents in the board or to an electromagnetic field. built inside the processor chip. The bus signal tracks may be reduced by insert- Insufficient immunity may lead to per- system radiating area, which is usually ing small series resistors at the signal manent destruction of one or more com- one of the noisiest parts of the design, driver outputs. Observe that these ponents, which requires manual repair will then be greatly reduced. resistors will result in a slight increase for the equipment to resume normal of the signal delays. - Do not use a clock controlled design if functioning. Less serious failures may other design principles can be utilized. - Series resistors may be useful on the lead to a manual reset or restart of the This is perhaps a naive suggestion, as most noisy bus signals; the read and equipment. Immunity can be designed most of today’s designs use micropro- write signals, and on the address bus’ into the system by using the same EMC cessors. least significant bit position A . This rules as used to limit emissions. In addi- 0 bus line has the highest frequency, tion to that, it may be necessary to add - Do not use multiplex controlled light because the address bus is incremented special, board space consuming, over- emitting diode matrices. These are in the normal binary code. The signals voltage protection components. It is often used in displays, and have large, mentioned are often noisy because important to be aware of the fact that the often unscreened areas carrying rela- they are routed to many components, overvoltage protection circuitry of the tively high current transients, causing

24 and thus have a large noise area and a load when its control current is shut - For the bus noise area to be as small as high capacitive loading. Observe that off. The protection may consist of a possible, the component placement series resistors may distort the signal shunt diode connected over the load. should be accommodated to a short wave form, especially if the parasitic The transient pulse may generate a and straight bus. capacitive loads are distributed along strong noise field. This can be limited - The placement and orientation of the the signal track. by mounting the diode close to the components should be accommodated load, to make the transient current loop - All connections from drivers to receiv- to a short and straight signal track lay- as small as possible. ers shall have correct voltage level out. matching. This is often ignored in - Noisy transient pulses may be limited - Low output impedance components modern designs where level incom- using components with TTL output should have short output layout tracks. patible bipolar and CMOS circuits are levels. This is especially valid for buffer-cir- used. The result is poor self compati- - Design the overvoltage protection as cuits of the “Advanced” types that are bility, which often may pass both sim- inexpensive as possible. The testing, capable of driving more than 100 mA ulation and test, whereafter circuit not the dimensioning, should dictate its into the receivers’ parasitic capaci- malfunction may appear after the prod- complexity. tances. uct is sold. - SMD capacitors have low parasitic - The decoupling capacitors should be - Correct impedance matching between series inductivity ∆L. They are there- located close to the ground connection driver and receiver will guarantee fore effective as high frequency noise of the ICs. Each and every IC should undistorted pulse shape, and thus cor- reducing decoupling capacitors. have its own decoupling capacitor. rect functioning of the circuit. Noise transients generated by switch- Impedance matching is often obtained - The use of discrete SMD transistors ing of the components can be substan- using pull up and pull down resistors at instead of transistor array – a multitude tial, but is easily controlled by ade- the receiving end of the connection. of transistors in an IC – will often quate decoupling. This will normally increase the signal result in a cheaper, smaller and less current, and may thus result in a higher noisy coupling area. - The analogue part of the circuitry noise level. A distorted signal can be should be separated from the digital - Avoid the use of IC sockets. They acceptable if it has stabilized before part. Intermixing analogue and digital have inductive pins that may introduce being clocked into further circuitry. In components is a common cause of a higher emission, higher cost, and such cases there is no need for costly, functional noise in the analogue sys- blocking of the air cooling. possibly noise generating, impedance tem, and of common-mode emissions. matching. - Limit the use of filtering components. - Interface circuits should be located Use as few, inexpensive, and as small - Avoid bus contention. Tristate bus close to its connectors. Noise from the components as possible. It is normally logic, where the output of two or more interface part of the logic often results not easy to determine the need for fil- circuits are connected together, may in radiated emissions from the cables, tering components at the outset. One generate heavy noise if the two circuits which can be hard to eliminate. may therefore possibly use too many, are delivering signals to the common and afterwards remove those not need- - Cable connector filtering capacitors output at the same time. The situation ed according to the test result. should be located very close to the may occur as a short-circuit glitch connector. The filtering performance when control of the common output - Filtering components are often needed will be substantially reduced if the dis- change from one driver to the other, close to the cable connectors. It is very tance between capacitor and its corre- and if one driver delivers a “0” and the important that they have short leads, sponding connector pin is increased other a “1”. The situation can easily be preferably SMD, and that they are con- only a few millimetres from its optimal avoided utilizing proper design princi- nected to a “clean” ground reference. location. ples. - The power supply voltage should be - The components for overvoltage pro- - Outputs from octal circuits – eight bit filtered on each card to avoid HF noise tection should be located close to the bus drivers – should not be used to to contaminate the backplane. line connector. This will increase the control edge triggered components. - All cabling to the equipment, except chance of correct functioning even The reason is the “ground bounce” the mains cable, should in most cases under exposure to overvoltages. phenomenon, which causes outputs to use shielded cables. have transients that may trigger edge - The components should be placed so triggered components. that even cooling over the board area is obtained. The cooling problem result- - Counters, such as the 74HC161, may Placement of components ing from the introduction of surface cause malfunction if their carry-output A sound placement of the components on mounting technology and the use of is used to control edge-triggered com- a board is vital to obtain a low-noise higher frequencies, has led to the need ponents. This is misuse of the counting printed circuit layout. But rules for low- of forced air cooling in many new component because of the decoding noise component placement may conflict designs. transients on its carry-output. with placement rules to obtain self-compatibility, and with placement rules to - The location and orientation of the - A circuit that controls an inductive obtain even cooling of the board. The components should be adapted to the load must be protected from the strong following is a listing of EMC-effective soldering process. This may be a prob- transient pulse that will be generated component placement examples: lem when components are to be by the stored inductive energy in the

25 mounted on both sides of the board; - Digital signal tracks should run only tight time schedules it may sometimes be this mainly because wave soldering, over digital power and digital ground, advantageous to first make a multilayer which is often used on the secondary and analogue tracks only over ana- board, as this will have a higher chance side of the board, may accept only logue power and analogue ground. Not of success getting the product as fast as resistors, capacitors, and transistors on observing this rule will increase the possible on the market. Later, when the that side. risk of malfunction and of common- product is selling, one may try to cut mode radiation. costs using fewer layers. Automatic routing of printed circuit lay- Printed circuit layout The main goal for the PCB designer is to out tracks, “autorouting”, has for some design a correct functioning board which design years been advantageous because of fast also meets the EMC specification at first The routing on a board consists of the results and low cost. Autorouting has, attempt. However, redesign is necessary signal tracks and of the power and however, traditionally not taken enough more often due to changes in the board ground system. It will be based on the care of EMC considerations, and practice specification, than as a result of insuffi- location of the components, and its task has shown that most designs based on cient self-compatibility or because the is to ensure the boards’ self-compatibil- such programs have resulted in systems EMC specification is not met. ity, so that it will function reliably. It that do not meet the relevant EMC speci- shall also be designed to secure the EMC It is interesting to observe, though, that fications. system specification. Its main goals the client usually accepts a layout that Autorouting programs that do take EMC should be: results in a board that does not function into account have to some degree been properly because of imperfect signal - All nodes of the schematic diagram available for some time. They are expen- wave shapes, or the board does not meet must be realized in the printed circuit. sive, and to achieve acceptable results the EMC specification, even if the layout with respect to EMC, they need the same - All signal tracks should have return designer should be to blame. The PCB component data, and other relevant EMC tracks that follow the signal tracks as designer is normally not made responsi- data as is used by “manual” PCB design. closely as possible. A signal track and ble for delays, and in addition he is even The result may be acceptable, but not its corresponding return path will form usually trusted to make a new try. When always cost effective. a noise radiating loop which has to be do we have a PCB designer who would minimized. The return path may be a guarantee his design, both with respect to It is not advisable to use autorouting pro- dedicated board track, or a less well a functioning board and to the EMC grams that do not take EMC considera- defined current path in the ground specification? tions properly into account. A typical plane. result is more via-holes, longer tracks, In his strive for success, the PCB design- and less control with critical signals with - The signal tracks should have low er may be tempted to use more board lay- respect to EMC. cross-coupling. This demands that long ers than necessary. The board, however, adjacent signal tracks should be should only have as many layers as the The result may be: avoided. The introduction of the SMD component density and the signal and - The high via-count will degrade the technology with closer component power track density demands. Only when uniformity of the power and ground connection pins (fine pitch), denser the EMC restrictions dictate more layers, planes, generating ground impedances component packing on the board, finer this should be done. This is because the and force the return currents to take tracks, and with higher system fre- board price increases as the number of uncontrolled return paths. quency, have increased the cross-cou- layers goes up. Four layers have twice pling problems. the cost of two layers, and for each suc- - Higher printed circuit board produc- cessive pair of layers the cost goes up by tion cost - Some signal tracks must be some 40 %. impedance-matched to its transmitting - Lower printed circuit board production and receiving components. This can It is not possible before start of a design yield fairly easily be accomplished by an to figure out how many layers a certain - Poor self compatibility of the logic cir- advanced CAD system, and normally board will need for the complete system cuit results in wider or narrower tracks than to pass the EMC test. If the system those used for connections without passes the test, it is seldom checked if - Uncontrolled and higher radiation impedance matching. one or more boards could have had fewer - Low overvoltage immunity layers. - The tracks must have sufficient insula- - Risk of more PCB redesigns tion distances. This may be a problem In most cases at least four layers will be when the board carries higher voltages needed. Practice has shown, though, that - Risk of costly repeated EMC measure- than normal, such as the mains volt- an EMC layout expert who co-operates ments age. with the circuit designer, may in some - Risk of introducing delays in the cases succeed with only two layers. For - Multilayer boards should have as few development phase low volume production, the board price vias as possible. Vias increase the has less significance justifying a higher - Risk of higher total development cost. board cost, the risk of scrapping the layer count (e.g. six layers). For medium board, and they reduce the quality of volume one may try to succeed with four The above listing may seem somewhat the power and ground planes. They layers, and for large volume, or if the exaggerated and indeed it is. Manual also introduce added inductivity to the EMC specifications are minor, one could designs are sometimes even worse, but tracks. try to succeed with only two layers. With

26 the listing illustrates the involved risks should be routed separately on one or when using autorouting. both sides along the entire clock signal 1 track. A mix between manual routing and autorouting, and between manual place- - Connections having high capacitive ment and “autoplacement” may give load. All signal receivers have a para- Area good results. Programs for autoplacement sitic capacitance to ground. Signal and autorouting are continually being voltage pulsing will generate noise refined. current transients that charges and dis- 1 charges these capacitances. Signal Some general layout rules lines having many destinations will Figure 2 Signal loop noise area have multiple capacitive loads, and The printed circuit tracks shall form corresponding higher noise current smallest possible current loop areas, see transients. Typical connections of this Figure 2. type will be clock signals, and read When discussing current loop areas, one and write signals. tends to forget that the total noise loop - Connections having low impedance 5V area covers the complete area, from the drivers. Most digital circuits have an power input through the signal track and output impedance at 5V V that back to the power output, see Figure 3. CC make them capable of driving signal 1 The noise areas in boards having “com- noise transients of up to some 25 mA. plete” power and ground planes (they are Fast buffer circuits from the never quite complete, due to lots of “Advanced” series, or from the BIC- Area through holes), will always be substan- MOS series have output impedances tially smaller than in a comparable two- well below 100 ohms at 5V VCC, and layer board. This is because the return are thus capable of driving signal noise 1 currents will obey Lenz’ law and run as transients of up to 150 mA or even close to the signal currents as possible, higher. All connections from these cir- thus reducing the area and the noise as cuits should be short and straight, and 0V much as possible (for once, the nature of have a nearby ground return path. Figure 3 Two-layer board. Large noise area electricity is on our side), see Figure 4. - All bus connections contribute signifi- The noise level from a four-layer board cantly to the total noise picture. The having power and ground planes and two address bus normally represents less signal layers, is some 20 dB lower than problems than the data bus, and is also from a comparable two-layer board. the easiest to cure. The address bus has usually only one source, and has also 0V All tracks on a printed circuit board, and 5V the lowest parasitic capacitive load, all return paths will contribute to the total whereas the data bus normally has noise spectrum and noise levels. When many alternating sources and a higher designing a low-emitting board it is not 1 parasitic capacitive load. The data bus possible to look at all noise generating has therefore a stronger and more paths. One has to give priority to those untidy noise spectrum. All buses Area considered to contribute most to the should be short and straight, and not noise. A PCB designer may not be located near the board edges. The skilled enough to know which tracks that ground return should be good, prefer- will contribute most to the noise, and ably a ground plane with as few via- 1 when making up his priority list he has to holes as possible. co-operate with the logic circuit designer. A priority list may look as follows: Decoupling of the integrated circuits with Figure 4 Four-layer board. Small noise area - Connections having high pulse fre- capacitors was originally used to ensure quency. The track carrying the highest the correct functioning of the logic cir- system frequency is usually the most cuits, and the capacitors were spread over noisy one because the noise emissions the board according to more or less well are proportional to the frequency defined rules. The change from TTL to squared. The clock frequency is usu- HCMOS requires a more stringent prac- ally the signal that has the highest fre- tising of those rules, mainly because mal- 5V 5V quency. It is of utmost importance that functioning of the logic circuits occur the clock signal has a ground return more and more frequently. The reason path that runs as close as possible to for this is the higher voltage pulse levels, 0V 0V the clock signal track. On multilayer and the faster edge rates generated from boards this will normally be ensured HCMOS compared to TTL. by the ground plane. On two-layer Better Decoupling rules should now be one boards, the clock signal return path capacitor for every integrated circuit, and Figure 5 Location of decoupling capacitor

27 5V 0V DSUB

0V 0V

5V 5V

Figure 6 Ground return current contaminating connector area

5V 0V DSUB

0V 0V

5V 5V

Figure 7 Diverting ground return current

for large ICs one should use several The decoupling transient current running The use of electrolytic decoupling capac- capacitors. The reason for this is the from the capacitor to the 0 V termination itors (LF-decoupling) was originally EMC problem, and that insufficient of the IC will have a short path if the introduced to stabilize the VCC of the decoupling will readily be observed on capacitor is located close to the 0 V ter- board, and to thus secure the correct the emissions plot. mination of the IC. This current will thus functioning of the board. contaminate the 0 V plane or the 0 V dis- The type of decoupling capacitor and its The high density of integrated circuits on tribution layout as little as possible. This location with respect to the IC is also modern printed circuit boards, and the will reduce the mixing of the transient very important. Capacitors having long use of one decoupling capacitor for every ground current with signal return currents leads, or being located a couple of cen- IC, often results in a total HF decoupling running in the close vicinity. This is an timetres from the IC are of limited or no per board of several microfarads. This example of how to keep the 0 V distribu- use. It is also important to locate the value is often sufficient LF decoupling, tion or the 0 V plane as clean as possible, capacitor close to the ground connection and because an electrolytic capacitor has which is believed to be very important in of the IC, not close to the power connec- no EMI-reducing function, it may be the strive to reduce the generation of tion, see Figure 5. omitted. It may have a function, though, common mode emissions.

28 5V 0V

Figure 8 Hindering ground currents in DSUB connector area

5V as a stabilizer of the VCC if an Intentional use of slits in a overvoltage should occur. Prac- ground plane may be used to tice has shown that a micropro- divert ground return currents cessor based system may fail due 1 1 off from a noise sensitive area, to electrostatic discharge if the see Figure 7. electrolytic LF capacitor is omit- Unwanted ground return cur- ted. 0V rents in the ground plane close It is on two-layer boards, as well Figure 9 Ground return bridge over slit in ground plane to a DSUB connector can also as on multilayer boards, impor- be avoided by a slit in the tant to know where the signal ground plane around the con- return currents will flow. The return cur- nector pins, or simply by removing the a noise emitting area, or the current may rent paths will be difficult to localize in ground plane in that area, see Figure 8. be forced into a noise sensitive area ground planes. They will spread out to where it may generate circuit malfunc- Slits in the ground plane made up by a some more or less predictable degree, tion, or where it may generate common row of consecutive circular insulation although they will obey Lenz’ law and mode emissions. A typical example is a areas, as in Figures 6 and 7, can be run as close as possible to the signal situation where a return current is forced avoided if the circular insulation areas tracks. to run in the ground plane close to a are made small enough to let the current If the natural return path in the ground DSUB connector, where it will generate pass through. Another method is to plane is broken by a slit, as formed by a common mode emissions that will be bridge the slit by a ground current strap row of consecutive through holes, the conducted by the cable connected to the close to the signal current track, see Fig- return current may be forced to run DSUB connector, see Figure 6. ure 9. around the obstruction and thus produce

29 Electromagnetic compatibility between power networks and telecommunication networks

BY KNUT J RAMLETH

1 Introduction Provided both source and receiver are nected to noise have not earlier been as working and operating as intended, it strictly regulated, but have now been Transmission of electric power as well as seems likely that the coupling between highlighted through the introduction of telecommunication signals on galvanic source and receiver should be reduced in the European EMC Directive. cables are necessary functions in a mod- case of compatibility problems. Coupling ern society. Transmission may take place and physical distance are usually two on the same cable pair or on separate 3 Structure and character- sides of the same issue, so an increase in cable installations. This article will dis- distance will reduce the coupling. On the istics of the power net- cuss the problems of compatibility other hand, there are the general environ- between separate cable installations. We work mental requirements to keep the number will further elucidate some problems on of cable routes and encroachments on 3.1 Alternating current noise and dangerous voltages which the nature at a minimum. Economic consid- networks power network may cause in the termina- erations also favour common cable tion points of the telecommunication A distribution network for electric power trenches, common poles and common cables. may in principle be constructed as out- routes. Besides, the telecommunication lined in Figure 1. Power networks, electrified railways and network is necessary for operating both telecommunication networks are com- the power network and the railway, and A three phase alternating current can be petitors in the selection of cable routes. finally, the telecommunication operator transferred on three separate conductors Legally, the three parties are equal in that and the power supplier have largely the as shown in Figure 2. the Principle of Priority (“First come, same customers and users. In other In a balanced three phase system the vec- first serve”) forms the basis for financial words, a close neighbourhood between tor sum of the three phase currents is settlements. A case of the telecommuni- cable networks and between apparatus zero, and the current in the three phases cation network acting as a noise source to and equipment cannot be avoided in seen together will therefore have no mag- power networks or electrified railway practical life. netic influence outside the near zone. operation happens rarely or not at all, and The lack of compatibility between power in questions of compatibility between the The example shown in Figure 3 repre- supply and telecommunications affects three parties the telecommunication side sents an ideal case. In practice, there will both disturbance in the telecommunica- is therefore always the “victim”. always be degrees of unbalance in the tion network and safety of personnel and vector diagram because of varying loads equipment connected to telecom services. (current outlets) on the three phases; the 2 General Safety is regulated by Norwegian law worst case being when there is an earth- Any problem of compatibility comprises through “Forskrifter for elektriske fault on one of the phases. The current in the three main elements forsyningsanlegg” (Regulations for elec- an unbalanced three phase network can tric power supply plants) and “Forskrifter - source (emitter) mathematically be decomposed into for elektriske bygningsinstallasjoner - receiver (victim) m.m.” (Regulations for electric installa- - a symmetric three phase system with - coupling path. tions in buildings, etc.). Problems con- normal phase sequence, the plus- sequence system - a symmetric three phase system with opposite phase sequence, the minus- G sequence system - a third three phase system where all three currents are in phase, the zero- Generator High voltage High voltage Low voltage transmission distribution distribution sequence system. Figure 1 Transmission of electric power

Neutral conductor/earth wire I L1 L1

o o Phase conductors 120 120 L2

120o L3 I L3 L2 Transformer Transformer Figure 3 Phase currents in a balanced Figure 2 Three phase system three phase system

30 The three decomposed systems are illus- 3.2 Direct current networks profitable to use direct current by intro- trated in Figure 4. ducing rectifiers/inverters at the cable The capacitive loss of power in long termination points. This applies e.g. to The individual plus- and minus-systems cables in alternating current network is the exchange of power between Norway are balanced three phase systems which considerable. As an alternative trans- and Denmark running through a 120 km do not cause magnetic disturbance to the mission system it would therefore be long sea cable between Kristiansand and surroundings. The zero-system, however, represents all the unbalance because the zero-sequence currents must return to the source (generator/transformer) through the earth and/or a possible fourth conductor (top line). In the case of an earth-fault on one of the phase conductors the sys- 3 3 tem earthing of the power network will 1, 2, 3 therefore greatly influence the effect on the telecom network in the area. The part 1 2 of the fault current which returns to the 1 2 source through the top line is in opposite phase to the inducing fault current and thereby has a reducing effect on the in- I+ currents I- currents I0 currents duced voltage in the telecom network. The resistance in the top line will be Figure 4 Decomposition of currents in an unbalanced three phase system decisive for its “reduction factor”. A well conducting top line has a good reduction effect; aluminium conductors in the top line will in other words be preferable to steel conductors. The reason why steel lines are still used is the relatively poor mechanical tensile properties of aluminium. In systems with the neutral point connected to earth, ref. Figure 5, the fault current (3 x Io) may be quite large, but as previously mentioned, the inductive effect on nearby telecom cable may be attenuated when the high voltage towers are connected with continuous and well Top line/earth wire conducting top lines. Tower Fault current (3xIo) A large earth current induces high volt- earthing ages in nearby telecom cables. However, Earth current the large fault current makes it technically possible to achieve a quick and safe Figure 5 Fault currents generated from an earth-fault in a three phase network with automatic disconnection of the fault, and directly earthed neutral the effect on the nearby telecom network will therefore be brief (less than 1 sec.). In a transmission system with isolated neutral the earth-fault current will normally be small. The whole of the fault current must return to “the healthy phases” through the conductor capacities, ref. Figure 6. This will normally limit the fault current to a few tens of Amperes. A small fault current in the power network will in turn induce small 50 Hz voltages in nearby telecom networks. However, the combination of a small Fault current (3xI ) fault current and existing relay technique C1 C2 o in the power network will render it impossible to achieve an automatic and C1 & C2: Phase (to earth) quick disconnection of the fault. The capacities duration of this type of earth fault may therefore be hours and days. Figure 6 Fault currents generated from an earth-fault in a three phase network with isolated neutral

31 electrode and may cause voltage in the nearby telecom network Sea cable 1 electrolytic corrosion of will occur. Particularly high inducing metal sheathed telecom currents are seen in connection with earth cables in the ground. faults on one of the phase conductors. The inducing fault current may then Direction of 4 Noise amount to tens of kA and cause danger- current ous voltages in telecom networks several sources in kilometres away from the power net- the power work. On the spot where the fault current goes to earth, ground potentials of sev- network eral thousand Volts may occur posing a Rotating AC generators threat to telecom networks in the area. and power transmission Direction of Agreements between the telecom and systems with transform- current power supply parties in Norway are ers and cables do not founded on the limits stated in the recom- themselves produce mendations in the CCITT’s “Directives noise in nearby telecom concerning the protection of telecommu- networks. It is power Sea cable 2 nication lines against harmful effects consumption and coupling tech- from electric power and electrified rail- Figure 7 Co-operation of 2 DC cables niques which “pollute the sinus curve” way lines.” These directives have been that cause the power network to act as a prepared in a co-operation between the coupling element between noise source International Telecommunication Union and the telecom network. As mentioned Jutland. Sea water may in principle act as represented by the CCITT, the Interna- above, noise sources are found in large return conductor, but as the loss of tional Union of Railways and the Interna- rectifier plants for power transmission, energy in sea water is considerable – in tional Conference for Co-operation in but also in similar plants for heavy indus- this case in the megawatt area – a cou- High Voltage Systems. Without further try. Static frequency converter plants also pling as shown in Figure 7 is used. national discussions, the stated limits produce noise in connected networks. have therefore been acknowledged by Polarities are chosen so that the return Other examples of pollution sources in NSB (Norwegian State Railway), the currents for the two DC cables are invert- the power network are power regulators power suppliers and Norwegian Telecom ed in the sea area between the sea elec- based on semi-conductors, coupling as a guide for determining when protect- trodes. If the cables have equal current mechanisms like thermostats, relays, etc., ive measures have to be implemented. load the current in the sea is zero, and the power switches for switching on and off transmission losses are thereby min- The following limits are used: large power flows, as well as occasional imised. faults involving poor or varying degree - 60 Veff permanent longitudinal volt- The rectifier/inverter installations in the of contact and arcing. Corona discharge age in a telecom cable. The limit may transmission system are a source of on conductors and coupling equipment be increased to 150 Veff in certain major noise problems to the telecom net- may sometimes produce interference in cases, provided parts to be handled are work. The thyristor-based the radio frequency band. clearly marked by danger signs. rectifier/inverter installations generate - 430 Veff longitudinal voltage lasting considerable amounts of harmonic com- 5 The power network as from 0.5 to 1 sec. ponents in the audible frequency band, and the introduction of effective smooth- a source of dangerous - 650 Veff longitudinal voltage lasting ing filters after rectifying is limited by voltage less than 0.5 sec. the fairly large operational currents. For A balanced three phase system will only - 1000 V peak during a contact to earth environmental reasons, the coupling have effect on telecom networks in the of one wire of a nearby DC power or installation with transformers and thyris- close proximity. As long as the power electrified railway line. tors on the Norwegian side was located transmission takes place with full balance some distance inland; however, at the - For capacitive coupling the permis- between the currents in the three phases, cost of the whole aerial route with pole sible limit has been set at 10 mA mea- only capacitive and galvanic couplings cables and electrode cables between the sured low-ohmic between a telecom will convey dangerous voltage from the plant and the landing spot near Kris- conductor and earth, or between a tele- power network to the telecom network. tiansand now being a powerful electro- com conductor and metal parts that In cases where the power network and magnetic polluter. Noise conditions are may be touched. the telecom network are very close, the particularly bad during monopole opera- difference in distance between the indi- tion with all the return current going via If the estimated or measured voltages and vidual phases in the power network and the sea electrode. currents in the telecom network exceed nearby telecom network may be so great the above limits, protective measures will Another negative side effect of the that we also get a magnetically trans- be implemented in the telecom network monopole operation is that the return cur- ferred voltage in the telecom network. or the power network, or in both net- rent does not only take the direct route In the case of unbalance between the works. The costs of the protective mea- between sea water and electrode: Parts of three phase currents in the power net- sures are shared according to the Princi- the return current act as earth current at a work, zero-sequence currents induced ple of Priority. distance of several kilometres from the

