Loading ...
Sorry, an error occurred while loading the content.
 

OT: large files, RAM, 16bit and 8bit

Expand Messages
  • Carel
    Now that I output all equirectangulars to max size, I keep running into memory problems (XP with 2GB of RAM). I have been Googling about PS memory usage and
    Message 1 of 11 , Jan 3, 2007
      Now that I output all equirectangulars to max size, I keep running into
      memory problems (XP with 2GB of RAM). I have been Googling about PS memory
      usage and max addressable RAM. 2GB of RAM seems to be the maximum that
      Photoshop can use with your standard 32bit XP.

      The below talks about a "middle way".

      http://www.extremetech.com/article2/0,1697,1910437,00.asp
      There is a middle way being adapted by digital photographers who work in Raw
      mode:

      1. Process the raw file in Camera Raw and then open it in Photoshop in
      16-BPC mode.
      2. Do any major tonal and color corrections.
      3. Save the file as a master file.
      4. Duplicate the document.
      5. Convert it down to 8-BPC.
      6. Do any major edits that require hefty resources, such as multiple
      adjustment layers, duplicate
      blending, layer styles, multiple type layers, and so on.
      7. Optimize the 8-BPC version for output.

      Untill now I did all adjustments in 16bit.
      Does the above quoted advise make sense for the typical
      Equirectangular>Quicktime cubic output?
      Are there some general rules of thumb about the sequence of adjustments to
      stay out of trouble in 8bit mode?

      Carel Struycken
      --
      View this message in context: http://www.nabble.com/OT%3A-large-files%2C-RAM%2C-16bit-and-8bit-tf2917940.html#a8154563
      Sent from the PanoToolsNG mailing list archive at Nabble.com.
    • Aaron Spence
      G day Carel, I can t comment on the 16bit work arounds you ve mentioned as I only use 16bit for HDR stuff on still photos not VR s. But I can comment on the
      Message 2 of 11 , Jan 4, 2007
        G'day Carel,

        I can't comment on the 16bit work arounds you've mentioned as I only
        use 16bit for HDR stuff on still photos not VR's.

        But I can comment on the 2gb RAM issue. I upgraded my 1gb dual Athlon
        MP 2800+ to 3Gb earlier in the year, and didn't see a huge improvement
        as PS will only use 1.7Gb of that RAM. That is until I used the '3Gb
        switch' in XP Pro to give windows 3Gb of Ram to play with. Then PS
        started using 2.7Gb of RAM, and some processes, like flattening a big
        equirectangular PSD to save as a tif, went from 7mins to 7 seconds.
        Big difference :)

        Anyways I tried the standard methods of using the switch (google it)
        and they didn't work for me. Then I came across some other parameters
        to use, which got my system working fine. Here is the line that
        worked for me. The 'userva=3030' is the one to play with, and that I
        didn't have at all in the beginning. If it doesn't work try a smaller
        number like 2900 etc. (You will of course need 3gb or more of RAM
        before this will work)

        multi(0)disk(0)rdisk(0)partition(1)\WINDOWS="Microsoft Windows XP
        Professional 3GB" /fastdetect /3GB /USERVA=3030

        Hope that makes at least a little bit of sense to you :)

        Thanks,

        Aaron.
      • Chris Thomas
        Carel. This is pretty much what I do, except for # 3. I don t save 16 bit Master files out of Photoshop because. I ve been told; someone correct me if I m
        Message 3 of 11 , Jan 4, 2007
          Carel.



          This is pretty much what I do, except for # 3.



          I don't save 16 bit Master files out of Photoshop because.



          I've been told; someone correct me if I'm wrong, but "16 bit" PSD files are
          really only 15 bit,

          with the lost bit sacrificed for file size. 16 bit Tiffs are ok..



          I save all my Master "Capture" files as Adobe DNGs.. the only open source
          RAW format.



          After I've set the numbers in RAW, I don't need to make layered adjustments
          . so I can go

          to 8 bit for additional work and save the regular layered PSD as a "Worked
          Master".





