26 Replies Latest reply: Dec 27, 2011 5:03 PM by Bollo1 RSS

    10-Bit color calibration problem

    Bollo1 Community Member

      I have the following setup:

       

      Dual HP ZR30w Wide LCD Monitors, each connected via Display Port to each below card.

      Dual ATI FirePro v7800 (FIreGL) display adapters

      Latest FirePro drivers from the AMD website (file name: FirePro_8.85.7.2_Win7x64_125182.exe)
      Win 7 Pro 64-bit
      Spyder3Elite 4.0.2

       

      When I enable 10-bit color in the ATI drivers, and use the Spyder 3 package to perform

      a calibration the profile is absurd --- light yellows render as light green, off whites become

      purple...  If, in 10-bit I switch to an old profile generated when running 8-bit I get reasonable

      but slightly off color.  Running 8-bit with an 8-bit calibration works perfectly to my eye.

       

      Colorvision claims there is no problem, that the above ATI/HP pipeline is really 8-bit.  Here

      is the quote from their tech support:

       

      "the option you deactivated is not a 10bit communication between your graphic card and your monitor.

      That's just a 10bit LUT/mode in your video card which will be reduced to 8bit output to your monitor.

      Please understand we can't give you a fix for a 10bit pseudo wide color depth mode of your video card.

      The output is 8bit and that's all. The only advantage is that you could tweak color with less reduction.

      But this has nothing to do with the industrial standardized ICC profiling system."

       

      Does what they say make any sense?  My understanding of the HP monitors and ATI cards is that

      it is a true 10-bit pipeline.

       

      If I do have a true 10-bit pipeline, what can I do to run in that mode and have well calibrated monitors?

       

      I contacted integrated color, as they have software that will work with my Spyder 3 sensor and they

      said I would not see any benefit from 10-bit monitor color and their software does not support 10-bit

      mode.  Does that ring true?

       

      Thanks in advance,

      Andrew

        • 1. Re: 10-Bit color calibration problem
          Noel Carboni Community Member

          Try one Photoshop config change that may help:

           

          Go into Edit - Preferences - Performance, see that the Enable OpenGL Drawing setting is checked, then click the [Advanced Settings...] button.  Change it to Basic mode then OK out and close and restart Photoshop.

           

          Assuming you were in Normal or Advanced mode before, this change will cause the color management transforms to move from being done in the GPU to being done in the CPU.  I'm not sure what may be wrong, but it would be interesting if this were to help.

           

          -Noel

          • 2. Re: 10-Bit color calibration problem
            Bollo1 Community Member

            Thank you Noel,

             

            I should have been more specific, the bad calibrations affects everything, not just PS.  This is a general

            color calibration problem, not a photoshop specific problem.

             

            Should I be asking this question in another forum?  I did not see a forum specific to monitor calibration,

            and thought that the folks here were liable to be very up on calibration issues.

             

            Thanks again,

            Andrew

            • 3. Re: 10-Bit color calibration problem
              Noel Carboni Community Member

              If it's affecting everything then I'd say it's most likely a display driver problem.

               

              Have you downloaded the latest drivers from the amd.com web site?

               

              -Noel

              • 4. Re: 10-Bit color calibration problem
                Bollo1 Community Member

                Yes, as stated in my setup list they are the latest drivers.

                 

                I think there is some conflict between the drivers and the Spyder software.

                 

                But I was looking for comments on the verity of the Spyder statement that there was

                not a true 10-bit pathway in the above described setup.  And of course for a workaround

                if I do have a true 10-bit pathway (which I paid a lot to get, and am annoyed that I cannot

                use).

                 

                Thanks,
                Andrew

                • 5. Re: 10-Bit color calibration problem
                  Noel Carboni Community Member

                  You'll find that there are VERY few folks who have actually achieved 10 bit pipeline operation, and information from companies like what you quoted above from Colorvision may be flawed.

                   

                  I can only suggest searching this forum for "10 bit".  There have been threads occasionally here, some of them long, about how people have achieved higher quality 10 bit display pipelines.

                   

                  -Noel

                  • 6. Re: 10-Bit color calibration problem
                    Hudechrome Community Member

                    Since I am un an upgrade path to monitors, and since the one to which I am zeroing has 10 bit capability I have been on a pursuit of any current information, and I have to say there is a dearth of solid info, even to the level of whether 10 bit capability exists in any of the cards on the market. I saw a list from ATI that said 10 Bit for all their current cards, which, ostensibly, includes my current card, but I doubt it, or if it does exist, cannot be invoked.

