Bits Understanding Colour Depth

[Tim S]

http://www.bit-tech.net/bits/2007/11...colour_depth/1

16b, 24b, 32b...10b, 8b....the list goes on and on. Don't know your CMYK from your RGB? How about 8bpc? It's time to get down and dirty with one of the biggest misunderstandings in graphics - and shed some light on colour depth.

[Zurechial]

Great article, guys - Very informative and interesting.

The defining difference for me between 16-bit and 32-bit colour in games was noticed whilst playing Deus Ex on a CRT a few years ago.
In 16-bit, any scene with shadows showed the shadows as having an overlaid pattern not unlike a 'Houndtooth' pattern which was more noticable in lower resoltuions.
Upon switching to 32-bit, that pattern disappeared.

I'm not quite sure why, but this undesirable effect of 16-bit was always most noticable to me in games based on the older iterations of Unreal engine (ie; Unreal, Na Pali, Deus Ex, UT99), presumably because of how the engine renders or handles colour depth.

[theprodigalrebel]

Another great article from Brett. I like articles that I have to read twice so that my feeble mind can process all the information.

[dgb]

Good article - thankyou.

[Zeali]

I had to register that i could say this. These are the kind of articles what makes bit-tech so good. I love reading these so keep up the good work.

[Paradigm Shifter]

Hehe. It's good to see that it isn't just me that hates the 'millisecond myth'. (Sometimes it seems that way...) And 6-bit-with-dithering as opposed to 8-bit colour.

Great article.

[Flibblebot]

Originally Posted by Paradigm Shifter

Hehe. It's good to see that it isn't just me that hates the 'millisecond myth'. (Sometimes it seems that way...) And 6-bit-with-dithering as opposed to 8-bit colour.

Great article.

It's the same with the whole "I can get 500,000fps from my graphics card" thang. I always find that anything above 60fps annoys me because of the shearing, so I set most games to turn the "disable vsync" setting off. I'd rather a game that looks amazing at 60fps than mediocre at 500fps.

Anyway, nice article. Strangely, I've just been on a training course that used exactly the same graphic for the LCD screen...

[Da Dego]

Originally Posted by Flibblebot

Anyway, nice article. Strangely, I've just been on a training course that used exactly the same graphic for the LCD screen...

Hehe...yeah, it's creative common license, and it was WAY better than I could create with my meager paint and photoshop skills.

Glad you guys enjoyed the article so far!

If you liked it, please digg to share:
http://digg.com/hardware/Understanding_Colour_Depth

Thanks for reading!

<Original Edit brought into article conclusion>

[Ramble]

*Hugs 8-bit display*

AFAIK the 'compound' in eyes (which I believe is a protein) doesn't break apart, what it does is introduce a potential difference across the membrane which then activesa nerve.

Got an artice on how to calibrate your monitor coming up? i've always wanted to do it properly but I'm too lazy and stupid to figure it out myself.

[aggies11]

The whole 60fps cap so anything under 16ms won't be noticed. Is that really true?

I mean, if your response time is OVER that length, then the pixel won't have changed before the next frame is displayed, so that obviously will have problems.

But even if it's less (lets say 8ms vs 1ms) there should be a visible difference, no? Maybe not if all colour transitions take the exact same amount of time. But if there is some variance in the times (eg. red to white, takes long than white to red etc), then that will affect the displays ability to output a uniform image. (ie. Some parts of the new frame will display faster/before others), which we see as ghosting/blurring. Presumably then the lower response time of the display, the lower the potential amount of variance between transitions, which means the less time "ghosting" of the previous image is displayed, which results in a lower level of perceived ghosting/blur.

Does that make any sense, or am I totally missing something on this one?

Aggies

[Veles]

Fantastic article, very informative

[Da Dego]

Originally Posted by aggies11

The whole 60fps cap so anything under 16ms won't be noticed. Is that really true?

I mean, if your response time is OVER that length, then the pixel won't have changed before the next frame is displayed, so that obviously will have problems.

