tips of using TDR probe

Hi,We're evaluating the performance of our TDR probe to determine if it can be used in VNA measurement. Here is some observations:1.       Although in most people's opinion the contact quality of probe is far less reliable & trustable compared to SMA, based on my experience, it's feasible to achieve a stable contact with probe, at least for the bandwidth up to 20GHz.2.       Besides measuring the sparam of the tline or other components on PCB without the trouble of fabricating the test fixture, another benefit of directly probing is the masking effect due to the fanout traces on the fixture is avoided-more details will be shown on the TDR profile.3.  Experiments showed that the contact of ground collar on the TDR probe can reduce the inductive peak at the very beginning of the TDR curve. It seems the ground conductor must be surrounding the signal - a single ground wire won't have much effect.4.  The deembeding of the probe is challenging - lots of glitches appeared on the deembedded sparam. We're still working on that trying to find out the best method.I would appreciate if you can share your experience with using TDR probe either for TDR measurement, or for sparam measurement.Best Regards,Sherman ChenSignal IntegrityEMC Global Hardware EngineeringTel: +86 21 60951100-3329
sherman.chen 5 years 8 months 8 days

16 answers


The best answer


You can select the best answer for current question!
Answered byhan.guobing 5 years 8 months 27 days
Hello Tom,Agree with you, to be sure not to probe in a middle of a trace to make theprobe point closer.As for the leading trace length, I think 3~4inch is acceptable if the Df isvery little, else 0.5~2inch is acceptable.BTW, I think that directly probe at the via pad is not recommended. :)Thanks,Robin2015-01-26 10:17 GMT+08:00 Tom Dagostino :> Robin>>>> There are practical limits to the location for TDR probing.  You should> not probe in the middle of a trace, you will get reflections off both sides> of the insertion point causing a confusing display that will be hard to> interpret.  Likewise probing at the DUT with 5 inches of trace from a> connector or whatever will give you hard to interpret data.  Pick the best> point that you can, preferably at the end of the trace leading up to the> DUT.  If there are connectors leading up to the final trace and DUT then> disconnect the other boards or cables and probe at the connector.   I.e.,> don't try to TDR through more trace than you have to .>>>> Tom Dagostino>>>> Teraspeed Labs>> 9999 SW Wilshire Street>> Suite 102>> Portland, OR 97225>>>> tom@xxxxxxxxxxxxxxxxx>> www.teraspeedlabs.com>>>> 971-279-5325 office>> 503-430-1065 cell>>>> *From:* Han, Guobing [mailto:han.guobing@xxxxxxxxx]> *Sent:* Sunday, January 25, 2015 5:36 PM> *To:* Chen, Sherman> *Cc:* tom@xxxxxxxxxxxxxxxxx; Charles.Grasso@xxxxxxxxxxxx;> heidi_barnes@xxxxxxxxxxxx; si-list@xxxxxxxxxxxxx> *Subject:* Re: [SI-LIST] Re: tips of using TDR probe>>>> Hello Sherman and Tom,>>>> I am afraid that I can't agree with point 1 to launch the step at the> closest location to the DUT, which will make the peeling effect worse.>> To be simplifed, the first reflection point is the probe, the second one> is the input point of DUT, the third is the second refection point of DUT.>> If the delay between probe and input point of DUT is too short, such as> 1~5ps, the come and back reflection of input point of DUT will superimpose> to the second and later refection point of DUT. Thus, make the peeling> effect worse.>>>> So, my point is just opposite of yours, it's not recommended to launch the> step at the closest location to the DUT by a probe.>>>>>> Thanks,>> Robin>>>>>>>> 2015-01-22 23:45 GMT+08:00 Chen, Sherman :>> Tom,>> Agree with you on all you said.> My point is:> a. launch the step at the closest location to the UUT thus to reduce the> peeling effect.> b. reduce the preceding peak or dips as much as could. Better to be even> with Z0.> Then you will obtain quite accurate impedance profile of UUT.> We taught our JDM this method and they applied it in measuring vias. The> results are pretty good.>> Best Regards,>> Sherman Chen> Signal Integrity> EMC Global Hardware Engineering> Tel: +86 21 60951100-3329>>>>> -->> Thanks,> - Robin (Han, Guobing)> TEL: 86-21-61094805> MSN: han_guobing@xxxxxxxxxxx>-- Thanks,- Robin (Han, Guobing)TEL: 86-21-61094805MSN: han_guobing@xxxxxxxxxxx
Answered byhan.guobing 5 years 8 months 27 days