                     

                    So then, should I consider the higher priced monitor over a similar monitor at less cost where the significant difference seems to be the 10 bit capability of the expensive version?

                    • 7. Re: 10-Bit color calibration problem
                      Noel Carboni Community Member

                      I had heard that "all their Fire-series cards" support 10 bit color.  Where did you see the more general statement?

                       

                      -Noel

                      • 8. Re: 10-Bit color calibration problem
                        Hudechrome Community Member

                        On the AMD website for ATI. I must confess I have not found it since, otherwise I would have posted a link.

                         

                        If I do this monitor, I also need to change cards to one that has DisplayPort. Eizo makes a significant point about it.

                        • 9. Re: 10-Bit color calibration problem
                          Noel Carboni Community Member

                          I wonder if someone was being overly exhuberant.

                           

                          This is what I saw, and it might help Bollo1 to read through it as well:  http://www.amd.com/us/products/workstation/graphics/software/Pages/adobe-photoshop.aspx

                           

                          -Noel

                          • 10. Re: 10-Bit color calibration problem
                            Hudechrome Community Member

                            I saw that as well some time ago. Interesting that one has to enable the 10 bit in Catalyst Control Center, which isn't installed here; only the driver.

                             

                            A FirePro is available for about $150 or so, which will add to the cost. The Eizo I am contemplating is on their list.

                             

                            I also looked at the banding image and frankly, I don't see banding here. Even going to 100% and using the move tool did not show any banding.

                            • 11. Re: 10-Bit color calibration problem
                              Noel Carboni Community Member

                              Doesn't sound like you really need 10 bit color.  I'd say try it with what you have now before assuming you do.

                               

                              Don't I recall that the EIZOs are calibrated internal to the monitor using high-accuracy tables?  If that's the case, and assuming I haven't got my concepts wrong, what *I'd* try to do is try to calibrate the EIZO to be a near-perfect sRGB reproduction device, then just use the (well crafted and tested) sRGB color profile on the computer, along with center-scale (default) video card calibrations.  That should conceivably use every possible level from within the 16 million possible colors.  That should minimize posterization to be nearly undetectable in most use.

                               

                              That said, I can clearly see the grayscale steps in Ramp.psd.

                               

                              -Noel

                              • 12. Re: 10-Bit color calibration problem
                                twenty_one Community Member

                                Hudechrome wrote:

                                 

                                I also need to change cards to one that has DisplayPort. Eizo makes a significant point about it.

                                And not just Eizo. DisplayPort is a requirement for 10 bit; DVI doesn't have the bandwidth.

                                 

                                What I'd like to know is whether it's possible to calibrate to the video card at all using 3rd party calibrators (as per the OP's original question). If the people at Integrated Color said their software doesn't support 10 bit I'd take their word for it (they know these things; calibration is all they do). And if not, you're limited to Eizo and NEC since they have hardware calibration to the monitor LUT, communicated via USB cable.

                                 

                                The other possibility, using the DDC protocol to calibrate to the monitor LUT, is not supported with DisplayPort. But DDC rarely works anyway, even when it's supposed to.

                                 

                                All in all, the safest option (again) seems to be Eizo ColorEdge or NEC PA.

                                • 13. Re: 10-Bit color calibration problem
                                  Noel Carboni Community Member

                                  D Fosse wrote:

                                   


                                  What I'd like to know is whether it's possible to calibrate to the video card at all using 3rd party calibrators (as per the OP's original question).

                                   

                                  It might be interesting, in this same vein, to know whether the Catalyst Control Center for the Fire-Pro card offers the user color calibration settings.

                                   

                                  As a side note, I've noticed the way AMD provides such settings is changing from release to release, by the way.  There's less calibration functionality directly available today than there has been in the past with the very same video card.

                                   

                                  -Noel

                                  • 14. Re: 10-Bit color calibration problem
                                    twenty_one Community Member

                                    It seems 10 bit color is still in its infancy and in most configurations it simply doesn't work for one reason or another. Perhaps better to just wait it out until the dust settles.

                                     

                                    On the Mac side it doesn't work at all. Apparently it hinges on Apple drivers. Some background: http://forums.adobe.com/thread/736938?tstart=0

                                    • 15. Re: 10-Bit color calibration problem
                                      Hudechrome Community Member

                                      All those pros and cons are exactly what I am weighing. Add to that the blogs proporting to know what they are doing spouting nonsense. Like an NEC monitor for $500 being better than the SX 24 model. The Eizo covers 98% of Adobe RGB, the NEC, 72%!

