Topics

Quick compare with HP 8753C...

Jeff Anderson
 

My NanoVNA finally arrived (took almost 3 weeks). Hurray!

For what it's worth, I decided to do a quick comparison between the NanoVNA and my HP 8753C, looking at the input to my antenna tuner (which has tuned my G5RV antenna at 3.865 MHz to a low SWR).
Attached is an image showing comparison plots of the .S1P files I stored for each device's capture.

Note that as Rho becomes larger, the measurements made from the two VNAs diverge more and more.

At 4 MHz, Z, measured with the 8753C, is 19.061 - j32.701 ohms, while Z measured with the NanoVNA is 20.208 - j33.049 ohms.

Not too bad. That's a variation in Rho (i.e. magnitude of Gamma) on the order of about 2.5% at 4 MHz. But closer to resonance (i.e. smaller rho), the more coincident the plots of the two captures become.

As to why there is a difference at all, I cannot explain (but hey, it's a $50 VNA). I will note that both units were cal'd with the same SOL standards. And repeated cal-and-capture cycles of both the NanoVNA and the 8753C show that the 8753C set of captures has very little divergence between the captures within its set, and the NanoVNA set of captures has very little divergence within its set. But there is a difference between the two sets, as exemplified by the two captures plotted in the attached image.

- Jeff, k6jca

Roger Henderson
 

One idea as to why the divergence:

Were the SOL standards defined in the 8753 in the same way as the NanoVNA?
If you were to define a user cal kit with all of the characterisation data
zeroed out then it just leaves the NanoVNA characterisation data.
I posted a while ago that I think there is a C0 value hardcoded into the
firmware for the open standard. So matching that in the 8753 cal kit
definition would be the idea.

Roger zl4rog

On Wed, 21 Aug 2019 at 10:26, Jeff Anderson <jca1955@...> wrote:

My NanoVNA finally arrived (took almost 3 weeks). Hurray!

For what it's worth, I decided to do a quick comparison between the
NanoVNA and my HP 8753C, looking at the input to my antenna tuner (which
has tuned my G5RV antenna at 3.865 MHz to a low SWR).
Attached is an image showing comparison plots of the .S1P files I stored
for each device's capture.

Note that as Rho becomes larger, the measurements made from the two VNAs
diverge more and more.

At 4 MHz, Z, measured with the 8753C, is 19.061 - j32.701 ohms, while Z
measured with the NanoVNA is 20.208 - j33.049 ohms.

Not too bad. That's a variation in Rho (i.e. magnitude of Gamma) on the
order of about 2.5% at 4 MHz. But closer to resonance (i.e. smaller rho),
the more coincident the plots of the two captures become.

As to why there is a difference at all, I cannot explain (but hey, it's a
$50 VNA). I will note that both units were cal'd with the same SOL
standards. And repeated cal-and-capture cycles of both the NanoVNA and the
8753C show that the 8753C set of captures has very little divergence
between the captures within its set, and the NanoVNA set of captures has
very little divergence within its set. But there is a difference between
the two sets, as exemplified by the two captures plotted in the attached
image.

- Jeff, k6jca



Jeff Anderson
 

Hi Roger,

I'm not sure what you mean by "defined in the 8753", so let me describe what I do...

I don't have calibration information stored within my 8753C. Instead, I calibrate my 8753C after first selecting the frequency range (in this case, 3.8 to 4 MHz), and then I invoke the calibration menu. Its S11 cal option then requires that I attach my short, open, and load standards and capture their data, in sequence, after which I press "done", at which point the cal coefficients are calculated.

I follow a similar procedure for the NanoVNA, using the calibration procedure in the Windows NanaVNA V 1.03 app (which requires two additional steps: isolation and thru).

In other words, calibration for either instrument is assuming that the SOL loads I am using are "perfect." And, although I would be very surprised if they were, for this purpose (of comparison), they should be perfectly adequate. After all, if I do a calibration on either instrument and then measure S11 for each of these three loads, they appear as little dots exactly where one would expect them to appear -- at the far left, center, and far right of the Smith chart.

