Pages: [1] 2   Go Down

Author Topic: Is i1Profiler scrambling random or deterministic?  (Read 8973 times)

howardm

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1984
Is i1Profiler scrambling random or deterministic?
« on: June 14, 2014, 10:34:27 am »

Using an i1Pro2 w/ i1Profiler 1.5.4 on a Mac.

For printer profiling, if I create a patch set, say 1224 (max # on 2 letter sheets) and I 'scramble' them and save the .tif targets. 

If I were to make another 1224 patch set (all params the same) and scramble them, am I assured that the scramble layout will or will not be the same?

Ie, I am asking if the scramble layout is random every time or whether it's the same (uses the same random seed/sequencing).  I ask because I did not save the original patch set and test chart assets but want to re-measure the existing charts and create a new profile for a different illuminant.  If the scramble layout is different, then I don't think I can use the existing charts but must start from scratch.

Thanks,
Howard

howardm

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1984
Re: Is i1Profiler scrambling random or deterministic?
« Reply #1 on: June 15, 2014, 11:41:25 am »

I think I have answered my own question.  I generated a 'new' 1224 patch set & test chart layout but then measured the *old* 2 sheet print out and all the measurements  were 'accepted' by i1Profiler and it generated a new profile which was, from a volume/shape POV, 99+% the same so I'll assume a patch set of X colors will always generate the same physical layout or at least i1Profiler can deal with it.

I'm guessing the measurements would have failed if the measured color was wildly off as it would be if it was scrambled differently.

I also made a profile using Illuminant A and will be printing up the Outback image and we'll see how it looks vs. the D50 under my display conditions.

darlingm

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 361
    • Westland Printworks
Re: Is i1Profiler scrambling random or deterministic?
« Reply #2 on: June 15, 2014, 07:29:23 pm »

As far as I've seen, the same number of patches on the same size page will deterministicly scramble.  (As if they, as you proposed, were randomizing but re-using the same seed - but who knows what goes on internally.)  As a side note, after switching to argyll, it allows you to randomize/scramble and specify a seed.  I always use seed 0 for repeatability, but still helping with measurement errors.
Logged
Mike • Westland Printworks
Fine Art Printing • Amazing Artwork Reproduction • Photography
http://www.westlandprintworks.com • (734) 255-9761

Some Guy

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 729
Re: Is i1Profiler scrambling random or deterministic?
« Reply #3 on: June 16, 2014, 12:38:47 am »

I believe you can save all of the various profiles and patch charts and call them up to read later.  Allows for some drying time that way too, or so they told me.  Lots of big image files though as well as data.  It uses a lot of different extensions as well.  Seems each screen allows you to save, although the extension and file type might also change too.

SG
Logged

howardm

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1984
Re: Is i1Profiler scrambling random or deterministic?
« Reply #4 on: June 16, 2014, 06:52:31 am »

I believe you can save all of the various profiles and patch charts and call them up to read later.  Allows for some drying time that way too, or so they told me.  Lots of big image files though as well as data.  It uses a lot of different extensions as well.  Seems each screen allows you to save, although the extension and file type might also change too.

SG


Yes, I had measured and created a 1224 patch profile for a certain paper (Canson Baryta) but then did not save patch/test chart/measurement set.   Then I wanted to create a new profile for a different illuminant from the same paper chart.  I haven't spent any time really looking through the resulting .mxf measurement or the xml files that define the patch/test charts

I'm starting to standardize on 1215 as compromise between using only 2 sheets at the smallest patch size and a patch set that seems to give significant number of 'near' grey patches along the grey axis (vs. the 1224 oddly enough).

howardm

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1984
Re: Is i1Profiler scrambling random or deterministic?
« Reply #5 on: June 16, 2014, 08:50:52 am »

Here is Xrite's response in case anyone wants to know :)   That last paragraph is really cool and unexpected.


The algorithms for scrambling the patch set are OS specific, which is to say a scrambled patch set made on a Mac, must be read by a Mac.

If you are concerned about being able to use a target cross-platform, we would recommend that you not choose the scrambled target layout.

We're not aware of any changes in the patches if you create two targets with identical patch counts and identical page layout settings.

If you save the original profile, the patch set and page layout information is stored in the profile itself. You could drag the original profile into the profiling workflow and it will populate all the settings you used originally.

samueljohnchia

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 498
Re: Is i1Profiler scrambling random or deterministic?
« Reply #6 on: June 16, 2014, 08:56:57 am »

If you save the original profile, the patch set and page layout information is stored in the profile itself. You could drag the original profile into the profiling workflow and it will populate all the settings you used originally.