Hello Sherman and Tom,I am afraid that I can't agree with point 1 to launch the step at theclosest location to the DUT, which will make the peeling effect worse.To be simplifed, the first reflection point is the probe, the second one isthe input point of DUT, the third is the second refection point of DUT.If the delay between probe and input point of DUT is too short, such as1~5ps, the come and back reflection of input point of DUT will superimposeto the second and later refection point of DUT. Thus, make the peelingeffect worse.So, my point is just opposite of yours, it's not recommended to launch thestep at the closest location to the DUT by a probe.Thanks,Robin2015-01-22 23:45 GMT+08:00 Chen, Sherman :> Tom,>> Agree with you on all you said.> My point is:> a. launch the step at the closest location to the UUT thus to reduce the> peeling effect.> b. reduce the preceding peak or dips as much as could. Better to be even> with Z0.> Then you will obtain quite accurate impedance profile of UUT.> We taught our JDM this method and they applied it in measuring vias. The> results are pretty good.>> Best Regards,>> Sherman Chen> Signal Integrity> EMC Global Hardware Engineering> Tel: +86 21 60951100-3329>>-- Thanks,- Robin (Han, Guobing)TEL: 86-21-61094805MSN: han_guobing@xxxxxxxxxxx
Answered bytom 5 years 8 months 28 days
RobinThere are practical limits to the location for TDR probing.  You should notprobe in the middle of a trace, you will get reflections off both sides ofthe insertion point causing a confusing display that will be hard tointerpret.  Likewise probing at the DUT with 5 inches of trace from aconnector or whatever will give you hard to interpret data.  Pick the bestpoint that you can, preferably at the end of the trace leading up to theDUT.  If there are connectors leading up to the final trace and DUT thendisconnect the other boards or cables and probe at the connector.   I.e.,don't try to TDR through more trace than you have to .Tom DagostinoTeraspeed Labs9999 SW Wilshire StreetSuite 102Portland, OR 97225tom@xxxxxxxxxxxxxxxxx www.teraspeedlabs.com  971-279-5325 office503-430-1065 cellFrom: Han, Guobing [mailto:han.guobing@xxxxxxxxx] Sent: Sunday, January 25, 2015 5:36 PMTo: Chen, ShermanCc: tom@xxxxxxxxxxxxxxxxx; Charles.Grasso@xxxxxxxxxxxx;heidi_barnes@xxxxxxxxxxxx; si-list@xxxxxxxxxxxxxSubject: Re: [SI-LIST] Re: tips of using TDR probeHello Sherman and Tom,I am afraid that I can't agree with point 1 to launch the step at theclosest location to the DUT, which will make the peeling effect worse.To be simplifed, the first reflection point is the probe, the second one isthe input point of DUT, the third is the second refection point of DUT.If the delay between probe and input point of DUT is too short, such as1~5ps, the come and back reflection of input point of DUT will superimposeto the second and later refection point of DUT. Thus, make the peelingeffect worse.So, my point is just opposite of yours, it's not recommended to launch thestep at the closest location to the DUT by a probe.Thanks,Robin2015-01-22 23:45 GMT+08:00 Chen, Sherman :Tom,Agree with you on all you said.My point is:a. launch the step at the closest location to the UUT thus to reduce thepeeling effect.b. reduce the preceding peak or dips as much as could. Better to be evenwith Z0.Then you will obtain quite accurate impedance profile of UUT.We taught our JDM this method and they applied it in measuring vias. Theresults are pretty good.Best Regards,Sherman ChenSignal IntegrityEMC Global Hardware EngineeringTel: +86 21 60951100-3329-- Thanks,- Robin (Han, Guobing)TEL: 86-21-61094805MSN: han_guobing@xxxxxxxxxxx
Answered bysherman.chen 5 years 8 months 1 day
Tom,Agree with you on all you said. My point is:a. launch the step at the closest location to the UUT thus to reduce the peeling effect. b. reduce the preceding peak or dips as much as could. Better to be even with Z0.Then you will obtain quite accurate impedance profile of UUT. We taught our JDM this method and they applied it in measuring vias. The results are pretty good. Best Regards,Sherman Chen Signal Integrity EMC Global Hardware EngineeringTel: +86 21 60951100-3329 -----
Answered bytom 5 years 8 months 1 day
ShermanI think we are in complete agreement.Tom DagostinoTeraspeed Labs9999 SW Wilshire StreetSuite 102Portland, OR 97225tom@xxxxxxxxxxxxxxxxx www.teraspeedlabs.com971-279-5325 office503-430-1065 cell-----
Answered byal 5 years 8 months 1 day