Please let me know if you think I should be doing this comparison (or cal) a different way.

Thanks,

- Jeff, k6jca

Roger Henderson
 

Hi Jeff,
In my 8753 it will default to one of the inbuilt calibration kit
definitions. I am not in front of it right now, but from memory it does not
(by default) assume the cal kit is perfect. It will use one of the
pre-defined HP cal kits by default.
I would have to define a cal kit with no characterisation data - i.e. no
C0, C1, C2, C3 values, no delay, no loss and with a Zo of 50 ohms for all
of the standards.

Now it is quite possible of course that you have done this, or that at the
frequency range in question the difference is extremely small and is
irrelevant.
Or it could be that it is just relevant enough to affect the traces you
posted, but not enough to be visible on the smith chart at these low
freq's.

As to the NanoVNA, I think the calibration is still being done on the
device and not in the software. The software is not open source so I am not
sure how to determine if it is done on the PC, or on the NanoVNA.
I found this in the NanoVNA firmware source code:
https://github.com/hugen79/NanoVNA-H/blob/master/main.c

Code copied below. The code listed is the calculation of the 'adjusted open
standard' or s11ao. s11aor is the real component and s11aoi is the
imaginary component.
You can see this term: *float c = 50e-15;*
I think that is a C0 term in Farads.

So, if the calibration is being done on the device, then as far as I can
tell, the open standard is _not_ assumed to be perfect as it has a
capacitance adjustment.
How much of an effect that has I don't know and have not checked.

So in a perfect test I would take this small C0 adjustment and enter it in
the - otherwise assumed perfect - cal kit definition in the 8753. Then they
should match.

Anyway, that is where I am coming from.

Roger



static void
eterm_calc_es(void)
{
int i;
for (i = 0; i < sweep_points; i++) {
// z=1/(jwc*z0) = 1/(2*pi*f*c*z0) Note: normalized with Z0
// s11ao = (z-1)/(z+1) = (1-1/z)/(1+1/z) = (1-jwcz0)/(1+jwcz0)
// prepare 1/s11ao for effeiciency
*float c = 50e-15;*
//float c = 1.707e-12;
float z0 = 50;
float z = 6.2832 * frequencies[i] * *c ** z0;
float sq = 1 + z*z;
float s11aor = (1 - z*z) / sq;
float s11aoi = 2*z / sq;

On Wed, 21 Aug 2019 at 11:03, Jeff Anderson <jca1955@...> wrote:

Hi Roger,

I'm not sure what you mean by "defined in the 8753", so let me describe
what I do...

I don't have calibration information stored within my 8753C. Instead, I
calibrate my 8753C after first selecting the frequency range (in this case,
3.8 to 4 MHz), and then I invoke the calibration menu. Its S11 cal option
then requires that I attach my short, open, and load standards and capture
their data, in sequence, after which I press "done", at which point the cal
coefficients are calculated.

I follow a similar procedure for the NanoVNA, using the calibration
procedure in the Windows NanaVNA V 1.03 app (which requires two additional
steps: isolation and thru).

In other words, calibration for either instrument is assuming that the SOL
loads I am using are "perfect." And, although I would be very surprised if
they were, for this purpose (of comparison), they should be perfectly
adequate. After all, if I do a calibration on either instrument and then
measure S11 for each of these three loads, they appear as little dots
exactly where one would expect them to appear -- at the far left, center,
and far right of the Smith chart.

Please let me know if you think I should be doing this comparison (or cal)
a different way.

Thanks,

- Jeff, k6jca



Jeff Anderson
 

Thanks very much for the explanation, Roger. Understood.

I'm usually dealing with frequencies in the HF range, so 50 femtofarads is pretty much "de minimis" for me, and thus its addition (or lack thereof) I'd think would not be a contributor to the divergence I'm measuring in the 4 MHz range.

