Terry Pinnell said:
Looking at this new illustration
http://www.terrypin.dial.pipex.com/Images/NoiseLayout2.gif
are you saying that the 'fairly noisy' and 'very noisy' circuits are
roughly as I've shown in green and blue respectively please?
Correct. As you show, the ADC is grounded via the parallel port cable,
which has to be one of the noisiest things you can imagine, since it's also
carrying nonperiodic high-f digital signals with relatively substantial
current. I would expect to see a few dozen mV at least from one end of
those wires to the other. In the 'very noisy' circuit, any voltage induced
across that cable shows up directly on the scope. Even in the 'fairly
noisy' circuit, any flaws in the internal grounding scheme of the ADC will
inject that noise into the scope. (Thus Chris C.'s observation about the
noise to be expected in PC scopes.)
The noise waveform looks nonperiodic: motor brushes and welders were
excellent guesses, and I think digital signal bleedthrough is a good
candidate too. Looking at a single trace with the ADC, you see individual
spikes; on the traditional scope, you're seeing many traces overlaid, so you
get the 'halo' you noted. But I also am going to guess - you've not told us
yet, I think - that on the traditional scope you're seeing less noise
overall. (Does it change depending on whether the ADC is on or off?) I
*don't* think that flourescent lights would explain the waveform you showed,
though with noise there are always surprises. IME it's more regular than
that.
I was envisioning the ADC as being a card plugged into the PC. Sounds like
that's not right. So, what I really mean is to use the Hameg to measure
from the ADC's ground to the mains ground. Although the same problems of
measurement apply - as a rule, it's hard to measure the voltage difference
between two points separate in space. Sort of like the relativistic problem
of synchronizing two clocks at different locations.
You might want to temporarily rearrange things so that you can reduce loop
area and see whether it makes a difference. I'm going to guess it won't
make much difference; I'm going to guess that whatever you've got is either
induced galvanically (that is, by a voltage source in series with your loop,
such as the PC), or capacitively (a signal outside the loop, coupled in by
leakage capacitances in cables or in mains transformers), rather than
inductively (electromagnetic signal inducing a current in the antenna formed
by the loop). But that's just an educated guess.
And by the way, I'll second Bill S.'s recommendation for Morrison. I've got
three of his books sitting on my shelf, including the one Bill recommends,
and have found them very helpful. I've also found Bill Whitlock's papers on
noise in balanced signal transmission, some of which are available on
Jensen's web site and others through the Audio Engineering Society, to be
helpful as a general primer. But Morrison contains much of the same
information.