Fitting under a 0.95 state norm limit reduced mean 7 Hz filter phase error by 14.26 degrees against a scaled guard in nine SEEG records with changing simulated gaps of 20 to 200 ms

Fitting under a state norm limit retained more phase accuracy than scaling state weights after fitting in the main simulated-gap test. We tested nine reused public SEEG rest records from OpenNeuro ds006065 and ds005670, version 1.0.0. Each record supplied four fixed pairs. The target was an intact causal 7 Hz filter. A convex fit changed 67 of 144 gap maps and kept every complete gap map norm at most 0.95. With changing simulated gaps of 20 to 200 ms, mean phase error fell from 65.24 to 50.97 degrees against the earlier scaled guard. All nine record means improved. The paired descriptive interval for the change was -17.70 to -11.66 degrees. State error also fell. Projection without refitting gave 65.19 degrees. The burst schedule had no mean phase gain against the scaled guard. The phase gain over held voltage occurred in only five of nine record means. The norm bound contracts initial state differences for arbitrary supported gap order under shared voltage histories; it does not bound error against the intact filter. This is an exploratory offline result on reused records. No neural phase truth, measured stimulation artifact, or stimulation effect was tested. The package preserves selected voltage, fitted models, code, a proof, figures, source receipts, independent checks, and reproduction instructions. Fresh public prefixes reproduced all 7.2 million selected voltage values exactly. A separate numerical rerun reproduced all 117 result files exactly.

SEEG · state-space filter · phase estimation · gap reader · spectral norm · ridge regression · stability