Common Mode Inductors for EMI Filters Require Careful Attention to Core Material Selection

POWER

Robert West, Magnetics, Division of Spang & Co., Butler, Pennsylvania

The common mode inductor is an integral part of most EMI filters; its very high impedance over a wide frequency range suppresses high frequency power supply spikes.

Switching power supplies gener- X and Y capacitors. The Y capaci- impedance over the switching fre- ate two types of noise: common mode tors and the common mode induc- quency range. and differential mode. Differential tors contribute to the attenuation of Common mode inductors are mode noise (Figure 1a) follows the the common mode noise. The induc- wound with two windings of equal same path as the input power. Com- tors become high impedances to the numbers of turns. The windings are mon mode noise (Figure 1b) is rep- high frequency noise and either re- placed on the core so that the line resented by spikes that are equal to flect or absorb the noise while the currents in each winding create fluxes and in phase with each other and capacitors become low impedance that are equal in magnitude but op- have a circuit path through ground. paths to ground and redirect the noise posite in phase. These two fluxes To suppress EMI, a typical filter away from the main line (Figure 2). cancel each other, leaving the core in will include common mode induc- To be effective, the common mode an unbiased state. The differential tors, differential mode inductors and inductor must provide the proper mode inductor has only one wind-

Line SMPS Line SMPS

(a) Differential Mode (b) Common Mode

Figure 1. Noise Types. ing, requiring the core to support the High Z entire line current without saturat- ing. Herein lies the great difference between common mode and differential mode inductors. To prevent Low Z saturation, the differential mode in- Line Noise Source ductor must be made with a core that has a low effective permeability (gapped ferrites or powder cores). The common mode inductor, however, can use a high permeability material and obtain a very high in- Figure 2. Common Mode Filter. ductance on a relatively small core.

Material Selection Total impedance of the common able impedance (Zs) over the entire Noise generated by switching mode inductor is composed of two frequency spectrum. power supplies is primarily at the parts, the series inductive reactance For the most part, ferrites are the unit’s fundamental frequency, plus (Xs) and series resistance (Rs ). At material of choice for common mode higher harmonics. This means that low frequencies, the reactance is the inductors and they are divided into the noise spectrum usually runs any- primary contributor to impedance, two groups: nickel zinc and manga- where from 10kHz to 50MHz. To but as the frequency increases, the nese zinc. Nickel zinc materials are provide proper attenuation, the im- real part of the permeability drops characterized by low initial pedance of the inductor must be suf- and losses within the core rise, as permeabilities (<1000µ), but they ficiently high over this frequency seen in Figure 3. These two factors maintain their permeabilities at very range. combine to help produce an accept- high frequencies (>100MHz). Man-

