N=?ISO-8859-1?Q?=FCv=FC?= EMCCD camera

[lechristophe]

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

I was quite surprized to read about the "most sensitive camera of the world"
in my mainstream web newspaper. Apparently a former PhD student from Canada
named Olivier Daigle has come up with a new type of EMCCD that has a lot
less noise that traditional ones, and has been bought "by the NASA" (quoting
the article here...). He has started a company named Nüvü that
commercializes the camera. You can find it on the company's website :
http://www.nuvucameras.com/en/en/Products.html
along with a spec sheet. I'm not a specialist so i was wondering what part
of it is hyped PR and what part is real innovation. So I'd be happy to have
the opinion of imaging specialists, what do you think ?

Thank you

[James Pawley]

*****
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
>*****
>
>I was quite surprized to read about the "most sensitive camera of the world"
>in my mainstream web newspaper. Apparently a former PhD student from Canada
>named Olivier Daigle has come up with a new type of EMCCD that has a lot
>less noise that traditional ones, and has been bought "by the NASA" (quoting
>the article here...). He has started a company named Nüvü that
>commercializes the camera. You can find it on the company's website :
>http://www.nuvucameras.com/en/en/Products.html
>along with a spec sheet. I'm not a specialist so i was wondering what part
>of it is hyped PR and what part is real innovation. So I'd be happy to have
>the opinion of imaging specialists, what do you think ?
>
>Thank you

Hi Christophe,

As I read it, this is just an EM-CCD with a photon-counting option.

Normal EM-CCDs can easily count the signal from
one photoelectron/pixel above the read noise, but
the penalty is that the single-photon pulse may
vary tremendously in size (multiplicative noise).
The photon counting version (That has in the past
been offered by Andor and others) merely detects
these single-photoelectron events, finds their
centroid and stores one count in the memory at
the correct location. Like all photon-counting
systems, it is subject to "Pile-up" errors (if
two, or 3 or 4) real PE are produced in the pixel
during the exposure time, they will still be
counted as a single event), so you must keep the
signal level so low that the chance of there
being 2 photoelectrons made during one read cycle
will be very low.  This implies a signal rate
that is really too low for most live cell
microscopy (the cell may change as you collect
your data).

Alternatively, you might employ some fancy signal
processing to separate real single PE/pixel
occurrences from parts of the image where
brighter objects are recorded. The single-PE
events might be discriminated with some (but not
total) accuracy by looking at nearby pixels or
the same pixels in earlier and later frames. As
Nyquist-sampled point objects must cover at least
roundish patch of 12 adjacent pixels, one might
be able to discriminate some big pules as
representing 2PE because for instance, after
processing may frames they could be seen to be
from pixels towards the center of the image of a
point object.

But this seems a lot of work for a small gain and
it is not clear that this camera tries any trick
of this type.

If I were looking for a new camera, I would look
at the new sCMOS camera. No multiplicative noise
(which effectively means that the QE is about 50%
higher than the effective QE of a EM-CCD without
photon counting or 50% less than one with photon
counting) and only 1-2 electrons of noise,
probably less than you would get from your level
of non-specific staining. As well, you get very
fast read out, massive dynamic range without
saturation and a lot of pixels.

Cheers,

Jim P.
--
***************************************************************************
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.

[Andreas Bruckbauer]

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


 Hi Jim,

from the Andor Brochure:
"The SNR superiority of sCMOS over even well-optimized Interline CCD technology can
clearly be observed, manifest as better contrast of signal against a
less noisy read noise background, resulting also in better resolution of
features. However, comparison of the two technologies against backilluminated
EMCCD (figure 2) at the weakest LED setting, showed
that the < 1 electron noise floor and higher QE of the EMCCD resulted
in superior contrast of the weak signal from the noise floor."

seems the sCMOS still has a way to go?
In the Andor webinar they had a slide stating that 1% of pixels have a read noise greater than 5 e-, so they employ a filter to reduce spurious noise.

best wishes

Andreas




 


 

 

-----Original Message-----
From: James Pawley <[hidden email]>
To: [hidden email]
Sent: Tue, 15 Feb 2011 22:12
Subject:  Re: Nüvü EMCCD camera


*****
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 
>*****
>
>I was quite surprized to read about the "most sensitive camera of the world"
>in my mainstream web newspaper. Apparently a former PhD student from Canada
>named Olivier Daigle has come up with a new type of EMCCD that has a lot
>less noise that traditional ones, and has been bought "by the NASA" (quoting
>the article here...). He has started a company named Nüvü that
>commercializes the camera. You can find it on the company's website :
>http://www.nuvucameras.com/en/en/Products.html 
>along with a spec sheet. I'm not a specialist so i was wondering what part
>of it is hyped PR and what part is real innovation. So I'd be happy to have
>the opinion of imaging specialists, what do you think ?
>
>Thank you

 
Hi Christophe,
 
As I read it, this is just an EM-CCD with a photon-counting option.
 
