Join 3,551 readers in helping fund MetaFilter (Hide)

Tags:

Never mess with CMOS
March 17, 2014 3:43 AM   Subscribe

CMOS for imaging is the technology behind billions of cameras ( smartphones,....). Yet that technology simply did not exist 19 years ago. Eric Fossum started its development within NASA's and Caltech's Jet Propulsion Lab but he also shepherded that technology through the Technology Readiness Level ladder from a single item to its current market thereby reducing then incumbent technology (CCD) to a niche market. That story is recounted in one of his recent video presentation entitled CMOS Image Sensors: Tech Transfer from Saturn to your Cell Phone.
posted by IgorCarron (28 comments total) 19 users marked this as a favorite

 
Exhibit A on why we need a space program
posted by LogicalDash at 4:55 AM on March 17 [7 favorites]


CMOS existed at least 20 years ago: this camera (PDF) was in use before 1994 (I worked with it).
posted by Songdog at 4:56 AM on March 17 [2 favorites]


Saying we need a space program because we got digital camera sensors is like saying we need the military because we got the internet.
posted by Elementary Penguin at 5:16 AM on March 17 [4 favorites]


The military isn't a research organization. More like saying we need DARPA because the internet. Which I could kind of get behind.
posted by LogicalDash at 5:43 AM on March 17 [7 favorites]


CMOS existed at least 20 years ago

I had the same sense because I'd built chips while at Caltech in the 80s that used CMOS for image detection. So I went ahead and read a few of the links, and it seems that the distinction is in active versus passive cells. Active pixel sensors have an amplifier for each cell. In the passive sensors, amplification didn't happen until after the values were read onto a common line and the result was much noisier. Your link makes pretty clear that there were only two amplifiers for the 256x256 array, so it was a passive pixel sensor.
posted by Slothrup at 5:45 AM on March 17 [4 favorites]


What this really means is that as soon as they work out whatever details for the google-brain-implant - we will all be cyborgs quite soon after.
posted by sammyo at 5:50 AM on March 17


I'm not sure that something like DARPA would work without the military. As Fossum says, necessity is the mother of invention. Without specific problems to solve, it is very difficult to just sit down and invent.

No-one sat down and said 'I am going to invent the internet today', rather the internet was created as a solution to a series of military problems. Just like a lot of other projects, the solution turned out to be useful for more than just military applications. It was a significant improvement to existing pornographic distribution mechanisms, for example.

Thanks for sharing the talk though. CMOS sensors are cool, and the process by which technology gets introduced to the marketplace is fascinating (and surprisingly slow).
posted by YAMWAK at 5:51 AM on March 17 [2 favorites]


Thankfully we can totally have a NASA without a military.
posted by LogicalDash at 6:15 AM on March 17


Yeah, you have to ignore some prior art to claim that he's the first one using CMOS at all -- e.g. these guys (which my employer used back in the days) had commercial sensors in 1987, and non-commercial variants had obviously been around for a bit. (They guys also put a SIMD on the same chip, which sounds pretty active to me :-)

(insert comment here about how almost every country has the one true inventor of the propeller/fridge/television/computer/any possible gadget or food stuff)
posted by effbot at 6:37 AM on March 17 [1 favorite]


@Slothrup thanks for the clarification.
@effbot maybe my characterization was too much.

Yet I think there is an enormous difference between getting something to work and shepherding the technology to the masses. I am not trying to belittle the previous achievements, but very often, in this day and age of binary thinking, not much thought is given to how one needs to "cuddle" a technology through the technology readiness level ladder. Fossum went from developing something in the lab to the very cash hungry startup stage and saw the explosion of the technology on his watch. I cannot think of technologies that exploded like this one in recent time besides computers.

My other take away from this is that every time you develop a technology that is in competition with CMOS (even crummy CMOS implementation initially) Moore's law will allow you to slide past the incumbent (slide 7 of http://ericfossum.com/Presentations/2012%20March%20QIS%20London.pdf ). It is a starck reminder to never mess with Moore's law and that if you really want to see the future, you have to do it through that "law".
posted by IgorCarron at 7:10 AM on March 17 [1 favorite]


Thankfully we can totally have a NASA without a military.

Totally.

"Between 1982 and 1992, NASA launched 11 shuttle flights with classified payloads, honoring a deal that dated to 1969, when the National Reconnaissance Office—an organization so secret its name could not be published at the time—requested certain changes to the design of NASA’s new space transportation system. The NRO built and operated large, expensive reconnaissance satellites, and it wanted a bigger shuttle cargo bay than NASA had planned. The spysat agency also wanted the option to fly “once around” polar missions, which demanded more flexibility to maneuver for a landing that could be on either side of the vehicle’s ground track.

“NRO requirements drove the shuttle design,” says Parker Temple, a historian who served on the policy staff of the secretary of the Air Force and later with the NRO’s office within the Central Intelligence Agency. The Air Force signed on to use the shuttle too, and in 1979 started building a launch pad at Vandenberg Air Force Base in northern California for reaching polar orbits. Neither the Air Force nor the NRO was ever comfortable relying exclusively on NASA’s vehicle, however. Delays in shuttle launches only increased their worry; even before the 1986 Challenger accident, they were looking for a way off the shuttle and back onto conventional rockets like the Titan."

