ACP Consultants

ACP scientific tools · Version 1.0 · 17 November 2026

Keep the sample rate.
Understand the signal.

Waveform, spectrogram and measurement records from original-rate WAV samples.

T12 · Measurements with provenance

Three views. One source.

Inspect amplitude over time, frequency through time and the mean spectrum of a selected interval. The workbench reads WAV sample values directly without browser audio resampling.

Formats, analysis definitions and step-by-step instructions →

01 / Preserve the source

Open a compatible original-rate recording.

Little-endian PCM16/24/32 or float32; mono/stereo. Select up to two seconds for each analysis. No recording is uploaded.

No WAV opened.

The tool cannot detect a header changed for slow listening or infer a time-expansion factor. It never silently multiplies frequencies.

02 / Select a reproducible window

Choose time, channel and frequency band.

Periodic Hann window · no zero-padding · no detrending · complete frames only.

Try a known example first. The source sample rate controls every time and frequency axis.

03 / Inspect the signal

Read the charts together.

Waveform, spectrogram and spectrum will appear after analysis.

On a small screen, scroll the chart horizontally to read its axes. Controls and summaries remain within the page width.

Read the measurement definitions and exact values

Waveform uses a minimum–maximum display envelope; metrics use all selected samples. RMS and rail fraction cover the selected channel across all frequencies. The band peak comes from mean unnormalised STFT power. The spectrogram uses its displayed band’s maximum; the mean spectrum uses its full-spectrum maximum.

04 / Record your judgement

A measurement log, tied to the source.

Each entry retains its settings, source fingerprint and selected-region duration.

A manually chosen region is not automatically a call. Its duration and band peak must be described in context. The log does not calculate species labels or count animals.

Practical questions

Why not analyse the slowed sound used for listening?

A listening copy may change the stated sample rate or time scale. Its displayed frequencies and durations then describe that copy. Use the original-rate recording for source measurements and document any transformation.

Does a larger FFT always improve the result?

No. A longer window makes the bins closer in frequency while spanning more time. Short changes can become less distinct. Window shape, noise and signal duration affect practical resolution; the bin spacing alone is not accuracy.

Can I identify bats from these plots?

These are measurement aids. Identification needs the call sequence, recording context, overlapping possibilities and competent review. Return to the R/Python resource for reusable analysis scripts and its links to acoustics and species guidance.