Comparing Microstrip and CPW Performance

Technieal Feature

By buildingabetter electromagneti.c (EM) si,mulation madel, whl,ch includes the fficts of a PCB's m.etal surface roughness, microstrip and coplanar u-;aaeguide, ci,rcuits can be closely compared to find the best fit for different applications.

f atchins a microware transmission- will be compared. Further analysis will be per- I tin" te?hnology to an application re- formed with the aid of electromagnetic (EM) l- quires carelul consideration of more models and EM simulation softwaie. The soft- than a few factors. Depending on the require- ware modeling will help validate the measured ments of an application, high-frequency circuit results and also show how effective software designers may be concerned with loss budgets, modeling can alleviate concer-rrs, when using propagation mode issues, radiation losses and new transmission-line approaches an&or cir- electromagnetic interference (EMI), and even cuit topologies. the printed-circuit-board (PCB) assembly lo- Microstrip and CPW formats are often se- gistics and the relative difficulty of adding com- lected over other high-frequency transmission- ponents to a PCB. Microstrip has been one of line options, such as stripline, due to their the most popular microwave transmission-line simolicitv.simplicity. StriolineStriplineStri can deliver excellent hish-high- formats for decades and is well characterized. [requency performance, with good noise im- Coplanar waveguide (CPW) transmission lines munity and isolation between adjacent circuit have also been used extensively in microwave traces. But it is also more dlfficult and expen- PCB applications, although they are not as sive to fabricate than microstrip or CPW Strip- well understood as microstrip lines. Typi- line is essentially a flat metal transmission line cally, conductor-backed coplanar waveguide between two ground planes, with the ground (CBCPW) circuits are often used in conjunc- planes separated by a dielectric substrate mation with microstrip in microwave circuit terial. The width of the transmission line, the designs. A common approach is the use of thickness of the substrate, and the relative di- CBCPW in the circuit's signal launch area, electric constant of the substrate material de- transitioning to microstrip for the remainder of termine the characteristic impedance of the the circuit to enable simple component place- transmission line. Difficulties with stripline ment and PCB assembly. To help designers understand differences between microstrip and CPW transmission- ]ouN CooNnon Iine approaches, measurement data from dif- Rogers Corp., Chandler, AZ ferent test circuits fabricated with the same, BmeN Reurro well-known commercial substrate material Sonnet Software lnc., North Syracuse, NY

