Kirkby calibration kit alternatives?

[Gandalf_Sr]

I got a floppy disk with calibration data on it when I bought the cal kit from Kirby.

[KE5FX]

OK, thanks for the advice.  I'm travelling right now but I will try to re-calibrate my VNA and then ask the questions along the lines of - Am I doing it right?

I have no idea why they're not on YouTube, but there are some videos and other material here that should be helpful:

http://rfic.eecs.berkeley.edu/142/labs.html

[technogeeky]

Does anyone here know how long it would take to record all of the relevant calibration coefficients, if someone purchased a open-short-load and sent it to another forum member with access to a good/great VNA?

At the end of the day, when is the modeling necessary? If it's only when you are making your own and trying to improve your design of DIY devices, then we can just bypass that step and send devices to people who can make the measurements for a small fee?

[Mechatrommer]

equipments needed should be in 5 digit price. the man should have access to company's equipments, or he must already invested his own money for the cost or borrowed from friend. not to mention knowledge need to be gained so he can answer all customers inquiries about the field. how many VNA customers potentially buying the service? 10? 100 a year? $600 seems reasonable already, imho....

fwiw, among other things, attached is attenuator plot provided from Kirkby to me. the result i got on my VNA is much noisier, so i know something went wrong. i'm ordering stuffs to further verify my VNA condition if its fits for measurement etc later. i hope they will not get lost during CNY...

[hendorog]

Does anyone here know how long it would take to record all of the relevant calibration coefficients, if someone purchased a open-short-load and sent it to another forum member with access to a good/great VNA?

At the end of the day, when is the modeling necessary? If it's only when you are making your own and trying to improve your design of DIY devices, then we can just bypass that step and send devices to people who can make the measurements for a small fee?

Yes the modelling would not be required to transfer a calibration from someone elses' cal kit.

In theory, all you need to do is send them the items you want to use as calibration standards and adapters. They just need to generate an S parameter file for each one on their calibrated VNA.

Then you run it through the script to determine the coefficients. How well that step works is unknown.

If you also sent a decent attenuator then that could be used to compare your new calibration with the calibration of the other user.

So they would need a VNA and a good cal kit. Then they would do a calibration first, and then make 5 or 6 measurements, transfer them to a PC and email them. Then ship back the items.

[hendorog]

fwiw, among other things, attached is attenuator plot provided from Kirkby to me. the result i got on my VNA is much noisier...

He has reduced the IF bandwidth from the default. That would reduce the noise. What VNA have you got?

[Koen]

I just wanted to point out these male SMA references available for about $20 in the USA.

They manufacturer link goes back to mRS miniVNA Tiny and Pro's website.  After more and more searching, it seems like these guys are the manufacturer

Amphenol 132360, 132331, 202112 maybe. I have both the kit you linked and these very-cheap three but I'm not knowledgeable enough to say anything except they look very similar.

[technogeeky]

Does anyone here know how long it would take to record all of the relevant calibration coefficients, if someone purchased a open-short-load and sent it to another forum member with access to a good/great VNA?

At the end of the day, when is the modeling necessary? If it's only when you are making your own and trying to improve your design of DIY devices, then we can just bypass that step and send devices to people who can make the measurements for a small fee?

Yes the modelling would not be required to transfer a calibration from someone elses' cal kit.

In theory, all you need to do is send them the items you want to use as calibration standards and adapters. They just need to generate an S parameter file for each one on their calibrated VNA.

Then you run it through the script to determine the coefficients. How well that step works is unknown.

If you also sent a decent attenuator then that could be used to compare your new calibration with the calibration of the other user.

So they would need a VNA and a good cal kit. Then they would do a calibration first, and then make 5 or 6 measurements, transfer them to a PC and email them. Then ship back the items.

I don't even own a VNA (though some day I would like to). But I would like information on how good a particular cal kit is, so I can understand the limitations of what measurements I can make with e.g. spectrum analyzer with tracking generator (where you normalize, so this is less important); but more importantly what the (lack of very good cal kit) reference imposed limitations are on the equipment I own (in this case, the Agilent E7459a). Most importantly, measurements of one port insertion loss and one port return loss should be calibrated, but the equipment I have was probably calibrated against a very good cal kit (more then -40dB across the almost 3 GHz spectrum) and I would like to get a sense of how far off this the instrument is, in its current state (last cal was 2011 due for re-cal in 2013). All of this in addition to using such devices for measurements.

