3 Replies Latest reply on Jun 3, 2008 1:36 AM by (Pike_P)

    Profiling targets 24x36 vs 4 x 5

      Hi guys,

      I'm profiling an Epson V750 flatbedscanner for filmscanning/transparency. I'm using PMP5 and film IT8 targets in the size of 24x 36 mm and 4 x 5" with the correct refsfiles for each of them.

      The It8 targets are actually from bundled packages, no fancy hutch-stuff this time:
      - Nikon filmscanner for the 24 x 36, which is a Silverfast/W.Faust target.
      - Monaco for the 4 x 5 target, which is Kodak and this came with the V750 scanner.

      Ok, we're scanning the 4 x 5 target using Silverfast scannerprogram in the HDR mode. This is a completely RAW scan mode in 16 bit. All adjustments are locked out at this setting, no CMS is lurking beneath. Pure, pure RAW at 800 ppi even though the scanner can deliver a fantastic quality at 6400 ppi.

      What I've found out is that this profile does a very good job when assigned to the RAW IT8 target itself. From which it was built. The white patch on the trans IT8 4x5 sits around 1% - 3%. But when assigned to smaller positive trans, like a 24x36, this profile lifts highligts so that these blow out in the small trans. It is lifting the the upper quarter tone range of the ramp. It's a general thing, not just when assigned to the smaller IT8..

      I've done a bunch of scans and profiles where I did make 180 degrees turns of the target at the same spot as well as moved the target to a new location on the flatbed (in the holder of course). No big differences here. I've built about 20 profiles from different spots. One could suspect uneven ligt distrubution, but no. Very even and not anything to be concerned of at this level of scanning.

      When assigning the two profiles on the 24 x 36 I get these numbers in procent (easier to read the diffs than R-G-B values and the channeldiffs are of no interest in this matter):

      Patch, 24x36 profile, 4x5 profile assigned to the24 x 36, 4x5 with own profile

      0 2 % 1 % 2 %
      1 11 % 1 % 7 %
      2 17 % 7 % 12 %
      3 22 % 14 % 20 %
      4 29 % 22 % 26 %
      5 32 % 25 % 31 %
      6 39 % 34 % 36 %
      7 42 % 37 % 42 %
      8 46 % 43 % 45 %
      9 50 % 47 % 50 %
      10 55 % 53 % 54 %
      11 60 % 58 % 62 %
      12 62 % 61 % 66 %
      13 67 % 66 % 70 %
      14 72 % 71 % 74 %
      15 75 % 74 % 78 %
      16 79 % 78 % 82 %
      17 84 % 84 % 85 %
      18 87 % 87 % 87 %
      19 90 % 90 % 93 %
      20 92 % 94 % 96 %
      21 94 % 97 % 97 %
      B 96 % 97 % 97 %

      The question is why this is happening when assigning scannerprofiles built from a large tran like 4x5. At first glance it seems to be a good thing thing as it would work like a natural avaraging effect, since collecting over a larger area.

      In the zonesystem this was typical when one compared where to expect whiteclipping on paper as small filmsize clipped earlier on the ramp, but it feels a bit strange to put that thinking in this. We're just dealing with a filmbase and a refsfile who tells where the lighter tones are in the upper part.

      Yes, we can expect diffs and variables in "batch measured" targets like these. However when comparing the same 4 x 5 profile over both the smaller trans and the larger when assigned, it seems it's not that kind of variable in this case.

      Any thoughts?



      Edit: Ok, the table format screws up completely. The first is the Patch on the IT8, the second column is the value on the 24x 36 mm with it's own profile assigned. The third column is the 24x36 again with the 4x5 profile assigned, the fourth column is the 4x5 target with it's own profile assigned..... Sorry for the messy post.
        • 1. Re: Profiling targets 24x36 vs 4 x 5
          Level 1
          A follow up:

          Done a profile from the 4x5 target through Monaco EZ color, bundled with the Epson V750 scanner. At first it seems that a much lower allover lightness is the result when the Monaco built profile is assigned. Much lower quartertone part of the greyramp. Very clean ramp too.

          When converted from this scannerprofile to workingspace (Adobe RGB) using perceptual rendering things goes lighter again. Not as light as using PMP5 scannerprofiling module. Using Relative this raise of quartertones doesn't happen, the lower allover lightness is kept when arriving into workingspace. As well as the lower white level. Bumping up the white level with curves/levels and still a bit lower quarterpart is kept a bit better than with perceptual rendering. The saturation is bumping up by nature but not as clean and punchy as with the perceptual. A bit more muddy in the darker patches of colors.