But even if it's less (lets say 8ms vs 1ms) there should be a visible difference, no? Maybe not if all colour transitions take the exact same amount of time. But if there is some variance in the times (eg. red to white, takes long than white to red etc), then that will affect the displays ability to output a uniform image. (ie. Some parts of the new frame will display faster/before others), which we see as ghosting/blurring. Presumably then the lower response time of the display, the lower the potential amount of variance between transitions, which means the less time "ghosting" of the previous image is displayed, which results in a lower level of perceived ghosting/blur.

Does that make any sense, or am I totally missing something on this one?

Aggies

You would *think* you should see a difference...but that's the hype talking. 8ms or 1ms, it's still beyond the realm of perception. See, you only NEED to change 60 times per second. As long as you do so within that time, it's "in sync" and is not able to be discerned.

The important thing you have to understand is what ghosting really IS - it's the CUMULATIVE existence of the previous frame interspersing with the existence of the next frame. Think of it like this - you're moving 60FPS, but if your response time is slower than 60FPS, pixels that have a large transition will only run at 40 or 50fps. THAT's when you see ghosting - certain pixels from the old frame creeping into the new frame.

As long as you're below 16.667ms, the frame will shift more or less uniformly - it can draw faster than the frame needs to be rendered. Your eyes can only see about 30FPS - we see ghosting over a cumulative series of frames where parts of those frames are unable to move quickly enough. It's the overall anomaly of part of one frame clearly inside part of another frame, and then that continuing over multiple frames, pulling the whole load out of sync. However, as long as you're able to draw faster than the frames need rendered, then those anomalies aren't present.

On a side note, lower response times mean nothing for that - as I mentioned, a lot of responses are measured in "gray to gray"...which is a fairly useless number. You don' know which grays you're shifting from or to, or how it's being done. It's pretty much an arbitrary point, which is why companies like to use it...it can be manipulated.

[C-Sniper]

Thanks for another great article!

[legoman666]

Originally Posted by Da Dego

You would *think* you should see a difference...but that's the hype talking. 8ms or 1ms, it's still beyond the realm of perception. See, you only NEED to change 60 times per second. As long as you do so within that time, it's "in sync" and is not able to be discerned.

The important thing you have to understand is what ghosting really IS - it's the CUMULATIVE existence of the previous frame interspersing with the existence of the next frame. Think of it like this - you're moving 60FPS, but if your response time is slower than 60FPS, pixels that have a large transition will only run at 40 or 50fps. THAT's when you see ghosting - certain pixels from the old frame creeping into the new frame.

As long as you're below 16.667ms, the frame will shift more or less uniformly - it can draw faster than the frame needs to be rendered. Your eyes can only see about 30FPS - we see ghosting over a cumulative series of frames where parts of those frames are unable to move quickly enough. It's the overall anomaly of part of one frame clearly inside part of another frame, and then that continuing over multiple frames, pulling the whole load out of sync. However, as long as you're able to draw faster than the frames need rendered, then those anomalies aren't present.

I disagree. While having <= 16ms response time will get rid of ghosting, you will still see motion blur. Response time (to me) is defined as how long it takes the display to change from 1 frame to the next. Regardless of how many FPS you have, each pixel still takes X amount of time to change from 1 color to another. The color doens't instantaneously change, it fades. For example, I have a 2ms viewsonic panel. When I move the mouse around against a white background, I can see blurriness. However, if I move my mouse to my 2nd monitor (an old 15" POS CRT), I zero blur when I move the mouse around. This is due to the difference in response time between the 2. Thus, even though my 2ms Viewsonic display is <16ms (1/60hz), I still see motion blur. Now, if the display was say... 0.2ms, there would be considerably less motion blur.

So you can be happy with your 16ms display and "wait" 16ms for the colors to fade in each frame and everything will appear blurry. I'll sit here with my 2ms display and see nice sharp images (comparatively anyway, if you want it as good as it gets, buy a nice CRT).

Also, you claim that the human eye can only see 30fps. Bullshit sir. Go to your 60" widescreen TV. Watch a big panning scene in LOTR and notice the "jumpiness." Or go to the movie theatre and watch another panning scene. You WILL see jumpiness becasue the image in each frame is something like 2ft from where it was in the previous frame. Now, go to the racetrack and watch a car go by at 200mph. You will see no space inbetween each "frame" because the human eye is an analog device. It simply cannot be measured.

[EQC]

Legoman666: I think it's difficult to compare "response time" on a CRT vs. an LCD because they work differently.

In an LCD, your backlight is always on. In a CRT, if you had a very high speed camera, you'd see the picture being dark most of the time with a brief flash of light for each frame. You can do this with a CRT without it appearing "dim" because the pixels are so bright anyway. Your eye perceives some sort of time-averaged brightness that still looks good.

I think they tried something similar with LCD's -- I think it might have been on the first 120Hz displays. Basically, they turned the backlight off during the extra 60 frames, and only had it turned on during the 60 real frames every second. Supposedly, flashing it on briefly 60 times per second in that way got rid of ghosting, motion blur, etc... but in the store, the display appears "dimmer" and most consumers perceive brightness as picture quality. So, the newer generation 120Hz displays just interpolate extra frames...which looks brighter, but still may suffer some motion blur, etc. for some viewers.

I'm currently trying to find the original article I'm remembering...
*Edit*
I can't seem to find the original article I read...I think it was on HDGuru, and linked by Engadget a few months to a year ago.
*end Edit*


Personally, I've got an LCD with a 16-25 ms response time (2003 Dell Ultrasharp), and I rarely notice any ghosting/blur in movies.
Even a 16ms response time for a 60-hz refresh rate confuses me -- seems like the pixels would be showing the "wrong" thing for 99% of the time, and then finally get to the actual correct color just before the frame starts changing again. I don't understand why my eye perceives the "correct" image then...but I guess it's more complex than that.

[Da Dego]

Originally Posted by legoman666

long narrative

legoman666, though I understand what you're trying to say, it's...well, simply incorrect. Most of what you stated in this post is exactly the type of marketing misconception that we're trying to debunk. There are limits to perception, full stop. You can believe you see things that take 2ms, but it's simply not the case. Further, your entire analysis of what causes motion blur is incorrect, as is your film analysis.

In fact, even the definition of response time is wrong in your post. You say "the definition of response time, to me," but it's not a to you - this is a real term with real meaning. You can't just redefine it and then argue your points based around it. Response isn't drawing the next frame in its entirety - if it was than the 2ms display you covet wouldn't be 2ms - that's a gray to gray measurement.

I really don't know where to begin, I'm sorry that this is the exact opposite of constructive disagreement, but you're just frankly saying "fact isn't fact, because I believe I can see a scientifically imperceivable difference." I wish I could put it any other way...If I have time, I'll try to come back and put this in a more constructive manner but I simply don't want other people being led astray on it.

[Ramble]

I've never noticed ghosting on any 16ms or under display. I believe 12ms is the limit of mechanical->brain human perception (i.e. the lag between switching something like a joystick and the action appearing somewhere else).
Old 25ms screens are pretty obvious in the ghosting department.

[aggies11]

Re: Da Dego

Just a quick clarification then (as I'm sure no one wants a nitty gritty explanation of how LCD's draw their images via pixels). When you do the standard "draw a red/blue/yellow/etc square in mspaint, then drag it around the screen quickly" and see the faint "blur" of colour (eg. a pink hue that follows a red circle). Is that ghosting/ a result of "response time"? (The reason I mention it is that, conventionally, I am under that (possibly incorrect ) belief. And moving from my "2ms" Vx922 to my 8ms LG 22", (and comparing it to a CRT on Clone'd displays). The CRT showed nothing, the Viewsonic an almost imperceptible amount, and the LG a noticable (only if you are really looking for it) "pink" blur. (Presumably the transition from red to white. As dragging a white circle on a red background causing the leading edge to ghost-pink).

Aggies

[stephen2002]

That was really an excellent article! I'm still using an 4-year-old ViewSonic LCD with 25ms response time. I can't find the color depth anywhere in the specs though.

[Woodstock]

hmmmm, some interesting articles coming of late, keep it up... is there a list anywhere of all 8bit panels? and is it true that as a monitor ages the refresh rate decreases?