MrcRowAVE JOURNAI I JULY 2012 k*hnFc*i Fsatur*

hybrid transverse- circuit at the same frequencies. A magnetic modes are review of the practical tradeoffs of also possible with via placement for CBCPW circuits is microstrip, but these available in the literature.r Figure 2 modes are some- offers an overview of signal loss (S21) times the result of performance for microstrip, coplanar- undesired spurious launched microstrip, and CBCPW wave propagation. circuits fabricated on 3O-mil-thlck L nlg. I Cross-sectional aieu of a microstrip line (a) and. three- In general, CBCPW R04350BrM circuit-board material dimensional oieu; o.f a CBCPW line (b). circuits offer propa- from Rogers Co1p. gation behavior sim- GCPWG refers to a grounded co- include ground planes that must be ilartothatof microstripcircuits. planar waveguide and is actually the shorted together, requiring electrical For both microstrip and CBCPW same configuration as CBCPW. The via connections between the tr,vo met- circuits, spurious parasitic wave prop- top ground microstrip configuration al groundplanes andthe lackof direct agation can be a problem. As a gen- is essentially a coplanar-launched mi- access to the signal layer for compo- eral mle, the circuit geometry (that is crostrip circuit - a microstrip circuit nent mounting. Stripline's second its cross-sectional features) for either with a CBCPW configuration in the ground plane also results in narrower transmission-line approach should connector signal launch area. The transmissionline widths, for a given be less than 45' long at the highest cuwe-fit data for microstrip and co- substrate thickrress and characteristics operating frequency of interest- For planarJaunched microstrip are taken impedance, than for microstrip. microstrip, the circuit parameters of from the literature.l The traces reveal In contrast, microstrip and CPW concern include the thickness of the some interesting traits to consider for circuits feature an exposed signal substrate (that is the distance between the different transmission lines. For layer, greatly simplifying compo- the signal and ground planes) and the example. CBCPW tlpically suffers nent assembly on the PCB, Figure I width of the signal conductor (trans- higher loss than microstrip or copla- shows simple drawings of microstrip mission line width). For CBCPW, nar-launched microstrip. The GSG and CBCPW transmission lines. The attention must be paid to those two configuration of the CBCPW copla- microstrip circuit has a signal con- parameters, as well as to the distance nar layer exhlbits hlgher conduitor ductor on the top of the dielectric between the GSG spacing on the co- loss than microstrip-based circuits. substrate and a ground plane on the planar layer. SUll, the loss for CBCPW follows a bottom. In a CBCPW circuit, a copla- For proper grounding, CBCPW constant slope, while the loss curves nar layer with ground-signal-ground circuits employ vias to Connect the for microstrip and coplanar-launched (GSG) configuration replaces the sig- topJayer coplanar ground planes and microstrip undergo slope transitions nal layer of microstrip. The CBCPW the bottom-layer ground plane. The at approximately 27 and 30 GHz, circuit's top ground planes are tied to placement of these vias can be critical respectively. These loss transitions the bottom ground plane by means of for achieving the desired impedance are associated with radiation losses. vias. CBCPW is sometimes known as and loss characteristics, as well as for With proper spacing and via spacing, grounded coplanar waveguide. suppressing parasitic wave modes. CBCPW can be fabricated with mini- In terms of wave propagation, When grounding vias are effectively mal radiation loss. microstrip transmissionline circuits positioned in a CBCPW circuit, a In wideband applications, disper- generally operate in a quasi trans- much thicker dielectric substrate can sion can be important. Microstrip verse-electromagnetic (TEM) mode. be used at higher frequencies than transmission lines are dispersive by Hybrid transverse-electric (TE) and would be possible for a microstrip nature: the phase velocity for EM waves is different in the air above the signal conductor than through the dielectric material of the substrate. CBCPW circuits can achieve much o less &spersion when there is tight coupling at the GSG interfaces on -t.oo the coplanar layer, since more of the -2.Cb. E-field occurs in air to reduce the ef-

a! -r.oo" fective inhomogeneity of wave travel ::4.OO through different media. ? *r.6O Using proper design techniques, .6,0O CBCPW circuits can achieve a much wider range of impedances mi- -ZOO than crostrip circuits. applica- .-8-00, In ad&tion, tions where crosstalk may be a concern, circuit performance can benefit ,r ,.'' r i , , . .r ,r , "' ., : ..'r.,1;', r.i;:tiilt:.:.ti=::.t' :::,ii-,ffF1l.:.ii!:li+:l#;:i+ii?,: from the coplanar ground plane sepa- rtg.flg. zZ Comparison.romparxson oJ resttest 2y2' CCPWG,2/2" anrl,2%" ration of CBCPW's neighboring signal ^L of ilatafor top grounil microstrip straight microstrip test boards conductors. Due to their significantly

