Problem calibrating MS Surface Pro 3

[WayneLarmon]

I have one of the new Microsoft Surface Pro 3s and I like it a lot but I think that I might have a problem with the monitor profile I created.  I calibrated with an i1 Display Pro and the most recent version of iProfilier.

The Surface's gamut looks like it is very close to sRGB:



But...after calibration, it doesn't look like my NEC PA241W running in sRGB emulation mode (calibrated with the same i1DP and Spectraview).    Neutrals on the Surface are mostly neutal but some hues have a greenish tint. And there isn't enough red saturation. (Compared with the PA241W.) I need to resolve this before I can do any serious image editing on it. (The Surface runs CC Photoshop very well.)

I calibrated and profiled it with an XRite i1Display Pro using the most recent version of XRite iProfilier. (The above gamut plots were made from the profile that iProfilier generated.)  Enabling ADC (Automatic Digital Control) in iProfilier caused it to crash, so I disabled ADC. The brightness of the Surface is set at the factory default of 50%.

I made screen shots of all the steps I took in iProfilier.  I'm not going to insert them in this post but they are on this page.

Can anybody tell me what I did wrong?  Or am I expecting too much from color management (matching two monitors)?

TIA.

Wayne

[digitaldog]

How did you build the profile for this tablet specifically getting the colors over to it and measuring it?

[WayneLarmon]

I left this out of my post (but added it while you were writing your reply.)  I used an i1 Display Pro with the most recent version (downloaded from XRite) of iProfilier.

I generated the gamut plots with Argyll iccgamut and viewgam using a method similar to this.  (Actually I use an Perl program I wrote so I can generate a lot of gamut plots to study.)   But it uses the same methods that I described, above.  For example, my PA241W running in sRGB emulation mode exactly matches sRGB (VRML.)

Wayne

[digitaldog]

What is set for sRGB emulation for the NEC in terms of brightness? Is it the same as the tablet? I don't know squat about Win 8 color management or what it's doing in terms of sending data to the display. Are there any ICC aware applications on it where you can ensure that something like a ProPhoto RGB document is using the profile and it looks reasonable?

[WayneLarmon]

What is set for sRGB emulation for the NEC in terms of brightness? Is it the same as the tablet?

The NEC is set for 120 cd/m2  The Surface measured as being 92 cd/m2.  But I wouldn't think that luminance differences would affect hues.

Are there any ICC aware applications on it where you can ensure that something like a ProPhoto RGB document is using the profile and it looks reasonable?

I downloaded the Digital Outback Printer Evaluation Image (that was mentioned in another post on this forum)

The image is in ProPhoto.  I loaded into Photoshop on the Surface and converted it to Adobe RGB (1998) and saved with a different file name.  Then undid that (back to ProPhoto) and converted to sRGB and saved that with a different file name.

The stock Windows image viewer is color managed in windowed (resizable) mode but unmanaged in slideshow mode.  When I used the image viewer to view the three images in windowed mode (color managed), they look (mostly) identical. (On the "almost" sRGB Surface.)  When I look at them in slideshow (unmanaged) mode they look drastically different.

CC Photoshop is, of course, ICC color managed.  (The Surface runs full Windows 8.1 and I installed CC Photoshop a few days ago.)

Wayne

[Royce Howland]

I'm not sure you did anything wrong in calibrating the Surface Pro 3, from what I can see. Except possibly in your selection of what type of display the device has. In your screenshot you appear to have selected CCFL as the backlight type, but I believe the Surface Pro 3 uses an LED backlit panel. I'm not sure if this will make a difference or not.

That aside, I wouldn't expect these two devices to calibrate to a visual match even when using the same tool. You're comparing a big, relatively power-hungry, colour-critical desktop monitor using a wide gamut (near Adobe RGB) IPS panel with CCFL backlight and a 14-bit LUT, against a low-power, consumer mobile display with a narrow gamut (less than sRGB) IPS panel with an LED backlight and an unknown LUT (but likely 8 bit at best). Unless you force the issue somehow (i.e. dumb down the NEC to the level the SP3 can reach), simply calibrating both of these independently, even against sRGB as the benchmark, likely can't produce a complete visual match.

