How do you calibrate a VNA (Vector Network Analyzer) cal kit?



How do you calibrate a VNA cal kit?
I'm talking about kits such as the Agilent/Keysight 85057B 2.4mm kit, or even a waveguide cal kit like the W11644A W-band kit. It's my understanding Agilent/Keysight just uses a common VNA which is in turn calibrated with a cal kit of exceptional uncertainty. I accept this, but I was considering a poor mans method...
If i use a non-calibrated kit to calibrate my VNA, then I use a calibrated verification kit to verify the VNA, doesn't that imply that the cal kit is good? The VNA wouldn't be able to correctly measure the verification kit without calibration from a good cal kit. The reason calibration exists is to answer the question, "how do i know it's good?" and it seems to me this method answers that.
Realistically, i would first calibrate/verify the VNA using a properly calibrated cal kit and verification kit. Then i would start over except use the non-calibrated cal kit. That way I know the VNA itself is not to blame should the verification fail.
I would say that is evidence enough, but, in addition to this, i could use a calibrated cal kit (same model as UUT) to cal the VNA, then proceed to measure the components from my non-calibrated kit. Is that 1:1? No? Then all I have to do is relax the test limits and make it at least 1:1. Label it as Limited/Special calibration, right? You might say the increased uncertainty could result in false pass components, but what does any of that matter as long as it can properly measure the verification kit components?
Any thoughts on this?


Hello there!!
Now, I first have to admit ignorance with the specifics of the process you're wanting to calibrate, so forgive any dumb questioms.

You have three items here in your measurement process : 1) the device you are wanting to verify, 2) some intermediate device (the item not calibrated), and then 3) a calibrated master/standard of some kind.

You will use 3 to verify 2; then use 2 to verify 1. Correct?

As far as uncertainty, are you building a budget? Have you estimated all the uncertainties in the whole process?

Sorry for all the questions.:eek: Hopefully I'll can help out somehow. :)


Hi BradM,
thanks for your interest in my question. I may have said too much and that complicated my question. Your understanding of my question didn't sound quite right to me. You seemed to outline a transfer standard sort of situation and I don't think that's what it is. Also, to answer your question, I have not attempted to create an uncertainty budget nor do I intend to. Creating uncertainty budgets, especially of something this complicated/advanced, is beyond me.

Rather, i was hoping that someone could tell me if the concept made sense and could be a plausible calibration method: essentially using a verification kit and VNA to calibrate a cal kit.

All I know is a cal kit allows a VNA to cancel out VNA systemic errors. Since no cal kit is perfect, I presume that the systemic error is not eliminated, rather, it is significantly reduced. How well the systemic error is reduced depends on the quality/precision of the cal kit. I believe that if a non-calibrated kit is used on a VNA to reduce systemic errors, and the VNA proceeds to make expected measurements on known/calibrated devices (a verification kit), then the non-calibrated cal kit's performance has been verified and can henceforth be considered calibrated. I've never heard of such a method being used, so I propose it here for scrutiny.


Trusted Information Resource
First, back to basics. If you know how to tune and flatten a waveguide, and remember Smith Charts (showing my age here), then you have the basic knowledge, and it is now application.

The comparison method is correct, except the continuous loop does wind up as - at best - a 1:1 as you point out.

Since you are using a 2.4, you are at much higher frequencies than X-Band. The simplest way is to get a mixer, and mix everything to 1GHz, then use an 8902A. However, you now must add several additional potential sources of VSWR. As long as that total does not exceed 0.05, you should be OK in the total characterization.

Otherwise, you need to plot Ro and Gamma of source and load, try to calculate any potential surface loss, and then you should Smith Chart manually to verify what the VNA tells you.

If you have access to another identical VNA then you can also run a separate comparison as a reality check.

This is not simple I realize, but then high-end microwave requires the rigor.

Hope this helps.


Trusted Information Resource
This is an interesting subject.

Years back I was fortunate enough to have been sent to the Hewlett Packard training school on 8510 Network Analyzer "calibration". Though the instruments have changed, the practices are basically the same.

First is the misnomer of the "Calibration Kit". This is used for measurement calibration which would be better referred to as optimization or normalization. It does correct for system errors, yielding a higher accuracy measurement. But this is not a true "calibration" standard - there is no measurement traceability through the calibration kit. You are correct in stating that the traceability goes through the Verification Kit.

One of the things HP taught us at that school was that the correct way to verify the performance of a network analyzer is by treating it as a system. That means verifying the analyzer, calibration kit, cables, adapters used, etc. as a system. That way you can discover if one of the system components was contributing to out of tolerance results.

If I went to a customer to calibrate their network analyzer and I used my own calibration kit, cables, etc. and the unit passes, that would not give an indication that it performed in tolerance with the customer's own accessories - I would be putting a sticker on a network analyzer that might not be in tolerance the way the customer typically used it. Also, there have been times that I have used a calibration kit where some element gets rejected when the kit was sent out for calibration, but the system passed when using the verification kit - the system was still accurate enough even with a compromised element in the calibration kit.

I look at the calibration kit as an item that does not require "calibration". I think of it as something like a test block that you use for setting up hardness testers - you use the cheap one for daily checks (to check for repeatability), and the certified block for annual calibrations. If you use the calibration kit to set up the analyzer and then measure a "golden" artifact you will have good confidence in the integrity of your system. As to whether or not you "need" to have the calibration kit calibrated... I think that you can come up with a strong argument to consider it an "accessory" that needs to be controlled and cared for, but not necessarily calibrated.

Some interesting readings on this subject:
Top Bottom