Temperature sensors inaccuracy

ESPHome
1

I was wondering if I am the only one that needs to calibrate temperature sensors.

Currently, I have 3 devices with 3 mcp9808 (apparently highly accurate) inside identical enclosures. (ikea vindrinking, it has vent holes). I do understand that you might get a slight higher temp reading if the sensor is inside an enclosure but what I don’t understand that it’s so far off and i have to apply different offsets.

the offsets rage from:

  • temperature_offset: “-4.3”
  • temperature_offset: “-6”
  • temperature_offset: “-2.8”
- platform: mcp9808
    name: "${device_name} Temperature"
    id: t1
    filters:
    - offset: ${temperature_offset}

I have calibrated and synced them to a “truth” in the summer and now they even start to drift, approx. 3 degrees

Are you guys also applying an offset to your temperature devices? If not, what sensor are you using that is accurate out of the box?

Thanks a lot

2

The original ones from microchip indeed claim +/-0.5°C acccuracy. :thermometer:

Question is did you got originals? Chances are always high that you end up with counterfeits buying @ aliexpress or ebay :shopping:

No

I make a lots of use of this hygro/thermometer from xiaomi. Quite a bargain in the past (like a pack of 4 for ~$12 including batteries :battery:)

3

I did buy originals, overpriced, from a local vendor. I am pretty sure it’s the real deal. I actually think it measures changes correctly, just the initial value is way too high.
I have a similar xiaomi device that I used as a truth :slight_smile:

I wonder if these breakout boards like the mcp9808 or bme280 are even factory calibrated.

Thanks @orange-assistant for your inputs, much appreciated.

4

I have BME280 + BME680+SCD40+SHT22+SHT11+Velux Sensor + Heatmiser … they are all different! I tend to find that room position makes an even bigger difference…

5

An easy way to test calibration is to put the sensor in a ziplock bag, squeeze as much air as possible out, and dunk it in ice water or boiling water. If you get 0C and 100C respectively, you know the sensor is accurate.

As long as the sensor can survive 100C, you can do both tests. I use the ice water test with every temperature sensor I get, since it’s the only convenient temperature reference that I can whip out consistently.

1 Like
6

I’ve heard the BME280 heats up as it senses.

7

Is the temperature difference significant for your uses?
What do you need them to read for?
You can try this PID controller of equipment you have that is already accurate here: GitHub - ScratMan/HASmartThermostat: Smart Thermostat with PID controller for HomeAssistant
There are thermal switches and wires, its incredible! They are the best kind of smart the ‘dumb’ kind of electronics that work automatically: they have usually on/off and different temperatures. https://www.amazon.com/dp/B019VSPFGA

1 Like
8

@CrackedHouse not significant. I mean, If I feel it’s hot or cold, I want to check my sensor to be sure I don’t have a fever or hypothermia :wink: As for instance one of my sensor was 6 degrees off, I rather relay on my gut feels then on these sensors. I am ok with one or two degrees difference, I don’t need them to be extremely accurate.

@JulianDH : i guess in the range of +/- 2 degrees C?

@Nick2253 : are must me kidding me :slight_smile: Are you really doing this?

I was more wondering, if you guys buy these breakout boards (mcp9808, bme280) and connect them you your ESPs, do you have to calibrate them? Shouldn’t they be factory calibrated? Sending accurate readings out of the box, especially the i2c ones. As already mentioned, I don’t need them to be super accurate.

9

Yup, haha. It’s really easy to do, and then I’m confident that the sensor are correct. It’s a trick I learned in college for field world. Most sensors I use can’t survive 100C, so I only do the ice water calibration.

After I posted, I was thinking about it. I’m usually using Zigbee sensors, and they don’t have wires to get in the way. With wired sensors, you’d have to be more careful. Probably leave the bag open at the top.

I’ve found that the cheap Zigbee sensors can be off by up to 5C, so this calibration is really useful.

10

@Nick2253 @op00

Rather difficult to do calibration for my little circuit boards… and so a bit of a reality check … the most sophisticated (expensive) sensor the SCD40 has a standard offset of 4C… you adjust it depending on the enclosure … the heater for the C02 sensor is significant making the temperature sensor a secondary by product.

Photo below of all my sensors on the same bench (but SCD40 is in the ceiling directly above the counter) and the Heatmiser which is on the wall! Humidity was the biggest variation

IMG_1662.PNG
IMG_1662.PNG1125×2436 239 KB

IMG_1664.PNG
IMG_1664.PNG1125×2436 262 KB

1 Like
11

IDK of any of these sensors that come calibraed. They’re made by children with little to no QC standards. you could buy a 5 pack of sensors and each one will need its own linear calibration and/or offsets. you didn’t use the same calibration settings for all them without doing each one did you? You also need to look at the data sheets, the information is right there, you just have to make the effort to go read it. It will tell you the +/- % accuaracy and you can’t very easily calibrate your way away from that realilty.

1 Like