With fixed integer conversion error, zero and edge padding produced last-trial target means more than 10 microvolts apart in 28 of 40 ERP CORE N170 records under 0.1 Hz and 30 Hz filters

We tested whether the EEG resample boundary rule changes trial means after fixed whole-record filters. The study used all 40 N170 records in ERP CORE, NEMAR nm000132 version 1.1.1. The fixed biosigio 1.2.7 conversion reproduced all 194,807,250 serving integer samples, scales, and offsets with zero differences. We compared zero, edge, and reflection padding at 1024 to 250 Hz. We kept integer conversion error fixed. With the saved 0.1 Hz high-pass and 30 Hz low-pass filters, zero and edge padding gave last-trial target means more than 10 microvolts apart in 28 of 40 people. The median absolute difference was 42.11 microvolts. After baseline subtraction, it was 3.86 microvolts; six people remained above 10 microvolts. The unfiltered padding difference was zero inside all measured trial windows. The later filters spread it into those windows. A linear proof and numerical controls show how a constant voltage offset can enter the filtered signal through zero padding. A secondary face-minus-car average changed by 0.0074 microvolts across the 40 people. The package includes code, source hashes, retrieval instructions, paired results, and a compact dataset. Saved-data replay matched all 12,800 trials. The result applies to the tested conversion, reference, windows, and filters. It does not establish a true neural amplitude or a true signal outside a record.

EEG · event-related potentials · N170 · signal processing · resampling · boundary padding · zero padding · edge padding · reflection padding · digital filters