Working notes from helping a colleague (Ruslan) get a DIY pocket thermal imager up and running. The original board kept producing a noisy frame with a fixed-pattern offset across rows.
Repo: github.com/georgeschaible/thermal-imager
The symptom#
Captured frames showed a clear horizontal banding pattern that didn’t go away with calibration. Looked like a clocking issue.
Diagnosis steps#
- Scoped the SPI clock — clean
- Checked sensor power rail under load — sagging ~80 mV during readout
- Added a 22 µF tantalum on the sensor side of the rail — banding gone
Reproduction#
This is a cat-able single file for now; the full schematic and build log will move to its own page once Ruslan finalizes v0.2.