Some specific examples of why. The CCFL vs. LED backlight could introduce a slight colour cast difference between the two panels, even if calibrating both to D65. Also the difference in LUT's may mean the SP3 can't be linearized to quite the same level of neutrality as the NEC, leading some neutrals to be slightly cast on the SP3. The displays' native gamut differences easily could explain why the SP3 doesn't have the same level of red saturation; for example from one gamut plot I found of an SP3 review, its red primary is noticeably shifted off the sRGB red primary. Since the NEC's colour primaries all exceed sRGB, you can expect more or less a 100% gamut match to sRGB for it whereas the SP3 doesn't fully achieve sRGB gamut, in particular in the reds.

You can't necessarily "resolve this", depending on what you mean by that phrase. It's simply a fact of life that every type of device has different colour reproduction characteristics. Calibration & profiling isn't about making different devices match each other, really, because you can't over-compensate for a device that is more limited than some other device. Unless you want to restrict all devices to the lowest common denominator of operating envelopes from the least capable devices, you're going to be living with devices that don't match.

The way I'd describe it is that calibration & profiling is really about: a) making a device relatively self-consistent over time (subject to a degree of drift as the device ages); b) making the device congruent to some reference baseline, within the limitations of the device; so that c) image colours that fall fully within the envelopes of multiple devices will reproduce relatively accurately on them all, while image colours outside the range of any specific device are handled in a defined way based on other factors in the colour management system.

[WayneLarmon]

In your screenshot you appear to have selected CCFL as the backlight type, but I believe the Surface Pro 3 uses an LED backlit panel.

I didn't know that.  In iProfilier, there is a choice between "White LED" and "RGB LED".  Do you know which one I should choose?

I can't argue with what you wrote in the rest of your post--you know more of the fundamentals than I do.  But it is a shame that "color management" apparently doesn't encompass making monitors match. 

I'm not much interested in prints.  I'm a web developer.  The reason I got the PA241W was precisely to use the sRGB emulation, with the hopes that images edited on that would look at least civilized on the widest variety of uncalibrated monitors.  (Or on mini-lab printers.)  On the assumption that uncalibrated monitors would (roughly) have a Bell curve distribution.  (Nobody will accept a monitor that displays purple skin tones.  Ditto for mini-lab prints.)   I want to be in the center of the Bell curve when I'm editing images.

But if what you say means what I think it means, then the Surface isn't in the center of the Bell curve, even after calibration.  (Or maybe the PA241W isn't either?)

Do you have any comments on the way I used the Argyll utilities to generate the gamut plots?   Are there any problems with them?

Wayne

[Royce Howland]

I didn't know that.  In iProfiler, there is a choice between "White LED" and "RGB LED".  Do you know which one I should choose?

I looked for detailed specs on the Surface Pro 3 panel & backlight, and couldn't really find much, including what type of LED backlight it's using. It might make a difference. Failing any details on the backlight, try both LED choices in i1Profiler and see if either one offers better accuracy (either according to stats or visually with test images).

[...] But it is a shame that "color management" apparently doesn't encompass making monitors match.

I know it's semantics but I'd argue that it's not really a shame because it's simply not a pragmatic goal. Colour management aims to manage colour, simply that. Managing colour means you know what it's doing and why it's doing it, and it works consistently. It doesn't mean any particular device looks the same as any other particular device; if they're not engineered to look like each other we can't expect colour management to make them do so. Claims to the contrary are usually just vendor marketing copy.

For most people most of the time, there's little point trying to force different devices to match when they don't match, unless you're prepared to dumb everything down to the lowest common denominator. That would be substantially more limited than even sRGB, especially when dealing with mobile device and laptop displays, the vast majority of which don't cover sRGB. Devices have different colour reproduction characteristics -- different white point, black point, tone response curves, colour temperatures and gamuts, among other things. Each one of these factors is limited potentially in a different way for each device; so to make all devices look the same you'd have to take the subset of all black points, all white points, all TRC's, all colour temps, all gamuts, etc. You wouldn't be left with much.