One company - MAGNETICS - offer the broadest range. Selecting the ideal Ferrite Cores: High frequency materials (10kHz - 2Mhz) for power transformers magnetic core and inductors; High permeability can be a problem. materials (to 15,000µ) for EMI filters, ISDN transformers and broadband transformers; and Temperature stable materials for telecom applications. Powder Cores: (Molypermalloy, High Flux and Kool Mµ®): for in-line filters, output chokes and flyback transformers. Strip Wound Cores: (Tape cores, cut c- cores, bobbin cores, and laminations) for high power transformers, audio transformers, magnetic amplifiers, ground fault interrupters and current transformers. For more information, technical brochures or application engineering assistance, contact: MAGNETICS has the solution! Division of Spang & Company ganese zinc materials, on the other IMPEDANCE VS. FREQUENCY hand, can attain permeabilities in excess of 15,000 but may start to “roll-off” at frequencies as low as 20kHz. Because of their low initial permeabilities, nickel zinc materials will not produce a high impedance at low frequencies. They are most often used when the majority of unwanted noise is greater than 10 or 20MHz. Manganese zinc materials, however, offer very high permeabilties at low frequencies and are very well suited to EMI suppres- FREQUENCY MHz sion in the 10kHz through 50MHz Figure 3. Impedance vs Frequency. range. For these reasons, the remainder of this article will focus on the high permeability, manganese zinc ferrites. High permeability ferrites come economically. Toroids require spe- filters with a 50Ω Line Impedance in many different shapes: toroids, E cial winding machines or must be Stabilization Network (LISN). This cores, pot cores, RMs, EPs, etc.; but wound by hand, making the per- has become a standard method for for the most part, common mode piece winding cost higher. Fortu- measuring filter performance but it filters are wound on toroids. nately, the number of turns on com- can lead to results that are quite There are two main reasons for mon mode inductors is usually quite different from those in real life. using toroids. First, toroids are gen- low, so the winding costs do not A true first order filter will pro- erally less expensive than the other become too prohibitive. vide an attenuation that increases by shapes because they are one piece, For these reasons, toroids are the -6 dB per octave beyond the corner whereas other shapes require two geometry of choice in common mode frequency. This corner frequency is halves. When cores come in two inductors and the remainder of this usually low enough so that the in- halves, they must be flat ground on article will focus on their use. ductive reactance is the primary con- their mating surfaces to make them tributor to the impedance, thereby smooth and to minimize the air gap Design Considerations allowing the inductance to be calcu- between them. Furthermore, high The basic parameters needed for lated: Ls=Xs/2πf. permeability cores often require an common mode inductor design are Once the inductance is known, additional lapping procedure to make input current, impedance, and fre- the remainder of the design involves them even smoother (this produces a quency. Input current determines the core and material selections along mirror-like finish). Toroids require size of the conductor needed for the with the calculations for number of none of these extra manufacturing windings. Four hundred amps per turns. steps. square centimeter is a common de- Quite often, the first step in the Second, toroids have the highest sign value for calculating wire size, design is to select a core size. If the effective permeability of any core but may be altered depending upon design has physical requirements, shape. The two-piece construction the acceptable temperature rise of then the largest core should be se- of the other shapes introduces an air the inductor. Single stranded wire is lected that will still stay within these gap between the halves, which low- almost always used because it is the requirements after it is wound. If ers the effective permeability of the least expensive and it helps contrib- there are no size restrictions, then a set (typically by about 30%). Lap- ute to