At fixed coupling and equal uncoupled variance, fast and slow noise reduced squared S cone signal scores by 13.8% and 0.7% in one passive human foveal model

Electrical coupling can lower voltage noise and mix colour signals. We tested how source-noise time scale changes signal detection in one archived human foveal graph. We kept all 333 named photoreceptor units and 1,141 contact rows. The passive model had equal leaks, independent Gaussian source noise and a coupling-to-leak ratio of 0.0545456 per recorded weight unit. Each noise model had equal uncoupled voltage variance. At a readout of the 16 labelled S cones, coupling reduced the squared detection score for S-only drive by 0.73% with slow noise and 13.85% with fast noise. Both cases had lower voltage variance. For common drive, the score rose by 34.62% and 16.34%. For opposite S and LM drive, it fell by 30.03% and 38.92%. Intermediate noise rates gave intermediate scores. All four assignments of two uncertain S labels gave the same time-scale effect. Reading all 333 voltages recovered the slow-noise S-only score but did not recover the fast-noise loss. Removing output covariance gave a false 42.45% slow-noise common-drive score gain at full readout. Independent parsers, sparse solves, Lyapunov equations, simulation and a fresh full run checked the results. The package includes the full source table, graph, code, 10,080 result rows, figures, provenance and repeat instructions. These are single-sample scores in a fixed linear model. They do not measure human colour vision. Active currents, feedback, nonlinear readout and measured tissue noise are outside the tested scope.

human fovea · connectome · electrical coupling · S cones · noise time scale · signal detection · noise covariance · passive model