AVS Forum banner
Status
Not open for further replies.
1 - 16 of 16 Posts

·
Registered
Joined
·
1,836 Posts
Discussion Starter · #1 ·
How do the "bits" of the flat displays affect performance. The more bits the more colours and shades of grey right.. so the more bits the better?


The pioneer mxe is 10bit but the hd 8bit, and that has a seeable pq difference and I read in the latest homecinemachoice mag that nec are coming out with a 12 bit plasma(which is also suppose to be the brightest plasma yet) and I think it was suppose to have 4096 shades of grey.


How much do the "bits" matter, and how many bits are the common plasmas and latest(and coming soon) LCD´s? How do they affect the colour gamut?


Even though a display has many bits.. how can they say it has so so many shades of gray, because if the display has bad contrast/black levels, then it can´t have that many shades of gray can it?
 

·
Registered
Joined
·
542 Posts
The more bits the better but you do eventually get to the point of what the human eye can perceive. Color depth and resolution is what we perceive to be a good image. Sometimes having more resolution does not guarantee a better image just the same as color depth. The question you ask is a very good one. The problem is it is a very difficult question to answer exactly because there are too many variables some of which you have pointed out. Bits do not determine the brightness or contrast ratio and even that is subjective. What the manufacturers claim is not usually based on any specific standard. When 3 tube projectors were more popular the manufacturers loved to spec peak lumens which meant nothing. Eventually, an ANSI lumen spec was created and amazingly enough the truth came out. However, ANSI lumens did not determine if there are artifacts or what the gray scale is, and so on.


Unfortunately, what most consumers experience when reading specs is marketing magic whereby they pull stuff out of their you know what to make it sound exciting. Everything has to work together to achieve the objective. If just one part does not perform the same or better than the rest of the parts then the entire system is compromised. Most marketing people do not take the worst technical spec of a chip, instead they take the best. In essence, they are not telling a lie ;) they are just not telling the entire truth.


What I always tell people before they invest in any equipment is first determine what features you will actually use not what sounds cool. Second, what is practical display size for the room it is to be mounted in. Bigger is not better if you do not have the proper viewing distance. Third, look at the display if possible or ask other (AVS Forum) before you buy it.

Everybody looks at images in a different way. Just because someone thinks it looks great does not mean it will look great to you or that you can even see the difference. When you look at a display, while it is nice to be 1 inch from the display to find the imperfections what does it look like at the proper viewing distance. Bottom line - don't drive yourself nuts with the technical bulls**t just go with what looks good to you. :D
 

·
Registered
Joined
·
71 Posts
pharris, you're right on.


But as for bits, the human eye can distinguish almost 8 bits of brightness. What? A contrast ratio of 256:1?!? It's true, but here's the kicker: The eye's dark and bright points are constantly self-adjusting.


If you've got an image with very bright and very dark spots, the eye will miss details in both areas, but if an image goes mostly bright, the iris can close pretty quickly and use the eye's 8 bits to look at perhaps 3 bits of range in an image. Opening up can take quite a while if you're coming into a darkened theater on a sunny day because the iris is still closed so the whole visual dynamic range of a darkened room gets lost.


Also, all the video standards I've heard of use 8 bits (but I wouldn't be surprised if some used 10 or 12) for each color. So if a display can show 10 bits, it can remove some of the artifacts of digitizing by filtering the image to use all 10 bits to show images.


And bits goes to gray scale exactly... so 8 bits means 2^8 or 256 levels. 10 bits is 1024 levels and 12 bits is 4096 levels. My scanner says it's 30-bit, but it counts 10 each for red, green and blue.


And no matter how bad CR or brightness are, you can divide what range there is into how ever many bits you want.


As for color gamut... you've seen the roughly rounded triangle shapes for such things, right? Well imagine it not filled in, but populated by dots. If you've got more bits in your display, the shape will have more dots (double the dots for each bit), so you can show more colors, but the outline of the shape (which is the definition of gamut) will not change.


