im never a fan of giving up data (ie filtering before recording), there is no reason you can't capture everything and filter it later if you know how to building a filter (me and lurker were talking about a differentiation filter). the problem then i picking your corner frequency on your filter (basically slope). idk, i'll try things out next week when i have some time.
If the filter is low-pass you won't lose any information that's relevant to the test. All measurable effects I'm seeing there are well below 500 Hz, and it looks like your sample rate is what, 1ksps? I don't see any line hum, so that's good, but It's probably below the sensitivity of your scope or buried in all that other crap.
There is a reason you don't want to filter after you've digitized your signal. Before it goes in the 'scope, it's an analog signal so it has all frequencies, from 0 to infinity. When you digitize that signal, every one of those contributions to the signal gets stuffed into that data point. But because your sampling rate is less than infinite, you can't distinguish between signal and noise at a higher frequency. You won't be able to get rid of any noise that's at a higher frequency than half your sampling rate. And even eliminating anything close to that frequency is difficult.
As a rule of thumb, when it comes to noise the earlier you can nip it the better. And with a properly built filter, you won't be losing any information. The other thing you could try is hooking your sensor up in differential configuration instead of single-ended (just an assumption on my part that it's not already).
If I'm guessing right at what a "differentiation filter" is, I don't think you'll have much luck because the "slope" for the noise is probably less than some of the signal itself, and greater than any details you're trying to get out of the noise. All said and done it is a pretty strong signal.
What sensor are you using (model # & MFR) - I might poke around the spec sheets to see what's up.
Edited by Egomaniacal, 03 July 2011 - 11:22 PM.