Leica Hyd vs. Zeiss GaAsP

[Ralf Zenke]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Dear Listers,

I am trying to find out technical properties of Zeiss GaAsP detectors
versus Leica Hybrid detectors. All I have is quantum efficiency, which
is around 45% for both devices. But what about dynamic range (or signal
to noise ratio)? Do GaAsP PMTs have the same SNR as conventional PMTs?
From what I know, Leica's hybrid detectors come from Hamamatsu, but I
can't find any in their product range. Does anybody know a product
number?

Thanks a lot for any kind of help!

Ralf


--
Ralf Zenke

Max Planck Institute of Biochemistry
Core Facility
Am Klopferspitz 18
DE-82152 Martinsried near Munich
GERMANY
Phone: (+49) (89) 8578 3798
Fax: (+49) (89) 8578 2847
www.biochem.mpg.de

[Csúcs Gábor]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Dear Ralf,


The ALMF people from the EMBL Heidelberg had a nice poster on this subject
at the latest ELMI meeting in Greece, so the may give you a more detailed
answer. As far as I recall, their take home message was that the Zeiss
GaAsP and the Leica HyDe perform virtually identically in terms of dynamic
range and S/N.

Greetings   Gabor


On 10/13/11 9:05 AM, "Ralf Zenke" <[hidden email]> wrote:

>*****
>To join, leave or search the confocal microscopy listserv, go to:
>http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
>*****
>
>Dear Listers,
>
>I am trying to find out technical properties of Zeiss GaAsP detectors
>versus Leica Hybrid detectors. All I have is quantum efficiency, which
>is around 45% for both devices. But what about dynamic range (or signal
>to noise ratio)? Do GaAsP PMTs have the same SNR as conventional PMTs?
>From what I know, Leica's hybrid detectors come from Hamamatsu, but I
>can't find any in their product range. Does anybody know a product
>number?
>
>Thanks a lot for any kind of help!
>
>Ralf
>
>
>--
>Ralf Zenke
>
>Max Planck Institute of Biochemistry
>Core Facility
>Am Klopferspitz 18
>DE-82152 Martinsried near Munich
>GERMANY
>Phone: (+49) (89) 8578 3798
>Fax: (+49) (89) 8578 2847
>www.biochem.mpg.de

[Sudipta Maiti]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Check the Hamamatsu H7422P series.
Sudipta

On Thu, 13 Oct 2011 07:11:49 +0000, Csúcs  Gábor wrote

> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> *****
>
> Dear Ralf,
>
> The ALMF people from the EMBL Heidelberg had a nice poster on this subject
> at the latest ELMI meeting in Greece, so the may give you a more detailed
> answer. As far as I recall, their take home message was that the
> Zeiss GaAsP and the Leica HyDe perform virtually identically in
> terms of dynamic range and S/N.
>
> Greetings   Gabor
>
> On 10/13/11 9:05 AM, "Ralf Zenke" <[hidden email]> wrote:
>
> >*****
> >To join, leave or search the confocal microscopy listserv, go to:
> >http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> >*****
> >
> >Dear Listers,
> >
> >I am trying to find out technical properties of Zeiss GaAsP detectors
> >versus Leica Hybrid detectors. All I have is quantum efficiency, which
> >is around 45% for both devices. But what about dynamic range (or signal
> >to noise ratio)? Do GaAsP PMTs have the same SNR as conventional PMTs?
> >From what I know, Leica's hybrid detectors come from Hamamatsu, but I
> >can't find any in their product range. Does anybody know a product
> >number?
> >
> >Thanks a lot for any kind of help!
> >
> >Ralf
> >
> >
> >--
> >Ralf Zenke
> >
> >Max Planck Institute of Biochemistry
> >Core Facility
> >Am Klopferspitz 18
> >DE-82152 Martinsried near Munich
> >GERMANY
> >Phone: (+49) (89) 8578 3798
> >Fax: (+49) (89) 8578 2847
> >www.biochem.mpg.de

Dr. Sudipta Maiti
Dept. of Chemical Sciences
Tata Institute of Fundamental Research
Homi Bhabha Raod, Colaba, Mumbai 400005
Ph. 91-22-2278-2716 / 2539
Fax: [phone]
alternate e-mail: [hidden email]
url: biophotonics.weebly.com

[James Pawley]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Hi all,

Great to see that we are beginning to get some
more advanced photodetectors in Live-Cell
Microscopy!

Although I don't own any of these marvellous new
instruments, I may be able to provide a little
background info (More in Chapter 12 of the
Handbook).

Hybrid PMTs have the high gain of normal PMTs but
the gain of the first stage (from the GaAsP
photocathode to the APD, is so high (say 1,000x
vs. 3-15x) that the multiplicative noise (cause
by the statistical limitations on the charge
multiplication process of the ensemble producing
a wide range of heights in single photoelectron
pulses.), is much lower (maybe adding 3% to the
intrinsic Poisson noise set by the number of
photons detected) than a normal PMT (that maybe
adds 30% to the Poisson noise) and VERY much
lower than that of the early mini-PMT arrays
(that add maybe 100% additional uncertainty.).

i.e., High multiplicative noise can double the
Poisson noise of the original signal! This has
the same effect on signal accuracy as reducing
the effective QE by 4x (i.e., a hybrid with the
same cathode might require only 25% as many
photons striking its surface to produce data with
a given uncertainty as a would be required by a
single channel of an analog mini-PMT).

