Re: sCMOS camera -global vs rolling exposure?

[Shalin Mehta]

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

Hi Michael,
We have demoed sCMOS cameras from Hamamatsu, Andor, and PCO.  I found
that Flash 4 has better sensitivity as compared to Andor or PCO's
cameras. I have been told by Hamamatsu's representative that the
Micro-manager driver is very useable, but it does not handle the top
frame-rate of Flash 4.

We are still undecided and I need to understand a couple of issues
before we make the decision.

For our target application (quantitative imaging and analysis of fast
moving specimens) the rolling shutter-only operation of Flash 4 is an
issue. To investigate the performance of global vs. rolling exposure,
we used Andor's Neo with the same settings in either mode. Here is a
dark-field image of an intact sea urchin sperm beating at about 30
waveforms per second:
http://www.mshalin.com/interesting-data/GlobalvsRollingShutter.png

From the images and intensity profiles we see that rolling shutter
introduces motion artifacts such as skew in the shape of head and
flagellum waveform in vertical direction. But, rolling exposure has
visibly better sensitivity and SNR.

I still don't understand the following completely. All inputs from
community and manufacturers will be appreciated.

a) How does global exposure work in Andor and PCO cameras? What
sources of noise have increased contribution when doing global
exposure?

 b) For any camera in rolling shutter mode, we can achieve global
exposure by strobing the light-source (such as LED or laser) so that
* the specimen is illuminated right after all the pixels have been
rolled active
* the source is turned-off right before the pixels are read out in
rolling fashion

The cameras take about 10 ms 'to roll the full chip'. So if we set the
exposure to desired global exposure + 20ms and strobe as above, it
seems we again have global exposure. Will noise be any better with
this custom arrangement?

Thanks,
Shalin
website: http://mshalin.com
(office) Lillie 110, (ph) 508-289-7374.

HFSP Postdoctoral Fellow,
Marine Biological Laboratory,
7 MBL Street, Woods Hole MA 02543, USA


On Sun, Aug 12, 2012 at 1:37 PM, Cammer, Michael
<[hidden email]> wrote:

> We demoed an Andor Neo camera a few weeks ago and was very impressed with it.  We couldnt quite get the sensitivity needed for single molecule imaging but were very impressed by its other features.  We'd like to get a camera with these capabilities that we may use both with Nikon Elements on one microscope and with MicroManager on another (currently one microscope has a DU897 and we want ta bigger field and wider dynamic range of view for some applications and the other has decade old Orca ER).
>
> Two part question.
>
> 1.  Is anyone developing a driver for MicroManager that will take advantage of this camera's many many features?
>
> 2.  While we really liked the Andor Neo, does anybody have data or feelings that other sCMOS cameras out there are better-- keeping in mind that drivers for both software packages is also important.
>
> Thank you!!
>
> _________________________________________
> Michael Cammer, Assistant Research Scientist
> Skirball Institute of Biomolecular Medicine
> Lab: [phone]  Cell: [phone]
>
>
> ------------------------------------------------------------------------------
> Live Security Virtual Conference
> Exclusive live event will cover all the ways today's security and
> threat landscape has changed and how IT managers can respond. Discussions
> will include endpoint security, mobile security and the latest in malware
> threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
> _______________________________________________
> micro-manager-general mailing list
> [hidden email]
> https://lists.sourceforge.net/lists/listinfo/micro-manager-general

[Andrew York]

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

 We've some experience using the pco.edge in both global shutter and
rolling-with-strobe mode. We generally don't use the pco in rolling shutter
without strobe, because of the artifacts you mention. Compared to global
shutter, we get higher framerates in rolling-with-strobe mode, and seem to
get lower noise, although I've done nothing truly rigorous to show this.

