We are facing a problem with our E8362B network analyzer. the problem is:
1. S21 is showing values in positive (some times very little like 0.005 dB and some times higher values like 0.7/0.8 dB).
2. Also, S11 is some times showing values which are not accurate and some times positive values.
3. Also, it was observed that some times just after calibration, either there was no error or it was very little but the next day results got bad with the very same calibration.
The temperature was in the range of 35-28 degree Celsius at different timings.
As per procedures, calibration was done at several different times but the problem is persistent.
The operator check was also performed satisfactorily but the issue did not resolve.
The NA cables were also changed, but the result did not show consistency; some times less error or no error but next time bad results with the very same calibration.
Almost without question, you have unstable cables. To verify the stabilty of the VNA, do a 1 port calibration directly on the hard-port (physical test port) of the VNA, with no cables of course, and put a short or open from a calkit (don't just leave it open, the radiation is unstable in S11). Put Data->mem and data/mem on S11 and S22. Leave it overnite and see if there is any change. If there is no change, it really is probably the cable.
To test the cable: Place the cable on one port of the VNA, set a short or open from a calkit on the other end. Don't cal. Just do data->mem and data-mem (that's right, data -minus- memory). Look at the result in dB. Immediately after the cal, the trace should show -70 or -80 dB return loss. Now flex the cable back and forth and up and down. The trace for a "GOOD" cable should be -50 dB; if the highest spot on the trace is above -30 dB it is a "BAD" cable: take a wire cutter and immediately cut the cable in 20. For me, I would not use a cable that is worse than -40 dB.
IF the cables are good, and the VNA is stable: the last place to look is at the VNA-to-cable connectors. If this comes loose and then the operator re-tightens it, or if it is not tight at the start, it can cause this kind of trouble.
The other recommendation is to enable Stepped Sweep and/or drop the IFBW to 1 kHz (the default is 35 kHz). This will clear up quite a few of the positive S21 measurements after calibration when everything is passive.