i1 profiler and 10.9.5

[Jeff-Grant]

I just tried building a profile for Metallic Pearl. Since I last used my i1Pro2 and i1io2, OS X has been updated to 10.9.5. I was using dual scan mode as the paper has OBAs. After about 10 tries I managed to get a complete measurement which produced a lumpy profile. Mostly, Profiler would hang on the first page at some point. I would get a Device Connected for the io and Profiler would give me a 'no device' error.

I just dug out another Mac with 10.9.4 and produced a clean profile without any errors. My take is that something is broken between the Xrite Device Services and Profiler after 10.9.5 is installed. I used Profiler 1.5.6 on both machines.

I have reported this to Xrite in Sydney and hope that it will get fixed soon.

[digitaldog]

I usually use an iSis but had to use an i1Pro-2 the other day for the same task, saw the same results you report. I'll try booting off another older system and see if I can replicate what you report.

[howardm]

Is this a generalized 10.9.5 problem or specific to Metallic Pearl (and maybe some others?).

I'm increasingly happy my old iMac is running 10.8.x 

[digitaldog]

So here's what I did. I have an Emergency 10.6 drive. Booted off that, installed an old version of i1P (1.2). Didn't know how far to go back so started there. It recognized the i1Pro-2 as the original (legacy mode) which should be fine. I was unable to measure two rows on page one without getting an error. Told to slow down, went incredibly slow and in between, no go, software refused to read the two rows. I wonder if the patch colors are too close together for it to see a difference. These ARE scrambled patches too. But they appear visually too close. Could be an error in the patch generation, not sure.

Then let the software update itself to latest version via the software update. Had to reboot. When I went back to measure, the newer i1Pro-2 was recognized and I told the software just to scan single mode to speed up testing. Software read every row of the three pages. NO error message about bad data. Only when I went to build the profile was I told the patch data was too close or some lame error message. So I'm not sure if this is an OS issue or a i1P issue or a bit of both. But I concur something isn't working correctly!

I think there is a bug in the way the software reports bad readings. That and maybe how patches are generated. Older software simply wouldn't read two rows, newer software did, then popped an error at the very end of the process. That's really not a good behavior! So at this point I don't know what to tell you other than try making a different target and see if you can built it such it's scrambled with large visual differences between each color. I've sent an email to the person I know inside X-rite to examine this post. I no longer do any beta testing for them so if this gets anyone's attention or how long it will take to fix it is anyone's guess. 

[Some Guy]

Fwiw, I got an email from x-rite saying something about Apple's newest OS messing up some of their hardware and software within the past week.  Said they expected to have a fix around Nov. 5th I think it was, although it might be later.  I don't do Apple so I tossed it.

Seems to be a normal event with each major OS update to have something get broken in color or hardware land.

SG

[Jeff-Grant]

Thanks folks.

Andrew, using 10.9.4 I got a profile without error. I used scrambled and dual mode so the io was positioning each patch. I managed to get a clean profile build with no error messages. The thing that has me scratching my head now is the comparison of the two profiles. In the attached screenshot, the canned profile is the wireframe and my build is solid. Looking at it, I'm left wondering what magic the canned one performed to get to that difference, and which one to use. I normally trust my own, and wonder if this is another paper with inter-batch variation.

[digitaldog]

No screen shot. But the question is, how does the profile preform? Mine was bad!

Did you measure one patch at a time with the i0 to get good data? That usually works. I'm thinking the issue is the target and scanning. I've built two new targets for my client where all the patches are quite different from one another for scanning. It was easy to do in ColorPort, ridiculous and difficult in i1Profiler. I'm hoping I'll get to sets of targets in the next day or two to see if they work. My suspicion is that the error checking in i1P is lousily (it always was to differing degrees). It's accepting the measurement in the current version when it shouldn't then builds a crappy profile. In the older version, I couldn’t read two rows on page 1 but the current version took it fine. What pisses me off is that you can't measure part of the target in scan mode, the problematic rows in patch. We were able to do this in MeasureTool, even save partially measured data, then come back and finish. Not in i1P, the design and architecture is way worse than what GMB built last decade. We should be able to patch measure any of the target during or after scanning if there's an issue. If you ask to go from scan to patch, the stupid software barfs and doesn't save the work you did in the past. Annoying! I'm hoping ColorPort will do the job this time out but I have an i1P target to test too.

[Jeff-Grant]

Thanks Andrew. The grab is now attached. I used dual scan mode which spot scans on the io. I'll print a test from each profile and get back with the results.

[Jeff-Grant]

Well, that's interesting. I just printed an evaluation image with both the canned and my own profile. The canned image has lost the yellow. In the attached image the marked yellow areas are white using the canned profile and yellow with mine. The yellows in the top row are significantly different between the two. There is a clear difference on the green patches too.

Whatever is going wrong is still producing a better profile than the canned one that I used.