This has been an interesting thread, thanks.   Comments are follow on from Heidi and Tom.A NIST traceable coaxial with cal kit (or probe and substrate) serves to establish a reference plane and secondly an error block that reduces system error to residual error.   The significant errors are source and load match, directivity, and transmission error.Each one of these errors relates to the device under test reflectance, which is why a broadband SOLT calibration uses a short and open (high reflectance) with a load and THRU (low reflectance and unity transmission).  It provides a calibration error block encompassing a wide range of reflectance.   I would also add Group Delay noise and dynamic range as another metric, especially  when you are using de-embedding.   Accordingly, if you calibrated the probes using some type of substrate, or modeled the error and performed some flavor of measure-modeled de-embedding approach you should evaluate the residual error of measured standards including the following test structures:1.  short transmission system, low loss, 2 ports2.  highly reflective structure such as open, being careful of fringing fields, 1 port3   resonant structures with no TEM issues way past your Fstop, such as Beatty Standards, offset stub like resonators, 2 ports4.  Time domain resolution structureWe added a whisker like structure, for our Channel Modeling CMP-28,  that tests time domain of either time domain transformed VNA S-parameter data or direct TDR for 10 and 30psec rise times.   The whisker geometry has specific resolving metrics.   WRT will be offering a complimentary Web Presentation on Resonant Structures that is low marketing and very technical after DesignCon.   What is so novel about Beatty standard it is both reflective AND non-reflective depending on frequency, material properties of the planar standard.    We believe these class of old school Microwave test structures are underutilized in the S.I. community.Ping me for anyone interested in the seminar.  For attendee of DesignCon we address some of these issues for our tue afternoon Tutorial we are collaborating on with Keysight and Speeding Edge.  tuesday 3:30pm, DesignCon, "PCB Materials, Simulations, and Measurements for 32 Gb/s�.    This tutorial includes de-embedding comparisons and approaches, along with a primer on Beatty Standards for material extraction.- Al NevesProducts for the Signal Integrity PractitionerAlfred P. NevesChief TechnologistOffice: 503-679-2429www.wildrivertech.com> On Jan 22, 2015, at 7:45 AM, Chen, Sherman  wrote:> > Tom,> > Agree with you on all you said. > My point is:> a. launch the step at the closest location to the UUT thus to reduce the > peeling effect. > b. reduce the preceding peak or dips as much as could. Better to be even with > Z0.> Then you will obtain quite accurate impedance profile of UUT. > We taught our JDM this method and they applied it in measuring vias. The > results are pretty good. > > Best Regards,> > Sherman Chen > Signal Integrity > EMC Global Hardware Engineering> Tel: +86 21 60951100-3329 > > > -----
Answered bysherman.chen 5 years 8 months 3 days
Who is Chris? :) Yes, the solution is a bit tricky but the concept is actually simple - just to keep the ground reference as continuous as possible. I'm writing an article on this right now. Will let you know when I am permitted to publish it. Best Regards,Sherman Chen Signal Integrity EMC Global Hardware EngineeringTel: +86 21 60951100-3329 -----
Answered bysherman.chen 5 years 8 months 3 days
Hi Tom,Thanks for sharing the valuable experiences. To verify my measurement results, I did numerous simulation in ADS. Here is what I got:1. an inductive peak or a capacitive dip at the launch will smear the peaks and dips at the following discontinuities. But looks they don't distort the average impedance much. 2. a lossy tline also has the similar impact. So the point here is: when there exist peaks or dips, or a segment of lossy tline in front of the UUT, the peaks and dips displayed on  the TDR profile are very likely distorted. Meanwhile  the average impedance of  the  UUT mostly is still trustable. As I said in my last mail, to accurately evaluate the impedance of a UUT, I will first try to set the launch point closest to the UUT. If it's not possible to do so, above conclusion can be used to estimate the average impedance of the far off UUT. Best Regards,Sherman Chen Signal Integrity EMC Global Hardware EngineeringTel: +86 21 60951100-3329 -----
Answered bytom 5 years 8 months 3 days
