[LOG 04] — Controlled Drift Test Suite: Formant Stability Across 3.2s Emotional Envelopes
Date: 2025-03-10
Time: 04:37 AM EST
Session ID: VF-0325-ΔCT-DRIFT4
Engineer: Z. Kanwar-Torge
Build: v0.82b (Unreleased)
Input Chain: Rode NT1-A > Focusrite Scarlett 2i2 Gen 3 > 44.1kHz/24bit Mono WAV
Test Prompt: “I am speaking without emotion. I am still speaking. I am no longer speaking.”
Duration: 3.237 seconds (manual trim applied post-capture)
⸻
OBJECTIVE
To measure the persistence and deformation of detected formant clusters across a fixed emotional trajectory, using constant SPL and neutral inflection, in the presence of an active Microformant Drift circuit (Intensity: 64.8%), Ultrasonic Emotion Drift (Frequency: 25.7kHz, Depth: 0.036), and Recursive Decay enabled (Decay constant = 0.97 static, no modulation).
⸻
PROCEDURE
1. Room conditioned to 21.0°C at 36% humidity, with floor/ceiling reflection minimized via foam hex-damping (12-point perimeter).
2. Vocal input recorded at 14” mouth-to-mic distance with static head angle (confirmed via 3-point chin calibration brace).
3. No vocal warmup prior to capture.
4. Software input buffer manually confirmed at 512 samples, fixed.
5. Plugin interface settings:
• Formant Drift: 64.8%
• Sentiment Drag: 18.7%
• Recursive Decay: 23%
• Bleed Depth: 61.5%
• Prosodic Reweaver: Disabled
6. Output rendered to floating-point .wav and visualized in SonicVisualiser v3.4
⸻
OBSERVATIONS
• Initial formant lock held stable for 0.81s.
• Drift began at the 0.93s mark, slightly earlier than projected.
• By 1.17s, the F2 band was consistently ~42Hz higher than baseline.
• No detectable SPL spike or vocal strain justifying this shift.
Waveform behavior:
• Output showed a pre-ring echo just after silence—lasting 211ms.
• This was not present in dry signal.
• Wavelet overlay confirmed the echo followed the emotional shape of the spoken content, not the phonemes.
Visual lattice (Emotive Harmonic Grid):
• Peak intensity plotted at (X: 2.97s, Y: ~1450Hz), corresponding to a non-vocalized memory bleed harmonic.
• Lattice cell did not resolve fully in decay cycle—suggests sustained re-entry.
⸻
NOTES
• Prosodic shift occurred even with the reweaver disabled—implies the phase-drift engine is imprinting emotion onto silence-adjacent signal space.
• Rendered output when played at 1.5x exhibits what sounds like a breath hitch, despite no intake being recorded.
• Began triggering transient skin response in listener 3 during passive monitoring.
• This is the third test in which emotional misalignment occurred between input affect and output affect.
• Plugin is not only interpreting sentiment—it is inferring it.
⸻
NEXT STEPS
1. Run the same session with all drift toggles disabled to confirm zero emotional residue.
2. Introduce false-emotion prompt (“I am happy” spoken flat) to observe whether the system prioritizes semantic or acoustic affect features.
3. Consider adding a discrete bypass meter for spectral residue monitoring.
4. Begin preparing controlled test phrases tagged with psycholinguistic baselines (e.g., ANEW dataset).