Normal EM-CCDs can easily count the signal from one photoelectron/pixel above the read noise, but the penalty is that the single-photon pulse may vary tremendously in size (multiplicative noise). The photon counting version (That has in the past been offered by Andor and others) merely detects these single-photoelectron events, finds their centroid and stores one count in the memory at the correct location. Like all photon-counting systems, it is subject to "Pile-up" errors (if two, or 3 or 4) real PE are produced in the pixel during the exposure time, they will still be counted as a single event), so you must keep the signal level so low that the chance of there being 2 photoelectrons made during one read cycle will be very low.  This implies a signal rate that is really too low for most live cell microscopy (the cell may change as you collect your data).
 
Alternatively, you might employ some fancy signal processing to separate real single PE/pixel occurrences from parts of the image where brighter objects are recorded. The single-PE events might be discriminated with some (but not total) accuracy by looking at nearby pixels or the same pixels in earlier and later frames. As Nyquist-sampled point objects must cover at least roundish patch of 12 adjacent pixels, one might be able to discriminate some big pules as representing 2PE because for instance, after processing may frames they could be seen to be from pixels towards the center of the image of a point object.
 
But this seems a lot of work for a small gain and it is not clear that this camera tries any trick of this type.
 
If I were looking for a new camera, I would look at the new sCMOS camera. No multiplicative noise (which effectively means that the QE is about 50% higher than the effective QE of a EM-CCD without photon counting or 50% less than one with photon counting) and only 1-2 electrons of noise, probably less than you would get from your level of non-specific staining. As well, you get very fast read out, massive dynamic range without saturation and a lot of pixels.
 
Cheers,
 
Jim P.
-- ***************************************************************************
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.

 

[lechristophe]

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

On the subject of sCMOS camera, I was told that (remember I'm not a
specialist at all so I may have misunderstood) the Andor technology (5 MPx
camera) uses independant amplification of each pixels and that this can lead
to important differences between each pixels. Whereas the Hamamatsu sCMOS
with fewer pixels (3 MPx) is different and somehow more reliable in terms of
constant gain across pixels. Does that makes sense to you ?

Christophe

On Tue, Feb 15, 2011 at 23:56, Andreas Bruckbauer <[hidden email]> wrote:

> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> *****
>
>
>  Hi Jim,
>
> from the Andor Brochure:
> "The SNR superiority of sCMOS over even well-optimized Interline CCD
> technology can
> clearly be observed, manifest as better contrast of signal against a
> less noisy read noise background, resulting also in better resolution of
> features. However, comparison of the two technologies against
> backilluminated
> EMCCD (figure 2) at the weakest LED setting, showed
> that the < 1 electron noise floor and higher QE of the EMCCD resulted
> in superior contrast of the weak signal from the noise floor."
>
> seems the sCMOS still has a way to go?
> In the Andor webinar they had a slide stating that 1% of pixels have a read
> noise greater than 5 e-, so they employ a filter to reduce spurious noise.
>
> best wishes
>
> Andreas
>
>
>
>
>
>
>
>
>
>
>
> -----Original Message-----
> From: James Pawley <[hidden email]>
> To: [hidden email]
> Sent: Tue, 15 Feb 2011 22:12
> Subject:  Re: Nüvü EMCCD camera
>
>
> *****
> 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
> >*****
> >
> >I was quite surprized to read about the "most sensitive camera of the
> world"
> >in my mainstream web newspaper. Apparently a former PhD student from
> Canada
> >named Olivier Daigle has come up with a new type of EMCCD that has a lot
> >less noise that traditional ones, and has been bought "by the NASA"
> (quoting
> >the article here...). He has started a company named Nüvü that
> >commercializes the camera. You can find it on the company's website :
> >http://www.nuvucameras.com/en/en/Products.html
> >along with a spec sheet. I'm not a specialist so i was wondering what part
> >of it is hyped PR and what part is real innovation. So I'd be happy to
> have
> >the opinion of imaging specialists, what do you think ?
> >
> >Thank you
>
> Hi Christophe,
>
> As I read it, this is just an EM-CCD with a photon-counting option.
>
> Normal EM-CCDs can easily count the signal from one photoelectron/pixel
> above the read noise, but the penalty is that the single-photon pulse may
> vary tremendously in size (multiplicative noise). The photon counting
> version (That has in the past been offered by Andor and others) merely
> detects these single-photoelectron events, finds their centroid and stores
> one count in the memory at the correct location. Like all photon-counting
> systems, it is subject to "Pile-up" errors (if two, or 3 or 4) real PE are
> produced in the pixel during the exposure time, they will still be counted
> as a single event), so you must keep the signal level so low that the chance
> of there being 2 photoelectrons made during one read cycle will be very low.
>  This implies a signal rate that is really too low for most live cell
> microscopy (the cell may change as you collect your data).
>
> Alternatively, you might employ some fancy signal processing to separate
> real single PE/pixel occurrences from parts of the image where brighter
> objects are recorded. The single-PE events might be discriminated with some
> (but not total) accuracy by looking at nearby pixels or the same pixels in
> earlier and later frames. As Nyquist-sampled point objects must cover at
> least roundish patch of 12 adjacent pixels, one might be able to
> discriminate some big pules as representing 2PE because for instance, after
> processing may frames they could be seen to be from pixels towards the
> center of the image of a point object.
>
> But this seems a lot of work for a small gain and it is not clear that this
> camera tries any trick of this type.
>
> If I were looking for a new camera, I would look at the new sCMOS camera.
> No multiplicative noise (which effectively means that the QE is about 50%
> higher than the effective QE of a EM-CCD without photon counting or 50% less
> than one with photon counting) and only 1-2 electrons of noise, probably
> less than you would get from your level of non-specific staining. As well,
> you get very fast read out, massive dynamic range without saturation and a
> lot of pixels.
>
> Cheers,
>
> Jim P.
> --
> ***************************************************************************
> 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.
>
>
>

[James Pawley]

*****
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
>*****
>
>On the subject of sCMOS camera, I was told that (remember I'm not a
>specialist at all so I may have misunderstood) the Andor technology (5 MPx
>camera) uses independant amplification of each pixels and that this can lead
>to important differences between each pixels. Whereas the Hamamatsu sCMOS
>with fewer pixels (3 MPx) is different and somehow more reliable in terms of
>constant gain across pixels. Does that makes sense to you ?
>
>Christophe

Hi again,

You are right. Every CMOS pixels has its own 3-5
transistor charge amplifier and reset circuit.
All the amplifiers for each column of pixels are
turned on at the same time and as a result, the
integration time or each amplifier is longer than
it would be on a CCD (that reads out each pixel
one-at-a-time in sequence) by a factor about
equal to the number of pixels in the column. This
is how CMOS can push the read noise so low.

But as you note, this approach also gives
tremendous fixed-pattern noise because the gain
and offset of each on-pixel amplifier is a little
different from that of its neighbors (and indeed,
each of the 10,000 A/D converters is also a
little different too.).

The good news is that, at a given temperature,
the gain and offset of each pixel (and A/D
converter) is relatively constant. Therefore, it
is possible to measure and store these parameters
(using images of first black and then white
fields) for every pixel. When the camera is in
use, these results are used to normalize
on-the-fly the signal coming from each pixel of
the sensor. Indeed, much of the remaining "read
noise" in the specs on these camera is  actually
due to errors in this re-normalization process.

So yes, it is a problem. But the problem has been
largely solved with on-the-fly normalization. The
results are visible in the pictures.