Per my calculations from the above formulas, at 4 MHz s11aor is 1.000 and s11aoi is 1.2566e-4. I.e. the angle of gamma is about 0.007 degrees, compared to a perfect 0 degrees.

By the way, it's interesting that this factor has been hard-coded into the NanoVNA software. Which, to me, implies that it would be applied to all cal standards, be they the supplied SMA standards, or something else.

- Jeff

Roger Henderson
 

Excellent, that sure is a tiny effect at 4MHz. So safe to rule that out.
However since it is a comparison, would still need to consider what the cal
kit definitions are in your 8753. I expect they are also tiny, but the
question remains.

And yes, I'm fairly sure that there is no code path which avoids applying
that C0 constant in the Nano. It might just be legacy from the original
version as the same code is in the original github repo.

Roger

On Wed, 21 Aug 2019 at 12:12, Jeff Anderson <jca1955@...> wrote:

Thanks very much for the explanation, Roger. Understood.

I'm usually dealing with frequencies in the HF range, so 50 femtofarads is
pretty much "de minimis" for me, and thus its addition (or lack thereof)
I'd think would not be a contributor to the divergence I'm measuring in the
4 MHz range.

Per my calculations from the above formulas, at 4 MHz s11aor is 1.000 and
s11aoi is 1.2566e-4. I.e. the angle of gamma is about 0.007 degrees,
compared to a perfect 0 degrees.

By the way, it's interesting that this factor has been hard-coded into the
NanoVNA software. Which, to me, implies that it would be applied to all
cal standards, be they the supplied SMA standards, or something else.

- Jeff



Dr. David Kirkby from Kirkby Microwave Ltd <drkirkby@...>
 

On Tue, 20 Aug 2019 at 23:26, Jeff Anderson <jca1955@...> wrote:

My NanoVNA finally arrived (took almost 3 weeks). Hurray!
Good

For what it's worth, I decided to do a quick comparison between the NanoVNA
and my HP 8753C, looking at the input to my antenna tuner (which has tuned
my G5RV antenna at 3.865 MHz to a low SWR).
Attached is an image showing comparison plots of the .S1P files I stored
for each device's capture.

Note that as Rho becomes larger, the measurements made from the two VNAs
diverge more and more.

Hardly surprising. Measurements are always going to be best near 50 ohms.
Other impedance measurement techniques like RF-IV work better over a larger
impedance range. But those are restricted to a narrower frequency range,
although well beyond 900 MHz.


I will note that both units were cal'd with the same SOL standards.


Em, what cal standards? Exactly how did you configure your 8753C?

The NanoVNA has a really dumb implementation of calibration kits which
means it assumes ideal text book standards, so it is best calibrated with
the kit supplied, but don’t use the open standard - just leave the
connector open. I should a graph a few weeks ago that indicates that the
open standard supplied causes phase errors greater than just leaving the
cable open.

*If you use a professional calibration kit with the NanoVNA you will just
make the calibration worse than using the cheap kit supplied. *

Unless you know the properties of the calibration kit supplied with the
NanoVNA, and set up the 8753 to have a user kit with those parameters, then
the 8753, will not be calibrated properly. I would suggest that you use

1) The supplied kit with the NanoVNA.
2) A professional calibration kit with the 8753, with the 8753 being
configured properly to use that kit.

Jeff, k6jca


Dave, G8WRB

--
Dr. David Kirkby,

Enrico Dona
 

Hi Jeff,Which SW are you using to plot/compare data? Is it a Matlab script?  Can you make it available?BR Enrico
https://www.hamqth.com/oe7aft

On Wednesday, August 21, 2019, 12:26:16 AM GMT+2, Jeff Anderson <jca1955@...> wrote:

My NanoVNA finally arrived (took almost 3 weeks).  Hurray!

For what it's worth, I decided to do a quick comparison between the NanoVNA and my HP 8753C, looking at the input to my antenna tuner (which has tuned my G5RV antenna at 3.865 MHz to a low SWR).
Attached is an image showing comparison plots of the .S1P files I stored for each device's capture.

Note that as Rho becomes larger, the measurements made from the two VNAs diverge more and more.

At 4 MHz, Z, measured with the 8753C, is 19.061 - j32.701 ohms, while Z measured with the NanoVNA is 20.208 - j33.049 ohms.

Not too bad.  That's a variation in Rho (i.e. magnitude of Gamma) on the order of about 2.5% at 4 MHz.  But closer to resonance (i.e. smaller rho), the more coincident the plots of the two captures become.

As to why there is a difference at all, I cannot explain (but hey, it's a $50 VNA).  I will note that both units were cal'd with the same SOL standards.  And repeated cal-and-capture cycles of both the NanoVNA and the 8753C show that the 8753C set of captures has very little divergence between the captures within its set, and the NanoVNA set of captures has very little divergence within its set.  But there is a difference between the two sets, as exemplified by the two captures plotted in the attached image.

- Jeff, k6jca

N8AUM
 
Edited

On Tue, Aug 20, 2019 at 05:26 PM, Jeff Anderson wrote:
TNX for the post Jeff, I wonder what the results would be if you doubled the span on the 8753 since the nano only has 101 data points while the 8753 has twice that which gives better resolution. I still haven't time to compare the nano between my 8711 or 8712es. For $50 I think its a good deal for something that u can carry in ur shirt pocket lol

73 N8AUM Vidas

Jeff Anderson
 

On Tue, Aug 20, 2019 at 09:29 PM, Enrico Dona wrote:


Is it a Matlab script?
Hi Enrico,

Yes, it is a Matlab script. Note that it also requires the "RF Toolbox".

You can download it from MATLAB Central's page of contributions by Dick Benson (who has posted many excellent tools): https://www.mathworks.com/matlabcentral/profile/authors/6564033-dick-benson

Scroll down the page to his submission titled "S-Parameter Utilities v1.0.1". It contains viewers for both .s1p and .s2p files.

(While you're there, you also might want to take a look at his technique for measuring Q using a VNA -- it's the "A Tool for Measuring High-Q Resonators" submission.. it contains a PDF describing the technique. Very interesting!)

- Jeff

Jeff Anderson
 

Hi Vidas,

I agree, for $50 the NanoVNA is a heck of a tool!

I purposefully set the 8753C to use the same number of points as the NanoVNA (101 points), so both instruments use the same set of frequencies for all 101 measurements (verified by looking at the S1P files). And I kept the frequency span small to minimize the jagged "piecewise-linear" effect one sometimes see on plots, for example when gamma sweeps broadly around the Smith chart.

So one sort of error that might explain what I'm seeing is an error in frequency between the two devices. But if this were the case, I would expect the two overlaying Smith chart curves to be coincident, but with one curve having different start and stop points, but still along the curve path.

- Jeff

Larry Rothman
 

The document referring to the Q measurement is no longer on the Matlab website but can still be found here:
http://www.walkingitaly.com/radio/RADIOSITO/za_fatti/tutorial/q/old/rfqmeas2b.pdf

Jeff Anderson
 

On Tue, Aug 20, 2019 at 05:32 PM, Roger Henderson wrote:

However since it is a comparison, would still need to consider what the cal
kit definitions are in your 8753. I expect they are also tiny, but the
question remains.
And on Tue, Aug 20, 2019 at 06:51 PM, Dr. David Kirkby from Kirkby Microwave Ltd wrote:

Unless you know the properties of the calibration kit supplied with the
NanoVNA, and set up the 8753 to have a user kit with those parameters, then
the 8753, will not be calibrated properly. I would suggest that you use

1) The supplied kit with the NanoVNA.
2) A professional calibration kit with the 8753, with the 8753 being
configured properly to use that kit.
Good morning, Roger and David,

Both of you raise an interesting question, which is "how much does a calibration kit's parameters affect measurements at very low frequencies?"

I don't have a quantitative answer for that, but in my case of measuring at 4 MHz, I believe it to be so minimal as to be visually imperceptible on a full-size (rather than a significantly zoomed-in) Smith chart. Here's my reasoning...

First, HP, in their app note: "Specifying Calibration Standards for the Agilent 8510 Network Analyzer" (https://literature.cdn.keysight.com/litweb/pdf/5956-4352.pdf?id=1000000237:epsg:apn -- a document referenced by the 8753C Network Analyzer Reference manual), state:

"At microwave frequencies however, the magnitude and phase of an “open” are affected by the radiation loss and capacitive “fringing” fields, respectively."

The app note states essentially the same thing, but in terms of residual inductance, for a reference "short."

Note their use of the phrase, "At microwave frequencies". In my opinion, 4 MHz is so far away from "microwave frequencies" as to be essentially DC.

Next, if one looks at the terms of HP's third-order polynomial equations for defining the residual capacitance and inductance of cal kits, the two terms that will have the greatest impact on these definitions, C0 and L0, are in terms of femtofarads and picohenries, respectively.

Anyway, I'm having a (very) difficult time believing that standards-residuals in the femtofarad or picohenry range have any significant effect on my measurements at 4 MHz, where, for my DUT, gamma's phase delta between the 8753C measurement and that of the NanoVNA is 1 degree.

Well, such are my thoughts, which brings me back to the question, "how much does a calibration kit's parameters affect measurements at very low frequencies?". Do either of you (or anyone else) have an answer to this for, say, frequencies in the HF range -- i.e. up to 30 MHz?

Best regards,

- Jeff, k6jca

P.S. David, per your phase-shift measurement of the supplied "open" load. Roger points out that there seems to be an "open" compensation factor of 50 femtofarads written into the NanoVNA code (e.g. the C0 term in the third-order polynomial) -- would this compensate for the additional phase shift you're measuring? And would it therefore be better to cal using the supplied standard, rather than leaving the connector open?

Jeff Anderson
 

On Wed, Aug 21, 2019 at 07:20 AM, Larry Rothman wrote:


The document referring to the Q measurement is no longer on the Matlab website
but can still be found here:
http://www.walkingitaly.com/radio/RADIOSITO/za_fatti/tutorial/q/old/rfqmeas2b.pdf
Thanks, Larry, I'll forward your link to Dick. Maybe he can update the Matlab site with the new link.

By the way, if one can download the zip file for the tool (I think you have to have a Matlab account, but I could be wrong), Dick's "user manual" has some interesting examples and fixtures that he has used.

- Jeff

Roger Henderson
 

Hi Jeff,
Ok, I've finally fired up the 8753 and done a test.

This was done a bit differently to yours, as the calibration is done
offline on a PC. Therefore it eliminates all of the issues around cal kit
definitions on the two devices.
It also eliminates the calibration code on the NanoVNA, there could be a
subtle bug in there.
Conversely your test is more comprehensive and tests more of the NanoVNA
than I have.

Outline of the process I followed:
3MHz - 300MHz sweep on both
101 data points on both
I set the 8753 to 100Hz IF filtering to drop the noise a bit.
Warm up time of about 1 hour for both devices.
NanoVNA was kept flat on a surface. My thinking was that would reduce air
flow due to convection.
NanoVNA was connected to the PC USB port. Again, trying to keep the system
static. Hoping that any noise introduced will not be significant.

Sweep each of the NanoVNA calibration standards and save the s1p files for
each one.
Sweep the reference attenuator from my Kirkby cal kit and save the s1p file.

Repeat the process on the 8753 and save all of the s1p files.

Sweeps were taken as quickly as possible to reduce drift. Time was about 5
minutes.

