Lock in Your Premium Membership Discount!

Your content will be up shortly. Please allow up to 5 seconds
Gear Rumors

Sony’s Rumored APCS Technology, Not All It Was Cracked Up To Be?

By Anthony Thurston on November 16th 2014

Today, we have some interesting new information regarding Sony’s APCS (Active Pixel Color Sampling) sensor, which we first told you about earlier this week. This information significantly changes, for now, the outlook of this new technology in my mind.

large_structure1

According to a new post over on Mirrorless Rumors, which includes a diagram of the new sensor and how it will work, this new technology has one very big drawback. That drawback is that, unlike what was previously thought, this sensor must take three exposures in order to take one picture: a green exposure, a red exposure, and a blue exposure. It’s time to take those rumor goggles (that make everything sound amazing) off and look at this objectively.

The fact is, as you can imagine, this would severely limit the camera’s ability to shoot anything with motion or even long exposures. I can’t imagine Sony would invest heavily in technology that would be hampered by things as simple as motion or long exposures, so there must be something here that we are not aware of yet (almost a certainty). But as is, this tech is not all it was cracked up to be when the patents/rumors first came to light earlier this week.

What Do You Think About Sony's New APCS Technology?

View Results

Loading ... Loading ...

Maybe since the sensor has a global shutter, it is able to take these RGB exposures in such quick succession that for shorter exposures, the time to take three exposures is negligible. But that still raises questions about longer exposures and motion, because any movement between exposures could really mess up an image if the sensor is really taking three separate exposures and merging them to get the final image. I can’t help but reiterate here that either something is going on here that we don’t know (a very likely situation), or this bit about taking three exposures is not entirely accurate. Either way, we will need to wait until the tech is released to find out.

28-megapixel-aps-c-cmos-image-sensor_2

The new APCS sensor technology is expected to debut in some new mobile phones in early 2015, with larger (APS-C and Full Frame sized) sensors coming in new camera models sometime after that.

[via Mirrorless Rumors]

Anthony Thurston is a photographer based in the Salem, Oregon area specializing in Boudoir. He recently started a new project, Fiercely Boudoir to help support the growing boudoir community. Find him over on Instagram. You may also connect with him via Email.

Q&A Discussions

Please or register to post a comment.

  1. Greg Geis

    I’m sure they could come up with reasonable technology to deal with long exposures and movement, like a continuous cycling chip while it is exposing. But if each wavelength is exposing separately, it seems like it would fundamentally kill the advantage of being able to use 3x larger photo sites. I’d like to see where it develops and what advantages/disadvantages the technology could yield.

    | |
    • Stan Rogers

      It’s not the 3X larger photo sites that’s the advantage, it’s the full-colour info at each site. Aliasing with a Bayer-type sensor (or anything that uses a spatially-displaced colour filter array; the problem isn’t specific to the Bayer pattern) happens at much lower frequencies than it should because the colours are arrived at by consensus in a committee of sensels, so to speak. As committees go, they do a pretty good job, but they do build the occasional camel where a horse is needed (things like moiré and mistaking fine texture for flat colour). There are two basic ways of solving the problem.

      The first is to throw more pixels at it, but you need a lot of pixels and a lot of bureaucracy (processing power, parallel reads,etc.) compared to the actual output you’re looking for. That’s where RED and the existing Canons and Sonys have gone so far.

      The other is to give each pixel total responsibility for its own info. Foveon does that, as do the Hasselblad multishot and three-chip cameras. Foveon is necessarily limited in sensitivity (and Sigma hasn’t been great at processing speed), the ‘Blad is way too slow, and three-chip cameras start running into alignment issues when the resolution gets high. Sony is essentially using Hasselblad’s trick, but using very short electronically-shuttered exposures rather than firing a physical shutter for each sub-exposure. Limit the “shutter speed” at the high end (flash need not apply) and relabel and restrict the “ISO”, and you can still pull decent performance out of the thing. (Think α7s-type noise performance, but with the 25.6K ISO marked as 6400, the 6400 marked as 1600, etc., since the exposure per colour will be about 1/4 of what it might have been with a fixed chip. And if they can move the base “ISO” setting down to, say, 25, that’ll eliminate a lot of ND filter play. As often as not, getting rid of excess light is the problem in video, not seeing in the dark.)

      (Three-chip — or three-tube, if you’re old enough — video cameras used to be the standard for direct-to-video production, with all of the light losses you’d expect from beam splitters AND long light paths that meant that wide-angle lenses needed to be severely retrofocal with HUGE front elements. Oh, and the beam splitters and chips need to be perfectly aligned, which gets much mor delicate and fragile as video resolution goes up. For stills, there are a few interesting 15MP Sigma DPx Merrill versus 36 MP Nikon D800E side-by-sides on teh intarwebz, and you can also check out the Hassy 50MP versus Hassy 50MP multishot — or the 200MS in 4-shot 50MP mode — just to see what dumping Bayer does for detail. If Sony can manage the trick at video rates and resolutions and reasonable equivalent ISOs with a good, gradable codec, this chip could make a helluva cine camera.)

      | |
  2. Stan Rogers

    That’s pretty much what I got out of the first report (and mentioned in my comment on that). The physical shifts needed are *tiny* and they’re moving far less mass than in the Hasselblad MS system, so they can be done very quickly (I’d assume at a fixed rate that works out to multiples of 3 exposures at most “shutter speeds”), but there is an upper limit on the (equivalent) shutter speeds you can use, and there will be a trade-off between luminance noise and colour accuracy (that will be more than acceptable at low-to-moderate equivalent ISOs, but will rapidly degrade into colour mush at higher equivalent ISOs — there’s no reason to think that they couldn’t coax reasonable performance out of a system like this up to, say, 6400, but don’t look for a clean 102K+ or anything like that). But it is something that would be happening at video rates, where the top shutter speed never has to be too very high (twice the frame rate, so 1/500 apparent at 240fps if they decide to go that high, which could easily be 6 “real” exposures at 1/8000 using the electronic shutter with time to move between shots, or something even faster — fast enough motion at a high enough rate to move most of the noise out of the audio spectrum). But it does look like it’s aimed squarely at video.

    | |
  3. Peter Nord

    Most ideas change in time. Sensors will change. Others will build on today’s ideas. Pretty much the way it’s always worked. I bet our grandchildren”s cameras won’t be much like todays. If they come out with a product from these ideas, like most product development the end product may be different from the product idea. Speculation is fun. Keep at it. It generates ideas. The what do you think statements in the poll are naive in asking for a conclusion about something that doesn’t exist.

    | |
  4. Arturo Mieussens

    I think the three charge-storage diodes below the actual photosensitive diode must have something to do with this. Maybe the filters are changing very fast and the three diodes add up the total charge for every color, so after the full exposure you can read the three values for every pixel. Just a guess but very interesting indeed.

    | |
  5. Gareth Roughley

    Sounds like there are some big kinks to work out but if they can the possibilities will be amazing

    | |