Pages: [1]   Go Down

Author Topic: Optimizing i1 Profiler profiles  (Read 5994 times)

Daverich

  • Full Member
  • ***
  • Offline Offline
  • Posts: 164
Optimizing i1 Profiler profiles
« on: September 05, 2017, 05:49:07 pm »

I've had quite a bit of personal and commercial experience with Monaco Profiler and Profilemaker 5 but I recently bought i1 Profiler as I no longer wanted to keep a separate Mac around just to run Profilemaker 5 as well as it being way past time to get up to date. I'm getting around it pretty well but have a question about optimizing profiles. All my testing to date has been on Ultra Premium Luster with 1457 patches. When I get to the point I'm not experimenting but making final profiles I intend to use 2371 patches so maybe this question is moot. One thing I've noticed is that the gamut volumes observed using the Colorsync Utility of the Epson profiles that came with my P5000 are noticeably larger in all directions than the ones I've made with i1 Profiler. When I tried the Optimize option using "All Patches" my profile filled out some of the edges of the gamut volume but didn't really make any changes I could discern in test prints. In another thread on LuLa (http://forum.luminous-landscape.com/index.php?topic=118476.msg982583#msg982583) Ethan Hansen posted some optimization files that he had created along with instructions for using them. I followed his instructions for using his 1200 patch file and everything seemed to work. Looking at the gamut volume, not only was it much larger than my plain profile, it was also much larger than the factory Epson profile. I thought I was home free but when I printed some test prints I saw immediately that what should be paper white picked up a tone to it and gray ramps were no longer smooth but had steps in them. Ethan is a really knowledgeable person and before Dry Creek quit making profiles for the general public he used to make them for me. Does anyone have a guess on what I've done wrong or am I misunderstanding what his files are intended to be used for? I know that some of this stuff is like how many Angels are on the head of a pin and not readily discernible in an average print but as long as the Optimize feature is in the software I'd like to understand how to use it. Thanks.

Dave
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Optimizing i1 Profiler profiles
« Reply #1 on: September 06, 2017, 03:42:46 am »

First, the gamut volume of Epson's canned profiles appears larger than that of a custom profile. This is normal and is not an indication the actual printable color gamut is larger but is a side effect of Epson's choice to bake in BPC in the profile's LUT tables. Doing this shifts the physical L* black from, say 3 or 4, down to 0. Of course it doesn't print at L*=0, but the profile pretends it does. This is not a correctly constructed profile according to the ICC. The net effect is to expand the reported gamut volume and said expansion is most overstated on matte profiles because of the lower DMax (higher L*).

Secondly, gamut volume, even when accurately reported says nothing about the quality of a printer profile. I generally ignore it.

Thirdly, I have had some very peculiar results with I1Profiler that I tracked down to targets where I had added neutral and near neutral patches.

http://forum.luminous-landscape.com/index.php?topic=114653.msg944691#msg944691
Logged

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
Re: Optimizing i1 Profiler profiles
« Reply #2 on: September 06, 2017, 07:53:28 am »

First, the gamut volume of Epson's canned profiles appears larger than that of a custom profile. This is normal and is not an indication the actual printable color gamut is larger but is a side effect of Epson's choice to bake in BPC in the profile's LUT tables. Doing this shifts the physical L* black from, say 3 or 4, down to 0. Of course it doesn't print at L*=0, but the profile pretends it does. This is not a correctly constructed profile according to the ICC. The net effect is to expand the reported gamut volume and said expansion is most overstated on matte profiles because of the lower DMax (higher L*).

Secondly, gamut volume, even when accurately reported says nothing about the quality of a printer profile. I generally ignore it.

Thirdly, I have had some very peculiar results with I1Profiler that I tracked down to targets where I had added neutral and near neutral patches.

http://forum.luminous-landscape.com/index.php?topic=114653.msg944691#msg944691

Well, like in so many of these things......."it depends". I'm using i1Profiler and an i1Pro2. I've profiled and studied the canned profiles of a large number and variety of papers from different providers over the past several years. I have not ever seen a profile reporting L*0 for the Black point. As for gamut volume, it depends - sometimes the canned profile has larger gamut than my profiles for the same printer/paper combination, sometimes not. I can't explain why these kind of apparently unsystematic variances, but I just observe them in the results I see. That said, these differences between canned and custom gamut volumes are usually immaterial - not large enough to make a difference in a print except in certain rare conditions which I've discussed previously. There are comparative cases where gamut volume is clearly an important indicator of what the printer/paper combinations can do - to take an extreme comparison which is nonetheless a real one: Japanese Kozo paper compared with Epson Legacy Baryta. You would make completely different kinds of prints with these papers and the gamut volume is an important parameter explaining why.

