Search the CONFOCAL archive at
http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal
Hi all,
I think that Vitaly may be telling only part of the story.
1. Using 16 micron pixels is something that you almost have to accept
if you want back-illumination (Because photoelectrons formed on the
"back" face of pixels that are much smaller but are 8-9 microns from
the charge-transfer electrodes and may drift sideways, to be
collected in neighboring pixels).
The peak QE of such sensors can be >80%, which is good but if you are
using such a sensor with the electron-multiplier gain turned on so
that you can count single photons above the background noise level of
the FET readout amp, then the multiplicative noise of the EM amp
effectively reduces the actual QE by half (i.e. to 40%. Good, but
nothing to brag about)
2. The "significant loss of light" is also rather undefined. If you
use a 2x coupling tube rather than a 1x, on average, each pixel
SHOULD have only 25% as much signal (assuming the same microscope
set-up, specimen, focus plane, illumination, and exposure time etc.).
In fact, it will be a bit less because the extra optics will have
some reflection losses what may reduce the signal to, say 22%. If
you get greater losses, something is wrong. Of course, as S/N depends
on signal level (in photons, not arbitrary computer units), the 2x
signal will seem to have more noise because of Poisson Noise, and if
the signal level is very low (<50 photoelectrons?) then the CCD read
noise will become more apparent.
However, if you wish to collect high-resolution information, this is
what you need to do .
3. The comment about confocal having "0.3 light collection
efficiency" seems a bit misleading. With a water lens, no more than
one fluorescent photon in 4 even strikes the glass of the objective.
Further losses can occur at the dichroic (up to 50%) and the barrier
filter. So far all of these factors affect widefield CCDs in the same
way that they affect whatever detector the confocal uses.
It is true that the confocal pinhole CAN be set too small, but if it
is set to be the same size as the first zero of the back-projected
Airy pattern, it will pass 80% of the light emitted from the plane of
focus. This is the signal you want. You may get more photons by
opening the pinhole, but only as long as there are fluorescent
substances above or below the plane of focus. Widefield operates as
though the pinhole is infinitely large. Although, from a 3D object,
more photons/pixel will strike the detector when using such an
"Infinite pinhole," the extent to which doing so provides more
information about the structure of the specimen is not so clear. This
is another very contentious topic but I think that most will agree
that photons emerging from objects more than a few microns from the
plane of focus will contribute more noise than signal.
If leave this topic and limit our consideration to the QE of the
detectors themselves, at red wavelengths, the effective QE of a
properly-operated PMT is no more than 10x worse than that of the best
CCD (i.e., 5x worse than an EM-CCD used in the EM mode). This
difference is 2x less in the green and less still in the blue. But
this reduction must be balanced against the fact that out-of-focus
light, and the Poisson Noise introduced by counting it, is also
substantially reduced.
With the disk-scanners, one can almost have the best of both worlds:
Good QE of the EM-CCD as well as the exclusion of far-out-of-focus
light by a set of pinholes. However, problems still remain in terms
of the difficulty of changing the effective pinhole size (i.e.
needing high-NA objectives of different mag), and difficulties in
spectral imaging.
EM-CCDs with smaller pixels are available now and more are coming. It
remains to be seen if they their CIC and read-noise performance will
match those of the present E2V 16 micron chips.
In the mean time, I think that the best plan is to buy a 2x camera
coupling tube. (and, if you are using this 16 micron sensor in the EM
mode, take some images with no excitation striking the specimen, just
to make sure that you aren't picking up a lot of stray light. Keep
turning off room lights until you only get 2-3 noise pulses per 512
pixel line.)
Cheers,
Jim P.
>Search the CONFOCAL archive at
>http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal
>
>Dear Michael,
>
>Thanks for the commercial input.
>
>I do not know who is working with the 3.5 micron pixels CCD camera.
>
>However, I have one argument.
>How many labs have an EM-CCD camera with 8 micron pixels? The anwser
>is rather obvious.
>
>Many labs have Roper's Cascade 512B or Hamamatsu EM-CCD with the
>"gigantic" pixels.
>
>I tell you it is impossible to avoid undersampling without
>significant loss of light and working at physiologically relevant
>concentrations of fluorescently tagged biomolecules.
>
>Anyway, it is known that a (confocal) microscope is a sort of a
>primitive light-collecting device (ca. 0.3% light collecting
>efficiency, isn't it?).
>Sensitivity is a serious issue here, but not at the cost of undersampling.
>
>I do hope that Roper's or Hamamatsu sales men would be less bullish
>and more objective when pushing on 16 um pixel CCDs.
>
>And I do believe that the "gap" in misunderstanding between the
>Roper's, Hamamatsu, etc. engineers and biologists would narrow down.
>
>Cheers,
>
>Vitaly
>
>NCI-Frederick,
>301-846-6575
>
--
**********************************************
Prof. James B. Pawley, Ph. 608-263-3147
Room 223, Zoology Research Building,
FAX 608-265-5315
1117 Johnson Ave., Madison, WI, 53706
[log in to unmask]
3D Microscopy of Living Cells Course, June 17-29, 2007, UBC, Vancouver Canada
Info: http://www.3dcourse.ubc.ca/ Applications due by March 15, 2007
"If it ain't diffraction, it must be statistics." Anon.
|
