Monitor calibration sensor evaluations

[shewhorn]

Argyll won't modify colorimeters LUT. At least not yet...

No, definitely not so it's of no benefit if you want to use another application but if you're staying within Argyll then you can take advantage of the correction matrix. Unfortunately even with the DispCal GUI it's not exactly what I'd call an easy to use app for the masses (actually I prefer the command line over DispCal GUI myself as I find it easier to use but... a well done GUI would certainly be a welcome addition).

Cheers, Joe

[Ethan_Hansen]

Very interesting - as a logical conclusion it seems that every NEC OEM sensor is individually calibrated...

I'd really like to have some kind of software, that could internally tune an i1d2 to a specific display using some spectroradiometer - I suppose even i1pro or ColorMunki could make things better, and maybe it also could fix the problem with i1d2's tendency to degrade over time.

Speaking of stability - any information regarding basICColor DISCUS testing?

I have not heard anything from X-Rite directly about how one can calibrate an Eye-One Display. The OEM vendors we spoke to did not know if this was covered under NDA, but the suspicion is that it was. It certainly would be revealing to try tuning an i1D2 to a particular screen and seeing how it could perform.

On to the BasICColor DISCUS. Thanks to BasICColor and CHROMiX, we have a unit in-house. The performance was impressive. For full details, read the updated article on our site or skip ahead to the conclusions.

The white measurement error on the wide-gamut display showed the DISCUS at its worst. This panel used RGB LED backlighting. On other wide-gamut monitors having CCFL backlights (e.g. Eizo, NEC PA-series. LaCie) the DISCUS measured in at between 0.75 - 1.4 dE-2000 from where our spectroradiometer claimed the monitor really was. Those values are simply impressive. The DISCUS' white point error is less than the across-screen uniformity of even most high-end monitors.

All our measurements used the DISCUS release of BasICColor Display (v4.2.4) to drive the instrument. There are many rough edges in this release - video LUTs not flushing consistently, DDC calibrations ending up at D50 no matter what white point was specified, missing i1D2 support, and, of course, the dispute with ICS making L* calibration unavailable unless you state you do not reside in the US. Even so, it did allow accurate measurements of monitor black and white points.

The DISCUS - at least the unit we tested - did show one notable flaw. The puck includes an ambient light measurement diffuser. Although measurements of monitors were very accurate, ambient measurements were not. We consistently obtained readings with too low a color temperature, the equivalent of at least 6 dE-2000 low. Also, the DISCUS has thermal compensation capability. For monitor measurements, this worked like a charm -- only minimal shifts with a 10C change in temperature. Ambient light measurements did change, with the DISCUS reporting the color temperature of a constant light source rising by over 300K as the puck heated up.

I do not know if this indicates a problem in our measurement methodology (light source too bright?), a QA flaw in the puck we measured, contamination or damage to the ambient light diffuser, a systematic problem such as thermochromic diffuser material, or something else entirely. Too soon to say, but if any of you DISCUS owners also have an i1 Beamer or even an i1D2 with the ambient head, try measuring a viewing booth or other constant, 5000 - 6500K light source with the DISCUS and the second instrument. See if there is a consistent offset, particularly if the DISCUS always reads low.

[Czornyj]

Ethan - problem with measuring chromatic coordinates of ambient light is that in basICColor display 4.2.4 there's only a correction matrix for JUST Normlicht CCFL spectra available. So it musn't be reliable with other light sources, I suppose.



As for temperature sensor - I've noticed, that it's not correcting measurements internally, but the correction is applied in the profiling software. It works while profiling, but not while you're measuring the x,y coordinates of the wtpt (in ambient or contact mode), the temperature change can cause a drift, so recalibration of the instrument is needed to counteract this behavior.

Anyway, basICColor display 4.2.4 is the first release that supports the DISCUS, so there may be some minor early age issues. And there's basICColor display 5 coming soon...

Thanks for sharing the test results!

[tony22]

I've been focusing my testing on The Spyder3Elite and EyeOnePro devices using ColorEyesDisplayPro, DispCal/Argyll, Syper3Elite, and i1Profiler. From the testing I've done thusfar, I have to say I'm pretty blown away and completely surprised with what I'm seeing from i1Profiler with an EyeOnePro. Not only does the process take less than 2 minutes (!!) but the shadows and smoothness are incredible, rivaling even DispCal's hour long calibration process. i1Profiler's QA analysis reports lower Average Delta E variations with i1Profiler's calibration than the others. I've been frustrated with shadow detail from the i1 Pro in the past but i1Profiler is doing a better job than I've seen with other solutions.

Ethan you wrote "No software wizardry will improve a spectrophotometer's accuracy at low light levels to be competitive with a colorimeter." But that's exactly what I'm seeing with i1Profiler (to my surprise as much as yours I'm sure!). Suddenly the shadows are all there statistically and visually and the performance with soft proofing is incredible.

I have been impressed with i1Profiler as well. X-Rite indeed managed to coax commendable performance out of a short measurement cycle. I assume they are using a method better than the simple averaging in Argyll to tease a signal from the measurement noise. (Serious late-night geek digression: The standard error of a measurement sample goes down with the square root of the sample size. In English, this means that taking 3 measurements cuts your random variability in half vs. 1 sample. To reduce it by half again, you need 10 measurements. Shrinking the error by half again requires 100 measurements, followed by 10000 measurements for another 50% reduction. You can see why the DispCal's high quality mode takes an hour to run).

Back to reality... Our tests of i1Profiler do indeed give more neutral and open shadows with an i1Pro than most other software. The calibration ain't perfect, however.

Ethan and Scott, I have found this discussion fascinating. Given that I'm still using i1Match, however, it leads to some questions. I've read some snippets suggesting that i1Profiler's monitor calibration software is not quite as flexible as i1Match's (setting user selected white and black point, for example). If this is true, ultimately how does the monitor calibration capability compare between the two? With all the good things said above, is there still a price to pay with less adjustability?

[ronker]

Both of these i1's turned in excellent performance on their respective monitors. Using the tuned sensor on any other monitor, however, was invitation to disaster. The results were ugly.

I have a NEC PA301W with a NEC OEM EOD2 (MDSV2) sensor. I've thought about use the sensor on other monitor especially with my NEC 2180UX. Do you think there will be also bad results on a 2180UX? I've read on the NEC homepage that the MDSV2 ist backward compatible to the old sensors.

[mjdl]

[...] Unfortunately even with the DispCal GUI it's not exactly what I'd call an easy to use app for the masses (actually I prefer the command line over DispCal GUI myself as I find it easier to use but... a well done GUI would certainly be a welcome addition). [...]

I'm just a newbie user of ArgyllCMS, dispcalGUI, and a DTP-94, but I really have to disagree that the dispcalGUI application is difficult to use: just set the desired display calibration parameters in the upper part of the application, select the kind of display profile you want, choose one of the provided colour target sets in the lower part of the application (and customize it if necessary, to produce a new target set), and that's pretty much it--make sure you deselect "Interactive display adjustment" (assuming you've already adjusted the display controls), so that you can go off and do something else for the hour or more it takes for a high quality calibrate+profile process to finish--it will finish unattended.

Of course I've had to do a fair amount of reading to understand the general principles of operation and to ensure the options I've selected in dispcalGUI make sense, but the latest version of the application (0.7.x.x) does have a detailed, long page of instructions.

Granted, my ThinkPad matte display is nothing compared to the kinds of LCDs mentioned already in this thread (it only has 2/3rds the colour volume of sRGB), but the verification results of one profile, attached, seem O.K. to me, and the actual visual performance a vast improvement over the uncalibrated state, i.e. now there is a neutral grayscale, good separation in shadows and highlights on all kinds of photos, etc.

What is really difficult, however, is going beyond the basic Argyll calibration parameters: e.g., I need to figure out how to correct the very slight over-warming of dark greys in the range RGB(10,10,10) to RGB(30,30,30) and of highlights RGB(235,235,235)+ compared to the middle grey values.

[Czornyj]

there's only a correction matrix for JUST Normlicht CCFL spectra available. So it musn't be reliable with other light sources, I suppose.

I guess I was right:
http://www.basiccolor.de/dry-creek-photo-testet-discus/

Einziger Kritikpunkt ist die Lichtmessung, hier wurde im Test ein nicht näher spezifiziertes Umgebungslicht gemessen, der basICColor DISCUS ist jedoch auf Normlicht kalibriert und liefert deshalb nur hierfür perfekte Messwerte

[stefohl]

Comparing Basiccolor Discus with KonicaMinolta Display Color Analyzer CA-210

Today I did a test on the Discus by comparing it to a CA-210. I've done several such tests in the past with EyeOne Pro, Spyder III and EyeOne Display. The DeltaE was below 1 when we compared the measurements from the Discus and the CA-210. When I tested the other instruments the Delta E on a calibrator that worked OK was between 3-5.

I've tested more than 15 EyeOne Displays now, and I estimate that more than 30% of them showed results that were unacceptable, with an average DeltaE over 10, compared to the measurements we got from our reference calibrator. I haven't tested any that was tuned to a specific monitor, so I don't know how good the results can be in that case. But I don't recommend any of the photographers that we are working with buying a generic EyeOne Display.

The results from the Discus was very impressive. I'm happy to see that we now have a calibrator that can do a very satisfactory calibration even on a wide gamut monitor.

Best/
Stefan

[trinityss]

The only device missing in this test is the new xrite display pro :-)
I would now like to see a comparison with the discuss!

Kr,

[pcunite]

The only device missing in this test is the new xrite display pro :-)
I would now like to see a comparison with the discuss!

Indeed, we know that the plastic i1 Display 2 was junk, but what about the i1 Display 3? It is getting good reviews here:
http://www.curtpalme.com/ChromaPure_EyeOneDisplay3.shtm

[digitaldog]

Indeed, we know that the plastic i1 Display 2 was junk, but what about the i1 Display 3? It is getting good reviews here:
http://www.curtpalme.com/ChromaPure_EyeOneDisplay3.shtm

I’d be worried by any review that can’t correctly name the product. There is no such thing as an EyeOne Display 3, at least as far as X-rite is concerned. No, its not officially called the i1 Display Pro III.

[WombatHorror]

Iliah - excellent point, and one i should have mentioned. We looked into this with a couple of monitors. Setting black to 0.5 - 0.8 decreased the mean measurement error from ~10dE to ~5. Big improvement if you are willing to live with lighter blacks levels. This is a valid option, particularly for anyone printing on fine art papers or other stock with relatively low print contrast.

AFAIK SV II used with NEC PA monitors writes but a simple matrix profile and it doesn't even use the probe to set the gray-scale, it simple tunes the color temperature using a brightish white patch and then the 3D 13bit LUT and color engine do all the rest. It does use the probes to show you afterwards how it measured the gray-scale but since it just writes a matrix profile I don't believe it actually write that to the profile. So the poor dark readings of the i1Pro should have no effect when using SV II and NEC PA series.

Of course it is also true that an i1Pro costs like $800+ and the NEC i1D2 like $200 and when I compared the values that each measured for color temp of a bright patch and for the primary locations most values were 0.000-0.001 apart, the same, might have a couple .002s I forget, at most a single .003 but I think .002 was the largest difference I got. (xy values of xyY) so if all you do is use the NEC and are not calibrated other things as well, then their rebadged and specially calibrated probe is heck of a lot less expensive way to get the same results.

[WombatHorror]

The answer is I don't know. The two OEM-adjusted sensors we have are a NEC sold with the Spectraview package and HP's DreamColor unit. Our assumption is that these two Eye-One flavors are adjusted differently. The NEC sensor performs well on the PA monitor, but is grossly inaccurate on either sRGB monitors, the HP (uses LED backlighting rather than CCFL), and even wide-gamut Eizo displays that are also CCFL lit. The HP puck has similar behavior: excellent performance on the monitor it is sold with, lousy on all others.

