Hey, now... just because I don't work in the field doesn't mean *my* chemical engineering degree counts for nothing.
Actually, being at 3000 psi *does* have some bearing on gas behavior. The higher density means the mean free path of the molecules in the cylinder will be reduced, which means diffusion effects may be slowed.
Take a very special room pressurized to 3000 psi and kept precisely at, say, 75°F. In this room, take an open cylinder (anchored securely in a massive, unmoving fixture) and fill it completely with 75°F nitrogen. Now, *very* slowly pour (through a straw, from the bottom up) enough 75°F oxygen into the cylinder to make it 21% oxygen inside. *Very* gently secure the valve. If you did this *just* right, you now have a cylinder with the oxygen on the bottom with the nitrogen floating on top. As everything is kept *precisely* at uniform temperature, there is no convection in the gases inside cylinder. As the cylinder is held "perfectly" still, there is no inertial "stirring". In this situation, the mixing inside the cylinder will be limited to diffusion effects only. It would, of course, be quite prohibitive to actually perform the experiment.
The thing to note, however, is that even in our very contrived example, we're not talking about the gases *separating*. The only thing we're doing is slowing down the rate at which the gases are *mixing* on their way toward the well-mixed state. Entropy (i.e. the randomness of the system) increases, as they say ("they" generally being people referring to the "second law of thermodynamics").
I *really* need to dig out my fluids texts and run the numbers. What I *can* say is this: If you're filling an 80 cubic foot cylinder at 300 psi/min, and if the incoming air is at 3000 psi and magically does not expand through the dip tube (basically, it behaves like a liquid), and if the dip tube's internal diameter is as wide as 4 mm, the air coming in through the dip tube will be moving at about 1.5 meters per second. Someone blowing moderately through a straw for 10 minutes ought to be enough to mix a small vessel of gases, eh?
(In practice, the linear velocity of the incoming air in our example will be greater with a less-full cylinder, gradually decreasing to that 1.5 meters per second by the time we get all the way to full. "300 psi" worth of gas is a fixed mass, at least when considered in the precision with which we're working here. A fixed mass of air at a lower pressure has more volume than the same mass at higher pressure, and when you pass that volume through a tube of a certain cross-sectional area in a given time, well, that's just arithmetic. :biggrin
*Anyway*, I don't have any CFD (computational fluid dynamics) software handy to do a numerical model of filling a scuba cylinder, but the thought that a single tank can be poorly mixed after filling *and* that simply rolling it will make it well mixed seems highly improbable on its face. (Rolling, after all, is about the second least effective way I can think of to mix fluids in a cylinder -- leaving it alone is the least.
)It seems *vastly* more probable that the analysis was faulty. Improper calibration, faulty procedures, and possibly thermal effects are my three guesses there. Was the analyzer calibrated with dry air from a scuba cylinder? (The "wave and smile" method is not reliable.) Was the proper flow rate across the sensor used, e.g. a metered inflator attachment? (If it's not metered, i.e. controlled by an orifice, it's "close enough" at best -- you can't precisely control flow rate through a tube pressed against the valve face.) If a metered inflator attachment was used, was adequate time allowed for the residual gas in the reg set to be exhausted before the measurement was accepted? Was the analyzer subject to significant temperature swings during the process? (If it was left in the sun and then used immediately, measuring the cold air coming from a cylinder, the sensor's temperature may change significantly from one measurement to the next as it cools down.) And of course, my favorite: If there was a dastardly knob, did I bump it *again*? (I've done more calibrations than accepted measurements on some bad days.
)Anyway, just a few more cents (a few likely being far more than two, considering this post's length).
