This is the circuit I have:
How long wire to sensors? Shielded cable? May be it would make a difference to put a cap close to inputs.
Your power supply should not be having such an effect and normally when this happens is because the impedance of the input to the analogue inputs is too high. When this happens you also find that one input appears to affect the others.
Does this PSU have a transformer in it or is it a switcher type? What is the scale of the voltages shown in the oscillograms? 20V/div - 2V/div - .2V/div. If it is as I suspect a scale of .02V/div the noise is normal and an artifact of the measurement setup rather than real noise. Or the results of the radiated magnetic field inducing voltage in the test setup wiring. The setup I describe below will answer why there might be noise from "Clean sources" as well.For Example... consider an Uno connected to a computer and an external good clean 1A 9V out switching PSU With my test board I measure 100mV -1V of Noise on the PSU + rail or output with a Tektronics 2213 oscilloscope. Unsatisfactory PSU ? NO it is ground common mode voltage or the slight difference in the Grounds between my O'scope and my Computer. The Proof for me is to disconnect the computer from the Uno leaving the PSU to power the Uno. 0 volts PSU noise is measured when the computer is disconnected.Doc
Cables are about 20cm, non-shielded. How big cap could be of use?
Actually I didn't see any interference between inputs, it just seemed logical that current could flow back through another input's resistor and affect it, not much of course.
Oh, it's short. Probably you don't need caps after all.I'd investigate this issue you have with a sparks . Definitely something wrong with electrical side. Are computer and PSU connected to the same outlet? Can you measure voltages between cases PC and PSU? Does PSU has 2-wire or 3-wire cable?
Not quite getting that, how do you thing that could happen.