seeking suggestions on new sCMOS camera

[Zulin Yu]

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

Dear All,
    I am writing to ask your suggestions about sCMOS camera on spinning disk
system. Recently, I  got a demo (Hamamatsu Flash 4.0) camera on our spinning
disk system (PerkinElmer Ultraview) and I am a little bit disappointed about its
performance  - it is a decent camera, but not as good as I expected.
    Based on the company's brochure, Flash 4.0 SNR exceeds that of EM-CCDs at
about 6 photon/pixel
(http://sales.hamamatsu.com/assets/pdf/hpspdf/e_flash4.pdf); several research
paper also claimed that the latest sCMOS is as good as EMCCD with more than 15
photons (http://www.ncbi.nlm.nih.gov/pubmed?term=EMCCD%20sCMOS).
However, we still got a much higher SNR (double) on EMCCD even we have more
than 50 photons/pixel.  Did anybody have experience with the high end sCMOS
camera and compared it with back-thinned EMCCD? I appreciate any input, Zulin

[Gerhard Holst]

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

Dear Zulin Yu,

I have seen Visitron efficiently using 2 sCMOS cameras pco.edge with their new yokogawa Spinning disk system.
In general the calculations suggest that from, let's say 10-12 photons the SNR of corresponding sCMOS cameras is better than the emCCD cameras, but this also means front illuminated compared to front illuminated image sensors. From the guys at Visitron I have heard, that they are happy with the synchronization of the sCMOS cameras, but they mostly used Photometrics Evolve cameras back illuminated emCCD cameras before.
Still, if you in the photon counting range, the BI emCCD would do better.

I have to state, that I work for PCO, but I conducted the sCMOS project on our side.

with best regards,

Gerhard Holst
_______________________________

Dr. Gerhard Holst
Science & Research
PCO AG
Donaupark 11
93309 Kelheim, Germany
fon [phone]
fax [phone]
mob [phone]
[hidden email]
www.pco.de


-----Ursprüngliche Nachricht-----
Von: Confocal Microscopy List [mailto:[hidden email]] Im Auftrag von Zulin Yu
Gesendet: Mittwoch, 16. Januar 2013 05:46
An: [hidden email]
Betreff: seeking suggestions on new sCMOS camera

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

Dear All,
    I am writing to ask your suggestions about sCMOS camera on spinning disk
system. Recently, I  got a demo (Hamamatsu Flash 4.0) camera on our spinning
disk system (PerkinElmer Ultraview) and I am a little bit disappointed about its
performance  - it is a decent camera, but not as good as I expected.
    Based on the company's brochure, Flash 4.0 SNR exceeds that of EM-CCDs at
about 6 photon/pixel
(http://sales.hamamatsu.com/assets/pdf/hpspdf/e_flash4.pdf); several research
paper also claimed that the latest sCMOS is as good as EMCCD with more than 15
photons (http://www.ncbi.nlm.nih.gov/pubmed?term=EMCCD%20sCMOS).
However, we still got a much higher SNR (double) on EMCCD even we have more
than 50 photons/pixel.  Did anybody have experience with the high end sCMOS
camera and compared it with back-thinned EMCCD? I appreciate any input, Zulin

[Zdenek Svindrych]

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

Hi Zulin,

So the CMOS is better than your EMCCD? Then go for it!

We are using Andor Neo, it is faster and has more pixels than EMCCDs. Also
the image looks OK. We are using the CMOS on SIM, where the light level is
not such an issue. We are consitering it also for STORM where it should be
comparable to EMCCD.

We also have a spinning disc with back-thinned EMCCD an ad I recall the
camera already saturates with (almost) minimum laser level at 30fps with 200
x gain (provided the sample is well stained, of course). So there is also a
lot of light and I assume the CMOS should do the job.

As to the noise of your EMCCD, I'm not sure how you calibrated the photon
flux. I am only able to calculate the photon flux from the poisson noise of
an image series, so the noise then always appears 'correct' with respect to
the photon count. So your photon calibration might be wrong.

If you suspect your camera has exrta noise, try to plot the standard
deviation of your pixels versus mean intensity for a broad range of
intensities. Then you should be able to identify the extra noise. You may
also try to record the same intensity distribution without EM gain and with
the other (non-EM, slow) port.

