[hvalentim]

Although it is annoying it is not unexpected/abnormal.

According to specifications those sensors should have a Relative Humidity Accuracy between +-2 (typical) and +-3 (maximal) at 25ºC.

Theoretically they could be anywhere between 4%(*) and 6% apart and still performing according to spec.s.

(*) Say the "real" humidity is 50%. Sensor A overreads 2% and shows 52%, sensor B underreads 2% and shows 48%.

To improve accuracy, perhaps one could use both and average the reading

[Mikko]

Having ±5% would be fine for this price class, but 10 - 30% is not. I believe that those cheap sensors are not meant to last outside for several months. I do have now a new Vantage Pro with SHT31 humidity sensor and that match quite closely to SHT11 readings, both are outside about 5 m apart. Will replace other sensors with SHT31 before winter time.
BME280 sensors will also go outside and it's interesting to see how they survive, since designed for automotive applications.

[spudstrawb]

...Will replace other sensors with SHT31 before winter time.

Mikko
Which SHT31 sensors are you looking at using? There are a number of different versions available on ebay, starting from around AU$10 to around AU$25. I suspect that the ones at the cheaper end of the scale might have the same accuracy issues that the current HTU21D sensors have?

[hvalentim]

Having ±5% would be fine for this price class, but 10 - 30% is not.

True. But I did pay 10 times the price of one of these for a Sensirion SHT15 which ended up with that sort of offset after just eight months outside.

Of course, being of Chinese provenience one has always the suspicion it could be a second grade product.

But if you do a search you will find plenty of complaints of Davis users ending up with humidity offsets in their Sensirions as well...

I fear that for a reliable lasting measurement something like the Vaisala you suggest may be the only option.

Short of that, perhaps paying a bit more and getting the Sensirions from a reliable electronics store would be wise.

[Mikko]

I bought these SHT31
There were not so many other versions earlier, but now I can find quite many SHT31 variations. Hopefully the difference is only pcb, not the sensor itself. SHT11 has been outside one year now and it's giving the same readings as new SHT31, therefore I have some faith on Sensirion (for now)...

[Mikko]

Here is the latest picture with fresh SHT31, Sensirion sensors match quite well.

[JT118]

I have found, doing many tests with many versions of the SHT 11,15 etc and 21, that if you connect them up, and put them over a cup of hot water ( about 5cm above) from the kettle and leave them for 30mins. They then read identical RH. I have done this with a number of combinations both genuine Sensirion and copies (2nd's?), It seems to work every time. The cheaper ones then seem to drift more quickly.

[spudstrawb]

Well my SHT31 arrived today (the one that Mikko mentioned) but I can't for the life of me get it working. I changed the TX unit configuration from HTU21D to SHT21 and then ran it but the TX led on the TX unit flashes twice on boot and then gives nothing. Nothing comes up in debug either - it all just sits there. I've tried a number of other things - running it off the I2C pins etc (as a HTU21D). At least the TX unit will transmit in that mode but it is giving temp and humidity values of -312.72 for both. So I don't know if:
* It was faulty on arrival.
* I've damaged it during installation.
* I'm missing a setting somewhere.

I was hoping to provide some more insight into HTU21D vs SHT31 accuracy but I'm back running with the HTU21D for now.

[werk_ag]

The SHT31 is a very new sensor from Sensirion and it is not yet included in the list of supported sensors for the WeatherDuino Pro2 system.
As far as I know, it requires a specific library, which means it can't be read using the SHT2x library.
As soon as I get one sample, I will try to add support for it. No promises.

[spudstrawb]

Okay, thanks Werk_AG.