32 6 Coupling mechanisms Table 1 Coupling impedance per km in V/kA for 2,500 ohm metre and 50 Hz frequency between power and tele- Distance 0 m 10 m 20 m 30 m 40 m 50 m 60 m 70 m 80 m 90 m com networks (6 m)

The following coupling mechanisms 0 m (421) 389 346 320 303 289 278 268 260 253 apply between the power network and the telecom network: 100 m 246 240 235 230 226 221 217 214 210 207 - Galvanic coupling 200 m 204 201 198 195 193 190 188 186 183 181 - Capacitive (electric) coupling 300 m 179 177 175 173 172 170 168 167 165 163 - Inductive (magnetic) coupling. 400 m 162 160 159 158 156 155 154 152 151 150 Galvanic coupling between two circuits 500 m 149 147 146 145 144 143 142 141 140 139 requires a common impedance, Zk, in both circuits, ref. Figure 8. 600 m 138 137 136 135 134 133 132 131 130 130 The current in circuit 1 produces a volt- 700 m 129 128 127 126 126 125 124 123 122 122 Zk age drop over the common impedance 800 m 121 120 120 119 118 117 117 116 115 115 Vk = i Zk x 900 m 114 114 113 112 112 111 110 110 109 109 Zk is the coupling impedance and Vk is the operating voltage in circuit 2. Zk may e.g. be represented by the resistance in a Z common earth electrode (ref. section 8), 2 the impedance in an earth connection cable, or similar. i1 Capacitive coupling between two cable Z1 systems over a common earth plan is out- C lined in Figure 9. I II K 2 Z U2 Transferred voltage from circuit 1 to cir- k ~ U cuit 2 may be calculated with the formula 1 Ι Z2 ~ V Z V V2 = V1 1 2 2 Zk + Z2 Normally, Zk » Z2, and in the case of sinusoidal current, the numeric value of V2 may be calculated with the formula Figure 8 Galvanic coupling Figure 9 Capacitive coupling V2 ≈ V1 2 π f Ck Z2 with good approximation. CK F/m From the formula for V2 we have that the transmitted voltage is proportional to fre- CK d quency (f), coupling capacity (Ck) and 15 impedance (Z2) between circuit 2 and a earth. The numeric value of Ck is approximate to a logarithmic function as regards the separation between the two networks. As can be seen from Figure 10, 10 quite a high reduction of attenuation may be achieved by increasing the separation when the networks are very close. At medium and large distances, on the other hand, variations in separation have less 5 importance. The inductive coupling between two aerial cables at a certain height above ground may be calculated by using quite 0 complex formulas. The complete formu- 1 5 10 50 100 500 1000 las given in the above mentioned direct- a ives from the CCITT are poorly suited to d manual processing. Mutual induction be- Figure 10 Capacitive coupling between two parallel conductors

33 V/km.kA M[µH//km] stretches of proximity are then divided into shorter lengths where the two net- 600 2000 works are approximately parallel. Each length is cal-

300 1000 culated separately, and the induced voltages some- Earth resistivity where in the network will 150 500 appear as the sum of the induced voltages on the 10000 part lengths between that 5000 spot and the reference 60 200 3500 1000 point. 300 150 For the purpose of induc- 30 100 100 tion calculation the two 50 ends of the power cable 25

15 50 10 are defined as plus and

5 minus, respectively. The 2.5 calculated part voltages 1 add up to a positive or neg- 6 20 ative sum, depending on the direction of the part lengths in relation to the 3 10 direction of the power 2 5 10 20 50 100 200 500 1000 2000 5000 10000 Separation in metres cable. An example of summation of part voltages in Figure 11 Mutual induction as a function of separation and ground resistivity for the 50 Hz an exchange area is shown frequency in Figure 12. The exchange has been chosen as reference point. U12 UE Galvanic and capacitive couplings are typical proximity phenomena, while + ÷ U inductive coupling occurs from quite C U11 U8 close, up to a distance of some kilo- U6 U U metres. UB 5 10 U 3 U4 7 Protective measures to U7 U9 U Exch D U reduce disturbing or A U2 U 1 dangerous effects in

U1 to U12: Estimated part voltages the telecom network If calculations or measurements show UA to UE: Sum of dangerous voltages that the power network may cause elec- U = U +U +U ; U =U +U -U +U ; U =U +U -U -U A 1 2 3 B 1 2 4 6 C 1 2 4 5 tric effects exceeding the stated limits in nearby telecom networks, protective UD = U7-U8-U9; UE=-U7-U8-U10-U11+ U12 measures will normally be implemented Figure 12 Inductive coupling. Summation of part voltages in the telecom network. Theoretically, protective measures might also be implemented on the power current side, but if the power network is already established, tween a power cable and a telecom cable For practical purposes we use pre-calcu- the most economical solution would nor- with earth being the return conductor for lated tables of coupling impedance in mally be achieved by implementing pro- the inducing fault current, may be Norwegian Telecom. The tables are tective measures on the telecom side. machine estimated by a complex function based on typically Norwegian earth resis- of frequency, earth resistivity, screening tivities when calculating induced danger- 7.1 Protective measures factor, separation, height above ground ous voltages in the telecom network. against noise and length of exposure. An example of Table 1 is an example of such a calcula- estimated mutual coupling M as a function table. We distinguish between two principally tion of distance for the frequency 50 Hz different generations of noise on the The numeric values in the table presup- and with earth resistivity as parameter, is power side. Rectifiers, power regulators, pose parallelism between telecable and shown in Figure 11. etc., represent unlinear loads which dis- inducing line. In real life, the cable routes tort the sinus curve in the power network. will be more or less at angles, and the

34 In addition to the 50 Hz component, volt- Low frequency IT 230V - Distribution system age components will occur with a fre- longitudinal volt- L1 quency as a multiple of 50 Hz, and ages are reduced amplitudes decreasing with increasing by introducing a frequency. In such cases, nearby telecom neutralising trans- networks will primarily be exposed to former in the tele- L2 noise in analogue connections in the com cable. The voice-frequency band. Measuring this principle of such kind of noise may be done with a a transformer is L3 psophometer, a broadband volt meter that the current with a weighting filter in the input cir- from a pilot wire cuit. The filter has characteristics simu- is induced from a House 2 House 1 lating the sensitivity of the human ear. primary winding to secondary In connection with physical couplings Tele windings in series (on/off switch, load adjustment, etc.) in with each cable an operating power network, various pair. The induced degrees of arcing will occur in manoeu- voltage is in op- vred switches and reversers. The arcing posite phase to will in turn generate impulse noise in the the influence form of singular bursts or transient from the outside. bursts. Impulse noise mainly disturbs Rj2 Rj A certain degree 1 digital telecom connections. of noise cancella- Noise voltages/currents in the power net- tion is thereby work are transferred to the telecom net- achieved, but the work as longitudinal voltages (common solution requires mode). A reduction in the interference vacant cable pairs can be achieved by reducing the coupling or an insulated between the two networks and/or reduce cable screen as the effect of the transferred noise volt- pilot wire, and Exchange ages. The coupling is reduced by using that low-ohmic telecom cables with earthed screen and, termination earth- if possible, increasing the physical dis- ing of the pilot tance between the networks. The use of wire can be Rj<< coupling reduction in existing installa- achieved. Earth- tions is quite limited, and if the noise ing of vacant source itself operates within permissible pairs in a cable Figure 13 230 V IT distribution system regulations and limits, the effect of the will often in- transmitted noise voltages should primar- crease the un- ily be reduced in the telecom network. balance of the other pairs in the In the efforts to reduce the effect of the cable, and this, transferred noise voltages in the telecom together with difficult earthing condi- and magnetic coupling from a nearby network, the balancing properties of the tions, renders this method of little interest power network. If unacceptable, lasting installations in the network are essential. in Norway. overvoltages occur in an established net- Introducing isolating transformers be- work, the telecom cable may be divided tween a telecom cable and its termination 7.2 Protective measures into shorter lengths with 1:1 isolating reduces the effect of unbalance in con- against overvoltages transformers. A major disadvantage of nected equipment; the same applies if the method of division is that it causes longitudinal chokes are introduced in the The most frequently used method of blockages for the transmission of direct cable pairs. keeping permanent longitudinal voltage current in the telecom cable. in the telecom network below the stated Cable- or insulation faults may cause a maximum limit is the use of distance During earth-faults in the power network weakening of the balancing properties of attenuation. When using joint trenches brief overvoltages will occur in the the telecom cable and thereby increase and tunnels for power and telecom cables nearby telecom network. In real life, such the network’s sensitivity to noise influ- it is important to keep in mind that the earth-faults are impossible to avoid, and ence from the outside. The same applies separation between telecom network and measures have to be taken in the telecom if the cable screen is poorly earthed. Both power network should be kept large network to reduce the overvoltages to examples represent cases which should enough for a normal load fluctuation in safe levels. The most frequently used be seen to before costly noise reduction the power network not to cause unaccept- protective methods in telecom networks measures are implemented in nearby able voltages in the telecom network. are power network or in a “noise source” The corresponding rule applies to aerial which may already meet the stated - Metal sheathed, steel reinforced cable networks where an unprotected telecom requirements. (cable with reduction factor) cable may be exposed to both electric

35 - Surge arresters coupled between cable the individual installation is denoted Rjx towards 230 Volts. In house 2 is shown a pair and earth in the figure. telecom installation with a gas discharge tube installed between the telecom cable - Division of exposed telecom cable by Should an insulation fault (line to earth) and local earth. If this arrester has a introducing isolating transformers with occur on phase L3 in house 1, a capaci- spark-over voltage smaller than the a high dielectric strength tive current of a few hundred mA will ground potential for house 2, the arrester flow through Rj1, via earth and the cable - Neutralising transformers. will operate and open for a 50 Hz current capacities in the low voltage distribution into the telecom network. The extent of network, back to phases L1 and L2. The The protective methods may be applied damages that such a current may have on magnitude of this current depends on separately or in combination, depending the telecom network may be indicated by whether aerial or ground cables are used, on the level of overvoltages. The differ- the fact that a fire at Frogner exchange in but measurements carried out show that ent protective methods have their advan- Oslo in the mid-1980s happened as a as a rule of thumb, the earth current, tages and disadvantages – discussing result of a 50 Hz current from a tele- measured in mA, is estimated at a magni- them is outside the scope of this article. phone subscriber, and that the material tude of the same number as the trans- damage in the exchange amounted to former size in kVA, i.e. in a distribution some NOK 100 mill. One should also 8 Low voltage network system with a 500 kVA transformer, the bear in mind that a gas discharge tube first earth-fault will produce a current of without an earth with a continuous flow of current will be some 500 mA. If Rj1 is less than 100 so hot that the arrester itself represents a reference ohms the ground potential of house 1 in potential fire hazard. Preventing a recur- Norway is just about the only country to relation to distant earth is below the dan- rence of the phenomenon which led to use the internationally regulated IT sys- ger limit of 50 Volts. Irrespective of the the Frogner fire needs requirements to tem for 230 Volts power distribution value of Rj1, the local danger of handling make sure that gas discharge tubes at the within large areas. While all other coun- parts will be taken care of, provided there subscriber end coupled between telecom tries use distribution systems with de- are no electrical breaks in the earthing cable and local earth, shall have a nomi- fined earth references (the TN or TT sys- installation in the house. nal DC spark-over voltage of at least tem), large parts of Norway has a low With a sustaining earth-fault on phase L3 420 Volts. voltage network without any permanent in house 1, an additional situation may earth reference. This leads to the peculiar arise of an insulation fault on phase L1 or Norwegian compatibility problems be- L2 in house 2, whereby there is a total 9 Conclusion tween telecom network and low voltage system voltage of 230 Volts between the A modern society needs both telecom power supply with earth-faults. The internal earthing installations in house 1 services and electric power in order to problem has been taken care of for new and house 2. Capacitive earth currents function. In the foreseeable future, both installations in that the 1991 issue of will now flow in both installations and in telecommunications and power supplies “Forskrifter for elektriske bygningsinstal- addition, we will have an earth current will make use of galvanic cable pairs. As lasjoner” (Regulations for electric instal- between two phases, only limited by Rj1 long as we operate on galvanic connec- lations in buildings) contains the require- in series with Rj2. Ideally, the earth cur- tions, electromagnetic compatibility ment that new electrical installation shall rent ought to cause a fuse break, but problems may occur between telecom have an earth-fault protection or insula- because of the difficult earth conditions and power. The problems may be limited tion protection. However, the regulations we have in Norway, the added sum of to noise and disturbance, but situations do not apply to previously erected instal- Rj1 and Rj2 is very rarely small enough may also arise where life and property lations, and we should therefore expect to for any current fuse to operate. On the are at stake. All compatibility problems live with our particular earth-fault prob- contrary, the usual case in large parts of may be solved technically; the practical lems in old IT installations until well into the country is that we will have an earth limitations are usually of an economic the next century. The difference regard- current of only a few Amperes between nature. Questions of electromagnetic ing earth-faults between IT and TN or TT house 1 and house 2. When a voltage of compatibility are legally taken care of by systems is that in an IT system without 230 Volts is distributed on the two resist- the Norwegian regulations. Conditions of earth-fault protection “standing earth ances Rj1 and Rj2, it is physically impos- rights and responsibilities are based on faults” may occur, raising the local sible to meet the requirement of the earth the Principle of Priority, provided the ground potential to more than 200 Volts. potential of the two individual installa- limitations in established regulations are This is not possible in TN or TT systems. tions to be less than 50 Volts, as stated in fulfilled. Compatibility problems between the tele- the regulations. Provided that the two com network and an IT power network earthing installations are otherwise in without automatic earth-fault protection accordance with the regulations, there is, may be illustrated by the example in Fig- however, still no voltage which is dan- ure 13. gerous to touch locally. In the most widely used IT system in We will now assume a situation as de- Norway there are three live phase con- scribed above with two earth faults with- ductors (L1, L2 and L3) with 230 Volts out automatic switch-off, and that Rj1 is between the phases. In order to prevent very small and Rj2 correspondingly equipment with insulation faults being large. The ground potential of house 1 dangerous to touch, local earth electrodes will then be a few Volts, while the earth- are established. The earth resistance of ing installation in house 2 increases

cables should have a screen which is connected to the room screen via a bushing sheet. The connection between the cable and the room screen is critical; it should be done with the minimum impedance and with contact to the whole circumference of the cable. 1000 km 2300 km Height of "EMP coverage" "EMP coverage" Height of 3.1 Building protection explosion explosion 100 km 450 km Depending on the construction of the building and at which stage the EMP screen is to be installed, the building’s own reinforcement may be used as protection, or an additional protection is installed, made of foil or wire mesh. In EMP screens with single reinforcement consisting of iron reinforcements 8 mm thick and with a mesh size of 150 mm, the attenuation in the frequency range 10 kHz – 10 MHz is around 28 dB, Figure 3 Coverage area of a detonation above the atmosphere over the North Sea while for wire mesh with a mesh size of 25 mm and wires 0.8 mm thick, the attenuation is around 40 dB. The attenuation applies to rooms with a minimum V/km. The phenomenon resembles room, we have several options as regards size of 4 m. Double reinforcement is very induction caused by geomagnetic storms the cabling within the installation. often used, producing greater attenuation which may cause great long-lasting in- Cabling inside a screened room may as a depending on the distance between the duced voltages on long communication general rule be done like an ordinary screens. cables. telecommunication installation. If the EMP screen is open, e.g. wire mesh is For a room of height h and longest side l, 3 Protection of telecom- used, screened cables have to be used the lowest resonance frequency is given munication installations when cabling close to the EMP screen. by the formula Use of and requirements to screened 1 1 The general method of protecting against cables are otherwise as for an ordinary f 0 = 150 2 + 2 radiated EMP is to protect the equipment telecommunication installation. h l with the help of a screen cage (Faraday’s All electrical connections with the out- where the room’s dimensions are given in cage). The attenuation requirements are side world, like communication cables, m and the resonance frequency in MHz. dependent on the equipment’s resistibil- power supply, ventilation ducts, etc., ity against NEMP. We have placed dif- By use of reinforcement mesh protection should thus be handled as specially. Long ferent types of telecommunication equip- a combination of bonding and binding is cable routes may act as antennas for ment in an EMP simulator in order to employed. In Figure 6 bonding points are EMP and actions must be introduced to check the intrinsic EMP immunity of the indicated by circles, while binding points reduce the harmful effects. Surge arrest- equipment and whether it can “survive” are marked by a slash. Regardless of the ers, filters and limiters may be used; all such a radiation. The critical situation type of screen chosen, good electrical mounted with the lowest possible arises when cables are connected. By connection at the joints must be assured. impedance to the screen. All lead-in placing the equipment in a screened

90 N o r

dB m 80 1 ) 2At a

h l m i / A 70 z s e - 0.9 d V Electric-field

c µ 60 spectrum u B

20 Log 2At m 10 r d

( 50 t u

h E l

0.5A a Normalized t

40 i cumulative v energy-density e

30 e 0.1A spectrum 0.1 n

0 e r

20 g time 4 5 6 7 8 9 tr 1 1 Frequency 10 10 10 10 10 10 y

πth πtr Frequency (Hz) d e n

Figure 4 Typical model pulse for EMP Figure 5 High altitude EMP spectrum and s i t normalized energy density spectrum y

38 Corner Wall armouring armouring

It is important that the screen is not Due to the thickness of the plas- “punctured”. Outer bolts e.g. should not tic insulation, damage is not be led through the screen. All outer metal expected on our standard constructions, like cable rack from the telecommunication cables, but antenna tower, work as an antenna for connected equipment may be EMP, and their chance of leading de- harmed. Currents in an earth structive currents into the installation cable as a function of the length should be prohibited. is shown in Figure 9.

3.2 EMP barrier Aerial cables may hold larger currents and as characteristic The ideal requirement of all cables being impedance for aerial cables is fed through a single entry combined with greater than for earth cables, the filters and surge protection placed in the voltage will also be higher and a barrier, usually has to be abandoned and flash over to suspension equip- replaced by compromises. A telecommu- ment may occur if the cable is nication installation may be an exchange not protected. with tens of thousands of connections. It For screened cables the voltage is hard to visualise a single entry with Figure 6 Reinforcement mesh screen will be reduced. We have cho- limiters and filters for tens of thousands sen screened lead-in cables of line pairs placed in the barrier! where coupling impedance is It is of great importance that the electri- reduced by increasing frequency. 0.2 cal connections between the screen room mm aluminium sheath or 2 mm lead and the outside world are concentrated in sheath is used. These types of cables one point; a so-called single entry. Hav- are shown as graph A in Figure 10. ing multiple separate entries in the screen As a comparison is shown the cou- core conductors may easily cause currents in the screen pling impedance for cables with an o wall which may destroy the effect of the aluminium foil cover (graph C) and 360 connector (correct) room screen. with a braided copper screen (graph B). One may clearly see that for The best discharge of sheath currents is cable screens which are not HF achieved when a copper braided-wire tight, the coupling impedance shield is thread on the outside of the increases considerably with high fre- metal sheath, so that the whole circum- quencies. ference of the cable is covered. If the screen termination is done as pig-tail, its If unshielded cables are used in a external pigtail impedance may be considerable (U = network it is still a requirement that (poor) ωLI or L ⋅ di/dt). All bondings or screen shielded cables should be employed connections should therefore be as short for the last 300 m leading up to the as possible and preferably done with a EMP protected installation. Lead copper lace/ribbon in order to limit the cables are preferred; not least inductivity. The inductivity, L, is approx. because it is relatively easy to have 1.5 µH/m for wire. In Figure 7 are shown sheath currents discharged to the internal pigtail examples of screen termination. barrier. (bad) Telenor has standardised the shield con- Outside the installation the shielded nector, so that the whole of the cable’s cables may be connected to long un- circumference is covered. The braided- shielded cables. By using shielded wire sleeves are shown in Figure 8. They lead-in cables, we have in principle are provided in dimensions from 16 mm introduced a low-pass filter, even if (worst) signal ground to 84 mm (inside measurements). the attenuation is not very high. Typical value for plastic insulated 3.3 Physical connections cables is 3 dB/100 m at 1 MHz, and the attenuation in dB is approxi- A telecommunication network may con- mately proportional to the square Figure 7 Examples of screen terminations sist of both screened and unscreened root of the frequency. cables and installed mainly as earth cables as regards protected installations. In order to achieve the wanted effect of plastic into a semi-conductor (ρ ≈ But particularly on the outskirts of a sub- the cable screen it has to be earthed; the 20 Ωm). scriber network there may be aerial best solution being a continuous dis- cables. charge to earth. Such discharge is achiev- Cables which are insulated from earth ed quite simply for lead sheathed cables should be earthed at various points some For earth cables an accumulated current without any outer insulation and for Al- 30, 60 and 90 m away from the screen of 1 – 1.5 kA may be estimated (1). For sheathed cables by the outer plastic room. In addition, the cable screens unscreened cables the voltage between sheath being added coal, thus turning the should be discharged to the screen wall conductors and earth may reach 100 kV.

39 by braided-wire sleeves in single entry. favourable than one long one. This par- The outer sheath of the cable is important ticularly applies to NEMP because of the for the screening effect. Lead cables short rise time of the pulse. The effective without any outer insulation are length and pulse impedance for one earth favourable from an EMP point of view, rod may be calculated by the following as they may be installed directly on formulas (2): earthed cable racks or put directly in con- l ≈ 0.9 t ρ tact with the ground. The attenuation of eff r the sheath current because of insulation 1 Rpulse = related to earth may perhaps be illus- Gleff trated by the following numbers (1): 2π 1 Figure 8 Examples of braided-wire sleeves G = - Approx. 2 dB/100 m on cable with ρln l insulated sheath, r - Approx. 4.5 dB/100 m on cable with where A conducting sheath in dry rock, and G is the conductance of the electrode - Approx. 33 dB/100 m on cable with (S/m) conducting sheath in moist ground. leff is the effective length of the electrode (m) 103 Dry rock Cable with conducting sheath laid directly in cable racks of metal and being l is the actual length of the electrode continuously earthed, also show favour- (m) able values for sheath discharge. t is the rise time of the pulse (µs) Moist earth r 3.3.1 Cable termination/protection ρ is earth resistivity (ohmm)

102 Large parts of the telecommunication r is the radius of the electrode (m). cable network is unshielded, and the volt- 1 10 100 500 ages fed into the installation can there- 5 References Length of conductor, m fore be great. In order to reduce the conducted EMP to levels which the equip- Figure 9 Current as a function of cable length 1 Elektromagnetisk puls, virkninger og ment may stand, all connections should vern. FFI/Rapport-82/3003. be equipped with surge arresters. For paired cable conventional gas discharge Z 2 Handbuch für Blitzschutz und κ tubes (GDT) are used, e.g. mounted in mΩ/m Erdung. Hasse/Wiesinger. 5 Krone/Siemens Trennleiste 71 or Krone 10 LSA + cable termination. The reason why we may use conventional GDTs is

4 the “filter” in the form of shielded cables 10 in front of the barrier. For protection of antenna cables are used coaxial mounted GDTs. For protection of power supply 103 and short connections for fire alarms and so on are used various types of varistors. C 102 3.4 Optic cables Fibre optic cables are well suited as con- B nections between the protected installa- 10 tion and the outside world. This is partic- A ularly favourable if the cable is metal free. If the cable contains metal, the 1 metal should be connected to the screen wall as for conventional paired cables. Regenerators or terminals may, however, -1 10 be influenced by EMP.