          Cheers

          chris



          Chris Thomas

          Photographer

          cell... 403-615-1212

          In North America

          call... 1-800-870-5110

          <http://www.christhomas.com/> http://www.christhomas.com



          -----Original Message-----
          From: PanoToolsNG@yahoogroups.com [mailto:PanoToolsNG@yahoogroups.com] On
          Behalf Of Carel
          Sent: Wednesday, January 03, 2007 11:14 PM
          To: PanoToolsNG@yahoogroups.com
          Subject: [PanoToolsNG] OT: large files, RAM, 16bit and 8bit




          snip
          The below talks about a "middle way".

          http://www.extremet
          <http://www.extremetech.com/article2/0,1697,1910437,00.asp>
          ech.com/article2/0,1697,1910437,00.asp
          There is a middle way being adapted by digital photographers who work in Raw
          mode:

          1. Process the raw file in Camera Raw and then open it in Photoshop in
          16-BPC mode.
          2. Do any major tonal and color corrections.
          3. Save the file as a master file.
          4. Duplicate the document.
          5. Convert it down to 8-BPC.
          6. Do any major edits that require hefty resources, such as multiple
          adjustment layers, duplicate
          blending, layer styles, multiple type layers, and so on.
          7. Optimize the 8-BPC version for output.
          [Chris Thomas] snip



          [Non-text portions of this message have been removed]
        • Jim Watters
          ... Internally Photoshop uses a 15bit+1 (0 to 32768), 32769 discrete values per channel system. 15bit+1 was used over 16 bit for the purpose of increased
          Message 4 of 11 , Jan 4, 2007
            Chris Thomas wrote:
            > Carel.
            >
            > I don't save 16 bit Master files out of Photoshop because.
            > I've been told; someone correct me if I'm wrong, but "16 bit" PSD
            > files are really only 15 bit, with the lost bit sacrificed for file
            > size. 16 bit Tiffs are ok.
            >
            Internally Photoshop uses a 15bit+1 (0 to 32768), 32769 discrete values
            per channel system.
            15bit+1 was used over 16 bit for the purpose of increased speed.
            8 bit (0 to 255) has 256 discrete values per channel.

            Files saved by Photoshop are full 16bit.


            --
            Jim Watters

            jwatters @ photocreations . ca
            http://photocreations.ca
          • Paul D. DeRocco
            ... I think the 16th bit is discarded in order to make it easier to do the calculations and detect overflow. Anyway, there s no reason to worry about the
            Message 5 of 11 , Jan 4, 2007
              > From: Chris Thomas
              >
              > I don't save 16 bit Master files out of Photoshop because.
              >
              > I've been told; someone correct me if I'm wrong, but "16 bit"
              > PSD files are really only 15 bit,
              > with the lost bit sacrificed for file size. 16 bit Tiffs are ok..

              I think the 16th bit is discarded in order to make it easier to do the
              calculations and detect overflow.

              Anyway, there's no reason to worry about the difference between a 15-bit
              file and a 16-bit file. The amount of noise in any real-world image, even
              from the quietest drum scanner, leaves less than 15 bits of real
              information. Indeed, digicams don't even reach 12 bits. Point-and-shoots
              barely manage eight.

              --

              Ciao, Paul D. DeRocco
              Paul mailto:pderocco@...
            • Roger Howard
              ... Exactly; the math, according to Photoshop engineers like Chris Cox, is radically easier (and therefore much faster) in 15bits than 16, and it would be a
              Message 6 of 11 , Jan 4, 2007
                On Thu, January 4, 2007 10:04 am, Paul D. DeRocco wrote:
                >> From: Chris Thomas
                >>
                >> I don't save 16 bit Master files out of Photoshop because.
                >>
                >> I've been told; someone correct me if I'm wrong, but "16 bit"
                >> PSD files are really only 15 bit,
                >> with the lost bit sacrificed for file size. 16 bit Tiffs are ok..
                >
                > I think the 16th bit is discarded in order to make it easier to do the
                > calculations and detect overflow

                Exactly; the math, according to Photoshop engineers like Chris Cox, is
                radically easier (and therefore much faster) in 15bits than 16, and it
                would be a stretch to find any image quality difference. And to be clear
                (to the original poster), this 15bits math happens regardless of the file
                format - both PSD and TIFF can and do store full 16bit samples, but the
                least significant bit is dropped during processing. So file format choice
                won't make any difference.

