[RASMB] XL/I Absorbance Problem (follow-up)

[Borries Demeler]

> 
> I think a couple of points regarding the use of pseudo-absorbance were
> not sufficiently brought out by either Bo Demeler or Peter Schuck:
> 
> (1) With pseudo-absorbance there are vertical shifts of the scans
> from one scan to another, somewhat similar to the TIN or 'jitter' in
> interference data. Treating this systematic noise as a true constant
> vertical displacement certainly helps, but if you look closely at the
> structure of these shifts they do not appear to be truly constant across
> the cell. One source of this effect is clearly the adjustment of the
> photomultiplier voltage based on the light levels at 6.5 cm, as Peter
> mentioned, and thus the shifts tend to be largest as the boundary moves
> past 6.5 cm.

John,

what we see on our instrument is a different scaling of the entire scan
if the gain setting changes mid-run and the absorbance is too high
(see attached image for a worst-case example - this one looks like it
has 5 different gain settings in it). Regarding radially invariant
noise (constant vertical offsets), you can also correct for that 
algebraically, but I don't find the need to do that with pseudoabsorbance
data. Most of the (small scale) RI noise we see in such data results
from variations of intensity in the lamp over the course of the 
experiment (lamp gets progressively dirtier, for example). To handle
that, I have implemented in UltraScan a correction approach that seems to
work quite well: Alongside the actual data you collect on reference channel
with water in it. For each experimental data scan, you also make a 
reference scan of one channel. Then you average a certain radial range 
of each baseline scan to get an average count for the intensity. This
value is used as your I_0 for the calculation of the pseudo-absorbance data.
In UltraScan, this correction happens automatically when you convert
to pseudo-abs data, and it seems to take care of the problem of drift
in the baseline quite satisfactorily. At least I cannot find any improvement
of the data when I fit either with or without RI noise correction.  

> (2) I've seen pseudo-absorbance data gathered by a number of users of
> DCDT+ and that strongly suggests that this approach really only works
> well when the absorbance in both sectors is pretty low. At higher ODs
> the systematic errors associated with (1) and other effects appear to
> get much worse.  Exactly how high an OD is reasonable to use depends
> on the quality of the data you need, so it is hard to give a general
> cut-off, but I would probably stay below ~0.5 OD. Anyone planning to
> use pseudoabsorbance should do their own testing to see what errors are
> acceptable for their particular application.

Agreed. In the machine I tested it looks like we are OK up to ~0.7 OD,
depending on wavelength, this may change, I only tried 230, which has
the highest intensity of all wavelengths in the xenon flash lamp.
 
> With respect to (1) it would be useful for Beckman to provide an option
> that suppresses the photomultiplier voltage adjustment after the first
> scan of each cell, and that also drops the assumption that the voltage
> should be initially optimized for measuring sample - reference.
 
That would be great! I asked our service rep about this and he said that
it was part of the embedded software in the eprom chips, so that may
not be so easy to change?

-Borries
-------------- next part --------------
A non-text attachment was scrubbed...
Name: pseudoabs.png
Type: application/octet-stream
Size: 46275 bytes
Desc: PNG image data, 697 x 835, 8-bit/color RGB, non-interlaced
URL: <http://list.rasmb.org/pipermail/rasmb-rasmb.org/attachments/20070823/9ec00ea9/attachment.obj>