2port measurement determining Zdut with S11 and S21
|
Hello Roger,
I thought that the method I used now for the resistors was working. And it seems to work at least for these 4 resistors (330, 985, 5160 and 21800);-) But I now did a 1:1 Guanella choke as dut and it seems to behave differently in a 2 port measurement... Here are the 2 port measurements NanoVNA Saver of my 1:1 Guanella choke (I also attached the 11Guanella-corebraid-1s21.s2p file): [image: afbeelding.png] As you can see the S21 R+jX has totally different form (e.g. peaks) than S11 R+jX. If I use the formula of (Page 10) : [image: afbeelding.png] I can determine with the 2 port measurements Zdut (Zx) in two ways through S11 and S21 2 port measurement. Z2pseries(S11)=(Zsource+Zload)*S11/(1-S11) Z2pseries(S21)=(Zource+Zload)*(1-S21)/S21 [image: afbeelding.png]The green curve should be close to the red (blue dotted or orange dotted) ones (as when using my method for the resistors), but with the choke they are totally different. I must be doing something (obviously?) wrong! But what? Any help is welcome. All the best, Victor |
|
Victor,
I am not sure how you are measuring the DUT using S11 reflection and S21 series methods. You have to measure them with the correct reference plane calibration and this usually requires two different test jigs. What are you using for S11 and S21 test jigs? Attached is one I use for S11 up to 30 MHz. I tack solder the DUT as shown. My example uses an inductor but you can solder the leads to your choke. The second one I built into a box and it works well too. I posted my S21 series jig in the last topic Roger |
|
Victor,
Let me preface my comments by saying that I have built several test jigs for measuring impedance using the S21 series method and tried several ways to compensate for errors. None have been very successful and I now just use the S11 reflection method which works up to several thousand ohms of complex impedance with a decent test jig. I believe the NanoVNA-H4 is just not the right tool for the job for several reasons that I have mentioned before.... - the source and termination impedances are not 50 + j0 ohms and vary with frequency - the current firmware and PC programs only use 6 term error correction not 12 term Others have found similar issues using the S21 series method using the NanoVNA and even with higher performance HP VNA equipment. Here are some interesting links. The K6JCA experiments find differences similar to what you have measured. In my experiments I found that parasitic capacitance had considerable affect on the S21 measurements and I tried several different test jigs to see what happened. The issue of where to establish the reference plane took some consideration. S21 Gain and S21 phase determine the complex impedance of the DUT and even with some type of source (Port1) and termination (Port2) correction the phase was often incorrect. Phase measurement was the problem and in my tests was due to an offset delay across the jig. Another issue I found was that for some reason high DUT impedance resulted in common mode current on the surface of the coax cables connecting the NanoVNA to the test jig. I tried snap-on ferrites and the results changed. You mentioned that you could do a S11 reflection and S21 series measurement at the same time. You will get different results when the DUT is in series between Port 1 and Port2 compared to when it is across Port 1 alone. Best case is that it will be 50 ohms higher. TIn reality the complex impedance measured when in series is the DUT plus the input impedance of the cable attached to Port 2 (which will deviate from 50 ohms due to Port 2 impedance not being 50 ohms) and any parasitic inductance or capacitance between the DUT and the Port 2 cable. I wish you good luck in your quest and hope you find a method that works for you and that you find a solution that eluded me. Roger |
|
On Sat, Jun 11, 2022 at 11:08 AM, Roger Need wrote:
In my experiments I found that parasitic capacitance had considerable affectHere is one of theS11/S21 test jigs I tried. It could use improvement but I don't have the time right now to continue with this project. Roger |
|
Thanks Roger and Owen,
toggle quoted message
Show quoted text
Remember that the S11 formula from and NanoVNA Saver (and my own Excel spreadsheets) already takes away the 50ohm (Zload) of port 2. The S11 2-port series through measurement (also described by k6jca, NaonoVNA Saver and earlier link), is always close to the S11 1-port measurement. My calibration plane is at the end of the middle/earth tips of the PCB SMA connector (also my calibration kit is at these tips: ). I solder the DUT directly to these pins aka calibration plane (so no real extension of the plane). This all to minimize variability of reference plane capacitance/etc. In my case Zload of port2 is very close to on average 50.8-0.1i ohm (300kHz to 30MHz), which is based on a calibration load of 49.7 ohm (using DVM, accuracy 1.5%). As my Calibration-load is a little smaller than 50~0iohm, the resistances measured by the NanoVNA will be a little too high. So my Zload comes closer to 50ohm. Anyway I compensate S21 for that Zload. In my case Zsource is significantly different (on average 43.6+3.2iohm) from 50ohm. So I compensate S21 for that (using Owen's method: ), S11 is not affected by this Zsource as Zsource is calibrated 'out' by Calibration-load. I put the DUT on a styrofoam underground (to minimize the effects of my wooden table top's dielectricum). I also put a clip-on ferrite on the port1 or port2 wire, or shortened the port2 cable (from 16cm to 2cm), or put it as far away from other equipment (the NanoVNA is though always some 16cm nearby: IMG_1643.jpg). All these changes don't really have an effect on 2port series through measurements of S11 or S21. Only slightly at the end of my measurement range (30MHz). The DUT (in 2 port series through) has *certainly* hand effect (aka adding stray capacitance), so I need to stay way;-). This hand effect is most prominent on the S21 (not much on the S11) 2-port series through. If I change the offset delay, this has almost no effect on S21 2port series through (as expected), but on the S11 2port series through (it shifts the R and X peaks to higher frequencies). If I put the offset delay to around -56ps equivalent to 7.5mm, the R&X peaks of S11 map the R&X peaks of S21. The below screengrab is with a offset delay of -56ps: [image: afbeelding.png] I still find it strange that while the S21 changes so much the S11 stays quite constant. If I understand k6jca ( ), he (as some other experts) likes S21 the best (also more stable than S11...) for high impedances (like for chokes at say 10kohm). This makes it even more confusing as others now tell me S11 is better;-) It would be nice to use Benson's Y method; but that needs a S11, S21, S12, S22 setup (and with that a 12-term calibration). This is not available with NanoVNA, so that does not help as advice. Perhaps in a few years prices will go down for such equipment;-) I will digest k6jca pages in more detail (I recognise some of what he is telling, looking at my own experiences). Thanks for all the feedback provided. If I gain more experience, I will communicate over the group. All the best, Victor Op za 11 jun. 2022 om 20:08 schreef Roger Need via groups.io <sailtamarack= [email protected]>: Others have found similar issues using the S21 series method using the |
|
I forgot to add that the NanoVNA measured Impedance (based on the S21 (or
toggle quoted message
Show quoted text
S11) 2port series through): [image: afbeelding.png] matches quite nicely the theoretical simulation of a choke: [image: afbeelding.png] <the theoretical simulation got one parameter (stray capacitance of the choke, not of the testing gear!) which was derived by matching the peak in the simulation with the peak in the measured S21 curve> I think I am more or less understanding what/how is measured with the NanoVNA. Will try to make a web page out of that. All the best, Victor Op za 11 jun. 2022 om 08:32 schreef Victor Reijs via groups.io <pe1atn.victor.reijs@...>: Hello Roger, |
Roger Need