10-2 4 Earth electrodes Telenor uses mainly earth rods in order 10-3 to avoid an increase in earth resistance 0.01 0.1 1 10 100 MHz caused by drought or frost. In connection with lightning protection it is well known Figure 10 Shielded cables that several short electrodes are more

40 Telecom overhead cables as antennas for long wave radio signals

BY ARNE THOMASSEN

Introduction effect (1, page 284). This is in fact also a Table 1 Maximum magnetic field strength in µA/m, result of the use of Millington’s method Southern Norway It has been observed over quite a long (1, page 282). period of time that telephone subscribers Locations 19 kHz 23.4 kHz in some areas were annoyed by false The signal strength is much attenuated telephone calls. At the same time the over land because of low conductivity. Arendal, ref 50 100 normal functioning of the telephone Lakes, fjords and valleys with rivers will Dølemo 35 60 equipment was disturbed, especially for give relatively good conditions for the Vrådal 25 20 frequency division multiplex (FDM) sub- signals, while especially mountains and scribers. Investigations in 1987/1988 woods may increase the attenuation. Kviteseid 30 20 showed that the noise problems came Table 1 and Figure 2 show some Brunkeberg 25 20 from a navigational transmitter in Germ- maximum field strengths in Southern Mandal 40 160 any, and the problems will here be dealt Norway. Vigeland 35 140 with in 9 sections. 4 22 kilovolt overhead Flekkefjord 35 140 1 The actual transmitter power lines as antennas Jøssingfjord 25 110 Stavanger 25 100 Among the navigational transmitters, a Long 22 kilovolt overhead power lines in Karmøy 45 140 transmitter in Germany gave us the the north-south direction may act as strongest received signals. The frequency travelling wave antennas. The induced Ølen 25 35 23.4 kHz was also well inside the Divis- currents are added to another all the way Røldal 20 15 ion Multiplex frequency band in the along the power line and may reach large Hønefoss 40 15 direction from the customer to the levels, and then it is induced into parallel Hægeland 15 45 exchange (bandwidth 19 – 37 kHz). In telecommunication cables. If the power addition, even the carrier was from time line is long enough the energy may be Byglandsfjord 40 60 to time amplitude modulated with 10 Hz transmitted from the power line to the Lillehammer 25 25 and interfering with our own 10 Hz telecommunication cable, or from power Vågsli 35 25 signalling. The transmitter also came on line via power cable to telecommuni- and off from time to time, and the trans- cation cable, ref. Table 2 and (2). Haukeli 25 25 mitted level seemed to vary. Ålvik 25 20 5 Travelling wave Br. Knarvik 25 110 2 Time propagation antennas Lindås 20 25 effects Overhead paired cables will act as Vadheim 25 25 travelling wave antennas and may even Florø 20 8 Because of the special propagation form quarter-wave antennas and common effects the level of the received signal Ramsdal 10 15 mode LC-resonators. Shielded cable could be several times higher than Førde 25 10 sections may play a part in resonances, normal about one hour after sunset and RC-attenuation and impedances to the Stryn 25 35 one hour before sunrise. Maximum ground. Signals may be reflected from N.fjordeid 35 35 values (“worst case”) was then obtained, power lines, paired cables, and big metal but even this varied with time, latitudes Ålesund 10 6 structures and be induced into the cables and longitudes (fading). It was assumed Namsos 4 20 with different phases and with different that it was advisable not to make routine levels. The result of all this is therefore measurements in these periods. A refer- that the condition is rather complicated Comments to the table: ence antenna was established in Arendal and difficult to predict. All measurements are related to worst case. 23.4 kHz is and maximum values were continuously the strongest frequency component most places. recorded for several days, and a multipli- The most convenient way is therefore to 19 kHz is varying somewhat different. A few values cation factor was then given to the measure the standing waves and then could be high because of reflection from metal struc- measuring teams travelling along the make some estimates. One thing that tures, but they should give a good idea of the situation coast controlling the FDM systems for seems certain, however, is that the as examples. one exchange at a time. induced currents travel with the same phase

3 Geographical propa- and the same direction Peak on all the single wires 75 µA/m gation effects in the cable (common Also a geographical kind of propagation mode). A change in effect was observed. A low frequency one of these cable signal has a tendency to bend (1, page pairs therefore also 365) to the lowest conductivity. There- affects all the other 20 µA/m fore, the 23.4 kHz signal could follow the cable pairs and alters coastline up to Karmøy and Bergen. In the standing wave addition, the signal on the south coast of patterns on all the 1.18 2.55 4.32 Karmøy had a higher received signal pairs. Individual than Stavanger because of the recovery variations will appear Figure 1 Peaking level a little before sunrise where the cable has taps to smaller cables. 41 5.1 Measurement at the filter Table 2 Magnetic field strengths (23.4 kHz) at Kristiansand location in Hægeland Magnetic field in the open fields 100 µA/m It is evident that low levels between cable pairs and ground is normal in the Magnetic field 5 metres from 230 V power line 200 µA/m exchange because of low impedance to Magnetic field 5 metres from 22 KV power line 500 µA/m the ground while levels far out on the cable can be high. This is due to standing Magnetic field 50 cm from 22 KV cable 5000 µA/m waves, impedances and mismatch. At such critical places only a slight All measurements in this table are related to worst case. impedance imbalance or mismatch can cause trouble. Transformers and common mode noise filters, correctly placed, can Table 3 Results in 10 exchanges from Kristiansand to Karmøy reduce the problem. The best, however, is to avoid FDM systems when such 235 customers had between 001 and 499 µV (Safe values) conditions occur. 30 customers had between 0.5 and 1.5 mV (Shadow zone) 6 Cable pair impedances 19 customers had between 1.6 and 3.9 mV (Danger zone) to ground 6 customers had between 4.0 and 6.0 mV (Some failure) Low impedances to ground at the 2 customers had 6.1 and 15 mV (Much failure) exchange is an important factor. At some 1 customer had (Hægeland) 53 mV (Much failure) places with lead cables into the station, signals to ground were completely extinguished while they were still high in the outer end of the attached overhead themselves, however, could also cause an cable network. Unfortunately, it is in this unbalanced condition to the ground so outer end that bad impedance balance in that the noise voltage (measured up to line filters made common mode currents 850 mVrms) would give severe prob- into signals between a- and b-wire. If lems. These filters were later removed there are several sections with shielded from problem locations and replaced by cable (see Table 3, line 1), the induced modified filters. 6 noise is greatly attenuated both because 8 of less induction and because of lower 7.1 Measurement showing impedance to the ground. Overhead mismatch and unbalance cables with lengths of 100 metres be- at filter 110 25 tween filter and customer did not cause The measurement shows that the mis- 15 any trouble. Cables up to 20 metres be- 15 match and unbalance are creating a tween filter and customer gave no signal between a-wire and b-wire at the 20 resonance problems at relevant frequen- 35 Hf-customer (90 mV). The signal may cies. Shielded overhead cable reduced then be attenuated on its way to the 140 the problems, but even distance between 20 demodulator in the exchange (50 mV). the grounding should be avoided because 100 of the danger of multiple resonance. 60 8 Susceptibility of the 110 60 140 7 Longitudinal balance and FDM 1+1 system 100 160 mismatch on cable pair Customer-HF demodulators on the FDM 140 1+1 systems use carrier on/off for The last model of the FDM 1+1 system Figure 2 Maximum field strength in signalling. (28 kHz gave dialling tone. had a line filter which had a serious µA/m, Southern Norway Short breaks in the 28 kHz gave the diall- longitudinal balance error. Induced coming signals). Of course, the exchange also mon mode noise on cable pairs were then receives tone signalling, but both systems transformed to differential mode and existed simultaneously. When the noise could then disturb the receiver at the also made complaints because they often voltages exceeded the receiver threshold exchange. The fault was in a common could not get the dialling tone. The noise of about 2 – 4 mV, this was treated as a mode filter-coil so that the longitudinal had activated the dialling tone for such a call and the dialling tone started. Keying unbalance came between input and out- long time that the exchange had deacti- was detected when level varied or the put of the filter and the common mode vated it. unwanted radio signal was keyed. This currents on the line then caused prob- caused automatic set-ups to other cus- lems. The filter was “floating” and had tomers. The customers who had been 9 Routine measurements no connection to the ground. This caused phoned automatically complained to our 293 FDM customers in Flekkerøy, mismatch in the filter. There was also a service department. They claimed that Randesund, Hægeland, Kvinesdal, Store- mismatch as well in the exchange as at someone was bothering them because kvina, Hjelmestad, Ferkingstad, the customer. Errors in the cable pairs nobody answered. The FDM customers Koparvik, Kjernagel and Sveio in the

42 Southern and South-Western part of 15 mV a/ground 450 mV a/ground 80 mV a/ground Norway, were measured. A lot of other a 530 mV places were later controlled (ref. Table <0.3 mV a/b HF- 3). The lines were controlled with the Phone b 530 mV help of a HP-3580 spectrum analyzer, 15 mV b/ground 450 mV b/ground 80 mV b/ground with high impedance and in parallel with communication signals between a-wire Exchange Line filter and b-wire. Not all customers had prob- location Filter removed lems, even at the worst locations. Only a few customers in some of the locations Figure 3 23.4 kHz levels between wires and ground had overhead cables with problematic lengths which could cause resonances. The measurements were calibrated using the antenna in Arendal simultaneously. Comments to Table 3: 340 mV a/ground 540 mV a/ground a 850 mV Worst exchange was Hægeland in Setes- HF- 50 mV a/b90 mV a/b 100 mV a/b dal with many unshielded overhead b 820 mV Phone cables. Most of the FDM customers were 400 mV b/ground 510 mV b/ground in trouble there. After changing from old filter to a modified one, nearly all cus- Exchange tomers had a disturbing signal level 400 mV a/ground below 0.5 mV, which was well below the LF- level at which the equipment would oper- <0.3 mV a/b ate. But the high levels between cable Phone and ground are still there. Just a slight 400 mV b/ground error in line impedance balance may Figure 4 23.4 kHz levels with filter included in Hægeland therefore bring the problem forth again. Randesund was a well operating exchange because of many shielded cables (the 500 KV-DC power line goes through the area on its way to Denmark). Mod/demod HF/LF Mod/demod Exchaange Only one out of a hundred customers had trouble there. HF-line Common line FDM- Line- FDM- HF The exchanges at Åna-Sira and Flikka unit filter unit sometimes had so big problems that the LF entire exchange was blocked. As a first (28 +/- 8kHz) (79 +/- 8kHz step the service-team had to disconnect Selektive some of the FDM equipment. Phone Phone Level meter An example in Åna-Sira exchange shows the resonance problems. From the exchange there was at first 3,500 m shield- HF.subscriber LF.subscriber ed overhead cable, grounded with even Figure 5 Method of routine measurement intervals. Then came a 450 metre unshielded overhead cable up to the line filter. From the filter to the customer the distance was 300 metres. All worked Kjernagel exchange in Haugesund had well, but then a new customer was overhead and underground cables with- References connected to this line. The only differ- out shielding. The levels here were com- 1 Stokke, K N. Radiotransmisjon: ence now was the fact that the distance paratively low. 5 years later, however, kabler, bølgeutbredelse, antenner. 4. between filter and customer was reduced some problems arose at this exchange. utgave. Oslo, Universitetsforlaget, to 200 metres. This caused the signal to Something had presumably happened in 1982. exceed the problem level. This shows the cable network. how difficult it is to predict these condi- Routine measurements should be done at 2 Stokke, K N. Travelling wave tions. It is absolutely necessary to make all exchanges from the Swedish border antennas. Telektronikk, this issue. measurements. up to Bergen because of all the uncer- Koparvik exchange at Karmøy had no tainties. Every measurement should be customers above 50 µV disturbing signal. referred to the reference level in Arendal. The lead shielded cable caused these Exchanges with many overhead cables very low levels. Some overhead cable along the coast should be measured first. outside the lead cable here did not seem Some places which are thought to be safe to cause problems. may not actually be so.

43 EMC-norms and regulations – the situation in Norway

BY SVERRE TANNUM

The regulations on electrical devices and As mentioned, the Norwegian regulations If we establish the same transitional systems are based upon the low voltage were put into force as of January 1, 1991. period as many other European countries directive and the EMC-directive of the In another regulation put into force on have, these years should not be spent European Union (EU). The substantial the same date it is stated that the self- without preparing for the coming content of this directive was implement- declaration shall contain a reference to situation. ed in the Norwegian regulations as of the relevant Norwegian regulations and January 1, 1991 (Forskrift om utførelse shall confirm that the equipment com- As a minimum, we should demand that og kontroll av elektrisk utstyr som tilbys plies with their requirements. This is all products which are developed now eller omsettes til bruk i lavspenningsan- valid for EMC when norms exist. In and in the coming years must comply legg). The EMC requirements are other words, Norwegian regulations with the requirements of the EMC incorporated in § 11. require compliance with national directive. This view is expressed by the Norwegian norms. leader of NEK (The National Traditionally, it has been demanded that Electrotechnical Committee of Norway), a number of products shall be controlled The situation now is that some 20 Mr. Bjørn I. Ødegård, and he recomm- in one way or another before marketing. European EMC norms exist. They are ends a positive attitude to the EMC qual- We say that a major part of all electrical already put into force as Norwegian ity requirements. equipment belong to the regulated pro- norms and consequently, they have to be duct area. complied with.

The situation now is that there is no Because Norway already has adopted the longer an obligation for approval tests of European EMC Norms as Norwegian electrical equipment before marketing in norms, we have in principle no transi- Norway, but the authorities have estab- tional period like many other European lished a registration system which re- countries have. quires full technical documentation of the product. For all electrical apparatus and The so-called generic norms for emission equipment which can be connected to the and immunity for domestic commercial low voltage network a declaration is and light industry equipment is very required from the producer identifying important in this respect. the product itself and the actual Norwegian regulations or norms which To put the new European Norms (EN) have to be fulfilled for this product. into force at an early stage without wait- ing until 1996 could make the Norwegian The EMC directive covers all electrical export industry more competitive, but on and electronic equipment with very few the other hand, it would reduce the turn- exceptions. over for Norwegian import firms who get their products from countries without the A short time after the approval of the same EMC requirements. directive it was agreed that there should be a transition period from January 1, These problems are now being discussed 1992 to January 1, 1996. In this period and the question is whether the economic the producers can choose if they want to interests will overrule the importance of fulfil the national requirements of each the electromagnetic environment until country or the EMC directive with CE- 1996. marking, according to the Cenelec European norms (the new approach).

44 An overview of Norwegian electrotechnical norms concerning EMC

NEK-EN 50 065-1 (1991) Signalling on low-voltage electrical installations in the frequency range 3 kHz to 148.5 kHz. Part 1: General requirements, frequency bands and electromagnetic disturbances. NEK-EN 50 081-1 (1992) Generic emission standard. Part 1: Residential, commercial and light industry. NEK-EN 50 081-2 (1993) Generic emission standard. Part 2: Industrial environment. NEK-EN 50 082-1 (1992) Electromagnetic compatibility – Generic immunity standard Part 1: Residential, commercial and light industry. NEK-EN 55011 (1989) Limits and methods of measurement of radio disturbance characteristics of industrial, scientific and medical (ISM) radio frequency equipment. Modified. NEK-EN 55 013 (1989) Limits and methods of measurement of radio disturbance characteristics of broadcast receivers and associated equipment. Amdt 1 (1983). NEK-EN 55 014 (1988) Limits and methods of measurement of radio interference characteristics of household electrical appliances, portable tools and similar electrical apparatus. Modified. Amdt 2 (1989) A2 to EN 55 014. NEK-EN 55 015 (1988) Limits and methods of measurement of radio interference characteristics of fluorescent lamps and luminaries. Modified. Amdt 1 (1989) A1 (1990) to EN 55 015. NEK-EN 55 020 (1988) Immunity from radio interference of broadcast receivers and associated equipment. NEK-EN 55 022 (1988) Limits and methods of measurement of radio interference characteristics of information technology equipment. Modified. NEK-EN 60 555-1 (1987) Elektrisk støybegrensning i forsyningsnett. Part 1: Definisjoner. NEK-EN 60 555-2 (1987) Part 2: Harmonics. Modified. Amdt 1 (1985). NEK-EN 60 555-3 (1987) Part 3: Voltage fluctuations. (+ Corrigendum 1990) A1 (1992) to EN 60 555-3. NEK-EN 60 601-1-2 (1993) Medical electrical equipment Part 1: General requirements for safety. 2. Collateral standard: EMC – Requirements and tests. NEK-EN 61 000-4-1 (1994) Part 4: Testing and measurement techniques Section 1: Overview of immunity tests. NEK-EN 61 000-4-7 (1993) Part 4: Testing and measurement techniques. Section 7: General guide on harmonics and interharmonics measurements and instrumentation for power supply systems and equipment connected thereto. NEK-EN 61 000-4-8 (1993) Part 4: Testing and measurement techniques. Section 8: Power frequency magnetic field immunity test. NEK-EN 61 000-4-9 (1993) Part 4: Testing and measurement techniques. Section 9: Pulse magnetic field immunity test. NEK-EN 61 000-4-10 (1993) Part 4: Testing and measurement techniques. Section 10: Damped oscillatory magnetic field immunity. NEK-HD 481.1 S1 (1987) Electromagnetic compatibility for industrial process measurement and control equipment. Part 1: General introduction. NEK-HD 481.2 S1 (1987) Part 2: Electrostatic discharge requirements. NEK-HD 481.3 S1 (1987) Part 3: Radiated electromagnetic field requirements.

45 EMC measurements

BY SVERRE TANNUM

Open Area Test Site (OATS) Turntable load: 5,788 kg The new Open Area Test Site (OATS) of Radiostøykontrollen (the Radio Interfer- Power supply: ence Service) is located at Maridalen 30 A 400 V AC three phase near Oslo. It is built for the normalized 30 A 230 V AC three phase measuring distances 3–10 and 30 metres and complies with the international 16 A 230 V AC single phase requirements for OATS. 100 A 48 V DC The test site must be free from reflecting By special arrangement objects within the ellipse having a major 160 A 230 V AC diameter 2 d and a minor diameter √3 d (d = measuring distance). Site attenuation: Within the specified CISPR The glass fibre weather protection covers Limits ± 4 dB. the 3 – 10 metres measuring sites. 30 m measurements is made without weather Electrical and electronic protection of the receiving antenna. consumer equipment as well as telecommunication The measuring site is located about apparatus must comply with 15 km outside the city boundary and the national or international topography gives good protection from EMC (Electro Magnetic the ambient noise which is at least 30 dB Compatibility) requirements. lower than in the city. EMC means the ability of a Specifications: device, unit of equipment or Weather protection: system to function satisfacto- 20 x 10 x 8 m (LxWxH) rily in its electromagnetic environment without introducing Ground plane: intolerable electromagnetic distur- 18.5 x 10 m bances to anything in that environment. Turntable diameter: Testing of EMC includes testing of both 4 m Figure 3 EMC test site, emission mea- emission and immunity. surement Radiostøykontrollen performs measurements according to the following norms: EN 55011 corre- CISPR sponding to Publication 11 test (EUT) which is placed upon a turn- CISPR table rotating 360 degrees. Publication 12 The field strength is measured via the

S EN 55013 corre- CISPR receiving antenna by a measuring receiv- A r

o sponding to Publication 13 er. The antenna height is adjusted be- n e l tween 1 and 4 metres above the conduct- e

T EN 55020 corre- CISPR

: ing ground plane for maximum reading. o t sponding to Publication 20 o h P Figure 1 OATS in Maridalen EN 55022 corre- CISPR Measurements of conducted sponding to Publication 22 interference EN 50081-1 Conducted interference (normally below 30 MHz) is measured either directlt on EN 50081-2 the signal terminals or via a Line Stabi- EN 50082-1 lizing Network (LISN) or an Artificial Mains Network (AMN). The interference and corresponding national norms. level is recorded by the measuring re- A short description of ceiver. The AMN and LISN are required to pro- some EMC measurements vide a defined impedance at high fre- S A r quencies and they have filters to isolate o Measurement of radiated n e

l the test circuit from unwanted radio fre- e interference T quency signals on the signal- and supply : o t o The most common method of measuring lines. The measurements of conducted h P emission of radio noise and interference interference are performed in a screened Figure 2 OATS interior, 3 and 10 metres measuring sites is to measure the field strength of the room to avoid influence from back- with weather protection emitted signal from the equipment under ground noise.

46 LISN “Clamp” measurements - Immunity from radiated fields (electromagnetic fields present at the equip- signal port An alternative measuring method to ment) radiation measurement is the so-called EUT AMN “clamp” method. - Screening effectiveness is the characteristic of a coaxial connector terminal to It is generally considered that for fre- mains port measuring attenuate the transfer of external cur- quencies above 30 MHz the disturbing receiver rents into internal voltages. energy produced by appliances and simi- EUT = Equipment Under Test lar devices is propagated by radiation. The EMC laboratory is equipped to per- LISN = LIne Stabiliizing Network Experience has shown that the disturbing form these measurements. AMN = Artificial Mains Network energy is mostly radiated by the portion Figure 4 Measurements of conducted interference of the mains lead near the appliance. It is Stripline therefore agreed to define the disturbing The immunity of broadcasting receivers capability of an appliance (EUT) as the from radiated fields can be tested in an power it could supply to the mains lead. open stripline device (TEM). The equip- This power is nearly equal to that sup- ment under test (EUT) is placed in the plied by the appliance to a suitable ab- open stripline with the wanted signal sorbing device (clamp) placed around supplied at the input terminal. The this lead at the position where the absorb- unwanted signal in the range 0.15 MHz ed power is at its maximum. to 150 MHz is supplied by a generator Immunity measurements via a matching network to the stripline

which is loaded with a terminating S A r

Immunity is the ability to maintain a impedance (150 ohms). The wanted sig- o n e specified performance when the equip- nal is specified according to the operat- l e T

ment is subjected to interference signals. ing mode of the receiver and the : o t o

Unwanted signals can attack the victim unwanted signal is amplitude modulated h P equipment in different ways like con- with 1000 Hz at 80 % depth. Figure 5 Measurement of conducted interference, the ducted disturbances, radiated fields The stripline is normally placed in a artificial mains network (AMN) down to the left (wanted and unwanted) from other equip- shielded room. The level of the unwanted ment or even from modules in the victim field is monitored equipment itself. Concerning broadcast- by a probe close to ing receivers we usually specify the fol- measuring the EUT, and the lowing: receiver level is increased - Input immunity (immunity from until interference is unwanted signals present at the detected in picture antenna input terminal) or sound quality. absorbing clamp - Immunity from conducted voltages The immunity test (unwanted signal currents present at based upon a large EUT 230 V the mains- and signal terminals of the number of single equipment) frequencies will be very time consum- - Immunity from conducted currents ing. A scan-method 5 m (unwanted signal currents present in is now evaluated to cables connected to the equipment) reduce test time and costs. Figure 6 “Clamp” measurements

~ S A r o n e l e T : o t o h P Figure 8 Open stripline – TEM device, basic configuration with matching network and Figure 7 Measurements by clamp in a screened room terminating impedance. The equipment under test is placed between the two plates

47 Table 1 Measuring instruments and antennas Cavity antenna The dimensions of the EUTs are limited Emission measurements by the stripline. For physically large EUTs it is necessary to expose the whole CISPR EMI Receiver: volume of the shielded enclosure to the Spectrum Analyzer 100 Hz – 22 GHz HP 8566B Hewlett Packard electromagnetic field. RF Preselector 20 Hz – 2 GHz HP 85685A Hewlett Packard Quasi Peak Adapter HP 85650A Hewlett Packard A cavity antenna placed on one of the EMI Test Receiver 9 kHz – 30 MHz ESVS30 Rohde & Schwarz walls will create an HF field, but because of reflections the homogeneity is poor. EMI Test Receiver 9 kHz – 30 MHz ESH3 Rohde & Schwarz To compensate for this, sensors are EMI Test Receiver 20 – 1300 MHz ESVP Rohde & Schwarz placed close to the EUT. The output from the sensors regulates the power input to Spectrum Monitor ESM Rohde & Schwarz the antenna to maintain a constant field EMI Software EZM-K1 Rohde & Schwarz strength close to the EUT. XYT Recorder ZSKT Rohde & Schwarz A fully anechoic room will give better possibilities to create a more homo- Spectrum Analyzer 9 kHz – 2.6 GHz R3261 Advantest geneous field over a large volume. Synthesized CW Generator 1 – 20 GHz HP 83711A Hewlett Packard Signal Generator 0.1 – 990 MHz HP8656A Hewlett Packard EMC measurements in situ Colour TV Pattern Generator PM 5415TN Philips When the equipment is too large to be tested in the laboratory or test site, in situ Programmable RF Generator 100 kHz – 1 GHz PM 5390S Philips testing has to be performed. This is often Process Controller PSA 17 Rohde & Schwarz the case when testing large industrial and offshore installations or cable-TV net- Royce Field Site Source 10 – 600 MHz 4610 EMCO works. Impulse Generator 91263-1 ETN Measuring equipment Active Loop Antenna 10 kHz – 30 MHz 6502 EMCO In addition to the described OATS in Dipole Antenna Set 28 MHz – 1 GHz 3121 EMCO Maridalen, Radiostøykontrollen has the Biconical Antenna 20 – 300 MHz HUF-Z2 Rohde & Schwarz following equipment: Logperiodic Antenna 200 – 1000 MHz HUF-Z3 Rohde & Schwarz Screened enclosures (not anechoic) Horn Antenna 1 – 18 GHz 3115 Rohde & Schwarz A Logperiodic Antenna 1 – 18 GHz HL 025 Rohde & Schwarz 4 x 6 x 3 m DOOR 80 x 200 cm Pulse Limiter 10 dB ESH3-Z2 Rohde & Schwarz B 2.4 x 3.8 x 2.4 m DOOR 80 x 200 cm Two-line V-network, 2 x 10(16) A, 50 ohm // 50 uH + 5 ohm 9 kHz – 30 MHz ESH3-Z5 Rohde & Schwarz C 2.25 x 1.7 x 2.3 m DOOR 90 x 227 cm Artificial Mains Network, 2 x 10(16) A, 50 ohm // 50 uH + 5 ohm 9 kHz – 30 MHz NNLA 8119 Schwarzbeck D Artificial Mains Network, 4 x 16(25) A A fully anechoic screened enclosure 50 ohm // 50 uH + 5 ohm 9 kHz – 30 MHz NSLK 8126 Schwarzbeck 8.96 x 6.08 x 4.56 m DOOR 250 x 250 cm Absorbing material Artificial Main Network, 2 x 6 A walls, ceiling and Delta, 50 ohm NNBM 8116 Schwarzbeck floor: Ferrite tiles Step Attenuator HP 355C Hewlett Packard Turntable: Air floating 4 m diameter Step Attenuator HP 355D Hewlett Packard Load: 6,000 kg Manual Step Attenuator 0 – 11 dB HP8494B Hewlett Packard Manual Step Attenuator 0 – 110 dB HP8496B Hewlett Packard Photometer Lumacolor J17 Tektronix Motorized turntable 4 m, 5988 kg 1081-4.03 EMCO Antenna positioning mast 1 – 6 m 1050 EMCO Positioning Controller 1050-84 EMCO Positioning Controller 101991B EMCO Plotter R9833 Advantest Plotter HP7440A Hewlett Packard

