Ten numbered balls · one native SPICE phase landscape
Drag slider to compare open/active timeNatural clock drift (5 balls)Vout-only trim (5 balls)White rings: sampled minima
Sampled valley 120°, 240°Lowest sampled static RMS 1.243 mV
Late basin residence · measured
19–23.5 ms| Seed (φ₂,φ₃) | Drift RMS | Trim RMS | Trim residence |
|---|
Trim better than drift —Trim within 35° valley —
Selected matched pair
Two real native Vout RMS histories. Late values are averaged over all switching cycles, not the displayed decimation.
Ten native switching-cycle RMS trajectories
Native 2.5 µs cycle RMS · 24 msNative Vout waveform microscope
Native absolute Vout from saved SPICE samples (40 µs). Superimpose compares Drift and Feedback for the same seed and window; waveform shapes are not computed.
Native switched-SPICE VRM: 12 V, nominal 1 V/40 A, 400 kHz, 200 nH per phase, 1500 µF output, idealized switches. Each of five initial states is simulated twice with identical natural frequencies 400000 / 400010 / 399990 Hz. Each local feedback agent observes common Vout and its own oscillator, using a fixed harmonic response model; no peer phase, phase map, or sampled optimum is supplied. R34 (100 kHz, duty 1/3) inspired the matched-start design but is a different electrical fixture. The 0.5ms-phase-point animation interpolates gate-derived native coordinates for display only. The script retains successes and failures; this is segmented ngspice/Python control, not MCU hardware or a proof of universal convergence.