Kirkby calibration kit alternatives?

[hendorog]

I entered a polynomial for the Lx corrections for my DIY cal kit SHORT and got the result below...

what i concluded from your reply #35 and #36 is, are you suggesting that using less quality cal kit in the same VNA will result in more ripple? if that is the case, there is possibility that kirkby did the attenuator report using agilent cal kit as his calibration standard? i'll need to confirm this back to him.

anyway i'm doing the test with cheap crap china sma connection and 50 ohm coax cable. maybe those also contributes to the rippling effect, i'm not sure. what i'm sure is the cabling still give consistent result even if i move them around a little bit... fwiw...

I just took a closer look at your attachments in the post above this one.

Few comments which hopefully will help:
* The attenuator return loss measurements, with the free port terminated, is probably not too bad given the device. The return loss of that attenuator with attached terminator is probably pushing the limits of your device.
* The Dr Kirkby data and method you have to use (i.e. using the terminator) don't seem to line up with what he has in his FAQ: http://www.kirkbymicrowave.co.uk/FAQ/How-do-I-verify-the-calibration-kit-is-working-properly/
* The S21 measurements are probably poor as the Deepace doesn't seem to support a 'full 2 port' calibration. Do you know if it supports a full 2 port calibration, or is it just a 'response' calibration? In The Signal Path youtube review it shows a simple response cal only with a through.
* A full 2 port cal can correct for many more errors than a simple response cal can. Since you are using a lower cost, compact device which doesn't have great directivity and port return loss, then you will really need this error correction to get decent results on the S21 test.
* If you can pull the raw data out of the device and get it into touchstone format, then you should be able to use scikit-rf or Metas to do the full 2 port cal.
* According the manual it supports using an external bridge/coupler. Sometimes high quality ones come up on eBay which have come from HP gear.

[Mechatrommer]

* The Dr Kirkby data and method you have to use (i.e. using the terminator) don't seem to line up with what he has in his FAQ: http://www.kirkbymicrowave.co.uk/FAQ/How-do-I-verify-the-calibration-kit-is-working-properly/

i did exactly according to that in my first report to him (see attached version 1 report fwiw). the S11 or S22 (just turning the attenuator upside down on the same VNA port 1) was totally out. in our email conversation he admitted the mistake of procedure in his link and  said he will edit the link, but it seems he havent. he asked me to do the test again with attenuator terminated. with attenuator's end dangling opened (like in the link), the reflection from opened end will affect to the port under test on the under end, something like that according to him.

* The S21 measurements are probably poor as the Deepace doesn't seem to support a 'full 2 port' calibration. Do you know if it supports a full 2 port calibration, or is it just a 'response' calibration? In The Signal Path youtube review it shows a simple response cal only with a through.
* A full 2 port cal can correct for many more errors than a simple response cal can. Since you are using a lower cost, compact device which doesn't have great directivity and port return loss, then you will really need this error correction to get decent results on the S21 test.

correct, there is no directional coupler (or power splitter) on the other port 2 of the VNA, so it can measure S21 only but not true S12. i have to turn the DUT around to simulate S12. i guess the VNA will try to solve half part of the full 4 port parameter matrix.

* If you can pull the raw data out of the device and get it into touchstone format, then you should be able to use scikit-rf or Metas to do the full 2 port cal.

what raw data? sorry noob question. but i dont think i can extract anything from the device other than its measurement result.

* According the manual it supports using an external bridge/coupler. Sometimes high quality ones come up on eBay which have come from HP gear.

thanks for reminding me this. i will study more on this.

ps: albeit more ripples on the output of the KC901V VNA, similar to G0HZU simulation on poorly tuned cal kit, i think we can still do some regression or averaging in our head on the result, since its still following the main path if we can see in G0HZU's report pictures reply #35, #36, #44 overlapped (2nd attachment). so i hope my VNA will give not entirely useless result. i just have to "assume" of lesser ripple in the final output. my further test and measurement will confirm this hypothesis true or false as i go, i just have to understand my tool better. ymmv.

