BUG: Hard-coded value for filter length?

[nikhilchandra]

Describe the issue:

The function kilosort.preprocessing.get_highpass_filter includes the following lines:

def get_highpass_filter(fs=30000, cutoff=300, device=torch.device('cuda')):
"""Filter to use for high-pass filtering."""
NT = 30122

What is the origin of the value 30122? This looks like 30000 + 61 * 2 to me, which would be the Neuropixels sampling rate + 2*(default spike width of 61 samples).. But what if I am using a different sampling rate and spike width? For example, Plexon equipment records data at 40 kHz. Forgive me if I am mistaken, but would this be a problem here?

Reproduce the bug:

Error message:

Version information:

N/A -- just looking through the code

[jacobpennington]

That is where the value comes from, but it's not a problem to use the same filter length for different sampling rates and spike widths for that step in the pipeline.

[nikhilchandra]

Okay, thank you.