48 Table 1 continued

Immunity measurements

Signal Generator 10 kHz – 2.7 GHz 2031 Marconi Signal Generator 10 kHz – 1000 MHz 2022 Marconi Signal Generator 10 kHz – 1040 MHz 2019A Marconi Broadband Amplifier 125 kHz – 250 MHz 10 W model 10L Amplifier Research Broadband Amplifier 0.01 MHz – 100 MHz 25 W model 25A 100 Amplifier Research Broadband Amplifier 1 MHz – 1000 10 W Amplifier Research Broadband Amplifier 10 kHz – 220 MHz 150 W Amplifier Research CISPR Stripline 80 cm 150 kHz – 150 MHz IFI-field Antenna EFG-3 10 kHz – 220 MHz Eaton biconical Antenna 96002C 20 MHz – 300 MHz AR log periodic Antenna AT-1000 150 MHz – 1000 MHz Vertical Antennas for HF, VHF and UHF AR Fieldprobe FP1000 10 kHz 1 GHz Cavity Antenna AT 2000 30 – 1000 MHz 3500 W Amplifier Research AR Isotropic field monitor model FM 1000 10 kHz – 1 GHz 300 V/m AR Levelling amplifier model 888 10 kHz GHz Audi Elektronik Spark generator SS 4361503 Absorbing Clamp 30 MHz – 1000 MHz MDS-21 Rohde & Schwarz Absorbing Clamp 30 MHz – 300 MHz MDS Rohde & Schwarz Instruments for Industry E-field meter EFS-1 1 v/m-300 v/m 10 kHz – 200 MHz Broadband sampling voltmeter 1 mV – 3 V 10 kHz – 1200 MHz 3406A Hewlett Packard Sennheiser audio voltmeter UPM-550 Microphone system for audiomonitoring Hagener light intensity meter 52 Function generators 10 kHz – 0.1 MHz S S A A r r o o n n e e l l e e T T : : o o t t o o h h P P Figure 9 Immunity test of TV receiver in the stripline Figure 10 Cavity antenna and field strength sensors in the screened enclosure

49 Electrical Medical Equipment Control and Vigilance in Norway

BY ERIK FØNSTELIEN

In the late sixties and the early seventies - Evaluate and discuss maintenance and creasing number of reports, at present it became increasingly clear to biomedi- repair and the proper management of 5–15 a year. The cause of the incidents cal engineers and safety technicians that equipment for diagnosis, treatment, is mainly that patients and visitors take hospitals and other health institutions and monitoring of patients. their telephones into the hospital and represented an important area of hazards thereby bring the transmitters close to - Monitor the development and attend to due to defective medical equipment and vulnerable medical devices, the anten- national and international standardisa- harms to patients by equipment users’ nas sometimes touching electronic tion work related to health and medical errors. Worth mentioning is also some pumps, ECG-electrodes, etc. The devices, notably both IEC and CEN- reports on adverse effects by the use of equipment accidents are, with few ELEC working groups on EMC and correctly functioning equipment (1). In exceptions, seen with infusion pumps medical devices. 1970, Dr. A.F. Pacela of Beckman and dialysis systems. Both equipment Instruments Inc. published a bibliography types regulate the flow or composition EMU publishes a quarterly magazine on containing more than 300 articles on of potent fluids to the patient’s body. technical medical safety matters and electrically induced accidents in hospitals tries, in various ways, to inform clini- - Handheld phones are recorded as being (2). The cited articles, of course not cians and hospital engineers on such top- able to inhibit fire alarms and surveil- being anywhere near the complete story, ics within the limits set by financial and lance systems. It has been observed nevertheless describe thousands of single personnel resources. that automatic clinical chemistry anal- incidents, many with fatal outcomes. ysis machines have changed their read- Until about 1985, device and component Electromagnetic compatibility problems ings when antennas are close to their defects due to sub-standard construction have always been a matter of concern by circuitry. and maintenance counted for some 80 % the use of electrical devices connected to of all accidents reported (3,4). Today, the the patient. Living tissue contains elec- - Electrical wheelchairs’ controls are accelerating use of highly complicated, trolytes and the geometry of the human susceptible to fields from transmitters miniaturised, and processor-controlled body makes it susceptible to electrical and similar electronic devices. The equipment and the development of clini- and magnetic fields. Sensors, transduc- Radiostøykontrollen (the Radio Inter- cal practice and the steady movement of ers, electrodes, and fixtures on and inside ference Service) has made both field the limits of the possible, some 80 % of the body are leads to electric currents and and laboratory (unpublished) tests on equipment-induced harms to patients, are liquid infusions are electrolytes as well. some types of battery driven, electroni- ergonomic or users errors problems. Inci- Some commonly used devices like defib- cally controlled wheelchairs – con- dents with EMC are recorded and statisti- rillators, pulse generators to restart a fail- firming that some types are susceptible cal analysis of the registered trends indi- ing heart, and high frequency surgical to electromagnetic fields from, for cates that the problems will be increas- diathermy, generate extremely high example, cash registers and mobile ingly important in the future. power densities, >104 W/cm2, when in telephones. use, and tend to disturb everything in the The Norwegian Electrical Safety Direc- neighbourhood. Most clinicians are Many accidents are believed to be torate established the Special Inspection aware of these unavoidable problems, induced by conducted interference, but for Electrical Medical Equipment, and hospital procedures are accordingly the facts and knowledge are scarce. Nor- STEM, in 1980. In 1992, STEM merged modelled. Serious problems arise when wegian findings compare well with the with the Electrical Safety Directorate as electrical medical equipment is used in ECRI, USA (6). the Unit for Medical Devices, EMU. adverse environments and situations. EMU has obligations and rights in The investigation of incidents involving Modern medicine tends to relocate more accordance with given instructions based EMC are inevitably complicated, the and more patient care and treatment out- on law to: reproducibility being very low due to side the traditional institutions. missing information on the actual geome- - Do technical inspections on installa- Typical reported EMC incidents are: try and electronic systems. The tions and equipment in all the hospitals Radiostøykontrollen has been most help- once a year and in other health institu- - Disturbance of neighbouring devices ful to assist the EMU with competence tions when deemed necessary. and navigation systems in ambulance and measurements, both in hospitals and aircraft on patient defibrillation. - Run a national vigilance system to in the laboratory to establish actual emis- receive and investigate reports on acci- - Disturbance on recorded and monitor- sion from mobile telephones and immu- dents and near-accidents in the health ed physiological signals, notably ECG nity of some widespread medical infu- and social welfare area. All personnel etc., from radio transmitters, power sion pumps and a typical dialysis categories are obliged to report inci- transformers, navigation systems, etc. machine (7). dents without delay, i.e. by telephone. The physiological signals are micro- The Radiostøykontrollen has carried out EMU receives some 200 reports a and low milliVolt levelled. tests on field emissions and conducted year. 10 % are fatal incidents. It is esti- - Portable telephones have become an RF from diathermy devices to establish mated that non-reporters may represent important source of electromagnetic levels of signal noise both in the neigh- commensurable numbers, primarily emission disturbing medical equip- bourhood of the active electrode and on minor harms and near-accidents (5,6). ment. EMU has registered an in- the power line (8). The levels and spectra

50 registered indicate that diathermy is an 5 STEM-report (in Norwegian). 10 års important source of potential harm to erfaring med undersøkelse av uhell almost any other type of instrumentation og feil på medisinsk teknisk utstyr i in hospitals, both battery operated and norske sykehus. 1990. power line connected. These results and those earlier reported (7) are one of the 6 Fønstelien, E, Nilsson, F (in Norwe- background data sources for the develop- gian). Uhellsetterforskning ved Det ment of the IEC 601-1-2 standard, the særlige tilsyn med elektromedisinsk main standard on EMC and medical utstyr. Tidsskrift for Norsk Læge- devices. forening, 109 (16), 1801–1803, 1989. This article gives a short overview of the 7 Fønstelien, E (in Norwegian). Felt- activities at EMU, the medical device immunitet og mobiltelefoner. Helstj. unit at the Electrical Safety Directorate in med. teknikk, 9, 10–12, 1991. Norway. The directorate has seen important co-operation with the Radiostøykon- 8 Fønstelien, E (in Norwegian). Om trollen, the EMC work at the EMU hav- nettbåren elektromagnetisk støy; med ing been completely dependent on the særlig referanse til forholdene under laboratories and the competence at bruk av diatermi. Helstj. med. Radiostøykontrollen. The findings, inci- teknikk, 6, 24–26, 1992. dent reports and general information on medical device EMC are regularly published to keep Norwegian hospitals, clinical workers, and engineers continuously up-dated to enhance safety to patients.

References

1 Bruner, J M R. Hazards of electrical apparatus. Anaesthesiology, 28 (2), 396–425, 1967.

2 Pacela, A F. Bibliography on electrical hazards, safety & standards in biomedical engineering. Report CRR-21A. Beckman Inst. Inc. Fuller- ton C 92634.USA, 1970.

3 Fønstelien, E. Comments from a government. Proceedings Workshop on the European standardisation of medical devices, Brussels, Dec. 1988, Brussels, CEN, 317–322, 1989.

4 Fønstelien, E. Equipment failures, patient safety and governmental control. Proceedings of the World congress on health technology standards, 1990, Dublin, IEC/ISO, 509–512, 1991.

51 Company EMC organization

BY AGNAR GRØDAL

Introduction The person in charge of the EMC activi- - He should check that the choice of ties should have a similar function as the components and the hardware design New electronic products have to meet the person in charge of the quality activities. principles have a fair chance of meet- EMC specification, and from 1996.01.01 The EMC activities may in fact be de- ing the system EMC specification. the EMC Directive will apply. One fined as a part of the quality work, and should be prepared for meeting the EMC - He should check that the placement of the two job functions may, especially in specification when writing the product components on the printed circuit smaller companies, be performed by the specification and during the development boards and the layout of the signal and same person. phase, rather than hoping for the best ground return tracks will have a fair during the EMC test phase. This as well It is important that the EMC responsible chance of meeting the EMC specifica- as other EMC related activities will person and his job function is backed by tion. involve many departments and individu- a mandate, to guarantee that the relevant - He should check that the mechanical als in the company. To succeed, all these EMC activities will be performed com- design will give the product a fair must co-operate. pulsorily for all products that are design- chance of meeting the EMC specifica- ed and manufactured by the company. The technical problems that must be tion. solved are complicated, have many It is likewise important that all EMC - He should check that the EMC issues facets, and require competence in various activities are documented, and that the are discussed during all design re- mechanical and electronic design topics, documents – especially the test docu- views. often of high-frequency radiotechnical ments – are filed and stored in a safe nature. manner and over a time period as speci- - He should check that all work per- fied in the EMC Directive. formed by consultants has taken EMC The investment in, and the use of EMC into account. measuring instruments, also need special The work burden on the EMC responsi- competence. ble person may well result in a full-time - He should check that all purchased job. He must thus be given adequate time sub-units have the relevant EMC cer- The field of EMC problems involves and budget resources, as the job seldom tificates. even legal aspects. It is necessary to can be performed as an extra job for an know which EMC norms and other - He should check that all EMC measur- employee already occupied with other norms and regulations in the vast jungle es are adequate, but to the lowest pos- duties. of such related documents that apply to sible cost. the actual product. It is difficult to measure or to document - He should check that the product is in terms of money how profitable the Documentation of the EMC activities is tested according to the relevant EMC work of the EMC responsible person will also a comprehensive, essential, and time norms. be. It is therefore essential that the com- consuming activity. pany management has a correct under- - He should be responsible for the cost Solving the various EMC related prob- standing of the necessity of the EMC analysis, evaluation of usefulness, and lems as they evolve, and deciding who work, and that it supports it. the purchase of company EMC test should be responsible in each and every instruments. The person appointed as company EMC case, will inevitably lead to frustrations responsible may have the following - He should be responsible for the writ- and delays in the development process. duties: ing of the products’ “suppliers declara- To succeed, it is therefore vital that the tion of conformity” document. - He should always be updated on EMC company has some kind of organization norms and regulations. - He should be responsible for new for the various EMC activities. EMC testing and for revision of the - He should know which literature that relevant EMC documentation when a can be recommended. EMC responsible product is modified. To succeed with the various EMC activi- - He should know where to find the best - He should be responsible for the CE- ties, it is important that the company consultancy. marking of the products. appoints one person responsible for all - He should know and be able to recom- EMC activities. He should not necessar- - He should be responsible for the mend the best test laboratory in each ily be competent in the whole EMC field, archiving of all EMC documents for particular case. nor should he himself perform all the rel- the products. evant work. - He should be responsible for the EMC training of the employees. It is important to realize that issuing a He should, however, see to that this work “suppliers declaration of conformity” is being done, and he should preferably - He should gather and spread informa- document, or CE-marking a product also be EMC competent and perform tion on relevant EMC courses and without having the relevant confirming central parts of the work. seminars. documents is a criminal offence. The An important part of his work should be - He should check that the functional documents shall clearly show that the to make the company aware of the EMC specification of a new product incorpo- product meets the relevant specifications. issue, as it most probably, at least to rates the correct EMC specification. One should observe that an external some degree, will be the concern of all - He should check that the clients’ EMC EMC test laboratory normally does not employees. demands are included in the product issue any kind of “certificate”, just a specification. measuring protocol. The laboratory is

52 therefore not responsible for issuing doc- ceptibility (EMS) are given in the rele- often be counted in months rather than uments implying that the relevant specifi- vant norms. The aim should always be days or weeks. cations are met. The laboratory is, how- CE-marking of the product. This marking The standardization of components must ever, responsible for having performed is mandatory from 1996.01.01. Some always take their EMC properties into the ordered tests correctly. The producer products on the market already have this account. The EMC responsible person is responsible for the contents and the marking. may in these matters often be more com- writing of the “suppliers declaration of petent than the local component responsi- conformity” document. He is also Standardization of ble person. responsible for having performed the relevant tests. components There has, over the last few years, been a Design review CE-marking of a product implies not radical increase in the number of new Design review of a circuit diagram and only that the EMC specification is met, component types; this despite some “exits accompanying parts list has shown to but also that it meets other relevant perts” believing that this evolution be a very efficient means to disclose Directives and norms. A typical example should have levelled out some years ago, design errors, both functional errors and is that a product cannot be CE-marked as more and more logic hardware was economic ones. without meeting the safety norm EN supposed to be hidden in VLSI (Very 50950. Design review is also very efficient for Large Scale Integration) components. securing design principles, choice of The EMC documentation shall be at dis- The new components, both active and components, placement and orientation posal on request from the relevant passive types, have changes in their inter- of components on the circuit board, and inspecting authority for ten years after nal technology as well as in their physi- design of the printed circuits so that the the last sample of the product was manu- cal size. generation of radiation will be suffi- factured. ciently low, and so that the product will Most new component types have better be sufficiently immune to incoming radi- EMC qualities than the former similar EMC specification ation and overvoltages – all at the lowest types, but some have worse or even sig- A product’s possibility of success in the possible cost. nificantly worse EMC qualities. It is market can be studied before the product therefore important to choose component Design review should be performed is realized by reading its system specifi- types with sufficiently good EMC quali- using documented guidelines, where cation. This document is, however, nor- ties, but above all, to avoid the worst EMC should be among the main topics. mally incomplete and erroneous. types. Design review is a part of the quality The project manager and the EMC standard EN 29001, and the review pro- The process of choosing the right compo- responsible person are responsible for cedure is documented in the standard nent types has, however, the last few checking that the EMC requirements are IEC 1160, Formal Design Review. years tended to be more difficult than incorporated in the product specification before. The reason for this is the increas- All product designs, whether they are before the start of the design phase. ed and vast diversity of component types, made by company employees or by an Neglecting this will inevitably have and a seemingly decreasing interest in external consulting company, should costly consequences. component technology among hardware undergo the same review procedure. The The product specification has two main designers. procedure should also cover the review parts, the functional specification and the of purchased OEM products, Original A new problem is the decreasing possi- EMC specification. Both will affect the Equipment Manufacturer products, sub- bility of getting components with equal logical design and the types and number units that will be incorporated in the function and quality from more than one of components that will be used. The company’s own product. vendor, the so-called “second source” product specification will also incorpo- policy. Each vendor prefers more and Succeeding with EMC during the design rate the safety specification and the envi- more often to give his particular versions phase requires that the company has an ronmental specification relating to of standard function components some appointed and competent design review mechanical and climatic endurance such specific attractive properties, which group of which the EMC responsible per- as vibration, bump, temperature and hopefully will increase the sale. These son is a member. humidity. components may then not have second It is very important that the leader of the Producer and client should agree upon source components in the market. Even if review group has sufficient mandate, so the product specification before the start components have functional second that the design review process can be of the development process. sources in the market, their EMC proper- performed for absolutely all designs and ties may be different. This can cause International standardization work is revisions of designs, and that this work products that meet the EMC require- continually aiming to put different elec- will be documented. ments to fail the same requirements if tronic equipment into different “product functional second source components are categories”, each with a specific set of used. Printed circuit board EMC requirements. When developing a The printed circuit boards, their types new product one must first decide to Components with superior EMC proper- and their number of layers are normally which “product category” it belongs. ties are often higher priced; their prices discussed for the first time late in the Based on this decision the detailed are often unreasonable for use in low cost design phase, often after the logical dia- requirements to Electromagnetic Interfer- products. Their availability may also be a grams are designed. However, as a rule, ence (EMI) and to Electromagnetic Sus- problem, and the lead times – the time it is better to consider the number of lay- between ordering and delivery – can

53 ers as early as possible when writing the ing machines, changing from single wave demands that the product has been tested product specification. This because an machines to double wave machines and on an open area test site (OATS) as de- unnecessary high number of layers will to infrared soldering machines. fined in EN 55022, or in a large anechoic be costly, and because one may never test chamber with properly controlled Newer component types with shorter dis- know if a lower number of layers would and acceptable unwanted reflections and tance between terminals, circuit tracks have been sufficient. resonances. with shorter insulating distance, and the High EMC requirements will as a rule mounting of components on both board The EMC responsible person should be require multilayer boards. Signal track sides may require even newer soldering well informed of all domestic and some routing on densely populated boards will principles and machines, which again foreign test laboratories, their strengths also as a rule require multilayer boards. may require new investments. and weaknesses, their accreditation status, their accessibility and their pricing Design of the signal tracks and of the This escalation of investments in produc- policy. ground return current system without tion machinery partly caused by the considering EMC will as a rule result in EMC requirements, may force smaller It is important that the EMC responsible boards that have a limited chance of companies to let the mounting of compo- person is well informed of the most used meeting the EMC requirements. One nents and the soldering process to be per- test norms, and of their practical execu- may in that case be forced to revise the formed by external companies special- tion. He should also be present during the board design one or more times until ized in production. testing to gather experience for future tests eventually and hopefully will show EMC work and decisions. that the EMC requirements are met. Instruments It is quite essential for an acceptable Companies developing and producing Domestic EMC result that the printed board designer is electronic products, no matter how small, competence centres trained in EMC board design. It is also should invest in EMC measurement The EMC responsible person should be important that he co-operates closely instruments, both for measuring emission well informed of all domestic compe- with the designer of the circuit diagram. and for testing the products’ immunity to tence centres such as: He should also be in close contact with overvoltages and static electricity. The manufacturers of printed circuit boards, type and number of instruments may - Testing laboratories and be able to evaluate their quality pro- vary, depending on the need and the pos- - EMC consulting companies file, their punctuality, and the total price sibility to borrow or rent. of a printed circuit board, dependent on - Publishers of normative documents A useful first set of instruments may con- its number of layers and the pricing pol- sist of a spectrum analyzer including test - Sales offices of domestic and foreign icy of the producer. probes, antennas, plus an electrostatic normative documents. The PCB designer must continually have discharge (ESD) pistol. The set may cost competence in evaluating the cost effec- about GBP 20,000. tivity of the used CAD system, and be Consultants A miniature test chamber, a TEM-cell able to evaluate if a two-layer solution Succeeding with EMC and other relevant (Transversal ElectroMagnetic cell), may can be used without getting too serious disciplines necessary to develop a new also prove helpful. The smallest cham- EMC problems, and also be able to eval- electronic product requires a very broad bers are priced at about GBP 10,000. uate if autorouting of the board tracks competence, not only in technical mat- can be used. One should observe, how- EMC measurements based on the above ters, but also in a complicated and con- ever, that most autorouting systems of mentioned instrumentation will give indi- stantly growing jungle of norms and reg- today result in a rather poor EMC perfor- cating measurements only, but will prove ulations. mance. very helpful for checking EMC modifica- Few small or medium sized companies tions on prototypes. Correct use of the Well screened circuit boards having can afford or hope to have internal com- ESD pistol may give absolute-value mea- moderate EMC requirements may be petence to handle all these questions. A surements, as done in a test laboratory. based on a two-layer solution. High vol- way out of the problems is to use special- ume production series, where a higher To achieve absolute-value emission and ized external EMC consultants. This cost printed circuit board design can be susceptibility measurements, a reflection- may, however, sometimes give rise to acceptable, may also be based on a two- free test room must be used. This is other problems, as many companies have layer solution. affordable for the largest companies only, decided to solve all problems with the but reasonably acceptable results can be use of internal workforce only. They Recent practice based on EMC knowl- obtained in a sufficiently large empty cel- have usually no budget for the use of edge has shown that some complicated lar room or other less valuable space. external consultants, and see this as extra boards with high EMC requirements may unnecessary expenses. be based on only two layers, or even on Investments in test equipment may seem just one layer. expensive, but one should not forget the This way of thinking may indeed prove necessary additional investments in test- to be expensive. ing competence and manpower. Soldering process The EMC responsible person should be The change from hole mounting technol- well informed of the consultants market, ogy to surface mounting technology has Test laboratory and be able to recommend a specific con- resulted in smaller board areas, and thus The writing of a “supplier’s declaration sultant for solving a specific EMC prob- to lower emissions. This change will nor- of conformity” and the CE-marking lem. mally require investment in new solder-

54 Approval of products The “Technical Construction File” route specifies a description of the product, a The “approval” of a product may have description of the procedures used to different definitions. “Approval” may warrant the relevant safety requirements, sometimes not be related to EMC, and plus a technical report or certificate sometimes the word is even used to issued by a “Competent Body” stating deceive potential customers. Foreign cus- that the relevant EMC requirements are tomers may have their own definition, met. sometimes more severe and sometimes less severe than the EMC Directive. The use of a “Competent Body” may imply that no tests will be performed. Equipment connected to the Public Telecommunication Network must have Both the “supplier’s declaration of con- “Type Approval” which in Norway is formity” and the “Technical Construction issued by the Norwegian Telecommuni- File” route qualifies for the use of the CE cations Regulatory Authority. Some mark. products are “subject to compulsory cer- The EMC responsible person will be tification,” and some “subject to compul- responsible for the “supplier’s declara- sory registration.” Certifications and reg- tion of conformity”, or the administering istrations are in Norway issued by the of the “Technical Construction File” Norwegian Electricity Inspectorate. All route, and of the CE-marking. type approved, certified, or registered products must comply with the EMC Directive for both emission and suscepti- Competence and bility. Companies producing equipment co-operation that has to comply with the Directive on It may seem as meeting the EMC Direc- Telecommunications Terminal Equip- tive, both technically and administra- ment (91/263/EEC), must also comply tively, may be complicated. Indeed it is, with the quality norm ISO 9001. and it requires a multitude of competence The EMC responsible person must know and duties on behalf of the EMC respon- before starting the design phase which sible person, and of the other involved approval will be relevant for the new employees. The key word to success is product, as this information may influ- competence, and above all, close co- ence the design of the product. operation between Compliance with the EMC Directive may - The client be documented by a “supplier’s declara- - The writer of the technical and the tion of conformity” document, or by hav- EMC specification ing the importer or manufacturer to generate a “Technical Construction File”. - The EMC responsible person The “supplier’s declaration of conform- - The quality manager ity” route is expected to be the most - The designer of the circuit diagram common method. The tests may in this case be performed by the producer, pro- - The component responsible person vided he have the facilities and technical - The circuit board designer competence to do the job. As only the largest companies can afford to have - The designer of the mechanics. acceptable testing facilities, the testing will normally be performed by an extern- EMC is the concern of all employees. al independent test house. The test house issues test result protocols, and based upon these documents the producer will issue a “supplier’s declaration of conformity” document, providing the test protocols show that the relevant EMC requirements are met.