Of course, if you add photon-counting circuitry
to any PMT, you can eliminate the multiplicative
noise (all single-photon pulses now will count
the same). The problem then becomes, "How fast
can you count single photon pulses?" The specs
for the Hamamatsu GaAsP photon-counting PMT

http://www.google.com/search?client=safari&rls=en&q=Hamamatsu+H7422P&ie=UTF-8&oe=UTF-8

lists a pulse-pair resolution of 40 ns. Pretty
fast. But even if you could arrange for your
(otherwise random) photoelectrons to be produced
every 25 ns, you could still only count 40 in 1
µs. And as they actually appear randomly, you
would loose 10% of them if the average count rate
was only 4/µs (and lose 20% at 8 counts/µs.)

A 1µs pixel time implies a normal scan time of <4
frames/second for 512x512. At video rates, pixel
times are closer to 0.1µs (0.4
counts/counting-channel before you begin to lose
10% of your counts.)

Clearly this is a problem. Solutions include
using many channels and somehow arranging the
optics so that all the photons hit one of, say, 4
photon-counting channels (and no photons miss
hitting some detector!). This would increase the
1µs pixel loss rate to 10% at 16 counts (in the
brightest pixel). However, designing these optics
requires considerable skill and so far, I haven't
been able to get any design details.

Alternatively, one could use 4 detectors set up
so that the optical path uses a series of 3, 50%
reflecting mirrors so arranged that the first
detector gets 50% of the light, the second gets
25% of the light and the third 12% (so it would
not be seriously 10% saturated until >8
counts/µs), 25% on the third (saturating at 16
counts/µs) and 12.5% on the fourth (saturating at
32 counts/µs). One would then use fast logic to
choose the signal from the first unsaturated
detector and record that.  So we would have a 8x
greater dynamic range but at the considerable
cost of throwing away more than 12% of the
photons. In addition, the signal from the
brightest pixels would have the same percentage
uncertainty as the fainter ones (normally the
percentage uncertainty decreases with the sqrt of
the signal  strength). So this is probably NOT a
perfect idea.

Finally, one could disperse the signal light with
a prism or diffraction grating and again use many
PMTs                      Normally we do this for
spectral imaging but the same system that can be
used to do photon-counting without wasting any
photons on ND filters but splitting the signal in
a spectral peak between many detectors and so
reducing the chance of pulse-pileup on any of
them. (One would sum the photon counts from all
detectors under the peak.)

The above discussion can also be applied to APDs
which have high QE but massive multiplicative
noise.  Pulse counting is almost always used
because high-gain APDs must be pulsed off to
quench each single-photoelectron avalanche (and
low-gain APD have VERY low effective QE because
most photoelectrons fail to propagate at all and
are lost) so you have the same pile-up problems
and possible solutions. (or does anyone have one
I haven't heard of?)

Summary: Hybrid PMTs have such low multiplicative
noise that you don't need pulse counting, so they
don't saturate. However, if you reduce the first
stage gain because the signal is large, then
multiplicative noise may reappear. They are not
cheap and so damaging one with too much input
light may be a memorable experience.

Multi-Mini-PMTs can do about as well  as long as they use pulse-counting.

In all cases GaAsP photocathodes are wonderful in
terms of QE but must be cooled to prevent high
dark counts.

If you are trying to make comparisons, I suggest
that rather than looking at test specimens, you
use light from the transmitted light source in
your microscope. You can use your confocal
emission filters to select a particular
wavelength band and ND filters to adjust
intensity. Set up Kohler illumination and turn
off the laser and scan the "bright" field. As you
will only be looking at one-pinhole's worth of
light at any time, it is easy to adjust matters
(pinhole size, NS filters, lamp power), to get a
few photons/pixel. And, as the total signal
coming through the plane of focus is about
1,000,000 times greater than that going through
the pinhole, you can measure it fairly easily (Be
sure to set the field diaphragm at a repeatable
setting.). This is a number of variables that one
can almost cope with.

Put a specimen in and you have another 30 or so.

Good luck,

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 10-22, 2012, UBC, Vancouver Canada
Info: http://www.3dcourse.ubc.ca/        Applications accepted after 11/15/12
               "If it ain't diffraction, it must be statistics." Anon.