I only paid $275 for the E7495a, and I certainly can not afford anything near the cal kit that would have been included which would probably cost $1000 or more now.

[Mechatrommer]

fwiw, among other things, attached is attenuator plot provided from Kirkby to me. the result i got on my VNA is much noisier...

He has reduced the IF bandwidth from the default. That would reduce the noise. What VNA have you got?

Deepace KC901V. i'm start to think its a toyish VNA, but i'm not done testing yet.

[G0HZU]

fwiw, among other things, attached is attenuator plot provided from Kirkby to me. the result i got on my VNA is much noisier, so i know something went wrong. i'm ordering stuffs to further verify my VNA condition if its fits for measurement etc later. i hope they will not get lost during CNY...

Is that attenuator (reference) plot taken with the 8720D VNA and your cal kit? Or is it a 'best' reference taken of the attenuator with the 8720 and a Keysight/Agilent cal kit? See the text below from the Kirkby website.

The Kirkby Microwave 7 GHz 85033 SMA calibration and verification kit is by far the most popular kit we sell. It can be used to calibrate virtually any network analyzer to enable SMA, 3.5 mm or 2.92 mm coaxial devices to be measured, irrespective of the test ports on the VNA (N, 3.5 mm, APC7, 2.4 mm etc). The connectors on the calibration kit should match the device under test (DUT) and not necessarily the connectors on the VNA. The frequency range of the SMA coaxial calibration kit is from DC to 7 GHz, although higher frequency versions can be built to order.

The kit is both a calibration and verification kit, since it supplied with an attenuator which has been measured on an HP 8720D (20 GHz) VNA using a Agilent 85052B 26.5 GHz 3.5 mm calibration kit. Therefore after setting up the calibration kit, calibrating the VNA, the user can check the VNA is performing properly by measuring the attenuator which is supplied, and comparing that to the measured results, which are on a USB stick supplied with the kit.

If a calibration kit is made using SMA F-F bullets and end caps there are a few things that will limit its performance. There will be some (extremely tiny) added inductance at the end of the SHORT and also the SMA bullet won't have a perfect 50R Zo. The SMA bullet is about 42ps in terms of delay and this is longer than a regular 85033 cal kit and so this can make any phase variation (due to imperfect 50R Zo) appear slightly worse. The 85033 cal kit uses precision 3.5mm connectors and an air dielectric and it is much better suited for use up at 7GHz.

It doesn't take much to get 4 degrees of phase error over 6GHz in a DIY cal kit. If you get several degrees of phase error and it isn't corrected for by the cal kit corrections then you can expect to see that reference plot degrade a lot by 5GHz and it will get quite noisy by 7GHz. Also, an inline test of a 3dB attenuator isn't a very critical test.

I can show you what my old DIY cal kit can do to 7GHz but I'll have to generate a polynomial to correct the SHORT. Otherwise it will only be good to about 3GHz. I'll have a go at changing my simulation/model to produce L0, L1, L2, L3 correction factors for my old cal kit. It should then do quite well in an attenuator test.

However, I see this process as polishing a turd once you get up to 7GHz. Much better to have a decent 85033 cal kit here.

[G0HZU]

I don't have any precision 3dB attenuators here at home but I have measured a basic SMA 3dB attenuator on my VNA and the data is below. This is after a full 2 port calibration with an Ecal module. The scale is 0.2dB/div and I used a BW of 500Hz. Note that this is after an Ecal calibration and not after a calibration with my DIY cal kit. A plot after calibrating with my DIY cal kit would have ripple on it up at frequencies above 3GHz.

The attenuator isn't very accurate and is more like a 2.8dB attenuator at LF but it is rated to a high frequency.

But you can see how low the ripple and noise is after a full 2port calibration for S11 S21 S12 S22. Note that I'm using Gore cables here that would have cost a 4 figure sum each here and a $13k Ecal module and a fairly decent VNA.