55 The Radio Interference Service in Norway – a historic review

BY SVERRE TANNUM

The beginning of EMC was radio noise For transportation in the coastal districts of faults on overhead power lines was and in most countries this turned out to a 34 feet boat was purchased after the easier. The power lines were also rebuilt be a problem when broadcasting started war. The main purpose, apart from radio and the permanent noise disappeared around 1925. Radio noise and interfer- noise suppression, was to investigate the after a while. ence had to be defeated and on Septem- coverage of long- and medium wave ber 25, 1925, the Norwegian Telegraph transmitters. Administration addressed a letter to the Interference from There was a great need for assistance in Department of Trade concerning a draft the remote areas and the coverage from transmitters of international rules for interference free the Norwegian AM transmitters was often Interference from citizens band (walkie- broadcasting. insufficient and the interference from talkie) and illegal pirate FM transmitters The act (“Lov om tilsyn med elektriske more powerful transmitters in the rest of became a problem. The radio amateurs anlegg”) from 1929 was modified pre- Europe was a problem. The boat was fre- usually operate within their specified liminary for a period of 3 years ending in quently used during the summer and the power- and frequency limits but never- 1936. The proposal of amendment of the crew was made up of different radio noise theless, they can produce interference regulations was put into effect some time inspectors. One of them, though, Mr. and noise because of unwanted detection later. Kåre Kristensen from Tromsø, was a of their signals in stereo sets and low fre- skipper, engineer, navigator, deckhand quency amplifiers even when they are Radiostøykontrollen (the Radio Interfer- and cook, all in one person. turned off. This problem occurred when ence Service) was formally established in radio tubes were replaced by transistors, 1939, and the aim was: The instruments for radio noise local- and even signals from far away broad- ization were very simple in those years, 1 As far as possible to reduce the radio casting stations could be picked up by mostly portable radios were used. The noise from electrical machinery and long loudspeaker cables and detected. noise sources were mainly rotating apparatus before marketing machines, a lot of local DC power plants Protecting broadcasting reception from 2 To localize and suppress radio noise of different types. The distribution net- the signals from neighbouring radio ama- from existing equipment works were worn and repaired by all teurs can be very difficult and often available means – copper and iron in a requires good co-operation from all parts: 3 To contribute to the best possible mixture. Generators from 12 to 2 x 220 the amateur, the listeners and the radio immunity of the receiving installations. volts had not been maintained for many noise inspector. years. A lot of these installations were Radiostøykontrollen was organised under Even broadcasting transmitters can be a dangerous and some had to be closed the Norwegian Telecommunication problem for radio reception. Particularly down. Authorities (Telegrafstyret). the high power transmitters on the island The municipal distribution of electricity of Kvitsøy off Stavanger created prob- The service was established in 6 control developed, the maintenance was better lems for the local population. Medium districts corresponding to the organiza- and in the early 1970s this problem dis- wave (1200 kW) and short wave (1000 tion of the Norwegian Electrical Safety appeared. kW) transmitters interfered with TV and Directorate who is acting as the legal radio reception, video tape recorders authority when injunctions have to be Some districts, like Sogndal and Åndal- (VCR), video cameras and telephones. executed. snes, had got equipment for transmitting Radiostøykontrollen was engaged from radio signals over the power distribution 1979 to 1983 to solve the interference network. Problems occurred with 50 Hz problems and the costs were estimated at The early years of interference caused by modulation of about NOK 270,000. broadcasting power frequency. The reason was usually corrosion which was difficult to locate, Despite the problems and the special sit- and this problem existed as long as the uation during the second World War the The electric tramway transmitters were operating. organization of Radiostøykontrollen was One of the most serious sources of radio continued and the first 10 persons were The high voltage overhead power lines noise in the first period of broadcasting engaged 1940 – 1941. After the confisca- also created radio disturbance. It could be was electric tramways. Broadcasting at tion of radio receivers during the occupa- permanent noise from insulators which that time was performed in the long- and tion, the engineers were employed on might be difficult to get rid of because of medium wave bands and electric transmitter stations and laboratories. the high cost of rebuilding. Tracking tramways and railways was a serious down such noise sources with the equip- problem in Norway, as in other countries. In 1945, bicycles and instruments were ment available at that time was very time put into service and the German supplies The tramway is normally driven by DC consuming. However, surprisingly often were searched for radio noise suppressor and the radio noise from the rectifiers it was a success. The managers of the material, because the noise sources were was transmitted through the wire along local power networks were often numerous and strong after 5 years of war the track and also generated in the motors impressed, but the ones who did not take and insufficient maintenance. and regulators. The most dominating rea- account of it had unexpected breakdowns son for radio noise was the contact Transportation was difficult and public which were much more expensive than a device (current collector) feeding the cur- communications poor, so the employees programmed cut-out. rent to the machinery. of Radiostøykontrollen were among the After the introduction of FM transmit- very first to get licences for buying cars. The generated radio noise is very much ters, the radio noise searching equipment dependent on the magnitude of the DC became more advanced and localization

56 current which is occasionally interrupted The problem with radio noise from the dry weather and almost no possible at the contact point. If the current is tramway disappeared after 1955 when reception during humid weather. This below the critical value, it generates the FM network was introduced. was also demonstrated by tape recordings sparks causing severe radio noise. The in the Court of Law. interruption of the DC current above the It was concluded that the degree of dis- critical value generates an arch which Radio noise from high turbance was dependent upon the type of does not disturb to the same extent. The voltage overhead power program transmitted. Speech and enter- critical value of the current is dependent tainment music required a signal to noise upon the voltage and the material of the lines ratio of 35 – 40 dB, while 20 dB was contact surface. The valuation cases of the 1960s considered sufficient for pop music! The noise problems were most severe Norway’s consumption of electrical The Supreme Court decided, however, with a moving car when the motors were energy is high, and we have a compre- that because of considerable investments not working. During the cold season the hensive distribution network. Long- and and intensive extension of the FM net- noise was reduced because of extra power medium wave reception is often disturb- work, no compensation should be given consumption for heating the tram cars. ed by radio noise from the high voltage for poor long- and medium wave recep- Reducing the noise was tried by in- overhead power lines. Especially the tion. A lot of people were after all given creasing the current with the help of extra corona discharge in humid weather is the FM modules by NVE (the State Electric- load resistors and by installing capacitors source of serious radio noise when re- ity Board). across the main circuit and by increasing ceiving AM (long and medium wave) the contact pressure. A certain reduction while FM reception is not disturbed by was accomplished, but the best results this type of radio noise. Instrumentation were achieved by use of broad and often The quality of the radio noise service Establishment of new lines caused a lot double contact pieces. Different materials depends very much on how experienced of valuation cases where Radiostøykon- were tested and carbon proved to give the the personnel is. Normally it takes 2 – 3 trollen acted as technical expertise in best result. The so-called Fischer phanto- years of practice before full competence courts of law. graph and extra load resistors introduced is reached. a great improvement to the radio noise People subject to disturbance were offer- problem, but resulted in increased costs ed a so-called FM module mounted on which had to be shared by the broadcast- the radio which made FM reception posing service and the tramway company. sible, or alternatively a compensation of NOK 350. Measurements were made and After the war, in the years 1947 – 1950, reports were given by Mr. Falk-Pedersen radio noise from the trolley buses in the to elucidate the problem. city of Drammen caused complaints, and the solution was the same as in the 1930s: It was proved that listeners closer than extra load resistors ensuring the current 100 – 200 m to the lines would have a drain to always exceed 2 – 3 ampères. degradation in their AM reception during S A r o n e l e T : o t o h P Figure 1 Trolley bus S A r o n e l e T : o

400 kV duplex t o h

300 kV simplex P Figure 2 Figure 3 Marconi radio noise instrument

57 But good and proper instrumentation is ment for this instrument and it is still in also very important. Today, we have a use. 30 years is a very ripe age for a selection of suitable instruments to mobile instrument and the reactions from choose from, which is a situation entirely today’s technical environment are not different from the early years. always positive when veteran equipment like this is exposed. A report from 1938 in Hammerfest gives the following description of what was The introduction of television in Norway

S available in the district: (around 1958) meant a challenge to A r

o Radiostøykontrollen concerning signal n 1 Marconi radio noise instrument (LW e l measurements. The signal strength alone e

T + MW)

: was not sufficient for quality judgement; o t o

h 2 Siemens radio noise instruments reflections could damage the picture P completely. Figure 4 The T.K. noise instrument 1 Tandberg radio noise instrument. Small portable TV measuring receivers These pieces of equipment were heavy for combined battery and 220 V supply and impractical. A modernised type were used for the purpose. called T.K. (Thor Kaltvedt) was used from 1940. The instruments were port- Colour TV and Text-TV have further able receivers, more or less modified. increased the demand for better measuring equipment, and during the 1980s From around 1950 the Kurér portable portable colour measuring receivers, radio was used supplied with universal oscilloscopes and frequency analysers instrument and megger. The David were supplied. Andersen portable receiver also had an instrument for signal strength indication. Special equipment for control of cable- S

A TV networks (CATV), ultra sound equip- r

o The FM transmissions in the VHF band n ment for localizing radio noise from e l were started around 1955. This intro- e

T overhead power lines, and equipment for

: duced a new technique because propaga- o t registration of magnetic and electric o

h tion of both wanted and unwanted signals

P fields have been introduced. Figure 5 Kurér portable radio was different. The EMC laboratory is also well The HUZ field strength meter from equipped with measuring facilities, such Rohde & Schwarz covered the frequency as instruments screened rooms and open range from 47 MHz to 225 MHz (Band I, area measuring sites. II and III). This instrument had the possibility to measure spark generated noise and was equipped with a special noise CATV (Cable-TV) detector (CISPR detector) in addition to In 1937 the first description of a network standard mean value detector. for the supply of antenna signal to concentrated flats was given by Christian The instrument had built-in calibration Nome. signal and a tunable dipole antenna mounted on a telescopic tube and it Before the TV age, in February 1956, S

A weighed 4 kg. The HUZ generation guidelines for communal aerial systems r o n lasted for 20 years and was succeeded by (fellesantenneanlegg) were edited. e l e

T an Italian field strength meter, PRES-

: These guidelines were based on research o t TEL, used from 1975 until a few years o

h done by Radiostøykontrollen and con-

P ago, but now succeeded by another Ital- tained the diagram shown in Figure 7. Figure 6 HUZ field strength meter ian instrument, UNAOHM. The transmission of TV signals along the For the lower frequency range (150 kHz Swedish border, especially in the south- – 3 MHz) the Siemens STTM was used eastern part of Norway, started an activ- from 1960, fitted ity in this sector in the 1960s. A docu- with an optional ment dated 26.08.1965 describes 6 net- Antenna ferite antenna works around Oslo containing 20,000 and the possibil- Transformer subscribers. ity to measure ca 300 Ω conducted radio It was evident that there was a need for noise via a probe guidelines and regulations to avoid inter- 1000 Ω or a line ference and to ensure that the subscribers Impedance Sta- were given a decent signal quality and A J A J A J A J bilizing Network good reliability. Earth (LISN). The technical specifications based upon It is difficult to recommendations from IEC are given by Figure 7 Diagram included in the guidelines for communal aerial systems find a replace-

58 Radiostøykontrollen in 1974 concerning signal levels, frequency allocations, 30 m attenuation between antenna outlets, amplitude response, frequency stability and intermodulation plus description of measuring methods. Field In 1975 these specifications were trans- Television stregth ferred into Norwegian norms (No. NEN receiver meter 59.75). However, they were not obliga- 3 m 3 m tory until 25.10.1980, when Norwegian Telecom edited regulations for cable-TV. x At that time, Norwegian Telecom was 1 m osc. the legal authority and the control was executed by Radiostøykontrollen. In 1983, a total of 385,000 subscribers Figure 8 were connected to CATV in Norway and 142,000 had more than 3 channels. Two networks had more than 10,000 subscribers and 15 networks had between 3,000 and 10,000 subscribers. The measurements of conducted and radi- National differences in the regulations In 1984 preliminary regulations were ated radio noise were performed by could lead to barriers of trade, so from given, and the government confirmed the Radiostøykontrollen and the formal this point of view international co-opera- technical regulations as being permanent approval was the responsibility of tion is also wanted. It started with a con- the same year. NEMKO. Radiated noise was measured at ference in Paris 1933 and this led to the a distance of 30 m (see Figure 8). foundation of the international radio The control today is performed by noise committee CISPR. The first meet- Radiostøykontrollen on behalf of The The national regulations were modified ing was held in Paris in 1934 by IEC Norwegian Telecommunications Regula- in 1983 in accordance with international with participants from the different na- tory Authority, which is legally in norms which included measurements of tional committees and representatives charge. conducted radio noise via a special from some international organizations. (delta) network, and the radiated signals were measured at a distance of 3 m. There were only European members in Satellite reception CISPR before the second World War, In the beginning, radio and TV receivers The large number of satellites now oper- and the most active countries were Great were the most numerous test objects, but ating have extended the offer of pro- Britain, France, Belgium, Germany and in later years different types of other grammes considerably. This has led to the Netherlands. The CISPR work was equipment are included, like video re- rebuilding and extension of the cable-TV resumed after the second World War in corders, TV games, electromedical networks. 1946. At a meeting in London it was equipment, satellite receivers and also decided that CISPR should be a commit- In 1989 CATV networks had a capacity some telecommunication equipment and tee under IEC with participation from of 24 channels – the majority being satel- personal computers (PCs). other international organizations. The lite programmes, because local pro- Since 1964, Radiostøykontrollen has also following organizations are now memb- gramme production is expensive. acted as a consultant engaged by the traf- ers of CISPR: The different national The CATV can be used as a general fic authorities for type approval of vehic- committees (NEK in Norway) and EBU, broadband network for transmission of les and for road controls. International OIRT, ECMA, CIGR, UNIPEDE, UIC, other services like alarm, monitoring, standardization has, however, reduced UITP, UIE and IARU. There is also a digital communication, etc. the problem with radio noise from cars close co-operation with CCIR and ICAO. and lorries to a minimum. The use of the network is mainly a politi- The work of CISPR is to establish norms cal question to be decided in the future. concerning radio noise (EMC), defining International collabora- measuring methods and limits. Approval test tion on radio noise Norway participates in CISPR through NEK (Norsk Elektroteknisk Komité). An important contribution to the reduc- reduction Radiostøykontrollen together with tion of radio noise is testing of consumer Since radio transmissions were started in NEMKO have played an active role. equipment to verify if the requirements Norway in 1923 it has become evident are fulfilled. The regulations on radio The Norwegian EMC norms are worked that radio noise from all kinds of electri- receivers based upon IEC/CEE norms out to the best possible compliance with cal equipment has to be fought interna- were put into force on January 1, 1954. the CISPR publications. tionally. Exact limiting values have to be The principal idea is that electrical equip- defined for all kinds of apparatus like The European Union has increased the ment shall not disturb broadcasting household appliances, lighting equip- requirement for co-ordination of norms reception (this is also valid for broad- ment, broadcasting receivers, medical and regulations. The European Commit- casting receivers themselves). apparatus, computers, etc. tee on Electrotechnical Standardization

59 (CENELEC) has established a set of Literature norms on radio noise/EMC (EN 55xxx) which are based upon CISPR publica- Isaksen S et al. 50 år mot radiostøy. tions. Radiostøykontrollen 1939 – 1989. Teledirektoratet, Fagenhet for radiostøy, Abbreviations Oslo, September 1989. CCIR International Radio Consult- ative Committee CENELEC European Committee for Electrotechnical Standard- ization CATV Cable Television CIGR International Conference on Large Electric Systems CISPR International Special Com- mittee on Radio Interference EBU European Broadcasting Union ECMA European Computer Manu- facturers Association EMC Electromagnetic Compati- bility IARU International Amateur Radio Union ICAO International Civil Aviation Organization IEC International Electrotechni- cal Commission NEK Norsk Elektroteknisk Komité OIRT International Radio and Television Organization UIC International Union of Rail- ways UIE International Union for Elec- troheat UITP International Union of Pub- lic Transport UNIPEDE International Union of Pro- ducers and Distributors of Electrical Power

Special

61 62

l=λ/2 l=λ/2 Telephone cables 2 and lines as 2 travelling wave

1.5 antennas 1.5 For example a telephone cable along a valley may 1 1 act as a travelling wave antenna. The cable is hung on .5 brackets of metal as sketched in Fig- .5 ure 3, and the capacitive coupling to the brackets and masts causes a leak- -1 1 -1 1 23age of energy to the ground. In addition, there -.5 are often several taps to -.5 subscribers along the cable, and this also reduces the -1 efficiency of the cable as -1 a travelling wave antenna. However, tele- -1.5 -1.5 phone cables are travelling wave antennas, and -2 when receiving at very -2 low frequencies, this effect may cause serious problems especially for frequency division multiplex (FDM) systems. Power lines as travelling wave antennas If the telephone cable is parallel to l= /2 λ a high voltage power line, we 2 may get another effect. The high voltage lines have very good insulators, that is, insulators 1.5 which are very effective also for high frequencies. These lines often go several tenths of kilometres before 1 there are any taps (to transformers, etc.). High voltage power lines are therefore very good .5 travelling wave antennas. The termination of such an antenna may be in a transformer, but we 6 -1 1 2 3 4 5 may also get a travelling wave effect because of -.5 the losses to ground along the line. If we have several conductors, for instance three as shown in Figure 3, the three -1 conductors get about the same electromagnetic energy, and may -1.5 here be considered as one line. As also indicated in Figure 3, -2 the current in the power line Figure 2 Antenna patterns for travelling wave antennas because of the received electromagnetic field, will give a magnetic field which may induce a We have a special type of travelling wave antenna when we use current in the telephone cable. We also have a capacitive cou- two conductor lines formed as a rhomb. If the conductors are pling, and when there is a mismatch and consequently standing several wavelengths, this antenna type is characterized by a very waves, we may have electromagnetic radiation from the power narrow main lobe in the antenna pattern. line to the telephone cable.

64 Partial mismatch for High voltage the travelling wave power line Equipment

Telephone cable Capacitive coupling Telephone cable Radiation Magnetic coupling

Figure 4 Termination of a telephone cable as travelling wave Figure 3 Telephone cable parallel to a high voltage power line antenna

When underground telephone cable is used, the problem of radi- If we have a standing wave on the outside of a properly termi- ation into the cable is highly reduced. However, the magnetic nated coaxial cable as indicated in Figure 5B, we get the recip- coupling is still there because ground normally gives very weak rocal effect. The standing wave will produce a current in the co- magnetic shielding. axial cable even if it is properly terminated. Here we must take into consideration that the coaxial cable is an unbalanced Mismatch for electromagnetic waves device. If in addition we have a mismatch, the phase conditions will cause a standing wave in the cable. It is now of interest to take a look at the problems concerning mismatch for electromagnetic waves. It is very important to be Concerning paired lines, a mismatched symmetrical line will aware of the difference in matching for a paired line in a tele- have standing waves as sketched in Figure 5C. These standing phone cable and the matching for the telephone cable consider- waves are then both between the conductors and on the two ed as a travelling wave antenna. conductors considered as a single line with ground return. The matching conditions for paired cables are well known. But If we have a standing wave imposed on the balanced line from concerning the travelling wave antenna, the matching conditions outside, we may have a perfectly balanced line (referred to will be more complicated. ground) and with perfect termination, as indicated in Figure 5D. Under these circumstances the standing wave will cause no sig- In Figure 4 are indicated the conditions when a shielded tele- nal between the two conductors. However, in practice we very phone cable is terminated at for instance a telephone exchange. seldom have good balancing, even if we have a twisted line. In The current caused by the travelling wave antenna effect goes addition, the influence of the surroundings causes changes in the via the equipment shielding and to the ground. The current may impedance of a paired line, often resulting in difficult terminat- also go through the electronic devices and thereby cause prob- ing problems. lems in the equipment. In Figure 5E is shown a mismatched balanced line. A standing However, in practice there are seldom good matching for a trav- wave imposed on the line from outside will here result in a sig- elling wave at the telephone exchange. The matching will be nal between the conductors because of the phase conditions at a dependent on the dimensions of the equipment, the earthing mismatch. In this context there is no difference if we use twisted conditions, etc. We therefore get a standing wave on the tele- line. phone cable, and this standing wave may cause severe interference problems especially for FDM systems. In order to stop or reduce the influence of the travelling wave antenna effect, it is possible to have filters for these signals in Also, when there is no shielding in the cable, we will have the every paired line. It is also possible to reduce the travelling same problems. The problems may even be worse because the wave by introducing leakage of energy via taps from the cable current then must go through the electronic devices and to the shielding to the ground. These taps should be done with a resist- ground. ance which will consume a part of the electromagnetic energy. If, as sketched in Figure 5A, we have a mismatched coaxial The problems may then be reduced, but often such attempts will cable with an input signal P we get standing waves in the cable. move some of the problems to other parts of the network. Inside the outer conductor we then have potential differences In principle, it is also possible to have taps to the ground for along the cable. The + and - signs indicate the direction for the high voltage power lines for instance by using high voltage con- potential differences. The result is that these potential diffe- densers. However, such condensers have large dimensions and rences will go through the conductor to the outer side of the are very expensive. In addition, the impedance of such a line outer conductor and cause a standing wave there. This effect varies with the power consumption, causing unstable conditions depends on the thickness of the outer conductor. for matching.

65 The conclusion is that when there are very strong received sig- ++ nals, some result may be obtained by using filters. But often the - - only way of solving the problems for the FDM system is to ++- - avoid using the disturbed channels. P Ro≠Zo Observed problems The problems concerning the travelling wave effect have been Z observed especially in the southern part of Norway. In some o areas the power lines along the valleys have a direction towards the European continent where there are several navigational transmitters in the 20 kHz to 30 kHz band. We here have to take into account the fact that the antenna pattern of a travelling wave antenna has maxima at certain angles to the antenna wire. ++ - - The received signal voltages have been rather high, up to more + - + - Ro=Zo than 0.5 V, and the results have been false calls and even blocking of telephone exchanges. Ro≠Zo When the antenna is directed towards the continent, we may have additional effect. Because of high atmospheric discharge activities such as lightning in continental areas, the natural noise

Zo in continental areas is higher than in coastal areas. The travelling wave antenna effect may therefore increase the noise in the network, and natural noise pulses may cause problems in the telephone systems. References

1 Stokke, K N. Radiotransmisjon: kabler, bølgeutbredelse, antenner. 4. utgave. Oslo, Universitetsforlaget, 1982.

P Ro≠Zo 2 Glazier, E V D, Lamont, H R L. The services’ textbook of Zo radio, volume 5: transmission and propagation. Her Majesty’s Stationary Office, 1958.

Zo Ro=Zo

Ro≠Zo Zo

Figure 5 Standing waves in cables and lines

66 XDR/RPC without TCP/IP Client-server communication with file exchange as transport mechanism

BY PER SIGMOND

1 Introduction 4 Substituting the transport protocol

External Data Representation – XDR (1) and Remote Procedure In order to allow client-server communication without using Call – RPC (2) are the names of two protocols used for client- TCP or UDP, we have to substitute the transport functionality server communication. XDR provides abstract type definition by some kind of file exchange or binary bulk-transfer. In the and concrete transfer encoding syntax. RPC provides a client- case of serial line connections the “modem community” use server message format, and is defined partly with XDR type several file transfer protocols which have proven to be reliable definition. XDR/RPC serves as a basis for the definition and and efficient, and there is no reason for not choosing one of implementation of various client-server protocols such as Net- them for a sample implementation. Among the candidates we work File System (NFS) and Network Information System find the [xyz]modem protocols and kermit. In a local network (NIS). The rpcgen code-generation tool makes it easy to specify environment the file exchange can be done by simple file- and implement customized client-server applications. sharing. But even if we substitute the transport protocol, we XDR/RPC is now regarded as an industry standard. want to preserve connectivity to existing RPC servers that use TCP or UDP for transport. The solution is a “transport bridge”. XDR/RPC is normally closely related to the TCP/IP (3)(4) protocol stack, using UDP (5) or TCP as the transport protocol. In the last few years Transport Independent RPC (TI-RPC) has Client machine Server machine evolved to make this relation less tight, but for the PC (DOS) most solutions use TCP/IP. This means that PC-users with no IP (or SLIP (6) or PPP (7)) network connection lack the ability to 1 5 Client Client Server Server use client-server communication based on the industry standard 10 stub stub 6 XDR/RPC.

This article presents a simple, transparent solution to this prob- 92 47 lem using file exchange as a substitute for the TCP transport protocol. Provided that the PC can exchange files with the 3 Transport Transport server machine, the user can do client-server communication 8 entity entity with his PC as the client. All existing TCP/IP-based RPC- servers are accessible via a special “transport bridge” at the host machine. Thanks to the modularity of the client-server model no Figure 1 The client-server model changes are necessary in the client-code; only the client-stub needs a little tweaking.

Client machine Server machine 2 Client-server model Server portmap starts on server- server Figure 1 shows the general client-server model as machine described by Tanenbaum (8). The main idea is that the client and the server is sheltered from the communication RPC- PMAP Server registers PROC protocol activity by the client- and server-stubs. The server _SET( program-number, pmap) stubs handle all client-server specific protocols (like version-number, protocol and : OK XDR/RPC) while the transport entity provides transmis- onse portnumber with Resp sion of the byte-stream across the network. portmapper

In the model there is a clear distinction between the transport entities and the stubs, suggesting that different (Client-server transport protocols can be used. In traditional XDR/RPC session starts here) RPC- one can choose between UDP and TCP. client Client asks portmapper which port-number the PMAPPRO C_GETPO 3 XDR/RPC client-server session server is registered RT(pmap) with A traditional client-server session using XDR/RPC is ber : portnum Response shown in Figure 2. A session normally has at least two Portmapper answers transactions. The first one starts with a request from the "Our client to the server-machine’s portmap-server. The rea- " RPC Client calls the -prot ocol.. son for this transaction is that the client needs to know server using this . what TCP port or UDP port a specific RPC-server listens portnumber and asks on. Knowing the full transport address of the server, the for a procedure service.... client now transmits the “real” request to the RPC-server optionally followed by other requests.

