T1000-E Temperature Sensor Accuracy

[Beeeerock]

I put this to the Discord troubleshooting channel and it was suggested that I post it here. My T1000-E device shows a temperature that is typically around 10 degrees C cooler than the ambient temperature. I only have the one device so I can't confirm whether it's a faulty sensor or an actual firmware issue. The initial thought on Discord was that perhaps the sensor needed an improved initialization.

[mrlmuk]

Mine is by the window of my office and says 13.4c, the TRV on the radiator nearby says 21c. They should not be the same but that seems like a similar offset so it's probably not a faulty sensor.

The firmware might be using the wrong sensor type or incorrect calibration value.

[seagull9000]

I have this issue too but haven't yet looked closely into it.

It's possible different hardware versions exist.

Mine is also reaching around 10 degrees C low

[Beeeerock]

My board has a black battery visible in the case. I've heard the battery colour might indicate hardware version - nothing relevant is visible on it otherwise.

[seagull9000]

Mine too has a black battery and the logo the front.

[mrlmuk]

I also have the black battery version, when I ordered a second it came with silver and did not work correctly. I believe the black is the later somehow improved one but I doubt they have changed the temperature sensor. They all seem to be around 10c lower so it seems like something that could be fixed in the code.

I think this file is the one /variants/t1000-e/t1000-e_sensors.cpp but I am not a developer, making sense of it is a very slow process and probably beyond what I can do.

[Beeeerock]

Has anyone tried this device with Meshtastic to see whether the apparent temperature error exists with that firmware too?

[mrlmuk]

Things like this likely contain shared example code so you would expect the problem to be present in other projects.

https://www.reddit.com/r/meshtastic/s/utHC24IbWv

You can easily find examples but the excuse made here is not valid as the offset is in the wrong direction, the device doesn't generate much heat and I tested by putting it in the fridge which is more than capable of dealing with it.

[seagull9000]

Looking at the code, it's an analogue read to get the temperature combined with heater temperature. Table lookups looks the same as meshtastic. I'll take a deeper dive later and update this.

Only does a single read is done after 10ms rather than several reads that are averaged, so if there is a difference in boards, it could be it takes longer to get the correct temperature.

Update:
I tried changing various things in the code but no luck. It always reads around 10 degrees too low.
I've not tried meshtastic again, but I believe that used to be more accurate.

[mrlmuk]

Mine is accurate when left alone over night.
T1000-e by window, 19.7c
NSpanel on desk, 21.1c
TRV on radiator, 21.5c (display rounds to .5c)

When holding it the temps quickly climbed to 24c .

After 15 mins of checking it reached 26c.
Internal battery resistance rises toward empty, perhaps this is enough to generate more heat just before the battery dies.

The temp dropped back to 21.5c in the app.
Charging generates a fair amount of heat IR thermometer shows 28c on the button/SoC RF shield and 25c on the battery.

Meshtastic is definitely not more accurate, if anything it is worse.
After install it went up to 29.7c, in the app probably due to high CPU usage during setup.|
After a few minutes the app showed 27.1c, but IR showed 25.5c on SoC and battery.
While charging the app showed 31.5c, IR 27c SoC 26c battery.
Clearly the sampling and update frequency are different, almost like it is just showing an internal temp.

Overall, I don't think there is anything specifically wrong with the results in Meshcore. while sometimes they are off it must be incredibly difficult to work out the ambient temperature from inside a tiny device with multiple heat generating objects.

Back in Meshcore, the temp immediately shows 18.2c after reinstalling and importing config while also charging. The numbers in Meshtastic are simply wrong.

[seagull9000]

Overall, I don't think there is anything specifically wrong with the results in Meshcore. while sometimes they are off it must be incredibly difficult to work out the ambient temperature from inside a tiny device with multiple heat generating objects.

It sounds like you might have a device that doesn't have the issue. The experience some of us are having is that the temperature consistently reads 10 °C below ambient. In my experience this is without charging, or doing much at all really.

Can you desribe the appearance of your device? It may provide some confirmation of which devices have the issue.

[mrlmuk]

I definitely have the problem of lower temperatures some of the time.

Now the battery is nearly charged.

Disconnected from charger
batt 4.00v
temp 15.1c

Connected from charger
batt 4.09v
temp 14.5c

Room is 21/22c.

Maybe this is something to do with the battery voltage?

I don't think there is anything special about my device.

SENSECAP logo on black battery.
SenseCAP Card Tracker T10000-E for Meshtastic
MAC: F7:64:2F:88:AF:xx
Model: T1000-E
Input: 5v 0.36A
FCC ID Z4T-T1000-E
S/N: 1149336925030xxxx

I did have a silver battery one which did not work properly. Meshtastic would install, Meshcore after 1.4 would not run.

[mrlmuk]

Possibly relevant comment if you can trace the same user on github.
https://www.reddit.com/r/meshtastic/comments/1gqo114/t1000e_temperature_data/

[black_kitsune]
(https://www.reddit.com/user/black_kitsune/)
[9mo ago]
(https://www.reddit.com/r/meshtastic/comments/1gqo114/comment/ly15dpz/)
A few of us have higher temperature readings than ambient. I am fixing that using a Sienhart-Hart equation instead of their default resistance = temperature mapping.
However, it is proving to be more challenging because it's wrong when I grab three points of data on the map to calculate the coefficients and plug them into the Sienhart-Hart formula. I believe their temperature-to-resistance map is wrong.
Hopefully, I can push an update to the Meshtashtic firmware soon.

[IoTThinks]

So the issue is at the code or the hardware of T1000-e?
My T1000-e has black battery and shows 10 degree C less than the environment.
E.g: Show 18*C instead of 28-29 *C

[danieldurrans]

Newly delivered (in last two weeks) T1000-E with the same issue... MeshCore iOS app is displaying the temp as circa 10'c lower than I would expect. Mine also has a black battery.

[jpds]

I've been running my T1000-E with the latest dev branch code and I think it's reporting more realistic temperature telemetry with the newer library versions on that.