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 Tue, Feb 15, 2011 at 23:56, Andreas Bruckbauer <[hidden email]> wrote:
>
>>  *****
>>  To join, leave or search the confocal microscopy listserv, go to:
>>  http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
>>  *****
>>
>>
>>   Hi Jim,
>>
>>  from the Andor Brochure:
>>  "The SNR superiority of sCMOS over even well-optimized Interline CCD
>>  technology can
>>  clearly be observed, manifest as better contrast of signal against a
>>  less noisy read noise background, resulting also in better resolution of
>>  features. However, comparison of the two technologies against
>>  backilluminated
>>  EMCCD (figure 2) at the weakest LED setting, showed
>>  that the < 1 electron noise floor and higher QE of the EMCCD resulted
>>  in superior contrast of the weak signal from the noise floor."
>>
>>  seems the sCMOS still has a way to go?
>>  In the Andor webinar they had a slide stating that 1% of pixels have a read
>>  noise greater than 5 e-, so they employ a filter to reduce spurious noise.
>>
>>  best wishes
>>
>>  Andreas
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>  -----Original Message-----
>>  From: James Pawley <[hidden email]>
>>  To: [hidden email]
>>  Sent: Tue, 15 Feb 2011 22:12
>>  Subject:  Re: Nüvü EMCCD camera
>>
>>
>>  *****
>>  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
>>  >*****
>>  >
>>  >I was quite surprized to read about the "most sensitive camera of the
>  > world"
>>  >in my mainstream web newspaper. Apparently a former PhD student from
>>  Canada
>>  >named Olivier Daigle has come up with a new type of EMCCD that has a lot
>>  >less noise that traditional ones, and has been bought "by the NASA"
>>  (quoting
>>  >the article here...). He has started a company named Nüvü that
>>  >commercializes the camera. You can find it on the company's website :
>>  >http://www.nuvucameras.com/en/en/Products.html
>>  >along with a spec sheet. I'm not a specialist so i was wondering what part
>>  >of it is hyped PR and what part is real innovation. So I'd be happy to
>>  have
>>  >the opinion of imaging specialists, what do you think ?
>>  >
>>  >Thank you
>>
>>  Hi Christophe,
>>
>>  As I read it, this is just an EM-CCD with a photon-counting option.
>>
>>  Normal EM-CCDs can easily count the signal from one photoelectron/pixel
>>  above the read noise, but the penalty is that the single-photon pulse may
>>  vary tremendously in size (multiplicative noise). The photon counting
>>  version (That has in the past been offered by Andor and others) merely
>>  detects these single-photoelectron events, finds their centroid and stores
>>  one count in the memory at the correct location. Like all photon-counting
>>  systems, it is subject to "Pile-up" errors (if two, or 3 or 4) real PE are
>>  produced in the pixel during the exposure time, they will still be counted
>>  as a single event), so you must keep the signal level so low that the chance
>>  of there being 2 photoelectrons made during one read cycle will be very low.
>>   This implies a signal rate that is really too low for most live cell
>>  microscopy (the cell may change as you collect your data).
>>
>>  Alternatively, you might employ some fancy signal processing to separate
>>  real single PE/pixel occurrences from parts of the image where brighter
>>  objects are recorded. The single-PE events might be discriminated with some
>>  (but not total) accuracy by looking at nearby pixels or the same pixels in
>>  earlier and later frames. As Nyquist-sampled point objects must cover at
>>  least roundish patch of 12 adjacent pixels, one might be able to
>>  discriminate some big pules as representing 2PE because for instance, after
>>  processing may frames they could be seen to be from pixels towards the
>>  center of the image of a point object.
>>
>>  But this seems a lot of work for a small gain and it is not clear that this
>>  camera tries any trick of this type.
>>
>>  If I were looking for a new camera, I would look at the new sCMOS camera.
>>  No multiplicative noise (which effectively means that the QE is about 50%
>>  higher than the effective QE of a EM-CCD without photon counting or 50% less
>>  than one with photon counting) and only 1-2 electrons of noise, probably
>>  less than you would get from your level of non-specific staining. As well,
>>  you get very fast read out, massive dynamic range without saturation and a
>>  lot of pixels.
>>
>>  Cheers,
>>
>>  Jim P.
>>  --
>>  ***************************************************************************
>>  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.
>>
>>
>>

--
***************************************************************************
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.

[Olivier Daigle]

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

Dear Christophe and Jim,
  I have been made aware that messages related to Nüvü Cameras' technology
were exchanged on confocalmicroscopy. As the CTO of Nüvü Cameras, I would
like to add my 2 cents to this discussion.

  We do build EMCCD cameras and we use the same EMCCD chips as every
other EMCCD camera maker (from e2v technologies). However, what
differentiates our cameras from the others is not only a photon counting
option: every camera maker has this function. The main difference between
our cameras and the others is the low level of spurious Clock Induced Charges
(CIC) that are generated during the EMCCD read-out, which is lower than
0.002 electron/pixel. And when it comes to photon counting, which, as Jim
noted, is restricted to low flux levels and/or high frame rates to avoid losses
by coincidence, the low CIC level is of extremely high importance as it is the
dominating source of noise. I must stress this: we do not filter the noise. We
just don't generate it. We achieve this by using an electronic controller (the
one that produces the electric signals used to clock the pixels out of the
EMCCD) that generates electric signals that are different from the controllers
used in the other cameras. By doing so, we generate less spurious charges
when we move the electrons around the EMCCD.

