Gen7 Board Temperature Sensor Validation

This report records the first thermal validation run on Gen7 hardware (earlier reports in this section are Gen6/C2). The Gen7 control assembly has three analog board-temperature inputs, TMP_SENSE_1..3, each an Amphenol ring-lug NTC thermistor bolted to the shared aluminum base plate. The objective is to validate the complete measurement chain — divider, RC filter, ADC, and firmware conversion — by showing that the three channels agree with each other at uniform plate temperature, agree with a thermal-imager reference, and correctly track an applied thermal gradient.

This run also closes out the firmware repair that made the channels readable at all: TMP_SENSE_1 was remapped to its actual pin (PF8 / ADC3_INP7, was PA5), and the ADC1 scan channels were connected to the sampling mux by setting the H7 PCSEL pre-channel-selection bits (OpenVVVF/RTE commit abe246a, branch coproc-flash-fixes). Before those fixes every ADC1-scanned channel converted a floating internal mux node and returned wandering garbage.

Test setup

• Hardware: Gen7 control board + gate-driver / DC-link stack on the shared aluminum base plate, bench powered, no motor, gate driver idle
• Sensors: 3× Amphenol Advanced Sensors RTS103C1R2M6L201 — 10 kΩ ±2 % NTC @ 25 °C, Beta 3977 K, M6 ring lug, 200 mm leads, bolted to the plate at three locations
• Divider (per channel): 10 kΩ pull-up to 3.3 V, NTC to GND, 3 V zener clamp, 1 kΩ + 10 µF RC filter at the MCU pin (fc ≈ 16 Hz)
• Firmware config: Hw.Temp.Bx: type NTC-beta, R25 = 10000, Beta = 3977, RSer = 10000, Orient = 0
• Instrumentation: Klein Tools TI250 thermal imager; 100 Hz telemetry session over USB CDC (temp_inv1_c, temp_inv2_c, temp_inv3_c)

Test conditions

Results

Baseline agreement (uniform plate, t < 60 s)

Baseline spread across the three channels is 2.3 °C (TMP1 sits ~2 °C below TMP2/3; expected contributors are the plate's idle gradient from the DC/DC converter and gate-driver side, plus the ±2 % resistance tolerance of the thermistors, ≈ ±0.5 °C each). The TI250 baseline cross-hair read 24.7 °C on the plate surface (scale 22.1–26.5 °C) — within ~3 °C of the NTC readings, with the imager expected to read low on bare low-emissivity aluminum.

Gradient response and re-convergence

With the torch on the TMP2/TMP3 edge, the channels separated exactly as their physical placement predicts: TMP2 climbed to a 68.9 °C peak, TMP3 followed at reduced amplitude (43.7 °C), and TMP1 on the far side saw only a slow ~6 °C rise. Peak channel-to-channel spread reached 37.5 °C at torch-off. Once the heat was removed all three channels re-converged monotonically (spread 9 °C and still falling at the end of the log) — no channel stuck, jumped, or dropped out at any point.

Open telemetry log: View the run in the Telemetry Viewer

Cross-check with the thermal imager

At the heated edge the TI250 showed a cross-hair reading of 50.0 °C with a frame maximum of 62.6 °C — bracketing the telemetry from the same region at the same time (TMP2 ≈ 60–69 °C, TMP3 ≈ 40–44 °C). Given cross-hair placement uncertainty and the low emissivity of bare aluminum, sensor and imager agree within a few °C.

Conclusion

All three Gen7 board temperature channels measure within ~2.3 °C of each other at uniform plate temperature and within a few °C of the TI250 reference, both at baseline and under an applied thermal gradient; the gradient response matches the physical sensor placement and re-converges cleanly after heat removal. The channels are validated as within a few degrees of reality and are fit for use in the over-temperature protection path (default limit Hw.Temp.Bx.CritC = 90 °C).

Artifacts

• Decimated telemetry log (JSONL, 1 s) - open in Telemetry Viewer
• Telemetry overview plot (PNG)
• Photos: test setup, live telemetry, torch under plate, torch + imager, baseline thermal, heated thermal

Notes

• The full-rate session log (~30 MB, 100 Hz) is decimated to 1 sample/s for the repository.
• Motor temperature and throttle channels were disconnected during this test.
• TMP1 is sampled on ADC3 (12-bit) and TMP2/3 on ADC1 (16-bit); no systematic difference beyond the baseline offset was observed.
• The DC-link current reference (~1.64 V = half-scaled 2.5 V) was also confirmed sane during this run; the earlier spurious CurrentSensorRef fault was a symptom of the same ADC1 mux (PCSEL) bug, not a sensor problem.