I agree with you that gamut volume and "quality" are not necessarily related, if by "quality" you mean accuracy of the printed result relative to the file values being rendered. But here again, it is inherently and necessarily the case that depending on the gamut volume some colours represented by their file values may well be out of printer/paper gamut, in which cases the Rendering Intent will remap the OOG values and the resulting printed values will differ from the file values. So then one is looking at the quality of the Rendering Intent performance and not the profile; however to the extent a profile can provide for wider gamut, the scope for OOG issues may be reduced.

As for your third point on peculiar results, I haven't tinkered with X-Rite's patch sets, but I find when I generate a normal patch set using one of the numbers that Ethan Hansen recommended in a recent article, the performance of both colours and grayscale is usually very satisfactory within the luminance gamut of the printer/paper combination being tested, so I don't generally see the need to tinker with these tables.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Optimizing i1 Profiler profiles
« Reply #3 on: September 06, 2017, 11:06:58 am »

Well, like in so many of these things......."it depends". I'm using i1Profiler and an i1Pro2. I've profiled and studied the canned profiles of a large number and variety of papers from different providers over the past several years. I have not ever seen a profile reporting L*0 for the Black point.
ProfileMaker 5 and I1Profiler correctly map BP. I've only seen the issue with canned Epson profiles. Canon's also have issues but they report BP correctly. What canned Epson profiles have you looked at?

Incorrect BP reporting dominates what I've seen from the 2200 to 2400 and 9800 Epson canned profiles that were created by Seiko profiling software. Here's two canned profiles supplied by Epson that were created by Seiko software. One for the R800 and one from the 9800/7800 canned sets. All of my other 9800/7800 canned profiles also show BP as L=0.

Profile: "C:\WINDOWS\System32\Spool\Drivers\Color\Pro9800 7800 DWMP_PK.icm"
Descr: "Pro9800 7800 DWMP_PK" Ver: 2.4.0   Copyright: "Copyright(C) SEIKO EPSON CORP. 2005"
WP:  93.6   0.6  -0.8     BP_PI:   0.0   0.0   0.0    BP_RI:   0.0   0.0   0.0

Profile: "C:\WINDOWS\System32\Spool\Drivers\Color\SPR800 Matte Paper-HW.icm"
Descr: "SPR800 Matte Paper-HW" Ver: 2.4.0   Copyright: "Copyright(C) SEIKO EPSON CORP. 2004"
WP:  95.5  -0.8   1.1     BP_PI:   0.0   0.0  -0.0    BP_RI:   0.0   0.0  -0.0

I have yet to see a canned Epson profile, manufactured by Seiko software, that didn't map BP to 0 and would be curious if you have or know of Seiko, Epson profiles that report BP correctly.


But Epson also sometimes provides non Seiko profiles and these, for the limited ones I've looked at, map normally like this for the 3880.

Profile: "C:\WINDOWS\System32\Spool\Drivers\Color\SP3880 ECM MK v1.icc"
Descr: "SP3880 ECM MK v1" Ver: 2.0.0   Copyright: "Copyright X-Rite, Inc."
WP:  96.5  -0.7   1.0     BP_PI:  20.7   0.7   1.2    BP_RI:  20.9   0.6   1.0

Quote
As for your third point on peculiar results, I haven't tinkered with X-Rite's patch sets, but I find when I generate a normal patch set using one of the numbers that Ethan Hansen recommended in a recent article, the performance of both colours and grayscale is usually very satisfactory within the luminance gamut of the printer/paper combination being tested, so I don't generally see the need to tinker with these tables.

I've found Ethan's patch sets to be quite good as well and appreciate the detailed breakdown he provided for I1Profiler's set of patch sizes. I've had very good results with them.
« Last Edit: September 06, 2017, 11:19:06 am by Doug Gray »
Logged

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
Re: Optimizing i1 Profiler profiles
« Reply #4 on: September 06, 2017, 11:13:51 am »

Well, that explains it - we've been looking at different profiles for different printers. The Epson printers I've been working with include the 3800/3880. 4900, SC-P800 and SC-P5000. The canned profiles I looked at for all these printers for a range of Epson papers are the ones made by Epson America. I *think* the ones you are looking at are made by Epson Japan. They aren't the same.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Optimizing i1 Profiler profiles
« Reply #5 on: September 06, 2017, 11:46:33 am »

Well, that explains it - we've been looking at different profiles for different printers. The Epson printers I've been working with include the 3800/3880. 4900, SC-P800 and SC-P5000. The canned profiles I looked at for all these printers for a range of Epson papers are the ones made by Epson America. I *think* the ones you are looking at are made by Epson Japan. They aren't the same.