76 MrcRowA\1E JOURNAL I JULY 2012 T**hnfcml ilem?*r*

fective dielectric constant increases as There is also a real-life issue af- ED = 3.0 pm RMS roughness ED = 1.5 pm Rlttl5 the surface of the copper fecting most PCB circuits and espe- RT = 0.7 pm RMS increases, as indicated by copper sur- cially CBCPW, which can cause more RT = 0.5 pm RMS faces with higher root-mean-square variation 49' in circuit performance due (RMS) roughness values. to standard fabrication effects. This 5tU. z. !l* In addition to obserued dielectric is the conductor trapezoidal effect, 66 constant effects, surface rough- "edge uZ the or profile," where the PCB >o ness of a microstrip is known to impact conductors are ideally rectangular in buql insertion loss performance.3-7 The cross*sectional r a view but the actual r :..{O:r::rr5p gl lO 2O 3O.: i.6o. i topology of the circuit may be more circuits are trapezoidal in shape. This FnEauENCY icHt'i or less prone to such copper surface can cause the current densiW in the roughness effects, simply due to cur- coplanar GSG area to vary; 'an ldeai A mg. S Effecti,ae dielectric constant of a I rent and distribution within mil LCP lami,nate uith a 50 dl microstrip line E-field rectangular conductor structure will with different surface roughness. the circuit. For example, the copper have more current density up the surface roughness has less effect on a sidewalls of the adjacent conductors reduced radiation losses, dispersion tightly coupled CBCPW transmission in this region, whereas the trapezoi- and parasitic wave mode propaga- line than on a microstrip. In a CB- dal structure will have more current tion, CBCPW circuits are often used CPW circuit, the current and E-field density at the base (copper-substrate at much higher frequencies than mi- are tightly maintained within the GSG interface). When there is more cur- crostrip circuits. At millimeter-wave on the coplanar layer. For a microstrip rent density at the base due to the frequencies, for example, it is often circuit, the field and current move trapezoidal effect, the copper surface that a simple wire-bonded air bridge more toward the bottom of the metal, roughness will have more influence on will be used to connect the ground where the roughness lies. losses and the propagation constant. planes on both sides of the CBCPW The trapezoidal concerns for CBCPW signal conductor, The air bridge ap- ,VIEASURING DIFFSRENCES PCBs are shown inFigure 4. proach serves as a "trap" for specific All of the circuits evaluated in this Figure 5 compares the effective frequencies of concern when spurious article were fabricated on a 254 pm dielectric constants for two different wave mode propagation is an issue.2 (10-mil) thick RT/duroid@ 5880 lanii- coplanar circuit tlpes and how they nate from Rogers Corp. The same di- are affected by two extreme cases of COPPER sI.IRFACE ROUGI{NEsS electric material was used in all cases, copper surface roughness. The phase The copper surface roughness of although with different copper gpes: response measurements that were PCB substrates has been known to rolled copper with surface roughness made for one data set of circuits em- affect conductor losses as well as the of 0.4 pm RMS, electrodeposited ployed a differential phase length propagation constant of the transmis- (trD) copper with surface roughness method.S Circuits were made in very sion line.s The effect on a transmis- of 1.8 pm RMS, and high-profile ED close proximity on the ,"*" pro""rJ- sion-linet propagation constant causes copper with surface roughness of 2.8 ing panel and the only difference for a circuit to have a different "apparent pm RMS. Table f provides details on the tr,vo circuits being measured was dielectric constant" than expected. Of the dimensions of the different cir- the physical length of the transmission course, the material parameter is un- cuits, along with their measured char- lines. changed by the roughness of the ma- acteristic impedances. The nominal The figure shows that the differ- terial's metal layer. Rather, the amount circuit dimensions noted in the table ence at 10 GHz for the microstrip of metal surface roughness causes the are per the circuit design; however, (cpw micro), for smooth vs. rough observed effects by influencing elec- typical PCB fabrication tolerances ap- copper, RMS = 0.4 vs. RMS = 2.8, tric field and current flow. As Fig- ply. On the actual circuits, the signal- respectively, is approximately 0.09 in ure 3 shows,3 the effective dielectric to-ground spacing for the coplanar terms of the effective dielectric con- constant can vary widely for the same layer of the CBCPW and the copper stant. The same consideration for the dielectric substrate when the copper thickrress had appreciable circuit-to- tlghtly coupled CBCPW is approxi- surface roughness is different. The ef- circuit variation. mately 0.06. Even though trapezoideLl

