New version of i1Profiler

[digitaldog]

And finally, it's 64 bit for the next major update of MacOS!
Who wants to be the first to update and see what breaks? 
Notes:



New for the 1.8.0 Release

The following features and improvements have been added for this release:

General Application

New Features

Made the i1Profiler and i1Profiler Tray applications 64-bit native to comply with future MacOS updates.

Added the ability to make ICC Max compatible profiles. This is a version 4 profile with an iccMax profile embedded.

Updated ICC version 4 profiles to the 4.3 specification.

Added support for the Japan Color Control Strip for i1iSis

Reworked the entire Japan Color Print Quality module for correct reporting and validation of the Japan Color standard.

General UI improvements to better show when measurements and profiles have been successfully completed.

Improved the spectral curve display for all lighting measurements.

Improved reporting of display Uniformity results by changing to ∆E method.

Bug Fixes

Fixed issue where the Print button for RGB printer test charts was not working on some Windows systems.

Fixed issue where wrong patches were being shown for Flare Correction in display.

Fixed issue where i1Profiler did not calculate XYZ correctly for illuminants other than D50.

Fixed issue where certain device values in CGATS measurements would cause profiling to fail.

Fixed issue where the black width setting was not working when creating device-link profiles.

Fixed issue where i1Profiler could not correctly profile a system with two identical displays attached.

[aaronchan]

Great!
But what’s the advantage by using the icc max over version 2.x for RGB printer?


Sent from my iPhone using Tapatalk

[Doug Gray]

Great!
But what’s the advantage by using the icc max over version 2.x for RGB printer?


Sent from my iPhone using Tapatalk

Good question, one thing that icc max allows is using different white points other than D50 in the PCS, IIRC. No idea if this is something I1Profiler can do. There's lots of stuff possible in V4 that I1 doesn't do. Being compatible just means they offset some subset.

[digitaldog]

Great!
But what’s the advantage by using the icc max over version 2.x for RGB printer?

I suspect zero and until they build V4 profiles that actually use the PRMG, seems pointless.

[Doug Gray]

I suspect zero and until they build V4 profiles that actually use the PRMG, seems pointless.

  Yeah.

I installed it on my Win 10 64x system. When it restarted after installation the monitors were black and stayed black. Hmmm.

I did a hard reset and rebooted again. This time it came up but my ColorNavigator had been bypassed and no longer listed the CG318 monitor. Lovely.

I did another reset and everything seemed to be working fine. Until, that is, the computer screen went to sleep. Another black screen that wouldn't come back.

Did another hard reset and disabled the XRite crap that was trying to control the monitor. Been working since.

Outside of that is seems to be working fine. Ran some measurements, made some profiles, everything else checks out.

One would think that XRite would be app aware of high end monitor control software like ColorNavigator. Hardly rare that peeps would be using Eizo stuff with I1Profiler.

Nothing really new on the menus. It retained all the info I had from 1.7.

[Doug Gray]

There are some differences in profiles generated by I1Profiler V1.7 and 1.8 but they are very, very small.

Test Process:
Make profiles from the same CGATs file from V1.7 and V1.8
Generate 100,000 random RGB triplets
Create 100,000 Lab values  using the AtoB1 tables from both profiles.

Result: Lab values are identical.


Next run these Lab values through the BtoA1 tables which reconstructs the original RGB values. On both profiles these differ somewhat (typically < .3 dE) from the originals since they go through two levels of interpolation.

Result, very slight differences between the two profiles with the maximum difference at  .1 (relative to 8 bit RGBs). This difference would not be measureable. 99.98% were identical. Mistake looking at histogram. Differences here are somewhat larger along gamut boundaries. Am investigating to see if the differences improve the roundtrip results of near boundary colors.

This indicates some change in the calculation of the BtoA tables while no change in the AtoB tables. Possibly this was to avoid crashing due to near singularities in the data. Or they could be trying to improve Lab mapping to device RGB which is particularly difficult at the gamut boundary. This was alluded to in the notes and I have seen occasional crashes with V1.7. Especially when making profiles for lighting different than D50.

The very small changes observed is consistent with using a more robust matrix inversion algorithm.

I have not examined the Perceptual tables, only the colorimetric so can't opine on any changes in those.