Using the sci-kit rf library I did two calibrations on the s1p files -
using perfect ideals. One for the 8753 data, and one for the NanoVNA data.
The results and code are in a jupyter notebook here:
https://github.com/hendorog/nanovna_test/blob/master/NanoVNA%20test.ipynb

The results are excellent I think, and show that the NanoVNA hardware and
firmware involved in extracting the data is capable of good performance.

[image: image.png]

Roger

On Thu, 22 Aug 2019 at 03:13, Jeff Anderson <jca1955@...> wrote:

On Tue, Aug 20, 2019 at 05:32 PM, Roger Henderson wrote:

However since it is a comparison, would still need to consider what the
cal
kit definitions are in your 8753. I expect they are also tiny, but the
question remains.
And on Tue, Aug 20, 2019 at 06:51 PM, Dr. David Kirkby from Kirkby
Microwave Ltd wrote:

Unless you know the properties of the calibration kit supplied with the
NanoVNA, and set up the 8753 to have a user kit with those parameters,
then
the 8753, will not be calibrated properly. I would suggest that you use

1) The supplied kit with the NanoVNA.
2) A professional calibration kit with the 8753, with the 8753 being
configured properly to use that kit.
Good morning, Roger and David,

Both of you raise an interesting question, which is "how much does a
calibration kit's parameters affect measurements at very low frequencies?"

I don't have a quantitative answer for that, but in my case of measuring
at 4 MHz, I believe it to be so minimal as to be visually imperceptible on
a full-size (rather than a significantly zoomed-in) Smith chart. Here's my
reasoning...

First, HP, in their app note: "Specifying Calibration Standards for the
Agilent 8510 Network Analyzer" (
https://literature.cdn.keysight.com/litweb/pdf/5956-4352.pdf?id=1000000237:epsg:apn
-- a document referenced by the 8753C Network Analyzer Reference manual),
state:

"At microwave frequencies however, the magnitude and phase of an “open”
are affected by the radiation loss and capacitive “fringing” fields,
respectively."

The app note states essentially the same thing, but in terms of residual
inductance, for a reference "short."

Note their use of the phrase, "At microwave frequencies". In my opinion,
4 MHz is so far away from "microwave frequencies" as to be essentially DC.

Next, if one looks at the terms of HP's third-order polynomial equations
for defining the residual capacitance and inductance of cal kits, the two
terms that will have the greatest impact on these definitions, C0 and L0,
are in terms of femtofarads and picohenries, respectively.

Anyway, I'm having a (very) difficult time believing that
standards-residuals in the femtofarad or picohenry range have any
significant effect on my measurements at 4 MHz, where, for my DUT, gamma's
phase delta between the 8753C measurement and that of the NanoVNA is 1
degree.

Well, such are my thoughts, which brings me back to the question, "how
much does a calibration kit's parameters affect measurements at very low
frequencies?". Do either of you (or anyone else) have an answer to this
for, say, frequencies in the HF range -- i.e. up to 30 MHz?

Best regards,

- Jeff, k6jca

P.S. David, per your phase-shift measurement of the supplied "open" load.
Roger points out that there seems to be an "open" compensation factor of 50
femtofarads written into the NanoVNA code (e.g. the C0 term in the
third-order polynomial) -- would this compensate for the additional phase
shift you're measuring? And would it therefore be better to cal using the
supplied standard, rather than leaving the connector open?



Stuart Landau
 

That is amazing correlation for the devices.
Stuart K6YAZ

-----Original Message-----
From: Roger Henderson <hendorog@...>
To: nanovna-users <nanovna-users@groups.io>
Sent: Wed, Aug 21, 2019 2:32 pm
Subject: Re: [nanovna-users] Quick compare with HP 8753C...

Hi Jeff,
Ok, I've finally fired up the 8753 and done a test.

