75 Ohm, 26 GHz Vector Network Analyzer And 30 Ps TDR - TDT

2m ago
451.04 KB
6 Pages

Application Note AN-17Copyright November 200475 Ohm, 26 GHz Vector Network Analyzerand 30 ps TDR - TDTJames R. Andrews, Ph.D., IEEE FellowPSPL Founder & former President (retired)INTRODUCTION Historically, 50 Ω coaxial systemshave been used for RF and microwave applications,while 75 Ω systems have been used for the video andtelevision industries. The original reasons for theseimpedance choices were that high RF power handlingwas optimized in 50 Ω coaxial cable, whereasminimum coax attenuation was obtained at 75 Ω. Therequirements for utilizing high RF power, and thus 50Ω coax, at microwave frequencies ( 1 GHz) datesback many years to WWII. The video/TV industryhistorically never required performance beyond 1 GHz.A typical color video camera's signal only extended upto 4.2 MHz. RF transmission of TV signals, eitherbroadcast over the air or within cable TV systems, wasin the spectrum from 50 to 800 MHz. Today, there arenew pressures on the frequency performance of 75 Ωcoax systems. The new high definition TV nowrequires sending 1.5 Gb/s HDTV digital data over 75 Ωcoax cables within TV studio plants. Also, interconnectcable impedances higher than 50 Ω are becomingmore desirable for various digital circuits operating atGigabit data rates. The compelling reason for 75 Ω vs.50 Ω here is the 1/3 reduction in current drive requiredto maintain the same logic voltage levels.A key test instrument used for evaluating manyelectronic components is the Vector Network Analyzer(VNA). 50 Ω VNAs are offered by severalmanufacturers to frequencies as high as 110 GHz. For75 Ω, only a few VNAs are available up to a maximumof 4 GHz [1]. They include Anritsu (to 2 GHz), Agilent(to 3 GHz), Advantest (to 3.8 GHz), and Rohde &Schwarz (to 4 GHz). They all use 75 Ω N connectors.Their N calibration kits are only rated up to 3 GHz.Several years ago, PSPL received a customer requestto build custom, 75 Ω bias tees capable of operating atfrequencies from a few kHz to in excess of 5 GHz. Weimmediately were faced with the problem of not havingsuitable 75 Ω coax connectors or test equipment,including a TDR or VNA. We were able to solve theseproblems, using 75 Ω SMA connectors and by buildingour own VNA, using our model 4015, 15 ps pulsegenerator and a 50 GHz, HP oscilloscope. We are nowusing our new, 7 ps model 4020 TDR/TDT pulsegenerator. This application note documents our 75 Ω,combined 30 ps (10 ps risetime with normalization)TDR/TDT and 26 GHz TDVNA instrument.SMA 75 Ω CONNECTORSHistorically, almost allRF and microwave test instruments have beenstandardized to 50 Ω coaxial interconnects. 50 Ωinstruments typically use BNC ( 1 GHz),N ( 18 GHz), SMA ( 18 GHz), 3.5mm ( 26 GHz),K ( 40 GHz), 2.4 mm ( 50 GHz), or V ( 65 GHz)connectors. 75 Ω test instruments have used BNC ( 1GHz), N ( 3 GHz), or rarely F ( 1 GHz), 75 Ω coaxialconnectors. A potential major breakthrough in 75 Ωinstrumentation is now possible with the introduction of75 Ω, SMA connectors. In the late 90s, the QMIcompany (Quality Microwave Interconnects) foresawthe need for higher performance, 75 Ω microwaveconnectors and developed an extensive line of 75 Ω,SMA connectors, adapters, cables, and SMA-N andSMA-BNC adapters [2]. The 75 Ω SMA items fromQMI were only specified up to 3 GHz because thatwas the upper test limit of their 75 Ω, type N,commercial VNA. PSPL has found, however, thatthese QMI, 75 Ω, SMA connectors work very well up to18 GHz and in some cases to 26 GHz, and arecomparable in performance to their 50 Ω SMAcounterparts. They form the basis of our 75 Ω, 26 GHzTDVNA and 30 ps TDR/TDT.Fig. 1 QMI 75 Ω SMA connectorsPICOSECOND PULSE LABS, P.O. BOX 44, BOULDER, CO 80306 USA TEL: 1.303.443.1249 FAX: 1.303.447.2236WWW.PICOSECOND.COMAN-30417, Revision 1, November 2004Page 1 of 6