ShermanThe resolution of any measurement system is dependent on the measurementbandwidth at the DUT.  Just about anything between the instrument and theDUT will cut the available bandwidth - and discontinuities can pack a bigpunch.  And in many cases you will have two, one at the input and one at theoutput of the DUT for insertion loss type measurements and once going in andonce going out for return loss measurements.  It is well known that TDR's ability to resolve features is related to therisetime of the TDR edge at the DUT, and that the TDR edge has to go throughthe system twice, once getting to the DUT and once getting back to thescope.The other thing you will run into the reflections in the system will makethe downstream measurements inaccurate.  This can be compensated for withIConnect from Tek.  At every discontinuity some of the energy is reflectedback to the scope.  This implies there is less energy propagating past thediscontinuity.  If you look at the equations for calculating the impedancebased on the amount of reflected voltage you will see that both the systemZo and the incident voltage are used.Vreflected = Vincident(Zx-Zo)/(Zx+Zo)Thus for accurate measurements the incident voltage must be known at theinput to every discontinuity.  The peeling algorithm in IConnect does thiscomputation to accurately calculate the impedance.Tom DagostinoTeraspeed Labs9999 SW Wilshire StreetSuite 102Portland, OR 97225tom@xxxxxxxxxxxxxxxxx  www.teraspeedlabs.com 971-279-5325 office503-430-1065 cell-----
Answered byCharles.Grasso 5 years 8 months 4 days
Hello Chris - I am intrigued. It sounds like you solved a tricky problem. Are you permitted to expand on your " enhancing the shielding" solution?  I am wondering how you avoided changing the impedance of theprobe.Best RegardsCharles GrassoCompliance EngineerEchostar Communications(w) 303-706-5467(c) 303-204-2974(t) 3032042974@xxxxxxxxx(e) charles.grasso@xxxxxxxxxxxx(e2) chasgrasso@xxxxxxxxx-----
Answered bysherman.chen 5 years 8 months 4 days
Hi Heidi,Thanks for sharing the tips. I was able to figure out the way how to accurately do TDR measurement by enhancing the shielding around the very ends of the probe tips. Now we are using this method for correlating simulation to measurement even on very short structures such as vias. And it works fine.For the method of using VNA to measure the sparam of the UUT,  I think although theoretically any types of probe can be used with ISS (impedance substrate standard) to calibrate themselves out, despite they are  highly reflective or not, due to the calibration error, a highly reflective probe may not deliver a satisfying calibration result. I will do some experiments to verify this when I get a chance to play with ISS.  Best Regards,Sherman Chen Signal Integrity EMC Global Hardware EngineeringTel: +86 21 60951100-3329 -----
Answered bytom 5 years 8 months 4 days
Hi Sherman and othersI've worked with TDR and all kinds of interfacing to the DUT.  Coaxial andmicroprobes tend to be best IF the launches into the test board are welldesigned.  If  not you have created a low pass filter ahead of your DUT.This low pass filter will either be an RC like filter if your launch ifcapacitive or a series inductance if your launch is inductive.  This filterdoes two things, first it limits the insertion loss measurement. Secondly iscan severely limit the return loss performance.  At higher frequencies youare seeing the launch not the DUT.  With a coaxial interface you cancalibrate out the effects of the launch if the launches have consistentperformance, i.e.,  they have identical TDR responses.   If you apply therule of thumb of insertion loss minus return loss should be greater than 15dB for a quality measure at a given frequency you can see how important thereturn loss effect of a poor launch can impact your measurements.I ran some experiments with some launches we have to illustrate this effect.There were three cases, a very good launch designed by Teraspeed, apublished launch by the connector vendor and a "hand soldered" launchtypical of someone trying to use a piece of coax soldered to the board.  Thediscontinuity cause by the launches were +2 Ohms, +15 Ohms and +40 Ohmsinductive respectfully if my memory serves me.  The corresponding effectivebandwidths were about 20 GHz, 6 GHz and 1 GHz.   I defined effectivebandwidth as a 15 dB spread between the insertion and return loss.With hand held probes you will run into a consistency issue and unless youdesign a consistent and effective return path you will have an inductivelaunch into your DUT.  And if you have a differential probe the bandwidth ofthat probe is determined by the spacing of the probe tips.  The wider thespacing the larger the inductive loop is at the probe/DUT interface.  Forsingle ended probes the launch will also have an inductive characteristiccaused by the loop area of the probe/ground return path.  