[hendorog]

* The Dr Kirkby data and method you have to use (i.e. using the terminator) don't seem to line up with what he has in his FAQ: http://www.kirkbymicrowave.co.uk/FAQ/How-do-I-verify-the-calibration-kit-is-working-properly/

i did exactly according to that in my first report to him (see attached version 1 report fwiw). the S11 or S22 (just turning the attenuator upside down on the same VNA port 1) was totally out. in our email conversation he admitted the mistake of procedure in his link and  said he will edit the link, but it seems he havent. he asked me to do the test again with attenuator terminated. with attenuator's end dangling opened (like in the link), the reflection from opened end will affect to the port under test on the under end, something like that according to him.

Yep makes sense.

* The S21 measurements are probably poor as the Deepace doesn't seem to support a 'full 2 port' calibration. Do you know if it supports a full 2 port calibration, or is it just a 'response' calibration? In The Signal Path youtube review it shows a simple response cal only with a through.
* A full 2 port cal can correct for many more errors than a simple response cal can. Since you are using a lower cost, compact device which doesn't have great directivity and port return loss, then you will really need this error correction to get decent results on the S21 test.

correct, there is no directional coupler (or power splitter) on the other port 2 of the VNA, so it can measure S21 only but not true S12. i have to turn the DUT around to simulate S12. i guess the VNA will try to solve half part of the full 4 port parameter matrix.

If you extract the data then you can still do better using scikit-rf. It supports a calibration type called 'Two Port one Path' or something like that. Basically you can do a full 2 port cal with only one directional coupler. You do need to swap around the DUT to measure both ports.
However to do this, first you need to be able to extract the raw data for the sweeps and transfer it to a PC. Then use software on the PC to do the calibration instead of using the software inside the KC901V.

* If you can pull the raw data out of the device and get it into touchstone format, then you should be able to use scikit-rf or Metas to do the full 2 port cal.

what raw data? sorry noob question. but i dont think i can extract anything from the device other than its measurement result.

Check the 'Data Saving' section of the manual. I found it online, apparently you can export s1p data files to an SD card in the smith chart mode.

Without doing a cal of the device, you do a sweep of your cal standards, and export them each into different files - i.e. short.s1p, open.s1p, load.s1p, thru.s1p and the device you want to test dut.s1p. Then you should be able to use those files containing 'raw data' with software like scikit-rf or Metas to do a calibration and measure the device on your PC.

Note I haven't done this myself - yet - so ymmv.

[in3otd]

That's why i hope somebody can recommend an alternative.

There is a description about using Rosenberger parts, similar to the ones in the SDR-Kits, as calkit, see http://www.hhft.de/index.php?page=competences&subpage=calibration. They just use the offset lengths to describe the standards and leave all the Cx and Lx to zero. It may not be so important to split the femtofarad if you go up to 5 GHz - any excess capacitance or inductance can probably taken into account by adjusting the offset length only.

[hendorog]

Metas download link has landed in my spam folder this evening (I use gmail)

So anyone that is waiting, might be worth checking there.

[hendorog]

That's why i hope somebody can recommend an alternative.

There is a description about using Rosenberger parts, similar to the ones in the SDR-Kits, as calkit, see http://www.hhft.de/index.php?page=competences&subpage=calibration. They just use the offset lengths to describe the standards and leave all the Cx and Lx to zero. It may not be so important to split the femtofarad if you go up to 5 GHz - any excess capacitance or inductance can probably taken into account by adjusting the offset length only.

Not sure if you have seen this info :- SDR kits actually publish reasonable delay info, and the part numbers of the Rosenberger parts they use. It used to be really difficult to find on their old website but its better now. There is also a document explaining how they got those values. Unfortunately they don't provide the s1p/s2p files or we could use them with your script to get the missing info.

https://www.sdr-kits.net/index.php?route=web/pages&page_id=36_36
https://www.sdr-kits.net/documents/Rosenberger_Fairview_Male_Cal_Standards.pdf
https://www.sdr-kits.net/documents/Rosenberger_Female_Cal_Standards_rev4.pdf

[in3otd]

Not sure if you have seen this info :- SDR kits actually publish reasonable delay info, and the part numbers of the Rosenberger parts they use. It used to be really difficult to find on their old website but its better now. There is also a document explaining how they got those values. Unfortunately they don't provide the s1p/s2p files or we could use them with your script to get the missing info.

yes, I should have started my message saying "in addition to the well know calkit from SDR-Kits, which comes wit some modeling data"  .
A few years ago I exchanged a couple of emails with Kurt Poulsen (which prepared the SDR-Kits calkit docs), when I saw a message he posted to the VNWA Users Group on Yahoo saying that had measured the S-parameters of the Amphenol Connex and Rosenberger calibration kits; I don't remember if he also sent me the related .s1p/.s2p files, need to check my old emails.
Anyway, it would make sense to tell the SDR-Kits people that making the measured calkit data available may help in selling more kits  .