78 MICROWAVE JOURNAT I JULY 2012 T**hr"ri*ml Featu*"*

cy for microstrip have more influence on insertion loss than for CBCPW. performance for CBCPW than for mi- Trapezoidal effects crostrip. are not considered in the data shown; 5tMt"fi-.ATtON5 AND however, CBCPW MEAST,IREMENTS circuits could have To better understand the perfor- slightly more disper- mance of the circuits studied in this sion than normal if article, models were constructed and trapezoidal effects analyzed with the help of Sonnet Suite are greater. Professional V13.54, a three-dimen- Figure 6 shows sional (3D)planar EM simulation soft- the insertion loss ware from Sonnet Software. Based on associated with the microsectional data from the circuits two different circuit tested, the simulation geometries, gpes and \ /ith dif- such as substrate thickness and metal ferent copper sur- surface profile, were entered into the face roughnesses. software. An optical coordinate mea- At 10 GHz, the dif- suring machine (CMM) was used to ference in loss for determine the circuit length accurate- effects will cause more variation on microstrip on rough copper versus ly. Figure 7 shows an image of one of the results of CBCPW than for mi- smooth copper is approximately 0.250 the CBCPW circuits as it appears in crostrip, the plot shows that the ef- dB/in. to 0.121 dB/in. For CBCPW, fect of copper surface roughness on the difference is about 0.280 dB/ propagation constant is much less for 0.167 loss RMS = 2.0 in. to dB/in. The insertion *CPWMICRO-CPWMICRO CBCPW than for microstrip. The fig- RMi = 0.4 performance of CBCPW is less af- * CBCPW TIGHT COUPLED RMS = 2.0 ure also shows a difference in disper- fected by copper surface roughness * CBCPW TIGHT COUPLED RM5 = 0.4 sion, where the effective dielectric than the insertion loss performance u. ',2.0 constant will vary more with frequen- of microstrip. Trapezoidal effects will d ,t,.,tzA gu ;]I'9 U i l-e o 6 r.o g,l s't.t5 t.q L) !I E 1^7 G lrt 5to e.. lerouixcr,{gfr.:Ii.i.:,:,,,,:i r ;.rr 'r ,. .., r-r: lr L flg. S Effecti,oe ilielectri.c constant for tlDo circuit tApes and tuo leaels of copper surface roughness,

Microstrip RMS = 0.4 Fm Tightly Coupled CBCPIY RMS = 0.4 pm Microstrip RMS = 2.0 Fm Tightly Coupled CBCPW RMS = 2.0 Hm

O I 2.:3 4:3.6,:7'.::::8!..t9 ' TPEQUEI{CY (€tlzl':'" ' rr , :,, i.,,.::,:ir. l.i L rlg. A Comparison of loss betueen microstrip and tightly coupled CBCPW circuits uith different copper surface roughness.

80 MICROWAVE JOURNAL r IULY 2012 Te*hnicm! F*m?ur*

the Sonnet software, ally using the software. Two infinitely whlle Figure 8 shows thin metals were used in this modei, a microphotograph separated by the physical thickness of the corresponding of the metal. The topJayer metal cross-section of the has a width equivalent to the top of circuit. the physical metal, and the bottom- While Sonnet con- layer metal has a width equivalent to tains a native support the bottom. The layers are then con- for modeling thick nected with edge vias. This serves to metal, Figure 7 shows effectively model the thickness of the a thick metal approxi- metal as well as the CBCPW trapezoi- L, ng. Z lsometric r:ieu of a CBCPW cross-secti.on in Sonnet. mation drawn manu- dal effects the bottom metal can be seen protruding- slightly past the edge via, providing the "sharpness" of the physical profile. A key to achieving success in the Cutting Edge GaN simulation of these tlpes of microstrip and CBCPW circuits is a recentlv in- for the Newest Applications troduced surface-roughness *oa"t See What's Next For: to V13 of the Sonnet software. The model, which was developed by Son- net Software's software engineers in collaboration with Rogers' material developers, represents a significant advance in metal profile modeling, ac- lnmarsat 1.6 GHz 15W CAIV, FTTH, & BTS / RRH Wideband 20-2000 N4Hz 10W counting for the effects on sur{ace in- i-rlii,lrrli MMIC ductance of current following paftial "loops" in a metal conductor's profile.e While it is possible to use the new surface roughness model on the top and bottom of a PCB, it is only applied to the bottom surface of the bottom metal. Roughness is intentionally added lnmarsai / kidium 50Q SMT GaN only to this physical surface, to aid ad- hesion to the PCB dielectric material. Figure 9 offers a comparison be- tr.lzeen a simulated model and the mea- Ultra Wideband 50Q SMT GaN sured data for microstrip transmission

G-PoN. FTTH, RFoG I A ng. e Microphotograph of a cross-secti.on of a CBCPW circuit.

Microstrip RMs = 0-4 pm Microstrip RMS = 2.8 ym Simulated Microstrip RMS = 0.4 pm . eq". .j ;,

:.aa:::t::::L::P. :C:r'.r.1::. ,S|ofrr '!-: ,.1,: B,'"e 6'to.,t" ga2 Z',-o2o #F#: ft4trVr' rrc.ir*".ri#n***i {q i;assi${r -o.25;rt:i-d lt*r*fi F*ir:iitie,i j jl Itrl : il;j : 1 -l :l- :1r17i I f :' i; : ri':rn i ijr::r : i: j:,.:!rrl i"J5.s. F*rjjj:].-ir . ,-. tl www rf hlc com A fig. S Simulnteil anil measureil microstrip insertion loss.

MICROWA\1E JOURNAL I JULY 2012 Teehnicai Feature lines. The simulation shows insertion setup, the measurement procedure loss for 1". of microstrip transmission and the vali&g' of the new SonneV Iine without the connector launch, in Rogers surface roughness model. Tightly Coupled CBCPW RMS = 0.4 pm Tigfttly Coupled CBCPW RMS = 2-8 Vm order to be a valid comparison with a Figure IO shows similar agreement Simslated CBCPW RJllS = O,4 gn differential length measurement. De- between simulations and measure- Simulated CBCPW RMS = 2.8 Fm spite working in a scale of only hun- ments for CBCPW structures. While dredths of decibels, good agreement the agreement for the CBCPW circuits was achieved between the simulated is not as close as that for the microstrip and measured results for both smooth geometries, it is well within the limits (0.4 pm RMS) and rough (2.8 pm of experimental error, confirming the RMS) metal surfaces. This close agree- accuracy of the model and the mea- ment provides reassurance of the test surements made in this article.

wlltrh*,ty*y**cH

A mg. rc Simulated anil measured. CBCPW insertion loss, Having established the validity of Portable PIM Analyzer the surface roughness simulations, it might be beneficial to see how they - GSM, WiBro, UMTS/|MT2000, DCS1800/PCS, can be further used in high-frequency PCS1900, AMPS/CDMA, EGSM, W|MAX, LTE, etc. circuit design. For example, a com- - Measurement Level : -32 dBm to -132 dBm mon issue with circuit topologies like (-75 dBc to -175 dBc) - Measure the Distance to Faulty PIM Position CBCPW is finding the desired imped- - Measure the Distance to Faulty VSWR Position ance. While many textbook formulas are available for this pulpose for con- * 19" Rack Mount Types are also available ventional microstrip circuits, it is less true for CBCPW circuits. Fortunately, EM simulators are suitable for finding RFID Protocol CBCPW geometries for the desired impedance for nearly any reasonable Analyzer/Simulator circuit topology. The impedance of - Emulate the Standard RFID Reader or Tag a CBCPW design for any PCB ma- - Measure and Evaluate RFID Tags'and terial can be broken down to three Readers' protocol. main parameters: conductor width, - Verify the compliance of all lSO18000 series material thickness and ground plane - Measurement of Frequency, Power, and separation. As Figure f f shows, the Modulation effects ofeach ofthese parameters on - Display Timing Waveform / Code Data CBCPW transmission-line impedance can then be readily analyzed within EM Radar Signal Generator the simulator environment. Once parameterization is com- Variety of Models (Poriable, Rack Mountable, plete, a simulation can be run, which or Customizing Models), Wideband (0.5-40GHz) automatically "sweeps" all combina- Multiple Signal Generation (1-48 signals or more, tions of the three parameters within simultaneous) a desired range. It is then convenient PRl, PW SCAN, FREQ, PHASE Modulation to plot all impedances on the same (Programmable or Pre-stored Library) AOA ouput for RWR/MWR Receiver Test & graph, allowing a designer to choose Evaluation the best geometry and impedance Visual Scenario Editing on 2D or 3D maps. from the results. Figure 12 shows an example of such an impedance plot. CONCLU'ION The performance levels of microstrip, CBCPW launched microstrip and CBCPW transmission Iines were evaluated under controlled conditions. Both measurements and