[Jeff-Grant]

Just to check whether the paper is having an effect, I just rescanned some Hahnemuhle Photo Rag Baryta to compare with a pre 10.9.5 scan. It went through in Dual Spot mode with no complaints unlike the metallic but when I compared the two results in CTP it was illuminating. I would expect very little difference to my original if everything was working to spec but the Delta E report shows a Max dE of over 50.

I think that we can assume that Profiler and 10.9.5 are to be avoided for now.

[digitaldog]

I think that we can assume that Profiler and 10.9.5 are to be avoided for now.

Something is way wrong here. I'll have to see if I can replicate that with an iSis.

[digitaldog]

Here's where I'm at thus far.

Yesterday I built two new targets using 918 patches defined in i1P. I made the patches larger (9mm). I saved the patch set so I can open them in ColorPort to build a target there too. What's great about CP is once you import the patch set, you can rearrange the colors manually. Why we can't do this in i1P is beyond me. Anyway, I moved all colors in the CP target such they were vastly visually different from the next. I feared that the original target wasn't being measured correctly due either to too small a patch or colors too close together. I wanted to make sure that as I scan the target, there are no similar colors to confuse the software. Not possible in i1P. The other nice thing about the CP targts is there are white or black lines between each patch. Only downside is it took three pages vs. two for i1P.

Client sent targets overnight, they arrived this morning. Measured in CP using i1Pro legacy mode (yes, it's hard to believe X-rite still hasn't updated CP to recognize a device they released copule years ago). So I'm not getting dual measurement mode, it's M0. No complaints from the software about the data. Saved a CGATs file to build the profile in i1P. It looks fine in 3D unlike the original bad profile that used with more but smaller patches. I then measured dual scanning in i1P which of course recognizes the device as a more modern i1Pro. No errors there, profile also looks smooth. Comparing the two profiles, they look slightly different in gamut but overall shape looks very similar. It's not an apples to apples comparison either, I've got dual scan data vs. M0. Now I need to hear back from client about how the two print. And difference between them.

I'm hoping the original target was the issue on this end. Will report back when I hear from client.

[digitaldog]

OK, client output tests using the two new profiles.

Profile from i1Profiler measurements of 918 target was no good.
Profile from ColorPort measurements was good!

So I am not sure it's specifically an XRD issue since I'd think the same hardware uses that protocol in each product. Don't really know, XRD is a huge mess.
It does appear that ColorPort under 10.5 and the same instrument, albeit handled like the older legacy Spectrophotometer works fine. So it's the data collection for sure, the i1P engine builds a good profile from data that comes from CP, not internally. The difference however was also dual scan. Maybe the bug is that? You might wish to try a single scan and see if you get a good profile or not.

[Jeff-Grant]

This just gets better. I rescanned the target in scan mode and compared in CTP. It's pretty impressive. I'm comparing the 10.9.4 Dual to 10.9.5 single and there are only 2 patches out of 2000 that are greater than 2.

The evidence is definitely pointing at Dual Scan mode. As an aside, there wasn't much between M0, 1 and 2 in the results anyway so M0 looks like it should be fine for metallic.

I then built a profile with the M0 data from 10.9.5 also attached. The wireframe is from the M0 10.9.5 scan and the flat from the Dual data.

To finish it off, I printed the test image from the new M0 scan, and it looks good.

[digitaldog]

The evidence is definitely pointing at Dual Scan mode.

What's odd is the first profile I did was single scan and I got a crap profile. So I'm thinking there are multiple issues here. In my case, I think the first target had patches that were too small (smallest the product will allow) and the error checking in the current version just sucks eggs. I say that because in an older version, it refused to accept the measurements and I was done. In the newer version it was happy to take that data and build a crap profile. The larger patch size (9mm) target sucked today because maybe I'm running into the bug you report with dual scan.
In a nutshell, i1P is simply not reliable in terms of scanning with an i1Pro, yet I've had no current issues with an iSis. I'm sticking with ColorPort for the time being but hopefully once this metal profile is complete, I can put the i1Pro back on the shelf and stick with the iSis. I still have to do an optimization of the good profile made today. I sent the CGATs data i1P built to CP and will measure there, then move that measured data back to i1P to build ten new profile. Don't know if it will work, but it should.

I should point out that the bad dual scan profile built to day looked good in terms of the gamut map and the soft proof. My client reported it was way off (way too yellow). So at least with this process, one has to actually make a print to know if the profile is OK or not. The older original profile gamut shape looked awful and it produced an awful print.

[Jeff-Grant]

From my perspective, I think that I'm OK as long as I avoid Dual Scan. The prints look fine. Of course without an Isis to do another set of measurements, that's just pure speculation on my part.

On the topic of optimisation, there's nothing in your book. Have you written anything on the subject recently, or is there another place I can look?

[digitaldog]

On the topic of optimisation, there's nothing in your book. Have you written anything on the subject recently, or is there another place I can look?

It's an option in i1P after you make the profile (little arrow to new workflow). I do a 2505 gray patch optimization. Helps on some devices for gray balance.

[Jeff-Grant]

Thanks Andrew. I was afraid that I may be missing something new.