I hope that helps answer your questions,

Chris
 

·
Registered
Joined
·
542 Posts
Torginator, I think we are both going in the right direction with this but a few things need to be added.

I think what you meant was 8 bits of gradients. Brightness & contrast is not determined by the amount of bits just the DC and level and the AC peak-peak. One thing I have found that in a real world design the more bits the more noise if you are not very careful. If a signal is .707 volts for example and you have 8 bits then 2.76mV steps. If you raise it to 12 bits then that number becomes 172.6uV which is lower than most power supply ripple. Trying to find chips as well as a circuit that can properly manage those levels is very difficult and errors can be introduced more easily. This is why I recommend people see what they are going to buy. Even though the eye can perceive only a certain range it gives it a better smoothness and depth to the picture. In other words the eye starts to see it contributing in other areas although not directly related such as resolution. Same example applies to audio. Most people can only hear up to 16kHz but yet a 20kz range audio system just seems to have the better sound even though you can't hear the additional frequencies (harmonics and other stuff). The worst joke is that when most (not all for some of you) people finally have enough money to buy the system of their dreams, they are getting close to their late 50's. The human hear starts to degrade to eventually a mere 8k of range and their site has diminished as well. Where is the fairness in that! My wish list of products to the forum is medical not electrical. I wish someone would make a product that retains our hearing and keeps our eye site perfect.:)
 

·
Registered
Joined
·
3,301 Posts
I'm not sure this is related-


But can someone explain to me how grayscale can double from 1024 to 2048 when switching to DVI input? This is what Panasonic claims.


Thanks,
 

·
Premium Member
Joined
·
3,133 Posts
The higher bits (9-10+) also helps eliminate "False-Contouring" and "Banding" on the lower end IREs. More ISF (calibrator) friendly for taking measurements down there as well.


Can't help ya Brucer- I really didn't give it much thought since I wasn't going the DVI route as advertised by Panasonic.


Dave
 

·
Registered
Joined
·
1,836 Posts
Discussion Starter · #7 ·
Is NEC the only plasma to offer 12 bit processing? What about the new large LCD´s?


I just saw an old sharp lcd, and some shades of black had a terrible glow to them??
 

·
Registered
Joined
·
518 Posts
Quote:
Originally posted by keyser
How do the "bits" of the flat displays affect performance. The more bits the more colours and shades of grey right.. so the more bits the better?


How much do the "bits" matter, and how many bits are the common plasmas and latest (and coming soon) LCD´s? How do they affect the colour gamut?


Even though a display has many bits.. how can they say it has so so many shades of gray, because if the display has bad contrast/black levels, then it can´t have that many shades of gray can it?
The bit depth is VERY important.


With 24 Bits Per Pixel (Three color channels: Red+Green+Blue, each which is 8 bits), we can show 2^8 * 2^8 * 2^8 = 2^24 = 16,777,216 Million colors. You're probably thinking: "Whoa! Lots of colors! We shouldn't need any more bits."


Unfortunately, 8 Bits only allows for gradients of (2^8) = 256 colors. This is for PURE colors only! The problem is the eye can certainly detect more then 256 shades of one hue! The technical terms for these artifacts are "Color Quantization", and "Mach Banding."


Color Quantization
http://w3.impa.br/~lvelho/hp/images/quant.html


Black and White banding
http://www.cquest.utoronto.ca/psych/...ch3/mb/mb.html


Illustration of what banding is about
http://www.edcenter.sdsu.edu/slides/...her/img047.JPG


If you have Pohoshop, create a New Image 512x512 pixels. Select the foreground color as pure green (0,255,0), and the background as pure black (0,0,0). Now use the Color Gradient tool - click in the top left, and shift-click in the bottom right. (If you want to really see the banding artifacts, set your display to 16 bit before the gradient fill !)


Similiarly,

2^10 = 1024 Gradients

2^12 = 4,096 Gradients


A higher bit deph allows for more non-pure colors to have a smoother ramp to pure white.


