Re: CMOS Camera - Commercial Reponse

[DENNIS Andrew]

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

Hi David, Prabhakar, Julio,
I work at Andor and can explain what's going on here. The Andor/PCO version of the Sensor is a 5.5MP sensor, and the Hamamatsu sensor is 4MP (they don't limit the number of useable pixels or use a mask). In addition to the number of pixels, these sensors differ in only one way...

The Andor version has a 5 transistor pixel design (5T design), this permits both Rolling shutter* readout mode (a fast, but transient mode of readout) and Global Shutter* readout (AKA snapshot or interline mode).

The 4MP sensor has one less transistor on each pixel (4T design), the result is two fold, 1.) there is a positive effect, an increase in QE (due to lack of the transistor circuit on the sensor pixel). And 2.) there is a negative impact, this sensor variant is limited to rolling shutter readout (transient readout only), so this variant does not permit global shutter /Snapshot mode.

The 'Gen II' label used to describe the 4MP/4T variant is just a label, the pixel design is identical, with the exception of the missing transistor.

In addition to appreciating the various modes of sCMOS readout the QE must always be considered in context of noise. As David comments CMOS sensors suffer from hot pixels, (more than CCDs), and there is a different distribution of noise across CMOS pixels compared with CCD. As you cool the sensor many of the hot pixels are "fixed" so less need to be corrected, the colder it is, the more pixels are corrected. If you have any further questions please feel free to ask me directly or through the listserv.

Andrew
[hidden email]

*Brief explanation of Global and Rolling shutter:
Global Shutter: the photoelectrons in each pixel are simultaneously moved into readout circuitry, where they are measured. This is a traditional shapshot acquisition and is very analogous to an Interline sensor readout.

Rolling shutter: the readout takes place as a wave traversing the sensor and travelling form a line across the centre of the sensor towards the outsides. The result is that the time of acquisition is different at different portions of the sensor. This can be avoided by switching off the light source during readout, but this has an obvious impact on data acquisition rates.

-----Original Message-----
From: Confocal Microscopy List [mailto:[hidden email]] On Behalf Of David Baddeley
Sent: 27 June 2012 02:10
To: [hidden email]
Subject: Re: CMOS Camera

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

Was just about to say the same thing - as far as I know they use a different chip (PCO and Andor both use the same chip, but Hamamatsu wasn't part of that consortium and decided to go it alone). The Gen I vs Gen II is probably mostly marketing hype ... In addition to QE I'd also pay attention to hot pixels and pixel to pixel gain/noise variations, as these tend to be the limiting features of sCMOS technology at the moment (you might get a mean read noise of ~ 1 e-, but a handful of pixels have much larger read noises). The manufacturers tend to offer soft/firmware solutions to mask these noisy pixels (often replacing the value by interpolation from neighbouring pixels), but whether this is desirable or not will depend on the application (it's one of the major concerns when evaluating sCMOS for PALM/STORM applications).

cheers,
David


________________________________
 From: Julio Vazquez <[hidden email]>
To: [hidden email]
Sent: Wednesday, 27 June 2012 12:11 PM
Subject: Re: CMOS Camera

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

Hi Prabhakar,

Do you know for a fact they use the same chip?  The Andor sCMOS camera's spec sheet reports 5.5 MP (2,560 x 2,160 pixels) while Hamamatsu reports 2,048 x 2,048 effective pixels for the Flash 4. Could still be the same chip but why would Hamamatsu waste 1.5 MP? Also, the QE curve of the Andor resembles the QE curves of what Hamamatsu calls "Generation I" sCMOS, while the Flash 4 uses a "Gen II" chip with higher QE.... I'd suggest asking the respective reps. Maybe Andor is just a bit behind and will release an sCMOS with 70% QE in the next few months, or they chose a different chip for some other reason...

