Saturday, October 29, 2005

Five-minute profiling with the DTP70

Apart from some tiny irritations listed below, Monaco Profiler has speeded up considerably after I installed it on my G5. I now need just about 5 minutes to make an inkjet profile with the half-page chart, and that time spans both measuring and computing the profile. I now feel that I've got this system sussed - it's time to do some real printing rather than just Colorcheckers and standard test pages to check out the profiles. I've been printing some portraits of a friend's son on Epson Premium Glossy, not at all recommended or supported on the Epson 2100. My most important criterion is skin texture, and the kid's skin tone looks ok on the prints.

I had a phone discussion with Steve Upton of Chromix yesterday, and he makes his inkjet profiles with at least a thousand patches on his slow Gretag Spectrolino/Spectroscan systems. I'm not sure that's necessarily a good thing, though - when I was younger I did a lot of neural net training and I learnt that when you use too many data points for training a net you can get overfitting. Once I've written the first Publish.com article about the hardware, I'll start to play around with changing parameters in the software, and test profile quality with large and small charts, dense cubes and sparse cubes.

Detailed DTP70 lab journal.:I installed Monaco Profiler on my G5 dual 2.5 with Tiger today - the computer I call "Fanjet". I encountered some minor software issues which I hadn't seen on the Powerbook under Panther. Profiler wasn't seeing the dongle. I found the dongle installer in a directory, installed it and everything started working properly. It seems that the dongle driver didn't install by default ! Also, there's some strange interaction between Monaco's screen profiling reminder and my screens, carefully profiled and matched with ColorEyes, are having their assigned profiles ignored at reboot. So I have to go into Display Preferences and reset the screen profiles to the ones I want. More entertainingly, the Xrite mail system has apparently blacklisted gmail, so I can't send those nice people mail easily and report these small glitches.

Oh, and by the way, I found my old Colorchecker under some socks, and it doesn't quite match the new one.

Thursday, October 27, 2005

DTP70 - a device for printing beautiful Colorcheckers !

Today I printed some beautiful Colorcheckers using that very first ICC profile I'd originally made within thirty minutes of unpacking the DTP70. On monday, I will give the DTP70 and Monaco Profiler a hard workout with a variety of printers, testing against GMB Profile Maker Pro with a fellow color geek in Paris. But I can already state that the first profile I made was very good; out-of-the-box experience excellent, and the profiling software is a breeze to use.

Regarding my earlier posts of problems: As I'd guessed, there was no problem with the profile. But printing the Colorchecker was real hard - for me. I had a lot of problems with neutrals going bluish etc. This was all my fault for not understanding the arcana of choosing a rendering when printing. I've taken a week to sort myself out. Even going to the extent of measuring the patches on my Colorcheker.

In the end I was told by the guys listed below to print with Absolute Colorimetric from a LAB-encoded Colorchecker file, and the result was beautiful !

I'd like to thank the color geeks on the colorsync list - especially Danny Pascale of Babelcolor fame, Steve Upton padrone of Chromix, and Roger Breton, for sorting this out for me.

Saturday, October 22, 2005

Comparing colorcheckers

Yes the red patch on my new Colorchecker was off by 3 deltaEab compared with Danny Pascale's average. And yes, you can compare images by superposing them on 2 layers and blinking. And yes, I'm going to have to look at how I'm printing, because the printed versions have those 3 colours which are off by 7 or 8 deltaE.

In the mean time, I used the DTP 94 and Profiler to profile a friend's LCD screens. Monaco Profiler Software and DTP94 hardware are neat, easy to install and easy use. Software has a feature for measuring performance with definable patches, I think the Colorchecker colors are wired in - I should use this feature in future screen profile tests, crosschecking with the i1.

How to make your own custom Colorchecker

The quick way to make your own ColorChecker is to start with a template eg. that from Danny Pascale's Babelcolor site. You go into i1 Share, and measure all the patches. Then you chhose Export and save out the data as either a Lab or AdobeRGB photoshop patch file, which you then read into Photoshop. Magic brush to select the patches from the template, then fill. Of course, if someone created a macro to do this it would be faster. If you want to do so, post a comment.

Inter-colorchecker, inter-instrument agreement and other issues

In this post I'm using Blogger as a lab journal.

The slight disparity of a few of the Colorchecker patches, between the original and the printed version is bugging me. Color geeks are perfectionists or they're nothing, and the bug has bit me too. The strange thing is, the match between paper types is very good, it's the match between my two profiled prints and the Colorchecker which diverges slightly.So I suspect that the only open part of the loop, the image data, might be bad.

So, l'll measure out my own Colorchecker. I will then update the image I got from DannyPascale's site with mty own data and reprint. Unfortunately, I cannot stuff the colorchecker into the DTP70, so I have to measure the chart with my EyeOne. This introduces the additional complication of inter-instrument agreement, which I will try to address in another set of measurements later by measuring some printed patches with both instruments.

Thursday, October 20, 2005

DTP70 progress notes

Today I set up the appointment for the GMB PMP comparison I want to do with the DTP70. We're going to do some CMYK with a BestColor RIP as well, on some real wide-format printers, we have a whole menagerie of them available !

And I profiled another paper,on my mini-wide-format Epson, which I call super-Tetenal. This is a 290g superglossy ceramic inkjet paper which I often use to make 13x18 prints. Give me some leeway, I'm using this page to dump my rough notes.

I haven't yet quite sorted out the paper feed of the DTP70. It rejected my A4 supertetenal sheets until I scissored off the leading margin up to the dotted line. Maybe I'll go in and edit the margin away in Photoshop and see whether matters improve.

Anyway, total time to get a profile is under 10 minutes: 2 minutes max to feed and read, 5 minutes to calculate on the Powerbook. Maybe it's time to move Profiler to the G5 - code name Fanjet- and let it earn its name.

I'm not totally happy with the profile quality today. The red patch, again, and surprisingly the foliage green. Strangely, the two sets of prints I have, matte and superglossy, these match quite well, they even match the Powerbook screen, but not the Colorchecker. However my PocketChecker (tm) matches the big new colorchecker. I'm going to hand-measure the colorchecker tomorrow and reprint. By now the installation phase should be considered over; it's time to go for results.

Today I had an epiphany: Black detail is much better on the matte than on the glossy. Hmmm. Maybe there is a reason why people print B&W on matte. It's not because it looks prettier, I always thought that was a convention, it's because the deeper blacks actually allow us to carve more detail into the dark tones.

As you can see, I have no problem showing how dumb I am on my blog - feel free to comment if you see something worth correcting !

Oh, and yes, the DTP-70 really makes profiling immediately before you print perfectly feasible. Maybe in a few years they'll bring out a really cheap model and everybody will have push-button profiling. The EyeOne is not something I use casually in the same way.

Wednesday, October 19, 2005

DTP-70 vs. EyeOne: US Automation as good as Swiss Elbow Grease.

Good news: It would seem that the DTP-70/Monaco Profiler 4.8 combo will immediately produce RGB profiles that are in the same league as those of my old EyeOne Match system. And, to my knowledge, nobody has ever faulted the quality of the EyeOne profile engine, although measurement errors sometimes crept in due to the ruler design (that's fixed, now).

I've been printing the Pixl Test Print rendered perceptually on my Epson 2100, with the Epson Matte profile I made with the DTP70, and an old i1 profile. Xrite manages to pull some more definition out of the shadows, Gretag yields perceptibly better yellows, but otherwise the two prints are really very close. I don't think anyone ouside color's magic circle -artists, designers, fashion people and us photographer chickens- would notice the difference.

However, the DTP-70 profile was generated with just one button-press to read the half-page with 343 patches, which took about a minute, while the EyeOne profile was made with two sheets of patches. That's a dollar more of ink and paper, one print more to run off and a decent amount of elbow grease to scan it in.

There are some much more sophisticated testcharts available in Monaco Profiler, but given the excellent initial results I haven'felt the need to use multi-page charts — yet: Epson's enhanced matte paper, with its reduced gamut, is not exactly the best candidate for a discriminating RGB test.

Also, I've been looking at yesterday's Colorchecker test in daylight. Yesterday's commentary is confirmed: Nice, with some imperfection on the hard red patch and on the light skin patch. Extremely good density match on the neutral patches.

I also did some due diligence by talking to Paris color consultant Gerard Niemetzky. He tells me his clients are happy with the DTP-70.

So far, so good, as Andrew Rodney wrote me once. Now, off to bed. Commentary on Apple's Aperture app should be up on Publish.com tomorrow.

UPDATE: I will be meeting with a fellow color geek in a week's time, he has the high-end Gretag package, and we will jointly run comparison tests with Profile Maker Pro. I prefer to do it this way, two pairs of eyes are better than one.

Tuesday, October 18, 2005

DTP70 Review Update 1: UV filter or no filter ?

While installing the DTP70 -the actual user's manual is on CD- I came across the instructions for moving the built-in UV-cut filter into or out of the spectro's optical path. Although I had noticed this filter in an earlier glance at a sample instrument , I'd forgotten all about it. A check showed me the filter of the unit I just got was in place.

Should I filter the UV, or not ?

Let's remember that the filter is not there to filter UV from the light reflected from the specimen. There's no point in doing this because the spectro can easily ignore any UV bands - after all it's a spectro and bandwidth discrimination is its job !

Indeed, the filter is there to remove UV from the illuminant . Because UV light hitting the specimen might cause fluorescence i.e.. light re-emission in a visible band. In such a case, the reflectivity of the specimen in that band would be remarkably enhanced, and might even rise above 1!

Now that we understand the reason for the filter, we have informed our decision as to its desirability. For proofing, we are simulating one substrate with another, and we really don't want some accidental fluorescence of the proofing paper to perturb our simulation of the printing paper. So we filter out the UV.

On the other hand, as a photographer, I expect people to view my prints in real-life conditions which may well contain some UV. So when making photo prints I would not want to filter out the UV.

The DTP70 is exceptional in that the UV filter can be flipped in and out of the illuminating path. But our discussion is also informative as regards the Pulse and the EyeOne: Printing-trade members might wish to purchase the more expensive UV-cut versions of these instruments, while the photo trade can safely employ the cheaper standard unfiltered models.

DTP70 and Monaco Profiler Review

I've been organizing review material for the fall. The first new item which has come in is Xrite's DTP70 and Monaco Profiler Platinum, together with a DTP92 screen calibrator. I expect it'll take a while for me to form an opinion about this profiling system, so information will get added here dribble by dribble. It will then be collated and summarized on Publish.com.

The DTP70 is Xrite's star product, an automatic chart reader that can gobble up whole A4 pages. Monaco Profiler Platinum is Xrite's high-end profiling software, and it can profile monitors, scanners, digital cameras, and RGB and CMYK printers. It also includes a profile editor.

What amazed me about the DTP70 package is how quickly it can all be set up. I had my first inkjet profile cooked one hour after opening the package, and it looks pretty good! The device is fuss-free, the documentation is not quite up to date but clearly written and seriously useful. And the software itself is truly user-friendly.

In fact, the remarkable thing about the whole product is that you don't get lost while using it. The software presents exactly the right amount of information and choices to allow you to advance through the profiling process, and also allows you to review and change your earlier choices if you want to backtrack. Did I say it's user-friendly ?

If you don't understand something while advancing through the software screens- I was wondering why I was being offered a choice of rendering intents when I thought all would be generated automagically — then the manual will supply the necessary information -apparently there is an indication in the profile of which rendering intent should be considered the default option.

I printed out a synthetic colorchecker on my Epson 2100 inkjet, and compared it with the real one, cutting up strips of the paper chart to superpose them on the paint squares. Viewed by tungsten lighting, the results was very nice color-wise, with just two squares slightly off - the perennially difficult red, and the "light skin analogue". Moreover, the gray square density of the print matched the original squares much better than in any profile I ever got from my Eyeone system.

Tomorrow I will go back and check the Colorchecker itself - I'll use the EyeOne to measure the squares, edit the values into the print file, and do another print and compare. The chart image and square values I used today were averages obtained from Danny Pascale's site, www.babelcolor.com

Some of you may wonder why I don't have a measured Colorchecker already, seeing the amount I write about CMS. Well, I do have data that adequately represented my old chart, but I've mislaid my old cardboard-backed chart. So I had to go out and purchase a new plasticky one. And, truth be said, the red square on the new one doesn't agree with my memory of the previous one !

I'm going to do some more user-testing of this system in the coming weeks and months. However, I beg the reader's indulgence, I am a journalist, and a color-management user as a photographer. I am not a color-scientist.