I know a couple of people on the list have mentioned having
their products tested for EMI interference and what they did
to stop it, hopefully they'll be able to give you some advice

> Used an old ATX power supply metal box,

Does this box have any holes in it ? If it's like every other
PC supply I'm guessing it does. The noise will find its way
in through those holes. Try blocking all the holes with
aluminium foil, see if that makes a difference. Ideally you
want an unbroken metal barrier all around the circuit, which
includes the wires

http://en.wikipedia.org/wiki/Faraday_cage

> and grounded the sensor, my circuit and the power source to
> the box. I got less noise, but still got it

Were the connections with shielded (eg good co-ax) wire ?

>     If i disconnect ALL the wires and sensors from the board,
> and let only the power supply and serial connection, the problem
> persists. Yes, if i "click" the piezzoeletrical ligth firer just near the
> metal box, the pic counts pulses! Without any wire connected!

Sounds like the sensor wire either isn't the problem or isn't the
only one. Have you looked at all the PIC pins with a 'scope to
see where the noise is ? Could be difficult to measure without
having a roll of kitchen cooking foil handy to seal any gaps
you make to get a probe in

>     what options more i have?

Do you use a crystal and have its case grounded ? In the product
of mine I mentioned, as unlikely as it sounds, the software doesn't
work properly without the crystal case grounded

Are you sure the PIC is counting pulses or is it resetting ? Or
even both ? MCLR can be sensitive and it doesn't take much
noise to cause problems. I'd focus on keeping noise off Vcc,
Vss and MCLR

-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist