Shimmer3R RTC Drift Test DEV-866 · 32k LSE crystal error vs host clock

Samples the device real-world clock against this computer's clock and fits the drift slope (ppm) — the same method as the Verisense console drift monitor (DEV-844). Prefer the dock/USB connection: wired round-trips have less jitter than BLE. A usable ppm estimate typically appears within 1–2 hours; leave running longer for tighter confidence. Keep this tab open, disable system sleep, and note that the fit automatically rebaselines if the host clock is stepped by NTP.

disconnected

Sample every s wake lock off

The RWC is driven by the 32k LSE, so its error shows here directly — the LSE-only ground truth that S3R_TEST_0028's differential (LSE − HSE) cannot give by itself. First hardware runs (2026-08-10, docked): stock 12 pF Shimmer3R ≈ −9 ppm (near-spec — unlike the Verisense, whose identical BOM ran +40…+65 fast), while a 22 pF-reworked unit ≈ −98 ppm (over-loaded). Docked units read a few ppm low from charge self-heating; a battery/BLE run at ambient gives the spec-comparable figure.

Samples

Elapsed

–

Drift fit

–

≙ per day

–

Offset (dev − host)

–

Last RTT

–

Device steps

Host (NTP) steps

Offset drift since first sample (blue dots) with least-squares fit (orange).

Log