--
Julio Vazquez
Fred Hutchinson Cancer Research Center
Seattle, WA 98109-1024


http://www.fhcrc.org/




On Jun 26, 2012, at 2:17 PM, B. Prabhakar Pandian wrote:

> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> *****
>
> Hello All,
>                We are planning on purchasing a CMOS camera from either
>Andor (Neo) or Hamamatsu (ORCA Flash 4). We are going to be demo them  before deciding. However, looking at the specs, ORCA claims 70% QE vs 55% for Neo. Since both of them have the same chip, is this information accurate.
> Any help in resolving this question is appreciated.
>
> Thanks,
>
> -Prabhakar

[http://www.andor.com/newsletter/footer/sig.jpg]<http://www.andor.com/newsletter/footer>

[Stephanie Fullerton]

Hi All,

Remarkably, George Foreman named all 5 of his sons George, starting with George Jr., and then moving down the line to George III, IV, V and VI.  To differentiate among his sons they all have nicknames, like “Monk” and “Big Wheel.”  Perhaps George did this as a marketing scheme to gain attention (and sell his grill). Or maybe he just really likes the name George.  In either case, the problem that he encountered is how to specify which son he means when he calls “George!”  

We encountered a similar problem in how to differentiate our sCMOS sensor from the other, older sensors available on the market; we chose to use “Gen II” which is in line with how other progressive technologies, such as gene sequencing and night vision sensors, have indicated significant leaps.  The fact of the matter is that our sensor is not the same sensor in the Neo, Zyla or Edge.   Actually, the "Gen II" moniker isn't hype so much as an absence of it since the Flash4.0 is unique in the following ways:

1. Higher QE.  Our pixels do indeed have a 4T (versus 5T) structure that delivers higher QEs. Our peak QE at 550nm is 72%.   Andor’s Neo and Zyla and PCO’s Edge top out at 57%.  Our high QE means that we have better SNR at all light levels.  Consider that it requires 26% more exposure time (0.72/0.57) to collect the same SNR with the Andor/PCO camera compared to the Flash4.0.
2. Rolling Shutter.  A 4T structure is only consistent with rolling shutter operation and this was a considered design choice.  Early in the development of these cameras, we recognized that global shutter has some major downsides including reduced frame rates (by 0.5x) and higher noise (by 1.4x).  We also recognized that few imaging applications involve the combination of large, fast-moving objects and short exposure times where rolling-shutter might be an issue.  
3. Improved Sensor Design.  Our sensor design, including the masks, is new and different from the Andor/PCO sensor.  It is more current and builds on the collective advancements the CMOS industry has made in pixel design and manufacturing.  Since every CMOS pixel has an individual amplifier that converts the photoelectric charge to voltage and column-parallel ADCs that convert the voltage into an output digital number, the uniformity of the output critically depends on the design and production process.  From our first sCMOS camera, the Flash2.8, to the current Flash4.0, we have made pixel uniformity, in terms of absolute response and read noise, a top design priority.
4. Hamamatsu’s experience.  A sensor’s performance is highly dependent on how it is implemented in a camera.  Our unique and improved sensor has the added advantage of being engineered into a camera by Hamamatsu scientists’ foremost in our minds.   The result is a camera that delivers minimal pixel gain variation (i.e. no stripes!), a user-switchable, real-time, FPGA-embedded hot-pixel correction and sustained (minutes!) of full-field streaming at 100fps in addition to the low read noise, high QE, high dynamic range and the robust performance of all Hamamatsu scientific cameras.  

Bottom line… our George is not the identical twin of George or the other George.  Perhaps we could have come up with a fancier name than “Gen II” that conveys the uniqueness of Hamamatsu’s latest technology.  But we had other priorities...  See our “Changing the Game” white paper and our collection of work related to Super Resolution [http://www.hamamatsucameras.com/flash4/index.php] -- our cameras are optimized for emerging applications where every photon counts.  And that's by design, not hype.

Many thanks to all for thinking about and discussing this very interesting topic.

Sincerely,
Steph

Stephanie Fullerton, Ph.D.
Manager, Hamamatsu Cameras