Yes, I know this and do this all the time, but it will not work for patch sets saved using V1.3.2 and loaded into V1.5 or later. The page layout information specifically is lost. You must manually key in the correct information again. Patches remain the same.

If you want OS independent scrambling use ColorLab to scramble and build the target, and load the resulting patch file as unscrambled in i1Profiler.
Logged

howardm

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1984
Re: Is i1Profiler scrambling random or deterministic?
« Reply #7 on: June 16, 2014, 09:27:24 am »

being a relative newcomer, I had to dig to find out what colorlab was  :-\

Seems like another piece of abandonware from Xrite/GM and it won't run on any vaguely modern Mac (10.7+).  I have a 10.4
laptop in storage but I suppose I'll use scrambling 'carefully'.

samueljohnchia

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 498
Re: Is i1Profiler scrambling random or deterministic?
« Reply #8 on: June 16, 2014, 09:36:58 am »

Sorry, I should have included a link. X-rite is not a fun playground. Since you are using the i1 Pro 2, as a fellow user, I'll stick with a non-scrambled target. Less measurement errors. Use larger patches. Scan slower (4 seconds per pass) and at a regular pace. Beware of the measurement bug with V1.5.4 (if you did an upgrade from an older version) where whites or light patches measure as significantly darker in scan mode than in spot mode.
Logged

howardm

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1984
Re: Is i1Profiler scrambling random or deterministic?
« Reply #9 on: June 16, 2014, 09:49:43 am »

Sorry, I should have included a link. X-rite is not a fun playground. Since you are using the i1 Pro 2, as a fellow user, I'll stick with a non-scrambled target. Less measurement errors. Use larger patches. Scan slower (4 seconds per pass) and at a regular pace. Beware of the measurement bug with V1.5.4 (if you did an upgrade from an older version) where whites or light patches measure as significantly darker in scan mode than in spot mode.

Thanks.  I'd like to try to keep it to as many patches as feasible on 2 sheets.  Seems like the absolute most is 1225 at minimum size.  I was not aware of the v1.5.4 bug.  Is there a workaround w/o resorting to spot (which I don't yet have a grip on)

samueljohnchia

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 498
Re: Is i1Profiler scrambling random or deterministic?
« Reply #10 on: June 16, 2014, 10:29:07 am »

Thanks.  I'd like to try to keep it to as many patches as feasible on 2 sheets.  Seems like the absolute most is 1225 at minimum size.  I was not aware of the v1.5.4 bug.  Is there a workaround w/o resorting to spot (which I don't yet have a grip on)

If you do not wish to use more or larger sheets of paper (a pity) I would recommend going for a smaller patch set where the smart patch generator maximises the number of neutrals. 1225 does not. For example try 815. Then go to 816 and see that all the neutrals suddenly disappear. You have to slowly work out what all the neutral-maximising patch counts are. I had it worked out before but can't remember the formula at the moment. Trial and error then.

I've been at war with X-rite on this for months. In the end it was easiest for me to reformat my computer to ensure a clean install. I'm back on V1.3.2 and its corresponding XRD which seems to be free of this issue as far as my experiments go, and all my old measurement files which I religiously saved since 2012 also indicate that it is problem free. V1.5.4 works fine for me on my laptop (it was an upgrade install, not a clean install), but one user here seems to have the bug still, which I think is related to an XRD problem.

You can also use Argyll to measure, or ColorPort which only measures in M2, but allows you to save out spectral data in a format that i1Profiler will accept.
« Last Edit: June 16, 2014, 10:30:42 am by samueljohnchia »
Logged

howardm

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1984
Re: Is i1Profiler scrambling random or deterministic?
« Reply #11 on: June 16, 2014, 11:29:51 am »

Right, I noticed that 1225 patches did NOT have much in the way of grey so I backed the count down to 1215 (?) and that is when the greys just re-appeared.  Adding even 1 more patch again made the grey's disappear.  Also, it seems like the chart generator won't really make any smart usage of 13x19 paper for an i1Pro2 and just wastes most of the sheet.  It appears as though 2 8.5x11 is a bit of a sweet spot.

digitaldog

  • Sr. Member
  • ****
  • Online Online
  • Posts: 20646
  • Andrew Rodney
    • http://www.digitaldog.net/
Re: Is i1Profiler scrambling random or deterministic?
« Reply #12 on: June 16, 2014, 04:08:18 pm »

