How do I overcome this type of data break in between my TRL calibration standards? This mostly happens on my second line transition from LINE1 to LINE2 when measuring my DUT's.
The break is because your two line standards are NOT the same impedance. Please make them the same impedance. If you cannot make them the same impedance, measure their impedance and update the calibration kit with the proper impedance.
Hello Dr. Joel,
Thanks for the response. The lines are etched on the exact same substrate with the exact same GCPW ratios along with the same connectors. How can I check the differences in impedance, measure both S11 and overlay on a Smith Chart?
I see now. Is there where I adjust it on the Calkit?
That’s the value you would change. The most proper way would be to do a time domain transform and look at the reflection coefficient in the step mode and compute the impedance from that. OR, a simple way is to take the second line, adjust the impedance up 1 ohm and repeat the cal…see if the step gets bigger or smaller. Then iterate until the step disappears. The s11 of the line includes many effects besides the line impedance, most predominantly the line length and the residual load match. The time domain transform doesn’t have such issues. You can see the details in pages 274-277 of my book.
This is a place where the Marks NIST style multi-line TRL works better. All line data is used at all frequencies, only the weighting of their contributions varies. The weighting change is a continuous process, avoiding these major steps in response which are really coming from an artificially high return loss when measuring a calibration line over the band that it was used as the calibration reference in traditional TRL. E.g., if you remeasure your shortest line after a traditional TRL cal you will see highest return loss at the highest frequency band.
In this conversation Line Breaks in Multi-line TRL there was a brief discussion about NIST TRL on PNA. The Form Factor (was Cascade Microtech) product WinCal has an implementation of the NIST TRL that could be what you want. I think that there is also an implementation in Python that is also freely available.
Retrieving data ...