  Another differentiating aspect of our cameras is the higher achievable EM
gain. This is of importance, as a higher EM gain allows, in photon counting, to
extract more photo-electrons out of the read-out noise. Its like having a
higher Quantum Efficiency (QE). There is no big magic here: it is only a
matter of producing a clock that has a higher voltage. However, a higher EM
gain also means a higher level of CIC. With out controller, we can reach a high
(>3000) EM gain while generating less CIC than our competitors.

  In photon counting, the lower dynamic range can be compensated by using a
higher frame rate and adding frames together by post-processing. Of course,
this requires a higher frame rate, which generates a higher total level of CIC
(CIC is read-out dependent, as compared to dark noise which is time
dependent). Once again, the low CIC level generated by our cameras has its
advantages.

  The use of this controller also has the advantage of yielding a higher charge
transfer efficiency (CTE). Some research fields, especially the ones using high
resolution spectroscopy, are rejecting the EMCCD technology because of their
less-than-optimal CTE. We hope that our higher CTE cameras will allow them
to re-think the usage of EMCCDs for their applications.

  For higher flux applications, the EMCCD can be operated at a lower gain to
avoid saturation, and get sub-electron read-out noise. Of course, you are then
plagued by the Excess Noise Factor that has the same effect on the SNR as if
you would halve the QE. However, if you get a single photon in a pixel, you
will see it with >5 sigma confidence level (read-out noise wise), which is
impossible to do with a sCMOS or a conventional CCD. At the same time, you
will be able to image the brighter regions of your image and be shot noise
limited there. EMCCD cameras also have the advantage of being operated at a
lower temperature than sCMOS, and benefit of a lower dark noise (about a
hundred times lower). sCMOS also have a lower QE (peak at 55%) while the
EMCCD peaks at around 92%. Divide that by 2 and you are not that far from
sCMOS, have lower dark noise, lower read-out noise, better pixel response
uniformity, Gaussian noise, etc. Of course, sCMOS can reach hundreds of FPS,
and it is 3 to 5 Mpix. It is just a matter of time before larger format EMCCD
will be out. Still, for some applications, sCMOS is better than EMCCDs. The
reverse is also true.

  These were my 2 cents. If you have questions, feel free to write back to me
at odaigle __at__ nuvucameras dot com.

[Mike Buchin]

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

The review of the NuVu emccd provided by Olivier was well worth the 2 cents.
I will like to add 2 cents to the discussion as well.  Another technology
that is often overlooked for single photon imaging and detection is the ICCD
camera.  An optimized dual MCP configuration with a GaAsP photocathode
(40-50% QE across the visible to NIR spectrum) cooled to -20C will have
25-50 dark counts per 1.4 Mpixel frame per second.  This is 0.36 photons per
mm**2 per second or 0.000036 dark counts per pixel per second.  The
intensifier tube acts as a high gain (1,000 to 1,000,000) preamplifier ahead
of the CCD (typically SONY ICX285 or ICX 414 for higher speeds).  The net
result here is that each single photon event imaged can have a signal to
noise of 10-20x relative to the CCD readout noise.  The CCD readout noise is
easily thresholded, leaving only the single photon image against an
absolutely black background.  Real time processing of single photon events,
such as subpixel localization, counting, centroid calculation of photon
clusters etc.. is simplified since readout noise is virtually zero. Only the
photon energy packet is imaged and enters into the math.  Another advantage
of being able to threshold CCD read noise with minimal single photon signal
loss is that that you can increase readout speeds/frame rates without adding
any noise factor whatsoever.   High gain also allows coupling to less
sensitive image sensors (e.g. 1.4 Mpix 500 fps CMOS) without losing single
photon detection capability in the electronic noise floor.

Running at high gain reduces dynamic range (same as the emccd), but going
faster to reduce photon coincidence or reducing gain for gray scale imaging
works in an ICCD as well.  We always use RAM summation vs. on chip
integration in the single photon domain.  After thresholding and removal of
cosmic ray events (real time), we can sum into virtual pixel wells that are
limited in depth/capacity only by the bit depth of the summation.

ICCD cameras may also be gated for synchronizing single photon detection to
laser/led pulses at high rates (we have worked at speeds up to 5 MHz).  We
also use autogating (hardware or software controlled) to allow the ICCD to
range over 6-7 decades of input light level when quantitative/fixed gain
data is not the objective.