the noise attenuation through core size can be selected at random. ping improves this but does not elimi- high frequency skin effect losses. The next step is to calculate the nate it. Because toroids are made as The impedance of the inductor is maximum number of turns that will one piece, they do not have an air normally specified as a minimum fit on the core. Common mode in- gap and do not suffer a reduction in value at a given frequency. This ductors require two windings, nor- effective permeability. impedance, in series with the line mally single layer, with each wind- Toroids do have one disadvan- impedance, will provide a desired ing on opposite sides of the core to tage – their high winding cost. Bob- noise attenuation. Unfortunately, the provide isolation. Double layer and bins, which are available for the other impedance of the line is rarely bank windings are occasionally used, shapes, can be wound quickly and known, so designers often test their but they increase the distributed ca- pacitance of the windings and this Table 1 lists typical design infor- Inside diameter = 13.72mm decreases the high frequency perfor- mation and an example calculation ±0.38mm = 13.34mm minimum mance of the inductor. using the AL value. Because the wire size has already 4. Calculate inner circumference been determined by the line current, For this example: (I.C.) and maximum number of an inner circumference can be calcu- turns possible: lated based upon the inner radius of J material (5000µ) is chosen with I.C. = π (core diameter - wire the core minus the radius of the wire. an A value of 3020 L diameter) The maximum number of turns can then be calculated by dividing the N is given as 20 turns, so: I.C. = π (13.34mm - 1 mm) wire diameter, with insulation, into I.C. = 38.76mm that portion of the circumference L = 1.208mH occupied by each winding. Note: Maximum turns = (160°/360°) * Allowing for isolation between the If this minimum inductance is too (38.76mm)/(1mm/turn) windings, each winding will typi- low for the design, then a higher cally occupy 150° to 170° of the permeability material, or a larger I.C. = 17.2 turns, or 17 turns inner circumference. core, can be selected. However, if 5. Calculate minimum inductance Once the maximum number of the calculated inductance is well for 17 turns: turns have been calculated, the next above the design limit, then a smaller steps are to choose a material and core with fewer turns could possibly AL = 3020 ±20% be substituted. 1/2 determine the inductance. Material N = 1000 (L/AL) choice involves many factors; oper- 1/2 17 = 1000 (L/3020 - 20%) ating temperature, frequency range DESIGN EXAMPLE: and cost, to name a few. However, L = 0.698 mH minimum the first issue is to verify the core An impedance of 1000Ω is needed size that was selected; the other fac- at 10kHz. RMS input line current is The resulting value is consider- tors can be resolved later. To do this, 3A. ably lower than the 1.59 mH needed, so a modification must be made. The a moderate permeability material 1. Choose wire size: should be chosen and the inductance options available for change are core calculated. 3A at 400 A/cm² yields a wire size, material permeability and wire Most ferrite manufacturers list area of 0.0075cm² size. A larger core will provide a bigger inside diameter so more turns inductance factor (AL ) values for #19 AWG is chosen with a wire their cores, which provides an easy can be wound on the core (larger area of 0.007907cm² (1 mm cores may also have a higher A method for calculating the induc- diameter), including insulation L tance. The relationship between the value). A higher permeability material will, naturally, raise the induc- number of turns and inductance is: 2. Calculate minimum inductance: tance and a smaller wire size will N=1000 (L/AL) ½ L minimum = 100Ω /2π(10,000 allow more turns to fit on the core where: Hz)=1.59mH (but this will also increase the copper losses). N = Number of turns 3. Choose a core size and material Continuing with the previous ex- L = Inductance (mH) from the table: ample, if it is decided to keep the 42206-TC size, then new turns cal- J-42206-TC is chosen AL = Inductance factor in mH/ culations must be made for each 1000 turns A L= 3020±20% material.

Table 1 lists typical design information and an example calculation using the AL value.

Core Type O.D. (mm) I.D. (mm) Height (mm) A L (mH/1000 turns) le (cm) Ae (cm²) Ve (cm³) J 20% W 30% H 30% (5000µ) (10,000µ) (15,000µ) 42206-TC 22.1 13.72 6.35 3020 6040 9060 5.42 0.250 1.36 J material (5000µ):N = 1000 frequency increases. The higher the (1.59mH/3020-20%) ½ =25.6 turns permeability, the lower the frequency where this roll-off occurs. Fortu- W material (10,000µ):N = 1000 nately, these materials also become (1.59mH/6040-30%)½ = 19.4 turns very lossy at high frequencies, and these resistive losses keep the total H material (15,000µ):N = 1000 impedance of the inductor high, out (1.59mH/9060-30%) ½ = 15.8 turns beyond 100MHz. Figures 6 and 7 show how the series inductive reac- If the J material is used, then a tance (Xs) and series resistance (Rs) smaller wire size is definitely needed, change over frequency for the three whereas the original wire size should high permeability materials (J, W R = Inner Radius of Core fit nicely on the H material. The and H). Figure 8 displays the total turns required for the W material are r = Inner Radius of Core Minus impedance versus frequency for each only slightly greater than the maxi- Wire Radius material. The measurements were mum calculated previously in step 4 Inner Circumference = 2 π r made with 10 turns on 42206-TC (17 turns). Sample windings should Figure 4. Toroid Inner Circumference. size cores. be tried on this core to determine if a The graphs indicate that H mate- smaller wire size is necessary. rial has a distinct advantage over W The steps mentioned above for beyond the corner frequency must and J at low frequencies. However, core selection can be quite time- be understood. between 100kHz and 200kHz its consuming. To speed the selection Manganese zinc ferrites exhibit permeability has dropped low process, a “Core Selector Chart” is high permeabilities at low frequen- enough so that the total impedance given in Figure 5. To use it, simply cies (<500kHz), but roll-off as the has fallen below the W material im- multiply the RMS line current by the required inductance (in mH), and locate this point on the abscissa. Move up the chart until the appropri- (with current density of 400 amps/cm²) ate diagonal material line is crossed, then continue upwards until the very next horizontal “size” line is reached. This line corresponds to a certain core size located on the ordinate of the graph. J, W and H materials are included on the chart. Naturally, H material yields the smallest core sizes. This graph assumes a current density of 400A/cm² and single layer windings on the core. Using different current densities will require some guesswork (the Wµ line can be used for Jµ at 200A/cm²). This chart is only meant as an aid to core selection; the final design may be slightly larger or smaller.