Quote
The algorithms for scrambling the patch set are OS specific, which is to say a scrambled patch set made on a Mac, must be read by a Mac.
That's just idiotic!
Quote
If you are concerned about being able to use a target cross-platform, we would recommend that you not choose the scrambled target layout.
That too is idiotic as a fix to something that should not have been designed that way (or no one noticed).
Quote
If you save the original profile, the patch set and page layout information is stored in the profile itself. You could drag the original profile into the profiling workflow and it will populate all the settings you used originally.
Yes, that's been there from day one and carries over from GretagMacbeth's ProfileMaker Pro. Now you know why their profiles are so big too (lots of EXIF data you can strip out).
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

Some Guy

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 729
Re: Is i1Profiler scrambling random or deterministic?
« Reply #13 on: June 16, 2014, 09:56:25 pm »

I've been having issues with different readings of the white patches in spot verse scan mode with iProfiler 1.5.4 on just a B&W 21-step gray scale.

I sent in a head diagnostic log off their newest i4Diagnostics (April version) which passed the test.  Suggestion today was to use a better (more current) powered USB port to drive the i1 PhotoPro 2 head.  Okay, so I got a 3 amp one at Radio Shack for $30.  Same results:  In spot mode the whites read higher (L=94), and in scan mode they read lower (L=86).  I also tried placing a thin black bar between each patch and the scan mode improved to L=90, but still the spot mode wins for consistency and accuracy (I hope!).

How much error this works out to in analysis of the 4 pages of color patches for normal ICM profile making I don't know, but probably not good.  Seems a lot of my most recent profiles made have a very condensed wire-frame gamut over some older ones in the iProfiler viewer.

Something is weird somewhere, and this was with Windows 8.1 64-bit too, fwiw.

Others need to report any issues to x-rite to get this fixed.  I gave it my shot today.

SG
Logged

samueljohnchia

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 498
Re: Is i1Profiler scrambling random or deterministic?
« Reply #14 on: June 16, 2014, 10:37:41 pm »

Right, I noticed that 1225 patches did NOT have much in the way of grey so I backed the count down to 1215 (?) and that is when the greys just re-appeared.  Adding even 1 more patch again made the grey's disappear.  Also, it seems like the chart generator won't really make any smart usage of 13x19 paper for an i1Pro2 and just wastes most of the sheet.  It appears as though 2 8.5x11 is a bit of a sweet spot.

I don't know why you are finding that smaller sheet sizes are better, since I can happily squeeze 1457 patches into one 13x19 sheet, minimal wastage. Smallest patch size though. Small sheets waste a lot more due to the margin requirements.

I looked at 1215 patches again and it contains unevenly sampled intervals in the RGB cube, 10x11x10. I prefer to use patch sets where the RGB sampling is even, like in the case of 1457, its 11x11x11, or 1105, 10x10x10.

That's just idiotic!

Thanks Andrew! Enough said.
Logged

samueljohnchia

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 498
Re: Is i1Profiler scrambling random or deterministic?
« Reply #15 on: June 16, 2014, 10:42:40 pm »

Others need to report any issues to x-rite to get this fixed.  I gave it my shot today.

Thanks Some Guy. I am staggered that the thousands (or tens of thousands) of i1 Pro/i1 Pro 2 users out there haven't already experienced this problem and complained yet. Maybe other folks just don't have the tools to look at their profiles properly. Maybe they don't know how to read the measurement data. I discovered this late because I am usually slow to upgrade to later versions of software if the current one is working. So I was on V1.3.2 for much too long, oblivious to the measurement bug.

For the record, I've sent X-rite two emails in the last fourteen days and they have gone ignored.
Logged

Some Guy

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 729
Re: Is i1Profiler scrambling random or deterministic?
« Reply #16 on: June 16, 2014, 11:13:22 pm »

Thanks Some Guy. I am staggered that the thousands (or tens of thousands) of i1 Pro/i1 Pro 2 users out there haven't already experienced this problem and complained yet. Maybe other folks just don't have the tools to look at their profiles properly. Maybe they don't know how to read the measurement data. I discovered this late because I am usually slow to upgrade to later versions of software if the current one is working. So I was on V1.3.2 for much too long, oblivious to the measurement bug.

For the record, I've sent X-rite two emails in the last fourteen days and they have gone ignored.

You're welcome.

I only happened to get into this due to some B&W step chart I was having issues with printing.

