[SID-6921]

What

compute_amplitude_cutoffs()'s own num_histogram_bins default stayed at 500, even though its own docstring claims default: 100, AmplitudeCutoff.metric_params (the default used by the standard quality-metrics pipeline) is 100, and the low-level amplitude_cutoff() helper it wraps also defaults to 100.

#4415 fixed amplitude_cutoff()'s signature and AmplitudeCutoff.metric_params back to 100 after a prior regression (#4353 had bumped it to 200), but didn't touch compute_amplitude_cutoffs()'s own signature default, which was never changed and stayed at 500.

With amplitudes_bins_min_ratio=5, this silently raises the spike-count floor a unit needs before amplitude_cutoff stops being NaN from 500 spikes (the documented/intended behavior) to 2500 spikes, for anyone calling compute_amplitude_cutoffs(sorting_analyzer) directly without an explicit num_histogram_bins override. I suspect this is part of what's going on in #3605 ("much of the data I get is NaN"), though that thread settled on a different (also real) cause.

Repro

from spikeinterface.metrics.quality.misc_metrics import amplitude_cutoff
import numpy as np

rng = np.random.default_rng(0)
amps = rng.normal(100, 20, 3000)
observed = amps[amps > 85][:2317]  # a realistic, moderately-sized unit

amplitude_cutoff(observed, num_histogram_bins=500)  # compute_amplitude_cutoffs()'s actual default
# -> nan (needs 2500 spikes at bins=500, amplitudes_bins_min_ratio=5)

amplitude_cutoff(observed, num_histogram_bins=100)  # the documented/intended default
# -> 0.2318 (needs only 500 spikes at bins=100)

Fix

Set compute_amplitude_cutoffs()'s num_histogram_bins default to 100, matching its own docstring, AmplitudeCutoff.metric_params, and amplitude_cutoff(). Also fixed amplitude_cutoff()'s docstring, which still said default: 500 after #4415 changed the actual signature default to 100.

Testing

Added test_amplitude_cutoff_default_num_histogram_bins_is_consistent, which asserts the three declared defaults agree on 100, and demonstrates the behavioral consequence (same spike count: NaN at the stale 500 default, finite at 100). Confirmed it fails against the pre-fix code and passes after the fix.

Ran the full metrics/quality/tests/test_metrics_functions.py suite: 32 passed (including the new test and the two pre-existing amplitude_cutoff tests), 4 pre-existing failures (test_compute_new_quality_metrics, test_metric_names_in_same_order, test_save_quality_metrics, test_unit_id_order_independence) reproduced identically on an unmodified checkout, unrelated to this change.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota. Try requesting a review at a lower effort level.