Unstable reading on ADC

Greetings, all.

I have a Picxe 18M2 I am using to read a temperature from a thermopile, running the thermopile through a basic op-amp to get the voltage to a usable level. For both readadc and readadc10 the data varies widely, by a range of 15 or more in debug on the readadc command (i.e. 8-bit resolution). The voltage from the op-amp is much steadier than that when tested with a voltage meter, though it does vary by 2-3 thousandths of a volt. (e.g. from .752 - . 755 volts). I have read that the FVR can vary by a bit, so I wondered if that might be it, but a potentiometer hooked in to the chip on readadc is rock solid, and a different sensor (a Hall effect sensor) run through the same op-amp is also quite stable.

So, any ideas on how to stablise the reading from the thermopile, or at least any thoughts why a thermopile would be so wonky? Specific chip data is listed below if that helps any.

Courteously,
The Professor (601)

Picaxe 18M2
ZTP135 thermopile (optical remote temp sensor)
LM358N dual channel op amp
470k (total) to 1k resistor multiplier on the op amp
All running on 4.5v battery pack

Welcome to the PICAXE forum.

The multimeter may be averaging the voltages read so it may be giving the appearance that there is more stability in the through op-amp reading than there actually is. Do you have an oscilloscope which would allow you to get a better insight into its stability ?

Post a picture of your circuit (especially the op-amp). There may be somewhere you can place a capacitor to reduce the high-frequency (Ie noise) gain.

Hi,

Yes, a (small, non-electrolytic) capacitance across the 470k would be wise. And just to confirm, you have configured the PICaxe to use FVR_2048? By default, PICaxes use the supply rail as the ADC "reference" and neither FVR_1024 nor FVR_4096 are recommended as the ADC reference (at your supply voltage).

Cheers, Alan.

I am an idiot. The ideas you offered were all excellent ones, and as I began to explore them I learned a really important bit of data: nothing works terribly well without a good solder connection. Reconnected the input wire to the board and now have things stable to + / - 2 which I consider pretty good for a thermal read. I doubt I could make a proper thermometer out of it without a lot of averaging but for what I want as a relative temperature measure it works.

Thanks for the ideas.

Courteously,
The Professor (601)