i add interpolation like on NanoVNA
See PR https://github.com/nanovna/NanoVNA-V2-firmware/pull/46
Also loat calibration data from slot 0 on startup

You can test it work.
Made 2 calibration:
min to 140MHz
40MHz to max
On 201 points

Select calibration for work interval. Change interval inside it. Calibration should be correctly rebuild (yes possible small errors, but it allow fast change frequency range, points count and see good result)

On Sun, Sep 20, 2020 at 01:58 PM, DiSlord wrote:

>
> i add interpolation like on NanoVNA
> See PR https://github.com/nanovna/NanoVNA-V2-firmware/pull/46
> Also load calibration data from slot 0 on startup
>
> You can test it work.
> Made 2 calibration:
> min to 140MHz
> 40MHz to max
> On 201 points
>
> Select calibration for work interval. Change interval inside it.
> Calibration should be correctly rebuild (yes possible small errors, but it
> allow fast change frequency range, points count and see good result)

DiSlord,
I had the following observations measuring a 20dB attenuator after installing the new firmware.  I used the same HP 70dB step attenuator rated DC-4GHz for each test:

Tindie V2 with 4" Display. No trace toggling but stays ~1.5dB             SAA-2N  Trace toggles between correct level and ~1.5dB
high.    480x320 interpolation test.bin                                              high.   480x320 interpolation test.bin

SAA-2 with 2.8" Display. No trace toggling. Works as expected.       Same as to the left but reduced stop frequency to 1.5GHZ to test
320x240 interpolation test.bin                                                              Linear interpolation feature.  First point error returned.

It appears there is something going on with the 480x320  firmware that causes an ~ 1.5dB S21 measurement error after 10-20dB of step attenuation.  The error was measured for both the Tindie V2 with 4" display and SAA-2N.  Kurt Poulson first reported this issue.

The S21 measurement error was not detected on a V2 with the 320x240 firmware installed while the HP attenuator was stepped from 0-60dB.

Limited testing of the new interpolation feature, using the 320x240 firmware, showed a first point error when reducing the stop frequency from 3GHz to 1.5GHz.

- Herb

First point error:

>
> Made 2 calibration:
> min to 140MHz
> 40MHz to max
> On 201 points
>

At begin (up to 40-50 MHz) on long range only 1-2 calibration point. And not possible made correct interpolation by it. Need as minimum 2 intervals (or more calibration points at begin).
So if need measure in min-140MHz range - made calibration in it (min-140MHz) and select any range in: 50-80MHz for example or 90-100 and so.
If need measure in high range made calibration in 40-Max and use this calibration.

>
>
> It appears there is something going on with the 480x320  firmware that
> causes an ~ 1.5dB S21 measurement error after 10-20dB of step attenuation.
> The error was measured for both the Tindie V2 with 4" display and SAA-2N. 
> Kurt Poulson first reported this issue.
>
>
>

You can try disable calibration, if you see some error - this not interpolation issue, i use last git version for add this feature.

>
>
>

On Sun, Sep 20, 2020 at 08:34 PM, DiSlord wrote:

It appears there is something going on with the 480x320  firmware that causes an ~ 1.5dB S21 measurement error after 10-20dB of step attenuation.  The error was measured for both the Tindie V2 with 4" display and SAA-2N.  Kurt Poulson first reported this issue.

You can try disable calibration, if you see some error - this not interpolation issue, i use last git version for add this feature.

==========================================
DiSlord,
I believe the issue may have existed in the original st7796 firmware and only noticed as the 4" SAA-2N begin to recently start arriving.  I tried your beta 480x240 FW dated 2020-21-7  and the S21 issue was also there.

Not related to the new beta interpolation feature.

- Herb

I do not think that the 4` display is the cause of this error (although it may be due to the higher current consumption for the backlight, but unlikely) in most cases code some, only little difference on init and data amount for draw.

On Sun, Sep 20, 2020 at 08:56 PM, DiSlord wrote:

>
> I do not think that the 4` display is the cause of this error (although it
> may be due to the higher current consumption for the backlight, but
> unlikely) in most cases code some, only little difference on init and data
> amount for draw.

