I thought about that but decided not to because i was unsure about it leaking back into the output. If a simple diode takes care of it then i guess the rectifier would do the same thing.

I worked out that witout a capacitor or anything that the PIC could get up to 0.38V less than what it should do. That equates to about 78 levels difference out of 1024 or a 7.6171875% error. All that is based on my maths being correct but 7.6% sounds about right.

I just had an idea. dhouston mentioned zero-crossing. I already know how to detect that thanks to the PWM circuits ive made before. What i could do is wait for a zero crossing then wait a certain amount of time then run a few samples. That should put all the samples closer to the peak.

The capacitor idea sounds really simple though. If the capacitor can keep the level high enough between peaks then in theory i could sample each one once every second. I would like a single chip to sample all 8 if possible and send the data to the PC through serial so i think i do really need the capacitor.

What is the purpose of a "bleed" resistor? Is that to go in series or parallel with the capacitor? I would need the capacitor to empty pretty quick but not as quick as the frequency of the input. I dont want to have to wait 10 mins while it realises the output isnt drawing as much current anymore.

Edit: I plan to buy an oscilloscope with these components so it will be easier to test and see the sinewave