Thread: Sony A7S III - Supreme Compact 4K

Reply to Thread
Page 6 of 10 FirstFirst ... 2345678910 LastLast
Results 51 to 60 of 93
  1. #51  
    Senior Member Karim D. Ghantous's Avatar
    Join Date
    Oct 2011
    Location
    Melbourne AU
    Posts
    1,966
    Quote Originally Posted by Bruce Allen View Post
    Ah, haven't used a Micro 4/3 camera in ages... why do you still use it? I'm interested!

    BTW, what's your bet that inside this is a 48mp quad bayer hard-wired to 12mp mode? Would explain the latitude if some of the pixels were at different sensitivities similar to Alexa / Canon dual-gain.
    I'm not sure that dual-gain has anything to do with the Bayer filter. I think its has to do with duplicating the output of the sensor before it is debayered, then amplifying one dataset, then merging them together, preserving the highlight data of the non-amplified dataset. Or some shit.

    As the for the Quad Bayer pattern - it could be the case. I don't know if anyone has said anything about it yet.

    Micro 4/3 as a whole is a terrific system. Smaller lenses (not always); very small telephotos; small bodies (though not the smallest); terrific IBIS; nice features that eliminate the need for post-processing; fast operation; amazing macro. I also like Sony and Fuji S35, though.

    I have a theory that says that the wider the AOV, the larger the sensor should be, and the narrower the AOV, the smaller the sensor should be. Ideally you'd shoot, say, a Leica S for interiors and an E-M1 for birds. Actually I might post a separate thread about that, even though it applies more to photography.
    Last edited by Karim D. Ghantous; 07-29-2020 at 06:45 PM. Reason: Typo
    Good production values may not be noticed. Bad production values will be.
    Pinterest
    | Flickr | Instagram | 1961 (blog)
    Reply With Quote  
     

  2. #52  
    Quote Originally Posted by Karim D. Ghantous View Post
    I'm not sure that dual-gain has anything to do with the Bayer filter. I think its has to do with duplicating the output of the sensor before it is debayered, then amplifying one dataset, then merging them together, preserving the highlight data of the non-amplified dataset. Or some shit.

    As the for the Quad Bayer pattern - it could be the case. I don't know if anyone has said anything about it yet.
    This is not the sensor in the a7s3 but from the datasheet of a similar Sony sensor (for a full frame consumer DSLR that hasn't been released yet) - take a look at how they do the "HDR" readout RE: conversion gain. The Quad Bayer setup makes it easier to output the combined result as one "RAW, HDR" sample.



    I suspect (but could be wrong!) that the a7s3 has a 12mp / 48mp quad bayer variant of this sensor design, hardwired to 12mp.

    Will we see future Sony DSLRs with higher res sensors and Quad Bayer high res stills too? Seems like a logical a9 mark 3 to me because then you can sell the Associated Press a high-speed sports camera that can also shoot great latitude video AND can pop off high-res stills too if you switch modes.

    I am sitting out the a7s3 unless the right project comes up - but future a9 or a7 cameras with that nice EVF, high speed cards (which will be cheaper in the future anyway) and new sensors could be tasty!

    Bruce Allen
    www.bruceallen.tv
    Reply With Quote  
     

  3. #53  
    Senior Member Jeffery Anderson's Avatar
    Join Date
    Jul 2012
    Location
    San Francisco
    Posts
    682
    Quote Originally Posted by David Rasberry View Post
    This is an old misconception. The finished resolution of a 4k Bayer image that is processed to RGB is still 4K. If it is processed to 4:4:4 RGB color space fully sampled, it will still be a full pixel for pixel 4K image with every pixel a fully defined RGB color pixel. There may be very slight loss of red/blue channel spatial density which is perceived as a blur, but nowhere near the loss of 4:2:2 or 4:2:0 sub sampled video.
    Sophisticated Bayer color filters all have some overlap in filter values such that a red pixel is not pure red but has some blue and green value. Green has blue and red values, and Blue has red and green values. Modern Debayer algorithms and advanced color science can accurately interpolate those value differences for full luminance resolution and nearly full color resolution.

