开云体育

It has been stated, by multiple members of this group, that due to hardware
deficiencies (port 2 is not a perfect 50Ω) and lack of 12 term error
correction,
the "S21 method" can't be used to measure high impedance reliably.
I decided to test that and see how inaccurate is it really.

On the first graph the red and green traces are the measured values (R&Z)
and the blue and black traces are a simulation of a 10K resistor with 50fF
of parasitic
capacitance (figure from some Vishay paper).

[image: image.png]

The DUT is a resistor of unknown origin from China that measures (at DC)
9.995K ± 10Ω.
The VNA (Nano-H4) was calibrated using the DIY "standards", no compensation
of the test
jig during calibration, offset delay of -4.5ps was used during measurement.
Test jig was made from cheap SMA connectors from AliExpress, cables (RG405)
also.

[image: 2021-02-14 19.56.05.jpg]

Draw your own conclusions.

Hi,
which firmware have you used? I found out that the newest cloned firmwares for the SAA2 has a worse quality than older. The firmwares until Sept. 2020 are OK. Later versions are not usable. I posted it but no of the developers are interested in it. Unfortunately only one user has answered.

--
Klaus, DL5KV

Hello Klaus,
how about the ojisan firmware.
I have added a patch (2.8 inch LCD only) to make a screenshot with a Python program,
see:

73, Rudi DL5FA

This is on H4 not SAA.
There is no such thing as "cloned firmware" so not sure
what you are refering to.

On Sun, Feb 14, 2021 at 10:56 PM, Klaus W?rner wrote:

Hi,
which firmware have you used? I found out that the newest cloned firmwares for
the SAA2 has a worse quality than older. The firmwares until Sept. 2020 are
OK. Later versions are not usable. I posted it but no of the developers are
interested in it. Unfortunately only one user has answered.

Its strange, my V2 and modded V2 to V2plus work fine.

Last time i check impedance measure in 2-30Mhz range vs Agilent Impedance Analyzer E4990A (use H4 and V2Plus and V2Plus4)
Measure laboratory equivalent of HF antenna.
Exist small difference in measure small R (less 1Om) and big (bigger then 10k)
V2plus4 show best perfomance (it can correct measure 10k Om)
H4 show only 8k in this case

But all other measure good (phase, Aplitude, Resonances)

Hi DiSlord,

The info about updating a V2.2 to V2plus you were so kind send me recently shows a hardware change of 3 components only, one resistor and 2 capacitors.

If I do understand well, this update makes it possible to increase the scan speed? in the firmware and does not enhance the measuring results at all.

If so, this update is of less importance to me.

I also learned that there are more hardware differences between the two versions, do you know what these are and if it is possible to implement them in the V2.2 ?

I have been looking around, but sofar did not succeed to find the schematic of the V2plus.

Thanks for your time and best regards,

Jos

V2 is open source, the later versions are not.

On Sun, Feb 14, 2021 at 12:55 PM, Dragan Milivojevic wrote:

It has been stated, by multiple members of this group, that due to hardware
deficiencies (port 2 is not a perfect 50Ω) and lack of 12 term error
correction,
the "S21 method" can't be used to measure high impedance reliably.

Yes this is true. The "S21 series method" is a way of calculating the complex impedance (R+jX) of a device under test (DUT). Unfortunately it does not work on a NanoVNA-H or -H4 which are 2 port uni-directional analyzers for the reasons you outlined in your post.

If you try it you will find that you can calculate the magnitude of a high impedance |Z| DUT fairly well but the phase angle is not correct except at very low frequencies. You need the correct phase angle in order to convert from polar (Z&angle) to polar (R+jX) format.

Users interested in trying this should use a SMD resistor as the DUT which should have a S21 phase angle very close to zero. You will only get this at very low frequencies. A small capacitor as the DUT should have a phase angle near 90 degrees but this will be incorrect as you increase frequency. I have seen this on several test jigs that I have built.

If users want to try these measurements be very careful with parasitic capacitance on the test fixture. The capacitance across the terminals where you connect the DUT has to be well under a pF or the S21 loss in dB will be incorrect.

Roger

Calibrated with "0 length through", test jig connected to port2 directly:

X:

[image: image.png]

R:

[image: image.png]

Z:

[image: image.png]



On Tue, 16 Feb 2021 at 01:38, Roger Need via groups.io <sailtamarack=
[email protected]> wrote:

On Sun, Feb 14, 2021 at 12:55 PM, Dragan Milivojevic wrote:

It has been stated, by multiple members of this group, that due to

hardware

deficiencies (port 2 is not a perfect 50Ω) and lack of 12 term error
correction,
the "S21 method" can't be used to measure high impedance reliably.

Yes this is true. The "S21 series method" is a way of calculating the
complex impedance (R+jX) of a device under test (DUT). Unfortunately it
does not work on a NanoVNA-H or -H4 which are 2 port uni-directional
analyzers for the reasons you outlined in your post.

If you try it you will find that you can calculate the magnitude of a high
impedance |Z| DUT fairly well but the phase angle is not correct except at
very low frequencies. You need the correct phase angle in order to convert
from polar (Z&angle) to polar (R+jX) format.

Users interested in trying this should use a SMD resistor as the DUT which
should have a S21 phase angle very close to zero. You will only get this
at very low frequencies. A small capacitor as the DUT should have a phase
angle near 90 degrees but this will be incorrect as you increase frequency.
I have seen this on several test jigs that I have built.

If users want to try these measurements be very careful with parasitic
capacitance on the test fixture. The capacitance across the terminals
where you connect the DUT has to be well under a pF or the S21 loss in dB
will be incorrect.

Roger

I'm not sure, but I think there are clones for H4 firmware, too. This ist only a notice of me. Maybe this problem doesn't exist on the V1. I haven't updated the firmware on V1 for a long time.

--
Klaus, DL5KV

It's a cool firmware. But I think we have to discuss this in an other group. It's not for V1.

--
Klaus, DL5KV

On Mon, Feb 15, 2021 at 06:43 PM, Dragan Milivojevic wrote:

Calibrated with "0 length through", test jig connected to port2 directly:

You did not provide any detail in your post about your test setup or how you derived R, X and Z from the S21 measurement. I do not recognize the plot formats that you are using. Also you did not describe what component you are measuring.

For those that are not familiar with the S21 series method here is an introductory paper.


Keysight gives a derivation of the conversion formula. (note S21 is a complex number)


If we are to have a discussion on this topic please let the group know a bit about what you are doing...
- What VNA is being used
- A description of the test jig
- What software program or method you are using to convert S21 to complex impedance (R,X)
- What component you are measuring. I suggest a 1K SMD resistor of around 1000 ohms because it is easy to see measurement errors and compare results. Another is a 10 pF SMD capacitor or very short lead ceramic type.

- When I measure (with a NanoVNA-H4) a 1K resistor on a microstrip with 50 ohm impedance my S21 phase starts at 0 but goes to 17 degrees by 100 MHz. . It should stay close to 0 degrees for a component that has minimal inductance.

Roger

Firmware clone, especially when it is an open source project
is a meaningless term. Hope that clears things up.

This user group is for both versions (I believe) but a new topic would
be appreciated.

Most of it is in the initial post, except for software (SimSmith).
Plot commands for SimSmith were:

Plot("R1", S1.z.R, "Ω", "y1");
Plot("X1", S1.z.X, "Ω", "y2");
Plot("Z1", S1.z.M, "Ω", "y1");

I have attached the simsmith circuit and the s2p file if someone wants to
play with it.
The second set of measurements were done with the test jig directly at
port2, in the first
test the jig was connected via 30cm of RG405. I was lazy, could not be
bothered to
solder a jumper on the jig so I did not calibrate out the test jig, just
used a -2ps offset.


