I’ve noticed I get some low frequency interference when I’m viewing the data from the shake live in swarm or the browser. More annoying was these periodic low frequency spikes about every 122s. After some troubleshooting it turned out to be the helicorder render cadence on the PI itself causing the spikes which were visible when streaming with RSUDP.
I effectively disabled them by changing the “UpdateInt” in /opt/settings/dataC/tpl/heli_ewII.d.tpl from 2 to 10080 (minutes). Modifying the file in /tpl/ was persistent across reboots.
This fixed the issue. Claude came up with this explanation:
"Mean ≈ 122.4 s, nothing outside 121–125. So it was a metronome, not sporadic network loss — which is what pinned it on the Pi rather than the link. UpdateInt 2 nominally means 2 minutes; the extra ~2 s is the render itself overrunning. Raising it to 10080 was the fix.
The spike’s own duration — 250 ms per lost packet, 0–4 Hz
Each render starved the UDP forward long enough to drop packets, and rsudp’s merge(fill_value=‘latest’) holds the last sample across the gap. At 100 sps a Shake packet is 25 samples = 250 ms, so each artifact is a flat plateau exactly 25 samples wide (or a multiple, for consecutive drops). A 250 ms zero-order hold puts the error spectrum’s main lobe at 0 to 1/0.25 = 4 Hz — hence low-frequency spikes in the filtered trace and the bright 0–5 Hz spectrogram blob, with no coda."
Not sure why this is happening, but it’s interesting behavior.