Actions
Forum Thread Details
  • State Verified Answer
  • Replies 38 replies
  • Subscribers 338 subscribers
  • Views 5294 views
  • Users 0 members are here
Related

Temperature sensor accuracy

ntewinkel

Hi all,

I have two temperature sensors near each other, outside. In the winter I was using one to keep track of the nectar temperature in a hummingbird feeder, but I just left it hanging there for the summer.

Here is the typical result:

image

The top line is a dht11 or could be 12. It includes pressure and humidity. The bottom line is a ds18b20 in a metal tube.

Both are digital, but one is nearly always about 1 or 2 degrees higher than the other. Generally it’s fine to give me a close enough idea of what’s going on, but I think I should be more accurate when looking at the freezing point.

Do I need to calibrate them? How would I go about doing that? Ie, what can I trust as a proper baseline? I don’t have access to any really guaranteed kind of thermometer, I think. Thinking 

Thanks in advance for any wisdom you can send my way!

-Nico

  • 0 in reply to kmikemoo

    Yes, the genes are not necessarily your friend.

    Luckily on my mom's side, the men make it into their 90's, so my wife is hoping I take after that side of the family.

    Still there are extenuating circumstances, my grandfather died at 55, but he lost a leg in WWI and was diabetic but would not eat right.

    My father smoked for 60 years and then wondered why he had stomach cancer.

    My older brother was exposed to agent orange in Vietnam, plus he was in numerous motorcycle accidents. I was very surprised he lived as long as he did.

    So far I am trying to do the right things, keep my weight down, no nasty habits, and other than the heart attack, I am doing ok. I could probably do more, but I am a big believer in the quality of life, so I am not giving up my deserts or scotch any time soon.

  • 0

    I think   was onto something... it looks like the other sensor was maybe getting warmed up? Possibly because the Wemos D1 Mini is right next to it in the enclosure, even though it does have a separation.

    And I was wrong - the other sensor is a BME280, not the DHT I thought I was using in there. Funny, I guess it's been up there long enough for my short memory to keep up with it Laughing

    imageimage

    I tried an ice bath, but it didn't work well. It dipped down close to zero and then I think the bag started floating.

    imageimage

    BUT... once I removed it from the enclosure, the BME280 (light blue line, previously always on top) was recording predictably about a degree lower than the DS18b20 probe.

    I had it in the container of water overnight (with the bag clipped down better), and then removed it this morning with the two sensors being side by side in the air afterwards. This is the resulting graph:

    image

    So it looks like the BME280 is probably the accurate one (based on it hitting 0 in the ice bath), with the DS18b20 always recording a degree higher.

    I'll try the ice bath again, this time with more insulation all around, and I'm going to have to print a new enclosure that keeps the BME280 away from the Wemos D1 Mini. After that, if the ice bath result is stable, I'll update the code to just do a simple +/- adjustment.

  • 0 in reply to ntewinkel

    It looks like the gap is a tad narrower at lower temperatures, but it still seem to remain around a degree almost consistently across the board, so it's probably enough to just subtract that 1 degree difference from the DS18B20, as you said. Fancier math is not going to give you a lot more accuracy considering what we are seeing in those graphs.

  • 0 in reply to battlecoder
    battlecoder said:
    Fancier math is not going to give you a lot more accuracy

    That’s what I’m thinking too. 80-20 rule.

    Looks like I have a good zero degree bath set up now. I used an insulated cooler with ice packs and a yogurt container with ice and water in the centre.

    image

    Now to add the fine tuning adjustments…

  • 0 in reply to ntewinkel

    There's definitely a difference at lower temps but that's still a lot better than the initial readings. Whether that's a good fit or not will depend on the accuracy you need, and in this case, how often it's going to be close to freezing in their actual real locations :D

  • 0 in reply to battlecoder
    battlecoder said:
    There's definitely a difference at lower temps

    Yes, it went from a one degree difference in the 10 to 20C range to nearly 2 degrees at freezing.

    It’s only for my specific use case for the hummingbird feeder, and freezing is the key thing I’m interested in. Our area might go down to -15C and up to maybe +35C, and those are interesting but not critical.

    I might get more fancy later, but this should do the trick for now.

  • 0

    Here’s my ice bath setup. I’m using an insulated little lunch box type cooler, with a yogurt tub in it that holds the ice water. Under it and on both sides are ice pack, and I tossed in a frozen bottle of water.

    image
    closed up it holds the temperature really well.

    image
    image
    I adjusted the sensor readings on the server side (node red). It’s a bit of a cheat but it’s specific to that sensor and that sensor only reports to the one server. And, at least at the zero mark, they are both reliably reporting the same zero temperature for hours.

    • image
      Not sure what will happen as the temperature goes up, but that is close enough. Next I’ll have to print a separate enclosure for the bme280 so the warmth of the chip won’t affect it.