>*****
>To join, leave or search the confocal microscopy listserv, go to:
>http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
>*****
>
>Check the Hamamatsu H7422P series.
>Sudipta
>
>On Thu, 13 Oct 2011 07:11:49 +0000, Csúcs  Gábor wrote
>>  *****
>>  To join, leave or search the confocal microscopy listserv, go to:
>>  http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
>>  *****
>>
>>  Dear Ralf,
>>
>>  The ALMF people from the EMBL Heidelberg had a nice poster on this subject
>>  at the latest ELMI meeting in Greece, so the may give you a more detailed
>>  answer. As far as I recall, their take home message was that the
>>  Zeiss GaAsP and the Leica HyDe perform virtually identically in
>>  terms of dynamic range and S/N.
>>
>>  Greetings   Gabor
>>
>>  On 10/13/11 9:05 AM, "Ralf Zenke" <[hidden email]> wrote:
>>
>>  >*****
>>  >To join, leave or search the confocal microscopy listserv, go to:
>>  >http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
>>  >*****
>>  >
>>  >Dear Listers,
>>  >
>>  >I am trying to find out technical properties of Zeiss GaAsP detectors
>>  >versus Leica Hybrid detectors. All I have is quantum efficiency, which
>>  >is around 45% for both devices. But what about dynamic range (or signal
>>  >to noise ratio)? Do GaAsP PMTs have the same SNR as conventional PMTs?
>>  >From what I know, Leica's hybrid detectors come from Hamamatsu, but I
>>  >can't find any in their product range. Does anybody know a product
>>  >number?
>>  >
>>  >Thanks a lot for any kind of help!
>>  >
>>  >Ralf
>>  >
>>  >
>>  >--
>>  >Ralf Zenke
>>  >
>>  >Max Planck Institute of Biochemistry
>>  >Core Facility
>>  >Am Klopferspitz 18
>>  >DE-82152 Martinsried near Munich
>>  >GERMANY
>>  >Phone: (+49) (89) 8578 3798
>>  >Fax: (+49) (89) 8578 2847
>>  >www.biochem.mpg.de
>
>
>Dr. Sudipta Maiti
>Dept. of Chemical Sciences
>Tata Institute of Fundamental Research
>Homi Bhabha Raod, Colaba, Mumbai 400005
>Ph. 91-22-2278-2716 / 2539
>Fax: [phone]
>alternate e-mail: [hidden email]
>url: biophotonics.weebly.com

--

[George McNamara]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Hi Jim,

Thanks for your post. Counting a couple of items:

1.  splitting photons onto multiple APDs - Stefan Hell's lab uses equal
splitting;
   Westphal 2009 New J Physics (available on S. Hell lab website):
To enable high count rates, the fluorescence was divided with a 1:4
fiber coupler onto four avalanche photo diodes (APDs, SPCM-AQR13, Perkin
Elmer, Fremont, CA) featuring >65% quantum efficiency of detection. The
resulting bright images enabled high frame rates. Photon counting and
data-preprocessing were accomplished with a custom field programmable
gate array board.

2. The Hamamatsu R10467 series hybrid photo detectors are on PDF page 8 of
http://sales.hamamatsu.com/assets/pdf/catsandguides/p-dev_2007_TOTH0014E01.pdf
R10467-06 has a bialkali photocathode, peak QE 0.32, spectral response
160-650 nm, rise time 400 ps, transit time spread 110 ps.
R10467-40 has a GaAsP photocathode, peak QE 0.45, spectral response
300-750 nm, rise time 400 ps, transit time spread 120 ps.

Wolfgang Becker, in his 2011 Microsc Res Tech article and in
http://www.becker-hickl.de/pdf/hpm-appnote03.pdf   thoroughly evaluates
his company's HPM-100-40 with an R10467-40 (GaAsP) tube. His electronics
probably somewhat different from the Leica HyD product. most
interesting: low afterpulsing. Quoting the B&H appnote (page 4);

    The low afterpulsing results in a significantly improved dynamic
    range of fluorescence decay measurements. An example is shown in
    Fig. 6. It shows the fluorescence decay of fluorescein recorded at a
    laser repetition rate of 20 MHz. The signal was detected by a
    HPM-100-40 (left) and a H5773-1 photosensor module (right). Both
    detectors have approximately the same dark count rates.
    For the HPM-100, the dark count rate is the only source of
    background. Because the dark count rate is only a few 100 counts per
    second an extraordinarily high dynamic range is obtained. For the
    H5773-1 the background is dominated by afterpulsing. The background
    is substantially higher, and the dynamic range is far smaller than
    for the HPM-100.


3. As Jim notes, the Zeiss GaAsP PMT has 40 ns (what I will call) dead
time. So what? Jim's book refers to highest resolution confocal images
having max ~10 photons/pixel (maybe 20 now with higher QE detector?). If
you hit saturation of Leica HyD or B&H Hybrid or Zeiss GaAsP PMT, turn
down the laser power.

4. Having had a Leica CW-STED in my office for three weeks, I concluded
that the killer feature of the HyD is that I could choose photon
counting mode. The B&H hybrid is essentially always in photon counting
mode - B&H uses its TCSPC board (time correlated single photon counting)
to get the data to the computer (and is already FLIM). Zeiss offers
photon counting capability on its GaAsP. The photon counting will be
even better when FLIM is routinely available on these instruments. I
never used them, but (at least some) old Bio-Rad confocals had photon
counting mode. Since their demise, we've been stuck with arbitrary
intensity levels, further complicated by some trainers instructing users
to set the detector offset to black out the "background".