Cheers,
zdenek


---------- Původní zpráva ----------
Od: Zulin Yu <[hidden email]>
Datum: 16. 1. 2013
Předmět: seeking suggestions on new sCMOS camera

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

Dear All,
I am writing to ask your suggestions about sCMOS camera on spinning disk
system. Recently, I got a demo (Hamamatsu Flash 4.0) camera on our spinning
disk system (PerkinElmer Ultraview) and I am a little bit disappointed about
its
performance - it is a decent camera, but not as good as I expected.
Based on the company's brochure, Flash 4.0 SNR exceeds that of EM-CCDs at
about 6 photon/pixel
(http://sales.hamamatsu.com/assets/pdf/hpspdf/e_flash4.pdf); several
research
paper also claimed that the latest sCMOS is as good as EMCCD with more than
15
photons (http://www.ncbi.nlm.nih.gov/pubmed?term=EMCCD%20sCMOS).
However, we still got a much higher SNR (double) on EMCCD even we have more
than 50 photons/pixel. Did anybody have experience with the high end sCMOS
camera and compared it with back-thinned EMCCD? I appreciate any input,
Zulin"

[Arne Seitz]

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

Comparing cameras is always very difficult and the outcome will strongly depend on the application. My personal strategy is always to rather rely on own data than what was written in product brochures or papers. Thus if you get better images with an EM CCD camera this is the camera of choice.

I tested an PCO.edge sCMOS camera (no commercial interest) on the new Yokogawa W1 spinning disk head and I was really impressed. For me the main advantage of this type of camera is the large field of view in combination with the high acquisition speed. Also the small pixel is a plus as it allows imaging under nyquist conditions (or even doing oversampling) with low magnification objectives having a high NA. Nevertheless at low light application I prefer using a back illuminated EM CCD camera.

And on a conventional spinning disk system it is at least questionable whether a sCMOS camera is the best camera, as large areas of the chip remain unused.

Regards
Arne

> -----Original Message-----
> From: Confocal Microscopy List
> [mailto:[hidden email]] On Behalf Of Zulin Yu
> Sent: mercredi 16 janvier 2013 05:46
> To: [hidden email]
> Subject: seeking suggestions on new sCMOS camera
>
> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> *****
>
> Dear All,
>     I am writing to ask your suggestions about sCMOS camera on spinning disk
> system. Recently, I  got a demo (Hamamatsu Flash 4.0) camera on our
> spinning disk system (PerkinElmer Ultraview) and I am a little bit
> disappointed about its performance  - it is a decent camera, but not as good
> as I expected.
>     Based on the company's brochure, Flash 4.0 SNR exceeds that of EM-CCDs
> at about 6 photon/pixel
> (http://sales.hamamatsu.com/assets/pdf/hpspdf/e_flash4.pdf); several
> research paper also claimed that the latest sCMOS is as good as EMCCD with
> more than 15 photons
> (http://www.ncbi.nlm.nih.gov/pubmed?term=EMCCD%20sCMOS).
> However, we still got a much higher SNR (double) on EMCCD even we have
> more than 50 photons/pixel.  Did anybody have experience with the high
> end sCMOS camera and compared it with back-thinned EMCCD? I appreciate
> any input, Zulin

[Nico Stuurman-3]

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

John Oreoupolus did a nice comparison of EMCCDs and sCMOS cameras on spinning disk: http://www.spectral.ca/_files/file.php?fileid=fileWFBaXTDIaW&filename=file_4_CSU_Camera_Comparison__sCMOSvsEMCCD.pdf

B.t.w., to do a fair comparison, you will need to change the magnification to the camera.  Hopefully, your current setup is optimized for the 16 micron pixels in your EMCCD.  Swapping in the sCMOS without changing magnification will give you a lot of empty magnification that will adversely affect noise characteristics.  Binning does not really help since in a CMOS all pixels are read out individually anyways (i.e. hardware and software binning does the same thing, unlike CCDs).

Best,

Nico

On Jan 15, 2013, at 8:45 PM, Zulin Yu wrote:

> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> *****
>
> Dear All,
>    I am writing to ask your suggestions about sCMOS camera on spinning disk
> system. Recently, I  got a demo (Hamamatsu Flash 4.0) camera on our spinning
> disk system (PerkinElmer Ultraview) and I am a little bit disappointed about its
> performance  - it is a decent camera, but not as good as I expected.
>    Based on the company's brochure, Flash 4.0 SNR exceeds that of EM-CCDs at
> about 6 photon/pixel
> (http://sales.hamamatsu.com/assets/pdf/hpspdf/e_flash4.pdf); several research
> paper also claimed that the latest sCMOS is as good as EMCCD with more than 15
> photons (http://www.ncbi.nlm.nih.gov/pubmed?term=EMCCD%20sCMOS).
> However, we still got a much higher SNR (double) on EMCCD even we have more
> than 50 photons/pixel.  Did anybody have experience with the high end sCMOS
> camera and compared it with back-thinned EMCCD? I appreciate any input, Zulin

