Mcu Lockup -power issue

Hi All:
Recently I 'upgraded' a nokia AC 3U charger to deliver 13.4V as a trickle charger for a Gel Cell that runs a hydroponic controller based on a 16F886 chip.
Well, the AC 3u is a 2 prong, ungrounded switching supply which seems to freak out the 16F chip after a few mins. No amount of capacitor decoupling, ferrite beads, ferrite clip ons etc. did any good. If I remove the std. decoupling cap (o.1uf) the MCu locks up real fast - a few secs, otherwise it lasts up to a 15 min with it.

But if I try to probe with a 'scope, attaching the scope's gnd clip sends all the lockups away. Thus I can't see what's happening. Since the ground lead 'solves' the lockups, I imagine there's some 'noise' on the supply's ground that is suppressed when a true ground is connected.

I'd like to be able to function without adding a ground prong to the AC 3U supply.
So any thoughts?

If its ground problem try using a virtual ground. Might work.

This is resolved. It seems there is a relationship with the 'noise' on the PIC supply and settling time. Increasing the settling time of the adc channel switching did the trick.

If I use atod and more than 1 channel I write a interrupt driven routine, first iteration stores a converted analogue input value and selects the next atod analogue channel to convert, second iteration starts atod conversion, third iteration stores that converted value and selects the next atod analogue channel, fourth starts a conversion, and so on.
I've used this a few times, so long as the interrupt period is more than the atod conversion time there isnt even a need to check the go_done bit as you know it'll have completed.

Also If theres not been a 0.1uF cap accross the supply pins of the micro I've had it where the Go_done bit once set never clears, watch out for that it'll hang your program.