IIR module gain has to be set too small

[peteasa]

I have been using the IIR module to try and understand how to configure digital filters. I have a demonstration that is easy to read here. And have been exploring resonance here. However I seem to have to set the IIR module gain to rather small values to get a working filter. Being new to pyrpl I would welcome someone who has knowledge of this area of the code to review my findings and give me a hint about how to either debug the IIR module or fix my code!

[peteasa]

I have been doing more work on this issue. I have been successful with a notch filter with

however with low pass and high pass -
resonance with a pole and a zero -
resonance with just a single pole -
these all require very low gains for them to work. I have some demo code for all these filters that I am in the process of analysing but I will publish them soon at https://github.com/peteasa/pyrpl_testbench/tree/main/filters if anyone want to play and provide feedback on this issue!

[michaelcroquette]

Hello,

I had a quick look at your example. I'm not so surprised at the gain having to be set at a very low value, the transfer function cannot be higher than 0 dB. If I take a gain=1e-8 and loops=10, I can create a decent low-pass filter. But I agree that the behavior of the gain and the loops parameter is quite strange. If I multiply the gain by 10, I would expect the dc gain of the loop to increase by 10 dB and that is not the case. Also the loops parameter changes the dc-gain and I don't think it should. The continuous tf seems to behave normally but the discrete one is shifted by 140 dB so the issue might be here.