[hendorog]

Not sure if you have seen this info :- SDR kits actually publish reasonable delay info, and the part numbers of the Rosenberger parts they use. It used to be really difficult to find on their old website but its better now. There is also a document explaining how they got those values. Unfortunately they don't provide the s1p/s2p files or we could use them with your script to get the missing info.

yes, I should have started my message saying "in addition to the well know calkit from SDR-Kits, which comes wit some modeling data"  .
A few years ago I exchanged a couple of emails with Kurt Poulsen (which prepared the SDR-Kits calkit docs), when I saw a message he posted to the VNWA Users Group on Yahoo saying that had measured the S-parameters of the Amphenol Connex and Rosenberger calibration kits; I don't remember if he also sent me the related .s1p/.s2p files, need to check my old emails.
Anyway, it would make sense to tell the SDR-Kits people that making the measured calkit data available may help in selling more kits  .

Hehe

I was just trawling through his website after reading your post. In one of his older docs he actually points to that same website and shows some of their data

It would be great if they did. There are so many 8753's around that the prices have dropped even on eBay - which is the opposite of just about everything else there! There are just no cal kits to go with them (or test sets... sigh...) I'm sure they would sell quite a few of them.

My cal kit box has s/n #28 on it, from about 2015. I wonder what number they are up to now...

[suj]

There is one type of calibration kits, that you can easily prepare yourself. Below the X band, waveguides are rarely used, so rather unhelpful for 8753. I have 8510C which supports TRL calibration. I needed to measure the WR90/WR75 adapter. I had the calibration kit WR90 but I missed the WR75. That's what I decided to build. I ordered machining on Wire EDM. I made the finishing of the surface with sandpaper and polishing paste. The determination of coefficients is pure geometry. I measured the length of the 1/4 lambda line with a micrometer. In the most simple case (if we do not measure the insulation), such a kit consists of two elements - short (wall closing the waveguide) and a section of waveguide 1/4 lambda. Inexpensive, quite easy to do and work - I compared the results with Maury X7005E WR90 kit.
And the most important thing  It must have a wooden box.

[Mechatrommer]

If you extract the data then you can still do better using scikit-rf. It supports a calibration type called 'Two Port one Path' or something like that. Basically you can do a full 2 port cal with only one directional coupler. You do need to swap around the DUT to measure both ports.
However to do this, first you need to be able to extract the raw data for the sweeps and transfer it to a PC. Then use software on the PC to do the calibration instead of using the software inside the KC901V.

Without doing a cal of the device, you do a sweep of your cal standards, and export them each into different files - i.e. short.s1p, open.s1p, load.s1p, thru.s1p and the device you want to test dut.s1p. Then you should be able to use those files containing 'raw data' with software like scikit-rf or Metas to do a calibration and measure the device on your PC.

i think this is the missing piece for me to all of your discussions here, thanks i'll try to catch up. but the downloaded META VNA Tools II only works on Windows later than Vista, so my WinXP cant installed, another moment... i'll try install it in another Win10 PC.

Check the 'Data Saving' section of the manual. I found it online, apparently you can export s1p data files to an SD card in the smith chart mode.

correct. i'll study this more seriously when i got installed any VNA tools. thanks.

[hendorog]

If you extract the data then you can still do better using scikit-rf. It supports a calibration type called 'Two Port one Path' or something like that. Basically you can do a full 2 port cal with only one directional coupler. You do need to swap around the DUT to measure both ports.
However to do this, first you need to be able to extract the raw data for the sweeps and transfer it to a PC. Then use software on the PC to do the calibration instead of using the software inside the KC901V.

Without doing a cal of the device, you do a sweep of your cal standards, and export them each into different files - i.e. short.s1p, open.s1p, load.s1p, thru.s1p and the device you want to test dut.s1p. Then you should be able to use those files containing 'raw data' with software like scikit-rf or Metas to do a calibration and measure the device on your PC.

i think this is the missing piece for me to all of your discussions here, thanks i'll try to catch up. but the downloaded META VNA Tools II only works on Windows later than Vista, so my WinXP cant installed, another moment... i'll try install it in another Win10 PC.