DiSlord,
I you have a 10dB step attenuator, can check values above 10dB using the st7796.bin and see if you get the same ~1.5dB measurement error?  I'm not seeing the issue using the ili9341.bin, only the st7796.bin.

- Herb

I can test at evening, but have only
V2 from tindie vs 2.8 and 4` display module (from Gabriel)
V2 from Hugen vs 2.8 LCD

I have electronic attenuator 0-94.5dB and variable attenuator from 5 to ~60dB for test

Hi DiSlord and Herb

I documented the problem with a video in uncalibrated mode. The problem is only solid reproduceable when running thru 1dB steps and the it is only happens for full sweep at a specific attenuation as seen on the video at a number of 1dB steps for reduced span. It happens for 13, 14 and 15dB attenuation in uncalibrated ad exactly at 140MHz by a sweep from 130 to 150MHz. At attenuation higher than 15dB there is a permanent error of +1dB. So the transition from 15dB to 16 dB does not produce a different trave on screen but stays at same level and onward for more attenuation a +1dB error introduced
In calibrated mode it appears at a different level starting around 15dB attenuation.
It appears it is a matter of a critical input level to port 2. It is only a problem for the firmware v2_2-st7796.bin. The version from 09162020 released by Gabrille still have the bug but it removed the first point measurements error. It is shown for Herb using a original Tindie version with 4 inch screen an on the SAA-2N from Hugem

<http://www.hamcom.dk/SAA-2N/00190.MTS> http://www.hamcom.dk/SAA-2N/00190.MTS

Kind regards

Kurt



Fra: NanoVNAV2@groups.io <NanoVNAV2@groups.io> På vegne af DiSlord
Sendt: 21. september 2020 06:20
Til: NanoVNAV2@groups.io
Emne: Re: [nanovnav2] Linear interpolation for calibration data on V2



I can test at evening, but have only
V2 from tindie vs 2.8 and 4` display module (from Gabriel)
V2 from Hugen vs 2.8 LCD

I have electronic attenuator 0-94.5dB and variable attenuator from 5 to ~60dB for test

Made test measure:
Use V2 from Hugen on 2.8 LCD, calibrate in 40MHz to 3.5GHz on 201 point
Get variable attenuator and slow up and down it. No issues.
Change range to 40MHz - 1.5GHz. Slowly change attenuator from 5dB to 60dB. No issues. Interpolation work as need.
Change range 100MHz to 200MHz. Repeat measure. No problem. I see flat line in all cases. (Try change attenuation vs 0.2-0.3dB step and wait all scan)

Remove 2.8 display module. And install 4`. Flash 'V2 480x320 interpolation test.bin' firmware.
Made touch calibration.
Repeat all measures.
No differences. All work as expected. In all cases i see ~flat line (+- small 0.1 dB noise).

I connect constant 20dB attenuator.
Interpolation allow get correct result in all frequency range/points count change.
If reduce point count to 50, i get more fast, but more noise measure (in some cases +-0.4dB)

Sorry, i can`t test SAA-2N, i not have it. But try use it parts.

Hi DiSloard

Thank you very much for your testing. What firmware did you use please

Kind regards

Kurt



Fra: NanoVNAV2@groups.io <NanoVNAV2@groups.io> På vegne af DiSlord
Sendt: 21. september 2020 19:56
Til: NanoVNAV2@groups.io
Emne: Re: [nanovnav2] Linear interpolation for calibration data on V2



Made test measure:
Use V2 from Hugen on 2.8 LCD, calibrate in 40MHz to 3.5GHz on 201 point
Get variable attenuator and slow up and down it. No issues.
Change range to 40MHz - 1.5GHz. Slowly change attenuator from 5dB to 60dB. No issues. Interpolation work as need.
Change range 100MHz to 200MHz. Repeat measure. No problem. I see flat line in all cases. (Try change attenuation vs 0.2-0.3dB step and wait all scan)

Remove 2.8 display module. And install 4`. Flash 'V2 480x320 interpolation test.bin' firmware.
Made touch calibration.
Repeat all measures.
No differences. All work as expected. In all cases i see ~flat line (+- small 0.1 dB noise).

