# Results and Measurements

Measurements turn waveform data into design metrics. Specs turn those metrics
into pass/fail decisions.

```text
SimResult -> metric function -> Spec -> SpecTable -> corner/report view
```

## Measurement Functions

Monata groups measurement helpers by domain:

| Module | Use |
| --- | --- |
| `monata.measure.time_domain` | delay, rise/fall time, overshoot, settling |
| `monata.measure.freq_domain` | gain, bandwidth, phase margin, unity-gain frequency |
| `monata.measure.statistics` | histogram, yield, sensitivity, worst-case analysis |
| `monata.measure.spec` | pass/fail specification evaluation |

A metric function should accept a `SimResult` and return a scalar.

```python
def propagation_delay(result):
    time = result.sweep_var
    vin = result.waveforms["in"]
    vout = result.waveforms["out"]
    return delay(time, vin, vout, threshold=0.6)
```

## Specs

Use `Spec` for one measurement and `SpecTable` for a group of checks.

```python
from monata.measure.spec import SpecTable

specs = SpecTable()
specs.add("tpd", propagation_delay, max=80e-12, unit="s")
specs.add("voh", lambda r: max(r.waveforms["out"]), min=1.1, unit="V")
```

Evaluate specs after simulation:

```python
table = specs.evaluate_all(corner_results)
worst_corner, worst_value = specs.worst_corner("tpd", corner_results)
```

## Design Guidance

Keep metric functions small and explicit. They are the bridge between
simulation and optimization, so they should not hide simulator setup or mutate
the circuit.

Good metric functions:

- read only from `SimResult`
- return one number
- handle missing or failed data deliberately
- use clear units in the spec table

Avoid mixing backend-specific parsing into metric functions. If a backend
needs special handling, normalize it when creating the `SimResult`.