Added:
I looking closely at the areas where the BtoA1 table varies the most, it's at the gamut boundary. I'm going to run a test with a set of boundary only colors to see if the BtoA table for V1.8 is more accurate than V1.7. It will compose of a set of Lab values on the boundary that roundtrip with greatest errors. While I assume the AtoB table is the most accurate, this needs to be verified. My guess is that it's the most accurate but that BtoA algos have changed to improve accuracy. By testing only boundary colors with the largest roundtrip differences, measurements should be able to affirm or reject. Stand by.

[Doug Gray]

Summary:

The AtoB1 tables are identical
The BtoA1 tables differ very small amounts between V1.7 and V1.8. Less than .1 dE00 for a large, random sample of in gamut colors. Overall, it isn't observable or even measurable.

However, there are a few colors where the Lab values translate into device RGB values that print with slightly significant dE00 differences.

The following was done to explore where the differences were greatest:

1. Create a profile with each version of I1Profiler using the defaults except that profile quality was set at maximum in both.
2. Initialize an RGB set of 200,000 random, device RGB triplets.
3. Verify that all RGB triplets produced identical Lab values though the AtoB1 tables in both profiles.
4. Run these 200,000 Lab triplets roundtrip through each profile. The new roundtripped, Lab values from each profile were compared and a list of the largest 500 dE00s made.
5. Then a subset of the largest dE00s was selected such that each color was 10 dE or more from each other color with emphasis on the largest dEs. This produced 16 color RGB sets.
6. Then an RGB device set consisting of these values as well as the RGB values from roundtripping these RGB values through the two profiles which produced 48 triplets.
7. These were duplicated to fill up a 957 patch target for creating an iSis 16 bit tif file and matching CGATs file.
8. This was printed and scanned. The average for all the identical ( > 20 for each color patch) colors comparing the measured values and the values calculated from the profile were tabulated for each of the three cases.

The average dE00 from V1.7 was .64, for V1.8 it was .53, and for the initial 16 RGB values, .33

As expected, the AtoB1 tables were pretty accurate at .33 dE00 ave predicted error. Roundtripping those RGB values didn't significantly reduce the accuracy, although I would normally expect random, in gamut roundtripping to increase dE00 by less than .1

Here's the actual dE00's of the 16 colors

    RGB       V1.7         V1.8
    0.1655    1.4160    0.9470
    0.4291    0.4229    0.4402
    0.1824    0.2290    0.5502
    0.6334    0.7475    0.5438
    0.4690    0.6337    0.6240
    0.3660    0.6126    1.0589
    0.3049    0.5313    0.6553
    0.3675    0.3151    0.5422
    0.1126    0.8664    0.8561
    0.1784    0.2511    0.5667
    0.1581    0.1963    0.2130
    0.2677    0.0744    0.8312
    0.2648    0.4827    0.6347
    0.5755    0.0685    0.1099
    0.5611    0.7721    0.8215
    0.2375    0.8617    0.8668

[Doug Gray]

I have some new observations about V1.8

First, the generation of AtoB1 tables (RelCol), which is used primarily for proofing and is the most accurate table colorimetrically, is identical between V1.7 and V1.8.
However, BtoA tables differ. For in gamut colors the differences are negligible.

I investigated the effect of the Smoothness slider on a 2871 patch profile consisting of the baseline i1iSis Default set with about 600 added near neutrals and the rest filled in with optimizing patches from a second pass. The default, 50, as well as settings 0, 20, 40, 60, 80, and 100. Settings below 80 produced identical AtoB1 tables. However, all the settings produced different BtoA tables. As for colorimetric accuracy, settings between 40 and 60 produced essentially nearly identical accuracy. Smoothness settings of 100 were by far the worst with 0 being a close second. Overall, the default setting of 50 produced the best results so it seems their default setting is good.

There is one major improvement. The problem I detailed in the Perceptual tables seems to have been fixed. See http://forum.luminous-landscape.com/index.php?topic=114653.msg998584;topicseen#msg998584

The strange dark and light blotches when soft proofing, or actually printing, Lab circles in Perceptual has been eliminated. There are also slight shifts in the way out of gamut colors are mapped.

As an aside, I printed my neutral tone curve in Abs Col and the average dE00 turned out to be .38, or .34 when accounting for patch variation from printing. That's quite good on my 9800 where it's really hard to get  good neutral tracking due to a* and b* anomalies. Especially in the neutrals where, unlike more saturated colors, dE00 usually exceeds dE76. Even the worst cases were under 1.

