Why are the Multile TRL Calibration Results displaying some strange output?

[Md. Imtiaz Kamrul]

Hello, Good day, everyone. This group was extremely helpful to me. 

Today, I'd like to address one of my concerns regarding multiline TRL calibration. After calibrating a frequency, I occasionally did not get the expected result. If you look at my attachment, you'll notice that after calibration in some frequency ranges, it shows a sharp curve above 0dB (sometimes).

Is there anyone who can say me why this type of error occurred?

[Ziad]

Hi!

Such artifacts are typical for mTRL using the common line procedure discussed in [1]. Basically, at each frequency point the algorithm is trying to decide which pair combinations of the lines are the best. If you don't correctly design your line lengths, such discontinuities will arise. I recommend you to design your mTRL kit based on the microwaves101 guide: https://www.microwaves101.com/encyclopedias/trl-calibration 

One thing you can do is to check the normalized standard deviation. This is basically the elements of the covariance matrix the algorithm uses to combine the results of the pairs. If you see spikes there, then the problem is indeed the ill-designed lines. Skrf mTRL already offer this parameter. you check the example and explanation here: https://scikit-rf.readthedocs.io/en/latest/examples/metrology/Multiline%20TRL.html 

Lastly, if you want to test your data with a different mTRL implementation, you can try my code: https://github.com/ZiadHatab/multiline-trl-calibration

In my repo I offer two mTRL implementations: i) my own mTRL implementation (different approach from [1]). ii) the traditional multical mTRL (i.e., [1]). Try both of them and see if you get the same results as in skrf. Btw, the notation I use is similar to skrf, but not 1:1 the same (check the example).

hope this helps!

Ziad

[1] Multiline TRL revealed, https://www.researchgate.net/publication/4027015_Multiline_TRL_revealed 

[Md. Imtiaz Kamrul]

Thank you so much for your reply. I'll check your code to see if I can figure out what's wrong.