Could well be. They were the ones installed by default with the printer drivers. The ones that exhibit this also are subjects of Andrew Rodney's discourse on profiles: "Not All Profiles are Created the Same." where these map Bill Atkinson's "blue ball" to black. Pretty much the entire blue ball is OOG with L* < 1. Since the Seiko profiles rescale the BP to 0 on the Rel Col tables, black, or close to it because of their low L*, is the proximate gamut surface point for these OOG colors.

I haven't used canned profiles much since 2006 and don't have the problem with ones I made with XRite or Gretag MacBeth's PM5.  I really only noticed this later when using Abs Col to print specific colors.

As for Ethan's set of patch colors I1Profiler offers. I am getting slightly better results with patch sets that have equal steps for R,G, and B. as opposed to the next larger patch set with G being one more than R and B. Can't say for sure yet. Also, I have not checked the patch sets with max, near neutrals for the anomaly I linked to that shows up as a major problem in their Perceptual mapping. Perhaps I'll revisit that with the profiles I've made using the Isis XL.

I've made some targets that work for both the Isis 2 and I1Pro 2 by adding black bars on the Isis target sheet edges and using 8mm patches instead of the default 6mm. These provide registration for the I1Pro 2 but are ignored by the Isis which uses "diamonds" on each row to align vertically and horizontally. Very handy for comparing I1Pro 2 and Isis inter-instrument consistency.

Update: I just checked profiles made with standard patch sets using the Isis 2 and found some with major anomalies in Perceptual Colorimetric. I1Profiler has some odd behavior in places. I'll add some stuff in a zip file so peeps can check their own profiles for this. Mostly this affects people using wide gamut spaces like ProPhoto or LAB and have colors outside the printer's gamut. The problem is that it's really non trivial to determine when, where, or if these will show up. I just ran through all the profiles installed in my system (easy to do) and was rather astonished.
« Last Edit: September 06, 2017, 02:53:42 pm by Doug Gray »
Logged

Ethan Hansen

  • Full Member
  • ***
  • Offline Offline
  • Posts: 148
    • Dry Creek Photo
Re: Optimizing i1 Profiler profiles
« Reply #6 on: September 08, 2017, 05:41:54 pm »

Dave: How well did your original (pre-optimization) profile perform in terms of paper white neutrality and gray ramp linearity? If the unoptimized profiles appear well behaved then something went amiss in the optimization process and we can debug from there. If the original profiles showed problems, however, I would first try printing a second set of targets and checking that the measured values were close to the first set and then building another profile.

Daverich

  • Full Member
  • ***
  • Offline Offline
  • Posts: 164
Re: Optimizing i1 Profiler profiles
« Reply #7 on: September 09, 2017, 11:12:46 am »

Dave: How well did your original (pre-optimization) profile perform in terms of paper white neutrality and gray ramp linearity? If the unoptimized profiles appear well behaved then something went amiss in the optimization process and we can debug from there. If the original profiles showed problems, however, I would first try printing a second set of targets and checking that the measured values were close to the first set and then building another profile.

Hi Ethan,

Thanks for the response. The original profile had paper white and smooth gradients as did the optimized profile done using the i1 Profiler optimization targets. When I ran your optimization I ended up with a real light cast to something that should be paper white and banding in the gray ramps. In this image the top ramp was printed with a profile optimized with the i1 targets and the bottom ramp was printed with the exact same profile optimized with your 1200 patch target. It seems pretty likely that I did something wrong but I'm not sure what it was. Thanks again for your help.

Dave
Logged

Stephen G

  • Full Member
  • ***
  • Offline Offline
  • Posts: 173
Re: Optimizing i1 Profiler profiles
« Reply #8 on: September 12, 2017, 05:40:06 am »

I'm following this thread with interest because I have also recently started using i1Profiler to create printer profiles. I've made some lovely profiles using the patch set sizes that Ethan Hansen suggested (thank you Ethan) but when I've tested the optimization process I've gotten results similar to Daverich's. White point shifts and banding all over the place.