[Nico Stuurman-3]

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

John Oreopoulos did a nice comparison of EMCCDs and sCMOS cameras on spinning disk: http://www.spectral.ca/_files/file.php?fileid=fileWFBaXTDIaW&filename=file_4_CSU_Camera_Comparison__sCMOSvsEMCCD.pdf

B.t.w., to do a fair comparison, you will need to change the magnification to the camera.  Hopefully, your current setup is optimized for the 16 micron pixels in your EMCCD.  Swapping in the sCMOS without changing magnification will give you a lot of empty magnification that will adversely affect noise characteristics.  Binning does not really help since in a CMOS all pixels are read out individually anyways (i.e. hardware and software binning does the same thing, unlike CCDs).

Best,

Nico

[Andrea Latini-4]

CONTENTS DELETED

The author has deleted this message.

[Zulin Yu]

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

Hello All,
    Thanks for all replies!
    Here is a quick summary wrote by my senior colleague and I hope his summary
will clarify some questions.  "I have been using the standard deviation of a dark
area as the noise, the average of the dark area as a background and the average
of a high signal area as the “max” (SNR=(Max-Min)/noise).  I have checked, the
min and noise in min figures are the same in the completely dark region of my
sample as they are in a truly dark image.  The gain on the EM is 255 and the
amplifier is 5.  I am guessing a total gain of 550-750 which puts the photon count
per pixel somewhere between 40-60." Under the condition of 40-60 photons,
EMCCD SNR is 46, much higher than 12 on sCMOS.
    I had a phone call with Hamamatsu technical support today, based on their
experiment, the EMCCD should have a higher SNR once the signal is higher than 6-
15 photons. However, our results have big disagreement with them. Something
must be wrong, either our measurement or either the camera. So we send our
original data to Hamamatsu for further investigation. I will keep you updated  once
I hear something from Hamamatsu.
   Ps, the EMCCD we used is also from Hamamatsu (c9100-13).
   Thanks a lot, Zulin

[Gerhard Holst]

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

Hello Zulin,

maybe you should consider, as I mentioned, that your emCCD camera, in case it is a back illuminated emCCD sensor, has a higher quantum efficiency, which would result it 30% more signal for the same amount of photons.

Second if you compare, please make sure, that the same amount of photons reaches the same amount of pixels, for example, if you use the same optics and image your sample to a camera with 12µm and to a camera with 6µm pixels, then the amount of photons hitting the 12µm pixel will be shared by 4 pixels on the camera with the smaller pixels, which is not a characteristic of the sensor but the optical situation. And if you provide only a 1/4 of the signal the corresponding SNR will be smaller.

Further, in your noise calculation, you only consider the spatial noise, meaning the inhomogeneity between the pixels, the temporal noise (from image to image) is neglected. If this is the relevant noise to you, that's ok.

But anyway, you should use the camera which fulfills best the requirements for your application, nothing else.

with best regards,

Gerhard
_______________________________

Dr. Gerhard Holst
Science & Research
PCO AG
Donaupark 11
93309 Kelheim, Germany
fon [phone]
fax [phone]
mob [phone]
[hidden email]
www.pco.de

-----Ursprüngliche Nachricht-----
Von: Confocal Microscopy List [mailto:[hidden email]] Im Auftrag von Zulin Yu
Gesendet: Donnerstag, 17. Januar 2013 04:41
An: [hidden email]
Betreff: Re: seeking suggestions on new sCMOS camera

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

Hello All,
    Thanks for all replies!
    Here is a quick summary wrote by my senior colleague and I hope his summary
will clarify some questions.  "I have been using the standard deviation of a dark
area as the noise, the average of the dark area as a background and the average
of a high signal area as the "max" (SNR=(Max-Min)/noise).  I have checked, the
min and noise in min figures are the same in the completely dark region of my
sample as they are in a truly dark image.  The gain on the EM is 255 and the
amplifier is 5.  I am guessing a total gain of 550-750 which puts the photon count
per pixel somewhere between 40-60." Under the condition of 40-60 photons,
EMCCD SNR is 46, much higher than 12 on sCMOS.
    I had a phone call with Hamamatsu technical support today, based on their
experiment, the EMCCD should have a higher SNR once the signal is higher than 6-
15 photons. However, our results have big disagreement with them. Something
must be wrong, either our measurement or either the camera. So we send our
original data to Hamamatsu for further investigation. I will keep you updated  once
I hear something from Hamamatsu.
   Ps, the EMCCD we used is also from Hamamatsu (c9100-13).
   Thanks a lot, Zulin