I connect constant 20dB attenuator.
Interpolation allow get correct result in all frequency range/points count change.
If reduce point count to 50, i get more fast, but more noise measure (in some cases +-0.4dB)

Sorry, i can`t test SAA-2N, i not have it. But try use it parts.

I use last compiled from official github V2 firmware + my interpolation fixes (see this post https://groups.io/g/NanoVNAV2/message/226 )
For 4` (SAA-2N) need use *V2 480x320 interpolation test.bin
* *
* PS as i see Gabriel apply my PR, so on next release you can use this feature

Hi DiSlord

Thank you very much

Will test in few minutes

Kind regards

Kurt





Fra: NanoVNAV2@groups.io <NanoVNAV2@groups.io> På vegne af DiSlord
Sendt: 21. september 2020 22:32
Til: NanoVNAV2@groups.io
Emne: Re: [nanovnav2] Linear interpolation for calibration data on V2



I use last compiled from official github V2 firmware + my interpolation fixes (see this post https://groups.io/g/NanoVNAV2/message/226)
For 4` (SAA-2N) need use V2 480x320 interpolation test.bin

PS as i see Gabriel apply my PR, so on next release you can use this feature

On Mon, Sep 21, 2020 at 10:55 AM, DiSlord wrote:

>
> Made test measure:
> Use V2 from Hugen on 2.8 LCD, calibrate in 40MHz to 3.5GHz on 201 point
> Get variable attenuator and slow up and down it. No issues.
> Change range to 40MHz - 1.5GHz. Slowly change attenuator from 5dB to 60dB.
> No issues. Interpolation work as need.
> Change range 100MHz to 200MHz. Repeat measure. No problem. I see flat line
> in all cases. (Try change attenuation vs 0.2-0.3dB step and wait all scan)
>
>
> Remove 2.8 display module. And install 4`. Flash 'V2 480x320 interpolation
> test.bin' firmware.
> Made touch calibration.
> Repeat all measures.
> No differences. All work as expected. In all cases i see ~flat line (+-
> small 0.1 dB noise).
>
> I connect constant 20dB attenuator.
> Interpolation allow get correct result in all frequency range/points count
> change.
> If reduce point count to 50, i get more fast, but more noise measure (in
> some cases +-0.4dB)
>
> Sorry, i can`t test SAA-2N, i not have it. But try use it parts.

DiSlord,
Thanks for the feedback.  The only difference I am seeing from your measurements is that with:

Remove 2.8 display module. And install 4`. Flash 'V2 480x320 interpolation test.bin' firmware.
Made touch calibration.
Repeat all measures.
No differences. All work as expected. In all cases i see ~flat line (+- small 0.1 dB noise).

With a 4" display installed in my Tindie V2 and using 'V2 480x320 interpolation test.bin' I am also seeing a flat line but it is ~1.5dB too high from its known value; and from the value measured with the a 320x240 display installed in my Tindie V2 and using 'V2 320x240 interpolation test.bin':

320x240 FW:  Straight line -20dB across.                                                   480x320 FW: Straight line but about -18.5dB across.

Can you confirm that your measured S21 level using the 'V2 480x320 interpolation test.bin' is the same as when using 'V2 320x240 interpolation test.bin'?

Thanks again for replying.

- Herb

For test need open again my V2, i do it tomorow

But anyway, then i test this attenuator on my Hugen V2 2.8
This attenuator: aliexpress.ru/item/32951172012.html
It have self 3.65dB if set to zero. But after some time use (made up/down change) i see 4.8dB on zero. I thought this strangely, but did not attach any importance to this, deciding that this was a possible drawback of this attenuator. But maybe I'm wrong and this value is floating at NanoVna

Hi DiSlord

You are a hero. Thank you very much for the firmware
You have fixed the jumping 1 dB problem with the SAA-2N. I have carefully stepped all levels up to 50dB and it is indeed very accurate (so is my 1dB step attenuator) Done for 0.1 to 40MHz and also for 40 to 3000MHz and it start of course to be noisy at high frequency without averaging.