Also I did a crosscheck with a random set of 957 RGB triplets in device space to check the profile's performance across the entire gamut and the average dE00 was .44.

Another thing. The black point for RelCol measured 4.37 and is the same as the device BP with RGB 0,0,0. Perhaps it's the patch set. I've noticed other patch sets producing higher black L* than the device BP as much as .6 higher L*. However, the neutrals and near neutrals I added perhaps is why. I haven't investigated this and was actually surprised that the RelCol BP and device RGB (0,0,0) BP were the same. Anything under 5 is is pretty good for the 9800. It is over 10 years old. What a workhorse though.

So all in all, V1.8 looks like a good update.

[RobWignell]

Tried to install this new version on my MacBook Pro, 2017 model, and got the dialog message:

Update Error!
The update is improperly signed.

See attached screen image.

Can anyone suggest what to do next.

[RobWignell]

Problem solved.  The new version failed to install when I used the update option in V1.7.

I subsequently downloaded V1.8 directly from the X-Rite web site.  This version installed without error.

RW

[digitaldog]

Problem solved.  The new version failed to install when I used the update option in V1.7.

I subsequently downloaded V1.8 directly from the X-Rite web site.  This version installed without error.

RW

Win or Mac? I was able to update from the older version on a Mac.

[RobWignell]

Win or Mac? I was able to update from the older version on a Mac.

Mac

[DickKenny]

Somewhat disturbing message flashed up on my iMac today - updated yesterday to macOS Mojave - when I launched my i1 Display. The exact wording escapes me as the panel disappeared in seconds and despite all efforts I haven't been able to resurrect same. Gist of the message was that once Apple completes it move to 64 bit standard the i1Display measuring unit will not compatible - and is not able to be updated.

I emphasise that is not the actual wording of the message; but my best effort to recall its content.

Very happy to be corrected if I grabbed the wrong end of the stick.

[digitaldog]

Somewhat disturbing message flashed up on my iMac today - updated yesterday to macOS Mojave - when I launched my i1 Display. The exact wording escapes me as the panel disappeared in seconds and despite all efforts I haven't been able to resurrect same. Gist of the message was that once Apple completes it move to 64 bit standard the i1Display measuring unit will not compatible - and is not able to be updated.

I emphasise that is not the actual wording of the message; but my best effort to recall its content.

Very happy to be corrected if I grabbed the wrong end of the stick.

The next major OS from Apple will no longer run any 32-bit applications hence the warning. That means X-rite has to update all their products to 64-bit like everyone else who hasn't yet. We've got some time considering Mojave just came out this week and all upgrades to Mojave will run 32-bit applications but after that, nope. HOW LONG it will take X-rite to provide 64-bit before this occurs is up in the air but I expect they will do so before the next major OS release. They better! 

[BAB]

QuestionI'm using an old eye-one for monitor calibration I purchased this about 16 years ago Serial no 3,278-110137 rev A on a MacBook Pro (Retina, 15-inch, Mid 2014) i1Profiler version 1.7.2 copyright X-Rite, Inc. 2011-2017.My monitor Nec MultiSync PA302w uses it own calibrator.
Should I upgrade to
i1Profiler (i1Publish) v1.8.1
Should I be using a newer profiler device for my Macbook?PrintingI have been using the Epson profiles and not making my own for the Epson P8000 seems to print just fine...would I notice a significant difference making my own profiles again with newer equipment? (The only adjustments I make occasionally after printing a small test image out of LR is to brighten the image)
Thanks
Barry

[digitaldog]

If it ain’t broke...
That said, the newer hardware from XRite is better!
But if you have a newer Colorimeter for the NEC and/or you are happy, I would not change anything.

[Ethan Hansen]

Isn't an issue that the integrated dongle is 32-bit only? From the 1.8.1 release notes:
"There is no native 64-bit driver for the i1Profiler Hasp dongle. When Apple releases an update that forces 64-bit for all applications and components, the i1Profiler dongle will no longer function."

Typically when one uses i1Profiler only with an i1Display, the i1D2 acts as a dongle to unlock display profiling. I hope this does not mean X-Rite is moving to a per-machine licensing solution instead of per-device.