Black crush on 8740w with Firepro/Dreamcolor display

I've had my 8740w for a little while now, and I love it, but one little issue keeps nagging me: The Dreamcolor display severely crushes blacks, and it does so far worse than any of the crappy 6-bit TN panels I own. The problem could possibly be the Firepro on my build, but I'm pretty sure it's the display. (I recall reading a review of a desktop Dreamcolor monitor where the reviewer briefly mentioned an issue like this.)

I've read about uneven gradient ramps in the owners' lounges, but this specific issue is a little bit more than typical 8-bit imperfections. For those of you with Dreamcolor displays, check out the gradient ramp on this page. Save it, open it up in an image editor, and zoom in. Using a color picker, you will notice that all 256 shades of gray are represented in 8-bit color, with three columns for each.

Zoom in and look at the gradient from right to left. As you approach black, you will notice that the monitor displays the exact same brightness for increasingly wide ranges of values. I've verified that the same issue exists in both Windows 7 and Linux, but I'll give an example from Ubuntu 10.10 with proprietary ATI drivers and default brightness/contrast/gamma in the drivers (since it's convenient for me ATM):

• RGB 0-10 is pure black, with no differentiation
• RGB 11-14 is the next step up, which is significantly lighter.
• RGB 15-18 is the next step up, which is significantly lighter again.
• After that, brightness gets more fine-grained, until each consecutive RGB value is slightly brighter than the last.

Adjusting brightness/contrast/gamma in the graphics options can redistribute the brightness levels across RGB values, but there will always be a sizable brightness jump between the darkest, second darkest, and third darkest shades, and they still attempt to cover a large range of RGB values. It's almost as if the display itself is only capable of displaying a very small number of dark colors (which I hope is untrue). The only way to get decent detail in the darks is to increase driver brightness and decrease driver contrast until the image is washed out. Reducing gamma helps correct for that, but then much of the detail gain in darks is lost.

I understand that you can't expect a perfect gradient ramp on an 8-bit display, or a 10-bit display using 8-bit mode/images. There will generally be some duplicate brightness values across different RGB values, and that's perfectly normal. However, even my worst 6-bit TN display has a much better distribution of brightness levels near black, leading me to believe the Dreamcolor has a flaw somewhere in its implementation. Perhaps the lookup tables are messed up, or dithering isn't being performed when it really should be.

So, are the rest of you guys experiencing the same issue? If we're all affected, as I suspect we are, we might be able to ask HP about a possible firmware fix.

 

I'm not really up on this display stuff, so forgive me if I'm asking basic questions, but are you using the HP MDA (Mobile Display Assistant) at all? I notice that when I use the MDA to select, for example, the Full/Native colorspace, just about everything goes much darker than when I select sRGB (which is what I believe you should choose for non-color aware programs, correct?). I wonder if this might be the cause of part of what you are seeing? For what it's worth, I'm running in 8-bit, not 10-bit.

 

Selecting the sRGB colorspace in MDA in Windows will help mask the problem, but it will still exist. I'm not exactly sure what mechanism MDA uses to change colorspaces, but it seems pretty equivalent to adjusting graphics driver brightness, contrast, gamma, saturation, and possibly hue. At least in the darker part of a brightness gradient, the same shades are just getting shuffled around to different RGB values.

I just checked sRGB in Windows a few moments ago. From short-term memory (I could be a little off), it makes only the first couple RGB values pure black, and the second-dimmest shade covers about 4 RGB values or so. From there, things start getting finer-grained, and the third dimmest shade only covers 2 or 3 RGB values. That's a significant improvement in terms of stepping, but it's still worse than any of my 6-bit TN panels, and that's pretty sad for the best case scenario on such an expensive display. Besides, there's no reason native mode shouldn't have reasonably fine-grained stepping in dark shades too.

The fundamental problem also remains: The brightness gap between pure black, the next dimmest shade, and the next dimmest shade is very large. That is to say, the display itself seems very limited in the number of dark shades it is capable of displaying. Every other [crappy] monitor I've used can make a much larger number of gradual steps from pure black to dark grays.

 

First a caveat. I'm currently on my 8540w, the 8740w is turned off, but since they're both dreamcolor with ATI cards, I tried your test. Note the machine is in 8-bit FULL NATIVE MDA mode with luminance @ 210n and I'm viewing the gradiant @ 500% in Photoshop CS5 (RGB color - 8 bits per channel) in a very dimly lit room. My results for the first 4 levels of significant differentiation are:

• RGB 1-2 are the deepest black in the gradient, with no differentiation.
• RGB 3-4 are the next step up, which is slightly lighter.
• RGB 5-7 is the next step up, which is slightly lighter again.
• And RGB 8-9 is the next step up, and again slightly lighter.
• And so on, in increments of 2, 3 or 4 RGB steps.

The key word here is "slightly", but noticable changes. And while "better" than your results, increments of up to 4 steps are still rather unacceptable at this level of hardware. You'll find I can be demanding.

From a practical standpoint though, we have a number of nighscape images in our portfolio (one is my most well known image and is on the cover of my book) with moonlit billowing clouds or moonlit fog over stands of trees that display very nice gradations on both the 8740w and 8540w, configured as mentioned above. If you have the time, I suggest you try these settings and see if you get comparable results.

EDIT: I see you were able to get "better" results in RGB mode. Haven't tried that yet, but may find time Thursday and will let you know.

 

A pair of final questions, are you using a color managed/aware image editor, and have you calibrated your monitor? From what I've heard, a non-color managed/aware image editor wouldn't work properly with the full/native colorspace, and, well, monitor calibration might solve the problem entirely.

 

A little later then Thursday (tis the season), but as promised, an update. I retried the test in full across the board sRGB mode. Results came up exactly the same as in NATIVE. Actually I was hoping so, to give me comfort in judging for a consistant feel for output in different media.

Still, would like to be doing tests in 10 bit. I know it takes time for HP/AMD to get it right, but these machines are now about 6 months old!!!!!

 

Hi guys, yes, I notice the same. When I first tried it, I would swear I got the same results as CantankerousBlowhard. I don't know if my eyes adjusted after working with the images or were able to tune out the bright white, but now I get VisionLight's resutls above. Also on CS5, 500%, etc.

I am also on a DC2 display on an 8540W with the ATI FirePro 5800, and I tried it in full native and sRGB with the same results. I am in an extremely dark room so at 210 or 75 nits, I was able to notice the same results.

After a couple of days of trying to get my old calibrator to work with Win 7 x64, I have given up, so I am not able to offer results after calibration (till I bite the bullet and get the one to be paired with this display.

If there is something you want me to try, though, I'll certainly try my best. It would be interesting to see whether this happens on the nVidias too.

Cheers.