Just out of curiosity, and a little desperation, I am going to try to fit a LPF in there, with a 100K resistor, 50 nF capacitor, for a T of 5 ms and a cutoff frequency of ~30 Hz. I wanted it to be lower but I don't want a large T. See what happens. I will later have you know how that went.
André
EDIT: two low pass-filters at 10 Hz before the ADC (one for each sensor), a severe cleaning of the ph probe, and a new lm35 solved my problems. Didn't manage to find the noise source though. Thanks for all the help guys.
EDIT 2: Just remembered I hadn't updated the post, for future reference to anyone who might need to know. Actually I did find the source. Some variation still shows up (despite the 10 Hz LPF) in windy conditions, even with a turbulence shield. In fact, the responsible for the noise is the water tap that inserts water in the system for recirculation (not the tap particularly, but the water in itself), because if I close the tap, the noise is gone. I guess it has to do with the waves that form in the water bay where the water is captured through a pump, on windy conditions. My mech. engineer uncle suggested that the waves could change the pressure in the pump's inlet which would variate the work, and it somehow was interfering with the system, maybe a cavitation effect or a low frequency noise.