Figure 2 XDR/RPC client-server session

67 Client machine Transport bridge machine Server machine

1 8 Client Client Transport Server Server 16 stub bridge stub 9

15 2 41312 5 710

3 6 Transport1 Transport1 Transport2 Transport2 14 11 entity entity entity entity

Figure 3 Transport bridge in relation to the client-server model

5 The transport bridge 7 A sample implementation

The task of the transport bridge is to transparently transform the 7.1 The PC-library (client-stub) stub-data from one transport protocol to the other. From the clients-stubs point of view the transport bridge must look like a In an implementation we have to make some choices. The file server stub and vice versa. exchange is done via serial line by the zmodem protocol, and the PC XDR/RPC code is my own port of the SUNRPC-4.0 for The bridge is not supposed to change the format of the stub- the “Waterloo TCP” package called “WATRPC”. I have chosen data, so the XDR format applied to both transports must be the not to incorporate modem initialization and dialling-scripts into same. Because of the stream-nature of the file exchange trans- the XDR/RPC code. This connection set-up (and close) must be port, it seems natural to use TCP as the transport protocol across the IP network. The XDR-bulk therefore has to be Client machine Transport bridge machine Server machine encoded according to the Record Mark- ing Standard for byte stream transport portmap protocols as defined in RFC1057 (2). server

Figure 3 shows how the transport bridge PM RPC- APPRO relates to the client-server model. The C_SET server (pmap “transport1” entities perform the file ) K exchange, while the “transport2” entities Transport se: O espon are TCP-peers. Bridge R

6 XDR/RPC client-server RPC- session using the trans- client

port bridge RP C req uest o file ex ver chang An XDR/RPC client-server session using e-prot ocol... the transport bridge is a bit different than Extract prog. nr. and vers nr. in the TCP/IP case. From Figure 4 we PMAPPRO C_GETPO can see that the client is not issuing a RT(pmap) portmap call. The only reason for the client to call the portmapper is to find the number sponse: port TCP or UDP port. Given that the client Re does not use TCP or UDP, it does not R PC re need the port number. In addition, every quest over T RPC-request carries full information CP/IP about RPC program number and version P/IP er TC number. So the transport bridge can se ov espon extract these numbers from the RPC RPC r request and do the portmap call by itself. ver nse o respo ocol RPC e prot chang file ex

Figure 4 Client-server session using the Transport Bridge

68 done by some other PC-program or manually. When you are #include safely logged in to the host machine (the one with the “transport bridge”) the RPC client-stub takes over. The client itself does main( int argc, char **argv ) { not have to be changed in any way other than specifying “modem” as the transport protocol (the last parameter to the char recfile[ 30 ]; standard clnt_create() procedure). Special “dsz” parameters char sendfile[ 30 ]; like comport/speed settings go into the first clnt_create() param- char cmdline[ 100 ]; eter named “host” (normally carrying the IP host information). The portmap call is skipped (as noted before). if (argc != 2) { fprintf( stderr, "Usage: %s serverhost\n", argv[0] ); The changes in the client-stub limits to a new section in the file clnt_gen.c and a new file clnt_mod.c (builds on the file exit(1); clnt_tcp.c). } strcpy( recfile, tmpnam(NULL) ); 7.2 The host-programs (transport bridge) strcpy( sendfile, tmpnam(NULL) );

The host-programs were developed on a Unix machine. The sprintf( cmdline, "myrz -q -f %s", recfile ); program “recxdr” is started automatically by the sending zmodem program on the PC (the command “rz” is aliased to if (system( cmdline )) { “recxdr serverhost”). “recxdr” administers the zmodem file- unlink( recfile ); transfers via the public-domain rz and sz programs as well as exit(1); the program “rawxdr” which is doing the TCP/UDP part. In } order to avoid name-conflicts, I chose to tweak the rz program to ignore the received file-name and instead use the name sup- sprintf( cmdline, "rawxdr %s < %s > %s", argv[1], recfile, sendfile ); plied by a new -f option. The tweaked rz is called “myrz”. All file names are provided by “recxdr” via the tmpnam() routine. if (system( cmdline )) { You get a good overview of the program flow of recxdr by just unlink( recfile ); looking at the program code (Figure 5). unlink( sendfile ); exit(1); The “rawxdr” program includes a partial portmap-client and code to transmit/receive the client-stub data (XDR-encoded } RPC-message) on a TCP socket. The RPC program- and version-numbers are easy to extract as they are always located at sprintf( cmdline, "sz %s", sendfile ); the same byte offset in the request-message. if (system( cmdline )) { unlink( recfile ); 7.3 Program interaction unlink( sendfile ); To give an impression of how the programs interact, I have exit(1); included a chart showing where the different programs are in } operation (Figure 6). unlink( recfile ); 8 References unlink( sendfile ); exit(0); 1 Sun Microsystems, Inc. XDR: External Data Representation } Standard. RFC 1014, June 1987. Figure 5 The “recxdr” program 2 Sun Microsystems, Inc., RPC. Remote Procedure Call Pro- tocol Specification Version 2. RFC 1057, June 1988. 7 Simpson, W. The Point-to-point Protocol (PPP) for the 3 Postel, J (ed.). Transmission control protocol – DARPA transmission of Multi-protocol Datagrams over Point-to- Internet program protocol specification. RFC 793, Septem- Point Links. RFC 1331, May 1992. ber 1981. 8 Tanenbaum, A S. Computer Networks. Prentice-Hall 1989. 4 Postel, J (ed.). Internet protocol – DARPA Internet program ISBN 0-13-166836-6. protocol specification. RFC 791, September 1981.

5 Postel, J. User datagram protocol. RFC 768, August 1980.

6 Postel, J. Nonstandard for transmission of IP datagrams over serial lines: SLIP. RFC 1055, June 1988.

69 Client machine Transport bridge machine Server machine

portmap server

RPC- server Transport Bridge

RPC- client client invokes client-stub client-stub starts "dsz sz"

recxdr (aliased as rz) is started remotely by dsz recxdr starts myrz myrz exits

"dsz sz" exits recxdr starts rawxdr

client-stub starts "dsz rz"

rawxdr exits

"dsz rz" exits recxdr starts sz sz exits client-stub presents the result to client recxdr exits client gets result Figure 6 Program interaction during client-server session

70 Isolation levels in relational database management systems

BY OLE J ANFINDSEN AND MARK HORNICK

A shorter version of this article was published in Platinum - CS (cursor stability) allows an application to read only com- Edge Magazine, vol 1, No. 1, 1994. mitted data and guarantees that a row will not change as long as a cursor is positioned on it. This is useful, e.g., for an application that fetches a row from a cursor and then performs 1 Introduction some database manipulation based on the current row’s data values, before fetching the next row. In particular, we strongly Database management systems (DBMSs) evolved to allow mul- recommend that you always use at least isolation level CS for tiple users safe access to shared data. Safe access implies that cursors that you intend to use for UPDATE or DELETE oper- different user transactions do not interfere with one another. The ations (otherwise the row in question could change after being activity necessary to do that is known as concurrency control fetched but before execution of UPDATE WHERE CUR- (CC). Most commercially available DBMS we are aware of – RENT OF / DELETE WHERE CURRENT OF). including the DB2 Family, Ingres, Informix, Oracle, and Sybase CS makes sense only if you are locking pages or rows (as – use a CC implementation technique called locking. opposed to locking the entire table). The row/page read lock is kept until the cursor moves to the next row/page. DBMS locking mechanisms are typically complicated yet fasci- nating. To make information systems run smoothly, database - RR (repeatable read) allows an application to read only com- application programmers and database administrators working mitted data and guarantees that read data will not change until in high-transaction-volume environments need a solid under- the transaction terminates (i.e., a read that is repeated will standing of these mechanisms. However, this article will not return the original row unchanged). RR will not prevent the focus on locking per se, but on a higher level concept used by so called phantom row phenomenon, i.e., when a cursor is DBMSs to control the underlying protocols for acquiring and reopened a row not present the previous time may appear. releasing locks. Read locks covering rows retrieved by an RR application must be kept until the end of the transaction, i.e., until the 2 Isolation levels transaction either commits or aborts. - TC (transaction consistency; also known as serializable) Applications have different requirements when it comes to CC: allows an application to read only committed data and some need to execute as if they had the database all to them- guarantees that the transaction has a consistent view of the selves, others can tolerate some degree of interference from con- database (as if no other transactions were active). currently running applications. To meet the needs of different applications, DBMS vendors provide a way to specify different We prefer to refer to this isolation level as TC rather than SR concurrency properties. In the case of relational DBMSs (serializable) because a TC transaction may be part of a non- (RDBMSs), this is typically done by means of isolation levels. serializable schedule where other transactions use lower lev- Although one of the nice things about isolation levels is that they els of isolation. In other words, since serializability is a prop- are implementation independent, we will assume throughout this erty of a schedule of database transactions, attaching the label article that they are implemented by means of locking. Tradi- “serializable” to a single transaction is misleading. On the tional isolation levels include (from least to most restrictive): other hand, if all transactions in a schedule use TC, that schedule will be serializable. (A schedule is serializable if it - UR (uncommitted read; also known as read uncommitted, is equivalent to a serial schedule, i.e., a schedule where each read through, or dirty read) allows an application to read both transaction is allowed to run to completion before the next committed and uncommitted data. can start.) UR applications do not acquire read locks; they simply read All read locks acquired by a TC application must be kept without locking. However, they do acquire write locks (see until the end of the transaction, i.e., until the transaction either below). commits or aborts. If you want to use UR, you need be sure you know what you are doing since the DBMS cannot guarantee the quality of the Each of these isolation levels, except CS, are defined in the retrieved data with this isolation level. There are basically SQL-92 standard, see e.g. (Date and Darwen 1993) or (Melton two situations where UR might be attractive: (1) when you and Simon 1993). Given that CS is very useful for some applica- don’t need an exact answer, and (2) when you know that the tions, we fail to see any good reason for leaving that isolation data you want to read is not being updated by anyone else level out of the standard. We hope users will demand support for (although using table locks is probably a better solution in the all of the above-mentioned isolation levels from DBMS vendors. latter case). We also hope ANSI/ISO will include CS and QC (see below) in the SQL standard. Not only is every single isolation level men- - CR (committed read; also known as read committed) allows tioned in this article useful for certain applications, it is also very an application to read only committed data. much to RDBMS users advantage if different systems have uni- CR can be implemented by “zero duration” read locks. That form CC options (this is, e.g., an interoperability issue). is, if a CR application wants to read some database object, it suffices for the DBMS to check if a read lock could have Readers should be aware that some DBMS vendors (e.g., IBM) been granted. If the answer is yes, the desired object is read implement RR in such a way that phantom rows are disallowed, (but no lock is acquired). thus providing full TC support – “disguised” as RR. Because of this, CR is much cheaper than CS (the next isola- As pointed out in (Anfindsen and Hornick 1994), an important tion level) in terms of CPU cycles. Unless your application isolation level is missing from this set. Consider a banking needs the extra semantics offered by CS (many applications application that executes the query do not), CR is a far better alternative.

71 SELECT SUM(Balance) Figure 1 shows isolation level support in ten commercial systems. FROM Accounts

If isolation level CS or weaker is used and money is transferred 3 Conclusions between accounts while the query executes, SUM(Balance) may Conventional RDBMSs provide applications an incomplete set return an incorrect result. Specifically, money can be transferred of options for controlling concurrency. Partly because they sup- between an account read by the query (but no longer locked) and port only a subset of the standard isolation levels, and partly an account not yet read by the query (and thus not yet locked). because the SQL standard does not recognize the important isolation level query consistency (QC). From our discussion it is In general, isolation level CS can cause problems for any query clear that each of these isolation levels satisfies common appli- that uses aggregation and for any application that processes a cation requirements. The situation is unlikely to change unless cursor where the answer set must be consistent. Using RR or TC DBMS users tell vendors and standard committees that fine- solves this problem, but these isolation levels are unnecessarily granularity control over CC options is of great value to them. restrictive for such applications, since getting the same answer set more than once is not the issue. To provide necessary and Until your RDBMS has QC support added to it, you must use sufficient isolation, we introduce the isolation level query con- isolation levels RR or TC to ensure consistent results for entire sistency: queries, e.g., when using aggregate functions. This reduces - QC (query consistency) allows an application to read only potential concurrency for your applications and overall DBMS committed data and guarantees that all data accessed in a sin- throughput. gle query is consistent For readers interested in a deeper understanding of isolation lev- Since cursors are defined using queries, QC guarantees that all els and transaction management in general, we recommend rows in a cursor’s answer set are consistent. QC is also valuable (Gray and Reuter 1993). when performing statistical analysis, or when comparing or otherwise using together data values from different cursor row occurrences. QC is weaker than RR but stronger than CS. 4 References Anfindsen, O J, Hornick, M. 1994. Application-Oriented Trans- Implementing QC by means of locking is straightforward; all action Management. Kjeller, Norwegian Telecom Research read locks must be kept until the query is completed (until the (technical report TF R 24/94). cursor is closed). Date, C J, Darwen, H. 1993. A guide to the SQL standard. Oracle-specific remark: Oracle7 differs from most other com- Reading, Mass., Addison-Wesley. mercially available RDBMSs in the CC area. It uses multiver- sion CC which causes read-write dependencies to be avoided, Gray, J, Reuter, A. 1993. Transaction processing: concepts and and it does not use the isolation level concept. It is interesting to techniques. San Mateo, Calif., Morgan Kaufmann Publishers. note, however, that the two CC options offered by Oracle7 seem to correspond very closely with isolation levels QC and TC. Melton, J, Simon, A R. 1993. Understanding the new SQL: a complete guide. San Mateo, Calif., Morgan Kaufmann Publishers.

Isolation level DB2/MVS DB2/2 DB2/6000 DB2/400 Oracle Sybase Ingres Informix SQL/DS Rdb

Uncommitted read (UR) Committed read (CR) Cursor stability (CS) Query consistency (QC) W Repeatable read (RR) Transaction consistency (TC)

Supported Proposed isolation level Oracle read consistency corresponds to isolation level QC W Figure 1 Isolation level support in ten commercial DBMSs

72 Mathematical methods and algorithms in the network utilization planning tool RUGINETT

BY RALPH LORENTZEN

1 Introduction transmission systems have bit rate capacities measured in bits per second. Once a year Telenor makes plans for routing and grouping of circuits in the Norwegian long distance and regional networks It is assumed that the reader is familiar with concepts like rout- for a planning period consisting of several years. Up to 1993 ing and grouping of circuits. these plans were made partly manually and partly using the EDP-based planning tool RUGSAM. Groups can be routed on systems with the same bit rate as the group and on groups with higher bit rates than the group in In RUGSAM separate models were used for 64 kbit/s circuit accordance with a given group hierarchy. planning and for planning of circuits with bit rate ≥ 2 Mbit/s. The model for circuits with bit rate ≥ 2 Mbit/s consisted of two An example of a group hierarchy is shown in Figure 1. integer programming submodels, namely a routing and a grouping model, which were run sequentially. These submodels were The arrows show which lower order groups that can be directly large and required several hours running times on a powerful routed on which higher order groups, and the number beside the work station. The partitioning of the problem into a routing and arrows indicate how many lower order groups can be directly a grouping problem also lead to solutions which in many cases routed on a higher order group. had to be modified by the planners. Furthermore, SDH was not supported in RUGSAM. In order to simplify the presentation we consider a system to be a special type of group which we call a system group. A system When the need arose for an EDP-based tool for planning the uti- group is never routed on other groups. So with this notation a lization of PDH/SDH networks the availability of new, power- group may be routed on system groups with the same bit rate as ful optimization software was taken into consideration. It was the group. now possible to create a model which, in addition to treating a combined PDH and SDH network, could consider routing and Actually, RUGINETT works with a slightly more general con- grouping simultaneously. So, in lieu of extending the function- cept of transmission network where the nodes are the transmis- ality of RUGSAM’s models for planning of circuits with band- sion stations and the edges are groups which have the property width ≥ 2 Mbit/s to encompass SDH, it was decided to develop that they are never routed on other groups. These groups are a new model which got the name RUGINETT. called edge groups. Of course, edge groups which are not system groups will in general be routed on other groups. This rout- RUGINETT is thus an integer programming optimization model ing is, however, not modelled in RUGINETT. which attempts to route and group circuit demands in a given PDH/SDH transmission network for each year in the planning The box on top of Figure 1 represents the bit rate 2 Mbit/s. The period such that the cost of new multiplexer equipment is mini- number 30 in the box indicates the number of 64 kbit/s single mized whilst at the same time securing that the routing and circuits which can be routed on a 2 Mbit/s group. The T in the grouping plan for each year does not deviate unnecessarily from box indicates that there exist transmission systems with bit rate previous year’s plan. capacity 2 Mbit/s.

The current version of RUGINETT does not pretend to solve the cost minimization problem to optimality. However, the user interface allows the planner to make modifications to the solu- 2 Mbit/s (30) T tion found by RUGINETT and check feasibility and cost. 4

RUGINETT consists of several modules, each of which can be put into one of two categories: PDH 8 Mbit/s (120) T 162163 - Modules which enable the user to enter, edit and modify input/output data via tables and which prepare the data 4 into formats suitable for the optimization module, - An optimization module which finds solutions to the PDH 34 Mbit/s (480) T SDH 49 Mbit/s (630) routing and grouping problem by integer programming 3 4 algorithms, 3 1 - An optimization module for the determination of stand- PDH 140 Mbit/s (1920) T SDH 155 Mbit/s (1890) T by groups. 4 4 4 In this paper the integer programming models and solution algorithms which form the basis for the optimization mod- 16 ules are described. PDH F 565 Mbit/s (7680) T SDH F 622 Mbit/s (7560) T 16 2 The routing and grouping problem SDH F 2.5 Gbit/s (30240) T A PDH/SDH transmission network consists of a set of nodes which are called transmission stations and a set of edges which are called transmission systems or short, systems. The Figure 1 Example of a group hierarchy

73 The other boxes are given a label PDH or SDH depending on Amongst all routing and grouping schemes which satisfy given whether the bit rate belongs to the PDH or SDH subhierarchy. demands with their diversity and load sharing requirements the Again, a T in a box indicates that there exist transmission sys- problem is to find one which minimizes the cost of new multi- tems with the given bit rate capacity. plexer equipment and which at the same time does not deviate unnecessarily from previous year’s scheme. This routing and Some of the boxes have an F (for ‘fixed’) in them. This indicates grouping problem will be solved using integer programming. that the groups with the corresponding bit rate can be directly routed on a single system only. It will be sufficient for us in this The planner has a problem of a dynamical nature. In establish- case to model the group and not the system it is routed on. ing the best way of using the network he must take into consideration the development over time of network structure and The fat arrows show the direct routing possibilities RUGINETT circuit demands. So, ideally the problem should be solved currently will choose from when new routings are proposed. dynamically where network topology and capacity and circuit demands are taken into consideration simultaneously for all the The edge groups may have one or more numbers called reliabil- years in the planning period. This has been beyond our capability codes associated with them. The idea is that edge groups ity, so we have elected to apply a static model which is run sep- sharing a common reliability code can fail due to the same arately for each year in the planning period. A simple feature ‘cause’. Groups inherit reliability codes from the groups they which supports ‘dynamical use’ is, however, built into RUG- are routed on. INETT. The planner can label certain group sets as target group sets which are then given a bonus. It is recommended that the We shall now define group sets. Edge groups which connect the planner first make a ‘future run’ where the input represents the same two transmission stations, have the same capacity and have situation in a ‘target year’ some time into the future. The group the same reliability codes, make up a group set. Furthermore, sets selected by RUGINETT in this run can then be copied into groups which have the same capacity and which are routed on the regular RUGINETT runs and labelled as target group sets. groups belonging to the same group sets, make up a group set. It is important for the planner to be able to exercise a certain Thus, we can consider reliability codes as being attributes of control over the characteristics of the solution. There may exist group sets rather than groups. group sets which he would label as fixed and which he would require the solution to contain, and there may be other group A class of group sets having at least one reliability code in com- sets which he would label as recommended and which he would mon is called a reliability class. give a special bonus.

In addition to talking about groups routed on groups we can Topologically, group sets connect pairs of transmission stations now talk about group sets routed on group sets. and follow a path in the transmission network. A group set is said to have complete routing if the group sets it is routed on The routing and grouping problem is to establish a routing and completely covers the path of the group set. Otherwise, it is said grouping scheme for a given set of circuit demands with various to have incomplete routing. A group set is said to have no rout- bit rates ≥ 2 Mbit/s between pairs of transmission nodes. ing if it is not routed on any group set. Edge group sets have no routing. There may exist other fixed group sets and recommend- Circuit demands are satisfied or covered by demand group sets or, ed group sets which have incomplete or no routing. The new shorter, demand sets with corresponding bit rate. A 2 Mbit/s cir- group sets which RUGINETT suggests will, however, always cuit demand must thus be covered by 2 Mbit/s demand sets while, have complete routing. for example, a 155 Mbit/s demand must be covered by 155 Mbit/s demand sets. No group set can be routed on demand sets. 3 The integer programming model Circuit demands may be directed or undirected. The corre- 3.1 Definitions and notation for the formulation sponding demand group sets are also denoted as directed or of the constraints undirected. A directed demand set utilizes capacity in one direction only in the group sets it is directly routed on. First we give some definitions. An undirected demand may be subject to a diversity require- A slot in a group is the largest common divisor of the maximum ment. This means that the fraction of the corresponding demand number of groups of each type which can be directly routed on sets in the same reliability class has an upper bound which is the group. So every group set which is directly routed on anoth- less than one. This upper bound is called the diversity factor of er group set occupies an integer number of slots in the latter the demand. A typical diversity factor value is .5. group set.

Similarly, a family of undirected demands may be subject to a We write g∈u(g∈d) to denote that group set g is an undirected load sharing requirement. This means that demand sets for at (directed) demand set which contributes to the satisfaction of most one of the circuit demands in the family can belong to any the undirected demand u (directed demand d). particular reliability class. Such a family is called a load sharing family. Furthermore, we write g∈g’ to denote that a (directed or undirected) group set g is directly routed on group set g’. A demand cannot both be subject to a diversity requirement and be a member of a load sharing family. We also write g∈g’+ to denote that g is a directed demand set directly routed on g’ in positive direction, and g∈g’– to denote

74 that g is a directed demand set directly routed on g’ in negative g withFg > 0 in the same reliability class and direction. load sharing family as g (2) x ≤ R (3) Now we introduce the notation we need for the formulation of rg g bounds and constraints in the integer program. We will later xtg ≤ Tg (4) introduce some more notation which is required for the formulation of the cost function to be minimized. (1) together with (2) express that the number of groups in group set g cannot be lower than the number of fixed groups in the Subscripts: group set. u undirected circuit demand d directed circuit demand (2) sharpens the inequality to equality for group sets which can- g group set not be extended or where an increase in xg would lead to an r recommended increase in load sharing violation. t target c reliability class (3) limits the number of recommended groups in group set g to Rg. f load sharing family (4) limits the number of target groups in group set g to Tg. Constants: Constraints: Fg number of fixed groups in group set g (Fg > 0 for edge group sets) Σg∈u (xrg + xtg + xg) + wu ≥ Du for all undirected demands u (5) Rg number of recommended groups in group set g Σ (x + x + x ) + w ≥ D for all directed T number of target groups in group set g g∈d rg tg g d d g demands d (6) Du number of demand groups for undirected circuit demand u Hg’ (xrg’ + xtg’ + xg’) – Σg∈g’ \ (g’+∪g’ –) kgg’ (xrg + xtg + xg) Dd number of demand groups for directed circuit demand d – Σg∈g’ + kgg’ (xrg + xtg + xg) ≥ 0 for all group sets g’ (7) k number of slots a group in set g occupies in a group in gg’ H (x + x + x ) – Σ + – k (x + x + x ) set g’ g’ rg’ tg’ g’ g∈g’ \ (g’ ∪g’ ) gg’ rg tg g – Σg∈g’ – kgg’ (xrg + xtg + xg) ≥ 0 for all group sets g’ (8) Hg number of slots in a group in group set g –Σ (x + x + x ) / (P D ) + w ≥ –1 P diversity factor for undirected demand u g∈u∩c rg tg g u u uc u for undirected demands u subject to diversity 3.2 Problem variables requirements and reliability Here we define the variables used in the integer programming classes c (9) model. –Σ Σ (x + x + x )/D + w ≥ –1 x number of recommended groups which are used in u∈f g∈u∩c rg tg g u fc rg for load sharing families f group set g (generated only if R > 0) g of undirected demands xtg number of target groups which are used in group set g and reliability classes c (generated only if Tg > 0) where Σg∈f∩c Fg = 0 (10) x number of other groups which are used in group set g g (5) and (6) express that the number of groups in demand sets wuc violation of diversity requirement of demand u for covering demands u and d cannot be less than Du - wu and reliability class c Dd -wd respectively. w violation of load sharing requirement of family f for fc (7) and (8) express that the group sets directly routed on group reliability class c set g’ do not exceed the capacity of group g’. For those group wu unsatisfied part of undirected demand u sets g’ on which no directed demand sets will be considered to be directly routed, (7) and (8) will coincide and will of course w unsatisfied part of directed demand d d not be duplicated in the integer program. All x-variables are required to be integer. (9) expresses that the fraction of undirected demand u covered 3.3 Bounds and constraints by group sets in reliability class c should not be greater than Du. Here we list the bounds and constraints used in the model. (10) expresses (as well as possible within the linear relaxation) that for at most one demand in load sharing family f correspond- Bounds: ing demand sets may belong to reliability class c. x ≥ F if group set g can be extended (1) g g It is important to realize that the bounds and constraints given xg = Fg if group set g cannot be extended, or if g is a above are not sufficient to describe the problem completely. The demand set and there is a demand set constraints (7) and (8) put limits on the number of group sets which can be directly routed on a group set, but they do not

75 model the fact that a group set which is not an edge group set sate for this the user interface contains a wide variety of options must be directly routed on a sequence of consecutive higher for inspection and manipulation of the solution. level group sets. This has one advantage, namely that the integer programming problem permits the existence of user defined The solution method consists of a combination of group sets which have incomplete routing. This is useful when a - linear programming with dynamic generation of constraints group set is partially routed in a network where the planner has and variables no responsibility/authority. Furthermore, as will be seen later, the non-edge group set variables xg will be generated dynami- - variable generation and variable bounding heuristics. cally in the course of the solution process in a way that guarantees that they are directly routed on a sequence of consecutive We start by solving the linear programming relaxation of the higher level group sets. integer program.

