Accuracy of EmonTH

Emoncms
1

When I first got my EmonTH (DHT variety) it was a revelation to be able to get Temp & Humidity data.

Unfortunately, technology has moved on and, unless you really need a battery based sensor, there are better and cheaper ways of getting the data.

The crunch has come as I realise that the EmonTH is very inaccurate for temperature, reading some 10% lower than 2 other (different types) sensors. Humidity accuracy is pretty good.

So it is time to retire the old sensor. I could simply add an offset to the data, but I don’t really need to do that as the new Komfovent control panel matches the BME680 I have been using to check.

I remember Trystan doing some experiments when OEM decided to go for the Si unit in the EmonTH, so the lack of accuracy is not surprising really.

2

I just bought an emonTH along with my new SD card and have to say I’m disappointed with it. In my case the temperature readings (I bought an external sensor as well) seem to be reasonably accurate but the humidity is too high.

I’ve just put it alongside a couple of other cheap T+H meters:

             temp  humidity
emonTH    21.2+21.3  56.3
other1       21.8    50.0
other2       21.4    50.0

So I won’t be buying any more. I have an si7021 on a pi that I bought a few years ago, and that performed well when I tested it. Maybe I should retest it now.

3

You assume these cheap sensors are more accurate?

IME variation between sensors is usually quite high, but often a constant difference. Some sensors react differently (change more quickly) than others. A number years ago @TrystanLea did some fairly extensive testing of T+H sensors to choose a new one for the emonTH.

I’ve got about 15 sensors dotted around the house and if all set next to each other I’d have 15 different answers with varying degrees of difference even those of the same make.

4

Well no and yes.

Like you I’ve got a whole bunch of sensors spread around and I’ve found they all tend to give fairly similar readings. The readings in my house are pretty stable, so I don’t think there’s a rate of change issue - it’s obvious in graphs if there is. So whilst I agree that the specs aren’t any better, my experience has been that the measurements are closer than the theoretical limits. I did compare some against a calibrated unit when I built the house, and the results were pretty reasonable.