Electronista and Crunchgear quote subscription-only Nikkei article saying that Sony has developed a new 17.7MP sensor that can "convert multiple pixels into signals simultaneously", which results in a reduction of the conversion time by up to 75% when compared to existing sensors.
The new sensor boasts a processing speed of 34.8Gbps (that’s up to five times faster than previous Sony sensors) and shoots video with 17.7MP at 120fps. Apparently, Sony also managed to keep power consumption in check by revamping the circuit design in the sensor.
The Nikkei also says that Sony plans to use the new technology in digital cameras and cell phones in the future, without providing information on when exactly this will be the case.
this looks like the titanium based curtain shutter in SLR camera at 1990s. High speed readout will remove virtually all the obstacles in consummer imaging devices.
ReplyDeleteDoes "convert multiple pixels" mean column-parallel ADC or an additional level of parallelism?
ReplyDeleteWhile this is fast, and in consumer-land Sony seems to be the technology leader, Alex Krymski and team at Photobit did a 200 fps 4Mpixel sensor in 2000-2001 time period. (PBMV40). Since then much higher throughput than Sony have been realized by Cypress, for example. With that in mind, the Sony speed achievement is not so impressive to me.
On the other hand, I am glad Sony has brought this advanced technology to the consumer level, presumably at consumer electronics prices. That contribution IS impressive!
Another plug for Fossum's work by Fossum. Photobit did something and Cypress did something. Were those really high quality products? Sony is not just playing in the sandbox.
ReplyDeleteReally troll? Read the post again. I think it mentions Krymski and Cypress. And calling their work "sandbox" and implying it was not high quality is idiotic and insulting to those researchers. Grow some huevos and tell us what you have done besides fawning over Sony.
ReplyDelete>>"Does "convert multiple pixels" mean column-parallel ADC or an additional level of parallelism??
ReplyDeleteIt sounds like pixel binning. It seems to be a technology that is gathered a lot of demand for CMOS sensors as the consumer market has moved their interest to video. Obviously, CMOS pixel binning presents its own set of challenges.
Handling 120 frames of 17.7MP video is not very feasible for most people. So there needs to be an efficient method that can lower that resolution for different consumption.
Also, If you are referring to Cypress' sensor that they announced as the "world's fastest" it had a digital data throughput of 13.2 Gbps. This sensor is suppose to be 34.7 Gbps.
Either way, speed isn't everything. If this is meant for consumer and professional imaging equipment, then there is a lot more that needs to go into the sensor than outright speed. Consumers are fairly critical of the various nuances of image quality. That Cypress sensor for instance only had 8-bit ADC, which helps in speed, but is unsuitable for cameras that this sensor is suppose to inhabit.
Eric, I may be better if you post anonymous. It is not helpful to post your name.
ReplyDeleteWe have to give credit to Eric R Forssum and his Photobit team for their amazing work and allowing us to see current generation of imaging.
ReplyDeleteBut surely this bring a philosophical aspect of life that power and fame corrupts even the most humble person on this planet.
Good points Francis.
ReplyDeleteOther troll, 16:54 should just be deleted by ISW as a personal attack.
I envy Eric for his confidence with his own skin. He does present somewhat controversy points sometimes but everyone has their own perceptions. I applaud him for exercising the 1st with confidence. The others who wants to present their own controversy points need to grow up. It is easier to be polarizing from behind the curtain. Man up people when you wanna attack someone. Attacking someone anonymous is the same as backstabbing. Face your opponent.
ReplyDeletecan't say that a troll is idiotic or even call the troll a troll and then ask to be relieved of a personal attack. Fossum can post anonymously like everyone. No applaud or envy needed here.
ReplyDeleteIt is a shame when one troll can steer an entire thread. It is an immature exercise of free speech to personally attack someone by name by making false accusations and innuendo. If you say something untrue enough times, it raises doubt in people's minds. This is the principle of disinformation. It is widely practiced in ugly politics.