MrcRowAVE JOURNAT r JULY 2Or2 T**h*l**i F*atur*

computer simulations . 57. 56 were performed using 55 a commercial, low-loss Ii r microwave substrate P53 material with different 252 lD rr copper types. includ- ing different values =50 of 49 copper conductor sur- 48 roughness. 47 face The ef- .,:.::2iO:,:: 0 l.O .. .r!:O-i, fects of copper surface FREQUENff{Gfiz}rri]:r. roughness were evaluated compared, Lmg. n Example of a "parameter naeep" and L f4. U CBCPW geometrg can be broken into three key parameters. simul&tion. showing that greater roughness ty,pically means greater loss. Dif- ferent circuit topologies were compared through both measurements and simulations and, by properly ap- plying computer simulati,on ioftware, it is possible to reduce the difficulties often encountered with lesser-known circuit topologies. I re References 1. B.Rosas, "Optinilzing Test Boards for 50 CHz End Launch Connectors: Croatded Coplanar Launches and. Tltrouglt Line,s on 30-m.il Rogers RO1350B uitlt Conparison to Microstrip," Southwest Microu'ave Inc., Tempe, AZ, 2007, wwrvsouthu'estmicro- wave.coln. 2. R.N. Simons, "Coplanar Waxeguide Cir- cuits, Conlponents and Systems," John Wiley & Sons, NewYork, NY,2001. 3 JW Reynolds, PA. LaFrance, J.C. Rautio and A.F. Horn III, "Effect of Conductor Pro- fiIe on the Insertion loss, Propagation Con- stant and Dispersi.on in Thin High Frequen- cy Transmission Lines," DesignCon, 2010. 4. S.P Morgan, "Effect of Surface Roughness on Eddy Curent Losses at Microwave Fre- quencies," lournal of Applied Plrys.ics,YoI. 20, No. 4, 1949, pp. 352-362. 5. S. Groisse, I. Bardi, O, Biro, K. Preis and K.R. Richter, "Parameters of Lossy Cavity Resonators Calculated by Finite Elen-rent Method," IEEE Transactions on Magnetic,s, Vol. 32, Part 1, No. 3, May 1996, pp. 1509- 15I2. G. Brist, S. Hall, S. Clouser and T. Liang, "Non-classical Conductor Losses Due to Copper Foil Roughness and Treatment," 2005 IPC Electronic Ctrailts World. Con- .#.r* *,,*,, .JJ,,*"f Dention Digest, Vol. 2, S19, pp. 898-908. "L,r; "A . . .. .,_. ,q:.:r: *i:r.ri;=s.jgl-r,:,ji 7. T. Liang, S. Hall, H. Heck and G. Brist, Practical Method for Modeling PCB Trans- ;tJ g J*, * i*t"'--f;*t+Y lnission Lines with Conductor Roughness and Wideband Dielectric Properties," 2006 IEEE MTT-S Intemational Microuaue Sym- posium Digest.pp 1780-1783. N.K. Das, S.M. Voda and D.M. Pozar, "Two Methods for the Measurement of Substrate Dielectric Constant." IEEE Transactions on Microroaoe Theory anrl Techniques, YoI. MTT-35, No. 7, July 1987, pp. 636-642. A,F. Horn, J.W Relnolds and j.C. Rautio, "Conductor Profile Effects on the Propaga- tion Constant of Microstrip Transmission Lines," 2070 IEEE MTT-S International Mi- crouaDe Symposiwn Digest, pp. 868-87I.

MICROWA\aE IOURNAL I IULY 2012