3.4 The cost function 3.2 Linear programming with dynamic generation of rows and columns In order to formulate the cost function to be minimized in RUG- The number of group variables x and constraints of type (7) INETT we need to introduce some additional notation: g and (8) is so enormous that it is unrealistic to include them all term Cg termination cost of group g explicitly in the model. Furthermore, the number of constraints of types (9) and (10) is also very large, and only a few of them C cost of routing group set g on group set g’ gg’ will be binding in the optimum solution. We therefore solve the Fr bonus factor for recommended groups linear programming relaxation by using dynamic row and column generation. Ft bonus factor for target groups M 1 penalty factor for not satisfying diversity require- Initially, a restricted version of the linear program is set up and ments solved which has 2 M penalty factor for not satisfying load sharing - all variables of types xrg and xtg requirements - variables of type xg with Fg > 0 M 3 penalty factor for not satisfying undirected demands - all constraints of types (5) and (6) M 4 penalty factor for not satisfying directed demands - constraints of type (7) and (8) corresponding to sets g’ with F > 0 or R > 0 or T > 0 (no duplication if (7) and (8) The cost function to be minimized is as follows: g’ g’ g’ coincide for a given g’) Σ C x + Σ F C x + Σ F C x g g g g r g rg g t g mg - constraints of types (9) or (10) containing at least one of the 1 2 3 4 + M Σdc wdc + M Σfc wfc + M Σu wu + M Σd wd above-mentioned variables term - all relevant bounds. where Cg = Cg + Σ{g’;g∈g’} Cgg’ Based on the shadow prices associated with the optimal solution We will not treat in detail how the individual cost elements are of this restricted linear program new groups with associated derived from the input data to RUGINETT here, but limit our- variables of type x and new constraints of types (7), (8), (9) selves to the following remarks: g and (10) are generated, and a new (restricted) linear program is solved. This process is continued until it can be ascertained that C term reflects the cost of multiplexing equipment used in both g we have an optimal solution to the complete linear program. ends of group g, and depends only on the bit rate of g. In order to describe the details of this iterative process we C reflects the cost of multiplexing group g into group g’ and gg’ assume that we are at a stage where we have just found an opti- depends only on the bit rates of g and g’. mal basis for a restricted linear program which, in addition to the initial constraints and variables, may include Fr and Ft lie between 0 and 1. Small values of Fr and Ft imply high desirability of sticking to recommended groups and target - some additional variables of type xg groups respectively. - the constraints of type (7) and (8) corresponding to the additional x variables (no duplication if the constraints coincide) M1, M2, M3 and M4, are large positive numbers. M3 and M4 g’ should be much larger than M1 and M2 because it is more - constraints of types (9) or (10) which contain at least one of important to cover the demands than to satisfy diversity and the additional xg variables. load sharing requirements. First, we check our basic solution to see if it violates any potential constraints of type (9) and (10). If it does, such violated con- 3 Solution method straints are added to the linear program which is reoptimized. 3.1 General This process is continued until all potential constraints of type (9) and (10) are satisfied. The integer program is too large and complex to be solved to a theoretical optimum, so some heuristics are used. To compen- The following notation is used for the shadow prices associated with a basis for the restricted and the complete linear program:

76 Constraint type: Shadow price: G denote the group sets g’ for which (7) and (8) coincide and has been generated (5) πu (6) πd G denote the group sets g’ for which neither (7) nor (8) has (7) πg’+ been generated. (8) πg’ – (9) π In section 3.2 we stated that we could choose whether we would cu consider the slack variable in (7) or (8) corresponding to a g’ in (10) π cf G to be basic. When we establish the reduced cost for an undi- If (7) and (8) coincide, their common shadow price is denoted rected demand set variable directly routed on a g’ in G , (7) by πg’. and (8) corresponding to g’ coincide. When we establish the The optimal solution to the restricted linear program is obvious- reduced cost for a directed demand set variable xd routed on ly a feasible solution to the complete linear program. We shall g’+, we elect to consider the slack variable associated with (8) now establish whether it is optimal or whether we can find one to be basic and vice versa. This choice will make the reduced or more not yet generated xg variables with negative reduced cost as large as possible. We permit ourselves to denote the cost which we can introduce into the linear program together shadow price πg’+ or πg’– corresponding to the not generated with the necessary constraints and continue the solution process. constraint (7) or (8) whose slack variable is non-basic, and thus First, we shall extend our optimal basis for the restricted linear may be non-zero, by πg’. program to an imagined feasible basis for the complete linear The reduced cost for of the demand group variable x contribut- program. We imagine for a moment that we have added all the g ing to the directed demand d can be written: not yet generated variables and constraints to the restricted pro- term gram. We may consider the slack variables of the not yet gene- – πd + Cg rated constraints of type (9) and (10) to be basic. If, for a group + Σ + + (C + k π +) set g’, one of (7) and (8) has been generated but the other not, we {g’∈G ;g∈g’ } gg’ gg’ g’ consider the slack variable of the not generated constraint to be + Σ{g’∈G–;g∈g’–} (Cgg’ + kgg’πg’–) basic. If (7) and (8) coincide for a not generated group set g’, we + Σ{g’∈G;g∈g’} (Cgg’ + kgg’πg’) consider the corresponding xg’ variable to be basic. This implies that πg’ is non-negative. If they do not coincide, we still consider +Σ (C + k π ) {g’∈G ;g∈g’} gg’ gg’ g’ the corresponding xg’ variable, and, in addition, the slack vari- (11) able of one of the constraints of type (7) or (8) of our choice to be basic. We shall return to this choice later. This implies that whilst the reduced cost for the demand group variable xg con- one of the shadow prices πg’+ and πg’– is zero and that the other tributing to the undirected demand u can be written: is non-negative, which in turn implies that the reduced costs of – π + C term the slack variables in (7) and (8) are non-negative. u g + Σ{g’∈G+;g∈g’} (Cgg’ + kgg’πg’+) Since all not generated xg’ variables are basic, their reduced cost is zero. Furthermore, since the slack variables for the not yet + Σ{g’∈G–;g∈g’} (Cgg’ + kgg’πg’–) generated constraints of types (9) and (10) are basic, the shadow + Σ (C + k π ) prices for these constraints are 0 since the slack variables have {g’∈G;g∈g’} gg’ gg’ g’ reduced cost zero. Thus, the diversity and load sharing violation +Σ (C + k π ) variables have positive reduced cost. Since the shadow prices {g’∈G ;g∈g’} gg’ gg’ g’ for the constraints (7) and (8) are non-negative, the slack vari- + Σ π /(P D )(if u is subject to a diversity ables of these constraints have non-negative reduced cost. {c;g∈c} cu u u requirement)

Therefore, when we search for new candidates in the column + Σ{c;g∈c} πcf /Du (if u belongs to a load generation process it suffices to consider xg variables corre- sharing family f) (12) sponding to demand sets. The values of the shadow prices πg’+, πg’– and πg’ are known The demand u may be subject to a diversity requirement or for all g’∈ G+, G –, and G respectively. Furthermore, since the belong to a load sharing family f (but not both). slack variables for the not yet generated constraints of types (9) and (10) are basic, the shadow prices πcu and πcf for these are 0. We shall now establish the reduced cost for a new demand set It remains to establish the values of the relevant shadow prices variable for a given demand relative to this basis for the com- πg’ for g’∈G (if such g’-s exist). Since the corresponding xg’ plete linear program. variables are basic, they price out zero. Furthermore, since the 3.3 Finding reduced costs for demand sets g’-s cannot be demand groups, the xg’ variables do not appear in any constraints of types (9) or (10). This means that: Let H π – Σ k π G+ denote the group sets g’ for which (7) has already been g’ g’ {g’’∈G;g’∈g’’} g’g’’ g’’ generated – Σ{g’’∉G;g’∈g’’} kg’g’’πg’’ = Cg’ (13) G– denote the group sets g’ for which (8) has already been generated

77 network of group sets g’’’ connecting pairs of nodes in the We make the assumption that all directed demand sets corre- transmission network where the length of a group set g’’’ is sponding to a directed demand are directly routed in he same given by: direction on any given group set.

(1/Hg’)(Cg’g’’ + kg’g’’(1/Hg’’)(Cg’’g’’’ + kg’’g’’’ πg’’’)) To see that we really have obtained the optimal solution we consider a not generated group set g’ on which directed demand + Σ π /(P D ) {c;g’’’∈c,(9) already generated for (c,u)} cu u u sets contributing to a given directed demand d may be directly + Σ{c;g’’’∈c,(10) already generated for (c,f)} πcf /Du routed. We can have two situations: In an optimal solution no directed demand set contributing to This reasoning can be continued until we encounter shortest d is directly routed on g’. We can then ignore the reduced path problems in networks of edge group sets where the lengths costs for demand set variables contributing to d which cross are known. the not generated constraints (7) and (8) corresponding to g’. So, the procedure will be to start using Floyd’s algorithm to In an optimal solution directed demand sets contributing to d establish shortest paths between all pairs of nodes in networks are directly routed on g’ in positive (negative) direction. We where the edges are edge group sets at a given level. The results can then ignore the reduced costs for demand set variables are then used to set up a new network where the edges with contributing to d which cross the not generated constraints (8) their lengths correspond to the shortest paths with their lengths. ((7)) corresponding to g’. Then we again use Floyd’s algorithm to establish shortest paths for group sets at different levels which can be routed in this net- In both situations we see that we may assume that the non-basic work. This is continued recursively until we get to the establish- directed demand set variables have non-negative reduced costs ment of the shortest paths for our demand sets. This is done relative to a unique basis. using Dijkstra’s algorithm. 3.7 Integer programming heuristics For demands with diversity requirements the lengths of the edges in the networks may vary from demand to demand, so When the linear programming relaxation of the integer program that Floyd’s and Dijkstra’s algorithms must be run separately has been found, a lot of the group set variables will normally be for each demand. Likewise, for demands with load sharing fractional. In order to reduce cost the solution ‘cheats’ us by requirements the edge lengths may vary from load sharing fam- using fractional group sets. ily to load sharing family, so that Floyd’s and Dijkstra’s algorithms must be run separately for each family. The basic principle used in the heuristics is to solve the linear program and inspect the solution to see if it contains fractional We see that we pay for not having established the complete lin- group set variables or load sharing violations. If so, new bounds ear programming matrix by having to solve a very large number are added to the linear program. The linear program is solved of shortest path problems. Therefore, care has been taken to again and checked for fractional group set variables or load code the establishment of the networks and the shortest path sharing violations which give rise to additional bounds, etc. algorithms efficiently. This process is continued until all the group set variables are integer and all load sharing requirements are satisfied. 3.5 Introducing new variables and constraints into the linear program A typical situation is illustrated in Figure 2. Three system group sets of value 1 with the same bit rate meet at a transmission By the procedure described above we establish in general some node A. Group sets 1 and 2, both with the same bit rate as the demand set variables with negative reduced costs. These system groups and with values .75 and .25, are routed on the demand set variables, together with the variables corresponding system sets as indicated in Figure 2. This is illegal since there is to all sets the demands are routed on, are introduced into the lin- of course room for at most one of the group sets 1 and 2 on the ear program. Furthermore, all corresponding constraints (7) and system set to the left. The corresponding constraint of type (6) (8) relating to new variables xg’ are introduced. for the system set to the left is, however, satisfied. When no more demand set variables with negative reduced costs are found, it is checked whether all potential constraints of types (9) and (10) are satisfied. If not, the unsatisfied constraints x =.75 are established and introduced into the linear program which is 1 resolved again with dynamic generation of demand set variables A and corresponding constraints. This process terminates when all minimal reduced costs of demand set variables are non-negative and all potential constraints of type (9) and (10) are satisfied.

3.6 Optimality criterion for the linear program

x3 x =.25 The algorithm will terminate with the minimum reduced costs 2 corresponding to all demands being non-negative. Since the basis relative to which a directed demand set variable is priced out depends on the variable, it is not immediately obvious that we have obtained an optimal solution to the complete linear program. Figure 2 Example of an illegal fractional solution

However, 140 Mbit/s reserve group sets may be directly routed This integer program will be solved by first solving the linear on 155 Mbit/s group sets. The reserve group sets are perma- programming relaxation of the problem. This is followed by an nently set up and can be used by different stand-by group sets integer programming heuristic. corresponding to different failure situations. 4.3 Solving the linear programming relaxation 4.2 The integer programming model The number of x and x variables and type (15) constraints in An integer programming model has been implemented for r s the linear program is too large to be included in the model from stand-by group set planning. In order to describe this model the the outset. They will be generated dynamically during the solu- following notation is used: tion procedure. Subscripts: Initially, a restricted version of the linear program is set up and g group set for which a stand-by group set is to be found solved which has s stand-by group set r reserve group set - all variables of type xs with Ls > 0 c reliability class - all variables of type xr on which xs variables with Ls > 0 are directly routed Constants: |g| number of groups in the group set g - all variables of type xr with Lr > 0 g(s) the group set for which the group set s is stand-by - for each reserve group set level, at least one variable of type x between every pair of nodes where it is possible to estab- C cost per group in reserve group set r r r lish a reserve group set at the relevant level C cost per group in stand-by group set s s - all constraints of type (14) and (16) Fg’ number of free 140/155 Mbit/s channels in group set g’ - all constraints of type (15) involving xs variables with Ls > 0 NC number of reliability classes - all relevant bounds. Problem variables: Initially, there may, for each reserve group set level, exist node xs number of groups in stand-by group set s pairs with no xr on which xs variables with Ls > 0 are directly xr number of groups in reserve group set r routed and no xr with Lr > 0. An xr variable for such a node pair can if possible be established by finding a shortest path in the Some of the x and x variables may have lower bounds L and s r s network of group sets with edge lengths CR1. Lr respectively, which may have been carried over from earlier time periods or otherwise specified by the user: Based on the shadow prices associated with the optimal solution x ≥ L s s of this restricted linear program new groups with associated xr ≥ Lr variables of types xr and xs and new constraints of type (15) are generated, and a new (restricted) linear program is solved. This We write: process is continued until it can be ascertained that we have an s∈g if the group set s is stand-by for groups in group set g optimal solution to the complete linear program. s∈r if the group set s is directly routed on group set r In order to describe the details of this iterative process we s∈c if the group set s is stand-by for a group set in reliability assume that we are at a stage where we just have found an opti- class c mal basis for a restricted linear program which, in addition to the initial constraints and variables, may include The constraints of the integer program are: - some additional variables of type xr and xs Σs∈g xs ≥ |g| for all group sets g (14) - the constraints of type (15) involving the additional x vari- x – Σ x ≥ 0 for all reserve group sets r and s r s∈r,s∈c s ables. reliability classes c (15)

–Σr∈g’ xr ≥ –Fg’ for all group sets g’ on which The following notation is used for the shadow prices associated reserves may be routed (16) with a basis for the restricted and the complete linear program: Constraint type: Shadow price: All variables are required to be integer. (14) πu The objective function to be minimized is (15) πcr Σr Crxr + Σs Csxs (17) (16) πg’ The optimal solution to the restricted linear program is obviously We shall assume that the cost coefficients C and C have the r s a feasible solution to the complete linear program. We shall now following structure: establish whether it is optimal or whether we can find one or R0 R1 Cr = C + C × #(group sets which r is directly routed on) more not yet generated xs variables with negative reduced cost which we can introduce into the linear program together with the C = CS0 + CS1 × #(reserve group sets which s is directly s necessary constraints and continue the solution process. routed on)

Status International research and standardization activities in telecommunication

Editor: Tom Handegård

83 84 Introduction

BY TOM HANDEGÅRD

Table 1 List of contributions to the Status section This is the fifth issue of Telek- Issue Study area Editor network with no signalling. tronikk containing the Status No. Connection establishment thus section on international re- involves manual procedures search and standardisation 4.93, Service definitions Elisabeth Tangvik and management systems. The activities. The contributions 3.94 pilot may be followed by a made so far within the differ- 4.93, Radio communications Ole D Svebak switched network, but not deci- ent telecommunication areas 3.94 sion has been made on this yet. are listed in Table 1. In addi- The pilot currently involves 18 tion, overview papers on ITU, 4.93 Transmission and switching Bernt Haram public network operators, each EURESCOM, TINA-C and 4.93, Intelligent Networks Endre Skolt responsible for one or more ACTS have been provided. 3.94 crossconnect nodes. The papers all conveys information about important 4.93, Broadband Inge Svinnset In his paper on terminal equip- organisations, projects and on- 4.94 ment and user aspects, Mr going processes often difficult 1.94, Terminal equipment and Trond Ulseth Trond Ulseth covers four to acquire for those not direct- 4.94 user aspects ETSI/UTU-T issues: ly involved. 1.94 Signal processing Gisle Bjøntegaard - Channel aggregation, which is the technique used to In this issue, status reports on 1.94, Telecommunications Ståle Wolland implement applications in the following telecommunica- 4.94 Management Network (TMN) the ISDN requiring a band- tions areas are included: 1.94 Teletraffic and dimensioning Harald Pettersen width larger than 64 kbit/s, - Broadband 1.94 Data networks Berit Svendsen - File transfer in ISDN, where - Terminal equipment and 2.94 Languages for telecommuni- Arve Meisingset ETSI has standardised two user aspects cation applications protocol profiles, - Telecommunications man- 2.94 Message Handling and Geir Thorud - Pictograms for videotele- agement network (TMN). Electronic Directory phony, where ETSI now has a standard ready, and 2.94 Security Sverre Walseth The paper on broadband is - PSTN telephones, where provided by Mr. Inge Svinnset, ETSI is working on a stan- and presents the European ATM pilot network that was put into dard on test principles, enabling, if successful, providers to operation in November 1994, and the effort undertaken by use a single set of test procedures to test all PSTN telephones EURESCOM to provide specifications for a European ATM in Europe. network. EURESCOM has provided specifications for a VP crossconnect network and is currently working on specifications The last paper, provided by Mr Ståle Wolland, gives an up to for a VC switched network. Both sets of spec- date overview of the TMN-work in ETSI, list- ifications extends current standards ing all ETSI groups involved in TMN- c n g from ITU-T. The pilot network is o o work, and highlighting important i n i m s h currently a VP crossconnect s documents. R m i c t a i d u m e r S d s w r a d n e i n s a d e i o c n r c a f d a i v a r h t n i t c T n l s it t a i e i a ic o o e t nd n n H a a s ling B rm nd ry ro o ha cto ad f age dire ba in ess nic nd M tro elec

Intelligent Languages for telecommunication networks applications nd ic a t ff n tra ing Se e n e s S c l o t u Te si i rit n c g e m y m n i e d p D i a p a l u s p t q a a r N o e M r c l n T e e a e s s n t s i u w in o m d g r r n k e a s T

85 Eurescom work on ATM Network Studies and the European ATM Pilot Network

BY INGE SVINNSET

Background The VP crossconnect is like a switch but lacks signalling. The establishment of connections will thus take more time since Availability of high capacity fibre optics and advanced switch- manual procedures and a management system will be involved. ing technology is now making broadband service provision fea- To allow for use of already installed transmission systems the sible at data rates up to 155 Mbit/s (later 565 Mbit/s). This has minimum requirement has been set to bi-directional 34 Mbit/s resulted in intense international activity for standardisation and PDH links for interconnection of the international nodes. Addi- realisation of broadband services and networks in the context of tional links may use 140 Mbit/s PDH or 155 Mbit/s SDH. The B-ISDN (Broadband Integrated Services Digital Network) and international node in Norway is connected both to Gothenburg ATM (Asynchronous Transfer Mode). A B-ISDN is a network and to Copenhagen with a 34 Mbit/s PDH link. that can support services requiring a capacity greater than the primary rate of ISDN. ATM has been chosen by ITU-T as the The official opening of the Pilot will be 24 November this year. transfer mode for B-ISDN. In ATM the user information is The Pilot users will be selected by the National Operators. segmented into cells of 53 octets. Cells belonging to one con- Among others, various RACE1 testbeds are planned to be inter- nection do not occupy pre-allocated time slots but may arrive in connected. Since the Pilot will not offer a commercial service, random order. The information identifying which connection a the uses are picked out to gain experience with various services cell belongs to (the routing field) can be found in the cell header over an ATM network, and the users should themselves be which is the first 5 octets of the cell. Of the other 48 octets some interested in gaining this experience from an application or are used for adaptation layer functions, the exact number research point of view. Feedback from the users will be very depending on the service. valuable for the network operators when later launching an ATM commercial service. The routing field is divided in two parts, the Virtual Path Identi- fier (VPI) field and the Virtual Channel Identifier (VCI) field. In the Norwegian branch of the network, Norwegian Telecom The VPI field is 8 bits at the User Network Interface (UNI) and Research is interconnected to the Norwegian international ATM 12 bits at the Network Node Interface (NNI). The VPI identifies crossconnect with a 155 Mbit/s SDH link. In addition to this the Virtual Path (VP) that the cell belongs to and has only local Alcatel Norway, Backup Centralen, the University of Oslo and significance. The VCI field is 16 bits at both the UNI and the the Norwegian Computing Center are for the time being poten- NNI. It identifies the Virtual Channel (VC) within the VP that tial Pilot users. the cell belongs to and also the VCI has only local significance. The international ATM crossconnect node in Oslo is an Alcatel The bundling of VCs into VPs enables us to organise a two A1000 crossconnect. The other participating operators have level network where VC connections can be switched separately chosen crossconnects from Alcatel, AT&T, Netcom (GDC) and in a VC switch or the whole bundle of VCs in a VP may be Siemens. For the time being the Pilot is scheduled to last until switched altogether in a VP switch (‘bundle switching’). VP June 30, 1995, with a possible extension to allow ore users to connections may be established by the Network Operator as a take part. After that the Pilot may be followed by a switched tool for managing the VC connections in the network or may be ATM network, but no decision has been taken on that matter established end-to-end between users. The last case opens up for yet. The benchmark services for the network are SMDS/DBDS, possible Virtual Private Networks inside a public ATM net- Frame Relay, ATM circuit emulation and a direct ATM VP ser- work. vice which provides ATM to the customer premises.

During the lifetime of the ATM Pilot, the operators will use the The European ATM Pilot Network network themselves for service trials and for performance measurement experiments. The performance measurements have BT, DBP Telekom, France Telecom, STET (Italy), IRITEL two aspects. Firstly, we are interested in measuring live traffic (Italy) and Telefonica (Spain) signed in November 1992 the generated by the users. This means to register, at certain mea- ‘Memorandum Of Understanding among the participating Oper- surement points, the time between arrival of ATM cells belong- ators for the installation of at ATM Pilot’. The intention behind ing to a certain connection. This will be a basis for characteris- the launching of this pilot network was to support an evaluation ing the user generated cell stream, and will increase our knowl- of ATM technology, to check proposed ATM standards and the edge about how much traffic (i.e. how many users) we can specification work of the Eurescom project P105 (see below) allow to share the capacity of for instance and SDH link. This and, not the least, start the process towards the first trans-Euro- will again be a basis for dimensioning an ATM network know- pean broadband network. Norwegian Telecom joined the pilot ing how much traffic demand we can expect. It will also be a in March 1993 together with Belgacom, PTT Telecom Neder- basis for forming the contract between a user of an application land, Telecom Finland, Telefones de Lisboa e Porto, Telia AB and the Network Operator. This contract should be formed to (Sweden) and Swiss PTT Telecom. Later, Tele Danmark, Tele- regulate the intensity of the information flowing from the user com Portugal, Telecom Eireann, Austrian PTT and ATC Fin- into the network in an optimal way. This is not only important land have also joined the Pilot, making it a total of 18 partners. for the accounting, but also to avoid that a too heavy inform- Telefones de Lisboa e Porto and Telecom Portugal have later ation flow disturbs other users’ service because of buffer over- merged to form Portugal Telecom SA, and STET and IRITEL flow in the network. have merged into Telecom Italia.

Each participating country is in the Pilot responsible for one international VP Crossconnect node, but in most countries extra 1 Research and development in Advanced Communications Crossconnect equipment from Different manufacturers have technologies in Europe – the EU research project on telecom- been provided constituting a national branch of the network. munications.

