As far as I know, measurement errors introduced by the test-fixture is systematic, the calibration process is a means to cancel the systematic errors. So I think whatever the test-fixture is, as long as it does not introduce too much attenuation/ resonation such that the signal power is below VNA’s sensitivity, the VNA shall obtain reliable measurement results.
I am now trying to design a custom test-fixture and its corresponding TRL calibration kit. But I find the test-fixture has great impact on the TRL calibration. I don't know what went wrong. Can you help me to analyze this problem?
Yes, VNA calibration removes systematic errors such as loss of the cables and test fixture, reflections from the cables and test fixture transitions. The accuracy of a calibration is dependent on how good the setup is before calibration (the smaller the reflections in the system and lower the loss) and how well the mechanical repeat-ability is. So in the fixture design you want to have good transitions from your co-ax cables to your fixture. This means good quality connectors and you want to pay attention to the RF ground connections as discontinuity in the ground plane is a major problem in transitions. Obtaining repeatable results is a function of the mechanical precision of the test fixture - how accurate can the connector and test fixture connections be made. This last factor is the difficult one to get right in a good system.
Dr. Cornell Van Niekerk, yes, certainly, the lower the loss before calibration, the better. But does that mean the test-fixture topology will affect the calibration dramatically? As in my case, I am merely simulating it in the schematic, so the loss, connector, repeat-ability etc. should negligible, the only thing that seems to affect my calibration might be the test-fixture change. As long as it is systematic, why does it still has such a big influence on my calibration simulation?
One of the key factors for a good calibration is how small a reflection can be measured after calibration. This factor is influenced by what is referred to as the effective directivity of the couplers inside the VNA that is separating the forward and reverse travelling waves. Any reflections from your test fixture is limiting this performance level. When measuring passive networks, this will also influence the transmission accuracy and will appear as ripple in your S21 and S12 measurements.
As long as the loss in your simulation is moderate, I would focus on reflections as this is more of a problem than loss in calibration. Most sensitivity analysis on calibration will even show loss to help with certain accuracy aspects, as long as it is not so much that the dynamic range of your system is influenced.
For background material on VNA calibration accuracy refer to Keysight and the publication "NETWORK ANALYZER ACCURACY OVERVIEW" by Douglas Kent Rytting.
- Badges
- Science topic
- Similar topics
- Biological Science
- Bioecology
- Systems Ecology
- Landscape Ecology
- Connectivity