                > Anyway, there's no reason to worry about the difference between a 15-bit
                > file and a 16-bit file. The amount of noise in any real-world image, even
                > from the quietest drum scanner, leaves less than 15 bits of real
                > information. Indeed, digicams don't even reach 12 bits. Point-and-shoots
                > barely manage eight.

                Absolutely; the only way to even detect the difference is with synthetic
                test images which are in no way representative of the real world image.

                And finally, the degradation that occurs due to all the various edits you
                perform in PS will far outweigh any practical loss due to 15bit
                operations. And that's assuming you have any image with significant data
                in that 16bit - you don't!
              • Carel
                ... I often shoot bracketed, so that the resulting tif probably IS a real 16 bit. But ignoring the bit that PS discards, Is there some kind of consensus on
                Message 7 of 11 , Jan 4, 2007
                  Paul D. DeRocco wrote:
                  >
                  >
                  > Anyway, there's no reason to worry about the difference between a 15-bit
                  > file and a 16-bit file. The amount of noise in any real-world image, even
                  > from the quietest drum scanner, leaves less than 15 bits of real
                  > information. Indeed, digicams don't even reach 12 bits. Point-and-shoots
                  > barely manage eight.
                  >
                  > --
                  >
                  > Ciao, Paul D. DeRocco
                  > Paul mailto:pderocco@...
                  >
                  >
                  >

                  I often shoot bracketed, so that the resulting tif probably IS a "real" 16
                  bit. But ignoring the bit that PS discards,
                  Is there some kind of consensus on what adjustments should be taken in 16
                  bit and what in 8 bit, if at all. As the Cubic pano output is 8 bit jpg
                  anyhow (is this true?) there should be a point on the workflow where one can
                  safely switch to 8bit(....?). The problem I see is that often some
                  adjustments that might be done in 8 bit mode, might need some final
                  adjustment in saturation or so, which should be done in 16bit (....?)

                  Carel

                  --
                  View this message in context: http://www.nabble.com/OT%3A-large-files%2C-RAM%2C-16bit-and-8bit-tf2917940.html#a8165427
                  Sent from the PanoToolsNG mailing list archive at Nabble.com.
                • Chris Thomas
                  Thanks to all.. for the details on the uselessness of the 16th bit I didn t know the Tiffs were also being converted to 15.glad it doesn t matter! Cheers chris
                  Message 8 of 11 , Jan 4, 2007
                    Thanks to all.. for the details on the uselessness of the 16th bit



                    I didn't know the Tiffs were also being converted to 15.glad it doesn't
                    matter!



                    Cheers

                    chris



                    Chris Thomas

                    Photographer

                    cell... 403-615-1212

                    In North America

                    call... 1-800-870-5110

                    <http://www.christhomas.com/> http://www.christhomas.com



                    -----Original Message-----
                    From: PanoToolsNG@yahoogroups.com [mailto:PanoToolsNG@yahoogroups.com] On
                    Behalf Of Roger Howard
                    Sent: Thursday, January 04, 2007 11:37 AM
                    To: PanoToolsNG@yahoogroups.com
                    Cc: panotoolsng@yahoogroups.com
                    Subject: RE: [PanoToolsNG] OT: large files, RAM, 16bit and 8bit



                    On Thu, January 4, 2007 10:04 am, Paul D. DeRocco wrote:
                    >> From: Chris Thomas
                    >>
                    >> I don't save 16 bit Master files out of Photoshop because.
                    >>
                    >> I've been told; someone correct me if I'm wrong, but "16 bit"
                    >> PSD files are really only 15 bit,
                    >> with the lost bit sacrificed for file size. 16 bit Tiffs are ok..
                    >
                    > I think the 16th bit is discarded in order to make it easier to do the
                    > calculations and detect overflow

                    Exactly; the math, according to Photoshop engineers like Chris Cox

                    [Chris Thomas] snip



                    [Non-text portions of this message have been removed]
                  • Hans Nyberg
                    ... Adjustments to images containing large flat areas, blue sky, large walls in same color etc, will easy get banding if you do it in 8 bit. This may be
                    Message 9 of 11 , Jan 4, 2007
                      --- In PanoToolsNG@yahoogroups.com, Carel <cs@...> wrote:

                      > I often shoot bracketed, so that the resulting tif probably IS a "real" 16
                      > bit. But ignoring the bit that PS discards,
                      > Is there some kind of consensus on what adjustments should be taken in 16
                      > bit and what in 8 bit, if at all. As the Cubic pano output is 8 bit jpg
                      > anyhow (is this true?) there should be a point on the workflow where one can
                      > safely switch to 8bit(....?). The problem I see is that often some
                      > adjustments that might be done in 8 bit mode, might need some final
                      > adjustment in saturation or so, which should be done in 16bit (....?)

                      Adjustments to images containing large flat areas, blue sky, large walls in same color etc, will
                      easy get banding if you do it in 8 bit.
                      This may be visible first after you compress the final image/movie

                      Hans
                    • Rick Drew
                      Thanks - I always wondered about that - I had that happen in a really nice blue sky and spent a lot of time cleaning it up. I never thought about it being the
                      Message 10 of 11 , Jan 4, 2007
                        Thanks - I always wondered about that - I had that happen in a really nice
                        blue sky and spent a lot of time cleaning it up. I never thought about it
                        being the being the bit depth, especially since the original was 8 bits!

                        Rick

                        Adjustments to images containing large flat areas, blue sky, large walls in
                        same color etc, will
                        easy get banding if you do it in 8 bit.
                        This may be visible first after you compress the final image/movie

                        Hans



                        [Non-text portions of this message have been removed]
                      • Roger Howard
                        ... There are two factors that Hans mentioned, and both can have major impacts on image quality. Bit depth - it s highly dependent on subject matter *and* the
                        Message 11 of 11 , Jan 4, 2007
                          On Thu, January 4, 2007 12:37 pm, Rick Drew wrote:
                          > Thanks - I always wondered about that - I had that happen in a really nice
                          > blue sky and spent a lot of time cleaning it up. I never thought about it
                          > being the being the bit depth, especially since the original was 8 bits!

                          There are two factors that Hans mentioned, and both can have major impacts
                          on image quality.

                          Bit depth - it's highly dependent on subject matter *and* the extent of
                          the edits you perform. For properly-exposed material it's quite possible
                          to get great results working only in 8bits per channel. There's also a
                          tradeoff here with color space gamut - the wider the gamut your color
                          space, the more likely you are to need to work in 16 bits per channel. So
                          someone working only in sRGB may do fine (with somewhat reduced gamut) and
                          not see much if any banding working in 8bits, but the same image
                          originating in ProPhoto or another wide-gamut color space may well cause
                          banding much more easily. Banding and other negative effects of limited
                          bit depth are easiest seen in areas of subtle gradations as Hans says.

                          Compression - this is a toughy. if you're outputting to a compressed
                          format like JPEG, in all likelihood you're outputting 8bits per channel at
                          the most to begin with. But the real problem is that the nature of lossy
                          compression will often make it hard to preserve things like subtle
                          gradations - skies, walls, etc - without them chunking up into visible
                          artifacts. This is less related to bit depth than compression strategies
                          though, and there are ways to minimize it. For instance, adding a bit of
                          noise in a nice sky or other gradation will actually *help* in many cases.
                          Note that, inherently, minimizing these artifacts typically means
                          increasing your target file size. There's no such thing as a free lunch!
                        Your message has been successfully submitted and would be delivered to recipients shortly.