Here's a summary of results for the two gascell tests Margie mentioned:

Test 1274-1277:
--------------
This was the CO gascell test.  We got good complete info on the
ROVS temperature (299.5K) and CO pressure (20.0 torr).  As expected,
the CO only shows up in one module, M2b. There was some visible
CO2 contamination; looks like ~0.04 torr of CO2.

It is hard to say if this data is OK.  At least a few channels
are clearly off, but overall nothing jumps out as wrong.  But
when I tried comparing the measurement with a a calc'ed spectra,
it was immediately obvious something was very wrong with the
pathlength or gas amount.  We were expecting 1.47 meter (58
inches), but if the pressure is correct, the pathlength looks
to be more like 1.04 meters.

Fit results:
foffset    = 0.154 cm-1
pathlength = 1.041 m
res. mult. = 1.042



Test 1266-1269:
--------------
This was the CO2 gascell test.  We didn't get much info about
the gas cell conditions, just "CO2 at 10 torr in ROVS".  I
assumed temperature was ~300 K like the other gascell runs.

According to a "Test Performance Summry Report" from 30 Mar 99
by Margie: "A preliminary data reduction shows that some modules
were saturated, and unexplained level shifts in other modules".

A quick look at the transmittance data (from the ratio of hot &
warm with and without gas) looks pretty bad.  Many channels in
M11 are above trans=1 by a few percent.  M12 looks much better,
but still shows some problems.  The M2b trans looks very
suspicious, while the M1b data is pretty much garbage.  M12
is the only data worth messing with.

As with the CO, before I did the fits I compared the measurements
to the nominal calc'ed spectra.  Again, the pathlength or pressure
is way off.  After a little playing around, it looked like the
pathlength might be similar to what I got for CO.  However, that
still didn't fix things.  I did this stuff in a hurry, but it
looks like it'd be impossible to fit the data without applying
a significant stretch to the transmittance.  (For all the other
fits I've just done small 1% or less trans offsets when needed).
I ended up scaling the trans to reduce it by 5% (that is,
tau_new = tau * 0.95).  That was a rough first guess correction
and no doubt it could be improved, but what's the point...this
data is clearly not good.

The fit results were:
foffset    = 0.034 cm-1
pathlength = 1.043 m
res. mult. = 1.060


Scott.
Scott Hannon, Thu Jun 10 1999 (hannon@umbc.edu)