Galvanic skin response, mains hum

[Buriedcode]

Hi there,

I've spent a day fiddling/breadboarding various simple opamp circuits for measuring galvanic skin response.  Whilst absolute resistance (or rather, conductance, uS) would be nice, for now, I'm just trying to get a basic circuit together.  Twocircuits I've tried (didn't think much of them, but might as well test out what others have done):

http://produceconsumerobot.com/truth/

Same opamp so the input bias current is tiny.

With any high impedance sensor like this, mains hum will always be an issue - that circuit only has a single pole low pass.  The second circuit includes a high pass so it picks up slow changes rather than absolute.  In order to try and eliminate the 50Hz mains hum, I'm sampling with an Arduino at 50sps.  I also tried sampling at 200sps, and averaging four consecutive samples.  In both cases there seems to be a slow oscillation, period about 6 seconds.  Is this the difference between my Arduino's clock, and the mains frequency?  I realize they will never be exactly the same but without some sort of PLL so the sampling tracks the interference I don't see how others managed to get a relatively 'clean' signal.

As I said, this is basic stuff so far but googling seems to throw up either hobbyist results (basically what I've done) or pseudo-science (scientology anyone?).  Perhaps it would be prudent to use AC at a frequency low enough to pretty much just measure conductance, but high enough so the reciever can filter out mains and other noise.  If anyone has any idea's, experiences, or know how 'commercial' devices do it, I'd be grateful.

I'm also aware that there isn't much merit in the measurement, I don't give much credence to 'lie detectors' or 'wellness' products, but I was tasked with at the very least measuring it - letting others interpret it as they wish - and wanted something that hopefully will at least pick up a pain response (been jabbing myself with IC pins for the past 15 minutes, no dice).

[StillTrying]

Average a number of readings spread evenly over a time of 20,40,60,80 or 100 miliseconds.

[Buriedcode]

Well with heavy filtering (100n cap with 1Meg feedback resistor) and a lot of averaging as suggesting, the very slow oscillation has disappeared.  However, a new problem has arisen which is annoying.  It appears my skins resistance is in the order of 8-20Meg  

The back of the wrist isn't an ideal location, but as I mentioned I don't give much value to this measurement, it would be nice however to know roughly the range of resistances I'm looking for.  Various websites have graphics with *roughly* 0-10uS on the Y axis.  This paper is an example:

http://affect.media.mit.edu/pdfs/10.Poh-etal-TBME-EDA-tests.pdf

The graphs there indicate an average of 1-3 uS.  That equates to 330k - 1Meg resistance, and order of magnitude lower than anything I can get on my skin. With the obvious exception of the fingers, where it sits at, again, roughly ~1.2Meg.

I realize its a very 'ish' measurement, but I don't' want to set to to pick up 10-20Meg based on my dry skin, when others have a much lower resistance.  Any chance some of you could test your skins resistance with a multimeter?  I have yet to see any 'response' to a sharp intake of breath, or pain (its getting old now) and I suspect its just me, rather than the setup.