Frequency Characteristics The design method just described provides core size and material, but it leaves out many other details that must also be covered. For instance, common mode filters operate over very wide frequency ranges (gov- LI (mH - Amps) ernment regulations on EMI extend to 30MHz), so material performance Figure 5. Common Mode filter Core Selector Chart. pedance. The W material then has rial when the frequency spectrum of choke designer are the effects versus the highest impedance until 2MHz, the noise is known. permeability and flux density. Most where the J material takes over. Temperature has an effect on most materials increase in permeability as Curves such as these can help the ferrite material properties. Of pri- the temperature increases. Figure 9 designer to select the proper mate- mary interest to the common mode shows the curve for W material. Like- wise, a decrease in permeability should be expected when the temperature goes below 25°C. Worst case permeability fluctuations must be taken into account when design- ing for the minimum filter inductance. Temperature also affects saturation flux density. Figure 10 shows W material’s typical decrease in flux density with increasing temperature. This reduction in usable flux density can increase the likelihood of core saturation. In addition, all magnetic materi- FREQUENCY MHz als have a Curie temperature, the point where magnetic activity stops. Figure 6. Series Reactance vs Frequency. High permeability ferrites usually have Curie temperatures between 120°C and 175°C. It is important to know where this Curie point is and to maintain the core operating temperature below this limit. Ferrites are not damaged if the Curie temperature is exceeded (i.e. during wave soldering), but they will become non- magnetic if the Curie temperature is reached during operation. Finally, ferrite toroids are fre- quently offered with a dielectric coating (i.e. parylene, epoxy, etc.) to help insulate the core from the windings. These coatings have their own FREQUENCY MHz temperature ratings and can be damaged by the combination of heat and Figure 7. Series Resistance vs Frequency. strong cleaning agents used during the assembly process. Manufactur- ers’ data books should always be consulted for the appropriate information on core coatings, as well as the other material properties mentioned earlier. Ferrite materials are susceptible to mechanical stress, both compres- sive and tensile. High permeability materials are particularly affected and can exhibit large, negative changes in permeability under moderate stresses. There are two major causes of core stress: encapsulants FREQUENCY MHz and windings. An encapsulant causes stress if it Figure 8. Total Impedance vs Frequency. has a thermal coefficient of expansion that is different from the ferrite. Encapsulants should be chosen with expansion coefficients as close to the ferrite as possible, but even small differences can cause problems. Therefore, one possible remedy is to cushion the core with a “rubbery” material, like RTV, before it is pot- ted. This coating can help to distrib- ute some of the stress caused by the encapsulant during temperature fluctuations. Winding stress occurs when the wire is wound onto the core. Com- mon mode inductors are usually Temperature (Centigrade) wound with rather heavy conductors and these wires must be pulled tight Figure 9. Permeability vs Temperature. in order to fit properly around the core. The stress induced can be quite severe. Temperature cycling normally relieves most winding stresses. The cycle should range from -55ºC to + 150°C, with both extremes main- tained for 30 minutes to 1 hour. The rate of change should only be a few degrees per minute to prevent a thermal shock (cracking) to the ferrite. During the process, the copper wire will expand and contract, thereby relaxing the force exerted on the core. To show the effects of stress on ferrites, inductance measurements were made on four toroids of differ- Temperature (Centigrade) ent permeabilities (3000 µ, 5000µ, 10,000µ 15,000µ) as tensile forces Figure 10. Flux Density vs Temperature. were exerted upon them. The results are given in Figure 11, as percent of initial permeability versus force per core cross-sectional area. As ex- and the leakage inductance of the that are three to four times greater pected, the 15,000 permeability ma- winding. Because the leakage flux than the RMS current can easily satu- terial had the greatest reduction in leaves the core and is not canceled, it rate the core, again allowing com- permeability while the 3,000 perme- is possible for it to saturate the core mon mode noise to pass (Figure 12). ability material had the smallest. material under high line currents, or Also, as mentioned earlier, high op- at least shift the point of operation erating temperatures intensify these Core Saturation away from the origin of the BH loop problems by lowering the saturation Popular opinion states that com- to a point of lower incremental per- flux density. mon mode inductors cannot be satu- meability (µ∆). This lowering of the To show the effects of core satu- rated; the differential mode flux permeability results in a proportional ration, three cores (one core of each within the core cancels and the com- decrease in series inductive reac- material, J,W and H) were wound mon mode flux is so low that it is not tance (XS) and could allow unwanted like a common mode filter. Each a concern. Unfortunately, this is not noise to pass through the filter. core had two windings of 15 turns of entirely true. It has been shown by The problems of core saturation #18 AWG. A third winding of 10 others [4,5] that some amount of dif- are exacerbated in switching power turns was put on the core and con- ferential flux exits the core from supplies that do not provide power nected to an inductance analyzer. each winding. This leakage flux is factor correction. High capacitor The common mode windings were proportional to both the line current charging currents with crest factors then connected in series so that any current through them would create stantaneous line current that passes 24%, respectively), while the H opposing fluxes. These windings through common mode inductors. material dropped a little further (15% were then connected to a DC power As can be seen from Figure 13, all and 35%). supply. Inductance was measured the cores dropped in inductance as These results show that the com- on the 10 turn winding as the DC the current increased. The J and W mon mode inductance is affected by current was increased from 0 to 15A. materials were reduced by the same the leakage inductance and that core The DC current simulated the in- percentage at 10 and 15A (10% and saturation is possible under peak line currents. If it is found that an inductor is experiencing partial core saturation, then a switch to a higher permeability material may be needed. A higher permeability will offset the effects of core saturation by provid- ing a greater starting inductance, as shown in Figure 13, or will reduce the level of saturation by allowing a reduction in the number of turns, thereby lowering the leakage inductance. Different core sizes and winding techniques are other methods that can reduce leakage inductance.

TENSILE STRENGTH References Newtons per square centimeter 1. MAGNETICS Technical Bul- Figure 11. Inductance vs Tensile Strength. letin, FC-S2, 1994. 2. MAGNETICS Ferrite Core Catalog, FC-601, 1994. 3. Srebranig, Steven and Leonard Crane, Guide for Common Mode Filter Design, Coilcraft Publi- cation, 1985. 4. Nave, Mark, On Modeling the Common Mode Inductor, Pub- lisher and Date Unknown. 5. Nave, Mark, A Novel Differen- tial Mode Rejection Network for Conducted Emissions Diagnos- Figure 12. Core Saturation B-H Curve Due to High Crest Factor. tics, IEEE, 1989. 6. Crane, Leonard and Steven Srebranig, Common Mode Fil- ter Inductor Analysis, Coilcraft Publication, 1985. 7 . Knurek, D.F., Reducing EMI in Switch Mode Power Supplies, IEEE, 1988.

FC-S5 10D