ADC Voltmeter Nimh cell

Hello. I need some advice.
I'm building a custom nimh battery charger and i'm doing the ADC routines to detect the peak voltage of the nimh cell (independent cell charger).
The problem is that the readings are not stable/accurate.
I have tried Vref+ 1.800v but the reading are 20mv off, then with 4.5v i get the correct value sometimes, but it's not stable, even tried the average value but it's jumps all the time.

I tried a 15nP cap with gnd and a series resistor 1k in AN0 but it's the same.

What are the correct procedures to get a stable value?
Am i reading it too fast?

With vref+ 1.800v i get (1.8/ 1024) = 1,7578125 resolution witch is very good, but the adc reading get it's wrong by 20mV reading a single cell don't know why.

Some ideias plz.
Thanks

I would use 2.50V or 4,96V for V-ref
The best way woudl be to take 16 or 64 samples then work out the average and use that, If it's a PIc then these only do 10bit A/D which 20Mv is not bad at all
If you need beter accurate readng then why not use a 12/24bit A/D chip

wizpic said:

I would use 2.50V or 4,96V for V-ref
The best way woudl be to take 16 or 64 samples then work out the average and use that, If it's a PIc then these only do 10bit A/D which 20Mv is not bad at all
If you need beter accurate readng then why not use a 12/24bit A/D chip

Click to expand...

Hello wizpic. Thanks for reply.
Yes, it's a PIC 16F876 / 886 and i've got some 18F2550, but i think they are all 10bit.
I'll try your approach and see what i can get.