 You can calculate how long you have to wait 'til global exposure with the
pco.edge: ~4.577 microseconds for each horizontal line in your region of
interest. It's really twice this much per line, but the camera
simultaneously deals with one line from the top and bottom of the chip, so
effectively it's 4.577. If you're willing to work with the company's
engineers, you can also get the pco.edge to indicate when all lines are
exposing, vs if any lines are not exposing.

 We often use python to control our hardware, and I've written some code to
talk to the pco.edge:
http://code.google.com/p/msim/source/browse/hardware_control/pco.py
 MY CODE IS NOT ROBUST CODE, IT CONTAINS KNOWN BUGS, so there isn't much
here of value to the non-programmer. However, it's also the code we've used
for several of our papers, so it's not worthless, either. For other python
nerds, it might have some useful clues how to get the camera to perform at
high frame rates. The code is open-source (GPL), so feel free to use, copy,
and modify, as long as you keep it open-source.

 Overall we're extremely happy with the pco.edge, but to be clear, this is
currently a camera that rewards the careful user, and can punish the casual
user. A few things we learned are described in Note 3 of this document:
http://www.nature.com/nmeth/journal/v9/n7/extref/nmeth.2025-S1.pdf
I suspect the company has since addressed some of these issues, so check
with them if any of these points sound like dealbreakers.

 Hope someone finds this useful.

On Sun, Aug 12, 2012 at 3:44 PM, Shalin Mehta <[hidden email]>wrote:

> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> *****
>
> Hi Michael,
> We have demoed sCMOS cameras from Hamamatsu, Andor, and PCO.  I found
> that Flash 4 has better sensitivity as compared to Andor or PCO's
> cameras. I have been told by Hamamatsu's representative that the
> Micro-manager driver is very useable, but it does not handle the top
> frame-rate of Flash 4.
>
> We are still undecided and I need to understand a couple of issues
> before we make the decision.
>
> For our target application (quantitative imaging and analysis of fast
> moving specimens) the rolling shutter-only operation of Flash 4 is an
> issue. To investigate the performance of global vs. rolling exposure,
> we used Andor's Neo with the same settings in either mode. Here is a
> dark-field image of an intact sea urchin sperm beating at about 30
> waveforms per second:
> http://www.mshalin.com/interesting-data/GlobalvsRollingShutter.png
>
> From the images and intensity profiles we see that rolling shutter
> introduces motion artifacts such as skew in the shape of head and
> flagellum waveform in vertical direction. But, rolling exposure has
> visibly better sensitivity and SNR.
>
> I still don't understand the following completely. All inputs from
> community and manufacturers will be appreciated.
>
> a) How does global exposure work in Andor and PCO cameras? What
> sources of noise have increased contribution when doing global
> exposure?
>
>  b) For any camera in rolling shutter mode, we can achieve global
> exposure by strobing the light-source (such as LED or laser) so that
> * the specimen is illuminated right after all the pixels have been
> rolled active
> * the source is turned-off right before the pixels are read out in
> rolling fashion
>
> The cameras take about 10 ms 'to roll the full chip'. So if we set the
> exposure to desired global exposure + 20ms and strobe as above, it
> seems we again have global exposure. Will noise be any better with
> this custom arrangement?
>
> Thanks,
> Shalin
> website: http://mshalin.com
> (office) Lillie 110, (ph) 508-289-7374.
>
> HFSP Postdoctoral Fellow,
> Marine Biological Laboratory,
> 7 MBL Street, Woods Hole MA 02543, USA
>
>
> On Sun, Aug 12, 2012 at 1:37 PM, Cammer, Michael
> <[hidden email]> wrote:
> > We demoed an Andor Neo camera a few weeks ago and was very impressed
> with it.  We couldnt quite get the sensitivity needed for single molecule
> imaging but were very impressed by its other features.  We'd like to get a
> camera with these capabilities that we may use both with Nikon Elements on
> one microscope and with MicroManager on another (currently one microscope
> has a DU897 and we want ta bigger field and wider dynamic range of view for
> some applications and the other has decade old Orca ER).
> >
> > Two part question.
> >
> > 1.  Is anyone developing a driver for MicroManager that will take
> advantage of this camera's many many features?
> >
> > 2.  While we really liked the Andor Neo, does anybody have data or
> feelings that other sCMOS cameras out there are better-- keeping in mind
> that drivers for both software packages is also important.
> >
> > Thank you!!
> >
> > _________________________________________
> > Michael Cammer, Assistant Research Scientist
> > Skirball Institute of Biomolecular Medicine
> > Lab: [phone]  Cell: [phone]
> >
> >
> >
> ------------------------------------------------------------------------------
> > Live Security Virtual Conference
> > Exclusive live event will cover all the ways today's security and
> > threat landscape has changed and how IT managers can respond. Discussions
> > will include endpoint security, mobile security and the latest in malware
> > threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
> > _______________________________________________
> > micro-manager-general mailing list
> > [hidden email]
> > https://lists.sourceforge.net/lists/listinfo/micro-manager-general
>

