AnsweredAssumed Answered

TRL* calibration results

Question asked by linan0827 on Mar 4, 2013
Latest reply on Aug 11, 2013 by linan0827

I designed a TRL calibration fixture for personal use. The fixture includes input/output part, and a section of microstrip line of 90 degree at 2GHz. I implemented the standards as follows:

THRU: just by putting input/output fixtures together. In standard definition it has 0 loss 0 delay.
REFLECT: putting input/output fixtures away from each other. In standard definition it has 0 loss 0 delay, a perfect OPEN.
LINE: by inserting the fixture in between input and output. In standard definition it has 120ps delay(90 degree @2GHz, on my substrate).

After I did the cal with TRL*/2port, the results look good, but with some strange behaviors.

In results1.png the simulation result and the real measurement result are both displayed. Two are very close, showing that the measurement agree with the simulation. But, the measurement result is rotating somehow, around the simulation results, which is very bad.

If the measurement results are displayed as log/phase versus frequency, it is kind of ripple, like result2.png and result3.png.

I know TRL* is not an accurate cal method, cause it is assuming a same source/load match condition during forward/reverse excitation, which it is not normally. But I am also wondering maybe I am not characterizing the cal kit accurate enough, especially for the Reflect(Open) standard. Could someone shows me some other better method to characterize my fixtures, especially those are critical for the accuracy of the TRL/TRL* calibration. What kind of suggestion could you offer based the results I have shown.

Thank you very much for any help in advance.

Best Regards,