It was Take Your nanoVNA to Work day here, so I did:

https://ae5x.blogspot.com/2021/07/1-to-4-ghz-nanovna-2-plus4-vs-keysight.html

73,
John AE5X

On 27/07/2021 23:10, John AE5X wrote:
> It was /Take Your nanoVNA to Work/ day here, so I did:
>
> https://ae5x.blogspot.com/2021/07/1-to-4-ghz-nanovna-2-plus4-vs-keysight.html
> <https://ae5x.blogspot.com/2021/07/1-to-4-ghz-nanovna-2-plus4-vs-keysight.html>
>
> 73,
> John AE5X

Thanks for that, John, very encouraging!

73,
David GM8ARV
--
SatSignal Software - Quality software for you
Web: https://www.satsignal.eu
Email: david-taylor@blueyonder.co.uk
Twitter: @gm8arv

Hello,

This comparison is very ineresting but I think there is some room to improve the measurements.

First I noticed that the adapters (SMA(m)-N(f), N(m)-N(m), etc.) used with NanoVNA and Fieldfox are different. They look also different w.r.t the ones you have used during calibration (with the N(m) Calkit). These adapters change the response of the DUT of a few tents of dB, so you should always use the same adapters on both VNAs, For example if you calibrate using a N(m) calkit, the fieldfox has already the right port N(f) while you need to add a SMA(m)-N(f) adapter to the NanoVNA. In this way both VNA are calibrated to a N(f) reference plane. From this point on, for a right comparison always use the same additional adapters that you need to connect the DUTs on both VNAs. Never remove the SMA(m)-N(f) used on NanoVNA during Calibration.

Second, you use NanoVNA-saver for measuring S21 on NanoVNA. Unfortunately NanoVNA-saver is inaccurate in measuring S21 since it does not implement Ehnanced Response algorithm. This leads to up to 0.5 dB inaccuracy on S21 at microwave frequencies. To improve the comparison you need to compare the Fieldfox screen (or saved tracks) with the NanoVNA screen (of course, always keeping enabled on Nano the Ehnanced Response Calibration). Also the Filedfox should be correctly configured by always performing a full 2 ports calibration.

Last point: since the NanonVNA-saver does not implement Enhanced Response, it is obvious that S21 measurements are independent by the SOL calibration kit you use (the SOL kit only calibrate the S11 measurements). The NanoVNA-saver calibrates S21 by using only the thru measurement (and only using the S21 of thru measurement) so you will always obtain the same S21 results by using any SOL calkits (until they share the same thru). To really compare the quality of two SOL calkit, you need to measure also the S11 (Mod and Phase) of the DUTs,

Best regards, Marco.

Thanks for the reply, David & Marco.

Marco, people a lot more knowledgeable than me know that the terminations used make a difference in the results obtained with a VNA. Let's forget the nanoVNA for a minute...here's what I want to do and maybe you (or others) can suggest the best way to go about it.

I want to know how much of a difference the quality of the terminations make, preferably at higher frequencies. Considering only S11 and S21 parameters, what would be the best way to see these differences using the FieldFox calibrated first with Agilent terminations and then with the cheap SMA terminations that come with the nanos?

As you mentioned, we'll have to keep in mind that one set of standards has N connectors, the other has SMA.

Using these 2 sets of standards and the FF at say, 3-4 GHz, how can I make an actual measurement that would best illustrate the differences between the standards?

Hello,

as far as I know, it is very difficult to compare calibration standards having different connectors.

Let's say you have an Agilent Calkit 85032B/E. This calkit is defined by a model with parameters that you can see at http://na.support.keysight.com/pna/caldefs/85032BE.htm. When you calibrate the FieldFox, you specify the usage of Calkit 85032B/E. The FieldFox will automatically load the right Calkit parameters (preloaded in its Firmware) and the Open Short and Load standards will be used in the calibration measurements considering their models.

When you want to use the NanoVNA calkit, this calkit have no paramenters. So you must load a "Dummy" calkit model that will use standards (ideal Open, Short, Load). Now, up to 4 GHz this is a rough approximation but maybe still acceptable. The problem worsen a lot if you put an adapter in front of this calkit to convert them in Type-N: the apdater has a delay and a return loss that is not zero and this makes the ideal assumption much less acceptable.

I think the only solution is to procure a professional HP/Agilent/Keysight 3.5 or 2.92 calkit and then compare the calibration made with this calkit (using the right calkit model) with the calibration obtained with NanoVNA calkit (using an ideal model). In this case the comparison can be done measuring S11 and S21 (Mod and Phase) of different DUTs. Use DUTs having high and low return loss and high and low tranfer loss just to probe all possible cases.

Ciao, Marco.

Ciao Marco, and group,

Thanks for your post.

It would be helpful to understand starting from which frequencies (a
"ballpark" figure is good enough)
these discrepancies in calibration do make a significant difference.

E.g. I am using my 2 nanoVNAs (early model SMA and SAA-2N 4") only up to
abt. 900 MHz.

Mostly for VSWR checking and some filter response measurements (various).

73, Klaus, DK3QN


Am 02.08.21 um 17:22 schrieb mce66:

> Hello,
>
> as far as I know, it is very difficult to compare calibration
> standards having different connectors.
>
> Let's say you have an Agilent Calkit 85032B/E. This calkit is defined
> by a model with parameters that you can see at
> http://na.support.keysight.com/pna/caldefs/85032BE.htm
> <http://na.support.keysight.com/pna/caldefs/85032BE.htm>. When you
> calibrate the FieldFox, you specify the usage of Calkit 85032B/E. The
> FieldFox will automatically load the right Calkit parameters
> (preloaded in its Firmware) and the Open Short and Load standards will
> be used in the calibration measurements considering their models.
>
> When you want to use the NanoVNA calkit, this calkit have no
> paramenters. So you must load a "Dummy" calkit model that will use
> standards (ideal Open, Short, Load). Now, up to 4 GHz this is a rough
> approximation but maybe still acceptable. The problem worsen a lot if
> you put an adapter in front of this calkit to convert them in Type-N:
> the apdater has a delay and a return loss that is not zero and this
> makes the ideal assumption much less acceptable.
>
> I think the only solution is to procure a professional
> HP/Agilent/Keysight 3.5 or 2.92 calkit and then compare the
> calibration made with this calkit (using the right calkit model) with
> the calibration obtained with NanoVNA calkit (using an ideal model).
> In this case the comparison can be done measuring S11 and S21 (Mod and
> Phase) of different DUTs. Use DUTs having high and low return loss and
> high and low tranfer loss just to probe all possible cases.
>
> Ciao, Marco.
>

Hello,

up to 1 GHz, I think the SMA Open and Short from NanoVNA kit are good enough. The Load I got with my Nano (a clone...) had a return loss of 35 dB at 1 GHz so it was quite poor. With this poor load as calibration standard you can measure DUT with return loss down to 25 dB (i.e. with VSWR down to 1.12) with an error of about 1/2 dB. I bought for 10€ a better load with about 50 dB return loss at 1 GHz (https://it.aliexpress.com/item/32860709756.html): I do not know if these Loads have all the same performance on mine, anyway this new Load maybe worth the expense.

I've no experience on the quality of the Type-N calibration kit coming with SAA-2N 4.

Ciao, Marco.