Upshot: photon counting with the new GaAsP hybrids and PMTs is going to
enable confocal microscopes to be more quantitative than they have been
for the past decade, while we transition subcellular imaging to the
nanoscopes era.

Sincerely,

George
p.s. of course variations in laser power, room temperature, wrong
coverglass thickness, poor specimen preparation, and of course mis-use,
will consider to make quantitative microscopy more wishful thinking than
a standard of care.


On 10/13/2011 9:18 PM, James Pawley wrote:

> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> *****
>
> Hi all,
>
> Great to see that we are beginning to get some more advanced
> photodetectors in Live-Cell Microscopy!
>
> Although I don't own any of these marvellous new instruments, I may be
> able to provide a little background info (More in Chapter 12 of the
> Handbook).
>
> Hybrid PMTs have the high gain of normal PMTs but the gain of the
> first stage (from the GaAsP photocathode to the APD, is so high (say
> 1,000x vs. 3-15x) that the multiplicative noise (cause by the
> statistical limitations on the charge multiplication process of the
> ensemble producing a wide range of heights in single photoelectron
> pulses.), is much lower (maybe adding 3% to the intrinsic Poisson
> noise set by the number of photons detected) than a normal PMT (that
> maybe adds 30% to the Poisson noise) and VERY much lower than that of
> the early mini-PMT arrays (that add maybe 100% additional uncertainty.).
>
> i.e., High multiplicative noise can double the Poisson noise of the
> original signal! This has the same effect on signal accuracy as
> reducing the effective QE by 4x (i.e., a hybrid with the same cathode
> might require only 25% as many photons striking its surface to produce
> data with a given uncertainty as a would be required by a single
> channel of an analog mini-PMT).
>
> Of course, if you add photon-counting circuitry to any PMT, you can
> eliminate the multiplicative noise (all single-photon pulses now will
> count the same). The problem then becomes, "How fast can you count
> single photon pulses?" The specs for the Hamamatsu GaAsP
> photon-counting PMT
>
> http://www.google.com/search?client=safari&rls=en&q=Hamamatsu+H7422P&ie=UTF-8&oe=UTF-8 
>
>
> lists a pulse-pair resolution of 40 ns. Pretty fast. But even if you
> could arrange for your (otherwise random) photoelectrons to be
> produced every 25 ns, you could still only count 40 in 1 µs. And as
> they actually appear randomly, you would loose 10% of them if the
> average count rate was only 4/µs (and lose 20% at 8 counts/µs.)
>
> A 1µs pixel time implies a normal scan time of <4 frames/second for
> 512x512. At video rates, pixel times are closer to 0.1µs (0.4
> counts/counting-channel before you begin to lose 10% of your counts.)
>
> Clearly this is a problem. Solutions include using many channels and
> somehow arranging the optics so that all the photons hit one of, say,
> 4 photon-counting channels (and no photons miss hitting some
> detector!). This would increase the 1µs pixel loss rate to 10% at 16
> counts (in the brightest pixel). However, designing these optics
> requires considerable skill and so far, I haven't been able to get any
> design details.
>
> Alternatively, one could use 4 detectors set up so that the optical
> path uses a series of 3, 50% reflecting mirrors so arranged that the
> first detector gets 50% of the light, the second gets 25% of the light
> and the third 12% (so it would not be seriously 10% saturated until >8
> counts/µs), 25% on the third (saturating at 16 counts/µs) and 12.5% on
> the fourth (saturating at 32 counts/µs). One would then use fast logic
> to choose the signal from the first unsaturated detector and record
> that.  So we would have a 8x greater dynamic range but at the
> considerable cost of throwing away more than 12% of the photons. In
> addition, the signal from the brightest pixels would have the same
> percentage uncertainty as the fainter ones (normally the percentage
> uncertainty decreases with the sqrt of the signal  strength). So this
> is probably NOT a perfect idea.
>
> Finally, one could disperse the signal light with a prism or
> diffraction grating and again use many PMTs                      
> Normally we do this for spectral imaging but the same system that can
> be used to do photon-counting without wasting any photons on ND
