Sony IMX586 promises to put 48 effective megapixels in phones

Sony has announced a new smartphone image sensor that promises to produce some seriously impressive camera specifications for some future smartphones. The IMX586 performs some DSP magic on the raw bits from its unusual Bayer color array to arrive at an impressive-sounding 48-effective-megapixel resolution. That kind of resolution is more typical of high-end DSLRs, not handsets. Of course, comparing a DSLR sensor to an eight-millimeter-diagonal smartphone sensor isn't an entirely straightforward comparison, and the IMX586 is no exception.

Image: Sony

You see, Sony has filtered the IMX586 with 2×2 arrays of red, blue, and green pixels, rather than the traditional Bayer array of one red, one blue, and two green color filters per 2×2-pixel sub-array. The company says this “quad-Bayer” approach lets it pull off two tricks. In good lighting conditions, Sony's DSP magic conceptually lets the sensor perform array conversion on the raw pixel data to a traditional Bayer mosaic before further processing. That conversion is how the company arrives at that 48-MP effective resolution. In darker scenes, the data from the sensor simply isn't converted, resulting in a 12-MP effective sensor that Sony says has the light-gathering power of 1.6-µm sensor wells.

IMX586 sample on the right. Image: Sony

High resolutions on tiny sensors  usually mean disappointingly noisy images, but Sony claims that the industry-first 0.8-µm bit buckets on the IMX586 were “designed and manufactured with techniques that improve light collection efficiency and photoelectric conversion efficiency over conventional products,” possibly leading to crisp images despite the tiny individual pixel wells. Indeed, the single sample that Sony shared offers noticeably higher resolution while keeping noise levels low, although the image was clearly taken in sensor-friendly broad daylight. I'll be curious to see results from the IMX586 in a wider range of lighting conditions.

The IMX586 has a range of other interesting raw specifications, like DCI 4K video recording at 90 FPS, 1920×1080 at 240 FPS, and a cropped 720p capture at 480 FPS. Of course, smartphone makers will need to integrate image signal processors and buffers capable of handling more than a few seconds of footage at those frame rates in their finished product, but the sensor does sound like an impressive starting point.

Image: Sony

Sony plans to begin sampling the IMX586 to customers in September at a price of 3000 yen, or about $27 at today's exchange rates. Shipping products with this sensor inside will likely begin appearing sometime next year.

Comments closed
    • pogsnet1
    • 1 year ago

    This will make your phone space full in a short time

    • psuedonymous
    • 1 year ago

    With the individual subpixels split into 4, there are some other interesting things you can do (with sufficient light available for the small subpixel area). For example, each subpixel can have a different exposure time to provide an extremely wide dynamic range. Or you can stagger capture times to get an extremely rapid 4 shot burst. Or add a laser-pulse illumination and offset the subpixel capture by a portion of the frequency (ofsetting the captured phase) and use it as a depth-sensing camera with the same mechanism as the Kinect 2 (and if you did it with 4 phase segments per pixel rather than 2, you have better discrimination too).

    You can do fun things for video too: because you have multiple sensors available per pixel, you can decouple shutter speed from framerate. For example, you could capture at 120FPS, but with a shutter open time longer than 8.3ms. You get the benefits of the nice long motion blur (and greater light capture time) of a long shutter interval, with the high motion discrimination of higher capture rate.

    • CuttinHobo
    • 1 year ago

    Enhance… Enhance… Enhance.

      • chuckula
      • 1 year ago

      Just print the damn thing!

      • Aether
      • 1 year ago

      Definite up vote for the reference!

      [url<]https://www.youtube.com/watch?v=xbCWYm7B_B4[/url<]

    • Kretschmer
    • 1 year ago

    I’m more interested to see this technology applied to actual cameras. Imagine what today’s 1″-sensor pocketable camera could do with this tech!

    • Unknown-Error
    • 1 year ago

    It’s been how many years since Lumia 1020?

      • NoOne ButMe
      • 1 year ago

      I am eagerly awaiting the Nokia zero-oh-zero…. which is probably never going to arrive.
      (808->909-> 000?)