Some profiling software does let you attempt to force a match. This can be important for multiple workstations within a single studio or production shop, for example. But the matching is only going to happen by subsetting the displays against each other. Doing that with the entire population of uncalibrated monitors used by all your potential viewers is a pragmatic impossibility.

[...] The reason I got the PA241W was precisely to use the sRGB emulation, with the hopes that images edited on that would look at least civilized on the widest variety of uncalibrated monitors.  (Or on mini-lab printers.)  On the assumption that uncalibrated monitors would (roughly) have a Bell curve distribution.  (Nobody will accept a monitor that displays purple skin tones.  Ditto for mini-lab prints.)   I want to be in the center of the Bell curve when I'm editing images.

I've seen plenty of purple prints roll out of crappy photo labs. You've done just about the best you could in your choice and setup of the PA241W, but your results with the Surface Pro 3 show a concrete example of the limitations of the idea. It's not a Bell curve, it's going to be something else. Especially since most uncalibrated monitors of the roughly sRGB class are running far brighter than would be the case for any monitor calibrated for colour critical work. Most mobile and laptop displays in particular are among the worst of the bunch, but economy home & office desktop displays often are hardly better. There are a bunch of skew factors, to the point where I personally don't worry much about what people are seeing on monitors that by definition aren't set up for colour critical work. Unless you know what your viewers are using and can target your images for that, the next best thing you can do is what you're doing.

But if what you say means what I think it means, then the Surface isn't in the center of the Bell curve, even after calibration.  (Or maybe the PA241W isn't either?)

Actually, from looking at the some of the reviews, I expect that a calibrated Surface Pro 3 is probably better than most mobile devices, and perhaps better than many laptops as well. It's certainly much better than the Surface Pro 2 was. But being a good mobile device still doesn't mean it's going to match the NEC which is near the top in colour critical displays and should fully encompass sRGB as well as being excellent in all the other display qualities. I'm afraid your idea of "centering within a Bell curve" essentially is too simple; the distribution of low-end, uncalibrated (or even calibrated) displays is not a uniform distribution, at the factory or as installed.

Do you have any comments on the way I used the Argyll utilities to generate the gamut plots?   Are there any problems with them?

I looked at several of your VRML's and don't see any issues there. I don't use Argyll myself, but I know of it and have no reason to believe it's showing you bad info. Those plots simply confirm to me that the Surface Pro 3 is very good for a mobile device in terms of sRGB class displays. But still not as close a match for sRGB as your PA241W. There will be visible differences if you look for them, as indeed you've found.

[digitaldog]

But it is a shame that "color management" apparently doesn't encompass making monitors match.

It's not impossible but it's difficult and as Royce points out, you have to take the lesser of the two and match the better to it (we can't move the volume to 11). Depending on a lot of factors, it might still be impossible. And as to which is correct, I'll bet you dollars to donuts it's the NEC in emulation mode.

Your concept of using the best quality display emulating sRGB is a good one. Always start with the best quality you can. But what other's see will very likely not look like that, they are not necessarily calibrating to anything close to that sRGB goal (even though we tell people to use sRGB for the web). You want two displays to match without any visual differences? Setup two SpectraView's of the same model and calibrate them identically. Or two Eizo's, Artisan's, Barco, PressViews (reference display systems).

[WayneLarmon]

Your concept of using the best quality display emulating sRGB is a good one. Always start with the best quality you can.   But what other's see will very likely not look like that, they are not necessarily calibrating to anything close to that sRGB goal...

Yes, but this is the world we live in.  I oversimplified what I am trying to accomplish.  I am not only a web designer but I am also working on building a digital archive of family photos.  That will be viewed on uncalibrated monitors and (sometimes) printed from the most convenient mini-lab.  And will be viewed on devices that haven't been invented yet (with mass market 4K monitors at the head of the list.  My 200 PPI Surface gives me a taste of how images will display on 4K monitors.)

