[P1] Inspect source, approximation, rendered audio, and spectrum #12

Parent plan: #1
Priority / gate: P1 / M2
Requirement: F08 in the PRD
Design: Technical design
Depends on: #5, #6, #7, #8, #11, #9

Outcome

Make it clear how a drawing, its mathematical representation, and the sound relate to each other.

Scope

• Use explicit Waveform (horizontal time, vertical signed amplitude) and Spectrum (horizontal Hz, vertical component magnitude/level) views. Keep oscillator colors consistent; identify any dB reference. Envelope views use time/gain and time/pitch-Hz axes.
• Build distinct source/model/rendered waveform views with accessible legends and matching phase/time/amplitude units.
• Show spectrum, sample rate, repetition or duration, tone frequency where applicable, peak, RMS, DC, harmonic count, and reconstruction error.
• Add a What you hear summary for selected source, bandwidth filtering, DC removal, attenuation, and envelope/seam processing.
• Document spectrum window, FFT length, frequency spacing, and amplitude scaling; avoid applying analysis windows to actual audio.
• Use each oscillator's stable color across waveform/pitch/gain traces; show the final mix in a neutral distinct style. Provide per-voice and mixed analysis without conflating phase and time views.
• Provide reusable labeled source/envelope/result inspection and processing metadata interfaces for [P1] Modulate sounds with imported audio or saved clips #20, preserving time/amplitude versus gain/pitch axes and oscillator identities. F18 owns the effect-specific UI integration and tests.

Acceptance criteria

• A fixed-time waveform view represents increased frequency as closer cycles; the corresponding spectrum places its peak at the higher Hz coordinate.
• Amplitude scaling increases positive and negative excursions around zero, while a vertical translation is identified as a DC offset. Spectrum magnitude and waveform signed amplitude are not mislabeled as interchangeable volume readings.
• A sine with a DC offset shows the offset in the source while the audio view identifies DC removal.
• Changing repetition frequency updates playable harmonic counts and rendered output while preserving the source coefficients.
• Waveform overlays, statistics, and spectrum come from the same revision/settings and never mix stale/current results.
• Approximation errors name the comparison source and evaluation grid; displayed rounding does not alter exported numerical data.
• Color is supplemented by labels/patterns, and equivalent text/numeric information is accessible without reading the canvas.
• Periodic analysis and finite-window spectra follow their different documented conventions.
• The selected oscillator and its pitch/gain curves are identifiable by color plus label/line cues; hiding a curve does not mute sound.

Validation

Known-frequency/amplitude/DC fixtures, cross-view consistency checks, stale-result tests, and keyboard/screen-reader review.

Boundaries

Spectrogram editing and advanced laboratory measurement claims.

Implementation evidence

Source/model/output inspection, physical-axis labels, stats, dense-grid fit errors, white-dotted reconstruction overlay, neutral numerical output and Hann spectrum are implemented. Human screen-reader acceptance remains in #15.

Merged work: #24, #28, #29.

Validated CI run · Validation report. The report arrives with the final documentation PR; numerical/browser logs already exist. Release-wide human acceptance and the recorded performance/memory gaps remain explicit in #15 and #16.