I am very pleased with the performance I also ran attest for the entire span of a 20dB as well a 40dB 18GHz HP atennuation and they worked just perfect spot on.
I did a SOLT calibration with my HP85032B male calibration kit and sweept a 50cm UT141A rigid cable as seen on the image. It is of course not perfect as calibration kit is not ideal for short and open having slight different delays

A test with VNA-QT is also 100% OK and the result attached. The calibration kit data are done and tested for the male and the female kit test are delayed as I have lost track of the female female adaptor. So I need help from others to test the female kit

How do I see an impact from Enhanced Response ? I did not figure out what is does

Kind regards

Kurt



Fra: NanoVNAV2@groups.io <NanoVNAV2@groups.io> På vegne af DiSlord
Sendt: 21. september 2020 22:32
Til: NanoVNAV2@groups.io
Emne: Re: [nanovnav2] Linear interpolation for calibration data on V2



I use last compiled from official github V2 firmware + my interpolation fixes (see this post <https://groups.io/g/NanoVNAV2/message/226> https://groups.io/g/NanoVNAV2/message/226)
For 4` (SAA-2N) need use V2 480x320 interpolation test.bin

PS as i see Gabriel apply my PR, so on next release you can use this feature

Hello Kurt,
After SOLT calibration with HP85032B Kit in NanoVNA-QT, did you save the calibration? Could you kindly share the saved calibration file (.cal)?
Best Regards, Marco.

Hi Marco

Maybe, but I will do it just for you 😊

I also attached the HP85032B s1p and s2p files for calibration of the VNA QT software. Please note the s2p files for Thru calibration (full forward and reverse calibration) are with no delay, meaning no thru adaptors used, so the female port 1 mate the male adaptor of the test cable from port2 during male calibration and vice versa for female calibration where a male male adaptor fitted to Port1 and a female female adaptor fitted to male connector on the test cable from port 2

Do you have a SSA-2N ? and a HP85032B or E calibration kit ??

In case I might need you help

Kind regards

Kurt

Fra: NanoVNAV2@groups.io <NanoVNAV2@groups.io> På vegne af mce66
Sendt: 22. september 2020 10:42
Til: NanoVNAV2@groups.io
Emne: Re: [nanovnav2] Linear interpolation for calibration data on V2



Hello Kurt,
After SOLT calibration with HP85032B Kit in NanoVNA-QT, did you save the calibration? Could you kindly share the saved calibration file (.cal)?
Best Regards, Marco.

Hello Kurt,

I attach 3 figures obtained from a quick analysis of your NanoVNA calibration data.

First picture shows the error coefficients of your port 1 (e00=directivity, e11=port 1 match, e1001=port 1 tracking). You can see that you HW is quite good, with a port 1 match that is between 20 and 30 dB RL (right scale, left scale is only for e1001)

Second picture shows the error coefficients of your port 2+cable (e22=port 2 match, e1032=port 2 tracking). Also here, you can see that you HW is quite good, with a port 2 match that is below 25 dB RL at 1.5 GHz and about 20 dB RL at 3 GHz.

Third picture answers your question: " How do I see an impact from Enhanced Response?". Enhanced Response should eliminate two approximations, present in the Standard Response (and implemented in original NanoVNA FW):
1) neglect e22/e11 reflection during calibration of e1032 with THRU
2) neglect S11/e11 reflection during measurement of the S21 of a DUT

So far, latest FW (20200916) only solves approximation "2)" and then the approximation "1)" is still present.

Figure 3 shows in ORANGE the effect of approximation "1)" applied to your "good" HW. In practice you should have a residual ripple of up to +/-0.1 dB around 1 and 2.5 GHz. This approximation is always present since it is independent by DUT. As said, is not solved by latest FW.

Figure 3 finally shows in BLU the ripple limit that is expected in your HW due to approximation "2)". Since approximation "2)" depends on the S11 of the DUT, the figure assumes a DUT with an S11 of 10 dB (VSWR=2). This is,for example, the S11 of a T-Check (if you have one, you can try to measure it and see if the ripple on S21 is about +/- 0.3 dB at 1.5 and 2.5 GHz, as expected).This approximation "2)" is now solved by the Enhanced Response in latest FW, so enabling and disabling this feature should be an easy way to see the correction on S21 measurement in action. Again, this will be only visible with pretty mis-matched DUT and will not be visible with well matched DUT (like attenuators) for which S11 is about 0. At last, note that the period of this ripple will depend on the length of the cable between VNA TX port and DUT input port.