> filters but splitting the signal in a spectral peak between many
> detectors and so reducing the chance of pulse-pileup on any of them.
> (One would sum the photon counts from all detectors under the peak.)
>
> The above discussion can also be applied to APDs which have high QE
> but massive multiplicative noise.  Pulse counting is almost always
> used because high-gain APDs must be pulsed off to quench each
> single-photoelectron avalanche (and low-gain APD have VERY low
> effective QE because most photoelectrons fail to propagate at all and
> are lost) so you have the same pile-up problems and possible
> solutions. (or does anyone have one I haven't heard of?)
>
> Summary: Hybrid PMTs have such low multiplicative noise that you don't
> need pulse counting, so they don't saturate. However, if you reduce
> the first stage gain because the signal is large, then multiplicative
> noise may reappear. They are not cheap and so damaging one with too
> much input light may be a memorable experience.
>
> Multi-Mini-PMTs can do about as well  as long as they use pulse-counting.
>
> In all cases GaAsP photocathodes are wonderful in terms of QE but must
> be cooled to prevent high dark counts.
>
> If you are trying to make comparisons, I suggest that rather than
> looking at test specimens, you use light from the transmitted light
> source in your microscope. You can use your confocal emission filters
> to select a particular wavelength band and ND filters to adjust
> intensity. Set up Kohler illumination and turn off the laser and scan
> the "bright" field. As you will only be looking at one-pinhole's worth
> of light at any time, it is easy to adjust matters (pinhole size, NS
> filters, lamp power), to get a few photons/pixel. And, as the total
> signal coming through the plane of focus is about 1,000,000 times
> greater than that going through the pinhole, you can measure it fairly
> easily (Be sure to set the field diaphragm at a repeatable setting.).
> This is a number of variables that one can almost cope with.
>
> Put a specimen in and you have another 30 or so.
>
> Good luck,
>
> 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 10-22, 2012, UBC, Vancouver
> Canada
> Info: http://www.3dcourse.ubc.ca/    Applications accepted after 11/15/12
>            "If it ain't diffraction, it must be statistics." Anon.
>
>> *****
>> To join, leave or search the confocal microscopy listserv, go to:
>> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
>> *****
>>
>> Check the Hamamatsu H7422P series.
>> Sudipta
>>
>> On Thu, 13 Oct 2011 07:11:49 +0000, Csúcs  Gábor wrote
>>>  *****
>>>  To join, leave or search the confocal microscopy listserv, go to:
>>>  http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
>>>  *****
>>>
>>>  Dear Ralf,
>>>
>>>  The ALMF people from the EMBL Heidelberg had a nice poster on this
>>> subject
>>>  at the latest ELMI meeting in Greece, so the may give you a more
>>> detailed
>>>  answer. As far as I recall, their take home message was that the
>>>  Zeiss GaAsP and the Leica HyDe perform virtually identically in
>>>  terms of dynamic range and S/N.
>>>
>>>  Greetings   Gabor
>>>
>>>  On 10/13/11 9:05 AM, "Ralf Zenke" <[hidden email]> wrote:
>>>
>>> >*****
>>> >To join, leave or search the confocal microscopy listserv, go to:
>>> >http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
>>> >*****
>>> >
>>> >Dear Listers,
>>> >
>>> >I am trying to find out technical properties of Zeiss GaAsP detectors
>>> >versus Leica Hybrid detectors. All I have is quantum efficiency, which
>>> >is around 45% for both devices. But what about dynamic range (or
>>> signal
>>> >to noise ratio)? Do GaAsP PMTs have the same SNR as conventional PMTs?
>>> >From what I know, Leica's hybrid detectors come from Hamamatsu, but I
>>> >can't find any in their product range. Does anybody know a product
>>> >number?
>>> >
>>> >Thanks a lot for any kind of help!
>>> >
>>> >Ralf
>>> >
>>> >
>>> >--
>>> >Ralf Zenke
>>> >
>>> >Max Planck Institute of Biochemistry
>>> >Core Facility
>>> >Am Klopferspitz 18
>>> >DE-82152 Martinsried near Munich
>>> >GERMANY
>>> >Phone: (+49) (89) 8578 3798
>>> >Fax: (+49) (89) 8578 2847
>>> >www.biochem.mpg.de
>>
>>
>> Dr. Sudipta Maiti
>> Dept. of Chemical Sciences
>> Tata Institute of Fundamental Research
>> Homi Bhabha Raod, Colaba, Mumbai 400005
>> Ph. 91-22-2278-2716 / 2539
>> Fax: [phone]
>> alternate e-mail: [hidden email]
>> url: biophotonics.weebly.com
>
>