Check the 'Data Saving' section of the manual. I found it online, apparently you can export s1p data files to an SD card in the smith chart mode.

correct. i'll study this more seriously when i got installed any VNA tools. thanks.

Cool.

Now that I have Metas I think you are better off using scikit-rf for this type of thing. Metas is obviously targeted at Metrology and lab work, and as such is very 'rigorous' in its processes. Having worked in a lab (food not RF!) before I recognise the focus on recording everything, traceability of measurements, cables and adapters used for the measurement etc etc.
I am not even sure yet what it can and can't do! This is just my impression from using it once so ymmv.

scikit-rf is a much more techy to get installed and working, but has more flexibility so I will use that for experimentation.

[Mechatrommer]

scikit-rf is a much more techy to get installed and working, but has more flexibility so I will use that for experimentation.

i am heavy butt lazy arse nowadays that i expect a ready made exe, i dont have much more time for long step by step install of Anaconda, Phyton learning etc. but i'll try that if i cant work out META... thanks.

[hendorog]

scikit-rf is a much more techy to get installed and working, but has more flexibility so I will use that for experimentation.

i am heavy butt lazy arse nowadays that i expect a ready made exe, i dont have much more time for long step by step install of Anaconda, Phyton learning etc. but i'll try that if i cant work out META... thanks.

Yeah I can understand.

It actually was quite straightforward though as anaconda takes care of installing python, scikit-rf and all of the dependancies, including a simple IDE. Certainly it is much easier than back when anything python related had to be installed manually, package by package. Now that was a pain in the a$$.

[hendorog]

New thread with install guide for the IN3OTD script is here:
https://www.eevblog.com/forum/rf-microwave/vna-cal-kit-modelling-script-install-guide/

[hendorog]

In other news, on his Test Equipment group Dr Kirkby offered a couple of out of spec cal kits at a low price.

Since these come with s-params I grabbed one of them. Shipping is very high, but ultimately still worth it for me to have items which I can use as a calibrated reference.

https://groups.io/g/Test-Equipment-For-Sale-Wanted-or-Exchange

[TheSteve]

In other news, on his Test Equipment group Dr Kirkby offered a couple of out of spec cal kits at a low price.

Since these come with s-params I grabbed one of them. Shipping is very high, but ultimately still worth it for me to have items which I can use as a calibrated reference.

https://groups.io/g/Test-Equipment-For-Sale-Wanted-or-Exchange

The question is, did you pay soon enough to get one of the sets. I know I paid for one 14 hours ago.

[hendorog]

In other news, on his Test Equipment group Dr Kirkby offered a couple of out of spec cal kits at a low price.

Since these come with s-params I grabbed one of them. Shipping is very high, but ultimately still worth it for me to have items which I can use as a calibrated reference.

https://groups.io/g/Test-Equipment-For-Sale-Wanted-or-Exchange

The question is, did you pay soon enough to get one of the sets. I know I paid for one 14 hours ago.

Yeah I have one of them locked in. I was in a meeting at the time and was surreptitiously putting a payment through on Paypal while nodding my head and looking interested in what was being said. No idea what I have promised to deliver on, but hey at least I'll have a cal kit.
Then I screwed up the email address and had to cancel the payment and put it through again. It was about 23 hours ago.

[Mechatrommer]

Congratulation for your score. When i made the purchase, i wish i can have yellow case instead of black. But i choosed cheaper option of 'whatever color mostly in stock'. In the end i got... yellow case

[hendorog]

Congratulation for your score. When i made the purchase, i wish i can have yellow case instead of black. But i choosed cheaper option of 'whatever color mostly in stock'. In the end i got... yellow case

Thanks

[G0HZU]

Since these come with s-params I grabbed one of them. Shipping is very high, but ultimately still worth it for me to have items which I can use as a calibrated reference.

I'm not convinced you can treat the Kirkby kit as some form of calibrated reference if you want to start playing with Cx and Lx factors. I'm not an expert on SMA-SMA connector variability but you can't avoid having a subtle discontinuity between your SMA test cable and the Kirkby SMA cal connector because you can't expect two solid dielectric connections to mate perfectly without a tiny (uncontrolled) air gap and without some form of pin mating variability as well. So I don't know how much this will affect the phase response across a 6GHz bandwidth. A 3.5mm to SMA connection should be a lot better in this respect.