In our experience, an optimized ICCD offers some distinct advantages over
emccd cameras in applications such as BRET, CRET and both fast and slow
(24/7 circadian rhythm) luciferase based imaging.  We also see excellent
results in the high speed single molecule domain and TRFM (lanthanides
etc.).  We have been working with the new SCMOS products and have used
emccd's as well.  Each technology has a sweet spot and can be complimentary.
At some point in time, I am sure we will get to see how the NuVu and the
cooled cathode GaAsP ICCD cameras compare in the category "most sensitive".

Olivier, nice overview.

Mike

[hidden email]

 
Michael Buchin
Stanford Photonics, Inc.
Ph: 650-969-5991


-----Original Message-----
From: Confocal Microscopy List [mailto:[hidden email]] On
Behalf Of Olivier Daigle
Sent: Saturday, February 19, 2011 2:41 PM
To: [hidden email]
Subject: Re: Nüvü EMCCD camera

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

Dear Christophe and Jim,
  I have been made aware that messages related to Nüvü Cameras' technology
were exchanged on confocalmicroscopy. As the CTO of Nüvü Cameras, I would
like to add my 2 cents to this discussion.

  We do build EMCCD cameras and we use the same EMCCD chips as every
other EMCCD camera maker (from e2v technologies). However, what
differentiates our cameras from the others is not only a photon counting
option: every camera maker has this function. The main difference between
our cameras and the others is the low level of spurious Clock Induced
Charges
(CIC) that are generated during the EMCCD read-out, which is lower than
0.002 electron/pixel. And when it comes to photon counting, which, as Jim
noted, is restricted to low flux levels and/or high frame rates to avoid
losses
by coincidence, the low CIC level is of extremely high importance as it is
the
dominating source of noise. I must stress this: we do not filter the noise.
We
just don't generate it. We achieve this by using an electronic controller
(the
one that produces the electric signals used to clock the pixels out of the
EMCCD) that generates electric signals that are different from the
controllers
used in the other cameras. By doing so, we generate less spurious charges
when we move the electrons around the EMCCD.

  Another differentiating aspect of our cameras is the higher achievable EM
gain. This is of importance, as a higher EM gain allows, in photon counting,
to
extract more photo-electrons out of the read-out noise. Its like having a
higher Quantum Efficiency (QE). There is no big magic here: it is only a
matter of producing a clock that has a higher voltage. However, a higher EM
gain also means a higher level of CIC. With out controller, we can reach a
high
(>3000) EM gain while generating less CIC than our competitors.

  In photon counting, the lower dynamic range can be compensated by using a
higher frame rate and adding frames together by post-processing. Of course,
this requires a higher frame rate, which generates a higher total level of
CIC
(CIC is read-out dependent, as compared to dark noise which is time
dependent). Once again, the low CIC level generated by our cameras has its
advantages.

  The use of this controller also has the advantage of yielding a higher
charge
transfer efficiency (CTE). Some research fields, especially the ones using
high
resolution spectroscopy, are rejecting the EMCCD technology because of their

less-than-optimal CTE. We hope that our higher CTE cameras will allow them
to re-think the usage of EMCCDs for their applications.

  For higher flux applications, the EMCCD can be operated at a lower gain to

avoid saturation, and get sub-electron read-out noise. Of course, you are
then
plagued by the Excess Noise Factor that has the same effect on the SNR as if

you would halve the QE. However, if you get a single photon in a pixel, you
will see it with >5 sigma confidence level (read-out noise wise), which is
impossible to do with a sCMOS or a conventional CCD. At the same time, you
will be able to image the brighter regions of your image and be shot noise
limited there. EMCCD cameras also have the advantage of being operated at a
lower temperature than sCMOS, and benefit of a lower dark noise (about a
hundred times lower). sCMOS also have a lower QE (peak at 55%) while the
EMCCD peaks at around 92%. Divide that by 2 and you are not that far from
sCMOS, have lower dark noise, lower read-out noise, better pixel response
uniformity, Gaussian noise, etc. Of course, sCMOS can reach hundreds of FPS,

and it is 3 to 5 Mpix. It is just a matter of time before larger format
EMCCD
will be out. Still, for some applications, sCMOS is better than EMCCDs. The
reverse is also true.

  These were my 2 cents. If you have questions, feel free to write back to
me
at odaigle __at__ nuvucameras dot com.