86 The other aspect of the measurement experiments is to observe Specifications for VC switched network the amount of errored and lost ATM cells and how this affects the quality of service. This will give input to the quality of ser- In September 1993 the EURESCOM P302 project ‘European vice requirements for the various applications and will also have switched VC-based ATM network studies’ was started up as a impact on dimensioning of the necessary equipment. follow-up of the P105 project. The project has 16 partners and all the major public network operators are participating. The The VP crossconnect network project leader is Pedro Chas Alonso, Telefonica. specifications The P105 project treated only Virtual Paths set up by management procedures. Since then the first recommendations on sig- EURESCOM (European Institute for Research and Strategic nalling have been released by ITU. The recommendations studies in Telecommunications) was established in 1991. It is a Q.2931 for the B-ISDN access signalling and Q.2764 for the research and development institute jointly owned by 26 public network signalling covers capability set 1 of B-ISDN (see (3)). network operators. A brief description of EURESCOM and These recommendations build on the SS7 for ISDN and are lim- some of its activities can be found in (1) and (2). The work ited to the establishment and release of point-to-point VC con- within EURESCOM on ATM started in October 1991 with the nections within an already established VP connection. P302 will project P105 on European ATM Network Studies. The main build on these recommendations and the work in ITU-T SG 11 objective was to provide specifications beyond existing stan- to extend them to multi-connection calls and to allow for re- dards for the development of an early European ATM pilot net- negotiation of connection parameters (i.e. B-ISDN capability set work. The project had 13 partners and all the major public net- 2 step 1). work operators took part. The project leader was Pierre Adam, France Telecom. This project was limited to the specifications Of the areas of study in the P302 project we can mention sig- for a VP crossconnect network. A major effort in this project nalling interfaces, network elements, service identification, net- was therefore to provide specifications for handling of establish- work aspects of a direct connectionless service, management ment, modifications and release of VP connections by use of a aspects, resource allocation and network performance. management system. While the set-up procedure in the ATM Pilot involves a telefax The flexibility of ATM allows the user to ask for a wide range procedure, manual operation and the management system, set- of bandwidth or capacities, only limited by the maximum sys- up of a switched connection should be possible end-to-end in tem bandwidth and the available resources. The bandwidth only a few seconds by means of signalling. The new capabilities given to the user for a specified connection is part of a contract also include the possibility to change bandwidth during the life- between the user and the network. In this project and in the time of a connection by means of signalling and to set up multi- ATM Pilot the bandwidth specification is limited to the maxi- connection calls between two end users (i.e. multi-media appli- mum rate (peak rate) of the connection and the cell delay varia- cations). tion (CDV) tolerance. The CDV tolerance allows for network induced variations in the rate due to buffers in network compo- Another important aspect is to take care of services with differ- nents (including the effect of a local network). The user may use ent requirements for quality of service. Data applications are the bandwidth as he wants within the limits defined by the con- loss sensitive but the requirements on transfer delay and CDV is tract with the Network Operator. Traffic contracts between the not so strict. This may be taken care of by using long buffers in Network Operator and users, procedures for deciding if a band- the switching and crossconnect nodes. Voice and video services width demand can be met (i.e. Connection Acceptance Control are delay sensitive (apart from broadcasting), CDV sensitive but – CAC) and procedures for controlling that the actual used not so loss sensitive. These services should thus see a short bandwidth is within the contract (Usage Parameter Control – buffer. In addition to this a ‘best effort’ service has been a study UPC ) have thus also been a necessary and a very important part item in international forums (e.g. ATM Forum). This service is of the specifications. a very important candidate user of the network and should be a low cost service with low quality of service guarantee and using Dimensioning methods, routing and addressing were also stud- momentarily free capacity in the network. ied within this project and key parameters for charging and accounting were identified. Since the user naturally is expecting The implementation of different services with so different qual- a certain Quality of Service, a study was performed within the ity of service demands and services for which the information project defining ATM network performance parameters, i.e. flow rate of the connections varies in time (variable bit rate requirements for cell error ration, cell loss ration, cell transfer sources) is a great challenge for the future. To fully exploit this delay, etc. Measurement procedures for the identified perfor- we will have to introduce more advanced contracts between mance parameters were studied in a separate task. user and network than the one used in the Pilot. We will have to see buffer management schemes in the switching and crosscon- Another aspects that has been studied is the interworking with nect nodes to allow for different treatment of cells belonging to other networks such as ISDN, the potential use of satellites in connections with different service needs. This means that more the network, physical interfaces and protocols over the inter- advanced CAC procedures and parameter control procedures faces. Norwegian Telecom participated in various subtasks in will have to be defined in this project to allow for a good utilisa- the project. tion of the network resources.

The purpose of the P302 project is to provide specifications for a first European switched ATM network. This network can either be an enhancement of the existing Pilot or a new Pilot

87 network. The time schedule for enhancing the existing Pilot or launching of a new network is mid 1996.

The P302 project will end by March 1995 and will then be followed by another EURESCOM project (P515) on VC switched ATM networks with enhanced capabilities beyond capability set 2 step 1 (e.g. point-to-multipoint connections) taking care of the capabilities that the next signalling releases will yield. Oslo Helsinki Concluding remarks

The first international field trials with ATM tech- Gøteborg nology are now taking place. One of these field tri- Dublin København als is the European ATM Pilot network involving 18 public network operators. Amsterdam The first initiative for this Pilot was the London ATM i research and development institute Brüssel Köln EURESCOM, jointly owned by 26 public europeisk network operators, with the project on pilotnett Paris European ATM network studies. The Pilot Wien is intended to be followed up as the ATM Zürich technology emerges, and specifications are already being made within Milano EURESCOM for a switched ATM pilot Aveiro network. Madrid The need for interconnection of LANs is only one of the driving forces for a broadband network. Video telephony, video confer- ences, multi-media calls, video-on-demand, home shopping, network access to libraries and databases are some applications that we will see in the not so far future. The different services will have different requirements for quality of service. This will be a great challenge for the future if the services are to be carried by the same network. Today, services like telephony, data and video (broadcasting) are carried in different networks. Even though these services have very different bandwidth, real time and quality demands, the flexibility of ATM will make it possible to integrate such services in the same network. The economy of this integration together with the need for providing new services, will be a driving force towards a mature broadband network needing all the functionality that ATM can give.

References

1 Fogerty, K D, Katzeff, K. EURESCOM: a new opportunity for collaborative R&D in European telecommunications. British Telecommunications Engineering, 12, 294–301, 1994.

2 Skolt, E. A presentation of EURESCOM. Telektronikk, 1 (90), 196–198, 1994.

3 ITU-T. Study Group 11. Report on the meeting held in Geneva, September 1994.

88 Terminal equipment and user aspects

BY TROND ULSETH

Most of the standardisation work on terminal equipment is The parameter negotiation of the B-channels are not compatible addressing terminals and applications for the ISDN. How- with the ITU-T/ETSI standards for audio-visual services. Some ever, there are applications that will give better performance of the ‘Bonding’ modes can be used for audio-visual communi- at higher bitrates than 64 kbit/s, e.g. audio-visual applica- cation, but the main applications are data communications. tions. To offer to the customers higher bandwidth in the nar- rowband ISDN, channel aggregators are required. This pre- ISO/IEC JTC 1/SC 6/WG 6 has prepared a Committee Draft (CD sentation presents the status of ITU-T and ETSI work on 13871). The draft is based on the principles of the TIA standard. recommendations/standards on channel aggregation. Two new ETSI standards on file transfer in the ISDN, and an ITU-T SG 15 is working on a recommendation on channel ETSI standard on pictograms for point-to-point videotele- aggregation based on the principles used for audio-visual com- phony are also presented. Finally, status for the work on munication. The number of B-channels aggregated can be ETSI standards for PSTN telephones are presented. extended to 24 in a flexible system.

ETSI STC TE4 is working on a similar ETSI standard. The Channel aggregation European Commission has given a standardisation mandate to ETSI on channel aggregation (BC-T-310). The mandate recom- The bandwidth of a B-channel in the ISDN is 64 kbit/s. There mends to align the ETSI standard with standards and recom- are a number of applications that require larger bandwidth, e.g. mendations of ISO/IEC and ITU-T, a recommendation which is transmission of moving images. In the future B-ISDN will offer fully supported by ETSI. larger bandwidth, and there is also standardisation work on a N x 64 kbit/s bearer service in the ISDN. However, to imple- The ITU-T Recommendation/ETSI Standard defines four differ- ment applications requiring a bandwidth larger than 64 kbit/s ent modes of operation: today, terminal based channel aggregation is required. - Mode B1. This mode is corresponding to mode 1 as defined in ISO/IEC CD 13871. Even though a videotelephony connection can be established using one B-channel, there are three videotelephony modes that - Mode H2. This is a mode using the framing specified in ETS require use of two B-channels. This is the simplest form of 300 144. For all terminals intended for communication with channel aggregation, aggregating two B-channels. To aggregate terminals conforming to ETS 300 144, this mode is manda- these two B-channels the principles specified in ETS 300 144, tory. The in-band signalling occupies 1.6 kbit/s on each B- the ETSI equivalent to ITU-T Recommendation H.221, are channel. used. The present version of ETS 300 144 describes codepoints - Mode B2. This mode is corresponding to mode 2 as defined and a numbering scheme for aggregating up to 6 B-channels. in ISO/IEC CD 13871. For each B-channel 1.6 kbit/s is allocated to inband signalling and synchronisation of the B-channels. - Mode B3. This mode is corresponding to mode 3 as defined in ISO/IEC CD 13871. In North America an industry consortium developed a system which is known as “Bonding”. The system can aggregate up to The channel aggregation function may be implemented in the 63 B-channels, and split one B-channel into sub-channels. The terminal equipment, a Multiple Channel Equipment (MCE), or in system is standardised by Telecommunications Industry Associ- a separate unit, a Channel Aggregation Unit (CAU). The CAU ation (TIA). Five different modes are defined: will provide the necessary channel aggregation functions for Sin- gle Channel Equipment (SCE). An example of interconnection - Mode 1. After parameter negotiation no monitoring of the sta- between a Single Channel Equipment and a Multiple Channel tus of each B-channel is provided. The delay difference Equipment is described in Figure 1, and the interconnection between each B-channel is compensated. between two Single Channel Equipment is described in Figure 2. - Transparent mode. No equalisation or parameter negotiation takes place.

These modes are mandatory for a channel aggregation unit implementing the TIA standard.

The following modes are optional: SCE CAU ISDN MCE - Mode 0. After parameter negotiation no delay equalisation is performed. Figure 1 Interconnection between a Single Channel Equipment and a Multiple - Mode 2. This mode supports data rates that are multiples of Channel Equipment 63/64 of the bearer rate. A monitoring function of each B- channel provides a continuous check for equalisation of the delay and an end-to-end bit error rate test for each B-channel. - Mode 3. This mode supports data rates that are integral multi- SCE ples of 8 kbit/s. An inband monitoring function provides a SCE CAU ISDN CAU continuous check for delay equalisation and end-to-end bit error rate test. The overhead is provided by adding bandwidth, most likely an additional B-channel. Figure 2 Interconnection between two sets of Single Channel Equipment

89 Mode B1 Mode H2 Mode B2 Mode B3 File transfer over the ISDN

ETSI has standardised two file transfer profiles for use over Aggregation None 2.5 % in the 1.5 % 64/56 kbit/s ISDN. One profile is based on a full OSI protocol stack using overhead additional channel the file transfer protocol FTAM. The FTAM profile is described Dynamic rate No Yes Yes Yes in ETS 300 388. The second file transfer profile is based on the changes Eurofile file transfer protocol. The Eurofile profile is described in ETS 300 383. In the standardisation work attempts were Exact multiples Yes Yes No Yes made to agree on a single protocol to be standardised by ETSI. of 64/56 kbit/s (only framed It became obvious at an early stage that there was equal support according to among the members of ETSI for the two alternatives. It was ETS 300 144/ITU-T therefore decided to standardised both profiles. These profiles Recommendation (layers 1 to 7) are described in Figures 5 and 6. H.221) Interworking No Yes not No The lower layer (layer 1 to layer 3) protocols are as specified in with MCE applicable ETS 300 080 for the DTE/DTE connection using ISDN circuit (audio-visual) mode.

Figure 3 Mode properties Standardisation of Pictograms for point-to-point videotelephony

The mode properties are summarised in Figure 3. Modern telecommunication terminals may offer a number of functions to the user. The use of these terminals may be complicated. It The applicability of the different modes defined in the ETSI is important to create user friendly man-machine interfaces. standard is described in Figure 4. The work of ETSI TC HF is addressing these issues. Some The ITU-T Recommendation (H.244) is scheduled for resolu- Human Factors issues are not suitable for standardisation, but tion 1 approval in February 1995. recommendations and guidelines are presented in ETSI Techni- cal reports. The ETSI standard was approved at STC level (TE4) in Septem- ber 1994 for presentation to TC TE for approval to be submitted ETS 300 375 on pictograms for point-to-point videotelephony is on PE. the result of a number of experiments carried out by ETSI STC

Mode AV/non-AV Mode applicable when ...

B1 Audio-visual the remote end is not an MCE, and the remote CAU does not support Mode H2, and Mode B3 is considered too inefficient or is not available; bitstream treated as Unrestricted Digital Information only; No inband management communication between SCE and CAU, so external means must be used to set bitrate Non- exact multiple of 64/56 kbit/s required, absence of dynamic rate change tolerable, audio-visual B3 considered too inefficient

B2 Audio-visual not applicable – B2 does not provide suitable bitrates Non- exact multiple of 64/56 kbit/s not essential, dynamic rate change desired audio-visual

B3 Audio-visual the remote end is not an MCE, and the remote CAU does not support Mode H2; dynamic rate change more important than efficiency; bitstream treated as Unrestricted Digital Information only; No inband management communication between SCE and CAU, so external means must be used to set bitrate Non- exact multiple of 64/56 kbit/s required, dynamic rate change more important than efficiency audio-visual

H2 Audio-visual remote end is an MCE, or a CAU supporting Mode H2; Inband signalling according to ETS 300 144/ITU-T Recommendation H.221, so no external control is needed Non- not applicable (unless terminal equipment is conformant to ETS 300 144/ITU-T Recommendation H.221) audio-visual

Figure 4 Applicability of channels aggregation modes

90 HF1. In total 14 organisations have contributed to the work Layer D-channel B-channel either by providing pictogram proposals or by participating in the evolution process of the proposals. The standard defines pictograms for eight point-to-point videotelephony functions, 7 FTAM ISO 8571 profiled according to ETS 300 388 1 Videotelephone/telephone modes (switching between these ACSE ISO 10607 modes) 6 ISO 8823 (ITU-T X.226) 2 Videophone camera ON/OFF 5 ISO 8327 (ITU-T X.225) 3 Videophone microphone ON/OFF 4 Videophone selfview ON/OFF 4 ISO 8073 (ITU-T X.224) 5 Videophone still picture ON/OFF 3 ETS 300 102-1 ISO 8208 (Q.931) 6 Videophone document camera ON/OFF 2 ETS 300 125 X.75 (NOTE 1) 7 Videophone handsfree ON/OFF (Q.921) or 8 Videophone loudspeaking ON/OFF. ISO 7776 1 ETS 300 011 (I.431) The results of the evaluation that form the basis for this standard or are reported in ETSI ETR 113. ETS 300 012 (I.430)

The pictograms are shown in Figure 7. Figure 5 Protocol Pillar for FTAM PSTN telephones

Telephones for attachment to the Public Switched Telephone Network (PSTN) are an important part of the telecommunica- Layer D-channel B-channel tions terminal market in Europe. This fact is recognised by the European Commission, and large resources are spent on trying 7 ETS 300 075 profiled to harmonise the PSTN attachment standards in Europe. according to ETS 300 383 ETS 300 079 The specifications for PSTN telephones in Europe are different both in terms of characteristics and in terms of test methods. 6 null (NOTE 2) ETSI STC TE4 is working on standards for PSTN telephones. A 5 null standard on test principles is now ready for Public Enquiry. If this standard is approved, a harmonised set of test principles for 4 null testing of PSTN telephones has been developed. This is an 3 ETS 300 102-1 ISO 8208 important improvement for the suppliers, one single set of tests (Q.931) can be used for testing the PSTN telephones in Europe. The next step is to prepare a standard where all requirements are 2 ETS 300 125 X.75 (NOTE 1) based on the test principles of the test standard. The require- (Q.921) or ments standard will hardly be a harmonised standard, but efforts ISO 7776 will be made to try to limit the number of national variations. 1 ETS 300 011 (I.431) or ETS 300 012 (I.430)

Figure 6 Protocol Pillar for Eurofile NOTE 1: ITU-T Recommendation X.75 as modified in ETS 300 080 NOTE 2: The main purpose of layer 6, the conversion from the “abstract syntax” to the “transfer syntax” is not necessary, because in this case the abstract data syntax in layer 7 is identical to the transfer data syntax. Also all other features of the layer 6 are not used and therefore “null” is inserted for layer 6. The abstract syntax in layer 7 and the coding in layer 6 correspond to the data syntaxes DS I, DS II and DS III in ITU-T Recommendation T.101.

91 Used for switching between videop- Used for switching on and off the Used for switching on and off the Used for switching on and off the hone (sound and picture) and telep- transmission of the camera signal transmission of the microphone document camera hone (sound only) modes signal

Used for switching on and off the selfview function Used for switching on and off the still picture (screen freeze) function Used for switching on and off the Used for switching on and handsfree mode off the loudspeaking mode

Figure 7 Pictograms for point-to-point videotelephony functions

92 Telecommunications Management Network

BY STÅLE WOLLAND

Introduction - MAS (Definition of management architecture and services) - GOM (Generic Object Model) Telecommunications Management Network (TMN) is an archi- tectural concept for communication and processing of manage- - SOC (Support Object Classes for TMN interfaces) ment information for tele- and data communications networks. - OCA (Object Model for Customer Administration on TMN The objective is to be able to implement management capability interfaces) over a distributed management network (logically and/or physically) in a standardised way so as to be system and vendor inde- - ORM (Object Model for Routing Management on TMN inter- pendent. Interoperability and reuse of management resources faces) are also important objectives. - OTM (Object Model for Traffic Management on TMN interfaces) In Telektronikk 1/94 the ISO/IEC JTC1 STC21 work within the management area was described. The following section will - INM (TMN principles for IN). give an overview over some of the ETSI work on management. A number of Sub-Technical Committees (STCs) are actively Documents relevant to management under processing or issued doing work within Telecommunications Management Network by NA4 include: (TMN). - Final Draft prI-ETS 300 291 : 1994-05 – Functional Specifi- cation of Customer Administration (CA) on the Operations There is related management work going on in Eurescom, ITU- System/Network Element (OS/NE) Interface (presented to the T, Network Management Forum, RACE, ANSI, IEEE, TINA-C NA plenary in October 1994) and other bodies. - prI-ETS 300 292 : 1994-02 – Functional specification of call routing information management on the Operation System / ETSI Network Element (OS/NE) interface (not stable yet, further Work Items on TMN work to be done) - Draft prI-ETS 300 293 : 1993-05 – Generic Managed Objects ETSI has presently no work item pertaining to the X (inter - Work Item DI/NA43314 (proposed I - ETS) – Functional domain TMN) and F (Work Station) interfaces. Only the Q specification of Traffic Management on the NE/OS interface (intra domain TMN) interface of TMN is addressed. Work on the Network Element view has received most attention so far. - Work Item DI/43316 (proposed I - ETS) – TMN – Network The first proposals for a Network Level view were proposed Level Generic Object Classes during the last year. At the last meeting of NA4 in September - Work item DTR/NA43315 (proposed TC-TR) – TMN Guide- 1994 a first proposal for Managed Objects for a Service Man- lines (second issue of DTR/NA43303) agement view was input based on RACE results (this work has also been submitted to ITU-T SG IV). - Work item DI/NA43319 (proposed I-ETS) – Technology independent path set-up Presently, there is very limited security work additional to the - Work Item DI/NA43320 (proposed I-ETS) DE/TM-2209 – STAG work within NA. Efforts to establish a Rapporteur’s Resource Management Group on security in NA4 was made at the last meeting in September. There are also initiatives to establish a security - Work Item DI/NA43321 (proposed I-ETS) – Specification of group at the STC NA level. The outcome of these initiatives the usage information on the OS/NE interface remains to be seen. - ETR 008 : 1990-12 – The method for the characterisation of Sub-Technical Committees working on TMN the man-machine interface utilised by a Telecommunications Management Network (TMN) The following is a brief introduction to some of the groups - ETR 046 : 1992-07 – Telecommunications management net- within ETSI that among other tasks also address management. works information modelling guidelines Only the management part of their objectives has been - ETR 078 : 1993-12 – TMN interface specification methodol- described. ogy [CCITT Recommendation M.3020 (1992)] NA (Network Aspects) - ETR 037 : 1992-02 – Telecommunications Management Net- work (TMN) Objectives, principles, concepts sand reference NA4 (Network Architecture, Operation & Maintenance, configurations Principles and Performance) - Work Item DTR/NA43206 (proposed ETR) – Migration to TMN STC NA4 has overall responsibility within ETSI for co-ordination and general principles for operation, maintenance and net- - Work Item DTR/NA43207 (proposed ETR?) – TMN Stan- work management in telecommunications networks and systems. dardisation overview - ETR 047 : 1992-09 – Telecommunications Management Net- A number of so-called Rapporteur`s Groups are doing the work (TMN) Management Services detailed technical work in various areas: - ETR 048 1992-09 – Telecommunications Management Net- - MNT (ISDN Customer access maintenance), VTM (Vocabu- work (TMN) Management Services: Prose Descriptions lary for TMN)

93 - ETR 088 : 1993-07 – Time / type of day dependent sche- Documents relevant to management under processing or issued duling function support object classes include: - TCRTR 008 : 1993-08 – Telecommunications Management - Work Item DE/SPS-33019 – Management of ATM switches Network (TMN) Vocabulary of terms. - Draft prETS 300 376-1 – Q3 interface at the Access Network (AN) for configuration management of V5 interfaces and NA5 (Broadband Networks) associated user ports Part 1: Q3 interface specification - Draft prETS 376-2 – Q3 interface at the Access Network STC NA5 has responsibility for co-ordination of studies of B- (AN) for configuration management of V5 interfaces and ISDN within ETSI including CPN and services. associated user ports Part 2: Managed Objects Conformance Statement (MOCS) proforma Documents relevant to management under processing or issued include: - Draft prETS 300 377-1 – Q3 interface at the local exchange (LE) for configuration management of V5 interfaces and - Draft prETS 300 404 : 1994-03 – B-ISDN Operation and associated customer profiles Maintenance (OAM) principles and functions - Draft prETS 300 377-2 – Q3 interface at the Local Exchange - ETR 089 : 1993-08 – Principles and requirements for sig- (LE) for configuration management of V5 interfaces and nalling and management information transfer associated customer profiles Part 2: Managed Object Confor- - DE/NA-52210 B-ISDN – Management Architecture and mance Statement (MOCS) proforma Management Information Model for the ATM cross connect - Draft prETS 300 378-1 – Q3 interface at the Access Network - ETS 300 273 : 1994-03 – Metropolitan Area Network (MAN) (AN) for fault and performance management of V5 interfaces Medium Access Control (MAC) layer management. and associated user ports Part 1: Q3 interface specification - Draft prETS 300 379-1 – Q3 interface at the Local Exchange NA6 (Intelligent Networks) (LE) for fault and performance management of V5 interfaces and associated customer profiles Part 1: Q3 interface specifi- STC NA6 has responsibility for defining the TMN principles cation. for IN. TM (Transmission and Multiplexing) Documents relevant to management under processing or issued include: TM2 (Transmission Networks Management, - ETR 062 : 1993-05 – Baseline document on the integration of Performance and Protection) Intelligent Network (IN) and Telecommunications Manage- ment Network (TMN) STC TM2 has among other objectives responsibility for defining general principles for operation and maintenance of trans- - TC-TR/NA60801 – IN Intra Domain Management Require- mission networks. ments for CS-2, DTR/NA6101 – Service and service feature interaction: service creation aspects, service management Documents relevant to management under processing or issued aspects and service execution aspects include: - TC-TR/NA6- 61201 – Security Requirements for Global IN - DTR/TM-2207 SDH Network Level Information Model – Systems (approved by STC NA6 in September 1994 – now Configuration Management Ensemble for approval by TC NA). - prETS DE/TM-2209 : 1994 Operations and Maintenance of NA7 (Universal Personal Telecommunications) Optical Access Networks ( for STC approval in September 1994) STC NA7 has among other objectives responsibility for identi- - ETS 300 150 : 1992-11 – Protocol suites for Q interfaces for fying the management requirements for UPT. management of transmission systems - RE/TM-2202 – Revision of ETS 300 150 : 1992-11 Documents relevant to management under processing or issued include: - ETS 300 304 : 1993-04 – Synchronous Digital Hierarchy (SDH) Information model for the Network Element (NE) - Draft ETR NA-70306 – UPT Management Aspects, Draft view ETR NA-05001 – UPT Management Services and Manage- ment Information Model. - RE/TM-2213 – Maintenance and enhancement of ETS 300 304 : 1993-04 SPS (Signalling, Protocols and Switching) - ETS 300 411 : 1994-03 – Performance monitoring information model for the network element view SPS3 (Digital Switching) - ETS 300 412 : 1994-03 – Payload configuration information model for the network element view STC SPS3 has among other objectives responsibility for defining V5 interface management. - ETS 300 413 : 1994-05 – Multiplex section protection information model for the network element view

94 - ETS 300 371 : 1993-08 – Plesiochronous Digital Hierarchy SMG (Special Mobile Group) (PDH) information model for the Network Element (NE) view SMG6 (Operation & Maintenance) - RE/TM-2223 – Maintenance and enhancement of ETS 300 STC SMG6 has among other objectives responsibility for speci- 371 : 1993-08 fying the network management functions of the GSM, DCS - DTR/TM-2223 – Software downloading process for use on Q 1800 and UMTS systems. interfaces Documents relevant to management under processing or issued - DTR/TM-2221 – The application of ODP to the management include: of a transmission network level model - D-ETR/SMG50501 Objectives and Framework for the TMN - DE/TM-2210 – Management of SDH transmission equipment of the UMTS (to be finalised by the end of 1994). - DE/TM-2220 – SDH information model connection supervi- sion function (HCS/LCS) for the Network Element view SES (Satellite Earth Stations and Systems) - DE/TM-2218 – SDH Radio relay equipment information model for use on Q interfaces SES is responsible for all aspects relative to satellite communications including management. - DE/TM-2216 – Object oriented information model for the protection of SDH NEs Documents relevant to management under processing or issued - DE/TM-2207 – SDH network information model – configura- include: tion management ensemble - ETS 300 160 : 1992-11 – Control and monitoring functions at - DTR/TM-2222 – Management of Access Networks a Very Small Aperture Terminal (VSAT) - DE/T M-2224 – Management of fixed radio Access Networks - ETS 300 161 : 1992-11 – Centralised control and monitoring functions for VSAT networks. - DE/TM-2??? – Management of Access Networks providing interactive and / or distributive services ECMA TC32 (Communications, Networks and - DE/TM-2209 – Operation and maintenance of Optical Access System Interconnection) Networks (OANs). ECMA TC32 TG12 (Private Telecommunications BTC (Business Telecommunications) Networks – Management)

BTC1 (Private Networking Aspects of 64 kbit/s ISDN TG12 is responsible for developing technical reports and stan- Business Communications) dards for the management of Private Telecommunications Net- works (PTNs) by extending and adapting general TMN to STC1 has among other objectives responsibility for defining PTNs. general principles and requirements on Corporate Network Management functions including control by the private operator of such networks.

BTC4 (Private Network Aspects of High Speed Busi- ness Communications)

STC4 has among other objectives responsibility for integration and co-ordination of high-speed services (LAN-interconnection etc.) by identifying the requirements for signalling, security and management services.