Emerald 8M9 features a 2.8µm global shutter pixel and provides a 8.9MP resolution in a 2/3-inch optical format. The sensor is available in two speed grades: a standard speed model (47fps @10bits) and a high speed model (107fps @10bits). The new sensor has a readout noise of 2.8e- combined with 65% QE.
Vincent Richard, Marketing Manager at Teledyne e2v, said “Emerald 8M9 is designed specifically to address the demands of machine vision, high resolution surveillance and traffic intelligence. The sensor is unmatched in the industry because of its versatile feature set. For example, real-time High Dynamic Range mode allows high resolution capture of fast moving situations from daylight to night-time with minimum artefacts and blur effects.“
Samples and demo kits are now available and mass production is planned for Q1 2019.
I dont understand arguments for such a wide nearly 2:1 format field of view. Optics have a circle inside which they are optimized/usable. A 2:1 rectangle covers a far smaller area inside this circle than a square. Machines dont need to produce wide screen movies for people but measure/inspect/act. But maybe its different in "traffic intelligence" markets. In the robotics/machine vision applications I know, a square field of view offers the best match to the operating area of optics.
ReplyDeleteThe second point is the large package and the non-symmetrical placement of the die. The active area is maybe 11.5x6mm, the package to me seems to be something like 24x24. If you place the optical center in the cmount center, a 29x29mm camera package will be hard to realize. But also here - maybe this is not so important in the main target market of the sensor...
Good remark on the non-symmetrical placement of the die vs. 29x29mm camera housing...will be a nightmare.
DeleteI'm also puzzled by the 2:1 aspect ratio for machine vision...I agree with your reasoning ...for ITS I can imagine 2:1 makes sense (~ capture multiple lanes).
The sensor size appears to be 20mm x 21mm (according to their public flyer). So a 29x29 housing should be possible, but I guess you will need a quite low power camera electronics to avoid serious heat issues... (The sensor alone has up to 1.6W at full speed)
DeleteRegarding the aspect ratio, I guess it is simply market demand driven (e.g. I for my side prefer 16:9 or 4:3). People that really need a square image can always leave part of the sensor unused... I assume the volume for square sensor applications is too low to justify the dedicated sensor development.
people that really need 4:3 can leave part of a square sensor untouched ;-) not reading lines even brings readout time benefit while not reading colums usually doesnt
Deletei like cmv4000 or gmax0505 in this respect: square and little package overhead
About aspect ratio:
ReplyDeleteAs someone pointed out in another post's comment, most displays sold seem to have an ever widening aspect ratio, also the underlying PCB's follow this trend.
The OP further explains this sensor is aiming also at surveillance, where maybe the sky is of little interest. In ITS a wide aspect ratio helps seeing multiple lanes with 1 sensor.
Nevertheless, I follow your second point about the non-symmetrical die placement, which seems to be a result of the quite large single-sided readout under the pixel array - not sure what magic happens there...
The resolution of 4k by 2k is needed for 4k cinema, broadcast markets, visualization. If you need classic aspect radio of 4:3 in MV or 1:1 for microscopy, you can choose the larger Emeralds with 12, 16 or 67 Megapixel. This sensor here is one member of 4 different resolutions. And each sensor has its own market by application.
ReplyDeleteThat the optical center is not the sensor package center is caused mainly by the ADC structure, clearly visible. There are other sensors on the market, with split into top and bottom circuits, but this is trickier.
E2V advertizes it for machine vision application an traffic monitoring. Check the link above. I don't see anywhere cinema or broadcast...
Delete