Web designers (and the stakeholders that employ web designers) wish for the same thing: they want their web sites to look the best they can on the devices that people actually use.   This is one facet of Responsive web Design.

I respectfully disagree with both you and Royce that monitors don't fall into some kind of Bell curve distribution.  Yes, we sometimes see purple faces, but they are outliers and fit in one of the the extremely skinny long tails of the distribution.

After thinking what Royce said about primaries, I wonder if I would get closer to the center of the distribution curve if I had chosen the sRGB PA version instead of the wide gamut PA241W?  If I am aiming for sRGB monitors (and minilabs) exclusively.

Now, I'll recalibrate the Surface with the LED settings that Royce suggested.

Wayne

[digitaldog]

That will be viewed on uncalibrated monitors and (sometimes) printed from the most convenient mini-lab. 

Then it's a crap shoot (anything can happen). You are producing what you feel is idealized RGB values that will be output in some fashion. Your RGB values have a meaning and work within a well structured color management workflow. As soon as you add uncalibrated monitors or uncalibrated or calibrated but not profiled devices, this falls apart and there is nothing you can do about. If someone complains, it's their fault, you're using best practices they are doing the opposite.

In a prefect world, even with color management (which is far from universal) two displays should match. But in the real world, there are a slew of issues both hardware and software that doesn't make this happen. Take color management out of the equation, all bets are off.

[digitaldog]

After thinking what Royce said about primaries, I wonder if I would get closer to the center of the distribution curve if I had chosen the sRGB PA version instead of the wide gamut PA241W?  If I am aiming for sRGB monitors (and minilabs) exclusively.

How can you aim for something that isn't defined. Even with so called 'sRGB monitors' they are all different. They all age differently. They have differing backlight technology. If all said devices really did produce sRGB as spec'd (and keep in mind, this is based on a theoretical CRT with P22 phosphors), why would anyone ever calibrate or profile them? The are sRGB. But in fact, they are not. As for the mini lab, you're aiming at specifically what output device, substrate and processing? Based on what profile?

[Royce Howland]

[...] Web designers (and the stakeholders that employ web designers) wish for the same thing: they want their web sites to look the best they can on the devices that people actually use.

Then perhaps you need to consider something that most effective testing plans do in my experience, and incorporate tests on representatives of the equipment that your viewers actually have. If your only frame of reference is a synthetic standard like sRGB that most uncalibrated displays don't match, and your own high-end calibrated display, this doesn't tell you how your images will look on those viewers' monitors. In software terms, this would be somewhat like a developer telling me "well, it compiles and runs on my dev machine" when I have a flood of user reports describing problems on their own machines.

I respectfully disagree with both you and Royce that monitors don't fall into some kind of Bell curve distribution.  Yes, we sometimes see purple faces, but they are outliers and fit in one of the the extremely skinny long tails of the distribution.

You're free to disagree of course. My counter example is the scores of displays I've evaluated or worked with; in Andrew's case it's probably thousands. I know, that sounds like the old internet argument "I'm right because I have X, Y, Z experience." Still, I wouldn't treat the entire population of uncalibrated displays as a single uniform distribution in terms of their performance because that theory doesn't match the real experience I see all the time. The reason it's not a single curve is that it's in fact multiple curves that cluster around different skew factors: LCD vs. CRT (for those who still have them), age (performance doesn't degrade on a smooth curve over time), wide gamut vs. narrow, IPS panel vs. TFT (combined with viewing angle), CCFL vs. LED backlight, size of panel (contributing to good panel uniformity or not), mobile/laptop vs. desktop, ambient lighting conditions such as dark workroom vs. bright fluorescent lit room vs. bright natural light room, matte vs. glossy, resolution including high DPI scaling vs. not, 6-bit vs. 8-bit vs. still-rare 10-bit video pipeline, Mac (ColorSync) vs. Windows (no general system colour management), user-controlled settings like brightness or even calibration, etc.

You can try to target some kind of average or median in this soup of variables if you like. All it means is that chunks of the population that cluster elsewhere will be further out from your presumed centre of the curve.

