Compiled multi-packet adapter comparison
6/6 generated controls pass.
| Model | Processing | Result | Position gap | First-derivative gap |
|---|---|---|---|---|
| Globe + field | Zero removal | Pass | 0 m | 0 m |
| Globe + field | Profiled removal | Pass | 0 m | 0 m |
| Globe, zero field | Zero removal | Pass | 0 m | 0 m |
| Globe, zero field | Profiled removal | Pass | 0 m | 0 m |
| Flat disc | Zero removal | Pass | 0 m | 0 m |
| Flat disc | Profiled removal | Pass | 0 m | 0 m |
Each case checks the first64 generated packets and all 27 physical parameter directions. The one-micrometre tolerance compares software values and first derivatives, not sensor accuracy. No aircraft fits or original recordings are read.
The first-order native reference does not independently verify second or third derivatives; those retain separate known-answer checks. These controls do not qualify full-history prediction ranges, complete clock/rotation branches, native execution-error bounds or an observational error law.
Frozen limits · Trajectory components · Remaining prediction requirements.