Yeah, really, weekend! Yahooo! What a week; I trade stocks for a living, and it was a good week, but with a lot of white knuckle moments -- the equipment budget for next year is looking good!
I'd been thinking of what you're doing, and came to the same conclusion -- a semi sealed chamber with drierite, or silica gel in it.
I've also been looking at putting the Am into a small lead "pit" so the long range alphas are either eaten in the lead (no ions or very few), or fire down the tube long axis so they are fully absorbed in the gas, rather than a random mix as would be the case in a small chamber and an isotropic radiator (or a half sphere, which is more probably what you have).
The range of alphas in air is a few cM, and most of the ions happen at the slow end, far end, of that range. So making sure all the ones that count at all, count fully, should reduce the effective noise -- as you pointed out, you've got ion current to burn, so it might be good to burn a little in a good cause, getting the ones you get more accurately measured. The idea there would be to use a small diameter tube but long enough to get all the ions from an alpha -- say 3/4" by about 4".
But you are going to see some long term drift that isn't the fault of the electronics and could be either way. This is because of radon, and I've seen it in some test samples I made out of U2O3 here. Sealing them too well let the radon collect, and the daughters of that (which are all hot, but solid) also. There is a term mentioned in the books "secular equilibrium" where all the daughters and the parent finally reach some stable ratio, and you'll see a drift while that happens. The radon daughters unfortunately will tend to collect inside the chamber no matter what you do, unless there is some xfer with the outer air (and even then some will happen). Those are all solids and very hot, short life stuff, which then decays into longer lived stuff.
The effect in my small samples, basically some stuff sealed between microscope cover slips on a 3/8" washers, was on the order of 10% in a couple months. Big shift. Glad I recalibrated.
If I were designing one of these today, what I'd do would be to make a cylinder chamber (copper pipe or similar) with the central rod not quite reaching the end far from the feedthrough, then make a cap for the other end that held the sample on the inside, and some dessicant on the outside such that the aperature for desiccant was large for the chamber innards, and tiny for the outside air (like a 1" piece of .005" id capillary tubing). The idea is not to ruin the desiccant too quick, but to have all that gets into the chamber on the usual thermal cycles pass through it.
As you say, if the central rod is large, and smooth (most people forget that part) you're not going to have issues about gas gain unless the volts get pretty high. I use anything stiff, and like tungsten TIG rods. Heck, with one that has the thorium doping, you might be done right then! But then you'd have to make the thing large diameter so most of the alphas get absorbed in the gas before they hit the walls and knock out variable numbers of electrons. The Th in those isn't that much, so that might put you back in a bad signal to noise situation, just a thought.
Considering what you normally do for a living, am I making sense here? It's kind of the same, rooting out all the variabilities so you get a better result.
Yes, I had a good laugh on the keyboard wedgie -- but hey, if it works, it works. I've seen some projects that simply emulate the rather strange keyboad protocol too. But what I use for that kind of thing is this:
http://www.coultersmithing.com/OldStuff ... gmoth.htmlThere is windows software (last I checked, been awhile) to redirect a serial port to look like a keyboard, and in linux (which is what I use) well, its "mine to toy with". The board is pretty simple to use with our opsys and driver collection. Windows has recently made it a little tougher for an app written in a grown up language to have hardware access -- their security model is not to fix the insides but make it hard to get to the insides. Thus, custom drivers or even talking to a serial port are kinda getting hard to do there outside of VB and that activeX control, assuming that still "works" as well as it used to (not that well). The pic has two rs232 outputs, we send one to a PC and hold the other one back for printf debugging uses.
We are doing a lot with pics here, and a little perl shoves the rs232 coming out of them into a mysql database, and now open office can talk straight to that -- all separate processes, very nice results and flexibility. Perl's not great for big complex programs but it's the worlds most fantastic duct-tape. We do C and asm in the pics.
Posting as just me, not as the forum owner. Everything I say is "in my opinion" and YMMV -- which should go for everyone without saying.