But really all of this is sort of angels dancing on the head of a pin. The simple fact is that calibrated or not, different monitors don't match. That's observable fact. How far they mismatch, and why that is, is causing you concern. Okay. If you want to produce a single image that looks good on all displays, the best you can do is to constrain it to a low common denominator. And even then, it's not going to look good on certain clusters of displays, for certain values of "good". Just look at the display performance of almost the entire collection of all smartphones prior to the past 12 - 18 months or so. They were frankly horrid, and certainly as a massive population of image-viewing devices I'd argue they did not share a smooth curve with desktop displays.

After thinking what Royce said about primaries, I wonder if I would get closer to the center of the distribution curve if I had chosen the sRGB PA version instead of the wide gamut PA241W?  If I am aiming for sRGB monitors (and minilabs) exclusively.

If you wanted to focus on sRGB-only then you certainly could just use sRGB class displays on your workstation. But I don't think that would be particularly helpful to what you're trying to do. If the wide gamut PA241W is already nailing sRGB essentially 100%, which it is, using a narrow gamut NEC PA monitor won't really change anything for you in a material way. That monitor likely will nail sRGB pretty much 100% as well. Either way you'll still have a calibrated colour-critical display, and your viewers won't. And since wide gamut monitors are increasing in availability and you stated a competing goal of future proofing your images for coming new standards, targeting sRGB only opens you to the reverse issue of how they will look on improperly configured wide gamut displays with viewers who have those.

Back to the example you started with. The profile you showed for the Surface Pro 3 is one of the best I've personally seen for any mobile device, and better than most laptop displays I've seen. But I highly doubt it would match an sRGB NEC display any better than it matches the PA241W. It's not an issue of the PA241W being a wide gamut display. It's the issue that, even when calibrated & profiled, the SP3 is unable to fully cover sRGB and perhaps has other issues as well such as less than perfect neutrality due to the LED backlight, presumed 8-bit video LUT, or both.

But this is all theory. Try some experiments. If it's a priority, find a way to get ahold of some other monitors and set up a small test bench. Run through various scenarios and see what happens. I'm not trying to be argumentative just providing some expectation management and illustrate why you observed what you did in this case, and how it might generalize out. (Or not.)

Your earlier stated desire for colour management to make different monitors match is not going to happen -- not in the general case, and not even in most specific cases. Plus few of your viewers will be calibrating & profiling their monitors anyway. So if you back away from that, what's left? Probably something like producing the best images you can on a good quality, properly configured display, and then testing how they look on a variety of representative viewer displays if you truly care to make that effort.

I'll be curious to see whether setting i1Profiler to one of the LED modes improves the Surface Pro 3 a bit...

[WayneLarmon]

About distribution curve shapes, I'm not sure what the point is that you and Andrew are making.  I'm only describing the same process that all the web developers use: attempt to make the content look as good as it can on the widest variety of devices possible.   The fact that perfection is impossible is immaterial--we just strive for the best that we can achieve.

The profile you showed for the Surface Pro 3 is one of the best I've personally seen for any mobile device, and better than most laptop displays I've seen.

Yes.  After co-inventing sRGB (along with HP) all those years ago, Microsoft is finally dropping the other shoe.  The Surface (even uncalibrated) is closer to sRGB than many (if not most) other consumer screens.   I've measured all the monitors I could get my hands on and I know how much they deviate from sRGB.  Hopefully, other manufacturers will get the hint and (consumer) monitors in the future will be closer to sRGB.  (To tide us over until Rec. 2020 takes over.)

Pragmatically, I have noticed that images that I edit on my PA241W do look better on a wider assortment of uncalibrated monitors then images did when I was using a lesser monitor (that was calibrated.) 

Also pragmatically, I can use RGB curves in Photoshop (on the Surface) to make images on the Surface more closely resemble images on the PA241W.  I was thinking of seeing if I can converge of a single set of RGB curve adjustments that I could save and use as a crude kind of soft proofing.  If I am editing images in the field, I could do my edits and then apply the curve preset as a kind of soft proof.  If the image looked like I wanted it to, then I'd undo the curve preset and save the image.   If for some reason I didn't want to wait until I got back to my PA241W.

