ACCURATE THERMOMETER

[Repairman77]

Another project; not built into a box yet as components only arrived today.

As most of you know we all seem to be obsessed with accuracy; at least I am, and most of you will have discovered that thermometers in the same location (whether digital, mercury or alcohol) can be anything but accurate. They can easily be +/- up to 2 degrees. Not ideal for us weather types.

I came across a transistor temperature device that has quite tight tolerances. An LM35; data sheet here in pdf format...
http://www.ti.com/lit/ds/symlink/lm35.pdf‎

Easily available on ebay for under £2 (best to buy a couple while you're at it). I got mine here; but there are plenty of suppliers.
http://www.ebay.co.uk/itm/290971687287? ... 1439.l2649

The display comes from Hong Kong but only took 8 days to come. £4.69 including postage.
http://www.ebay.co.uk/itm/150592184505? ... 1497.l2649
There may be UK sellers but I think they were more expensive.

Quite easy for anyone to wire up although you do need two batteries, one for the display and one for the sensor. It doesn't seem to work with one battery with common negatives between the two sections. The display takes 6mA so you wouldn't want to leave it on all day; you just need it on for a minute or two to stabilise. The sensor hardly takes anything at all but still takes a minute or two to stabilise. I've put 2 small switches to turn on the two sections. I can draw out the circuit if anyone needs it but it's self explanatory. Just a supply to two of the contacts on the display (either way round as it will work on AC as well as DC, and two contacts; the negative and output of the sensor (via a 2.2Kohm resistor); either way round but display may read - instead of +V. The display in the picture is reading 19.37 degrees C. If you want the display to read above 20 degrees you will need to add a couple of resistors into the circuit.

Glad to answer any questions about this project.

I haven't done many tests yet but it does seem to work. It's certainly going to be more accurate than most thermometers out of the box.

Mike.

[tumutbound]

Any plans to try this sensor out with some sort of micro-controller?

My weather station plans are to build a small micro-controller based unit with something like the RFM12B transceiver for communicating back to a Raspbery Pi.
Still looking at which low power micro controller to use but information on suitable (cheap!) sensors always comes in handy.

Regards
Geoff

[Repairman77]

No plans to do anything like that Geoff.

My main reason for the project was to find some way of calculation the differences in accuracy of all my analogue and digital thermometers. Some of the very cheap Chinese ones can be miles out. Found one yesterday that reads 2 degrees high! Most of the others are +/- 1 degree or a bit less.

Have ordered another sensor from another supplier and hence a different batch (which will make three) so I can get an average between them.

Mike.

[AllyCat]

Hi,

Another "classic" temperature sensor is the DS18B20; a little more expensive, with a similar accuracy. But as it gives a digital output (SPI "one wire" bus) to 1/16 degree C resolution, there are no further potential errors of D-A conversion and/or reference voltages, etc..

Still looking at which low power micro controller to use

Have you looked at the PICAXE? Low power (<1 mA, 2.5+ v, with further reductions possible by sleeping) and an "easy" learning experience (a rather basic Basic). If you already have a "bits box" and a COM port on a PC (or suitable USB-serial adapter) then you can be literally "up and running" in a day, for less than £5.

Cheers, Alan.

[prodata]

Another "classic" temperature sensor is the DS18B20; a little more expensive, with a similar accuracy. But as it gives a digital output (SPI "one wire" bus) to 1/16 degree C resolution, there are no further potential errors of D-A conversion and/or reference voltages, etc..

The data sheet seems to suggest only 0.5C accuracy. Greater (programmable) resolution than that certainly, but that's obviously something completely different. But maybe I'm missing something?

[AllyCat]

Hi,

I wrote "similar accuracy" because AFAIK +/- 0.5 degree C is also the claimed accuracy of the OP's LM35 ("tested limit +/- 0.5 @25C, +/-0.25 linearity") ?

Cheers, Alan.

[Charlie]

Another "classic" temperature sensor is the DS18B20; a little more expensive, with a similar accuracy. But as it gives a digital output (SPI "one wire" bus) to 1/16 degree C resolution, there are no further potential errors of D-A conversion and/or reference voltages, etc..

The data sheet seems to suggest only 0.5C accuracy. Greater (programmable) resolution than that certainly, but that's obviously something completely different. But maybe I'm missing something?

To expand on Alan's comment, both seem to have the same initial accuracy, but the DS18B20 outputs a digital number so all possible errors are controlled within the device, and that 0.5C includes ALL errors. The LM35 outputs an analogue signal, so there are quantizing errors, reference voltage errors, line loss errors, self heating errors... plenty of opportunity to add additional errors in whatever method you use to measure the voltage, compare it to a reference, and display it somehow. Unless you are engineering the entire solution carefully, the DS18B20 will likely give you more accurate results.

There are online sources of the device for just a few pennies. You will need a method of reading the device, however which could certainly add to the cost if you are doing a "one off" design.