Make the length of the frequency response adjustable

[hyperkomplexe]

When calling H, w = freqresp(filter) the returned frequency axis w is fixed at 257 rows.

This is an issue when someone (like me for example) needs a higher resolution along the frequency axis.

Matlab has an optional argument n which is used to specify the length of the axis required.
https://ch.mathworks.com/help/signal/ref/freqz.html

Scipy also has worN as an optional argument allowing this to be specified.
https://docs.scipy.org/doc/scipy/reference/generated/scipy.signal.freqz.html

I would suggest adding something similar.

I suppose it should be simple enough to implement, just add the length as an optional argument to the freqresp call.

DSP.jl/src/Filters/response.jl

Lines 9 to 15 in 3e3524f

	"""
	H, w = freqresp(filter)
	Frequency response `H` of a `filter` at (normalized) frequencies `w` in
	radians/sample for a digital filter or radians/second for an analog filter
	chosen as a reasonable default.
	"""

Then pass it also into the _freqrange call.

DSP.jl/src/Filters/response.jl

Line 154 in 3e3524f

_freqrange(::FilterCoefficients{:z}) = range(0, stop=π, length=257)

Cheers

[martinholters]

One issue here is that we have freqresp(filter::FilterCoefficients, w), where w can be anything that can be broadcast over, including single numbers. So freqresp(filter, 2) already has a meaning -- evaluating the filter at ω=2. Maybe adding a kwarg to the one-argument method of freqresp (that gets then passed on to _freqrange) might be an option, so that one could do e.g. freqresp(filter; n=1000).

(Meanwhile, the work-around would be calling freqresp(filter, range(0, stop=π, length=1000) for z-domain filters.)