Business Wire: ON Semi/Aptina's
AR0230CS sensor is a 1/2.7-inch device with 105dB DR combining low-light capabilities with superior NIR sensitivity for security cameras. It supports HDR video with improved motion compensation, advanced local tone mapping (ALTM), and digital lateral overflow (DLO) to minimize motion artifacts and enhance HDR color reproduction. AR0230CS' dual-conversion gain 3.0µm DR-Pix pixel has two modes: a low conversion gain mode that is optimized to offer greater charge handling capacity in bright environments and a high conversion gain mode which delivers increased sensitivity and reduced readout noise in low light environments.
“
With demands for higher resolutions, the growth in professional and consumer 1080p security cameras has been tremendous and the 2.1 megapixel, 1/2.7 inch AR0230CS sensor is well positioned for this market,” said Alvin Wong, senior director of automotive and industrial CMOS imaging products, ON Semiconductor. “
This sensor’s extensive dynamic range, superior low light performance and NIR sensitivity set it apart from other imaging solutions on the market.”
The AR0230CS is currently sampling and is expected to be in mass production in Q1 2015.
The Press Release, and a Sample Image are here:
ReplyDeletehttps://www.aptina.com/news/press/on_semiconductor%E2%80%99s_new_2.1_megapixel_cmos_image_sensor_delivers_high_performance_1080p_hdr_video_for_security_cameras/ .
https://www.aptina.com//AR0230CS_Press_12914.jpg
Image Title: "ON Semiconductor’s AR0230CS image sensor for security cameras provides amazing near-infrared performance.".
Thanks! We can see a lot of transition FPN on the image. It's totally uselss.
ReplyDeleteAny transition fixed pattern noise would indeed be difficult to correct.
DeleteA bit of sharpening on that Image does indeed bring on the dreaded Tartan pattern in areas that are supposed to be one color, and the 'HDR Transition' (banding) is badly mottled. The haloing and Swiss cheese in the banding is also unwanted.
Running JpgQ ( http://www.mediachance.com/digicam/jpgq.htm ) on the Jpeg Image says it was compressed with a Q of 89. When I use Q 98 and sharpen a few times (on my own Images) I can see noise, so they ought not to have used as low a value as 89, and would have been better off with a Png compressed file.
This Sensor is destined for a 'Security Camera' or 'Automotive DVR' Application, it's not sCMOS nor particularly expensive, even it's 40 dB lower DR is a lot less than some of the newest Sensors.
The Application for the Sensor is not precise measurement, most likely (with 1080P HDR @ 60 FPS) they are trying to get a decent Image of License Plates and Faces in poor lighting conditions (so a hair out of place is not the issue).
A further point is that we do not know which of the two DR-Pix modes was used, hopefully the relevant one was used (despite the poor choice of using jpeg compression).
The AR0230 Flyer does not quote a SNR, the AR0130 claims: "Dynamic Range: 83.5 dB
SNRmax: 44 dB" (which is low, for my liking, on both points). High DR with low SNR does not make for the prettiest picture.
So while the AR0230 may not be 'the greatest' this is the "News" - that OnSemi produces new Chips after it's Aptina Acquisition, and this is one of them. Very useful for making Cameras with a Parts Bill of less than $100 (that is it's use).
The Omnivision OV10640 (which is in a similar Application Class as the AR0230) produces slightly lower resolution Video, in color, with greater (smooth) dynamic range.
In this Video you can see that the Image Processing Chip (which the AR0230 does not have) does not get 'fooled' by changing light conditions and virtually always produces a really nice color HDR Image in all lighting conditions (even with dark and bright in the same Scene).
https://www.youtube.com/watch?v=1ruotjfo6XM&list=UUUH9y35fQqOKZQwmwOGYR0g
44dB SNR for a 3u pixel is low for you? Are you kidding me?!?!? As of the image link you put, it is just a compressed web image. Come on guys, download a small size image and scale it to a full hd monitor, of course you see pattern noise. Would be nice from Aptina to upload the full resolution image just to avoid the little confusion here.
ReplyDelete44dB SNR => 25K FWC if the read noise is 1e. For a 3um pixel, is this possible?
DeleteSNR max=20*log(sqrt(FWC)) if you correct the PRNU. At high light you are photonshot noise limited at sqrt(FWC)= 158e noise. So, no need for 1e read noise to have high SNRmax.
DeleteI think that engineers give this image under the presdure of marketing guys.
ReplyDeleteWhat about color imaging, it would be nice to have a color image sample to check the "HDR transition".
ReplyDelete