VG248QE Color Accuracy with G-Sync - Pro Calibration

[MonarchX]

Helo to all Blur Busters!

I am a VG248QE owner but also an experienced calibrator and in the process of launching a website to start calibrating displays professionally. I use i1Pro (certified) spectrometer to profile my i1Display Pro colorimeter. I am really interested to know if someone out there has similar equipment and had taken any CalMAN RGB/5, HCFR, or whichever program measurements of the regular LightBoost and ULBM G-Sync modes? What about the original standard and the new G-sync modes?

I keep hearing that G-Sync considerably improves color accuracy of VG248QE, which is something I would love to put to the test as an aspiring calibration pro, interested in such subjects. Some say that G-Sync only makes the ULBM mode use the same colors as the original or new G-Sync (non-LightBoost) mode. I am not sure what to believe. I am finding calibrating of this monitor without G-Sync in both modes quite problematic. I apologize in advance if I sound like I'm trying to educate an already knowledgeable community.

Some issues I had:

1. You absolutely have to have an accurate spectrometer in addition to a colorimeter to make accurate measurements because colorimeters use manufacturer reference display type tables, but these reference display types may differ from your own display. So, its kind of hard to judge results other than those obtained with the right equipment...

2. No matter whether I profiled or whether I used CalMAN 5/RGB, i1Profiler (terrible...), or even the new king in town - ArgyllCMS, I kept getting bad WB accuracy for the darker/black region in LightBoost mode. The squares in this test were never neutral and always had either a red/brown, green, or purple tint - http://www.lagom.nl/lcd-test/black.php . TVs have manual WB controls, so I can always correct dark grays and blacks and make them neutral, but it is not possible with software LUTs/ICC profiles. I think it happened because of 2 things - LUT interpolation and bad uniformity. Thing is... non-calibrated blacks/dark grays even in non-G-sync LightBoost mode were somewhat neutral, so that meant uniformity was not the cause, leaving only one cause - LUT interpolation. LUT interpolation simply means that software did not calibrate each and every level of the 256 grayscale levels - it simply guessed some of them (incorrectly). Not even ArgyllCMS, which took more than hundred measurements, could provide neutral grays and blacks. Some were always reddish or greenish, etc. I never had a single TV like that...

3. I know that using any 8bit LUT will result in some grayscale gradient banding, but VG248QE calibration specifically results in purplish bands on the left upper and right lower areas of this ramp - http://www.lagom.nl/lcd-test/gradient.php - in the 0-7.5% of the blackest/darkest areas.

I am just wondering if anyone else had similar issues with native VG248QE and had them solved by installing a G-Sync module? Did it allow you to have accurate/neutral blacks AFTER calibration in that Lagom.nl test link I provided? What about grayscale ramp gradient - any non-neutral bands after calibration? What about the colorspace/color gamut? I would so appreciate if someone with i1Display Pro or ColorMunki Display colorimeter could run a default, non-calibrated grayscale measurements of:
1. Grayscale in 10% steps
2. RGBYCM 25, 50, 75, 100% saturation sweeps
using the latest and 100% free HCFR 3.1.0.7 that can be downloaded here - http://sourceforge.net/projects/hcfr/fi ... s/3.1.0.7/

I wanted to conduct a professional review regarding G-Sync improving ASUS VG248QE color accuracy myself, something I think nVidia and ASUS would be interested in having as it would most likely promote their G-Sync/monitor product even further. I asked if they could supply me a module or a monitor equipped with one, so I could take all the needed measurements and return all that they sent me, but that didn't happen - nVidia and ASUS could care less for video fidelity... I can only hope someone has some decent equipment and can take those measurements for me and provide the results...

[Chief Blur Buster]

-- The non-strobed color seemed to be mostly unchanged.
-- However, strobed color was much better (LightBoost on unmodified VG248QE versus ULMB on upgraded VG248QE) though lower contrast. The lower contrast is likely caused by the faster LCD scanout (done to create longer pauses between refreshes for GtG settlement before strobing the backlight).

With ULMB, you are able to do hardware adjustments in strobe mode (adjusting via monitor adjustments), unlike with LightBoost mode which forces you to use only software adjustments.

There's a user TSM, who used an i1 Pro and an i1 Match 3 to do adjustments:
http://forums.blurbusters.com/viewtopic ... t=10#p3241
He has attached an .icc file.

I'll tell him to look at this thread, and comment!
In addition, I also have an i1 Display Pro too now, so I'll see if I have some time in the near future (i.e. within a week) to attempt this.

[TSM]

One problem you are butting your head against is that your monitor is only using 6 bits per color. This means that nearly every part of your screen is dithered at all times. Due to the lack of granularity, you are going to especially have trouble with this when adjusting your gamma. The small number of steps available at the bottom of the luminance range are going to interact with the dithering inherent with the monitor. This is most likely what you are experiencing.