No, I do ot have SSA-2N or HP85032B/E...

Best regards, Marco.

Hello Marco

Thank you very much for you analyzes I will try to test a T adaptor I have several some better than others 😊
As switching the enhanced function on and off can only be done with the barefoot SAA-2N and I have a set of homemade female N cal standards with rear flange calibration with know delays and it might be interesting to see how these behaves as well.

I envy you insight in these matters I will have to study your feed back in details to see if I can grasp how you are doing these analyzes

Kind regards

Kurt



Fra: NanoVNAV2@groups.io <NanoVNAV2@groups.io> På vegne af mce66
Sendt: 22. september 2020 15:53
Til: NanoVNAV2@groups.io
Emne: Re: [nanovnav2] Linear interpolation for calibration data on V2



Hello Kurt,

I attach 3 figures obtained from a quick analysis of your NanoVNA calibration data.

First picture shows the error coefficients of your port 1 (e00=directivity, e11=port 1 match, e1001=port 1 tracking). You can see that you HW is quite good, with a port 1 match that is between 20 and 30 dB RL (right scale, left scale is only for e1001)

Second picture shows the error coefficients of your port 2+cable (e22=port 2 match, e1032=port 2 tracking). Also here, you can see that you HW is quite good, with a port 2 match that is below 25 dB RL at 1.5 GHz and about 20 dB RL at 3 GHz.

Third picture answers your question: "How do I see an impact from Enhanced Response?". Enhanced Response should eliminate two approximations, present in the Standard Response (and implemented in original NanoVNA FW):
1) neglect e22/e11 reflection during calibration of e1032 with THRU
2) neglect S11/e11 reflection during measurement of the S21 of a DUT

So far, latest FW (20200916) only solves approximation "2)" and then the approximation "1)" is still present.

Figure 3 shows in ORANGE the effect of approximation "1)" applied to your "good" HW. In practice you should have a residual ripple of up to +/-0.1 dB around 1 and 2.5 GHz. This approximation is always present since it is independent by DUT. As said, is not solved by latest FW.

Figure 3 finally shows in BLU the ripple limit that is expected in your HW due to approximation "2)". Since approximation "2)" depends on the S11 of the DUT, the figure assumes a DUT with an S11 of 10 dB (VSWR=2). This is,for example, the S11 of a T-Check (if you have one, you can try to measure it and see if the ripple on S21 is about +/- 0.3 dB at 1.5 and 2.5 GHz, as expected).This approximation "2)" is now solved by the Enhanced Response in latest FW, so enabling and disabling this feature should be an easy way to see the correction on S21 measurement in action. Again, this will be only visible with pretty mis-matched DUT and will not be visible with well matched DUT (like attenuators) for which S11 is about 0. At last, note that the period of this ripple will depend on the length of the cable between VNA TX port and DUT input port.

No, I do ot have SSA-2N or HP85032B/E...

Best regards, Marco.

Hello Kurt,

for the analysis, I just used the same excel attached to my last message on Enhanced Response Topic with a small addition. I just copied in the excel (columns from B to K) the raw measurements of the cal kit recorded in your calibration file (the .cal file saved by NanoVNA-QT). This allows the excel to calculate all error coefficients of port 1 and port 2. Since you are not measuring any DUT, I also copied the Thru raw measurement (column H to K) in DUT columns M to P as if you were measuring a THRU as a DUT. I attach the excel with your data to this message for your convenience.

I also made the S21 measurement of a "high" mismatched DUT: I used a non-resistive 3 dB splitter as DUT just leaving the secomd output port open (i.e. not loaded) to have some large mismatch on splitter input port. Attached are the results without and with Enhanced Response (as fas as implemented so far). You can see in the Standard Response a large ripple on S21. When Enhanced Response is activated, the majority of the ripple is over. In my "bad" clone NanoVNA I still have a large residual ripple (+/- 0.5 dB) due to the " neglected e22/e11 reflection during calibration of e1032 with THRU" you can clearly see between 1.5 and 3 GHz. Hope also this will be corrected in future FW.