    The brilliance of the Bayer pattern is that it closely mimics the spatial color sensitivity of human vision. We don't perceive red and blue values with the same density as green either.
    those are some pretty words, but don't you lose like 25% resolution?
    I have to try and change the landscape, whatever it is.
    -Robert Plant
    Reply With Quote  
     

  4. #54  
    Senior Member Jeffery Anderson's Avatar
    Join Date
    Jul 2012
    Location
    San Francisco
    Posts
    682
    Quote Originally Posted by Bruce Allen View Post
    Bruce Allen
    www.bruceallen.tv
    anytime you can bring the Yellow Submarine into a conversation, it makes learning so much easier
    I have to try and change the landscape, whatever it is.
    -Robert Plant
    Reply With Quote  
     

  5. #55  
    Senior Member rand thompson's Avatar
    Join Date
    Aug 2011
    Posts
    7,605















    Reply With Quote  
     

  6. #56  
    Senior Member
    Join Date
    Aug 2008
    Posts
    4,270
    Quote Originally Posted by Jeffery Anderson View Post
    those are some pretty words, but don't you lose like 25% resolution?
    Optical resolution of any format at the Nyquist limit is less than the pixel resolution, oversampled or not. Oversampling gains a little bit , but nothing like 25%, more like 5%. The difference would be less aliasing and better chroma resolution along edges.
    Reply With Quote  
     

  7. #57  
    Senior Member
    Join Date
    Aug 2008
    Posts
    4,270
    Quote Originally Posted by Bruce Allen View Post
    Bayer pattern is brilliant but not THAT brilliant. You can debate overlapping filter values but let's be clear - you're only capturing one value per pixel. It's not as bad as 3:1:1 but it's not exactly 4:4:4 either. And if you have too much of an overlap in filter values, you get into some very wonky color issues because you're subtracting colors from each other to get "pure" values.

    The other thing you're overlooking is that the OLPF has to be tuned to not produce horrid color moire... so the OLPF for a 4K Bayer sensor movie-optimized camera will be pretty soft.

    If you compare a7 mark 3 (6K bayer downsampled to 4K) vs a7s2 (4K bayer to 4K) for video mode... a7 mark 3 has more real detail, a7s2 is a bit either soft or oversharpened.

    Scroll down, - the tests are pretty clear:
    https://www.dpreview.com/reviews/sony-a7-iii-review/10





    Another good way of comparing is take a look at the Sony a7r3 which has two modes:

    - normal Bayer mode
    - and pixel shift mode where it shifts so you're sampling a color every pixel so it doesn't have to debayer.

    Really seriously big real-world difference:
    https://www.dpreview.com/news/564575...our-landscapes
    https://petapixel.com/2017/11/18/tes...ature-a7r-iii/

    Bruce Allen
    www.bruceallen.tv
    I don't dispute the advantages of oversampling, it does produce a more resolved image. But the differences are not nearly as great as people think. And a 4k image is still 4k, with minor differences in aliasing and chroma resolution if the debayer process is high quality. Something not typical of video cameras because it has to be done in real time. Processing raw images in post is different. With video the resolution loss from 4:2:2 subsampling is much worse.
    Reply With Quote  
     

  8. #58  
    Tangential but

    https://www.youtube.com/watch?time_c...ature=emb_logo

    Is this fake?

    Reason is there's a shot going to the lightbulb and it's hand held and both cameras are exactly the same, I think these 'tests' are slightly screwed around footage for views...

    See, the inter webs - can't trust anything...

    cheers
    Paul
    -------------------
    insta: @paul.inventome
    Reply With Quote  
     

  9. #59  
    Senior Member Kemalettin Sert's Avatar
    Join Date
    May 2009
    Location
    Los Angeles CA
    Posts
    3,329
    is there any a7siii vs komodo tests? there are like few hundreds of komodo owners out there and its so suprising nobody did a comparison test.
    Reply With Quote  
     

  10. #60  
    Senior Member Jeffery Anderson's Avatar
    Join Date
    Jul 2012
    Location
    San Francisco
    Posts
    682
    Quote Originally Posted by David Rasberry View Post
    Optical resolution of any format at the Nyquist limit is less than the pixel resolution, oversampled or not. Oversampling gains a little bit , but nothing like 25%, more like 5%. The difference would be less aliasing and better chroma resolution along edges.
    I always understood that OLPF combined with debayering of a CMOS sensor means you need to shoot at least 25% more resolution than what you finish in... isn't that a requirement for more production companies that have theatrical releases? (this is stuff I've heard 1st hand from Producers, DITs and DoPs...)

    I'm not talking about oversampling, like what the Panasonic EVA-1, Sony FX9, and A73... I was referring to shooting 5K+ to finish in 4K
    I have to try and change the landscape, whatever it is.
    -Robert Plant
    Reply With Quote  
     

Posting Permissions
  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts