Monitor calibration sensor evaluations

[Pictus]

I love the iterative nature of CEDP's process. CEDP and i1P with and EyeOnePro is a combination I'm studying on a variety of displays right now. I'm finding that i1P has an edge on some displays while, surprisingly, CEDP can have an edge on others. So there's a need to analyze the broader landscape and make educated, forward thinking recommendations to clients.

Probably with Eizo models as CEDP can access the monitor internal LUT.

[shewhorn]

I love the iterative nature of CEDP's process. CEDP and i1P with and EyeOnePro is a combination I'm studying on a variety of displays right now. I'm finding that i1P has an edge on some displays while, surprisingly, CEDP can have an edge on others. So there's a need to analyze the broader landscape and make educated, forward thinking recommendations to clients.

I've been revisiting things a bit. Today I've been playing with BasICColor and CEDP using the i1Pro. A few observations when paired with the i1P... BasICColor's calibrations run much faster where as CEDP tends to linger on each path a little longer. I'm not sure if that's a good thing or not, especially where shadow detail is concerned. Until tonight I've not noticed this. While it's well known that the i1P is noisier when measuring the shadows, I've never observed this to be a problem using Spectraview on my 2690 or CEDP on my other screens but when doing validations with BasICColor I've noticed DRAMATICALLY different results for the black point and as such, wildly varying evaluations for contrast ratio. On one pass I'll get 400:1 and on the next... 750:1.

Perhaps this behavior is also present in CEDP but I'm usually pretty observant and I think I'd catch variations of 0.18 cd/m^2 on 1 pass to 0.35 cd/m^2 measured on the next (well... at the very least I'd like to think that I wouldn't miss something like that ). That's just not something I've ever noticed with either the US version of Spectraview or CEDP (when using the i1P). When using the Spyder 3 with BasICColor the validations are consistent. I can only theorize as to why this might be but it would seem that BasICColor might not be taking as many measurements per patch as Spectraview (US version) and CEDP are thus resulting in erratic measurements in black point from validation to validation.

I'll have to pay attention to i1Profiler to see what it does as it obviously whips through the patches quite quickly.

Cheers, Joe

[Czornyj]

While it's well known that the i1P is noisier when measuring the shadows, I've never observed this to be a problem using Spectraview on my 2690 or CEDP on my other screens but when doing validations with BasICColor I've noticed DRAMATICALLY different results for the black point and as such, wildly varying evaluations for contrast ratio. On one pass I'll get 400:1 and on the next... 750:1.

i1pro gives inconsistent results when measuring the shadows, so changing the profiler may give better or worse results, but so or so it's still a lottery. Just take a look a that video - I've measured my 3090WQXi bkpt with i1pro and basICColor DISCUS simultaneously:
http://dl.dropbox.com/u/19059944/i1pro-vs-dscs.MOV

(the DISCUS measurement is the Y coordinate on the left (it reports 0,275cd/m^2), the i1pro measurement is the Y coordinate on the right (it's changing its mind each time the averaged measurement is done: 0,19 - 0,27cd/m^2)

[shewhorn]

i1pro gives inconsistent results when measuring the shadows, so changing the profiler may give better or worse results, but so or so it's still a lottery. Just take a look a that video - I've measured my 3090WQXi bkpt with i1pro and basICColor DISCUS simultaneously:
http://dl.dropbox.com/u/19059944/i1pro-vs-dscs.MOV

(the DISCUS measurement is the Y coordinate on the left (it reports 0,275cd/m^2), the i1pro measurement is the Y coordinate on the right (it's changing its mind each time the averaged measurement is done: 0,19 - 0,27cd/m^2)

I definitely don't dispute that it's noisy (and I'm seeing the exact same behavior with luminance jumping all over the place with the i1P when measuring the blackest blacks) but I think when considering practical real world use that I've found (and I believe Scott may be of the same opinion) that in many cases the benefits of using the Eye One Pro outweigh its noisy performance with dark patches. I have to double check but I believe I also set the black point target on my NEC to 0.3 or 0.35 cd/m^2 so it's possible that that might be high enough that I'm not seeing as much flakiness resulting from the i1P's noise floor. With the testing I'm doing right now I have everything set to minimum black levels.

I'd love to have that Discus you have there... if you'd like to send it to me 

Just a side note... while blacker blacks (I feel a Spinal Tap quote coming) are always desirable, when it comes to profiling my monitors I place more importance on being able to see each patch discretely. Of course increasing your black level above the absolute minimum that your monitor can achieve can definitely lob off some of that resolution my personal experience has been that regardless of whether or not I'm using a spectro, or a colorimeter, there always seems to be a bit of funk in the darkest shadows (funk in this case is a color cast) on monitors with really low black levels that is often mitigated by bumping up the black levels a smidge so in real world use the noise present in spectrophotometers hasn't been a show stopper. As mentioned before I actually prefer my Eye One Pro over my other pucks (most of the time but not all of the time) as on the whole it definitely renders profiles that are more neutral. I reserve the right to be fickle and change my mind of course. 

Cheers, Joe

[Czornyj]

Generally, I had also preferred my i1pro over the other pucks. But knowing its limitations, I simply find it inconsistent when measuring the shadows, and don't belive any specific profiling software can really help it. Even Argyll CMS with its own i1pro driver (that have larger integration times) doesn't cure the issue, as can be seen in Ethan's evaluation results.

[shewhorn]

Generally, I had also preferred my i1pro over the other pucks. But knowing its limitations, I simply find it inconsistent when measuring the shadows, and don't belive any specific profiling software can really help it. Even Argyll CMS with its own i1pro driver (that have larger integration times) doesn't cure the issue, as can be seen in Ethan's evaluation results.

No doubt... weakest link. etc... created a correction matrix for the Spyder 3. Waiting for Argyll to finish up. Lots of hurry up and wait! For this much waiting.... it had better be good! 

Cheers, Joe

[Mark Paulson]

Mark: If the offer still stands, I may take you up on it after we get a chance to run the fisrt sample through its paces.

Just let me know when you want it. Scott is here in San Antonio and he can also have a test with it if desired. I don't have time as I have another full time job.

[Ethan_Hansen]

Joe: What you are seeing with the i1Pro is indeed the limitations of its measurement capabilities. From our results, the i1Pro has a real-world noise floor of between 0.3 - 0.4 cd/m2. At lower luminance values, i1Pro numbers are more random than real. When you raise the blackest black from 0.11 to 0.35, you are getting more signal, less noise, and the green casts to the shadows go away. (In all likelihood, the true monitor black point is 0.05 - 0.1 cd/m2 higher than what the i1Pro is reporting).

CEDP increases measurement integration time in darker tones above what other (non-Argyll) programs do. The price you pay for BasICColor Display's speed is inaccurate shadow readings. Measure the same monitor with a DTP-94, and BasICColor should give more consistent black levels. The white point will be all over the place, however.

I spent more time with i1Profiler over the weekend. When choosing "native" contrast ratio, it gives a higher black point than does CEDP's relative black, no matter whether an i1Pro or i1D2 is used. At the black level CEDP chooses, a good monitor (e.g. Eizo CG243W) is capable of differentiating RGB (1, 1, 1) from (0, 0, 0). By setting a higher level, i1Profiler both gets around the limitations of the i1Pro and allows faster measurement times. Whether this works for you depends on how black you like it.

The absence or presence of color casts at the lowest luminance levels appears to be a function of the sensor being used. One calibration suite (CalPC) we are evaluating supported our PR-730 spectroradiometer. With a monitor that it could talk to via DDC, it produced dead-neutral shadows down to 0.15 cd/m2. BasICColor Display also supports fancy equipment, just in case anyone has a spare KM CS-2000 floating around. Once we get a Discus in-house, we will check BasICColor's performance down in the mud.

[shewhorn]

Joe: What you are seeing with the i1Pro is indeed the limitations of its measurement capabilities. From our results, the i1Pro has a real-world noise floor of between 0.3 - 0.4 cd/m2. At lower luminance values, i1Pro numbers are more random than real. When you raise the blackest black from 0.11 to 0.35, you are getting more signal, less noise, and the green casts to the shadows go away. (In all likelihood, the true monitor black point is 0.05 - 0.1 cd/m2 higher than what the i1Pro is reporting).

Well it's good to get some validation that the numbers I've come up with based on observation are grounded with some objective measurements. I think I did my testing with the Eye One Pro about a year ago.

CEDP increases measurement integration time in darker tones above what other (non-Argyll) programs do. The price you pay for BasICColor Display's speed is inaccurate shadow readings.

It was pretty darn noticeable with some whackadoodle results on the shadow side of things. As far as the Eye One Pro is concerned, purely from an observation standpoint it seems like CEDP (or i1Profiler) is a better match for the Eye One Pro.

Measure the same monitor with a DTP-94, and BasICColor should give more consistent black levels. The white point will be all over the place, however.

Sigh, wide gamut monitors offer something I need but I cried a little when I had to put the DTP-94 away. Although I've spent a fair amount of time evaluating Argyll in the past for print profiling, I hadn't really dug into its display profiling capabilities. I'm curious to know if creating a correction matrix for the DTP-94 when used in conjunction with a screen that has white LED backlights will help any? I suspect not as my guess is that the problems the DTP-94 suffers with white LED backlights has to do with the spectrum of the light produced. I do still have one screen though (with a white LED backlight... and it's a REALLY sucky screen ([COUGH]MacBook Pro[/COUGH])) that can use all the help it can get. It's actually profiled quite well right now with CEDP but if I didn't try to do better I'd have to turn in my nerd card...

At the black level CEDP chooses, a good monitor (e.g. Eizo CG243W) is capable of differentiating RGB (1, 1, 1) from (0, 0, 0). By setting a higher level, i1Profiler both gets around the limitations of the i1Pro and allows faster measurement times. Whether this works for you depends on how black you like it.

I know all the reviewers rave about a nice black point but to be honest, to me it's all relative. What's more important to me is seeing as many discrete steps as I possibly can. If I can do that with a lower black point, great but if not, I'm not all that bothered. Of course increasing the BP to improve BP performance means that you have to sacrifice resolution somewhere else... whether or not that will manifest visibly of course depends upon the design of the monitor (resolution of the panel, resolution of the monitor LUT and... ahem... in the case of Asus whether or not someone who knew what they were doing was responsible for writing the firmware  , with the NEC PA series and the Eizo CG series that's often trivial as it has the resolution to do that).

The absence or presence of color casts at the lowest luminance levels appears to be a function of the sensor being used. One calibration suite (CalPC) we are evaluating supported our PR-730 spectroradiometer. With a monitor that it could talk to via DDC, it produced dead-neutral shadows down to 0.15 cd/m2. BasICColor Display also supports fancy equipment, just in case anyone has a spare KM CS-2000 floating around. Once we get a Discus in-house, we will check BasICColor's performance down in the mud.

How I wish . I'll be interested to see what you find once you get your hands on a few Discuses... (Disci?). I suspect they probably offer the greatest bang for the buck in terms of monitor calibration. What's the next step up from there? Sencore OTC1000/X-Rite Hubble?

Cheers, Joe

[Scott Martin]

Just a side note... while blacker blacks (I feel a Spinal Tap quote coming) are always desirable....

That's fantastic - LOL!!

[gromit]

At the black level CEDP chooses, a good monitor (e.g. Eizo CG243W) is capable of differentiating RGB (1, 1, 1) from (0, 0, 0).

It's debatable how useful this is. With a default gamma of 2.2, this represents an L* value of 0.0045 ... namely something that will print indistinguishable from black.

[shewhorn]

It's debatable how useful this is. With a default gamma of 2.2, this represents an L* value of 0.0045 ... namely something that will print indistinguishable from black.

In which case it will be reflected accurately on screen when you apply the soft proofing profile and you won't have to wonder whether or not what you're seeing is accurate. The whole point of color management is quality control and having the ability to manufacture a product to a specific set of tolerances consistently and reliably over time. To that end, it is quite useful in my opinion. When something goes wrong having solid quality control procedures in place reduces the amount of time it takes to identify the problem which saves time and money. :-)

Cheers, Joe

[gromit]

In which case it will be reflected accurately on screen when you apply the soft proofing profile ...

If you're seeing differences this small, the representation of shadows on your monitor is bogus ... whether soft-proofing is invoked or not.

[shewhorn]

If you're seeing differences this small, the representation of shadows on your monitor is bogus ... whether soft-proofing is invoked or not.

I have some papers I've profiled that will show differences down to 8,8,8 and when soft proofing is enabled, this is accurately reflected in the soft proof. If things are set up properly, yes it will work and it will work quite well.

Cheers, Joe

[gromit]

I have some papers I've profiled that will show differences down to 8,8,8 and when soft proofing is enabled, this is accurately reflected in the soft proof. If things are set up properly, yes it will work and it will work quite well.

You're missing the point, this has nothing to do with soft-proofing. Soft-proofing goes through the same mapping to the display. If the shadow values are represented inaccurately, they'll be inaccurate whether soft-proofing is invoked or not.

[eronald]

So far the best profile I got for my MacBook Pro 2011 17" is ColorMunki.

I don't do print or contract print so I don't need super accurate, though.
I do need something pleasant to look at.

Edmund

[shewhorn]

You're missing the point, this has nothing to do with soft-proofing. Soft-proofing goes through the same mapping to the display. If the shadow values are represented inaccurately, they'll be inaccurate whether soft-proofing is invoked or not.

Just because a print on a particular paper made with a particular printer can or cannot represent detail in a certain area is not a reason to have the monitor mimick that medium exactly. What if the output medium is another monitor and I want to represent as much shadow detail as possible? If you do what you propose (not being able to distinguish between the difference between any two given steps) then you're undermining the usefulness of your tools as a reference medium. We profile our screens to a specific known standard. Your post is the first time I've ever seen anyone suggest that maintaining those standards are not useful.

Ehtan said At the black level CEDP chooses, a good monitor (e.g. Eizo CG243W) is capable of differentiating RGB (1, 1, 1) from (0, 0, 0). and you questioned whether that was useful. You then said "If you're seeing differences this small, the representation of shadows on your monitor is bogus ... whether soft-proofing is invoked or not." If what you're saying is true, then Eizo, NEC, Integrated Color, BasICColor, and X-Rite are all doing their jobs wrong.

An analogy... I have a piano. It has 88 notes and every note is tuned to a specific frequency so the piano is calibrated. When I play, not every song I play hits the keys in the bottom octave. The fact that that octave is there though, tuned, and accurate and capable of reproducing those notes doesn't mean that the rest of what I play on the piano is inaccurate if the music doesn't call for those specific notes.

Cheers, Joe

[gromit]

Just because a print on a particular paper made with a particular printer can or cannot represent detail in a certain area is not a reason to have the monitor mimick that medium exactly.

You're conflating two different things. I suggest you re-read what I said.

[shewhorn]

You're conflating two different things. I suggest you re-read what I said.

It appears that what you're saying is, if you can see a difference between 0,0,0 and 1,1,1, then your monitor is not capable of accurately representing a print, and that being able to see a difference between 0,0,0 and 1,1,1 is not useful. Is that not what you are saying? If not I may be misunderstanding what you are trying to say.

[gromit]

It appears that what you're saying is, if you can see a difference between 0,0,0 and 1,1,1, then your monitor is not capable of accurately representing a print, and that being able to see a difference between 0,0,0 and 1,1,1 is not useful. Is that not what you are saying?

For a gamma of 2.2 (monitor calibration and working space) ... yes. Not only will the representation of shadows in the soft proof be inaccurate but also that of detail in the image file itself. For L* it's a different story.