On Tue, 16 Feb 2021 at 18:16, Roger Need via groups.io <sailtamarack=
[email protected]> wrote:

On Mon, Feb 15, 2021 at 06:43 PM, Dragan Milivojevic wrote:

Calibrated with "0 length through", test jig connected to port2 directly:

You did not provide any detail in your post about your test setup or how
you derived R, X and Z from the S21 measurement. I do not recognize the
plot formats that you are using. Also you did not describe what component
you are measuring.

For those that are not familiar with the S21 series method here is an
introductory paper.


Keysight gives a derivation of the conversion formula. (note S21 is a
complex number)



If we are to have a discussion on this topic please let the group know a
bit about what you are doing...
- What VNA is being used
- A description of the test jig
- What software program or method you are using to convert S21 to complex
impedance (R,X)
- What component you are measuring. I suggest a 1K SMD resistor of around
1000 ohms because it is easy to see measurement errors and compare
results. Another is a 10 pF SMD capacitor or very short lead ceramic type.

- When I measure (with a NanoVNA-H4) a 1K resistor on a microstrip with 50
ohm impedance my S21 phase starts at 0 but goes to 17 degrees by 100 MHz.
. It should stay close to 0 degrees for a component that has minimal
inductance.

Roger

Dragan,

Thank you for clarifying what you did on your tests. I took a look at the .s2p file that you posted and see that you get the same large S21 phase shift that I got in my tests. I posted my plot of your S21 data below. For a tiny SMD resistor like the one you used the phase shift should be minimal (at least on the lower frequencies.). The S21 magnitude loss is very close to what it should be for a 10K resistor but that phase shift will give a lower R value and indicate reactance that isn't there if I understand this correctly. Do you agree?

I also did a conversion from serial to parallel impedance and your S11 measurement was surprising close to 10K ohms. For a VNA designed for 50 ohm system impedance this was very good. Plot attached.

Roger

I guess that depends on what you consider as minimal
and if you are modeling an ideal resistor or a more
realistic one.

For example: on the first image is the plot of voltage
angle at the load (port 2 in this case), calculated from the
S21 file if the load is an ideal 50Ω.
The black trace is for a model of a 10K resistor with 50fF
of capacitance (reasonable value, according to Vishay for a
0805 capacitor) also into an ideal 50Ω.

[image: image.png]

The second image shows the same plot but this
time instead of an ideal 50Ω termination at port2,
I used the measured values for port2.
Not much of a change.

[image: image.png]

Last image is the port2 as measured, on a Smith chart (zoomed in quite a
bit).

[image: image.png]


Keep in mind that the calibration was done with
DIY "standards" and jig all made from cheap SMA
connectors from AliExpress, where the center
pin moves if you tighten the connectors enough.
DYT also from China with unknown properties.
Quite a sloopy setup.








On Tue, 16 Feb 2021 at 20:47, Roger Need via groups.io <sailtamarack=
[email protected]> wrote:

Dragan,

Thank you for clarifying what you did on your tests. I took a look at the
.s2p file that you posted and see that you get the same large S21 phase
shift that I got in my tests. I posted my plot of your S21 data below.
For a tiny SMD resistor like the one you used the phase shift should be
minimal (at least on the lower frequencies.). The S21 magnitude loss is
very close to what it should be for a 10K resistor but that phase shift
will give a lower R value and indicate reactance that isn't there if I
understand this correctly. Do you agree?

I also did a conversion from serial to parallel impedance and your S11
measurement was surprising close to 10K ohms. For a VNA designed for 50
ohm system impedance this was very good. Plot attached.

Roger

This just became a bit more interesting.
It seems that there is a discrepancy in how
SimSmith calculates the angle depending on
the circuit block used. For an S block it shows
the same curve as in NanoVNA Saver and also
in Rogers graph (green trace).
But for a impedance block angle is different (red trace)
and much closer to the modeled resistor (blue trace).

[image: image.png]



On Wed, 17 Feb 2021 at 03:52, Dragan Milivojevic via groups.io
<d.milivojevic@...> wrote:

I guess that depends on what you consider as minimal
and if you are modeling an ideal resistor or a more
realistic one.