I have continually noted ripple in the KC901V S21 results I get at higher frequencies for attenuation measurement, even though I have calibrated the unit for zero. The KV901V is marketed for enthusiasts, so I am probably expecting a bit much of the design for S21, as I am comparing the results to those from old top of the range HP test equipment. However the purpose of this theme is to demonstrate how S21 results can be greatly improved and the portability of this unit makes such a capability in the field invaluable. The only downside is that the solution does subtract from the units overall dynamic range however for low attenuation value testing this is not an issue.
SOME FACTORS TO CONSIDER
1/ Cables and connectors used must be of a quality sufficient for use up to 10GHz.
2/ The KC901V VSWR specifications for Ports 1 and 2 show relatively high VSWR figures at 3GHz and I can only guess at the figures above 3GHz. I suspect this is the main issue.
3/ I wondered whether 3150 sample points in calibration were enough to truly compensate for the source generators output level variations over the full range, so I conducted tests over a reduced frequency range of 5 to 6GHz where ripple is worst and hence improvements noted would be greatest.
TESTING TO DEMONSTRATE IMPROVED RESULTS
A/ Testing involved using high quality cables with a 1dB/10dB HP DC to 18GHz stepped attenuator set combination to give me the sample attenuation range. First I did tests with just the unit above as shown in picture 1
The second picture shows the result for calibration sweep (in blue) and the result after a 1dB step. Note marker 1 on max peak and marker 2 on min peak shows a total variation of 1.6dB. Clearly the ripple masks an average 1dB (approximately) step.
B/ I then added fixed attenuators 3dB on Port 1 (S21 in port) and 6dB on Port 2 (generator out) before calibration to normalise the system and minimise mismatch effects and repeated tests. The first result shows zero cal and the 1dB step with much less ripple.
This second result compares the 1dB step to the 4dB step. Results are much better though not perfect. (But for a device of this cost - amazing.)
As indicated at the start of the theme, these results are over a reduced frequency range to find out if the sample points are also an issue. So what about the before and after results over the full range?
"Normal" S21 test results showing cal result in blue then 4dB step result in yellow. You can easily see that the ripple is excessive. I suspect this is caused by the operation of the stepped attenuator which by its switching action varies the "effective" electrical length of the path between ports 1 and 2 thus rendering useless the zero calibration which removed the mismatch effects. I used the markers to show the minimum and maximum points of the ripple.
I then added the 3 and 6dB attenuators and repeated the test for a 4 then 6dB step.
The results are much smoother. Strangely the maximum point automatically selected first was a peak at the start on the sweep and I will investigate this further later.
This improvement carries over to insertion tests when using the field strength mode. It is much quicker to use if you only want results at a set frequency and proved to be quite accurate at the 5.5GHz test frequency I used for steps up to 10dB after applying the auto cal (insertion offset) calibration. The result shown below is for a 10dB step.
Happy testing- regards Noel