SHT11 & TMP36 reading bust

hi, i've connected a sht11 and a tmp36 next to each other on a breadboard and interfaced both to an arduino. but there is a bust between readings, the sht11 reading about 3 deg higher than the tmp36. the tmp36 value is spot on to that of a small merc thermometer on my desk. nothing is soldered so i think the sht11 should still be 'conditioned' ok. can anyone comment on any other sht11 or tmp36 (or other) comparisons that they might have made? i expected the temp readings to be much closer. the sht11 value is more stable, but i believe the tmp36 value is more correct. i am using jon oxer's sht1x library, but have checked against another library (sensirion) and the sht11 value is the same. one of these components is giving the wrong temp! any ideas appreciated. cheers

How do you know they aren't BOTH giving you the "wrong" >temperature? Just because one of them seems to agree with your >"small merc thermometer on your desk". When was the last time >you had your "small merc thermometer" calibrated?

That's just stupid. Of course the thermometer was not 'calibrated'. But taking it to the control room of interest it agrees after a short period to within a degree of three larger high quality liquid thermometers therein.

You may be expecting too much of mass-produced sensors. Few >mass-produced things are absolutely accurate. What is your >application? Does a KNOWN 3 degree offset really make a significant >different in what you are trying to do?

I was expecting better accuracy than 3deg. yes. and i need to control average room temperature using intake and exhaust fans to within 1 degree. so i do need better readings than +/- 3 deg.

We are assuming that you have researched this question in the >forums (including another discussion of SHT11 concurrent with this >one).

Who is we? I find no posts that deal with SHT11 accuracy, just how to get them working which is not my issue. I was hoping for some constructive comments, not some pretentious drivel.

SHT11 has an internal heater. That might account for the difference. See the datasheet for information about the heater.

(I'll say no more... thus giving you less to yell at ME about. )

I have posted this in another topic on this forum but this is identical to my problem. I am also using Sensirion SHT11 and I am getting around 3 degrees C too high temperature. My air conditioner shows 22 degrees C, the same goes for an old school thermomether and temperature probe on my multimeter. Only my Sensrion says 25 degrees C.

I know that Sensirion has an integrated heater but that should be off by default - also I have noticed somewhere that with the heater on it returns around 5-10 degrees higher temperature. Since I only get around 3 degrees higher temperature I don't think that that is the problem.

Is it an "accurate" about 3 deg higher? i.e. does it follow the tmp36, maintaining the same gap?

If so, the easiest fix would be to compensate for it in software. If not... sorry, can't help.

I'm using s/w offsets to tune 3x TMP36's so they all read within 0.1 degree of each other; it works well enough, so long as you average a few readings from each.

The easiest way to check temp sensors is put the kettle on for a nice cup of tea. At the same time fill a glass with cold water and icecubes. wait untill the water boils.

Hang in the temp sensor and it must read 100C=212F by definition. After that make tea from the boiling water. (hot coco is also an option)

In the time spent doing this first test and making the tea the icecubes and water temperature will be stabilized to 0C=32F again by definition. So hang in the sensor to check.

The only further test you can do is stick the sensor in your mouth as this is normally 37C=100F - unless you just sipped from the tea or the icewater.

All these test assume sea level air pressure etc.

So now its time for a nice cuppa tea..

So now its time for a nice cuppa tea..

I think I'll have a cold Rum and Pepsi, instead.

Well I don't drink tea but I did a few tests. I put a sensor in a plastic bag and into boiling water. It reached 95C before it reached 615C and stopped working. Proabably because plastic bag almost melted and glued to it. Multimeter probe gave me 99C in that water. And then I went outside and filled a cup with snow. Stuck both sensors in and waited. Multimeter probe gave me 0C in a few seconds and stayed there for the whole duration of the test. While the sensor slowly (it reached around 4C in a minute) reached 2.5C (this took 10 minutes or so). I could simply substract 2.5 from every reading but I can't be sure if this is accurate enough. I never tryed putting it directly into the snow - it was allways in a thin plastic bag.

But I did put in into my mouth without any bag around. It reached around 35C before giving me 615C and then 0. Proabably it doesn't like water that much :).

EDIT: There is a slight problem now. 15 minutes later sensor still shows around 22C in my room - as does my multimeter. And this is realy getting on my nerves. I can't have a sensor that is accurate once a day.

I can't have a sensor that is accurate once a day.

Smack it with a big hammer. Then, you won't need to worry about it being accurate again. :)

I haven’t tried boiling or freezing mine (yet?) but it tracks very closely to an Oregon Scientific wireless sensor sitting next to it.

How frequently does your sketch perform temp / humi measurements? Sensirion recommends the sensor be active no more than 10% of the time to avoid self heating. That means no more than one temperature measurement (320 msec) and one humidity measurement (80 msec) every 4 seconds in the default high resolution mode. I’ve noticed that the SHT1x library measures both temperature and humidity in its readHumidity() function (presumably to do temperature compensation) so you’d need to increase the cycle time to ~7.5 seconds if using that library.

I have also tried turning on the internal heater to see how it works and quickly saw the temperature readings shoot up 10 deg C before I pulled the plug.