[Stephanie Fullerton]

Hi Shalin,

Thank you for taking the time to consider our camera. It’s great to hear that the Flash4.0 gave you increased sensitivity. While this may be a surprise to some readers, we would expect this since our camera uses a Gen II sCMOS sensor with higher QE. To achieve this higher QE, we specifically designed the sensor with a 4T pixel structure, thereby sacrificing global shutter readout mode. But, fear not! All is not lost for high speed imaging applications such as yours. As you allude to in your post, there is another way to achieve “snap-shot” imaging with a rolling shutter CMOS sensor… this is called global exposure synchronization. Below is an excerpt of text from page 17 of our recently released “Hamamatsu Life Science Cameras Brochure, Summer 2012.” This describes the difference between global shutter readout mode (which has higher dark current and read noise and slower frame rates) and global exposure synchronization for rolling shutter. The entire brochure can be found at http://www.hamamatsucameras.com/cameras.php.

How We Accomplish Global Exposure

While the vast majority of experiments work perfectly with rolling shutter, there are a handful of protocols in which the signal from all pixels must be captured at exactly the same time. Of the two ways to accomplish this timing in CMOS cameras, Hamamatsu has implemented the solution which offers the fastest acquisition and best image quality possible.

Global shutter

A CMOS pixel read out mode that allows all pixels to be exposed at the same time. This mode requires each pixel to have an additional transistor. The complexity of this pixel design and its operation lowers the QE and increases dark current. Furthermore, for quantitative data, a reference frame must be acquired with each image, halving the effective frame rate and increasing the read noise by 1.4x. Because of these tradeoffs, Hamamatsu sCMOS cameras are designed without global shutter type sensors.

Global exposure synchronization

A method of driving a rolling shutter CMOS so that there is a time when all the lines are exposed simultaneously, thus emulating a global shutter CMOS. This acquisition option is built into our ORCA-Flash series of cameras. To enable global exposure the camera is synchronized with a pulsed light source or fast external shutter. In the camera, the exposure window is expanded to be slightly longer than time the light source is on. This makes possible a slice of time in which all lines are receiving light simultaneously. The precise hardware triggering is handled in software, allowing the user to easily achieve the same temporal synchronization as global shutter while still offering low noise, high QE and fast frame rates.

Direct answers to your questions Yes. It is possible to use a strobed light source to expose all lines at the same time. Yes. Using rolling shutter with global exposure sync and strobed light will avoid all the “negative side-effects” of global shutter, plus provide all the benefits of the high QE Flash4.0. This combination offers the best SNR performance for your application.

It is my understanding that you spoke with Megan, our regional rep this week and I’m confident we can work out the details for your application, as they will be specific to your samples and your analysis goals.