This was done a bit differently to yours, as the calibration is done
offline on a PC. Therefore it eliminates all of the issues around cal kit
definitions on the two devices.
It also eliminates the calibration code on the NanoVNA, there could be a
subtle bug in there.
Conversely your test is more comprehensive and tests more of the NanoVNA
than I have.

Outline of the process I followed:
3MHz - 300MHz sweep on both
101 data points on both
I set the 8753 to 100Hz IF filtering to drop the noise a bit.
Warm up time of about 1 hour for both devices.
NanoVNA was kept flat on a surface. My thinking was that would reduce air
flow due to convection.
NanoVNA was connected to the PC USB port. Again, trying to keep the system
static. Hoping that any noise introduced will not be significant.

Sweep each of the NanoVNA calibration standards and save the s1p files for
each one.
Sweep the reference attenuator from my Kirkby cal kit and save the s1p file.

Repeat the process on the 8753 and save all of the s1p files.

Sweeps were taken as quickly as possible to reduce drift. Time was about 5
minutes.

Using the sci-kit rf library I did two calibrations on the s1p files -
using perfect ideals. One for the 8753 data, and one for the NanoVNA data.
The results and code are in a jupyter notebook here:
https://github.com/hendorog/nanovna_test/blob/master/NanoVNA%20test.ipynb

The results are excellent I think, and show that the NanoVNA hardware and
firmware involved in extracting the data is capable of good performance.

[image: image.png]

Roger


On Thu, 22 Aug 2019 at 03:13, Jeff Anderson <jca1955@...> wrote:

On Tue, Aug 20, 2019 at 05:32 PM, Roger Henderson wrote:

However since it is a comparison, would still need to consider what the
cal
kit definitions are in your 8753. I expect they are also tiny, but the
question remains.
And on Tue, Aug 20, 2019 at 06:51 PM, Dr. David Kirkby from Kirkby
Microwave Ltd wrote:

Unless you know the properties of the calibration kit supplied with the
NanoVNA, and set up the 8753 to have a user kit with those parameters,
then
the 8753, will not be calibrated properly. I would suggest that you use

1) The supplied kit with the NanoVNA.
2) A professional calibration kit with the 8753, with the 8753 being
configured properly to use that kit.
Good morning, Roger and David,

Both of you raise an interesting question, which is "how much does a
calibration kit's parameters affect measurements at very low frequencies?"

I don't have a quantitative answer for that, but in my case of measuring
at 4 MHz, I believe it to be so minimal as to be visually imperceptible on
a full-size (rather than a significantly zoomed-in) Smith chart.  Here's my
reasoning...

First, HP, in their app note: "Specifying Calibration Standards for the
Agilent 8510 Network Analyzer" (
https://literature.cdn.keysight.com/litweb/pdf/5956-4352.pdf?id=1000000237:epsg:apn
-- a document referenced by the 8753C Network Analyzer Reference manual),
state:

"At microwave frequencies however, the magnitude and phase of an “open”
are affected by the radiation loss and capacitive “fringing” fields,
respectively."

The app note states essentially the same thing, but in terms of residual
inductance, for a reference "short."

Note their use of the phrase, "At microwave frequencies".  In my opinion,
4 MHz is so far away from "microwave frequencies" as to be essentially DC.

Next, if one looks at the terms of HP's third-order polynomial equations
for defining the residual capacitance and inductance of cal kits, the two
terms that will have the greatest impact on these definitions, C0 and L0,
are in terms of femtofarads and picohenries, respectively.

Anyway, I'm having a (very) difficult time believing that
standards-residuals in the femtofarad or picohenry range have any
significant effect on my measurements at 4 MHz, where, for my DUT,  gamma's
phase delta between the 8753C measurement and that of the NanoVNA is 1
degree.

Well, such are my thoughts, which brings me back to the question, "how
much does a calibration kit's parameters affect measurements at very low
frequencies?".  Do either of you (or anyone else) have an answer to this
for, say, frequencies in the HF range -- i.e. up to 30 MHz?

