Repository navigation
Add method="fir" to highpass_filter and bandpass_filter - #4856
Open
galenlynch wants to merge 6 commits into
Open
galenlynch wants to merge 6 commits into
galenlynch wants to merge 6 commits into
Conversation
fir_filter(recording, taps) applies any FIR kernel, shared or per channel, by overlap-save. Its margin is the kernel's support, so its output does not depend on chunking. Channels are filtered in blocks sized to keep each block's buffers in the L2 cache. design_fir_bandpass and design_fir_highpass design Kaiser-window kernels from a stated spec: -6 dB at the cutoffs, a given transition width at each edge, and a given stopband attenuation. Keeping design separate from application lets kernels be inspected or combined before use. The defaults (175 Hz lower transition, 45 dB stopband) pass at most 0.03% of the probe's specified noise power below 150 Hz, and no more than the default Butterworth between 150 and 300 Hz, on every channel of a Neuropixels 2.0 recording. A 400 Hz lower transition passed 2.4 times the Butterworth's power between 150 and 300 Hz. check_chunk_invariance in core.testing compares chunked reads with one whole read, so any preprocessor's seam error can be tested directly.
The core test jobs run without scipy and collect every test module, so a module-level scipy import failed collection there.
method="fir" applies a Kaiser kernel from design_fir_bandpass or design_fir_highpass by overlap-save. freq_min and freq_max stay the -6 dB points, as for the default Butterworth, and transition_low_hz, transition_high_hz and stopband_db set the rest of the response. The output does not depend on chunking, so margin_ms is ignored. The recording serializes the design parameters, not the taps. IIR-only arguments raise when set with method="fir" rather than being silently dropped. The default stays "iir", since the FIR is a different filter and switching would change every user's output.
galenlynch
force-pushed
the
feat/bandpass-fir-method
branch
from
October 12, 2026 22:08
680adf3 to
4728196
Compare
This was referenced Oct 12, 2026
alejoe91
added
preprocessing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Builds on #4855, which must merge first.
highpass_filter(rec, method="fir")andbandpass_filter(rec, method="fir")apply the kernels fromdesign_fir_highpassanddesign_fir_bandpasswithfir_filter, so users get the FIR without designing a kernel. The default stays"iir". Speed and frequency response are as in #4855.freq_minandfreq_maxare still the −6 dB points, as for the default Butterworth.transition_low_hz=175,transition_high_hz=2000(bandpass only) andstopband_db=45. These are serialized, not the taps.method="fir", settingfilter_order,ftype,filter_mode,coeffordirectionraisesValueError.margin_msis ignored, because the FIR reads exactly the margin it needs.method="fir"output equalsfir_filterwith the same design exactly. The tests also cover chunk invariance, save and load, and the errors.Effect on spike templates
Four units across the amplitude range, on each unit's peak channel. The difference, magnified 10 times, is a slow oscillation near 300 Hz, where the two filters' transitions differ; each panel's scale bar reads a tenth of its label for it.
Templates of 216 units from 10 min of one Neuropixels 2.0 probe, at the sorter's spike times (up to 2,000 spikes each), after
phase_shift,highpass_filter(method=...)andcommon_referencein float32. On each unit's peak channel, the IIR and FIR templates differ by 0.27 µV rms (median), or 0.34% of the peak and at most 1.0%. The FIR's peak is 0.39% larger (median, −0.2% to 0.9%), because the Butterworth's passband sags.