However, a 3dB attenuator test isn't exactly a demanding proof of the integrity of a VNA calibration. There are much tougher tests than this that will show up any sloppiness in the cal kit corrections.
Even an error of 0.2ps in a cal kit delay correction can be highlighted with some tests.

[G0HZU]

I entered a polynomial for the Lx corrections for my DIY cal kit SHORT and got the result below on the simulator after a quick manual tweak. It could be better than this but I'm not really that bothered to polish it any further.

I then did a full 2 port calibration using the DIY cal kit (rather than the Ecal) and remeasured the 2.8dB attenuator. So this was calibrated using my first attempt at some decent cal kit correction factors for my DIY kit. Note that this cal kit also includes my Cx corrections to get a similar flat response on the OPEN to 6GHz.

See below for the result. This is actually better than I expected. If I hadn't done the Lx corrections the trace would have a LOT of ripple on it by 5GHz.

[G0HZU]

Here's a plot of what my DIY kit looks like with only crude corrections in it. There are no Lx corrections for the SHORT and only a basic C0 correction for the OPEN. This would still be good to 3GHz as long as the various cal kit correction factors for DELAY and LOSS were correct.

That's why I think anyone could make a decent DIY cal kit to 3GHz assuming they are prepared to tweak/optimise their cal kit after some tests that help to optimise it.

I could put together a cal kit file with this data in it and show you the 3dB test if that helps? It would have some obvious ripple above 3GHz. However the 3dB test is a really soft test anyway.

[Mechatrommer]

By the way... can anybody explain how the "reflection bridge" of the KC901V exactly works?
I guess it is the structure that looks like a wilkinson power divider?

its a power splitter as described by TheSignalPath. there must be detailed explanation, concept and math behind it. i will read them up when i want to make my own diy VNA or if i figure out my VNA is broken and need a tear down. fwiw i guess this type of topology that makes KC901V gives false reading esp impedance value at very low end of frequency 5KHz - 1MHz or so.

fwiw, among other things, attached is attenuator plot provided from Kirkby to me. the result i got on my VNA is much noisier, so i know something went wrong. i'm ordering stuffs to further verify my VNA condition if its fits for measurement etc later. i hope they will not get lost during CNY...

Is that attenuator (reference) plot taken with the 8720D VNA and your cal kit? Or is it a 'best' reference taken of the attenuator with the 8720 and a Keysight/Agilent cal kit?

he claimed he tested all the cal kit sold to me by his own hand. incl the attenuator. so the plot you see should be unique only to my attenuator. i expect other kirkby cal kit owner should have different plot. it took long time (about a week or more) for the calibration to be made, he claimed he do this process in some quantities, to quicken his process. going to lab with my cal kit alone to do the process will take him even longer time, i take and understand this as his economical point of view. furthermore he was not well when i made the purchase i almost cancel the order because it took so long for him that i thought he gave less interest in my order. but i'm glad to support him.

If a calibration kit is made using SMA F-F bullets and end caps there are a few things that will limit its performance. There will be some (extremely tiny) added inductance at the end of the SHORT and also the SMA bullet won't have a perfect 50R Zo. The SMA bullet is about 42ps in terms of delay and this is longer than a regular 85033 cal kit and so this can make any phase variation (due to imperfect 50R Zo) appear slightly worse. The 85033 cal kit uses precision 3.5mm connectors and an air dielectric and it is much better suited for use up at 7GHz.

It doesn't take much to get 4 degrees of phase error over 6GHz in a DIY cal kit. If you get several degrees of phase error and it isn't corrected for by the cal kit corrections then you can expect to see that reference plot degrade a lot by 5GHz and it will get quite noisy by 7GHz. Also, an inline test of a 3dB attenuator isn't a very critical test.

i'm not sure how to reply to this, just as how i'm going to reply to Kirkby in my last test, i'm not knowledgable much that i need to gain some info, do some test (or even need to have access to much expensive gears) to be able to comprehend some of the meaning. i only can hope or rely on some expert to do the testing and give some promise that it should work.

