Posted by
James Pawley on
URL: http://confocal-microscopy-list.275.s1.nabble.com/GaAsP-PMTs-tp5949611p5951733.html
*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy*****
>*****
>To join, leave or search the confocal microscopy listserv, go to:
>
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy>*****
>
>Before people get too carried away with these
>hybrid devices, detectors with APDs are
>non-linear at higher light levels. For confocal
>with (say) a 1 us dwell time this means you must
>arrange to have <10 photons per pixel. A second
>issue is gain loss with age in APDs although
>with most of the gain being provided by the
>cathode-APD acceleration voltage this may be
>less of an issue. This count rate limit may be
>overcome with array APDs but they introduce a
>loss of quantum efficiency and 'after pulsing' .
>I guess what I am saying
>is be careful in detector selection, they all
>have +/- points. But the improvement in QE for
>the new photocathodes is impressive (albeit at
>much higher dark count rates) .
>
>Cheers Mark
Hi all,
I echo Mark's cautions. There are long
discussions of these matters in Chapter 12 and
Appendix 3 of the Handbook. With respect to the
URL Mark sent out, ultra bialkali with a maximum
QE of about 43% looks very good BUT:
1) It occurs at a wavelength of 350 nm, well into
the near UV where we really seldom have need for
a detector in confocal-type micrsocopy.
2) Although no details are given, there is no
indication of how these curves were measured.
However, it is common to make such measurements
in terms of the current in nA leaving the
photocathode when a known flux of photons in a
given narrow wavelength band strikes it. The
ratio of the number of electrons/s in the current
to the photons/s in the light is the QE.
This sound good but:
a) Not all photoelectrons leaving the PC,
actually strike the first dynode. The 20-30% that
do not, fail to multiply and this represents a
direct proportional loss of QE
b) Not all of the PE that strike the
first dynode actually produce secondary
electrons. Partially this is just due to Poisson
noise: if the average first stage gain is only
say, 3, then for about 10% of arriving PEs, it
will be zero. It is actually more complex than
this and different parts of Dynode are likely to
have different SE coefficients. Again this lost
signal reduces the effective QE.
c) Such QE curves usually represent the
best that can be obtained. However, as the PC
must be evaporated onto the inside of the glass
after each end-window tube has been evacuated and
pinched off, there is considerable variation in
the thickness and even the detailed atomic makeup
of this film (and hence it's QE: thicker PCs will
have higher QE in the red, lower in the blue).
Even selected tubes may have a QE 20% lower than
the published specs (i.e., maybe 37% rather than
43%) and unselected tubes can be as much as 50%
less.
And then there is the matter of multiplicative
noise. Even on the best tubes set up in the best
way, (usually obtainable only when voltage
between the PC and Dynode 1 is 5-10x higher than
that between the other sets of dynodes) this adds
20% to the Poisson noise, and can only be
"compensated for" by using 40% more signal in the
first place (Because Poisson Noise is
proportional to the sqrt of the signal, to
improve the S/N by a factor of 2, you must
increase the signal by a factor of 4). In other
words, the signal out the back of the PMT acts as
though the QE is only about 70% of what it would
have been after taking into account all of the
processes listed above.
Multiplicative noise can be substantially
eliminated by using pulse-counting circuitry, but
as Mark notes, pulse-counting tends to saturate
at the signal rates common in confocal microscopy
(i.e., the levels recorded in the brightest parts
of the image, (where the dye is) will be less
than they should be, perhaps much less.)
The reason for this tedious detail is that the
"QE" performance of avalanche photodiodes is not
usually measured in the same way (The exception
being so called linear-APDs). APDs have so much
multiplicative noise (and lost signal from PE
that don't avalanche) that they are almost always
used in a pulse-counting mode. As a result, the
"QE" performance of pulse-counting units is
usually measured in terms of Photon Detection
Efficiency (PDE, there are other terms). A PDE
of 30% means that 30% of the photons of a given
wavelength that strike the detector will give
exactly one count in your image memory. Clearly a
PDE of 30% can give you a far more accurate
measure of the signal related to a given pixel
than one would get using a PMT having a raw QE of
30% but which is then subject to all the other
problems noted above.