Application Note AN-17Copyright November 200475 Ω MEASUREMENTS WITH 50 Ω INSTRUMENTSTo make measurements on 75 Ω items with a 50 Ωinstrument typically has required the use of either a50/75 Ω impedance transformer, or a minimum loss50/75 Ω "L" matching pad, Figure 2, [3]. sformers offer very low insertion loss, but areusually only useful up to about 500 MHz. If carefullymade, the resistive minimum loss pad is useful toconsiderablyhigherfrequencies.Commercialminimum loss L pads with type N connectors areavailable up to 3 GHz. The major drawback to theminimum loss pad is its inherent insertion loss of 5.7dB and attendant loss in measurement systemdynamic range. If one doesn't account for this loss inthe calibration, then a 0 dB return loss in a 75 Ωsystem becomes -11.4 dB when measured in a 50 Ωsystem.Fig. 3 Basic TDR, TDT and TDVNA MeasurementSet-upτ Vtrans (t) / Vinc(t)(1)S21(f) FFT[Vtrans(t)] / FFT[Vinc(t)] (2)ρ Vrefl (t) / Vinc (t)(3)S11(f) FFT[Vrefl (t)] / FFT[Vinc (t)] (4)Fig. 2 Minimum Loss 50 / 75 Ω "L" Matching PadTD VECTOR NETWORK ANALYZERFigure 3shows the basic test set for a Time Domain VectorNetwork Analyzer (TDVNA). It is also the basic test setfor making TDR and TDT measurements. PSPL'sTDR/TDT and TDVNA application notes, AN-15 andAN-16, [4 & 5] should be referred to at this time asthey compliment this application note. Figure 3 is alsothe classical setup used in the frequency domain formaking S parameter insertion and reflectionmeasurements of S21 and S11. A sine wave generatoris used as the signal source (Vg, Rg) for nts, a pulse generator is used as the signalsource, and an oscilloscope is used to observe andmeasure the resultant waveforms, Vtdr(t) and Vtdt(t).For frequency domain S parameters, the Fast FourierTransform (FFT) is used to transform the time domainTDT and TDR waveforms into frequency domain data.TDT / S21 Test SetPSPL has the same problemthat other instrument manufacturers have in building a75 Ω VNA. That is that all our really high bandwidthproducts were all designed as 50 Ω systems. Ourfastest pulse generators that we used for this TDVNAare our 50 Ω models 4020/22 with risetimes of theorder of 5 to 10 ps and 2.4 mm, 50 Ω connectors.Likewise the oscilloscope we used was an HP 50 GHzscope with 50 Ω, 2.4 mm connectors. We thus need tomodify the basic set-up of Figure 3 to include someform of 50/75 Ω impedance matching or adaptation.See Figure 4. PSPL uses three different Z matchingtechniques, depending upon the desired upperfrequency limit. In order of increasing bandwidth, theyare the L pads, Figure 2, one-way 25 Ω seriesmatching resistors, Figure 5, or abrupt junctiontransitions, Figure 6.Fig. 4 75 Ω TDT and S21 TDVNA, using Z matchingnetworks and 50 Ω test equipmentPICOSECOND PULSE LABS, P.O. BOX 44, BOULDER, CO 80306 USA TEL: 1.303.443.1249 FAX: 1.303.447.2236WWW.PICOSECOND.COMPage 2 of 6AN-30417, Revision 1, November 2004