ReplyDeleteAnyway, let's not let the troll steer this forum and instead let's talk about current news, technology and when appropriate, past technical history.
If you agree with me, please respond!
the thing i like about Dr. fossum is that he's able to draw on his own vast experience and communicate it effectively so that a non-engineer like myself can get maximum benefit from his analysis. one of the reasons i like this blog so much besides vlad canvassing the internet for excellent articles and his input on them, is that Dr. fossum is here to comment on them.
ReplyDeleteDr. fossum, when you've done as much as you have in the cis space, you're bound to have jealous people attack you. it's just an irrational cost of your status. please continue to post like you have, i'd really appreciate it.
EK
The speed is not unique as EF said, and anybody who paid attention would remember these:
ReplyDeletehttp://image-sensors-world.blogspot.com/2010/04/fast-sensors-standoff.html
The pixel rates have been far exceeded, but what is truly unique, indeed, is that Sony has brought this to the consumer.
Yes, I agree that personal assault should be prohibited. The only question is how to enforce it. Now I'm trying a new tactic and block IP addresses where these comments come from using Blogger spam filter. Let's see if it works. Meanwhile the 16:54 comment will disappear in a minute.
ReplyDeleteDr. Fossum's comment is unfair to other engineers and researchers having worked hard on CIS. Science follows always helicot ascension. He has to be more humble and please don't give comments without estimate for others' work. I still remember that Cypress guys have designed a 16-bit flash ADC in one of their sensors, it should also be considered as impressive in circuit design ... The past is past and we have to look ahead, new era with new heros!
ReplyDelete@ 20:35 These are analog out sensors, whereas the Sony sensor spoken off is digital out - quite important difference.
ReplyDeleteAnd then the shutter type of the Sony sensor is rolling, just like the old and mighty PBMV40 - also an important difference in design and operation with already mentioned CY (OnSemi) VITA25k and CMOSIS CMV12000 global shutter sensors...
I see more and more high resolution-high speed sensors...Is there any objective to reach (e.g. ultra high definition television with 33Mpixels, 240f/s)? What are the potential applications which justify the higher costs?
ReplyDeleteAdi:
ReplyDeleteI'd imagine the demand is for contrast detection autofocus at full resolution.
@ 23:58 Cypress guys have designed a 16-bit flash ADC in one of their sensors
ReplyDeleteThis is just no true. They had 10 bits flash ADC with hardly 7 bits ENOB at full speed.
Back to the topic, any link to an image actually taken at full speed with the Sony sensor?
Apparently this is the same sensor presented at ISSCC few days ago:
ReplyDeletehttp://image-sensors-world.blogspot.com/2010/11/isscc-2011-abstracts-published.html
A 17.7Mpixel 120fps CMOS Image Sensor with 34.8Gb/s Readout
A 17.7Mpixel CMOS image sensor with a 27.5mm optical format realizes 120fps at 12b using 90nm CMOS. This sensor achieves 2.75e-rms random noise at 12b, 120fps with a maximum data rate of 34.8Gb/s. The 16 channels of scalable low-voltage signaling I/F with embedded clock operate at 2.376Gb/s each and the single-slope ADC ramp generator runs at 2.376GHz.
May be Sony has shown the images at the conference.
If its 27.5mm sensor, this will be an APS-C/Super 35mm sensor.
ReplyDeleteI agree with Eric.
ReplyDelete@ "the single-slope ADC ramp generator runs at 2.376 GHz"
I'm skeptical about this. Even if this figure is the counting speed, with a 12-bit count this implies a 580 kHz fundamental frequency for the ramp generator. That's in the ballpark for 120 fps with digital CDS for arrays of this size, but I think it would be challenging to get sufficient ramp linearity at 580 kHz.
On further thought, it is probably a challenge that is overcome with a large area and power budget. Very interesting stuff, this.
ReplyDelete