Define a test toolkit: technical tools + signal generators for instruments/plugins

[grammaton]

Summary

We need a curated list of technical measurement tools and signal generators for testing the SEAM-LTM instruments and plugins — both during development (alongside the SDK-free DSP tests and the SDK validator) and for evaluating finished plugins in a host.

This needs an in-person discussion before we scope/implement it. Opening now so we don't lose the thread.

Motivation

The current verification layers are:

• SDK-free doctest DSP cores (offline, deterministic).
• The Steinberg SDK validator (interface conformance).
• Manual visual/auditory load in a DAW (e.g. Reaper).

What's missing is a shared, documented set of signal sources and analysis tools to characterise behaviour systematically — sweeps, phase/group-delay measurement, spectrum/spectrogram, null tests, etc. Several of these are directly relevant to evaluating the new hilbert plugin: e.g. measuring the −90° phase difference across the band, comparing the RBJ vs polyphase topologies, and checking the all-pass magnitude flatness on a real host signal path.

Candidate scope (to discuss)

• Which signal generators we standardise on (log/linear sweeps, impulse, white/pink noise, multitone, two-tone IMD, fixed sinusoids).
• Which analysis tools (phase-difference / quadrature error meter, group delay, spectrum/spectrogram, null/difference test against a reference).
• Whether any belong in-repo as small SEAM utilities vs. documented external tools (plugins, Reaper JSFX, Faust scopes, Python notebooks).
• How this connects to the per-plugin verification checklists and the topology study.

Notes

• Discuss face-to-face before scoping.
• Relevant to: hilbert: add FIR Hilbert (analytic-signal) topology #5 (hilbert FIR topology) and hilbert topology A/B verification generally.

[grammaton]

Concrete reference objects (from the hilbert verification session in Reaper)

While validating hilbert we leaned on two off-the-shelf tools that map directly onto the two halves of the toolkit we want to build ourselves:

1. Test signal generator (source side)

Reference: Reaper's bundled JS: Tone Generator — shape (sine/…), base frequency, note + octave + fine-tune, wet/dry mix.
What we'd want in a SEAM version: log/linear sweeps, impulse, white/pink noise, multitone and two-tone (IMD), plus fixed sinusoids with precise frequency/level — a single, scriptable source we standardise on for every plugin test.

2. XY / mid-side analyzer (analysis side)

Reference: MeldaProduction MStereoScope — goniometer/Lissajous (XY), correlation/width, and M/S level meters.
This is exactly how we saw the quadrature relationship: feeding the hilbert I/Q pair into the XY scope draws a circle (90° phase difference across the band); 0°/180° would collapse it to a diagonal line. A SEAM analyzer would give us this Lissajous + a phase-difference / quadrature-error readout we control and can cite pedagogically.

Why this matters for hilbert specifically

The on-screen max err readout matched our offline tests at both rates (48k: RBJ 1.36° / Poly 1.67°; 96k: RBJ 0.51° / Poly 2.57°), and the XY circle confirmed the −90° relationship audibly/visually. A home-grown generator + XY-MS analyzer pair would let us reproduce this characterisation for every plugin without depending on third-party (and partly paid) tools.

Still to discuss face-to-face — this comment is just to capture the two concrete reference objects before we scope.