They’re perfectly fine for measuring indoor temps as long as you isolate them from the heat from the ESP board and allow a lot of airflow. The humidity is accurate enough too (+/- 5%).
I suspect the other types I’ve purchased are so poor because they were factory seconds that don’t meet spec. sold on Aliexpress.
I went to Xiaomi’s meters lately (LYWSD03MMC) They are surprisingly accurate: i’ve had 4 of them at same location for a few days and temperature was the same to the decimal, while humidity was inside a few percents. They don’t need wired power supply (only CR2032) and they are nice looking (and cheap, too). I set up three ESP32 as BT receivers around my house and currently i have 8 of Xiaomi’s with custom pvvx FW.
But, back to TMP117: i found out that too frequent update interval can cause showing too high temperature. I’ve had update set to 10 seconds, when i lower it to 60 seconds i’ve got a few decimals lower temp reading. So, i guess that self-heating is a problem when reading sensor too often.
Comparing with Xiaomi’s TMP117 shows the same temperature, so it’s additional proof that Xiaomi’s are accurate.
They were one of the sensor types I tried. The update rate I got from them was way too low though. Like one update an hour (using an ESP32). Also the readings were way off compared to my Fluke meter.
Hm…i kinda doubt that any multimeter’s temperature sensor can be held as a reference…
But, update rate is not so bad - i get a new value every few minutes.
EDIT: i agree, update rate is pretty low with original FW… but if you change it, it’s quite a different song…
OK so whatever I choose, I should not buy it from aliexpress?
At one stage I had a K-type thermocouple, alcohol thermometer and TMP117 all set up. They were all within a degree of one another at temperatures of 10 to 25°C. I broke the thermometer not long after and just kept using the meter as it was easier to set up than the TMP117. Just lucky they all pretty much agreed… like Segal’s law says, “A man with one watch always knows what time it is. A man with two watches is never sure”.
I did not realise that was possible. Interesting. Got a link?
It’s the only common denominator that could explain the terrible luck I had with BMEs and SHTs while others swear by them. Paying a few extra bucks by purchasing from somewhere reputable like Spark Fun or Adafruit would rule it out.
Sure:
Original “custom” FW
and
a branch from above, improved
With new FW you “loose” original bind key (not needeed anymore), so connecting to ESPHome is even easier.
Changing FW actually can’t be easier: on page above you have flasher, connect xiaomi, click “upload”… done. Then you can do some changes in settings if you wish.
There were some problems with different hardware, it seems that chinese change HW pretty frequently in these devices, but so far guys on that page managed to solve all
Regarding “chinese” sensors: i have quite some of them (BME280, BMP280, SHT31), and after calibration (done with “original” TMP117) they seem to show ok. However, i didn’t measure accuray across a certain band, like form -20 to +30 degrees. The most accurate turned out SHT31 sensors. BME280 were more “off”.
Would anyone venture a guess why this happens - a 100% reading sometimes for days in a row after it rains? The BME280 is outside, but not exposed to rain in a plastic box that’s thoroughly ventilated and also hosts a nodemcu running esphome…
When it’s raining humidity is (always) 100%, i think it’s quite normal…?
In my experience, humidity sensors don’t last long. I’m in a really humid winter and equally extreme dry summer climate (PNW US) and I get 3 years max out of them.
I’m currently using an Si-7021 (connected to an XBee3 ZigBee transceiver) with a 5V micro fan directed at it. In my testing the last several years of using that sensor, the fan has added at least a couple years to its life. I previously replaced them every year or so. My theory is the fan reduces the chance of dew or frost forming on the sensor, which eventually damages it.
so out of all Chinese Xiamoi is the most accurate?
well, it’s hard to “just claim” that… i guess it depends on what you get - it can be fake, but it can be original, but from a batch which didn’t pass quality check - note that devices, which didn’t pass quality check CAN be quite ok, since quality check is many times not made on every part manufactured, but only every xy made. And when one part on that check fails then all devices back to last good checkpoint are discarded. In that series theoretically 90% - or only 10% - of them can be good - depending when failure occured. measuring all of them again is just not worthed - too expensive. i guess that such devices are then collected by “ebay” and “aliexpress” sellers dirty cheap (or even free…) and sold to us.
For accuracy it’s best to place all sensors close together with one know accurate “reference” sensor (i use tmp117 for that) and leave them overnight. Then check graph, correct “liars” with offsets and you’re set.
I’ve had 4 of my xiaomi’s together for …a week or so, and they were equal to one decimal. So, it seems that these xiaomi’s are pretty darm accurate, yes.
DHT22, SHT30 and all the Velux Sensors were next to each other, the SCD40 was above and Heatmiser on the opposite wall. I found that location and enclosure made a big difference.
Well, bear in mind that most of these sensors are stated for ±0.5 or even ±1 degree error - and that goes for official ones, not ones bought on ebay or aliexpress… with these you never know real accuray until you calibrate them and even then it’s not guaranteed that they will be accurate over whole range or that accuracy won’t fluctuate with time. I’ve had one bme280 which “lied” over 2 degrees.
That’s why a very accurate sensor is a good reference to have. As said, TMP117 has accuracy 0.1 degrees, which is excellent for calibrating purposes.
But, as said, since i experienced that 4 of my xiaomi’s (mijia, ref. LYWSD03MMC ) were exact to one decimal point i could say that they are pretty good reference point.
As humidity goes, i found out that here sensors are not that accurate as with temperature…it can vary a “few” percent…
Most of them are spec’d at +/- 5% or worse. They are easy to calibrate with a jar of salt solution though.
Who will be dipping these sensor in salt of jar (read that correctly) 
I have similar situation with DHT 21 in my outdoor and greenhouse setup. Almost all this winter time humidity on both sensors is 100%. So, I made another device for controlling other greenhouse and use SHT41 sensor and turn his heater on. This sensor work properly, humidity is never 100% although temperature is never below 15°C. But same time those DHT21 sensors reads 100% outside and in first greenhouse.
@flega. Wait for a really sunny day. Open all greenhouse doors and vents fully. Measure humidity and calibrate. If your DHT is still 100% on a sunny day then it is damaged by over / permanent saturation. I have an external SHT31 on a roof weather station which spends most of time at 100% at damp winter. My greenhouse is not well ventilated and is super humid. I would expect 100%.
@JulianDH . Few day ago was a few very warm days and sensors goes beyond 100%.
But did you read my reply fully? I have SHT41 sensor in other greenhouse and that sensor have heater inside and it is build for outdoor/very humidity use and it is so far under 100%