btw attached (1st file) is my translation of the parameters extracted from kirkby's data relevant to my kit and VNA. for open standard, he provided varying offset delays (that i translated from offset delay ps metric in his file into mm). its varying in sub ps range or max deviation (not std deviation, maybe variance) of 0.162 mm throughout the BW. so i just prefer the mean value of 17.357 mm @ 3.5GHz (57-58 ps) to punch in the VNA if i want to do quick testing. fringing capacitance effects should already be lumped in the figures during calculation or modelling). as for short he only provided a single offset delay so i suspect the variance due to inductance effect should be negligible? i never heard of fringing inductance effect for short standard he should provide the complete data if this is a concern. anyway this only based on trust to the expertise because i'm not at his location during the modelling. if he lied, then i lied

the 2nd attachment is the zipped s1p for my SOL kit, i dont know how to read this and not sure if its relevant to this discussion, i need time to gain on this. 3rd pdf file is my latest test report to him to seek advice. you can see S11 and S22 plot of the terminated attenuator (after calibration with his SOL kit) are quite spot on except at the area of notches which he claimed of high uncertainty even if measured with his 20GHz VNA since attenuator is terminated hence reflecting very little signal for processing. his arguments are quite rational though... and i can accept a $2K VNA will be of no match to the $17K VNA. the concern in the report is S21 you can see the noises i was talking about. he stated the attenuator should not jump up and down 5dB variance in very short BW range. i'm yet be able to reply on this before i can do more testing..

However, I see this process as polishing a turd once you get up to 7GHz. Much better to have a decent 85033 cal kit here.

that Agilent cal kit if well within spec and in full complete data, should be about 2-3X the cost of my VNA. i can see used price is $2K already without any data. i would not give my money betting on that. ps: i'm not sure why Kirkby called his cal kit as 85033E similar to agilent naming. maybe its tracable to the agilent cal kit he owned? and his cal kit is modelled/compared after that? i'm not sure.

[Mechatrommer]

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...

[G0HZU]

I think your problems may be due to the type of VNA you are using. I've not seen one of those models before and it may have quite poor performance.

Can you post up your Kirkby cal kit correction file or list its corrections for Cx and delay etc?

I tried creating a model based on the default corrections for delay and Cx on his website here:
http://www.kirkbymicrowave.co.uk/support/Kits/85033/

...and then I compared this to the OPEN S1p file I found in your zip files. I assume this S1p file came from Dr Kirkby as it looks like a decent measurement to me. When I compare my model to his s1p file the phase response is very good. I'm using the Cx and delay from the web link above but you may have a tweaked version of this with your cal kit? Either way the phase response is very good and I won't post it up here because I don't want to give the impression I can measure this stuff more accurately than DrKirkby. But I will say that the results are very close! Not much to worry about here I think..

I also looked at the SHORT s1p file and the results are very close to the model that is based on his default cal kit corrections on the website. This is very good indeed and it kind of shows that the SHORT in his kit is a good one I think.

I can suggest subtle differences in the default Cx corrections but you probably have this info already. His kit data looks VERY good to me

[Bud]

Do not use crappy Chinese connectors directly on your VNA connectors, you will be sorry but it will be late.

"Less quality cal kit" - Dr. Dunsmore of HP/Agilent/Keysight, the 8753 designer, was saying he could calibrate a VNA using a brick, as long as the brick is accurately characterized  with calibration coefficients.

[G0HZU]

"Less quality cal kit" - Dr. Dunsmore of HP/Agilent/Keysight, the 8753 designer, was saying he could calibrate a VNA using a brick, as long as the brick is accurately characterized  with calibration coefficients.

I think my Ecal module 'works' because it has a lot of correction coefficients. Otherwise it would be only useful as a paperweight...

[G0HZU]

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.

For sure you will get more ripple with an inferior cal kit or with one that has not been corrected properly. I can make my ENA VNA look REALLY bad if I deliberately use a poor cal kit.

But I think the Kirkby kit looks OK based on the bundled S1p files. The SHORT, OPEN and LOAD s1p files look very good to me against a model based on the default cal kit data. Assuming you are setting up the transmission calibration correctly you should get very good results with minimal trace ripple in the S21 and S12 attenuator test.

