# INA219 Calibration Test Plan (Somatic Ledger v1.0)

## Objective
Validate signal-to-noise ratio under thermal load for power tracing in the Copenhagen Standard framework.

## Test Parameters
- Device: INA219 shunt (I2C bus, 16-bit ADC)
- Load: GPU compute workload (CUDA kernel launch >=100 kWh)
- Sampling Rate: >=1kHz (aligned with ts_utc_ns resolution)
- Noise Floor Threshold: <0.5% of measured power

## Test Scenarios
1. Baseline: Idle GPU, 25C, 10 min, <30W expected.
2. Thermal Load: Sustained compute @80%, 30 min, 200-400W expected.
3. Transient Spike: Kernel launch burst, 1ms, <1ms pulse.

## Expected Output
CSV with columns: ts_utc_ns, voltage_12v_rms, current_amps, power_watts, temp_celsius.

Noise floor calculation: std_dev(voltages) / mean(voltages) must be <0.5%.

## Decision Criteria
- Pass: Noise floor <=0.5%, stable 120Hz acoustic delta signature.
- Fail: Signal variance exceeds threshold, requires recalibration or hardware revision.

## Next Steps
1. Execute test under controlled conditions.
2. Compare against NVML baseline (nvidia-smi polling).
3. Document discrepancies in Somatic Ledger audit log.

Aligned with Copenhagen Standard: No Receipt = Unrewardable Action.