A quick breakdown of 6bit vs 8bit that I found:
http://www.tftcentral.co.uk/articles/co ... t_8bit.htm

The upgrade kit is also 6 bit, so in that sense it will still have a similar problem. With a 6 bit monitor you just aim for the best compromise as you will never reach a perfect result.

[MonarchX]

I numbered specific questions I would love to hear answers to and would appreciate your contribution

Yes, it is true. If you take in consideration TN uniformity and vertical gamma shift - you cannot consider such a display accurate regardless of the color gamut accuracy.

1a. How in the world do you take dithering into consideration??? The native white balance and gamma do result in more or less stable neutral blacks or maybe ones with a tint, but a tint that is applied to all those levels - there is no rainbow effect going on. I also tried to manually adjust LUT data and I was somewhat successful at making a few levels neutral, but its a hell of a tedious process to do it for all the problematic levels and it was screwing up the gamma curve considerably. 1b. How can it be that dithering creates a rainbow effect on the darker grays and blacks, but NOT on the mid-high level grays? My colorimeter can read very low light levels accurately and VG248QE has very high black levels, so its not the hardware that is the issue.

Even though my i1Pro spectro was checked against a reference-level accuracy JETI 1211, my i1D3 profiled with my accurate i1Pro measures those manually adjusted levels and the results show too much blue for those blacks... 2. Why is that? I also think there is a possibility that this monitor is not capable of neutral grays on some important dark gray levels - they are ALWAYS purple, dithering or not, calibration or not... I don't care for color gamut accuracy so much, but I just can't stand a non-neutral grayscale. 3. Is it truly impossible to achieve due to dithering? 3.Have you been successful at eliminating a purple tint on the darker gray levels of 10-20? Its 2nd or 3d row of squares on Lagom.nl test. Past those levels my grayscale is quite neutral.

Another interesting thing - standard mode and LightBoost mode have identical color gamuts, but they look very different, even after a calibration. 4. What in the world makes LightBoost mode look so nasty? Just the grayscake (gamma, WB, luminance?).

As far as manual controls go - I saw the screenshot and its just the most basic RGB white balance gains, no separate gamma controls, RGB white balance offsets, no colorspace controls... - nada! 5. Or I wrong, are there other important controls? A software LUT must be created for this display.

6. Any luck with taking those default/stock ULBM measurements? I can take measurements of my non-modified VG248QE, but I would love to compare them to G-Sync modified, stock ULBM and G-Sync modes to see if there is a change in gamma or WB or anything at all.

An important note on gamma calibration: Only and ONLY BT.1886 gamma curve should be used for this monitor (and just about any display), but this one and those with high black levels are particular the ones that benefit the most. I could go into detail as to why, but in general its a perceptual gamma that takes display's black level into consideration and changes the curve based on that. A plasma TV with BT1886 gamma will have a completely different curve than ASUS VG248QE with the same BT.1886 gamma. They say nothing is mastered to BT.1886 standard, but the reality is that mastering takes place on reference TVs with extremely low blacks, not even VT60 is capable of producing. Thus, a 2.2 power-law gamma on a reference TV will not look the same on consumer TVs calibrated properly with 2.2 power-law curve. BT.1886 eliminates that problem. If you look at a grayscale gradient ramp with power-law 2.2 gamma, you will see how the darkest/black portion is not gradual, but kind of lumped. BT.1886 gets rid of that lump and creates a very gradual rise from black to white. Almost all color scientists and subject experts recommend BT.1886 and expect it to be the future of film mastering. Games benefit from it just as much though. Power-law 2.2 on VG248 completely crushes first 2 black levels.

[Chief Blur Buster]

4. What in the world makes LightBoost mode look so nasty?

Some of this is covered in the LightBoost FAQ.

TL/DR:
- LightBoost was originally designed for 3D stereoscopic use, rather than 2D motion blur elimination as the primary use.
- Scan-out is accelerated (~1/200sec) to create longer pauses between refreshes for pixel settlement time (GtG) before strobe flash.
- Loss of luminance is caused by the longer black periods between strobe flash cycles.
- Reddish/purplish color may be compensation for green-tinted 3D shutter glasses.
- Contrast ratio loss is likely caused by accelerated scanout

The accelerated scanout means less time for LCD pixels to refresh, which means color degradation. Going to 60Hz->144Hz causes noticeable color degradation, so a scanout going as fast as 1/200sec causes even further color degradation.

- The VG248QE (LightBoost) is simply the lower-quality LightBoost ones.
- The LightBoost on most 27" has better color (especially my ASUS VG278H (non-E suffix) which has MUCH, MUCH better LightBoost than VG248QE)
- And EIZO's equivalent of LightBoost found in EIZO FG2421, called Turbo240, doesn't have the color degradation. Still 5000:1 contrast mode in their own LightBoost equivalent.
- ULMB greatly improves quality relative to LightBoost on the same panel, though is still has a smaller contrast ratio than non-ULMB.

