i.MX6UL: temperature monitor accuracy

[pierluigi_p]

Hi,

we are using a custom kernel based on branch imx_4.14.78_1.0.0_ga and we are reading the core temperature with both a IR thermal camera and the junction temperature from Linux kernel.

While the IR camera (under same load conditions) is showing more or less the same temperature (+/- 1*C), the junction temperature range from +4 to +15 °C compared with the IR camera.

We were actually expecting a difference of about 1.5°C (dissipation about 0.6 mW).

As side effect, the SoCs with higher temperature difference enter in thermal protection even if apparently far from the critical thermal thresholds.

Can this "variance" be anyhow originated by an incorrect formula in the TMU driver? Is there any suggestion to mitigate this behaviour?

Thanks

BR

Pier

[igorpadykov]

Hi pierluigi_p

one can refer to sect.3.1. Accuracy study AN5215

i.MX 6 Temperature Sensor Module - Application Note

https://community.nxp.com/t5/i-MX-Processors/i-MX6-Thermal-fuse-data-Dispersion/m-p/812123

Best regards
igor

[pierluigi_p · Accepted solution]

Hi  Igor,

thanks for the clarification.

Summarizing, I'm assuming that

- kernel 4.14.x already support the latest formula to estimate the junction temperature

- the calibration data is stored in the fuses during the factory calibration process

- the calibration data cannot be fine tuned

Is this correct?

Thanks

Regards

Pier

[igorpadykov]

Hi Pier

yes this is correct.

Best regards
igor