If these two OEM flavors of i1 are actually identical, that would imply a truly worrying lack of consistency. We asked both HP and NEC for details. None were forthcoming. Without a reasonable sampling of OEM pucks to characterize there is no way to be sure. X-Rite, to the best of my knowledge, has not published accuracy and resolution specs for the Eye-One colorimeters like they do for the i1Pro spectros.

As I stated in another post I got basically the exact same results on a NEC PA using either an i1pro or the NEC i1D2 for bright gray color temp measurements and primary location measurements. When I used the NEC i1D2 on my HDTV the results seemed pretty far off from my DTP94 though. Only 1 sample of each here but these seem to match exactly to what you found.

[WombatHorror]

I've been focusing my testing on The Spyder3Elite and EyeOnePro devices using ColorEyesDisplayPro, DispCal/Argyll, Syper3Elite, and i1Profiler. From the testing I've done thusfar, I have to say I'm pretty blown away and completely surprised with what I'm seeing from i1Profiler with an EyeOnePro. Not only does the process take less than 2 minutes (!!) but the shadows and smoothness are incredible, rivaling even DispCal's hour long calibration process. i1Profiler's QA analysis reports lower Average Delta E variations with i1Profiler's calibration than the others. I've been frustrated with shadow detail from the i1 Pro in the past but i1Profiler is doing a better job than I've seen with other solutions.

Ethan you wrote "No software wizardry will improve a spectrophotometer's accuracy at low light levels to be competitive with a colorimeter." But that's exactly what I'm seeing with i1Profiler (to my surprise as much as yours I'm sure!). Suddenly the shadows are all there statistically and visually and the performance with soft proofing is incredible.

Interesting. I didn't try the new i1profiler yet, the original x-rite profiler software that came with my i1pro was awful IMO, way worse than ColorEyes (one unfortunate thing about ColorEyes is that you can't chose an sRGB tone response curve).

One thing you can do as far as poor i1pro shadows with some programs is use the i1pro on bright colors to train say a DTP94b. I know Calman for HDTVs allows that and some other programs do as well, mostly more HDTV oriented ones I think. I think the free open-source monitor one does too though now actually.

[WombatHorror]

One last comment. The high quality mode in Argyll/DispCal has a potential gotcha: temperature drift. The ii1Pro is fairly sensitive to temperature. Even under controlled ambient temperature there can be shifts from the i1 heating up as is rests on the screen. We saw this firsthand when trying ridiculously long integration times to see what the absolute noise floor was for an i1Pro. We started with a warmed-up monitor, and saw the i1 readings progressively change over the course of 20 minutes. Our reference PR-730 is both temperature-compensated and was being used in non-contact mode. After the first 20 minutes, the i1 readings did not shift for the 24 hour duration of our test. Moral of the story: preheat your i1Pro by hanging it on your screen while you surf the web for 20 minutes or so before you calibrate.

yeah temperature drift with many of the instruments has scared me a bit about the profilers that take a long time
hopefully i'm wrong but it seemed that my i1pro drift again and again over time not just for the first 20 min and then stable

[WombatHorror]

Ethan, just to follow up. I'm continuing this testing and am finding that on some displays (like a Samsung 245T and Dell 24") I'm seeing dramatically improved results (visually and statistically) using an EyeOnePro device instead of a DTP94, even when the same software is used.

One of the interesting things that I see but suspect most people don't is how different two different types of monitors can look when calibrated the same way. Put a Cinema Display and a Samsung monitor on the same Mac and calibrate them both using the exact same settings and marvel at the disappointing differences you'll see. Better yet, take a client like Whole Foods World HQ where they've got 50+ designers and video professionals all in one area using a hodge podge of different brands of displays. I'm finding that if they are all calibrated using a colorimeter (Spyder3 or DTP94) when you stand back and look at all of them in one room it's kinda surprising how much inconsistency there is between them. Calibrate all of them with a spectro (using CEDP or DispCal) and they are visually perfectly consistent! Combine that with i1Profiler which does a better job at handling the shadows with a spectro than anything else and you've got a truly superior solution. Problem solved.

While lots of users may only have one or two displays these are the real world challenges (10+ different types of displays all side-by-side in one room) that my business faces every day. My hands-on testing is showing that spectros have advantages over even the best colorimeters in some situations and with i1Profiler I don't see any problems with the shadows like we've seen elsewhere. I'm not seeing any color crossovers or drawbacks with my particular pair of EyeOnes.

I'm going to stick to my guns here and suggest that, for now, i1Profiler with an i1Pro seems to be the answer to the question "when displays are calibrated with a variety of devices and applications which *combination* consistently yields the best results within a reasonable timeframe?"

there is also the problem of metamerism where the different spikes in the primaries of different monitors, particularly wide gamut vs. standard gamut may lead to measured results being identical but visual results differing. My Samsung 244T and Samsung C650 looked more similar than my NEC PA 241PA due to metamerism. (comparing to real world color checker chart under D65 this looked different yet again but it seemed closer to how my eye sees things to the NEC PA wide gamut actually) and more complicated is that the degree and nature of the metamerism seen in these cases is thought to vary somewhat person to person

[Ernst Dinkla]

One thing you can do as far as poor i1pro shadows with some programs is use the i1pro on bright colors to train say a DTP94b. I know Calman for HDTVs allows that and some other programs do as well, mostly more HDTV oriented ones I think. I think the free open-source monitor one does too though now actually.

With a spectrometer ArgyllCMS can create a correction matrix for a colorimeter, both used on the same monitor. In profile creation that correction matrix will adapt and correct the colorimeter to the specific monitor used and keep the colorimeter darker color measurements quality. It would be nice to see the colorimeter test widened up with test that have that correction included. Creating the correction matrix is a fast and easy process.

met vriendelijke groeten, Ernst


Dinkla Gallery Canvas Wrap Actions for Photoshop

http://www.pigment-print.com/dinklacanvaswraps/index.html

[Czornyj]

AFAIK SV II used with NEC PA monitors writes but a simple matrix profile and it doesn't even use the probe to set the gray-scale, it simple tunes the color temperature using a brightish white patch and then the 3D 13bit LUT and color engine do all the rest. It does use the probes to show you afterwards how it measured the gray-scale but since it just writes a matrix profile I don't believe it actually write that to the profile. So the poor dark readings of the i1Pro should have no effect when using SV II and NEC PA series.

SVII measures the native TRC of the display in 8, 16, 32, or 52 steps, and puts correction curves in high-bit LUT of the display. So poor dark readings of the i1Pro have obvious and noticable effect.

[RichWagner]

You understand our motivation perfectly. We are out to update our reviews of monitor profiling systems. Looking at hardware-agnostic packages (ColorEyes, BasICColor, CalPC, etc.) that support a huge range of sensors is daunting. Yes, it is pretty cool to be able to hook a Minolta CS-2000 up to BasICColor or our PR-730 to CalPC, but this doesn't tell you much about real-world performance.

Unfortunately, the number of devices that these excellent software packages support will likely decrease in the near future, thanks to new policies from X-Rite.

A few weeks ago X-Rite announced the availability of two new colorimeters: the i1 Display Pro and the ColorMunki Display.

Here is the current status on third-party software support for these X-Rite products, as noted by a third-party software developer:

ColorMunki Display

As in the past for previous Munki instruments, there is no third party Software Development Kit (SDK) available. An SDK is essentially the drivers and software libraries that enable third party developers to communicate with the instrument. This does not mean that the SDK does not exist, it only means it is not available; in other words, this is a business decision! Thus, until such an SDK is available, third party developers cannot provide support for the Munki series of instruments.

i1 Display Pro

This model replaces the i1 Display2. As with the i1 Display2, there are three Categories of the i1 Display Pro available:

Category (a)- The instruments sold by X-Rite under the X-Rite name (also called the retail version). In the past, for the older i1 Display2, the SDK could be used for instruments in this category; this is NOT possible with the SDK of the i1 Display Pro.

