• June 7, 2023, 6:39 a.m.

    About full-frame Foveon sensor:

    CEO Yamaki-san revealed that it is still actively in development. Although the company has assigned multiple engineers to this project, they continually run into technical issues.

  • June 7, 2023, 8:32 a.m.

    I'm not sure that I buy that at all. Past Foveon sensors, whilst the processing is more complex, haven't taxed the image processors of the time, which were much less powerful than todays. I would guess that what it's code for is that any new Foveon sensor camera will be extremely expensive and to justify the cost will be kitted out with all the 'pro' features, including integrated grip and big battery.

  • June 7, 2023, 8:49 a.m.

    Today you need to process 8K in realtime, there are likely no off-the-shelf processors for Foveon specific processing needs available - IMO this means that in addition to sensor development they struggle with processing Foveon data too.

    I have written some code to process older Foveon image data - this is quite nasty business actually; you require some different operations than Bayer processing needs.

  • Members 62 posts
    June 7, 2023, 10:04 a.m.

    But when asked about whether we'd see a Foveon camera in an fp form factor (4:20 onwards), he actually says "I really hope so".

  • Members 40 posts
    June 8, 2023, 4:16 a.m.

    From the video shared in the link...the discussion on the Foveon/X3 sensor is from timestamp 2:55 - 5:10

    Gist of the discussion: "We will never give up" [3:20]

    That for me sums it up. I am committed and in the Sigma camp 😇

    Some interesting observations from the discussion:

    Yamakai San hesitates in using the Foveon name...and instead corrects himself to use X3 sensor instead. [3:41]

    Minor problems in new X3 sensor design. [4:00]

    Yamaki San didn't like FP form factor. [4:32]

    Foveon/X3 sensor needs powerful processor and backend system. [4:44]

    All in all, its an optimistic view!

    And there is no discussion about battery?? At least, i could not hear it.

    Regards,
    Sattva

  • Members 62 posts
    June 8, 2023, 5:38 a.m.

    I believe this should be "really like", which would explain why he uses it every day.

  • Members 40 posts
    June 8, 2023, 7 a.m.

    Agree with you.

    Lost in translation. Inside Yamaki San's head 😃

    I listened to the interview more than once with headphones...and...

    At 4:34, Yamaki San clearly mentions "I didn't like the form factor of the FP"
    At 4:38, "Actually i use not FP everyday. I put the FP in my bag"

    So, i guess if Yamaki San communicates in Japanese, it will be clear and without ambiguity.

    Regards,
    Sattva

  • Members 509 posts
    June 8, 2023, 1:02 p.m.

    My most recent Foveon camera was a while ago, the DP2 Merrill. It was a very small and light camera with the tiniest little battery. The camera took about 7 secs to process and save a raw file. You could do nothing useful while the red light blinked. The battery lasted for about 40 shots fully charged, more like 20 shots on a cold day.

    I think it is fairly obvious the processor needed to do a lot of work before the raw got written to the card.

  • Removed user
    June 8, 2023, 7:38 p.m.

    As to speed and/or heat dissipation Sigma has used the Milbeaut PGLA in their cameras, e.g. the "True-II" processor.

    I don't know whether that continued in the Quattro series (True-III?).

    I have little interest in their X3 for my own purposes but it would be sad to see the Foveon technology die ...

  • Members 2 posts
    June 8, 2023, 9:19 p.m.

    All in all, I think this is good news.

    I’m happy to wait 2-3 more years for a good FFF from Sigma.

    —Tom

  • Members 9 posts
    July 13, 2023, 1:32 a.m.

    I agree. In a recent thread at DPreview I argued at length about this. The fact is the little DP Merrills, which are now like ten years old (I just checked, and the DP2 Merrill was announced February 8, 2012), have to process almost as much information as the FFF camera will (45 MP vs 60 MP), and processors have multiplied in speed/capacity since that time (more than doubling). This means they can indeed put a FFF sensor into an fp, with its bigger battery and special heat dissipation capability. They just need to be willing to do it. I don't think they are . . . yet. I think they just want to put the new sensor (when it is finally ready) into a form factor that screams, "I am a flagship camera!" The little fp body doesn't do that. I think they'll put the FFF sensor into the fp about a year or so after they put it into a nice body, which has an integrated viewfinder, with a bigger battery, which has an optional battery grip that can hold two batteries, and that larger camera body with have a dual-processor of some type, so it is significantly faster than any previous camera ever made with a Foveon sensor in it. Since it will be the first L mount camera with an X3 sensor, it will sell reasonably well, I think, and when they put the same sensor into an fp body, I believe many DP series camera owners will finally buy an L mount camera. Sigma has the little lenses available, and while the fp with a lens is not quite as small as a DP Merrill, it will be a faster small camera with an optional electronic vewfinder, which is something that has never existed in the world of Foveon sensors. It will still be significantly smaller than a full-sized camera, and it will be full-frame, so it will have an excuse for being bigger than the DP Merrills.

    BTW, for those who doubt that the fp can handle the FFF sensor, please tell me how the fp L can handle a 60 MP sensor? It seems to do it just fine, because I don't see a lot of complaints out there about how slow that camera operates. It's getting old now, and any newer model of fp will no doubt get a new, much faster processor (i.e. something 50% faster or even twice as fast as the one in the fp L). Now I know that Bayer pattern CFA cameras have the advantage of being able to take advantage of all sorts of circuitry that has been embedded into the modern processors that are made for compact cameras and smart-phones, but don't forget that the little DP2 Merrill worked just fine (though slowly) with a processor that was made for the world of CFA cameras too. That was more than ten years ago now, and today Sigma can easily get a processor that is more than twice as fast as that processor from the little DP2 Merrill.

  • July 13, 2023, 7:04 a.m.

    Let's continue arguing :)

    Does DP2 Merrill do 4k/8k video? No. For nowadays cameras you need hi-res video and doubling (or quadrupling) processor performance isn't nearly enough. You need some atypical (different than for CFA) processing even for raw video, off-the-shelf components and algorithms (optimised for CFA) can't handle that efficiently.

    For still images processor performance is likely not a problem - but this does not help.

  • July 13, 2023, 8:17 a.m.

    I think that's a key point. Most of the image processing chips in cameras have dedicated hardware pipelines for image and video encoding. The volumes of cameras, even in the heyday of digital photography, weren't enough for these chips to be designed from scratch for the purpose - they were always variants of commercial 'systems on a chip', mainly originally designed for phones or video systems. Sigma's volumes are tiny, their processors are likely completely off-the-shelf. Their imaging pipelines are designed for Bayer, and the raw processing needs to be done programatically, using the CPU.

  • Removed user
    July 13, 2023, 2:17 p.m.

    Sigma was using MilBeaut PGLAs - are they really designed only for Bayer ??

    en.wikipedia.org/wiki/Milbeaut

  • Members 1737 posts
    July 13, 2023, 2:35 p.m.

    Programmable Logic Gate Arrays aren't designed for anything, until they're programmed.

  • Removed user
    July 13, 2023, 6:12 p.m.

    Some time ago, I commented on the other Site that Sigma was going to drop the Foveon name for the new sensor.

    Many people disagreed.

  • July 13, 2023, 7:26 p.m.

    Actually those Milbeauts were processors, not just PLGA-s. SD14 (and probably earlier ones) used PLGA's.
    Many Foveon/X3 image processing operations are similar to CFA ones (matrix operations, some kind of noise removal, jpeg compression) - but this True-II was likely more hack than specifically created processor for X3, even raw file compression was optimised for CFA pattern and was not optimal for X3.
    About who produced processors for next Sigmas, I have no knowledge.

  • Members 40 posts
    July 21, 2023, 4:51 a.m.

    The Sigma SD14 used Analog Devices "Blackfin" processors.

    I do vaguely remember reading about them when i was researching my camera purchase way back in 2007. If i search my backups, i may probably find the relevant datasheets!

    Not sure about the Sigma DP2s...which i believe uses the same image sensor as the SD14.

    Regards,
    Sattva

  • July 21, 2023, 7:50 a.m.

    Thank you for correction!
    I managed to find one lost page on wayback machine about SD14 internals, it contains couple of interesting things :)

    ... the SD14 has inside:
    Foveon X3 14.1 MPixel sensor
    ASDP BF651 Blackfin DSP
    Xilinx Spartan-3 XC3S200 FPGA
    Analog Devices AD9228 A2D converter
    4 Samsung K4S511632D-UC75 (512Mb SDRAMs)
    MegaChips DSC-3H
    2 Samsung K4S561632H-UC75 (256 Mb SDRAM)
    Toshiba TC58FVM6B5BTG65 (64 Mbit NOR flash)
    Casio LCD

  • July 21, 2023, 10:13 a.m.

    Usually the PGAs are just used for interfacing. The computational stuff needs a processor.