PR Newswire: Vision Research unveils Phantom v1610 camera said to be the world's fastest 1 megapixel camera, 60% faster than any other camera on the market, up to 16,000fps at full resolution. The camera features custom designed 1280 x 800 resolution CMOS sensor with 28um pixels. The sensor's I/O speed is 16 gigapixels/second.
Impressive. I wonder if the sensor has on-chip ADC or not? I think earlier sensors for the Phantom line were analog output.
ReplyDeleteAlso, I wonder who designed the sensor?
@ "The sensor's I/O speed is 16 gigapixels/second."
ReplyDeleteThe maximum array read rate is 16 gigapixels/second, but the peak data rate it much higher. The release cites 12-bit A/D conversion (with no indication of whether this is in-pixel, array-external on-chip, or off-chip) for a raw data rate of 192 Gbits/second. I suspect this is passed to the local memory, with sustained rates limited by the 10 Gb Ethernet port or whatever the "CineMag" add-on allows. The 96 GB memory can store 4 seconds of raw data that takes 76.8 seconds to jam through the Ethernet port!
This release reminds me of a bittersweet moment from my past. On my one away college visit as a high school student, to Cambridge, MA, I was very excited to find that the campus tour would go by Harold Edgerton's office and that he would personally come out to distribute prints of some of his amazing photographs that I'd seen for years in my school science books. Everyone on the tour was disappointed to find a note on the office door indicating Prof. Edgerton was out sick. Shortly thereafter I learned that he'd passed away. He would certainly have loved to read this announcement and to get his hands on one of these cameras, data bottleneck or not.
They are big shoulders we all stand on, even if we don't think about it all that often.
That should say "is much higher".
ReplyDelete"Both cameras are based on Vision Research proprietary sensors offering not only high speeds, but larger 28-micron pixels that allow for superior sensitivity when shooting in low light, which is often a problem with high-speed imaging." From Vision Research's website
ReplyDeleteI doubt Vision Research designed the sensors. They probably spec'd them and some design group designed them.
ReplyDeleteShould be calest ,?
ReplyDeleteBy increasingthe io no you can increase a lot the frame rate. The system is complex, only interest for some very nich market.
ReplyDeleteMaybe this theory put to practice ?
ReplyDeletehttp://www.imagesensors.org/Past%20Workshops/2009%20Workshop/2009%20Papers/037_paper_meynants_cmosis_framerate_limits2.pdf
so what ???
ReplyDeleteI think he is suggesting that mr meynants (CMOSIS) designed this sensor.
ReplyDeleteIt's a joke.
ReplyDeletepeople in high speed imaging know who designed this sensor and also that this is not the highest speed CMOS sensor out there. the so-called limits presented in Ogunquit had already been broken years before they were published.
ReplyDeleteIn any case, when resorting to stitching, more I/O can be placed on-chip, more fingerpads can be placed on a single bonding deck etc...maybe other metrics are required to quantify the challenge? What about pixelrate/mm^2? (making this sensor about the same speed as sensors 4 years ago) or maybe pixelrate/mm (on a side) to indicate the difficulty of getting data out of the sensor and array.
From a camera perspective the challenge is considerable, nonetheless, and different paradigms are required to move much further.
Nice metrics. They need to account for the process, however. In 40nm process 5Gbps per diff pair I/O is almost a commodity. 10Gbps is available and even 25Gbps is around the corner. On the other hand, 180nm process is much more restrictive in that sense.
ReplyDeleteYour metric should account for this.
I am still wondering why the designers are secret. Seems like they should get credit for their contribution. How could it possibly hurt Vision Research to acknowledge where the sensor design came from? These days it is common, even a matter of pride, to advertise things like Carl Zeiss optics, or Intel-inside, or even firmware. Why is there no prestige in announcing "Sensor Design by Forza" (or in this case, maybe some European group)?
ReplyDeleteAs image sensor specialists, we should insist on such "brand recognition" and not be banished to be some anonymous designers.
Anyway, if it is common knowledge among people in high speed imaging, why not post it definitely and anonymously here? I could have guessed from the beginning but I don't know for 100% sure.