For those of you reading this post, still wondering if rolling shutter is an issue for your application, please check back with us in a week or so. We will present a detailed look at rolling shutter and how to determine if it matters for you: “Jello, Goodbye: Resolving the Myths about Rolling Shutter in Microscopy”. http://hamamatsucameras.com/assets/pdfs/MythsAboutRollingShutter.pdf

Sincerely,

Stephanie Fullerton

Manager, Hamamatsu Camera Products Group

[Ian Dobbie]

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

Stephanie Fullerton <[hidden email]> writes:

> surprise to some readers, we would expect this since our camera uses a Gen
> II sCMOS sensor with higher QE.  To achieve this higher QE, we specifically
> designed the sensor with a 4T pixel structure, thereby sacrificing global
> shutter readout mode.  But, fear not!  All is not lost for high speed

This is a bit of a stretch of the facts. The Hamatasu sensor is made by
Fairchild imaging like the Andor and PCO sCMOS sensor and is not
fundamentally genII in any way. Yes it has slightly higher QE and lower
noise by eliminating a transistor at the loss of the global shutter
mode.

The rumor mill (ie several people have told me this but I haven't tested
this so I believe it to be true but it may be wrong) says that the
global shutter mode has significantly more noise than the extra 2^0.5
expected from theory. We have some PCO sCMOS cameras but with our
software/hardware combination we cannot run in global shutter mode so I
cannot test this in person.

Ian
--
Ian Dobbie
Micron Imaging Unit Manager,
Biochemistry,
University of Oxford,
South Parks Road,
Oxford
OX1 3QU
Tel: [phone]
Email: [hidden email]

[Shalin Mehta]

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

Hi everyone,

I concur with Ian about read noise in global shutter based on very
helpful explanation by Andor's camera system engineer (Geoff Martin).
The spec sheets do account for this. My understanding is: with on-chip
global shutter there is reset noise (there is slight fluctuation when
all pixels are reset), extra read noise, and some additional dark
noise because the charge sits on the sensor for a short (but
exposure-independent) time before being read out.

Based on the inputs from manufacturers, I did some careful reading of
spec sheets to estimate noise sources when cameras are operated in
on-chip global exposure or rolling-with-strobe exposure.
I have posted what I understood about sources of noise and SNR
analysis  on my blog:
http://www.mshalin.com/blog

If I am mistaken in reading some of the noise figures, corrections are
appreciated.
It seems that rolling-with-strobe does slightly better than on-chip
global exposure in terms of SNR at low light. Hamamatsu's higher QE
does look attractive.

Best
Shalin
website: http://mshalin.com
(office) Lillie 110, (ph) 508-289-7374.

HFSP Postdoctoral Fellow,
Marine Biological Laboratory,
7 MBL Street, Woods Hole MA 02543, USA



On Thu, Aug 16, 2012 at 7:36 PM, Ian Dobbie <[hidden email]> wrote:

> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> *****
>
> Stephanie Fullerton <[hidden email]> writes:
>
>> surprise to some readers, we would expect this since our camera uses a Gen
>> II sCMOS sensor with higher QE.  To achieve this higher QE, we specifically
>> designed the sensor with a 4T pixel structure, thereby sacrificing global
>> shutter readout mode.  But, fear not!  All is not lost for high speed
>
> This is a bit of a stretch of the facts. The Hamatasu sensor is made by
> Fairchild imaging like the Andor and PCO sCMOS sensor and is not
> fundamentally genII in any way. Yes it has slightly higher QE and lower
> noise by eliminating a transistor at the loss of the global shutter
> mode.
>
> The rumor mill (ie several people have told me this but I haven't tested
> this so I believe it to be true but it may be wrong) says that the
> global shutter mode has significantly more noise than the extra 2^0.5
> expected from theory. We have some PCO sCMOS cameras but with our
> software/hardware combination we cannot run in global shutter mode so I
> cannot test this in person.
>
> Ian
> --
> Ian Dobbie
> Micron Imaging Unit Manager,
> Biochemistry,
> University of Oxford,
> South Parks Road,
> Oxford
> OX1 3QU
> Tel: [phone]
> Email: [hidden email]