                                       

                                      Huh?

                                       

                                      Thanks for the reminder about DisplayPort. I did read that it is necessary due to bandwidth limitations.

                                       

                                      I want a near perfect Adobe RGB, and not stop at sRGB. I've seen the difference, and so does the printer. So, I will continue with user calibration, for which the i1 works well with their software, (so they tell me!) and pick the color space in Photoshop. I can't imagine a well calibrated system displaying close to 100% Adobe RGB will be hobbled at sRGB. Again, the choice is in Preferences, where it belongs, imho. .

                                       

                                      Looking into the gradation problems especially at the low end of the scale, I am better off with a dense LUT, like 12 or 16 bit, than worrying about LED vs CCFL.

                                       

                                      Of course I need my cash flow restored! More later, hopefully!

                                      • 16. Re: 10-Bit color calibration problem
                                        Noel Carboni Community Member

                                        Hudechrome wrote:

                                         


                                        I want a near perfect Adobe RGB, and not stop at sRGB.

                                         

                                        Same concept applies, just substitute Adobe RGB 1998 where I wrote sRGB.  The point being that (assuming you want to WORK in the Adobe RGB color space) that no transform need be done in the CPU or GPU.

                                         

                                        -Noel

                                        • 17. Re: 10-Bit color calibration problem
                                          Hudechrome Community Member

                                          How is that different than doing a user calibration and selecting whatever color space in I wish in Color Settings? I actually have been using ProPhoto but experience shows that Adobe RGB does all I can see with the printer I now use.

                                           

                                          As you may recall, I am wary of setting a monitor profile to any published color space as a calibration/profiling opertion. It is far preferable to obtain the widest native gamut of the monitor, consistent with the profiling standards applied. Even now, with the crt on it's way out, I obtain a dE average at 1 or better, with no peaks exceeding 2.

                                          • 18. Re: 10-Bit color calibration problem
                                            Noel Carboni Community Member

                                            Hudechrome wrote:

                                             

                                            It is far preferable to obtain the widest native gamut of the monitor

                                             

                                            You can't make that statement as an absolute.  I can make arguments otherwise, but I'd rather not go down that road right now.

                                             

                                            But I'm perfectly willing to accept that what you've written is YOUR goal.

                                             

                                            You have stated your prospective monitor's gamut is very close to Adobe RGB 1998.  If in-monitor calibration makes it a good match to Adobe RGB 1998, and in the context of an 8 bits/channel color pipeline, then no calibration or profile-based transform would be required when you edit in the Adobe RGB 1998 color space.  Thus, given the limited bit depth, the opportunities for those processes to cause (visible) 2 level jumps in the output would be eliminated, and you'd have the best possible display an 8 bit pipeline could produce.

                                             

                                            I don't know how the whole Eizo in-monitor calibration process works; what I described may be how it works right out of the box.  It would make sense.

                                             

                                            -Noel

                                            • 19. Re: 10-Bit color calibration problem
                                              Hudechrome Community Member

                                              I see your point, and it's well taken.

                                               

                                              The out of box spec at 98% Adobe and 102% (or so) NTSC is just that, out of the box. Reading some of the better blogs like Luminous Landscape, I note that careful profiling can produce a gamut beyond Adobe RGB, and that only makes sense: If I am going to guarantee a measurement accuracy say at 3%, I'm calibrating to a much tighter spec. In order to achieve any defined color space, the available gamut to do so needs to exceed that, as is visible in color space diagrams defining the various (sRGB etc) color space triangles.

                                               

                                              That is the basis for my philosophy, honed over many years of instrument calibration where these numbers must be backed up by at least secondary standards.

                                              • 20. Re: 10-Bit color calibration problem
                                                Bollo1 Community Member

                                                Using the 8-bit path, the Spyder 3 elite software reports a final gamut very close to Adobe RGB.  The graphs

                                                of my gamut versus Adobe RGB are slightly offset, but at least to the eye, the same size.  Both widely engulf

                                                the sRGB.

                                                 

                                                Actually the software reported the when doing the calibration running in 10-bit mode, but as I said at the beginning

                                                of this thread, the profiles were totally absurd. 

                                                 

                                                Thanks to all above, I will follow the links mentioned and report back if I ever have any luck on the 10-bit path.