Maybe limited by their NDA/contract???
ReplyDeleteWell, I guess if you are "brand name" design house, you might have the clout to insist on recognition at product release before you sign the contract. This will be easier when it becomes common practice. We really need to elevate our profession.
ReplyDeleteif they are limited by their NDA/contract, it's clear their engineers are better at their jobs than their business folks
ReplyDeleteIt's really a value declaration. If the design house can bring market notority to thr camera maker, the camera maker will mention volonteerly the design house name. Otherwise they will not do this. Not all can have the marketing power as intel to impose 'intel inside' logo!
ReplyDeleteThere are some examples of camera companies disclosing their sensor vendors. In machine vision there is sCMOS partnership between Fairchild Imaging, PCO and Andor. Security camera vendors publish "Powered by Pixim" statement. In consumer space GE said it uses Aptina sensors in its cameras.
ReplyDeleteIt's just in this particular case Vision Research likes to keep it in secret.
Vladimir, please don't forget ALL claim Sony EXView HAD !!!
ReplyDeleteConsidering yesterday's announcement of Google offering to buy Motorola Mobility for USD 12.5 billion, the small shop pride-in-your-work seems like wishful thinking. If the image sensor in a smart phone appears to be a crucial differentiating feature AND there's the possibility of encumbering smart phone competitors through actual or suggested assertion of image sensor patents, the big smart-phone players with deep pockets - Google, peers, and would-be peers - will probably go after the free-standing image sensor companies that have either key individual patents or impenetrable-thicket patent portfolios, the offer amounts having little to do with whether the target companies have excellent design teams and marketing groups or not.
ReplyDelete@ CDM maybe its my reading comprehension, but can you be more lucid in a layman's term explain what exactly your point is??
ReplyDeleteCDM's comment is sort of non-sequitur, but I found his point clear enough. He thinks image sensor patent portfolios, thick or thin, might be going up in value. The increased IP valuation has little to do with the company's design strength, rather it is driven instead by IP market forces.
ReplyDeleteI don't agree with this completely. For example, we recently saw the very large Kodak image sensor IP portfolio bought up for a relatively small amount by Omnivision after being shopped around for a very long period of time at a various higher prices.
Well suppose you disclose where your superduper fast sensor comes from, it might be that your competitor goes to the same design house asking for the same design, basically. The design house, being a pure player, just recycles the design, adds some minor changes (just to be not an exact copy and hence preventing lawsuits) and sells it to your competitor...
ReplyDeleteThereby all your effort to get the best performing camera first on the market is made worthless as your competitor might just lag a quarter or two.
To my own comment just above: See the NY Times article today on the patent bull market
ReplyDeleteSeems to lend credence to CDM's comment. The article speculates that a smart phone might infringe about 250,000 patents.
To anon above: This situation is easily, and usually covered by contract. Also, I don't think it is possible to get a sensor modified, fabbed, qualified, into production and into a new, competing product in a quarter or two, but I understand your point.
ReplyDelete@anon
ReplyDeleteThe high speed camera vendors also know the design houses that can do this, so the non-disclosure doesn't really help much. Moreover, the benefit works in both directions.
There are improvements in quality and speed on every iteration and everybody benefits.
@EF
"covered by contract" : this is why anon mentioned "adds some minor changes"
but you are right, it takes more than two quarters...and lagging by more than that requires the incremental improvement mentioned above
everyone recycles the NMOS and PMOS transistors in his design ...
ReplyDeleteHere are two phrases to put in a design contract:
ReplyDelete"Substantially similar"
"Liquidated damages"
These will ensure that the same chip, or one with minor changes, never goes to a competitor of the customer. Of course no CEO in his right mind will accept a liquidated damages clause unless it is a huge business deal. Nevertheless, something like this is in most design contracts, whether or not the design house gets branding rights.
Some customers prefer limited exclusivity (e.g. 2 years after product release) for reduced NRE costs.
Exclusivity is a thing that should be very well understood and written.
ReplyDeletePutting such vague terms in your contracts, is asking for troubles. You will be out of business quickly, the legal way (no more customers), or illegal way (getting sued).