I'll be curious to see whether setting i1Profiler to one of the LED modes improves the Surface Pro 3 a bit...

It did,  "White LED" made it somewhat better (less color deviation and less posterization in the shadow portion of a gray gradient.)  The "RGB LED" setting made it worse and I only looked at that for a few minutes before recalibrating with "White LED". 

For the record, these are two of the images that concerned me:




For the first image, on the PA241 the wood and sand had a more robust reddish tone on the PA241W.  The Surface has a greenish tinge on the wood and sand that is much less on the PA241.   For the second image, the blue on the building has a greenish tinge that doesn't appear on the PA241.  (Or in reality--I've been there many times over the years.)  I elaborate more on my updated Surface Calibration page.  After calibrating with the "White LED" setting, these effects are less pronounced. 

Which was the point of me starting this thread.   Thanks to both you and Andrew.

Wayne 

[Royce Howland]

[...] After calibrating with the "White LED" setting, these effects are less pronounced.  

Which was the point of me starting this thread.   Thanks to both you and Andrew.

Starting with a positive note. That's a good result, and I'm glad it helped. Based on that I'd say it's likely the Surface Pro 3 has a white LED backlight, not a GB-R one.

About distribution curve shapes, I'm not sure what the point is that you and Andrew are making.  I'm only describing the same process that all the web developers use: attempt to make the content look as good as it can on the widest variety of devices possible.   The fact that perfection is impossible is immaterial--we just strive for the best that we can achieve.

As I stated last post, it's sort of an angels dancing on the head of a pin kind of discussion, so it may not be too productive to continue this line of the conversation. If you feel there's a single sRGB-like performance curve with a single median you can target to make images look their best on the widest array of devices, by all means pursue it. I don't believe that's the case. I think it's a multi-variable cluster situation and if you optimize for one or a few factors you have to sacrifice others, and not just obvious "purple face" outliers. In other words I believe it would be some flavour of this:
http://en.wikipedia.org/wiki/Cluster_analysis
and not this:
http://en.wikipedia.org/wiki/Normal_distribution

Andrew's point and part of mine too is that if most of your viewers have uncalibrated displays that don't match sRGB to varying degrees, it's kind of pot luck anyway. I don't worry about it myself, but my advice if you do want to worry about it was to actually test how images look on a sampling of real devices. Rather than trying to guess based on sRGB (a synthetic standard not closely matched by a large number of sRGB-class devices), or your calibrated high-end monitor.

The only reason the validity of your idea matters is if you're going to set something in motion because of it, and care about the outcome. What I do to make something better comes directly from my theories about why & how things work. If my theories are wrong or even just suboptimal, I probably need to adjust them so I don't waste time counter-productively chasing tactics that can't give me the results I want. That's my key point on this part of the subject.

[...] Hopefully, other manufacturers will get the hint and (consumer) monitors in the future will be closer to sRGB.

I would say the panel manufacturers don't really care that much because 99% of their target market also doesn't care that much. Bang-on accurate sRGB performance will not sell extra monitors or allow a higher price for them to the vast majority of the market. Personally, I think it's more likely that if/when colour reproduction performance changes substantially it will be driven by the increased adoption of near-Adobe RGB gamut panels, blowing past sRGB and rendering conformance to it a moot point. I know other people have different priorities, but I am past the point of worrying that much about a narrow gamut standard predicated on 1950's technology like cathode ray tube displays.

Pragmatically, I have noticed that images that I edit on my PA241W do look better on a wider assortment of uncalibrated monitors then images did when I was using a lesser monitor (that was calibrated.)

Yes, this is why I use NEC monitors myself and recommend them. They do take a premium price but the money is buying you something. That's not always true, but it is in this case, for my money at least. A better quality display lets you produce a better quality image at source; you can't view this as a bad thing for any downstream viewer, really. And that's why I don't think you should look at swapping out the PA241W as your editing monitor. It's not a problem, and switching to something else won't fix what the issue is.