But as Mark says, because APDs have to count
pulses, they are just not yet suitable for
"normal" confocal, where stain levels and other
variables mean that we are often surprised by
higher signals than can be handled by the
counting circuits.
Finally, the Subject line of this theme talks
about GaAsPMTs (but I could not find the first
post). GaAsPMTs are interesting because their
high-QE performance extents far into the red.
Unfortunately, this performance relies on being
able to create a PE using a low-energy photons
which in turn implies very dark current unless
the PC is either very small or is cooled (or both)
Cheers,
Jim Pawley
***************************************************************************
Prof. James B. Pawley,
Ph. 608-238-3953
21. N. Prospect Ave. Madison, WI 53726 USA
[hidden email]
3D Microscopy of Living Cells Course, June 11-23, 2011, UBC, Vancouver Canada
Info:
http://www.3dcourse.ubc.ca/ Applications due by March 15, 2011
"If it ain't diffraction, it must be statistics." Anon.
>On 23/01/2011, at 1:59 AM, George McNamara wrote:
>
>>*****
>>To join, leave or search the confocal microscopy listserv, go to:
>>
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy>>*****
>>
>>Hi Tom,
>>
>>See see Wolfgang's MRT article at
>>
http://onlinelibrary.wiley.com/doi/10.1002/jemt.20959/full>>
>>
http://www.becker-hickl.de/pdf/hpm-appnote03.pdf
>>(pdf page 6 - much larger area than an APD
>>results in somewhat higher photon counts ... so
>>much for simple QE curves! Example is from a
>>confocal microscope operate with 3 Airy Unit
>>pinhole - difference may be even bigger with MP
>>excitation and non-descanned detection).
>>
http://www.becker-hickl.de/pdf/dbhpm04.pdf>>
http://sales.hamamatsu.com/assets/pdf/catsandguides/p-dev_2007_TOTH0014E01.pdf
>>(pdf page 8, bottom half)
>>
>>If you have or are thinking of getting a Leica
>>confocal, multiphoton, and/or STED, ask your
>>Leica rep for info on the HyD detectors -
>>available internally on the SP5, or NDD for MP,
>>or on the X1 port (X1 usually uses APD's).
>>
>>Enjoy,
>>
>>George
>>
>>
>>
>>
>>On 1/21/2011 5:54 PM, Phillips, Thomas E. wrote:
>>>*****
>>>To join, leave or search the confocal microscopy listserv, go to:
>>>
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy>>>*****
>>>
>>>While searching the confocal archive about
>>>GaAsP PMTs, I came across Jim Pawley's
>>>authoritative discussion (appended below but
>>>note that I took the liberty of highlighting
>>>one sentence in red) of why the real world QE
>>>of these PMTs might not really be 40% but I
>>>was left wondering just how much better are
>>>they than the conventional PMTs on a Zeiss or
>>>Leica confocal? Jim says they are "much better
>>>than that of the more common S-20
>>>photocathode" . Is the ballpark sensitivity of
>>>a GaAsP unit about 2x higher? I would
>>>appreciate any insights or comments about the
>>>usefulness and limitations of these new
>>>detectors in core facilities. Tom
>>>
>>>Thomas E. Phillips, Ph.D
>>>Professor of Biological Sciences
>>>Director, Molecular Cytology Core
>>>2 Tucker Hall
>>>University of Missouri
>>>Columbia, MO 65211-7400
>>>573-882-4712 (office)
>>>573-882-0123 (fax)
>>>
[hidden email]<mailto:
[hidden email]>
>>>
>>>
http://www.biology.missouri.edu/faculty/phillips.html>>>
http://www.biotech.missouri.edu/mcc/>>>
>>>
>>>----- Original Message -----
>>>From: James Pawley<
[hidden email]<mailto:
[hidden email]>>
>>>Date: Wednesday, March 10, 2010 11:58 am
>>>Subject: Re: Zeiss or Olympus
>>>To:
>>>
[hidden email]<mailto:
[hidden email]>
>>>
>>>>Just to clarify, the 780 has a GaAsP (Gallium
>>>>Arsenite Phosphate) detector, not GaAs, the
>>>>difference in quantum efficiency can be seen
>>>>e.g. in the Webb multiphoton review (Nature
>>>>Biotechnology 2003, 21, 1369). The drawback
>>>>is that GaAsP QE drops dramatically for
>>>>wavelength> 700 nm, but they put a normal
>>>>PMTs as the two additional channels on the
>>>>780, to cover the extended range. By the way
>>>>GaAsP detectors are PMTs as well, it is just
>>>>a different material of the photocathode,
>>>>afterwards the photoelectrons are multiplied
>>>>in the same way. GaAsP detectors reach 40%
>>>>quantum efficiency which is about twice as
>>>>sensitive as a normal PMT. APDs have 60-70%
>>>>and a back-thinned CCD about 90%., so still a
>>>>lot of signal is thrown away, not to mention
>>>>the losses on the way to the detector.
