Having been the lone voice in the wilderness promoting 12 bit when everyone
else was 8, this thread is just too tempting.

Disclaimer: Commercial vendor, potentially but not necessarily biased...

If you are doing the same thing dozens of times to gather statistically
significant data, extracting the maximum amount of data from each scan is
attractive and helpful.  Our very first laser scanning instruments
(ACAS/Ultima) offer the choice of 8 bit and 12 bit, primarily because they are
heavily used for automatic quantitation, cell sorting and image cytometry.
However, if you want pretty pictures to add sparkle to a paper, extra bits
become a waste of space and a computational burden.

If the validity of an observation hinges on one in 256 grayscales let alone 1
in 4096 it is not worth publishing.  And the subtleties of image quality are
not going to show on thumbnail journal images.  Most of the variance seen in
confocal data sets comes from specimen preparation (dilution errors,
incubation time, laser power, temperature etc.) and effort spent on that side
of the experiment is where rewards are greatest.

With our latest offering (the TR) we collect from the PMTs at 10 bit, and
after averaging, shunt the data around and process it at 8- bit resolution.
This keeps the data compact enough that we can collect up to five channels
simultaneously, process it quickly, and store it economically.  We use a
similar trick with our InSight ocular viewing system, saving the most
significant 8 bits from a 10 bit or 12 bit digital camera.  This economy is
especially important when using high pixel densities and multiple channels
because Z-series and kinetics series contain tens to hundreds of individual
images.  Zip disks start looking like floppies.

David Carter
oQQQQQ@

InSight Biomedical
010 Filley Street
Lansing Mi
48906

tel. 517 327 0950 x23
fax 517 327 0959