开云体育

Hello Roger,

I try to have zero extension after the calibration planes (at least for
30MHz). So I use SMA connectors with open, closed, load (2-100ohm SMD in
parallel) and through pins, see:

I do of course all my (Excel) calculations in the complex environment. And
I get the same results as the NanoVNA saver program for all values. I
plotted the |Z|, as It provides enough detail (certainly when having 10%
deviation;-).
For my S21 evaluation; my DUTs are normal resistors soldered on centre of
female SMA connectors (I don't care if the leads are long, all is my DUT)
and for comparing I use low frequencies (to minimise stray effects).

I use a slightly extended formula:
ZDUT(s) = 2*(Zsource+Zload)*((1/S21)-1)
Zload is the one measured with through connection to port 2 (S11
reflection).
Zsource can be manually varied in Excel for the evaluation (default 50ohms,
but can be any complex number).
Besie this, I also have Zo which is the reference of pure 50ohm (used in
the calucations/definitions). My Zo calibration load is 49.7ohm using DVM.

[image: afbeelding.png]

I need to measure impedances larger than 3 or 4kohm. as I am looking at
choking devices and I hope they have impedances above 5 or more kohm. And
frequencies upto the end of the HF bands (30MHz).
If I look at my DVM comparison with 1port S11 (reflection) results; even
21kohm provides ok-ish results (within 2%, which is good enough for me).

But I still want to understand why my 2port S21 Zduts are off some 9%.

When doing the 2port measurements, I get (as expected) similar impedance
values for impedance measuring Series S11 (minus Zload) and 1 port S11
reflection measurement. So this (S11) part of the 2port measurements can be
mapped nicely on the 1 port measurements.

Op di 7 jun. 2022 om 01:09 schreef Roger Need via groups.io <sailtamarack=
[email protected]>:

Next I attached the cable to Port2 (CH1) and measured the complex
impedance R +/-jX fro 10 kHz to 250 MHz. The attached plots show that
there is a considerable change in resistance over this range and an
unwanted capacitive reactance.

So in my case the difference is (using the through connector of my
calibration kit and from 330kHz to 30MHz).
[image: afbeelding.png]

My results are comparable to your 6dB attenuator results (I though have
this dip near low freqeuncies, I don't see this in a 50ohm calibration
load, so this is soemthing of my port2, I think...).

What I have not tested is how the impedance of Port1 (Ch0) varies with

frequency. This is not as easy a measurement and I have not done it. The
attenuator "trick" could be used here as well but needs further
investigation.

Owen Duffy mentions a method (). I also
did not do it (I though variate in my Exel to see the results and see if I
can map the NanoVNA measurmeent with the Excel).

What happens with the S21 series method is that as you increase the DUT
resistance you get more error in the S21 gain AND S21 phase. This
translates to larger differences in the expected complex impedance of the
DUT.

But 9%!? I am not able to simulate this in my Excel spreadsheet (except by
lowering Zload or Zsource to some 40ohm)... And would you not see the same
type of error in S11? (while that one is ok-ish).

I believe there are many sources of error such as the following:
- Port1 and Port 2 impedance is not 50 +j0 as pointed out above
- Stray inductance and capacitance of the test jig have a considerable
effect on the magnitude and phase of the S21 measurement. I have tried
several different test jigs and keep trying to improve. The attached photo
shows one I use but still needs work because the S11 reference plane is
too far from the DUT and I get an unwanted phase shift

For now I keep the freqeuncies of interest (for me) low: upto 30MHz.

- To calculate S21 the NanoVNA needs to know the power level generated on
Port 1. I have a hunch that this power level fluctuates as the Return Loss
decreases due to a higher DUT impedance.

Something like that must indeed happen.
The behavoir looks to be similar if the Zload is paralleled (so on port2
side) with an impedance (at least that is what I can see/simulate in my
Exel spreadsheet). But I don't think the port2 impedance does not change
much due to Zdut (I really hope!).
I tried to simulate volttage change of port 1 by varying Zsource (so a
change on port1 side): I can get a Zdut(2pseriesS21) equal to
Zdut(2pseriesS11) when Zsource=40ohms.
But the Zrefl (using 1 port S11 measurement) also will be changed due to
this change in Zsource. While that one was measured ok-ish...
So it is something, IMHO, on the port2 side...

I tried a 10dB at port2 to see the effects on Zdut. After recalibration I
get thee for 2port measurements:
[image: afbeelding.png]

The right bottom graphs compares Zdut with 10dB (red/darkblue) and without
10dB (orange/light blue). No significant change, so still this 9% error
compared to Zdut using S11 reflection measurement...

For more discussion on this topic here are some old post links...

/g/nanovna-users/topic/67738316#7993

/g/nanovna-users/topic/hardware_deficiencies_when/80639862?p=

With care it is possible to get S21 results that compare well with S11
measurements. Here is a post I made earlier on the subject.
/g/nanovna-users/message/24390. But I just use S11
method most of the time for ease of use.

Will digest these links.

Thanks.

All the best,


Victor