It be better of we could do away with the illusion that NASA is a research institute.
posted by three blind mice at 7:12 AM on March 17 [1 favorite]


CMOS was around a lot longer. In the 80s, CMOS was more-or-less synonymous with the BIOS. (ie the "hit esc to configure" prompts at boot time )
posted by k5.user at 7:23 AM on March 17


@three blind mice

I am sure you are not trying to hijack this thread but it would be quite a naive view to think that walking on the Moon and the rest of the Apollo program did not have a strategic value beyond PR. Congress signed on the Shuttle because a similar DoD program had died earlier and it is known that the Shuttle requirements were not entirely for civilian use.

Now, JPL has very little to do with the manned space program and most of its missions are exploratory in nature. The pressure of getting cameraq that did not have the problems of CCDs yielded an avenue for Eric Fossum to develop CMOS. These constraints may have had a usefulness for the military but it was primarly a exploration one for this technology development. In effect, this particular technology development is just not one of those very sensitive with dual-use purposes from the very start. I, for one, like that story.
posted by IgorCarron at 7:27 AM on March 17 [1 favorite]


@k5.user CMOS for imaging.
posted by IgorCarron at 7:28 AM on March 17


NASA is totally a research institute - as long as you want to research Russia.
posted by GuyZero at 7:40 AM on March 17 [1 favorite]


I guess that qualifies the NSA as a research institute.
posted by blue_beetle at 7:45 AM on March 17


CMOS was around a lot longer. In the 80s, CMOS was more-or-less synonymous with the BIOS. (ie the "hit esc to configure" prompts at boot time )

Complementary Metal Oxide Semi-conductor technology has been around since the early 60s, active CMOS imaging as a replacement for much more expensive CCD sensors is more recent.
posted by atrazine at 8:09 AM on March 17 [2 favorites]


This makes me think back to that move where the NRO/Perkin-Elmer fucked over NASA with the Hubble Space Telescope's main mirror issue. OOPS. Known design flaw is CLASSIFIED so fuck you NASA...
posted by mikelieman at 8:14 AM on March 17


As Fossum says, necessity is the mother of invention. Without specific problems to solve, it is very difficult to just sit down and invent.

Difficult, but not impossible ... A fact for which cats everywhere should be grateful.
posted by ZenMasterThis at 8:39 AM on March 17 [1 favorite]


Now, now, let's not bicker and argue about who preceded who.

At the very least, I think this is a good example of the general principle that valuable applications often (if not usually) derive from research in unrelated fields. At a time when basic research is being gutted, and we are being offered in its place a model of practical (according to the local billionaire) science, I think this underscores the notion that science works more like an ecosystem (and perhaps we should support it accordingly).
posted by bjrubble at 8:48 AM on March 17 [2 favorites]


Now, now, let's not bicker and argue about who preceded who.

Sorry, sorry. I'm afraid when I'm in this idiom, I sometimes get a bit, uh, sort of carried away.
posted by three blind mice at 9:45 AM on March 17


Whatever happened to Carver Mead's Foveon sensors? They were supposed to take the world by storm, but don't seem to be used much.
posted by benito.strauss at 10:27 AM on March 17 [1 favorite]


You can still get Foveon sensors in Sigma's cameras, but they're the only ones who use it and they never got very popular. Basically they're just too expensive. They initially had to be built at lower resolutions than Bayer arrays, so their increase in color detail was offset by their decrease in luminance detail.

They've taken to counting each of the color layers separately for marketing purposes, so their "46 megapixel sensor" produces the same number of actual color pixels as a 15 MP sensor. And they're still expensive - their SLR with an APS-C sized sensor costs $2100, as much as the cheap full-frames from Canon and Nikon.

Cool technology, but the cost/benefit ratio never leaned their way.
posted by echo target at 11:06 AM on March 17 [2 favorites]


We actually just talked about Foveon in one of my classes—apparently they have bad color accuracy. The real benefit, as far as I can tell, is that demosaicing the bayer-filtered raw camera output isn't an issue for Foveons, but demosaicing has gotten really good in the last decade, so maybe it's less of an issue for regular cameras, too?
posted by Maecenas at 1:07 PM on March 17 [2 favorites]


Necessity is the mother of invention, they need a problem to solve to create, and then make some money with it. This whole set of ideas is a monument to both human endeavor, and the blind side approach to how we treat our planet, how we spend our planetary resources. If only we could see the right problems, and find solutions that serve the entire blue marble, and all its inhabitants.

But, I love CMOS sensors, I am glad Canon got on board, I love the images my camera brings me, and I appreciate Mr. Fossum, for a camera I can hand hold for virtually every image I shoot. Oh yeah! I can't wait to see even more detailed images from space, from earth, from the moon.
posted by Oyéah at 1:54 PM on March 17 [1 favorite]


Noting that NASA has helped launch classified payloads isn't actually a counterargument to the thesis that we can have NASA without a military.
posted by LogicalDash at 6:45 PM on March 17 [1 favorite]


http://www.swtpc.com/mholley/PopularElectronics/Feb1975/PE_Feb1975.htm

1975: 1Kpix images from a 1Kbit MOS chip (Altair 8800 interface available!)
posted by nickzoic at 4:52 PM on March 18


The Cyclops used a 1024-bit MOS dynamic memory chip with the cover pried off (0.001 megapixels, yay!), so not really CMOS (both CMOS and CCD use MOS technology).

And as noted above, this is really about active CMOS sensors.

(I guess I should write something about how many (most?) big steps forward in engineering happen when a lot of people have been exploring a space for quite some time, and then the right person appears at the right time and has the right amount of patience and the right amount of luck, but I gotta run now so that will have to wait for some other day.)
posted by effbot at 7:43 AM on March 23


« Older When was the last time you got something in the ma...  |  Rrow itself, let it be sorrow;... Newer »


This thread has been archived and is closed to new comments