Attached is a screengrab of Photoshop showing the same gray ramp softproofed with a profile for a 300g fine-art matte. The original profile (straight) and the optimized version of it.

Logged

Ethan Hansen

  • Full Member
  • ***
  • Offline Offline
  • Posts: 148
    • Dry Creek Photo
Re: Optimizing i1 Profiler profiles
« Reply #9 on: September 12, 2017, 07:44:36 pm »

Very strange results. I've mostly used the optimization with the 2500+ patch target. I'll need to give the smaller targets a try with the most recent i1Profiler version to see if I can reproduce the problems. As we don't use i1Profiler for printers often at all, the majority of my tests were with older versions. Given X-Rite's complete lack of documentation on much of anything, let alone change logs, there's no way of telling whether a more recent update altered optimization behavior.

Give me a week or so and I'll get back with more details.

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
Re: Optimizing i1 Profiler profiles
« Reply #10 on: September 13, 2017, 01:49:24 am »

................Given X-Rite's complete lack of documentation on much of anything, let alone change logs, there's no way of telling whether a more recent update altered optimization behavior.


That's because we're dealing with a company that doesn't see fit even to produce a basic manual for allowing customers to readily understand how to use all the features in their own application, and even when they can confirm application bugs that users report, they fail to issue dot releases in a timely way for correcting them. What makes you think that such a totally irresponsible and uncaring management group would get into the refinement of change logs?
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

Stephen G

  • Full Member
  • ***
  • Offline Offline
  • Posts: 173
Re: Optimizing i1 Profiler profiles
« Reply #11 on: September 13, 2017, 07:24:43 am »

Thanks again Ethan, the effort is appreciated.

I'll add that I also optimized a profile for a Baryta semi gloss paper and the optimized profile was, to me, indistinguishable from the source profile.
I used the larger 2525 patch set for this paper, and the 1200 patch set for the matte paper I mentioned previously.

If I get a chance I'll optimize the Baryta profile with the 1200 patch set too, to see what happens there.
Logged

Alan Goldhammer

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 4344
    • A Goldhammer Photography
Re: Optimizing i1 Profiler profiles
« Reply #12 on: September 13, 2017, 08:39:41 am »

That's because we're dealing with a company that doesn't see fit even to produce a basic manual for allowing customers to readily understand how to use all the features in their own application, and even when they can confirm application bugs that users report, they fail to issue dot releases in a timely way for correcting them. What makes you think that such a totally irresponsible and uncaring management group would get into the refinement of change logs?
Well we now have the wonderful example of Equifax to hold a mirror up to.  They won't answer anything regarding the recent data breech. :-[
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Re: Optimizing i1 Profiler profiles
« Reply #13 on: September 13, 2017, 10:20:39 am »

That's because we're dealing with a company that doesn't see fit even to produce a basic manual for allowing customers to readily understand how to use all the features in their own application, and even when they can confirm application bugs that users report, they fail to issue dot releases in a timely way for correcting them.
Some of the features don't work; difficult to document that kind of software.
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
Re: Optimizing i1 Profiler profiles
« Reply #14 on: September 13, 2017, 01:46:17 pm »

.............difficult to document that kind of software.

Only if it's XRite.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

Daverich

  • Full Member
  • ***
  • Offline Offline
  • Posts: 164
Re: Optimizing i1 Profiler profiles
« Reply #15 on: September 13, 2017, 05:46:51 pm »

Thanks again Ethan, the effort is appreciated.

I'll add that I also optimized a profile for a Baryta semi gloss paper and the optimized profile was, to me, indistinguishable from the source profile.
I used the larger 2525 patch set for this paper, and the 1200 patch set for the matte paper I mentioned previously.

If I get a chance I'll optimize the Baryta profile with the 1200 patch set too, to see what happens there.

It never occurred to me to try any of the other patch sets besides the 1200, I just assumed that I had done something wrong. I tried the 2525 patch set on the Luster profile I had tried with the 1200 patch sets and got much better results. The very light tone in the highlights is still there but most of the banding in the gray ramps is gone. There is one band left, almost in the middle of the ramp. I doubt very much that it would ever show in a normal print but knowing it's there and not in the regular Luster profile would bother me so I probably won't use it.

Dave
Logged
Pages: [1]   Go Up