NOTE: I also have the ASUS VG278H as well as the EIZO FG2421 sitting here as well.

[MonarchX]

OK, that makes sense. But that should only apply to native hardware characteristics, like contrast ratio, luminance, colorspace, native gamma, white balance, etc. Calibration should fix all of that, but it doesn't. 6500K is 6500K and when its achieved using balanced R, G, and B - the reddish.purplish tint should go away, but it doesn't.

Also, how is it that ULBM prevents the purplish, brownish tints when 3D glasses still maintain the same green tint?

Also, going back to dtiehring - it can't be the cause because the resulting rainbow effect (one level reddish, then 2 levels neutral, then next level is purplish, etc) because those levels STAY that way without shifting colors completely.

Would you say that G-Sync ULBM makes VG248QE perform on the same level as the VG278H LightBoost mode? Does NON-ULBM still looks better than ULBM?

Sorry for the interrogation, I am just trying to get as much info as I can. I would hate to buy G-Sync and then learn that it will also have uncalibrate-able purple grays.

[Chief Blur Buster]

Would you say that G-Sync ULBM makes VG248QE perform on the same level as the VG278H LightBoost mode? Does NON-ULBM still looks better than ULBM?

There's huge variances in LightBoost quality between different LightBoost monitors. But quick glance for games, simple contrast-ratio-wise, VG278H LightBoost still looks better than VG248QE strobed anything (ULMB or LightBoost) because you still have the severe contrast ratio limitation to contend with. However, less banding in gradients for VG248QE ULMB. The panel in VG278H is an excellent panel, better than the one used in the VG278HE and many others, able to maintain contrast ratio better. There is no purple tint in VG248QE ULMB. Turning on/off ULMB doesn't execute nearly as noticeable color change (ULMB 100% brightness versus non-ULMB 0% brightness), except a dimming & a sudden reduction in contrast ratio.

If strobing is your priority, have you considered the Eizo FG2421? It has the known VA gamma nonuniformity issues in dark colors, but so many other attributes are much better.

[TSM]

You have to remember that a 6 bit color palette only has 64 steps. When you attempt to modify your gamma curve, you are running into a granularity problem at the lower end of the scale. For example (making up numbers) normally working with 8 bit color your 20% target may be 5,7,3 and 30% may be 12,17,9. With 6 bit color your 20% target may be 2,3,2 and 30% may be 6,7,5. Trying to hit D65 exactly at the low end is going to be literally impossible. Then you add in the dithering. Dithering doesn't magically give you more dynamic range. What it attempts to do it so add more granularity through averaging. You may get a 2,2,1 / 3,4,4 checkerboard pattern instead of 2,3,2 and so on. Then when you take this logically a step further, you realize that a solid checkerboard field is the best case scenario. When you look at ramps for instance, the monitor does it's best to dither adjacent pixels of slightly differing luminances together as best it can. This is most likely the culprit for the tinting you are seeing. When you are attempting to micromanage the lower end with a LUT, you are for instance making 20% white slightly blue, 25% slightly red, 30% slightly green, etc. When the monitor is then mixing all these together with dithering, you are probably getting a tint from the interaction.

There are only 64 steps per color at most, and that the act of trying to change the white point lowers this even further for at least 2 of the primaries. At best you will get an approximation of what you want. This is where your professional skills will come into play. You will have to determine what the most acceptable compromise is. The meter does it's best to approximate human vision, but in the end you have to determine what looks most natural to the human eye.

Take what I say with a huge grain of salt as I'm not a professional calibrator. I highly recommend heading over to the avs forum display calibration area. It's one of the greatest resources online for calibrators:

http://www.avsforum.com/f/139/display-calibration

As for your questions about using ULMB, you would probably be best to direct them at either the forum I linked to or the company that you purchased your probe and software from. The strobing back light may cause the probe's results to be inaccurate. If so, this may be addressed by calibration software that takes this behavior into account.

[srsbsns]

This may seem a little overly obvious but has the op considered reducing the contrast on the VG248QE for the grayscale purple tint? I noticed for some reason it has a tendency to cause purple tinting. May or may not help in that section of your grayscale chart but I was speaking overall.

I dont think you are going to achieve a truly neutral grayscale with these monitors though. Ive never been able to with any TN. lot bit depth panel I assume.

[TSM]

Calibrating a 6 bit TN panel is a matter of figuring out what compromises you can live with. If quality image reproduction is important, TN panels are absolutely not the way to go. Ideally you'd want at least a 10 bit IPS or VA panel as even a quality 8 bit panel can leave you with banding when properly adjusted for gamma.