A power bound identified unchanged decisions under no clipping in eight fixed SEEG readers with uniform voltage codes

We tested uniform voltage codes in eight fixed readers from public SEEG recordings. A band-power bound gave a conditional check for an unchanged native decision. The check requires a flag for no clipping in the complete buffer and a known sample-error bound. At 12 bits per bipolar pair and a range of six training standard deviations, it accepted 47.71% of decisions as a mean across people. All accepted decisions matched the native reader. At eight bits, it accepted no decisions. The original 8-bit test compared contact coding before subtraction with pair coding after subtraction. The mean Brier-error difference was 0.035867. The descriptive 95% interval was -0.000649 to 0.089909. The sign-flip value was 0.148438. This comparison remains uncertain. The study reuses eight people and existing readers from OpenNeuro ds004473 version 1.0.2. The proof uses real arithmetic. The software uses numerical guards and checked all native intervals. The package contains exact selected source windows, code, source records, all scores, four figures, and replay instructions. The check concerns decision fidelity. It does not establish correct task labels, live control, or brain stimulation.

voltage quantization · SEEG · signal processing · decision fidelity · power bounds · brain-computer interfaces