--


George McNamara, PhD
Analytical Imaging Core Facility
University of Miami

[Wolfgang Staroske]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Hi all,

we have a Zeiss LSM780 and had the chance to test the Leica HyD
Detector. In the moment we compared only the HyD Detector to Leicas
standard PMTs (to justify an upgrade) and measured an increase in signal
to noise by a factor of 1.6 in the green region (Alexa488) and 2.2-2.5
in the far red region (Alexa633). So a clear yes for upgrading every SP5
with one to two HyDs. We didn't perform test on the LSM780 yet, because
when we bought it, there was no alternative available, and the rest is
lack of time.

I would like to comment on the software implementation of the photon
counting mode of systems, where Zeiss made a much better job than Leica
in my eyes.

In photon counting mode the saturation of a pixel depends on scan speed,
because the photon counting detector has a maximal count rate up to
which the detector can be assumed to be linear (not loosing to much
photons, due to dead time). Even when the Zeiss GaAsP has a dead time of
40ns corresponding to a maximal count rate of 25MHz, that means the
linear regime is much lower, because the probability that a photon hits
the detector in the dead time window, where it is not detected
increases, when you come closer to the 25MHz limit.

Zeiss has chosen a linear limit of 2MHz, corresponding to 50 photons at
scan speed 4 (25.6µs pixel dwell time (pdt)). This maximum value is
displayed in the range indicator and in the histogram, and all brighter
pixels are marked as saturated (red pixels in the range indicator mode),
even when you can read the detected photons for example as 80 or 100 in
the image. This saturation value changes according to the scan speed (25
for scan speed 5 -12.5µs pdt, 100 for scan speed 3-51.2µs pdt....). This
means the software makes sure, that the user drives the GaAsP in photon
counting mode always in the linear regime.

Leica has nothing like that. In the photon counting mode the saturation
value depends only on the chosen bit depth. There is no hint and we
weren't even able to get any number from Leica what the linear regime of
the HyD detector is. In our demo data we measured a maximum of 2250
photons/pixel with 92Hz line rate, corresponding to a dwell time of
28.3µs (256px/line) . This is a saturation count rate of 80MHz. The
linear regime seem to be up to 200Photons/pixel at same scan rate,
meaning around 7Mhz. It is known to us that the HyD can measure higher
count rates (maybe can even discriminate between multiple photon
events). So by the software it is very easy to use the HyD detector in
the non-linear regime, without noticing.

Additionally the Zeiss GaAsP detector can be put in a real PMT mode,
where you can adjust a real HV voltage gain, to detect also bright
signals with the spectral (32 channel) GaAsP detector. In contrast the
Leica HyD always runs in photon counting mode, even in the mode, which
they call normal and the gain there is a pure digital gain. This is as
we understood the Leica people and this was also our impression, the
Leica HyD always runs in photon counting mode, even in the mode, which
they call normal and the gain there is a pure digital gain. The Leica
people stated that the count rate of the HyD is higher compared to the
Zeiss GaAsP (80MHz to 25Mhz).

The rise and transit times stated by George, doesn't tell much about the
dead time of an APD (what is the core of the HyD) because the dead time
is mainly influenced by the active quenching of the electron avalanche
inside the APD, which is in principle shutting off the HV and on again
by a fast electronic circuit. The rise and transit times are good to
estimate the time resolution of the detector but not the saturation limit.

I have holiday next week - so I will reply to any comments the week after.

Bye Wolfgang

Dr. Wolfgang Staroske

Single Molecule Specialist
Light Microscopy Facility
EuroBioimaging PoC-Site

Technische Universität Dresden
Biotechnology Center
Tatzberg 47/49
01307 Dresden, Germany

Tel.: [phone]
Fax.: [phone]
E-Mail: [hidden email]
Webpage: www.biotec.tu-dresden.de

Am 20:59, schrieb George McNamara:

> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> *****
>
> Hi Jim,
>
> Thanks for your post. Counting a couple of items:
>
> 1.    splitting photons onto multiple APDs - Stefan Hell's lab uses equal
> splitting;
>     Westphal 2009 New J Physics (available on S. Hell lab website):
> To enable high count rates, the fluorescence was divided with a 1:4
> fiber coupler onto four avalanche photo diodes (APDs, SPCM-AQR13, Perkin
> Elmer, Fremont, CA) featuring &gt;65% quantum efficiency of detection.
> The
> resulting bright images enabled high frame rates. Photon counting and
> data-preprocessing were accomplished with a custom field programmable
> gate array board.
>
> 2. The Hamamatsu R10467 series hybrid photo detectors are on PDF page 8 of
> http://sales.hamamatsu.com/assets/pdf/catsandguides/p-dev_2007_TOTH0014E01.pdf
> R10467-06 has a bialkali photocathode, peak QE 0.32, spectral response
> 160-650 nm, rise time 400 ps, transit time spread 110 ps.
> R10467-40 has a GaAsP photocathode, peak QE 0.45, spectral response
> 300-750 nm, rise time 400 ps, transit time spread 120 ps.
>
> Wolfgang Becker, in his 2011 Microsc Res Tech article and in
> http://www.becker-hickl.de/pdf/hpm-appnote03.pdf       thoroughly
> evaluates
> his company's HPM-100-40 with an R10467-40 (GaAsP) tube. His electronics
> probably somewhat different from the Leica HyD product. most
> interesting: low afterpulsing. Quoting the B&amp;H appnote (page 4);
>
>        The low afterpulsing results in a significantly improved dynamic
>        range of fluorescence decay measurements. An example is shown in
>        Fig. 6. It shows the fluorescence decay of fluorescein recorded
> at a
>        laser repetition rate of 20 MHz. The signal was detected by a
>        HPM-100-40 (left) and a H5773-1 photosensor module (right). Both
>        detectors have approximately the same dark count rates.
>        For the HPM-100, the dark count rate is the only source of
>        background. Because the dark count rate is only a few 100
> counts per
>        second an extraordinarily high dynamic range is obtained. For the
>        H5773-1 the background is dominated by afterpulsing. The background
>        is substantially higher, and the dynamic range is far smaller than
>        for the HPM-100.
>
>
> 3. As Jim notes, the Zeiss GaAsP PMT has 40 ns (what I will call) dead
> time. So what? Jim's book refers to highest resolution confocal images
> having max ~10 photons/pixel (maybe 20 now with higher QE detector?). If
> you hit saturation of Leica HyD or B&amp;H Hybrid or Zeiss GaAsP PMT,
> turn
> down the laser power.
>
> 4. Having had a Leica CW-STED in my office for three weeks, I concluded
> that the killer feature of the HyD is that I could choose photon
> counting mode. The B&amp;H hybrid is essentially always in photon
> counting
> mode - B&amp;H uses its TCSPC board (time correlated single photon
> counting)
> to get the data to the computer (and is already FLIM). Zeiss offers
> photon counting capability on its GaAsP. The photon counting will be
> even better when FLIM is routinely available on these instruments. I
> never used them, but (at least some) old Bio-Rad confocals had photon
> counting mode. Since their demise, we've been stuck with arbitrary
> intensity levels, further complicated by some trainers instructing users
> to set the detector offset to black out the "background".
>
> Upshot: photon counting with the new GaAsP hybrids and PMTs is going to
> enable confocal microscopes to be more quantitative than they have been
> for the past decade, while we transition subcellular imaging to the
> nanoscopes era.
>
> Sincerely,
>
> George
> p.s. of course variations in laser power, room temperature, wrong
> coverglass thickness, poor specimen preparation, and of course mis-use,
> will consider to make quantitative microscopy more wishful thinking than
> a standard of care.
>
>
> On 10/13/2011 9:18 PM, James Pawley wrote:
> <blockquote type=cite>*****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> *****
>
> Hi all,
>
> Great to see that we are beginning to get some more advanced
> photodetectors in Live-Cell Microscopy!
>
> Although I don't own any of these marvellous new instruments, I may be
> able to provide a little background info (More in Chapter 12 of the
> Handbook).
>
> Hybrid PMTs have the high gain of normal PMTs but the gain of the
> first stage (from the GaAsP photocathode to the APD, is so high (say
> 1,000x vs. 3-15x) that the multiplicative noise (cause by the
> statistical limitations on the charge multiplication process of the
> ensemble producing a wide range of heights in single photoelectron
> pulses.), is much lower (maybe adding 3% to the intrinsic Poisson
> noise set by the number of photons detected) than a normal PMT (that
> maybe adds 30% to the Poisson noise) and VERY much lower than that of
> the early mini-PMT arrays (that add maybe 100% additional uncertainty.).
>
> i.e., High multiplicative noise can double the Poisson noise of the
> original signal! This has the same effect on signal accuracy as
> reducing the effective QE by 4x (i.e., a hybrid with the same cathode
> might require only 25% as many photons striking its surface to produce
> data with a given uncertainty as a would be required by a single
> channel of an analog mini-PMT).
>
> Of course, if you add photon-counting circuitry to any PMT, you can
> eliminate the multiplicative noise (all single-photon pulses now will
> count the same). The problem then becomes, "How fast can you count
> single photon pulses?" The specs for the Hamamatsu GaAsP
> photon-counting PMT
>
> http://www.google.com/search?client=safari&amp;rls=en&amp;q=Hamamatsu+H7422P&amp;ie=UTF-8&amp;oe=UTF-8 
>
>
>
> lists a pulse-pair resolution of 40 ns. Pretty fast. But even if you
> could arrange for your (otherwise random) photoelectrons to be
> produced every 25 ns, you could still only count 40 in 1 µs. And as
> they actually appear randomly, you would loose 10% of them if the
> average count rate was only 4/µs (and lose 20% at 8 counts/µs.)
>
> A 1µs pixel time implies a normal scan time of &lt;4 frames/second for
> 512x512. At video rates, pixel times are closer to 0.1µs (0.4
> counts/counting-channel before you begin to lose 10% of your counts.)
>
> Clearly this is a problem. Solutions include using many channels and