Note that I don't get as good a phase response as that from an Agilent/Keysight 85033 kit though. I get a virtually perfect flat phase line with the Agilent 85033 kit for the OPEN if I use the Agilent cal corrections for Cx etc in my comparison model. Note also that I'm just doing this in my workroom at home and using an RF simulator to create the comparison model. But my model seems 'perfect' when compared with the decent $$$ Agilent cal kit data and a measurement of a real 85033 cal kit on an HP8753ES as I get a flat line for the phase as per the second plot in post #15

I'd be happy to review a Kirkby cal kit and I now suspect the results will be better than I first thought. However, I'd be giving it some fairly strict tests, way beyond this simple 3dB attenuator test. I'll try and borrow one of these kits from my ex colleague and put it through its paces

[G0HZU]

I created a flawed DIY SMA cal kit correction file with performance similar to the phase plots in post#37 and see below for the result of the 3dB attenuator test after my VNA was calibrated with the same DIY SMA cal kit but it now has flawed correction data for Cx and Lx.

You can see that the response has ripple above about 4GHz. Note that this has the thru delay correct and also has the loss and the basic delay correct up to about 3GHz. So it works quite well up to about 3GHz.

[G0HZU]

I can have a play at making a quick and dirty cal kit using commercial SMA F-F bullets and a shorting end cap and a commercial SMA load and show you what happens if you just try and guess the delays based on caliper measurements and/or the typical delay in an SMA F-F barrel? This would represent about the crudest cal kit you could make for maybe <$20. The performance won't be very good but it might be fun to try this. My current SMA DIY kit uses precision end launcher connectors and this needed some adjusting with a rotary tool to get good results. But the quick and dirty cal kit described above would be made in the time it takes to screw the SMA connections together...

It might not be tonight though. Any interest in this?

[Mechatrommer]

Can you post up your Kirkby cal kit correction file or list its corrections for Cx and delay etc?

maybe the attached file is what you looking for. its for agilent VNA if using my cal kit with them. its not relevant to my VNA though.. fwiw...

Note also that I'm just doing this in my workroom at home and using an RF simulator to create the comparison model.

what rf simulator? is this simulator accessible to mortal? can i learn? ... for the rest of your post, i need time understanding that...

I think your problems may be due to the type of VNA you are using. I've not seen one of those models before and it may have quite poor performance.

now isnt that golden? (from your link above http://www.kirkbymicrowave.co.uk/support/Kits/85033/)



its been announced in public

[G0HZU]

I'm using an old version of Eagleware SW for the simulator. It's old and obsolete but it supports some features that aren't in the later Keysight version of this SW and I'm using one of these (script/equation based) features to model the Cx and Lx coefficients. So it won't run on the latest version below.

https://www.keysight.com/en/pc-1297125/genesys-rf-and-microwave-design-software?cc=US

I checked your custom correction data for your specific Kirkby cal kit and the results for the OPEN are very good when tested for the simulator mode against the measured/supplied S1p data for the actual OPEN in your kit. I get a flat line as below

The results for the short aren't as good because there aren't any Lx corrections in the data. But it is still good considering it is using SMA based connectors in the SHORT. It would be easy to correct this for VNAs that support Lx coefficients and if a VNA supports this feature then I guess these would be included in the data?

Note that my attempts to model this stuff are fairly quick and basic. But because the results for the OPEN are spot on for both the Keysight 85033 and the Kirkby kits I suspect I've got this nailed OK

[technogeeky]

...

Just to make sure you've seen it, have you ever looked at this
 guy's this DIY SMA female and this SMA male open/short/load kits? I notice that he never publishes the phase charts that you are showing.

[G0HZU]

Thanks. It's interesting to see how people approach this in different ways. I had a look at the links you gave and the IN30TD cal kit corrections have some very odd values in them. I guess there are no rules here but the numbers do seem very strange.

Back when I made/optimised my DIY kit I wanted flat phase to 3GHz and low loss in both open and short. I went to fairly nerdy lengths to achieve this so I could measure/model SMD parts accurately in a test fixture with my old HP8714B VNA. I kept on refining it until it gave the performance I wanted.

My cal kit cost nothing in parts because I salvaged the SMA parts but it cost a lot in time spent optimising it both electronically and with a rotary tool.