allow PySEP to use causal/acausal prefilter during instrument response removal

[bch0w]

Currently PySEP applies filtering during the instrument response removal (https://docs.obspy.org/packages/autogen/obspy.core.trace.Trace.remove_response.html), which does not have the typical ObsPy zerophase option for choosing causal or acausal filter.

pysep/pysep/pysep.py

Lines 1084 to 1092 in 3f5d7ab

	tr.remove_response(
	inventory=self.inv,
	water_level=self.water_level,
	pre_filt=_pre_filt,
	taper=bool(self.taper_percentage),
	taper_fraction=self.taper_percentage,
	zero_mean=self.demean,
	output=self.output_unit,
	)

Instead remove response applies the 4-corner prefilter in the frequency domain using a cosine taper (https://docs.obspy.org/packages/autogen/obspy.signal.invsim.cosine_sac_taper.html). To my understanding this is an acausal filter (?).

To allow Users more flexibility (i.e., to use a causal filter on pre-filtering), we can stop using the ObsPy prefilt option and instead perform prefiltering within PySEP, prior to instrument response removal. This is how it was done in 'old Pysep'

pysep/util_write_cap.py

Lines 1202 to 1223 in d5f594b

	def prefilter(st, fmin, fmax, zerophase, corners, filter_type):
	"""
	pre-filter traces. Filter options: bandpass, lowpass, highpass
	"""
	for tr in st:
	station_key = tr.stats.network + '.' + tr.stats.station + '.' + \
	tr.stats.location +'.'+ tr.stats.channel
	print("--> station %14s Applying %s filter" % (station_key, filter_type))
	if filter_type=='bandpass':
	print("pass %.3f to %.3f Hz (%.3f to %.3f s) zerophase %s corners %i" % \
	(fmin,fmax,1/fmax,1/fmin,zerophase,corners))
	tr.filter(filter_type, freqmin=fmin, freqmax=fmax, \
	zerophase=zerophase, corners=corners)
	elif filter_type=='lowpass': # use fmax; ignore fmin
	print("pass f < %.3f Hz (periods > %.3f s)" % (fmax,1/fmax))
	tr.filter(filter_type, freq=fmax, zerophase=zerophase, \
	corners=corners)
	elif filter_type=='highpass': # use fmin; ignore fmax
	print("pass f > %.3f Hz (periods < %.3f s)" % (fmin,1/fmin))
	tr.filter(filter_type, freq=fmin, zerophase=zerophase, \
	corners=corners)

This would give us the option to apply zerophase or not and control the causality of the filter. Particularly useful in cases where you want a causal filter.

Motivated by an email from @liamtoney, and discussion with @carltape

[aakash10gupta]

I think the zerophase option would also be useful with the PySEP record section plotter.

[liamtoney]

Instead remove response applies the 4-corner prefilter in the frequency domain using a cosine taper (https://docs.obspy.org/packages/autogen/obspy.signal.invsim.cosine_sac_taper.html). To my understanding this is an acausal filter (?).

According to my postdoc supervisor, the cosine taper in the frequency domain is indeed an acausal filter.