e2v announces the
Onyx EV76C664 1.3MP global shutter CMOS sensor. The new sensor features an embedded active imaging system which synchronises external illumination and internal electronic shuttering. This allows discernible color images to be produced at extremely low light levels and through diffusing environments (e.g. fog and smoke) by range-gating, making it ideal for Intelligent Traffic Systems (ITS) and security cameras.
The sensor has an innovative 10um-large pixel with extended sensitivity and performance in NIR, and can also cater for ‘all-light’ environments (typical of outdoor camera applications) where wide dynamic range is also needed. The new sensor is available in sparse colour and monochrome filter (EV76C664ACT) and black and white (EV76C664ABT).
Is that a typo? 5e-rms in global shutter with delta-double sampling? From SNRmax I get ~10ke- FWC. Assuming 1V swing I end up with 16e-rms read noise contribution from reset noise not being cancelled by uncorrelated DDS... I guess the description for 5e-rms and 17e-rms is mixed up?
ReplyDeleteThe 18e- rms are true for the GS. It seems to be a 5T pixel cell similar to the ev76c560 sensor. CDS only in RS mode.
ReplyDeleteWith two images captures and DDS in an external memory, you can archive 6e- rms.
Datasheet says in 6.3 page 9:
The Digital Double Sampling (DDS) is a function similar to the Correlated Double Sampling (CDS) but done in the digital domain
(after the signal is converted by the ADC) rather than in the analog main (before the ADC). It allows a readout noise reduction
especially when the Global Shutter mode is used. The sensor outputs 2 consecutive images: a first image with the all the
pixel reference levels followed by an image with all the pixel
signal levels. Subtraction product of these two images will generate a low readout noise global shutter image. An off - chip image processing with external memory will be required to perform this subtraction.
OK, understood. Thanks for the explanation. What confused me appeared to be the abbreviation DDS which is also known as "delta-double sampling". This technique, conversely to digital CDS, does not remove reset noise but only offset FPN and will, as explained above, result in a higher noise level. So thanks again for the clarification.
DeleteAnd what about the 1/f noise ? Between the two signals you need for the digital CDS as you suggest, a relative long time is needed and the 1/f noise will become more than important ...
DeleteNot only 1/f, but also FD leakage. So question is - at which frame-times and/or integration times are those 5e-rms achieved?
DeleteThis technique has been used since a decade or so. They must have mastered it. E2v is not known as a liar company...
DeleteNobody claims anybody to be a liar. We just raised technical questions - which is what this blog is about, no? Technical discussions between imaging people. Technically, those aspects will have impact on the read noise and thus it is interesting to ask under which constraints this performance can be expected. These questions preferably should be answered with numbers and not with criticism...
DeleteIf you have a motion bright object, the image will be amazing.
ReplyDeleteWhy would this sensor be better than SONY IMX250, which has twice more resolution?
ReplyDeleteDesperately no breakthrough on global shutter technologies...
ReplyDelete