I have a filter, which shows a mismatch between input return loss (S11) and output return loss (S22). They are not equal at the frequency of interest, which is around 36 GHz.
Would this affect the performance of the filter?
\$\begingroup\$
\$\endgroup\$
4
-
1\$\begingroup\$ It IS the performance of the filter. Do you mean 'would I expect S21 to be different from S12?' ? Would it matter if it was, as long as the S21 was suitable for your application? Is the filter really asymmetric, or are you using different length leads so the Sxx look different? Is the filter designed to be lossless? Is it designed to be reciprocal? \$\endgroup\$Neil_UK– Neil_UK2025-09-26 16:27:45 +00:00Commented Sep 26 at 16:27
-
\$\begingroup\$ Hi @Neil_UK, S21 of the filter looks flat and reasonable. I was thinking that the fact that S11 is different than S22 would indicate that there is an impeidance missmatch in the filter. It is a commerical filter so I am not sure why S11 and S22 are not symmetric markimicrowave.com/products/connectorized/filters/fh-2600/… \$\endgroup\$Amro Goneim– Amro Goneim2025-09-26 16:57:18 +00:00Commented Sep 26 at 16:57
-
2\$\begingroup\$ Now I've seen the graph that Andy has presented (which you should have presented, and I should have asked you to present), as an engineer, I would say that S11 IS EQUAL to S22, within the limitations of measurement error and device under test reproducibility in this frequency range. The apparent 'large' difference at 36 GHz is due to both measurements being good, and being presented on a log scale, as dB. Basically anything better than 20 dB return loss is perfect, and to be celebrated. 10dB is usually adequate for use in communications rather than metrology. \$\endgroup\$Neil_UK– Neil_UK2025-09-26 19:27:36 +00:00Commented Sep 26 at 19:27
-
1\$\begingroup\$ To emphasise the effect of the dB scale, consider what's happening at 31 GHz and 36 GHz. At 31, we have -12 and -14dB values, at 36 there are -23 and -30 dB values. At 31 GHz, those are 0.25 and 0.2 linear, with a best difference of 0.05, and worst (depending on phase) of 0.45. At 36 GHz, those are 0.071 and 0.032 linear, with best and worst differences being 0.04 and 0.1. The clue is in the 'sharp' dip that S22 takes at 36 GHz, once the linear figure is nearly zero (say less than 0.1), the dB value can hurtle between -20 and -infinity very quickly. \$\endgroup\$Neil_UK– Neil_UK2025-09-27 09:21:50 +00:00Commented Sep 27 at 9:21
Add a comment
|
1 Answer
\$\begingroup\$
\$\endgroup\$
I have a filter, which shows a missmatch between input return loss (S11) and output return loss (S22)
I agree; they are not very equal at 36 GHz (as per the data sheet): -
But they are both pretty good values nevertheless.
Would this affect the performance of the filter?
Yes of course but by a usually miniscule amount.