[hendorog]

Since these come with s-params I grabbed one of them. Shipping is very high, but ultimately still worth it for me to have items which I can use as a calibrated reference.

I'm not convinced you can treat the Kirkby kit as some form of calibrated reference if you want to start playing with Cx and Lx factors. I'm not an expert on SMA-SMA connector variability but you can't avoid having a subtle discontinuity between your SMA test cable and the Kirkby SMA cal connector because you can't expect two solid dielectric connections to mate perfectly without a tiny (uncontrolled) air gap and without some form of pin mating variability as well. So I don't know how much this will affect the phase response across a 6GHz bandwidth. A 3.5mm to SMA connection should be a lot better in this respect.

Hmm, that is an interesting point and something I don't know enough about to understand how to mitigate.

However, I do have some nice Gore 3.5mm M-M cables (I have been too scared to use them so far) which I can use to mate with the female type standards. Should I get some 3.5mm F-F barrels for the male type standards? I need some anyway as I have an old 8515 test set I want to play around with.

I assume the other side of this is what was used on the other end when they were swept. I can ask the question.

[G0HZU]

Also, an SMA female inner expands as it is mated so the Zo (subtly) changes depending on mating depth and male pin diameter. The 3.5mm connector has a solid outer so this problem goes away.

I just think that once you get past 2 or 3GHz you reach the point where you are trying to polish a turd if you want to try and 'calibrate' an SMA based standard using Cx and Lx corrections.

I've always had decent results with my DIY female SMA cal kit but this was only up to the 3GHz limit of my old HP8714B VNA. Also, a lot depends on how accurate it all really needs to be. I suspect that you can get away with using an SMA based cal kit up to 7GHz if you just want to make casual measurements and can accept a lot of uncertainty.

Has anyone tried doing an 'insertable device' calibration using both the male and female parts of the Kirkby kit up to 7GHz and them done some critical tests? Even the basic 6dB attenuator test would be interesting to see.

[hendorog]

Also, an SMA female inner expands as it is mated so the Zo (subtly) changes depending on mating depth and male pin diameter. The 3.5mm connector has a solid outer so this problem goes away.

I just think that once you get past 2 or 3GHz you reach the point where you are trying to polish a turd if you want to try and 'calibrate' an SMA based standard using Cx and Lx corrections.

I've always had decent results with my DIY female SMA cal kit but this was only up to the 3GHz limit of my old HP8714B VNA. Also, a lot depends on how accurate it all really needs to be. I suspect that you can get away with using an SMA based cal kit up to 7GHz if you just want to make casual measurements and can accept a lot of uncertainty.

Has anyone tried doing an 'insertable device' calibration using both the male and female parts of the Kirby kit up to 7GHz and them done some critical tests? Even the basic 6dB attenuator test would be interesting to see.

Yes that makes sense. I was going to bypass the Cx Lx issue by doing calibrations offline on a PC where I want the best precision, but it will still be limited.

While on the subject of 3.5mm  - I wonder how much better 3.5mm F-F would they be as cal standards if you used them instead of the SMA F-F bullets in post #86??
The end cap could still be SMA I assume as it won't ever be moved. Do you have any you could sweep?

They are much more expensive of course, 100 USD on digikey/mouser. But still that is cheap in relative terms.

[G0HZU]

I've got one at work but someone has borrowed it. Also, I haven't inspected it for quite a while to see how healthy it is. It's very old and has seen a lot of use. I'll try and find it tomorrow.

[G0HZU]

OK I found it at work today. I had to walk 1km into a horizontal blizzard to get it as it was being used at another company site

It's a 3.5mm F-F connector but not the same grade as the 3.5mm connectors in a typical Agilent cal kit as it uses expanding female contacts rather than the delicate solid centre (with inner petals) as per an Agilent cal kit. So it might not be that much better than an SMA F-F connector, especially as this one is very old and probably quite worn. But I've cleaned it and I'll try it later today. It's really cold and snowy here in the UK and we've been sent home from work early today and told that we don't have to go back until Monday 

I'll have an initial look at this connector once my VNA has warmed up.