And who knows whatthe characteristics of the DUT's interface looks like.  Unlike a designedlaunch you have with the coax interface you may be probing a couple of testpoints 0.100" apart.  They will in all likelihood look inductive or if thepads are large and use large drill sizes - capacitive.If you used something to keep the impedance constant between the probe andthe DUT then you will have a better shot at a meaningful measurement.  Butthis has to be consistent from test to test and during any de-embeddingcalibration you do.  If they do not then the difference between the twomeasurements (calibration and measurement) will be placed between yourmeasurement system and the DUT.Many of the TDR probes are measure for differential measurements and don'treally use a ground return.  These can do an excellent job of makingdifferential insertion or return loss measurements if the interface to theDUT is electrically clean.  And they are do a very respectful job ofdifferential Zo measurements.I've made test jigs to hold the hand held probes so that I got  consistentplacement of the probe onto the circuit board.  This helps, the geometry isthe same from measurement to measurement and it frees up the hands to pushbuttons and capture the waveforms.  And trying to hold a probe on a testpoint consistently while waiting for a VNA sweep or multiple average TDRmeasurement is a pain.As Heidi points out for quality measurements you need quality cables.  Thesame things applies to the interface between the probe and the DUT, it hasto be consistent.Hope this helps.Tom DagostinoTeraspeed Labs9999 SW Wilshire StreetSuite 102Portland, OR 97225tom@xxxxxxxxxxxxxxxxx  www.teraspeedlabs.com 971-279-5325 office503-430-1065 cell-----
Answered byheidi_barnes 5 years 8 months 7 days
.... yes my spell checker wasn't working for my fast typing, however same difference escaping fields/gas you get the picture.-Heidi-----
Answered byheidi_barnes 5 years 8 months 7 days
Hi Sherman,Investing time in set-up and training to use probes can pay off when it comes to making measurements, but it does require a methodical process for repeatability.1) Connecting to probes with an instrument tends to leave the connecting cables unsupported with multiple bends.  This means that high performance phase stable cables (and new ones that have not yet been damaged) are needed to avoid errors in the calibration due to phase changes from cable movement.  Investing in the setup to minimize cable movement and provide support for the cables can reduce the phase errors which should help with a cleaner de-embedding.2) Effervescent waves love to travel on the outside of a coax cable ( or probe tip) so getting the fields to go into the PCB and not back up the cable does require as continuous a ground as possible.  GSG probes  are better than GS, and if you look at some of the vertical launch SMA's that are compression mount, they could be considered the ideal  probe with a continuous 360 ground and no possibility for fields to travel on the outside jacket.  Some manufacturers add polyiron at the probe tip to reduce this problem, others try to improve the coax to planar transition of the probe.  4) Clean surfaces are also important to achieve repeatable contacts with the same force and flexing of the probe tip.  Ideally, one should verify each probe landing by looking at the TDR to confirm the impedance profile is the same as that used for calibration.  5)Using a 2-Tier calibration with NIST traceable coaxial calibrations to the end of the cables and then de-embedding of the probes is my preferred method.  This way I always have the NIST traceable calibrated data set, and I can always come back later and improve the probe model for de-embedding if needed.  De-Embedding probes is not easy since the probe is low loss and measured S-Parameters can suffer from passivity and causality problems (usually cable movement).  However, one can also look at creating deconstructed measurement based models in simulation to provide adjustable probe S-parameters to match with a given measurement.  Goodluck with the probing,Heidi BarnesEDA Software for SI/PI ApplicationsKeysight Technology-----
Answered byjeremy.webb 5 years 8 months 7 days
