Read locally; never uploaded.
0.0 s
Arrow keys page through the recording. Click a trace to highlight it everywhere.
Spectrum
Head map
Relative band power per channel
share of 1-45 Hz power · alpha peak searched 7-14 HzThings to try
- Find alpha. Open s01, click O1. Read the alpha peak from the table. Switch to s02 and compare the alpha column at O1 and O2.
- Kill the hum. Open s03 with every filter off and the gain at 1000 µV. Note the 60 Hz column. Add the 60 Hz notch, then a 1 Hz high-pass, then a 40 Hz low-pass, checking the trace and the spectrum each time.
- See the phase cost. On s06, find a blink at Fp1. Switch the filter type from zero-phase to causal: the blink shifts later and changes shape. Real-time neurofeedback has to live with the causal version.
- Watch a reference move a blink. On s06, switch the reference from "as recorded" to average and to Cz. Which channels gain a blink they did not have before?
- Mean vs median. On s09 with the whole recording, switch the spectrum average from median to mean. The mean is dragged up by the drift and movement bursts.
- Judge a cleaning pipeline. Step through s06, s07 and s08 with the same filters. What did rejection remove, and what did the full pipeline repair?
The browser filters are 4th-order Butterworth sections (notch Q = 30), not MNE's default FIR filters, so numbers will differ slightly from the Python labs. The ideas are the same.