Also pragmatically, I can use RGB curves in Photoshop (on the Surface) to make images on the Surface more closely resemble images on the PA241W.  I was thinking of seeing if I can converge of a single set of RGB curve adjustments that I could save and use as a crude kind of soft proofing.  If I am editing images in the field, I could do my edits and then apply the curve preset as a kind of soft proof.  If the image looked like I wanted it to, then I'd undo the curve preset and save the image.   If for some reason I didn't want to wait until I got back to my PA241W.

That's also a possibility. These kinds of closed-loop expedients are always options. When I'm dealing with devices of my own, colour management will baseline them to be self-consistent over time and congruent to some other defined reference point. Using methods like you describe, I can then more reliably & consistently account for any remaining differences triggered by technology. In fact I do precisely that for printing. I calibrate & profile my monitors and my Epson printers, but of course they don't precisely match even so. I have standard curves that I apply to most print master files to account for some of the differences. Of course this may not (likely does not) generalize out to any other systems besides my own. But it does at least let me have a more productive workflow on my own gear, and that's a legitimate goal.

Some of the colour differences you describe in the two test images make sense to me purely based on the difference between CCFL and white LED backlighting. Similar effects are observable even if you put a CCFL-based NEC PA241W next to a GB-R LED-based PA242W.

[WayneLarmon]

Andrew's point and part of mine too is that if most of your viewers have uncalibrated displays that don't match sRGB to varying degrees, it's kind of pot luck anyway. I don't worry about it myself, but my advice if you do want to worry about it was to actually test how images look on a sampling of real devices. Rather than trying to guess based on sRGB (a synthetic standard not closely matched by a large number of sRGB-class devices), or your calibrated high-end monitor.

I'd already addressed this when I said

Pragmatically, I have noticed that images that I edit on my PA241W do look better on a wider assortment of uncalibrated monitors then images did when I was using a lesser monitor (that was calibrated.)

My theory (several years ago) was that, if I edit on a high end monitor, such as the PA 241W, then my images should look at least decent on more uncalibrated monitors than they did when I edited them on a $350 HP monitor.  This has been borne out in practice.

I know that a more precise way of preparing content that will be viewed on a vast sea of unknown monitors is to have a larger collection of monitors to test on, with the sample set biased towards the target demographic.   This is what the large software companies do.  I am not a large software company, so I make do with what I have.   I am not planning on doing anything new; all I am trying to do is see if it is possible to calibrate my Surface so that it looks more like my PA241W.  Based on what I have found during the course of this thread, the answer is "maybe." I need to do more pragmatic testing.

Personally, I think it's more likely that if/when colour reproduction performance changes substantially it will be driven by the increased adoption of near-Adobe RGB gamut panels, blowing past sRGB and rendering conformance to it a moot point. I know other people have different priorities, but I am past the point of worrying that much about a narrow gamut standard predicated on 1950's technology like cathode ray tube displays.

You don't believe that Rec. 2020 will be the next standard?   Adobe RGB is based on an even older technology: CMYK printing, which predates CRTs by several decades (at least.)



Modern inkjet printers exceed Adobe RGB by a wide margin

Red River UltraProSatin on an Epson R3000 (VRML).  Heck, my wife's old HP 7300 OfficeJet (VRML) stuck out of Adobe RGB a little bit.

I don't print much so I don't have a large collection of inkjet printer gamuts.  But the Red River gamut plot all by itself make me yearn for a 4K Rec.2020 monitor, in case I ever decide to get a printer of my own.   (But not right now--see next paragraph.)

But I'm skeptical that the mass market will adopt wide gamut monitors at all, based on the glacial pace of addressing color management at all.  The Windows desktop isn't color managed, which makes for radioactive icons when using a wide gamut monitor natively.  Right now the computer industry is pushing high PPI counts (24-27 inch 4K monitors are shipping now) and the abysmal adoption of font scaling (cough, Adobe, cough) is painful enough.   (CC Photoshop menu items are tiny on my 200 PPI Surface.  Good thing that I had my eyes refurbished and calibrated several years ago (cataract surgery.))

