Inspecting Recordings

Clicking any row in the Library’s Recordings tab opens the recording inspector — a Quick View panel with six signal visualisation tabs. Use it to confirm signal presence, assess quality, and spot problems before committing recordings to a curation run.

Opening the inspector

Click Library in the top navigation and select the Recordings tab.
Use the Repository, Directory, or Branch filters to narrow the list.
Click any row. The inspector panel opens on the right (or below on narrow screens).

Visualisation tabs

Spectrogram

A time-vs-frequency power map rendered from the IQ samples. This is the fastest single view for a quality assessment.

What to look for:

Signal clearly visible as a distinct band above the noise floor
Consistent power level across the capture duration — no unexpected dropouts or gaps
Signal occupies the expected frequency range with no heavy spurs alongside it

Signs of a problem:

Flat or near-flat image — signal may have been absent, receiver was off-frequency, or gain was set too low
Bright saturation bands — gain was too high, ADC is clipping
Unexpected narrowband lines that don’t belong to the signal of interest

Constellation

An IQ scatter plot — plots real vs imaginary parts of each sample. Useful for identifying modulation type and spotting phase or amplitude errors.

What to look for by modulation:

Modulation	Expected shape
BPSK	Two tight lobes on the real axis
QPSK / DQPSK	Four clusters near the corners
8PSK	Eight clusters on a circle
QAM-16	4 × 4 grid of clusters
FSK	Ring or arc pattern depending on deviation
OOK	One cluster at origin, one on the real axis

Signs of a problem:

Blurred or widely spread clusters — low SNR or phase noise
Rotated or offset clusters — carrier offset not corrected, or DC offset present
Unexpected cluster count — wrong modulation label, or signal is a mix of types

PSD (Power Spectral Density)

A single-frame frequency view showing power as a function of frequency. Useful for confirming occupied bandwidth and spotting out-of-band interference.

What to look for:

Signal occupies the expected bandwidth with a clear spectral shape
Noise floor is flat outside the signal band
No unexpected spurs or secondary signals within the capture bandwidth

Time Series

Raw I and Q amplitude plotted over time. Use this to check for clipping and to confirm the signal envelope behaves as expected.

What to look for:

Amplitude stays within a consistent range — no flat tops (clipping) or sudden level changes
Both I and Q traces are present and roughly equal in peak amplitude
No long stretches of zero amplitude mid-capture (unless you expect silence)

Signs of a problem:

Flat tops on peaks — gain was too high, samples are clipped and the waveform is distorted
Near-zero amplitude throughout — gain was too low, signal not captured

FFT / Frequency Spectrum

A single-frame discrete FFT view — similar to the PSD but showing raw magnitude rather than estimated power density. Useful for a quick frequency-domain check on a specific moment in the recording.

3D Spectrogram

A depth-enhanced version of the spectrogram that adds a third axis for power magnitude. Useful for visualising signal dynamics and comparing power levels across time and frequency together.

Quick quality checklist

Before including a recording in a curation run:

Check	Pass	Fail
Spectrogram	Signal band clearly visible	Flat, near-zero throughout
Constellation	Clusters match expected modulation shape	Blurred to a blob, or missing entirely
PSD	Signal in expected band, flat noise floor elsewhere	Spurs, unexpected signals, or flat line
Time Series	No clipping, consistent envelope	Flat-topped peaks (clipping)

Next steps

Slice and export segments — Once quality is confirmed, use the Recording Inspector to mark keep/discard regions on the spectrogram and export clean segments to a repository.
Label recordings — Add or verify signal labels in Reviewing and Labelling Recordings.
Curate a dataset — Take reviewed recordings to the Curator to slice and package them.