Marco Berzborn
Marco Berzborn
I just triggered a re-run of the tests. Let's see what we get.
Seems like it was due to an issue with CircleCI, everything is fine.
Also noticed in #639
For the multichannel base class it would also be nice to implement an analogue to [np.squeeze](https://numpy.org/doc/2.2/reference/generated/numpy.squeeze.html).
I'm not sure what the impulse response thing has to do with this issue? I think we should focus on the problem ambiguous input parameters of the filter functions.
Wouldn't it then be easier to simply deprecate and remove `signal` as an input parameter from the functions we have? ```python lp = pf.dsp.filter.butterworth(N, frequency, btype='lowpass', sampling_rate=44100) filtered = lp.process(signal)...
It may indeed be something to consider changing in a major version jump further down the road.
The currently favored option is to implement child classes for specific filter types as suggested by @sikersten: > I would suggest to add a class for each of the filter...
Thanks for your suggestion. I however have the same impression as @sikersten here. The use case for filters which are re-used more than once is to return and store the...
Hi @jf-som, I'm not sure if I'm fully getting what you mean. The idea behind the persistent filter class would be more like this: ```python def process_signal(signal, filter): signal =...