Anyway, thanks again for your advice.

Wayne

[digitaldog]

I know that a more precise way of preparing content that will be viewed on a vast sea of unknown monitors is to have a larger collection of monitors to test on, with the sample set biased towards the target demographic.

Whatever, your aiming for a very large side of a barn. That's the bottom line.

But I'm skeptical that the mass market will adopt wide gamut monitors at all, based on the glacial pace of addressing color management at all.

Has no more to do with color management than the current sRGB system. IF the vast majority of displays had that wide gamut side of the barn Adobe RGB (1998) like behavior, instead of sRGB, we'd be in the same boat as we are now. Non color managed app's would have to assume the data is Adobe RGB and get it half right, like non color managed app's assume sRGB and are half right today. But we'd have the  benefit of a wider gamut side of the barn to work with. I think we will see that in coming years (sRGB like behavior goes away, wide gamut, Adobe RGB like behavior replacing it). Then we'll be in are real mess while some people are stuck using old sRGB like behavior while others are using the newer Adobe RGB wide gamut like behavior.

The Windows desktop isn't color managed, which makes for radioactive icons when using a wide gamut monitor natively.

Again, because the side of the barn assumption is something akin to sRGB not that those display are exactly producing sRGB as I said. Adobe RGB and sRGB like display's are two large barns!

Just because there's a very well defined recipe for sRGB and we have a specified RGB working space based on it, what our displays produced, even sRGB-like displays are all different. That's why they have to be calibrated and regularly, that's why the world sees your sRGB numbers differently. Doesn't matter how many you gather up and view, you're still only going to have one set of RGB numbers which fall somewhere on a very large side of a barn.

[WayneLarmon]

Just because there's a very well defined recipe for sRGB and we have a specified RGB working space based on it, what our displays produced, even sRGB-like displays are all different. That's why they have to be calibrated and regularly, that's why the world sees your sRGB numbers differently. Doesn't matter how many you gather up and view, you're still only going to have one set of RGB numbers which fall somewhere on a very large side of a barn.

Win, lose, or draw, web developers need to put images on the web.   If you were writing a new book, say,  Color Management for Responsive Web Design, addressed to web developers, what would you suggest for best practices? 

Second question.  I feel frustrated with iProfiler.  My old Monaco Optix/DTP-94 calibrator (that isn't supported on anything newer than 32 bit Windows XP) had what amounted to Photoshop's RGB curve adjustment for optionally adjusting the monitor profiles that the Monaco software generated.  iProfilier doesn't have any such facility.  Is there any way to edit an ICC profile generated by iProfilier like I used to be able to do with the Monaco software?  I can deal with curves.  I haven't had good luck with moving points around a color wheel (like how the editor in Adobe DNG Profile Editor works.)  But I suppose moving points around a color wheel would be a good addition to a Curves type adjustment.  (Just like how I sometimes need to move hues around in PS's Hue/Saturation tool, after I've done all I can do with RGB curves.)

Wayne

[digitaldog]

Win, lose, or draw, web developers need to put images on the web.   If you were writing a new book, say,  Color Management for Responsive Web Design, addressed to web developers, what would you suggest for best practices?

Exactly what you did getting the NEC and viewing in sRGB emulation. That's something I can also see identically to you because of color management. Anyone else who doesn't follow that will see something different. Make the image look good to you, rendered for that sRGB emulation and move on. If the other party cares about how the image should appear, they will calibrate and profile their display.

Forget profile editing with very rare exceptions (edit white point). It's pointless.

[Henrydavis]

Have you tried to uninstall the original Intel Driver with latest Intel Graphic Driver instead?

I found that grayscale calibration performance on the latest Intel Driver is much better than the default driver. My guess would be that loss of grayscale performance exchange for saving battery power.

Also I encountered numerous issues with Basiccolor Display which requires you to adjust its through windows firewall under the following "How to allow an app through the Windows® 8 Firewall".