Different result if connecting 9V?

The TMP36 output is 10mV per degree C and is designed to be 750mV at 25C.

The sensor is specced for -40 to +125, which means a voltage range at the output of 0.1V to 1.75V. That range is designed to allow operation off of a low voltage supply like 3.3V or less, but it does mean that in this case you are only using about a third of the A/Ds range.

I'm not really familiar with the Uno, but it looks like the A/D is 10-bit and is operating ratiometrically [taking the supply as its reference voltage]. Let's try calculating the reading for 25 degrees C (ie an input of 0.75V) for two different supply voltages [I've just arbitrarily chosen plus or minus 4% from a nominal 5V]

supply 5.2V, reading 148  (0.75 / 5.2V) * 1024
supply 4.8V, reading 160  (0.75 / 4.8V) * 1024

Each step from the A/D is equivalent to approximately 1/2 a degree, so that [quite modest] change in the supply constitutes about 6 degrees C in the calculated temperature. Since you are seeing about 2 degrees, it means your supply change is less than that.

Point is, the design is very sensitive to the supply voltage.

If you want it to work accurately regardless of the supply voltage then there are several options:

1) use the internal 1.1V reference of the ATMEGA328. That will be accurate and won't vary with the supply voltage. Limits you to 60 degrees C if you connect the sensor directly. You could get the full range if you use a potential divider on the sensor output.

2) provide your own reference and set the processor to use that rather than the supply. A 2.5V reference into AREF would be appropriate. [It might be that the Uno software would conflict with having an external reference, so have a look around and see if this is generally done or not. If you are still not sure, put a resistor between the reference output and the AREF pin so that you don't have two outputs driving each other directly and look at the voltages either end to see if there is any difference before your own code runs and sets up the processor to use the external reference.]

3) A final alternative would be to amplify the output from the sensor so that it makes better use of the A/D's input range. The effect of any supply change would then decrease in proportion. So if you amplified it 3 times, the effect from the supply change would be a third. If you don't need the full -40 to +125 degree range, you could amplify it even more with just the range you were interested in giving a full 0 to +5V at the A/D pin.

All would require you to change the way the software calculates the temperature from the result- so only try it if you are confident programming, though I'm sure you'd get a lot of help from people here if you wanted to give it a go.

And the TMP36 could do with a decoupling capacitor across its supply pins.

The same problem is elsewhere too - I just saw this which uses the exact same TMP36 part but with a Raspberry Pi:

http://raspi.tv/2016/using-mcp3008-to-measure-temperature-with-gpio-zero-and-raspio-pro-hat

The layout shown at that link uses the TMP36 and an MCP3008 but no decoupling : (

The £25 'experimenters kit' with these ICs contains no decoupling capacitors either.

My basic 'experimenter's kit' is still a bulk pack of through-hole resistors and capacitors from Maplin, and loads of SMD Rs and Cs from Farnell

and this very nice SMD component containerthis very nice SMD component container - I have several of them. They are the best I've found, if they are dropped the parts don't get

mixed up like other containers, nor do they leak from gaps if turned upside down. I use small stickers from a stationery store to label them.

I just left them (on raspi.tv) a post asking about the decoupling caps - I feel it's such a shame that so many people jumping on the "education" bandwagon seem to know so little about basic practical electronics.

MK

Hi Michael,

Good idea, I hope he corrects it, I just saw there was a corresponding kickstarter which achieved >1k backers so quite a number of beginners will be watching those videos..

Hi Michael,

Good idea, I hope he corrects it, I just saw there was a corresponding kickstarter which achieved >1k backers so quite a number of beginners will be watching those videos..