Best regards, Marco.

>
> Can you confirm that your measured S21 level using the 'V2 480x320
> interpolation test.bin' is the same as when using 'V2 320x240
> interpolation test.bin'?

On Mon, Sep 21, 2020 at 02:25 PM, DiSlord wrote:

>
> For test need open again my V2, i do it tomorow
>
> But anyway, then i test this attenuator on my Hugen V2 2.8
> This attenuator: aliexpress.ru/item/32951172012.html
> It have self 3.65dB if set to zero. But after some time use (made up/down
> change) i see 4.8dB on zero. I thought this strangely, but did not attach
> any importance to this, deciding that this was a possible drawback of this
> attenuator. But maybe I'm wrong and this value is floating at NanoVna

I made again measure:
Constant 10, 20, 30, 40dB attenuators, and no found any difference on use 2.8` display module or 4`.
On 500MHz
10dB - get  9.90dB
20dB - get  20.01dB
30dB - get  30.04dB
40dB - get  39.74dB
On both displays +-0.05dB difference

On Tue, Sep 22, 2020 at 10:24 AM, DiSlord wrote:

I made again measure:
Constant 10, 20, 30, 40dB attenuators, and no found any difference on use 2.8` display module or 4`.
On 500MHz
10dB - get  9.90dB
20dB - get  20.01dB
30dB - get  30.04dB
40dB - get  39.74dB
On both displays +-0.05dB difference
============================================

DiSlord,

Thanks for taking the time to check.  Possibly an error has worked its way into my set-up and I need to track it down. Fun and games.

- Herb

Hi Kurt
Thank you very much for publishing  your results and calibration files.
I think in this way we can compare different loads with your HP85032B  load that otherwise we would not have access.
I used your male calibration file and did 2 measurements: one with the original SAA2N firmware  and another one with latest DisLord firmware.
However it seems that my load is worse than yours.
Kind regards,
Corneliu

Hello Marco

Thank you, great.
I have some more to test, using you spreadsheet, which will be educational

Kind regards

Kurt



Fra: NanoVNAV2@groups.io <NanoVNAV2@groups.io> På vegne af mce66
Sendt: 22. september 2020 18:13
Til: NanoVNAV2@groups.io
Emne: Re: [nanovnav2] Linear interpolation for calibration data on V2



Hello Kurt,

for the analysis, I just used the same excel attached to my last message on Enhanced Response Topic with a small addition. I just copied in the excel (columns from B to K) the raw measurements of the cal kit recorded in your calibration file (the .cal file saved by NanoVNA-QT). This allows the excel to calculate all error coefficients of port 1 and port 2. Since you are not measuring any DUT, I also copied the Thru raw measurement (column H to K) in DUT columns M to P as if you were measuring a THRU as a DUT. I attach the excel with your data to this message for your convenience.

I also made the S21 measurement of a "high" mismatched DUT: I used a non-resistive 3 dB splitter as DUT just leaving the secomd output port open (i.e. not loaded) to have some large mismatch on splitter input port. Attached are the results without and with Enhanced Response (as fas as implemented so far). You can see in the Standard Response a large ripple on S21. When Enhanced Response is activated, the majority of the ripple is over. In my "bad" clone NanoVNA I still have a large residual ripple (+/- 0.5 dB) due to the "neglected e22/e11 reflection during calibration of e1032 with THRU" you can clearly see between 1.5 and 3 GHz. Hope also this will be corrected in future FW.

Best regards, Marco.

Hi Corneliu

I have just published the complete story

Kind regards

Kurt



Fra: NanoVNAV2@groups.io <NanoVNAV2@groups.io> På vegne af Corneliu via groups.io
Sendt: 22. september 2020 20:45
Til: NanoVNAV2@groups.io
Emne: Re: [nanovnav2] Linear interpolation for calibration data on V2



Hi Kurt
Thank you very much for publishing your results and calibration files.
I think in this way we can compare different loads with your HP85032B load that otherwise we would not have access.
I used your male calibration file and did 2 measurements: one with the original SAA2N firmware and another one with latest DisLord firmware.
However it seems that my load is worse than yours.
Kind regards,
Corneliu