AnsweredAssumed Answered

8720ES calibration problem

Question asked by linan0827 on Oct 31, 2012
Latest reply on Nov 13, 2012 by Dr_joel

We have a 8720ES network analyzer. Initially, we calibrated it with a Maury 2.92mm calibration kit. After calibration the results look good. Insertion loss with the the thru cal kit is within 0.02dB scale.

However, when we tested an RF cable with some adapters connected, the insertion loss of the cable(plus the adapters) became very strange. I attach the S21 data as a picture here(IMAG0014.jpg). It is quite strange that around 2.5GHz and 4.7GHz there is two step of the dB value of S21.

We tried re-cal the analyzer several times, also with other kind of cal kits, but the results did not improved at all. It always had a good thru characteristic with the thru standard, but a step response of S21 when measuring the cable with adapters.

Then we just checked other settings of the analyser. With the set-up above we have such settings:
Start - Stop frequency: 50MHz - 20GHz.
number of points: 201
sweep time: 400ms

We changed the number of points settings from 201 to 401, re-cal and re-measure the insertion loss of the previous cable. The result became better, with lesser step(but the step itself is still there). Also the insertion loss at the flat part decreased(from -6dB to -2dB, rougnly) See attachment IMG_20121029_152444.jpg.

Changing number of points from 401 to 801 improve the results even more.

Increasing the sweep time can also make results better. After we changed the number of points to 801, sweep time to 2s, the step is almost vanished.

Interestinglly, we did not change the IF bandwidth value during the above process. It is constant all the time.

We understand that narrow IF bandwidth and enough sweep time can improve the accuracy of the measurement, lowering the effect of noise, etc. What is confusing us is that if we have done some setting wrong, why there is such a step in the reponse, rather than some noisy response? And these steps are located at some specific frequency point.

Could anybody help us to loot at these issues, and give us some explanation? We are wondering if there is some hardware problem of the analyzer...