Stable ADC readings

Hi
I'm trying to use a BV500 to read 2 temperatures. I'm using MCP9700's connected to 3.3v & Gnd with the outputs being read in on ADC2 & 3. It's set it to sample 1000 times and then divide by 1000 to give an average. This appears to work fine until 1 of the sensors is warmed and then both reading increase, the warmed sensor increases from (say) 1800 to 2400, but the other sensor increases from 1800 to near 1900. It appears that the 2 ADC channels affect each other.
Are there ADC channels that you would recommend, so as not to cause any cross effect or is there a different method you would suggest? Thanks

I don't know what is causing the problem, ADC is not always as straight forward as it should be.

I can suggest a few pointers

1) verify that there really is cross talk between the two channels by using two potentiometers (or 4 resistors) with a total impedance of less then 10k. If the ADC is okay then the output impedance of the MCP9700 is too high, although according to the data sheet is only 20R

2) If there is still a problem when using the pots. then look at the sampling rate and acquisition time, slow these right down or speed up see the effects.

3) Filter the output of the MCP9700 using a capacitor, say 10uF, assuming that temperature does not change that fast.

Hope that helps