> somehow arranging the optics so that all the photons hit one of, say,
> 4 photon-counting channels (and no photons miss hitting some
> detector!). This would increase the 1µs pixel loss rate to 10% at 16
> counts (in the brightest pixel). However, designing these optics
> requires considerable skill and so far, I haven't been able to get any
> design details.
>
> Alternatively, one could use 4 detectors set up so that the optical
> path uses a series of 3, 50% reflecting mirrors so arranged that the
> first detector gets 50% of the light, the second gets 25% of the light
> and the third 12% (so it would not be seriously 10% saturated until &gt;8
> counts/µs), 25% on the third (saturating at 16 counts/µs) and 12.5% on
> the fourth (saturating at 32 counts/µs). One would then use fast logic
> to choose the signal from the first unsaturated detector and record
> that.    So we would have a 8x greater dynamic range but at the
> considerable cost of throwing away more than 12% of the photons. In
> addition, the signal from the brightest pixels would have the same
> percentage uncertainty as the fainter ones (normally the percentage
> uncertainty decreases with the sqrt of the signal    strength). So this
> is probably NOT a perfect idea.
>
> Finally, one could disperse the signal light with a prism or
> diffraction grating and again use many PMTs
> Normally we do this for spectral imaging but the same system that can
> be used to do photon-counting without wasting any photons on ND
> filters but splitting the signal in a spectral peak between many
> detectors and so reducing the chance of pulse-pileup on any of them.
> (One would sum the photon counts from all detectors under the peak.)
>
> The above discussion can also be applied to APDs which have high QE
> but massive multiplicative noise.    Pulse counting is almost always
> used because high-gain APDs must be pulsed off to quench each
> single-photoelectron avalanche (and low-gain APD have VERY low
> effective QE because most photoelectrons fail to propagate at all and
> are lost) so you have the same pile-up problems and possible
> solutions. (or does anyone have one I haven't heard of?)
>
> Summary: Hybrid PMTs have such low multiplicative noise that you don't
> need pulse counting, so they don't saturate. However, if you reduce
> the first stage gain because the signal is large, then multiplicative
> noise may reappear. They are not cheap and so damaging one with too
> much input light may be a memorable experience.
>
> Multi-Mini-PMTs can do about as well    as long as they use
> pulse-counting.
>
> In all cases GaAsP photocathodes are wonderful in terms of QE but must
> be cooled to prevent high dark counts.
>
> If you are trying to make comparisons, I suggest that rather than
> looking at test specimens, you use light from the transmitted light
> source in your microscope. You can use your confocal emission filters
> to select a particular wavelength band and ND filters to adjust
> intensity. Set up Kohler illumination and turn off the laser and scan
> the "bright" field. As you will only be looking at one-pinhole's worth
> of light at any time, it is easy to adjust matters (pinhole size, NS
> filters, lamp power), to get a few photons/pixel. And, as the total
> signal coming through the plane of focus is about 1,000,000 times
> greater than that going through the pinhole, you can measure it fairly
> easily (Be sure to set the field diaphragm at a repeatable setting.).
> This is a number of variables that one can almost cope with.
>
> Put a specimen in and you have another 30 or so.
>
> Good luck,
>
> 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 10-22, 2012, UBC, Vancouver
> Canada
> Info: http://www.3dcourse.ubc.ca/          Applications accepted after
> 11/15/12
>                                "If it ain't diffraction, it must be
> statistics." Anon.
>
> <blockquote type=cite>*****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> *****
>
> Check the Hamamatsu H7422P series.
> Sudipta
>
> On Thu, 13 Oct 2011 07:11:49 +0000, Csúcs    Gábor wrote
> <blockquote type=cite>   *****
>    To join, leave or search the confocal microscopy listserv, go to:
>    http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
>    *****
>
>    Dear Ralf,
>
>    The ALMF people from the EMBL Heidelberg had a nice poster on this
> subject
>    at the latest ELMI meeting in Greece, so the may give you a more
> detailed
>    answer. As far as I recall, their take home message was that the
>    Zeiss GaAsP and the Leica HyDe perform virtually identically in
>    terms of dynamic range and S/N.
>
>    Greetings       Gabor
>
>    On 10/13/11 9:05 AM, "Ralf Zenke" &lt;[hidden email]&gt; wrote:
>
> &gt;*****
> &gt;To join, leave or search the confocal microscopy listserv, go to:
> &gt;http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> &gt;*****
> &gt;
> &gt;Dear Listers,
> &gt;
> &gt;I am trying to find out technical properties of Zeiss GaAsP detectors
> &gt;versus Leica Hybrid detectors. All I have is quantum efficiency, which
> &gt;is around 45% for both devices. But what about dynamic range (or
> signal
> &gt;to noise ratio)? Do GaAsP PMTs have the same SNR as conventional PMTs?
> &gt;From what I know, Leica's hybrid detectors come from Hamamatsu, but I
> &gt;can't find any in their product range. Does anybody know a product
> &gt;number?
> &gt;
> &gt;Thanks a lot for any kind of help!
> &gt;
> &gt;Ralf
> &gt;
> &gt;
> &gt;--
> &gt;Ralf Zenke
> &gt;
> &gt;Max Planck Institute of Biochemistry
> &gt;Core Facility
> &gt;Am Klopferspitz 18
> &gt;DE-82152 Martinsried near Munich
> &gt;GERMANY
> &gt;Phone: (+49) (89) 8578 3798
> &gt;Fax: (+49) (89) 8578 2847
> &gt;www.biochem.mpg.de
> </blockquote>
>
> Dr. Sudipta Maiti
> Dept. of Chemical Sciences
> Tata Institute of Fundamental Research
> Homi Bhabha Raod, Colaba, Mumbai 400005
> Ph. 91-22-2278-2716 / 2539
> Fax: [phone]
> alternate e-mail: [hidden email]
> url: biophotonics.weebly.com
> </blockquote>
>
> </blockquote>
>
> </body>
> </html>
> </html>

--
Dr. Wolfgang Staroske

Single Molecule Specialist
Light Microscopy Facility

Technische Universität Dresden
Biotechnology Center
Tatzberg 47/49
01307 Dresden, Germany

Tel.: [phone]
Fax.: [phone]
E-Mail: [hidden email]
Webpage: www.biotec.tu-dresden.de