      Especially as HMD seems to have camera as the weak part of their phones 😐

      • freebird
      • 1 year ago

      Still got mine! but it is mining now… (shush…..)

    • strangerguy
    • 1 year ago

    I gonna call bullshit on that image comparison, when most flagship phone reviews say the image quality in broad daylight conditions between them are minimal at best, even when taking into account devices that are 2-3 generations behind.

      • adisor19
      • 1 year ago

      It’s probably zoomed in quite a bit in order to showcase the difference even more.

      Adi

    • DPete27
    • 1 year ago

    48MP smartphone camera image clarity finally equal to 5MP from a proper camera?

      • Froz
      • 1 year ago

      Not unless you put proper camera lens on it.

    • chuckula
    • 1 year ago

    The pixel wars have come to this. Now it’s “effective” pixels?

    Don’t give Intel ideas for how to bloat its “effective” core count via marketing here people!

      • blastdoor
      • 1 year ago

      Presumably it would be easier to inflate the ineffective pixel count. These pixels aren’t just effective, they’re [b<]in[/b<]effective!

        • chuckula
        • 1 year ago

        Effectiveness-light pixels.
        JUST ONE CALORIE!

      • DPete27
      • 1 year ago

      AMD already sailed that ship with Bulldozer.

        • chuckula
        • 1 year ago

        Intel has shamelessly stolen AMD’s innovations in the past. I expect that trend to continue!

          • blastdoor
          • 1 year ago

          Perhaps next year intel will brag about how their chips come with twice the number of nanometers

            • chuckula
            • 1 year ago

            After all these decades Intel will finally be competitive in TR’s nanometers-per-dollar scatter chart!

            • blastdoor
            • 1 year ago

            Pixels for nothin’ and the chips for free.

            (…I want my fast PC…. )

      • Froz
      • 1 year ago

      I would say Intel did this already years ago by introducing multithreading. It doubled the “effective cores”.

        • hansmuff
        • 1 year ago

        “Hyperthreading”, but you’re correct otherwise.

          • BurntMyBacon
          • 1 year ago

          “Hyper-threading technology (HTT)” if you want to refer specifically to Intel’s branded implementation of it.

          “Simultaneous multithreading (SMT)” if you want to refer to any implementation of the technology (including Intel’s HTT).

          “Multithreading” can be used with SMT implementations, multicore processors, systems with multiple single core processors, or even single core systems with accelerators that can execute their own thread.

        • jihadjoe
        • 1 year ago

        Nah, “effective resolution” has been a thing for a long time before Hyperthreading especially with stuff like Laser printers. (REt on the HP LaserJet III back in 1990!)

        Stuff like varying the dot size (oh those halftone patterns) to change the effective resolution despite the number of addressable points remaining the same.

        Edit: have a look at the vintage ad!

        [url<]http://www.hpmuseum.net/upload_htmlFile/PrintAds/Ad1990Aug_LJIII_InformationDecisions-41.jpg[/url<]

        • freebird
        • 1 year ago

        It didn’t matter too much until M$ & pORkACLE started charging for those phantom cores.

      • hungarianhc
      • 1 year ago

      TBH I think the pixel wars are actually on pause / dead. They’re going to OEM this to phone makers, and it seems like Samsung / Google / Apple have figured out that there’s diminishing returns on making images much larger. We’ve been at the 12MP range for a while, Google Photos does auto compression, etc.

      • psuedonymous
      • 1 year ago

      For cameras, the pixel-wars have always* been over ‘effective’ pixels. Each R, G and B element (subpixel) of the Bayer-patterned sensor has been counted as an individual pixel.
      In the professional space, this is why Red got a lot of shit from the rest of the industry when they launched: pro cameras quoted their resolution in terms of 1 pixel per RGB (or RGBG depending on array used) cluster. Red quoted them per subpixel, inflating their spec-sheet pixel count for an identical sensor. This is why they have such noticeably poorer light capture capabilities compared to other cameras, and why you end up with the extreme colour weirdness that needs to be compensated for in low light/high framerate (high framerate = faster shutter speed = less light gathered) like when shooting The Hobbit.

      * The exception in the consumer space is the stacked sensors from the now-defunct Foveon.

Pin It on Pinterest

Share This