Category (b)- The instruments sold by a third party (called Original Equipment Manufacturer, or OEM). The SDK could, and can still be used for instruments in this category. However, in the past, such instruments had the same communication interface specifications as the X-Rite retail model; this is not the case anymore.

Category (c)- The instruments sold by an OEM, which have communication interface specifications DIFFERENT from the Category (b) model. The SDK cannot be used to communicate with those instruments.

For example, the custom i1 Display2 recommended by NEC for many wide-gamut NEC monitors (called the MDSVSENSOR2) is a Category (b) model while the HP-branded i1 Display2 (HP DreamColor Advanced Profiling Solution, KZ300AA) sold by HP for use with the HP DreamColor monitors is a Category (c) model.

(Note that while the NEC MDSVSENSOR2 has the same communication interface as the standard i1 Display2, it has a different set of sensor filters which is better adapted to the wide gamut primaries of many NEC monitors.)

With the older i1 Display2, the X-Rite SDK could be used with the Category (a) and (b) instruments; this explains, for instance, why third party software can be used with the X-Rite i1 Display2 and the NEC i1 MDSVSENSOR2, but not with the HP DreamColor i1 Display2.

With the new X-Rite policies, the i1 Display Pro sold by X-Rite can ONLY be used with i1 Profiler; it CANNOT be used with the i1 Display Pro SDK. This means that third party developers cannot support this instrument. Third party developers can only support the Category (b) and (c) instruments (Category (c) instruments will also require a separate agreement between the third-party developer and the licensed OEM). In addition, Category (b) and (c) instruments CANNOT be used with the i1 Profiler software (or other X-Rite software); if you want to use i1 Profiler, you need to buy a retail i1 Display Pro from X-Rite.

These new policies severely limit customer benefits since the new colorimeters can only be used with the bundled software they are sold with (from either the OEM or from X-Rite). This really stinks.

These changes do not affect the i1 Pro (a spectrophotometer that provides spectral data) as it is not currently being replaced by a new model, and is still supported by the older third party X-Rite SDK.

For those concerned about the implications of these new X-Rite policies - in particular, if you think that you could benefit from using third party software with an X-Rite branded ColorMunki Display or i1 Display Pro -  I suggest you communicate with your X-Rite representative or file a complaint via the general support page:
http://www.xrite.com/contact_us.aspx?reasonid=3

Personally, I will not buy another X-Rite product until these policies change.

--Rich Wagner

[RichWagner]

With a spectrometer ArgyllCMS can create a correction matrix for a colorimeter, both used on the same monitor. In profile creation that correction matrix will adapt and correct the colorimeter to the specific monitor used and keep the colorimeter darker color measurements quality. It would be nice to see the colorimeter test widened up with test that have that correction included. Creating the correction matrix is a fast and easy process.

The procedure to derive a color correction matrix for colorimeters for specific displays using a reference instrument like an eye-one pro is described here:

http://www.astm.org/Standards/E1455.htm

I believe the original research was described in:

Four-Color Matrix Method for Correction of Tristimulus Colorimeters
Yoshihiro Ohno and Jonathan E. Hardis
Proc., IS&T Fifth Color Imaging Conference, 301-305 (1997)

and

Four-Color Matrix Method for Correction of Tristimulus Colorimeters – Part 2
Yoshi Ohno and Steven W. Brown
IS&T Sixth Color Imaging Conference, (1998)

PatchTool can also generate and use a correction matrix. Danny Pascale has written a nice note on the process:

http://babelcolor.com/download/AN-9%20How%20to%20derive%20and%20use%20a%20Color%20Correction%20Matrix.pdf

Rich Wagner