# What's the interpretation of this Low pass filter spec?

From the filter characteristic mentioned in the table, is it correct to interpret following for LPF: Fs=192Khz Passband edge=43.8Khz; passband ripple=5mdB Stopband edge=112Khz; stopband attenuation=72dB As per my understanding stopband edge has to be less than fs/2(because that's the max nyquist freq and we draw freq response from 0 to fs/2), in this case it should be 96Khz. But here its mentioned as 112Khz. Is the spec wrong or my understanding is wrong?

• What note 8 says? Commented May 21, 2022 at 17:50

For a real filter the unique frequency range is from $$f=0$$ to $$f=f_s/2$$ (Nyquist). However the stopband edge for the anti-alias filter can be beyond Nyquist as I detail further in this post. The primary concern is not about avoiding any aliasing anywhere in Nyquist but specifically to not allow aliasing into our occupied BW, as set by fpass.

We can assume the usable passband is from 0 to 43.8 KHz given the specification (possibly higher if some passband roll-off within our signal is acceptable), assuming that, with 198KHz sampling we want a stopband to start at least at 198 KHz - 43.8 KHz = 154.2 KHz plus some margin for realizable digital filtering downstream for final signal selection. A 112 KHz stopband is more than adequate for this purpose.

As David mentioned in the comments, the stopband corner would be better specified as a max frequency, not a min frequency; we want it to be at least the value given and not higher.

• I would add that the stopband frequency should be a maximum rather than a minimum. Commented Feb 15, 2023 at 21:07
• @David ah thank you! You made me think about this old post and I need to revise my answer —- the stopband can and typically is above f_s/2 as I detailed in another post I will link. We’re not concerned about aliasing anywhere in Nyquist but concerned about aliasing into our occupied BW as set by fpass. Commented Feb 15, 2023 at 23:13