Hi Heidi,Did you mean to say "evanescent waves" when you said "effervescent waves"?Regards,JeremyOn Friday, January 16, 2015,  wrote:> Hi Sherman,> Investing time in set-up and training to use probes can pay off when it> comes to making measurements, but it does require a methodical process for> repeatability.>> 1) Connecting to probes with an instrument tends to leave the connecting> cables unsupported with multiple bends.  This means that high performance> phase stable cables (and new ones that have not yet been damaged) are> needed to avoid errors in the calibration due to phase changes from cable> movement.  Investing in the setup to minimize cable movement and provide> support for the cables can reduce the phase errors which should help with a> cleaner de-embedding.>> 2) Effervescent waves love to travel on the outside of a coax cable ( or> probe tip) so getting the fields to go into the PCB and not back up the> cable does require as continuous a ground as possible.  GSG probes  are> better than GS, and if you look at some of the vertical launch SMA's that> are compression mount, they could be considered the ideal  probe with a> continuous 360 ground and no possibility for fields to travel on the> outside jacket.  Some manufacturers add polyiron at the probe tip to reduce> this problem, others try to improve the coax to planar transition of the> probe.>> 4) Clean surfaces are also important to achieve repeatable contacts with> the same force and flexing of the probe tip.  Ideally, one should verify> each probe landing by looking at the TDR to confirm the impedance profile> is the same as that used for calibration.>> 5)Using a 2-Tier calibration with NIST traceable coaxial calibrations to> the end of the cables and then de-embedding of the probes is my preferred> method.  This way I always have the NIST traceable calibrated data set, and> I can always come back later and improve the probe model for de-embedding> if needed.  De-Embedding probes is not easy since the probe is low loss and> measured S-Parameters can suffer from passivity and causality problems> (usually cable movement).  However, one can also look at creating> deconstructed measurement based models in simulation to provide adjustable> probe S-parameters to match with a given measurement.>> Goodluck with the probing,> Heidi Barnes> EDA Software for SI/PI Applications> Keysight Technology>> -----
Answered byjwwebb 5 years 8 months 7 days
Hi Heidi,Did you mean to say "evanescent waves" when you said "effervescent waves"?Regards,Jeremy> On Jan 16, 2015, at 1:45 PM,  >  wrote:> > Hi Sherman,> Investing time in set-up and training to use probes can pay off when it comes > to making measurements, but it does require a methodical process for > repeatability.> > 1) Connecting to probes with an instrument tends to leave the connecting > cables unsupported with multiple bends.  This means that high performance > phase stable cables (and new ones that have not yet been damaged) are needed > to avoid errors in the calibration due to phase changes from cable movement.  > Investing in the setup to minimize cable movement and provide support for the > cables can reduce the phase errors which should help with a cleaner > de-embedding.> > 2) Effervescent waves love to travel on the outside of a coax cable ( or > probe tip) so getting the fields to go into the PCB and not back up the cable > does require as continuous a ground as possible.  GSG probes  are better than > GS, and if you look at some of the vertical launch SMA's that are compression > mount, they could be considered the ideal  probe with a continuous 360 ground > and no possibility for fields to travel on the outside jacket.  Some > manufacturers add polyiron at the probe tip to reduce this problem, others > try to improve the coax to planar transition of the probe.  > > 4) Clean surfaces are also important to achieve repeatable contacts with the > same force and flexing of the probe tip.  Ideally, one should verify each > probe landing by looking at the TDR to confirm the impedance profile is the > same as that used for calibration.  > > 5)Using a 2-Tier calibration with NIST traceable coaxial calibrations to the > end of the cables and then de-embedding of the probes is my preferred method. >  This way I always have the NIST traceable calibrated data set, and I can > always come back later and improve the probe model for de-embedding if > needed.  De-Embedding probes is not easy since the probe is low loss and > measured S-Parameters can suffer from passivity and causality problems > (usually cable movement).  However, one can also look at creating > deconstructed measurement based models in simulation to provide adjustable > probe S-parameters to match with a given measurement.  > > Goodluck with the probing,> Heidi Barnes> EDA Software for SI/PI Applications> Keysight Technology> > -----