Channel error-budget diagnostic
Completed 46/46 diagnostic identities in 2.11 seconds. Implementation checks passed for 46/46. No original measurements, native integrations or observed fits were used.
Of 108 candidate/trajectory checks, 99 pass the local-linearity screen and 90 pass the trajectory-truncation screen; 81 pass both. These are assumed-error development screens, not scientific detections or calibrated coverage.
All candidate results retain every model and window. Noise levels are declared control assumptions, not measured receiver performance. No window is selected from these results.
Acceleration averaging
| Rotation in one second (degrees) | Mean-packet loss (m/s²) | Summed-vector loss (m/s²) |
|---|---|---|
| 0 | 0 | 8.8817842e-15 |
| 0.1 | 3.1127101e-11 | 1.2451237e-06 |
| 1 | 3.1128096e-09 | 0.0001245119 |
| 5 | 7.782023e-08 | 0.003112513 |
| 20 | 1.2451237e-06 | 0.049729146 |
| 90 | 2.5213736e-05 | 0.97789694 |
These exact rotating-force integrals show that averaging body vectors and averaging their magnitudes are different operations. Neither becomes an exact instantaneous force without a timing/averaging model.
Noise bias in magnitudes
| Assumed isotropic noise σ (m/s²) | Magnitude bias (m/s²) | Small-noise prediction (m/s²) |
|---|---|---|
| 0.001 | 1.019368e-07 | 1.019368e-07 |
| 0.01 | 1.019368e-05 | 1.019368e-05 |
| 0.1 | 0.001019368 | 0.001019368 |
| 0.5 | 0.0254842 | 0.0254842 |
Averaging more magnitudes does not remove this nonlinear mean shift. Actual noise, calibration, quantization and correlation still need an observational model.
The main 53-stretch findings and the earlier unresolved height-removal sensitivity result are unchanged. These controls do not qualify a new recorded-data likelihood.
Frozen matrix and limits · Next gate · Measurement contributions.