                                                 

                                                Andrew

                                                • 21. Re: 10-Bit color calibration problem
                                                  Bollo1 Community Member

                                                  OK, I went to the AMD site (http://www.amd.com/us/products/workstation/graphics/software/Pages/ado be-photoshop.aspx) and I had used

                                                  their instructions for enabling 10-bit.

                                                   

                                                  I downloaded the referenced ramp.psd (which has an embedded sRGB profile for some reason) and did indeed see banding running in 8-Bit

                                                  mode (I do have OpenGL mode enabled in PS CS5.1.  I do not think I will have time to switch to the 10-Bit mode until tomorrow or Saturday,

                                                  but I think if the banding goes away that will prove the 10-Bit path works.

                                                   

                                                  That will isolate my problem to calibration problems.

                                                   

                                                  Does anyone see a flaw in my logic?

                                                   

                                                  Thanks,

                                                  Andrew

                                                  • 22. Re: 10-Bit color calibration problem
                                                    Noel Carboni Community Member

                                                    Sounds right to me, Andrew.

                                                     

                                                    -Noel

                                                    • 23. Re: 10-Bit color calibration problem
                                                      twenty_one Community Member

                                                      It should also be noted here that banding is potentially a bigger problem with wide gamut, since the values are farther apart. So while 10 bit is perhaps not strictly needed, it has become more desirable.

                                                       

                                                      I see clear banding on that test image, on my wide gamut/8 bit Eizo S2243.

                                                      If in-monitor calibration makes it a good match to Adobe RGB 1998, and in the context of an 8 bits/channel color pipeline, then no calibration or profile-based transform would be required when you edit in the Adobe RGB 1998 color space.

                                                      This is where I still don't agree (but we probably shouldn't go there again ) I concede that my previous position was a bit...ah, inflexible. Mainly that's because I've since realized that there are many variables that need to be brought under strict control before you can make any claims to accuracy.

                                                       

                                                      So let's just say it's about peace of mind. There are many unknowns in the equation above, and it seems to me the best approach is to pin down as many as you can.

                                                       

                                                      As I write this, I'm on my way to look at proofs for a book I've supplied 80% of the photographs for. I've done what I can, I have peace of mind, and still...

                                                      • 24. Re: 10-Bit color calibration problem
                                                        Hudechrome Community Member

                                                        D Fosse wrote:

                                                         

                                                        ''''there are many variables that need to be brought under strict control before you can make any claims to accuracy.

                                                         


                                                        Great truth in that, as any engineer who is responsible for the measurement claims of their instrument.

                                                         

                                                        I once did a series of measurements in laser diodes. We needed to know the output power, among others. I had three meters, all calibrated and certified to be within 3% that, nevertheless showed a spread of 10% between them! Worst case should have been 6% and even then I would have grumbled.

                                                         

                                                        I became a very sharp thorn in the side of the calibration lab. We had to resort to calorimeter measurements to decide who what right.

                                                         

                                                        They were all outside their spec.

                                                         

                                                        It's called CYA!

                                                        • 25. Re: 10-Bit color calibration problem
                                                          Noel Carboni Community Member

                                                          D Fosse wrote:

                                                           


                                                          This is where I still don't agree (but we probably shouldn't go there again )

                                                           

                                                          No worries, I'm not interested in arguing over fine points of color management either.  Bygones are bygones, and I'm happy you're back here contributing, Dag.

                                                           

                                                          -Noel

                                                          • 26. Re: 10-Bit color calibration problem
                                                            Bollo1 Community Member

                                                            >OK, I went to the AMD site (http://www.amd.com/us/products/workstation/graphics/software/Pages/ado be-photoshop.aspx) and I had used

                                                            >their instructions for enabling 10-bit.

                                                             

                                                            >I downloaded the referenced ramp.psd (which has an embedded sRGB profile for some reason) and did indeed see banding running in 8-Bit

                                                            >mode (I do have OpenGL mode enabled in PS CS5.1.  I do not think I will have time to switch to the 10-Bit mode until tomorrow or Saturday,

                                                            >but I think if the banding goes away that will prove the 10-Bit path works.

                                                             

                                                            OK, sorry for the delay, busy season and all that.

                                                             

                                                            I finally did the test, and sure enough there is absolutely no visible banding on ramp.psd when viewed in PS CS5.1 when running the ATI

                                                            drivers in 10 bit mode.

                                                             

                                                            I think this proves that DataColor's techs are either ignorant or lying.

                                                             

                                                            Thank you all for your help. 

                                                             

                                                            Andrew