Application Note AN-17Copyright November 2004Fig. 5 One-way 25 Ω series matching resistorFig. 6 Abrupt junction 75 / 50 Ω transitionThe classical technique is to use minimum loss L padson either side of the 75 Ω Device Under Test (DUT).We have built experimental SMA, L pads, using 0603chip resistors and QMI connectors. Our L pads have arisetime of 23 ps and a -3 dB bandwidth of 15 GHz.(BW 0.35/Tr). For a pair of these L pads, the risetimeand bandwidth would be about 33 ps and 11 GHz.Thus we feel they are useful for VNA measurementsfrom DC up to 5 GHz.Fig. 7 75 Ω to 50 Ω series matching resistor. The 25Ω chip resistor is mounted between the center pins ofthe 75 Ω and 50 Ω coax connectors.A higher bandwidth arrangement is to use one-wayseries matching resistors on either side of the 75 ΩDUT. These are a single 25 Ω resistor mounted inseries between a 50 Ω coax line and a 75 Ω coax line,Figure. 5. When the 50 Ω connector is terminated in50 Ω, then a matched impedance of 75 Ω is seenlooking into the 75 Ω connector. There is, however, amismatch of 100 Ω (ρ 1/3) when viewed from the50 Ω side. If the signal generator, Rg, and thetermination, Rl, are relatively well matched to 50 Ω,then serious, multiple reflections will be suppressed.Oscilloscopes usually are rather well matched to 50 Ω.(Caution — One manufacturer's 70 GHz sampler is asevere mismatch of 70 Ω!) Pulse generators areusually not as well matched to 50 Ω. However, if onehas a higher than necessary pulse generatoramplitude, then a good quality, coax attenuator can beattached to the pulse generator output to provide goodimpedance matching. We built experimental 25 Ωseries matching resistors in a coaxial housing, Figure7. The resistor is a 0603 chip. Extended Teflondielectric, flange mount, SMA connectors were used.The 75 Ω SMA connector is a QMI p/n 3-E926-190-12(jack) or 3-E926-110-10 (plug). The 50 Ω SMAconnector is a CDI p/n 5220CC (jack) or 5340CC(plug). The risetime of this unit is 11 ps. The risetimefor a pair would be about 16 ps, with a resultant -3 dBbandwidth of 22 GHz.The highest bandwidth arrangement is shown inFigure 6. In this case, no attempt is made to matcheither the 50 Ω source or 50 Ω termination to the 75 Ωcables or DUT. If the 50/75 Ω abrupt transitions arebuilt properly, then the minimum risetime degradationis introduced and, thus, the bandwidth is maximized.There is, however, a definite lower frequency limitimposed by the necessity to avoid measuring any ofthe many multiple reflections that will occur in this setup. The reflection-free time window of observation isestablished by the two way transit time of the cables,DL1 and DL2. We have built experimental abrupttransitions with the same SMA connectors and coaxhousing as shown in Figure 7. The risetime of a single50/75 Ω transition is 5 ps. The risetime and bandwidthfor a pair would be 7 ps and 50 GHz.TDR / S11 Test Set TDT and S21 measurements arethe easiest to perform, and the maximum bandwidth isachieved for them. There is added complexity to TDRand S11 measurements. They are complicated by thenecessity to pick off the signal, Vtdr. The extracomponents required for TDR tend to slow down thesystem risetime and lower the S11 bandwidth. Figure 8shows how to use a 75 Ω , 6 dB power divider teealong with a pair of 25 Ω series matching resistors. Tominimize the risetime degradation, the two separate 25Ω resistors shown in series could be consolidated intoa single 50 Ω resistor. See Figure 9. A special 75 ΩTDR tee was built for the PSPL 75 Ω TDR and S11TDVNA, using chip resistors in the circuit of Figure 9 ina coax tee housing similar to the in-line housing shownin Figure 7.PICOSECOND PULSE LABS, P.O. BOX 44, BOULDER, CO 80306 USA TEL: 1.303.443.1249 FAX: 1.303.447.2236WWW.PICOSECOND.COMAN-30417, Revision 1, November 2004Page 3 of 6

Application Note AN-17Copyright November 2004Fig. 8 75 Ω TDR using a 75 Ω , 6 dB TeeFig. 9 Special 75 Ω TDR Tee for use in a 50 Ω system75 Ω Calibration StandardsA set of 75 Ω, SMAcalibration standards were built to use with thisTDVNA. In both sexes, they included offset shorts,offset opens, 75 Ω terminations, and equi-phase sexchange adapters. The shorts, opens, and equi-phaseadapters were all built using the extended Teflondielectric, flange mount, QMI, SMA connectors and thecoax housing shown in Figure 7. The time windowisolation lines, DL1 and DL2 were either the equi-phaseadapters or short lengths of 0.141", semi-rigid coaxcable (Haverhill, EZ-141-75). The 75 Ω referencetermination was a long length of EZ-141-75 coax cableterminated at the far end by a QMI 75 Ω termination(p/n 3-E809-790-11).PULSER and OSCILLISCOPEAll measurementswere performed using a Hewlett-Packard model 54750oscilloscope with a 50 GHz bandwidth (9 ps risetime)sampling head. The pulse generator used was aPicosecond Pulse Labs model 4020 TDR pulser. Thepulse head used was the positive polarity,transmission head, model 4020RPH-TP, whichproduced a 2.2 V, 7 ps risetime, 38 ns duration, steppulse. Figure 10 shows the leading edge waveformfrom this pulser as measured on the HP 50 GHzo'scope.Fig. 10 PSPL 4020 TDT pulser measured on HP 50GHz oscilloscope, 0.5 V/div and 10 ps/div75 Ω TDT and TDR PERFORMANCE Figures 11-13demonstrate the typical performance of this equipmentfor performing TDT and TDR measurements in 75 Ωcoax. For the TDT measurements, the time windowisolation lines were a pair of PSPL 75 Ω equi-phaseadapters, which gave a reflection-free window of 450ps. Figure 11 shows the leading edge of the TDT testpulse, using the three different Z matching networks ofeither the 50/75 Ω abrupt junctions, 25 Ω seriesresistors, or L pads. The measured risetimes were 17ps, 26 ps, and 37 ps, respectively. The equivalent -3dB bandwidths were 20 GHz, 13 GHz, and 9 GHz.Figure 12 shows the 75 Ω TDR test pulse as reflectedfrom an open, short, and 75 Ω termination at the endof a 3", 75 Ω semi-rigid coax, which was used as thereference line, DL1. The TDR arrangement used thespecial 75 Ω TDR tee (see Figure 9). The falltime ofthe reflected short was 29 ps. The equivalent -3 dBbandwidth is thus 12 GHz. It is possible to effectivelydecrease the TDT and TDR system risetimes down to10 ps, using deconvolution and digital filtering. Thiscapability is firmware built into the HP and Agilentoscilloscopes [6] and is called 'normalization'. Figure13 shows the result of normalizing to 10 ps risetimethe TDT and TDR pulses from Figures 11 and 12.Fig. 11 75 Ω TDT test pulses with various Z matchingnetworks, 100 mV/div and 20 ps/divPICOSECOND PULSE LABS, P.O. BOX 44, BOULDER, CO 80306 USA TEL: 1.303.443.1249 FAX: 1.303.447.2236WWW.PICOSECOND.COMPage 4 of 6AN-30417, Revision 1, November 2004

Application Note AN-17Copyright November 2004The hi-bandwidth S21 arrangement of Figure 6produces useful measurements even further outbeyond 26 GHz, but is troubled by higher ordermoding in the SMA connectors.Fig. 12 75 Ω TDR — reflections from open, 75 Ω, andshort; 60 mV/div & 20 ps/divFig. 14 75Ω TDVNA system self-test for dynamicrange; S21 (green) and S11 (blue)Fig. 13 10 ps normalized TDT and TDR test pulses,20 ps/divTDVNA PERFORMANCE For the tests shown here,th