CMOS binning and IMX249 vs ICX829 comparison for low-light imaging

Post Reply
User avatar
Hibou
Posts: 81
Joined: Thu Jul 05, 2018 3:25 pm
Location: French Alps
Contact:

CMOS binning and IMX249 vs ICX829 comparison for low-light imaging

#1

Post by Hibou »

On-chip binning is unavailable for CMOS chips, and because pixels are often small for high resolution, intensity can be insufficient (and resolution excessive) for low light imaging. Off-chip binning can then be useful to improve signal/noise. Isolated hot pixels should first be removed using eg Process/Noise/Despeckle in https://imagej.nih.gov/ij/ and then binning can be achieved with Image/Transform/Bin with either a single image or a stack.

The Sony CMOS IMX249 (FLIR Blackfly camera) can then be binned 2x2 to reduce the resolution to 960x600, comparable to the 752x580 resolution of the Sony CCD ICX829 (SX Lodestar-X2). Both cameras were mounted in identical black boxes with f/1.2 (IMX249) and f/1.0 (ICX829) lenses and 200ms exposures of a luminous disk compared. With 2x2 binning, the image obtained with the IMX249 http://neutronoptics.com/resources/IMX2 ... -2xbin.jpg has similar intensity to that obtained with the ICX829 http://neutronoptics.com/resources/ICX8 ... -2xbin.jpg.

In fact the larger CMOS chip is a little brighter, given that a larger but less efficient f/1.2 lens was needed for the larger chip. Of course dark current noise is very much higher for the CMOS chip (hot pixels filtered out), making it suitable for shorter exposures than with the CCD. (This was only a "quick and dirty" qualitative test).

It would be good to have real-time hot pixel filtering combined with off-chip binning in CMOS capture applications such as FlyCap and SharpCap.
User avatar
oopfan
Posts: 1321
Joined: Sat Jul 08, 2017 2:37 pm
Location: New York
Contact:

Re: CMOS binning and IMX249 vs ICX829 comparison for low-light imaging

#2

Post by oopfan »

I know that my lowly ZWO ASI120 has 2x2 binning available from the SharpCap user interface. Many astrophotographers use the ZWO ASI1600 for deep-sky, some of them are members of the forum. Here is an excerpt from the User Manual of the ASI1600:
The ASI1600 camera supports hardware and software bin2, bin3 and bin4 mode. Hardware
binning is supported by sensor but is done in digital domain like software binning and use 10bit
ADC. The only advantage of hardware binning is faster fps. We recommend customer to use
software binning if you don’t care speed
I would suggest contacting Sam at ZWO for a better explanation of what that actually means. I have seen many photos on AstroBin with the monochrome version using 2x2 binning in the RGB channels.

Not all cameras have binning. I love my Altair 290M but I am annoyed that it doesn't have binning nor a temperature sensor. Unfortunately you really need to dig deep to find this info before purchasing. Everyone in the supply chain does an awful job of adequately informing the consumer.

Brian
User avatar
Hibou
Posts: 81
Joined: Thu Jul 05, 2018 3:25 pm
Location: French Alps
Contact:

Re: CMOS binning and IMX249 vs ICX829 comparison for low-light imaging

#3

Post by Hibou »

I guess some CMOS cameras like ZWO may have software binning built in, but hardware (on-chip) binning before readout is a unique advantage for monochrome CCD technology. Atik have a nice explanation https://www.atik-cameras.com/news/binnn ... s-and-ccd/ and Robin gives more details on viewtopic.php?f=35&t=262. FLIR (Pt Grey) talk about CMOS binning in "mode 7" readout, but I have yet to understand how to implement that, and in any case it is best to filter isolated pixel noise before binning. Noise "reduction" using blurring makes it difficult to filter hot pixels later and is not advisable. Yes, it is better to collect raw data and correct off-line, and yes you can make signal/noise arguments for not binning, but often you want to be able to see real-time images, even if they are not optimal, so hot pixel filtering followed by optional software binning would be useful IMHO.
User avatar
admin
Site Admin
Posts: 13173
Joined: Sat Feb 11, 2017 3:52 pm
Location: Vale of the White Horse, UK
Contact:

Re: CMOS binning and IMX249 vs ICX829 comparison for low-light imaging

#4

Post by admin »

Hi,

yes, several brands of camera have CMOS software binning built in - either in the driver software running on the PC or in the firmware of the camera so that the binning happens before the image is downloaded to the PC ( the latter is called 'hardware binning' by the manufacturer!).

CCD hardware binning (adding electrons from adjacent pixels pre A->D conversion) can best be seen as a way to deal with high CCD read noise by increasing the signal and hence the S/N ratio. Because of the high CCD read noise, CCD style binning carries a S/N advantage over software binning. Since modern CMOS cameras can achieve much lower read noise figures there would be much less advantage to having CCD style pre-readout binning to the point of any advantage being marginal - I think this means that there is little likelihood of this feature being added to CMOS cameras in the immediate future.

Interesting point about hot pixel removal being best done pre-binning - makes sense.

cheers,

Robin
User avatar
Hibou
Posts: 81
Joined: Thu Jul 05, 2018 3:25 pm
Location: French Alps
Contact:

Re: CMOS binning and IMX249 vs ICX829 comparison for low-light imaging

#5

Post by Hibou »

Thanks for confirming that. For me, binning is what you do to trade resolution for intensity. I guess that's why some CMOS cameras emulate CCD binning in firmware/software. But you might do better by first filtering hot pixels prior to CMOS software binning.

Actually, there is not a huge difference in read noise between modern CCD (ICX825) https://eu.ptgrey.com/grasshopper3-14-m ... ony-icx825 and CMOS (IMX174) https://eu.ptgrey.com/grasshopper3-23-m ... ius-imx174 cameras. Both have ~7e- read noise (FLIR call it temporal dark noise), with the ICX825 also found in the Atik Infinity and SX-superstar. Yes, some FLIR CMOS cameras have less than half that, but the difference is not an order of magnitude, unlike the difference in dark current noise, which becomes important for longer (or stacked) exposures.

Cheers, Alan.
Post Reply