          The interesting thing is that this difference between perceptual vs relative is not a case with Profilemaker 5. These profiles is more like been through the effect the perceptual creates when moved into a workingspace! So, when converting such a file from a scannerprofile built in PMP5 into workinspace no bigger difference can be seen regardless of perceptual or relative rendering.

          So far, it looks like this; Monaco EZ color seems to do a little bit better job using the 4x4 target than PMP5. Not blowing out the lighter tones as much when assigned to the smaller 24x36 tran which has been my main concern. Still, not as good as I would expect. The strange thing is the difference between relative vs perceptual when Monaco is the host for building the profile. It also seems that PMP has a harder time to deal with yellows mostly in the more saturated parts. Something that we know happens very much in printprofiling with Profilemaker. It looks like the non-pro Monaco EZ beats Profilemaker Pro 5.

          I'm would be happy to hear anyones thoughts and experiences with these effects.

          • 2. Re: Profiling targets 24x36 vs 4 x 5
            Level 1
            ProfileMaker 4 scanner profiles supported Perceptual rendering when converting but v5 profiles do not. It's not a really big deal as the difference was very small and the v5 profiles are better overall. For overall accuracy, the best results are generally from PM5 and a HutchColor target. I use the Velvia Hutch on my Howtek and it works very well.

            A little confused over your use of percentages in your first post. What are you referring to? You should be reading Lab value not percentages and what are your percentages of???
            • 3. Re: Profiling targets 24x36 vs 4 x 5
              Level 1

              using perceptual rendering when converting with the monaco created profile gives a lightness bump while relC doesn't. There is a diff between the two renderings in Monaco but not in the Profilemaker conversion. The funny thing is that it's the Perceptual in Monaco that looks closest to Profilemaker, while Perceptual isn't supported in PMP5. ?? Should be the other way around, no?

              Sure, Labvalues are what to read, the procentvalues are just a take from the info palette set on greyscale while moving over the greyramp of the open IT8 target in Photoshop. An old habit to have a % readout set up in one corner of the palette. I didn't bother to switch over when I began. I just kept using the procent when it got hot. A bit embarrassing. Doesn't matter with any exact lab values, just a comparison of the relation-diff in the upper part of the ramp between the profiles when assigned to the targets. Simple as that. But I've redone it with L values.

              P 24x36 target with own profile 24x36 with 4x5 profile 4x5 with own profile

              0 98 99 98
              1 90 98 94
              2 85 94 90
              3 80 88 84
              4 74 81 77
              5 71 78 73
              6 66 70 68
              7 62 67 62
              8 58 61 59
              9 54 57 53
              10 49 51 50
              11 44 46 46
              12 41 42 41
              13 36 37 37
              14 30 31 32
              15 26 27 28
              16 21 23 23
              17 14 17 17
              18 13 15 14
              19 10 9 9
              20 6 5 4
              21 4 2 1
              22 2 1 1
              B 1 1 1

              The things that surprised me were that the 4x5 target based profiles built in PMP5 blow up the higher end of the greyramp on smaller trans. Why? It did a good job when assigned to the 4x5 target which to me seems to confirm that the refsfile isn't what's screwed here and the cause of the ******** (scuse me, coulden't resist ;) on the smaller trans 24x36. I'm interested in to hear if this is something anyone else is dealing with too. I'm comparing this to the profiles built from the smaller 24x36 IT8 target. It shoulden't be this diff. I just can't get it together.

              A part from that, again, a Monaco created profile keeps the upper part of the ramp a bit better better and gives much smoother transitions in, especially, the yellows where the PMP5 profiles actually gives some banding. I'm really surprised. Considering this is from the big 4x5 target, too. I have not tested the "larger" profile option, yet , in Profilemaker. So far, the default sized setting has been in use. Anyway, as we shoulden't edit scanner/inputprofiles they shoulden't show up these bigger differences. Differences in transitions and smoothness, fine & ok. But not the neutral tone ramp and where the lightness of the same patches is located in a standard IT8. I doubt that this is a matter of the quality of the target itself or the refsfile.

              After all, we're doing a characterizing of a scanner with fixed settings, fixed lightsource. To find a "true" rendering from the scanner as far as possible. The variables by nature of printer profiles or cameraprofiles are something else. It's much less involved in scannerprofiling.

              I'm just a bit surprised, I thought the 4x5 target would be the way to go with PMP5. I'm happy to hear the Hutch target works fine on the drum. I can't understand why the cheaper target doesnt work better than they do with PMP5 on a decent flatbed like the v750. I'll use the Monaco profile for now. The Monaco EX color pogram refuses the refsfile for the smaller 24x36 since it's not pure Monaco stuff, but Lasersoft. So, I couldent test that one. That's the crazy stuff in the Monaco camp. It's always been more closed and not so openminded as Profilemaker.