I was doing spot readings for so long (50 per paper change and brand) and then decided I would try Keith Cooper's 21 step-scans for the i1 software to speed up the process, and get away from slower spot reading method as used with Colorport or the spot readings in i1Profiler's newest software.   Then I found the readings were a lot lower than what I normally see in papers via the scan, way down to L=86, and someone else said that was too low on here.  L should be around L=94 or so for white borders, and it was with the spot readings.

I've been wondering if this new software is creating smaller ICC gamuts too since it seems to have issues with scanning, and probably with their own four pages (default) of color charts as well.  Dunno, but my older ICC profile wire-frames in their software comparison window are a lot larger than what I get with the newest version, almost shrunk by half too.  Seems messed up somewhere.

If you burrow into the software, the default patches in four pages of color patches for color profiling are a lot smaller than the narrowest column (patch) you can set manually too.  Seems like you should be able to match their size, but software doesn't allow it?

SG
Logged

samueljohnchia

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 498
Re: Is i1Profiler scrambling random or deterministic?
« Reply #17 on: June 17, 2014, 12:05:21 am »

Some Guy,

I replied in your other thread about my observations on my side for scanning. Smoothly varying patches are the problem still. If you use a scrambled strip target your measurements should be more accurate.

Given that X-rite's patch generator lays out patches with significantly varying adjacent patches it's no wonder that no one caught the problem yet. I build my own custom targets for i1Profiler and have been running several experiments of late so that's how I discovered the issue.

I can't tell if the gamuts are smaller than PM5 or MP, I've never used either. As far as I can tell, the gamut plots with my iPF8400 (before its chronic gamut shrinking problem started, unrelated) is huge. ColorThink calculates it at 966, 234 with Harman Gloss. That's on the standard inking with the optimal media setting. The paper can take even higher inking which sends the gamut volume up even more, by about 5%.

If you burrow into the software, the default patches in four pages of color patches for color profiling are a lot smaller than the narrowest column (patch) you can set manually too.  Seems like you should be able to match their size, but software doesn't allow it?

I have no idea what you're saying here? You can definitely set the patch width to whatever you want in 0.1mm increments. Depending on your measuring device selected you are limited to its minimum default. What size are you matching? What default patches are you referring to?
i1 Pro:7.3mm
i1 Pro 2:7mm
i1 Pro 2 with i1iO: 6mm
iSis: 6mm
Logged

howardm

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1984
Re: Is i1Profiler scrambling random or deterministic?
« Reply #18 on: June 17, 2014, 07:39:55 am »

I don't know why you are finding that smaller sheet sizes are better, since I can happily squeeze 1457 patches into one 13x19 sheet, minimal wastage. Smallest patch size though. Small sheets waste a lot more due to the margin requirements.

I looked at 1215 patches again and it contains unevenly sampled intervals in the RGB cube, 10x11x10. I prefer to use patch sets where the RGB sampling is even, like in the case of 1457, its 11x11x11, or 1105, 10x10x10.

Thanks Andrew! Enough said.

Valid points regarding the 'regularity' of the RGB grid although in the real world, I wonder what the perceptible differences are. And I'd like to understand where the extra 126 patches (on the 11^3 grid) are.  How exactly did you 'look' at the patch set?  I'd like to be able to visualize it (w/o dropping $400 on Colorthink :) )   Also, from a waste POV, 1215 patches on 2x 8.5x11 seems like it's more efficient (in my newly created 'patches per sq/in' metric) (6.5) than 1457 on 13x19 (5.9)

samueljohnchia

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 498
Re: Is i1Profiler scrambling random or deterministic?
« Reply #19 on: June 17, 2014, 09:48:38 am »

That's true, because 13 inches is a bit wider than the width span that the measurement ruler allows, so the wastage in that sense is higher. Sorry, I forget about these width issues with cut sheets because I only print off rolls these days. But try arranging 1457 patches across three 8.5x11 sheets. 5.2 patches/sq inch.

The perceptible differences, with i1Profiler's ingestion engine, is small. Measurement errors do cause far more significant issues. So for most its not worth bothering about.

I simply looked at the RGB reference values. You can save out from the "Patch Set" or "Test Chart" module. Visualizing with 3D modelling software is tricky. For example, 128, 127, 128 will look pretty close to 128, 128, 128 in a roughly evenly sampled RGB cube with the samples plotted as points in space, I doubt you can pick that out. Looking at the actual values itself is also highly informative of other under-the-hood calculations X-rite is making for their "smart" targets.
« Last Edit: June 17, 2014, 09:53:03 am by samueljohnchia »
Logged
Pages: [1] 2   Go Up