This page summarizes how compression ratio (CR) trades off against signal- and physiology-level fidelity for the v1 RVQ goldens. The Effective CR column is populated when an entropy-prior run is available; otherwise, the codec CR is the release-facing operating point. CRs are reported alongside the noise-aware metrics so it is easy to inspect the difference between waveform faithfulness, clean-reference fidelity, and physiological metric preservation.
Data source: measurements preserved from the original documentation.
| CR | Codec CR | Effective CR | bits/tok | N | Faithful PRD% |
|---|
| 02x | 2.00 | 5.76 | 2.78 | 1000 | 2.31 |
|---|
| 04x | 4.00 | 9.73 | 3.29 | 1000 | 3.63 |
|---|
| 08x | 8.00 | 15.24 | 4.20 | 1000 | 6.03 |
|---|
| 16x | 16.00 | 26.88 | 4.76 | 1000 | 9.72 |
|---|
| 32x | 32.00 | 59.87 | 4.28 | 1000 | 14.59 |
|---|
| 64x | 64.00 | 118.06 | 4.34 | 1000 | 21.05 |
|---|
| CR | Truth PRD% (clean) | PRDN-noise% | HR MAE (bpm) | QRS-band PSD err | Coherence |
|---|
| 02x | 1.38 | 0.00 | 0.48 | 0.0081 | 1.0000 |
|---|
| 04x | 2.33 | 0.00 | 0.95 | 0.0104 | 1.0000 |
|---|
| 08x | 3.84 | 0.00 | 1.39 | 0.0144 | 1.0000 |
|---|
| 16x | 7.28 | 0.04 | 2.05 | 0.0246 | 1.0000 |
|---|
| 32x | 13.27 | 0.78 | 2.34 | 0.0497 | 1.0000 |
|---|
| 64x | 20.49 | 3.53 | 3.00 | 0.0835 | 1.0000 |
|---|
| CR | Seam ratio |
|---|
| 02x | — |
|---|
| 04x | — |
|---|
| 08x | — |
|---|
| 16x | — |
|---|
| 32x | — |
|---|
| 64x | — |
|---|
Tertiles are formed from a band-power noise estimate over each input recording; clean is the lowest-noise third, noisy is the highest.
| CR | Tertile | N | PRD% | PRDN-noise% | HR MAE |
|---|
| 02x | clean | 333 | 2.09 | 0.00 | 0.40 |
|---|
| 02x | median | 334 | 2.26 | 0.00 | 0.45 |
|---|
| 02x | noisy | 333 | 2.57 | 0.00 | 0.59 |
|---|
| 04x | clean | 333 | 3.21 | 0.00 | 0.77 |
|---|
| 04x | median | 334 | 3.62 | 0.00 | 0.91 |
|---|
| 04x | noisy | 333 | 4.05 | 0.00 | 1.18 |
|---|
| 08x | clean | 333 | 5.02 | 0.00 | 1.25 |
|---|
| 08x | median | 334 | 6.11 | 0.00 | 1.27 |
|---|
| 08x | noisy | 333 | 6.97 | 0.00 | 1.67 |
|---|
| 16x | clean | 333 | 8.15 | 0.12 | 1.67 |
|---|
| 16x | median | 334 | 9.76 | 0.01 | 2.19 |
|---|
| 16x | noisy | 333 | 11.27 | 0.00 | 2.28 |
|---|
| 32x | clean | 333 | 12.78 | 1.93 | 1.93 |
|---|
| 32x | median | 334 | 14.70 | 0.31 | 2.22 |
|---|
| 32x | noisy | 333 | 16.28 | 0.11 | 2.86 |
|---|
| 64x | clean | 333 | 18.59 | 7.98 | 2.57 |
|---|
| 64x | median | 334 | 20.97 | 1.75 | 3.09 |
|---|
| 64x | noisy | 333 | 23.58 | 0.88 | 3.32 |
|---|
| CR | Tertile | QRS-band PSD err | Coherence |
|---|
| 02x | clean | — | 1.0000 |
|---|
| 02x | median | — | 1.0000 |
|---|
| 02x | noisy | — | 1.0000 |
|---|
| 04x | clean | — | 1.0000 |
|---|
| 04x | median | — | 1.0000 |
|---|
| 04x | noisy | — | 1.0000 |
|---|
| 08x | clean | — | 1.0000 |
|---|
| 08x | median | — | 1.0000 |
|---|
| 08x | noisy | — | 1.0000 |
|---|
| 16x | clean | — | 1.0000 |
|---|
| 16x | median | — | 1.0000 |
|---|
| 16x | noisy | — | 1.0000 |
|---|
| 32x | clean | — | 1.0000 |
|---|
| 32x | median | — | 1.0000 |
|---|
| 32x | noisy | — | 1.0000 |
|---|
| 64x | clean | — | 1.0000 |
|---|
| 64x | median | — | 1.0000 |
|---|
| 64x | noisy | — | 1.0000 |
|---|
- Faithful PRD compares reconstruction with the recorded input; Truth PRD uses the fixture’s clean reference. Their difference depends on the input noise and reference construction.
- PRDN-noise is a noise-normalized error measure. Interpret it with waveform and physiological errors, not as proof of denoising by itself.
- HR MAE and band error describe the measurements in this evaluation. Acceptance thresholds depend on your application.
- Seam ratio reports boundary energy relative to the window center. A missing value means no result was reported.
- Sample counts and noise-bucket composition can differ across ratios. Preserve that context when comparing candidates.
See the validation scorecard for definitions and customer evidence for a compact comparison.