>>>>
>>>
>>>>Andreas
>>>>
>>>
>>>>Indeed, the GaAs and GaAs phosphide QE curves
>>>>are very impressive. However, it is important
>>>>to remember what is actually measured to make
>>>>these curves. PMT curves refer to the
>>>>fraction of photons striking the photocathode
>>>>that produce photoelectrons (It is usually
>>>>measured by allowing a calibrated amount of
>>>>light to strike the photocathode and using a
>>>>nano-ammeter to sense the total photoelectron
>>>>current between the photocathode and all the
>>>>other electrodes in the PMT). However,
>>>>depending on the electrode geometry, 10-30%
>>>>of these photoelectrons don't actually hit
>>>>the first dynode (D1), and therefore do not
>>>>contribute to the PMT output.
>>>>
>>>
>>>>Of those photoelectrons that do hit D1, a
>>>>reasonable fraction fail to excite any
>>>>secondary electrons, and again, do not
>>>>contribute to the PMT output. There are many
>>>>reasons for this but one is just Poisson
>>>>statistics. If the average gain is say 4,
>>>>then about 8% of the collisions will result
>>>>in zero electrons being emitted. However,
>>>>this effect is again multiplied by
>>>>geometrical factors where SE produced in
>>>>different parts of D1 have better or worse
>>>>chances of actually striking D2 and producing
>>>>a SE. Signal loss in this way depends a lot
>>>>on the actual voltage between the
>>>>photocathode and D1: it will be less when the
>>>>voltage is higher. Unfortunately, few
>>>>confocals seem to have been set up in such
>>>>way that this is always true. On average
>>>>signal loss by failure to propagate after
>>>>collision with D1 will be an additional
>>>>20-40%.
>>>>
>>>
>>>>Finally, the same type of Poisson effects
>>>>that cause some signal to be lost entirely,
>>>>cause the amount by which the remainder is
>>>>amplified to be highly variable (10-90%).
>>>>This variation degrades the accuracy of the
>>>>output signal by introducing what is called
>>>>multiplicative noise. Because this extra
>>>>noise can only be "overcome" by counting more
>>>>photons, its presence effectively reduces the
>>>>effective QE of the device. In the best case,
>>>>this reduction is about 40% and in the worst
>>>>case (an exponential gain distribution,
>>>>approximated by some micro PMTs) 75% (i.e.,
>>>>the QE is reduced to 60% or 25% of what it
>>>>would have been if all photoelectrons were
>>>>equally amplified).
>>>>
>>>
>>>>As a result, while the peak effective QE of a
>>>>PMT with a GaAs or GaAsP photocathode is
>>>>indeed much better than that of the more
>>>>common S-20 photocathode, in terms of its
>>>>effectiveness in providing an output current
>>>>that is proportional to the input photon
>>>>signal, the QE is more in the range of 3 -10%
>>>>(depending on dynode geometry and
>>>>first-dynode voltage) than 40%. (The 60%
>>>>numbers are for APDs rather than for a GaAs
>>>>or GaAsP photocathode on a PMT.)