Best regards,

- Jeff, k6jca

P.S. David, per your phase-shift measurement of the supplied "open" load.
Roger points out that there seems to be an "open" compensation factor of 50
femtofarads written into the NanoVNA code (e.g. the C0 term in the
third-order polynomial) -- would this compensate for the additional phase
shift you're measuring?  And would it therefore be better to cal using the
supplied standard, rather than leaving the connector open?



Jeff Anderson
 

Roger, excellent!

Of the s1p files on your github site, which two files are the corrected attenuator files for the 8753 and for the NanoVNA measurements? I thought I would plot them out on a Smith chart. (I'm assuming they are "nano attenuator2.s1p" and "kirby attenuator.s1p", but I'd like to be sure).

Thanks!

- Jeff

Roger Henderson
 

Ah thanks. Yes I do have a double up there. Both of those you mentioned are
NanoVNA measurements, but done at different times.

The 8753 measured attenuator is:
8753 attenuator.S1P
<https://github.com/hendorog/nanovna_test/blob/master/data/measured/8753%20attenuator.S1P>

The NanoVNA measured attenuator to use is:
nano attenuator2.s1p
<https://github.com/hendorog/nanovna_test/blob/master/data/measured/nano%20attenuator2.s1p>


Those are the raw measurements though, so they will look very different
when you chart them.
To extract the 'post-calibration' I'll need to change the code to output
s1p files for the error corrected traces.

Cheers,
Roger

On Thu, 22 Aug 2019 at 10:01, Jeff Anderson <jca1955@...> wrote:

Roger, excellent!

Of the s1p files on your github site, which two files are the corrected
attenuator files for the 8753 and for the NanoVNA measurements? I thought
I would plot them out on a Smith chart. (I'm assuming they are "nano
attenuator2.s1p" and "kirby attenuator.s1p", but I'd like to be sure).

Thanks!

- Jeff



Roger Henderson
 

I've added the output files, thanks to scikit-rf. Hopefully the

The 8753 error corrected output
8753_attenuator_output.s1p
<https://github.com/hendorog/nanovna_test/blob/master/data/output/8753_attenuator_output.s1p>

The NanoVNA error corrected output
nano_attenuator_output.s1p
<https://github.com/hendorog/nanovna_test/blob/master/data/output/nano_attenuator_output.s1p>


Directory link
https://github.com/hendorog/nanovna_test/tree/master/data/output

Roger

On Thu, 22 Aug 2019 at 10:22, Roger Henderson <hendorog@...> wrote:

Ah thanks. Yes I do have a double up there. Both of those you mentioned
are NanoVNA measurements, but done at different times.

The 8753 measured attenuator is:
8753 attenuator.S1P
<https://github.com/hendorog/nanovna_test/blob/master/data/measured/8753%20attenuator.S1P>

The NanoVNA measured attenuator to use is:
nano attenuator2.s1p
<https://github.com/hendorog/nanovna_test/blob/master/data/measured/nano%20attenuator2.s1p>


Those are the raw measurements though, so they will look very different
when you chart them.
To extract the 'post-calibration' I'll need to change the code to output
s1p files for the error corrected traces.

Cheers,
Roger

On Thu, 22 Aug 2019 at 10:01, Jeff Anderson <jca1955@...> wrote:

Roger, excellent!

Of the s1p files on your github site, which two files are the corrected
attenuator files for the 8753 and for the NanoVNA measurements? I thought
I would plot them out on a Smith chart. (I'm assuming they are "nano
attenuator2.s1p" and "kirby attenuator.s1p", but I'd like to be sure).

Thanks!

- Jeff



Jeff Anderson
 

Thanks, Roger. Apologies for the extra work.

Smith chart plots are attached. One is "full-view", the other is zoomed-in.

Even zoomed-in, I cannot tell any difference between the two. In fact, they are so identical that I at first suspected the files might accidentally have been the same, but they are not.

- Jeff