Measured samples only.
A visible transfer cost.
A fixed weighted least-squares quadrature estimator removes assumed padding and gives an explicit conditional error budget. Its chosen window passes two of six synthetic cases.
A local error interface.
The estimator fits cosine, negative sine and a constant baseline in a fixed 2049-sample Hann-weighted window. Its complex coefficient preserves quadrature phase. Every valid output uses measured samples only, with no whole-record demeaning, extension or taper division.
For independently justified envelope variation L and input uncertainty ε, complex error is bounded in exact arithmetic by d = H₁L + H₀ε. The numerical kernel gives H₀≈2.000000000043 and H₁≈0.000311801908507 seconds. Formal deployment needs numerical enclosures and calibrated source/input bounds.
| Valid outputs | 14136 per 16184-sample record |
|---|---|
| Window radius / availability delay | 1024 samples / 524.288 microseconds |
| Constant phasor and baseline controls | Maximum numerical error below 2.23×10⁻¹³ V |
| Synthetic recovery gates | 2 / 6 passed |
Locality does not imply broadband recovery.
The exact carrier and a 44.65 Hz offset pass. Smooth amplitude/phase modulation, turn-on, narrow pulse and omitted-sideband full-state recovery fail. The pulse relative RMS error is about 0.549; its fast envelope is strongly smoothed.
The explicit transfer law includes both main response and conjugate-image leakage. At a 600 Hz offset the main gain is about 0.7694; at 1000 Hz it is about 0.4637. These sampled responses are not a certified continuous passband. No post-result tuning or transfer correction is applied.
Transfer and control results →Six-case metrics ↓Freeze the source class before redesign.
Next specify envelope bandwidth or regularity, noise, gain/skew bounds, allowed delay, passband distortion and image rejection. Then test whether one fixed local kernel can satisfy the joint budget using a continuous-frequency certificate.
The theorem supplies a conditional measurement interface. Source regularity, full energy, absolute native support, timing calibration and action normalization remain open.
Frozen estimator contract →Reproduction code ↓Preserved boundary comparison →