>>>>
>>>
>>>>The performance can be improved somewhat on
>>>>the few confocals that allow single-photon
>>>>counting as this procedure eliminates
>>>>multiplicative noise. (see below about the
>>>>limitations imposed by photon counting)
>>>>
>>>
>>>>This tedious recital is I hope justified by
>>>>noting that, at least when EG&G was the major
>>>>APD supplier, APD performance was not
>>>>specified in terms of QE but as Photon
>>>>Detection Efficiency (PDE). Although APDs can
>>>>be operated in a low gain, proportional mode,
>>>>their PDE under these conditions is very low
>>>>(because APD multiplicative noise is very
>>>>high and at low (non-avalanche breakdown)
>>>>gain, by far the most likely gain of the
>>>>initial photoelectron is zero).
>>>>
>>>
>>>>Therefore, high PDE (or high QE) AOD units
>>>>tend to operate at high bias (high, avalanche
>>>>gain) and this requires circuitry to quench
>>>>the avalanche breakdown and count the
>>>>single-photon pulses. Modern units contain
>>>>both the sensor itself and the pulse counting
>>>>and avalanche quenching circuits needed for
>>>>counting the single-photon pulses. In other
>>>>words (assuming that Hamamatsu follows the
>>>>EG&G precedent), their QE figures for
>>>>single-photon counting units already include
>>>>any losses for non-propagation or
>>>>multiplicative noise. Therefore, a quoted PID
>>>>of 60% really does mean that 60% of the
>>>>photons (of the specified wavelength) that
>>>>strike the center of the active surface will
>>>>be accurately counted.
>>>>
>>>
>>>>This is about 4-10x better than the
>>>>performance of a similar GaAs or GaAsP
>>>>photocathode on a PMT.
>>>>
>>>
>>>>This good news is tempered by the fact that,
>>>>because of the high capacitance of the AOD
>>>>itself, it is hard to count much faster than,
>>>>say 30MHz. As 30MHz comes out to an absolute
>>>>maximum of 60 counts during a 2 µs pixel,
>>>>this means that at least 50% of your counts
>>>>will be lost due to pulse pileup when 30
>>>>counts arrive per pixel and 10% will be lost
>>>>at only 6 counts/pixel. In other words one
>>>>has to be very careful to adjust the
>>>>excitation intensity so as not to "clip" the
>>>>brightness of those parts of the image that
>>>>contain a lot of fluorophor.
>>>>
>>>
>>>>Lots more on this in The Handbook,
>>>>
>>>
>>>>Cheers,
>>>>
>>>
>>>>Jim Pawley
>>>> **********************************************
>>>>Prof. James B. Pawley,
>>>>Ph. 608-263-3147
>>>>Room 223, Zoology Research Building,
>>>>FAX 608-265-5315
>>>>1117 Johnson Ave., Madison, WI, 53706
>>>>
[hidden email]<mailto:
[hidden email]>
>>>>3D Microscopy of Living Cells Course, June
>>>>12-24, 2010, UBC, Vancouver Canada
>>>>Info:
http://www.3dcourse.ubc.ca/
>>>>Applications due by March 15, 2010
>>>> "If it ain't diffraction, it must be statistics." Anon.
>>>>
>>>
>>>
>>>
>>>
>>>Thomas E. Phillips, Ph.D
>>>Professor of Biological Sciences
>>>Director, Molecular Cytology Core
>>>2 Tucker Hall
>>>University of Missouri
>>>Columbia, MO 65211-7400
>>>573-882-4712 (office)
>>>573-882-0123 (fax)
>>>
[hidden email]<mailto:
[hidden email]>
>>>
>>>
http://www.biology.missouri.edu/faculty/phillips.html>>>
http://www.biotech.missouri.edu/mcc/--
***************************************************************************
Prof. James B. Pawley,
Ph. 608-238-3953
21. N. Prospect Ave. Madison, WI 53726 USA
[hidden email]
3D Microscopy of Living Cells Course, June 11-23, 2011, UBC, Vancouver Canada
Info:
http://www.3dcourse.ubc.ca/ Applications due by March 15, 2011
"If it ain't diffraction, it must be statistics." Anon.