In the PC gaming world, we've (finally!) moving to 48 BPP (16-bit integer channel * 3), 64 BPP (16-bit integer channel * 4), 96 BPP (32-bit floating-point channels *3), and 128 BPPs (32-bit floating-point channels *4)!! 8 BPP is almost useless when you are trying to do real-time additive blends, such as haze, fog, smog, and other translucent effects.


Black Levels are for the most part, independent of (color) gradients. You can "effectively" have more black steps by paying attention to gamma: That is, our eyes, and displays are non-linear. Gamma is how we convert perceived brightness to voltagei.e. If you double the voltage, the image is NOT twice as bright.

http://www.wizardnet.com/musgrave/gamma.html


Here's an quick example:


Let's pretend we're a display manufactor, that is learning about gamma.


Naively, we could just linearly map the brightness (0 to 255) onto voltage (0.0 - 1.0) i.e. When we have a pixel of pure black (with intensity 0), we could set the voltage to 0. And when we have one that is full white (intensity of 255) we set the voltage to 1.0. If have a color of 128*, we would normally set it's voltage to 0.5.


* Technically 127.5, but we're assuming an 8-bit integer brightness scale.


The problem is as we display a monochrome (gray-scale) ramp, it would look way too bright in the middle.


What to do?


Well, instead we lower the voltages the closer we get to black. i.e. our 128 intensity ends up with 0.333v instead. Similiarlly for the other colors.


i.e.
http://www.bberger.net/gamma.html


Now we have a nice monochrome ramp that has a smooth brightness!


(Note: If you were into computers a few years back, the old 3Dfx Voodoo cards were only 16-bit devices (5/5/5 for RGB), but effectively were 22-bits.)


I forgot to mention one little detail -- the gamma curve varies for devices! BY tweaking the gamma curve, we can get better blacks.


BPP doesn't effect the gamut -- at least to my knowledge. (Please correct me if I'm wrong.)


Don't confuse gamma, and gamut: The latter is the effective dynamic range of ALL possible colors a device can show, the former how quickly brightness approaches white.


Cheers
 

·
Registered
Joined
·
1,836 Posts
Discussion Starter · #9 ·
Holy cow.. I´ll need some time to sink those answers through my skull.


How many bits are the latest LCD display´s.. theyre only 8 or 10 aren´t they?
 

·
Registered
Joined
·
542 Posts
_Michaelangelo_


Damn your thorough. I usually avoid those type of answers because every manufacturer has their own opinion on what is important and have already brain washed the end user. Trying to prove theory in real world just does not work out these days. Trying to explain why the theory does not hold true to an end user is asking for punishment. I agree with your answer, what I don't agree with is the manufacturer's results. They tend to spec the chip under ideal conditions instead of the real condition it is under. Technically, a manufacturer can say 10 bits per color but do they tell you what the bit error is. There can be a 1-2 bit error or even higher depending on their circuit. Even bandwidth can be stifled with a poor layout. The semi can call out 300MHz but the effects of the connectors capacitance, vias in the board, trace pairing and widths can take 300MHz down to 50MHz if you are not careful. I would like to see a spec that shows input to output that is measured in the circuit not just ideal chip specs. This applies for everything - bits, bandwidth, etc. I stick to the same old line - "See it for yourself and don't get caught in the spec game." :)
 

·
Registered
Joined
·
518 Posts
Quote:
Originally posted by pharris
_Michaelangelo_


Damn your thorough. I usually avoid those type of answers because every manufacturer has their own opinion on what is important and have already brain washed the end user.
Well, I wanted to point out the theory to illustrate what the problems are, and the common solutions. I've done 2D and 3D work over the years, and these annoying problems keep cropping up so I thought I would share what I've learnt. It's important that people are educated, since traditional video and computer graphics are slowly merging. If you know what to look for, you can make a better choice for yourself.


Couple of caveats:


- Don't forget that I'm coming from the computer graphics world where banding and moire patterns are much more noticable on a hi-res monitor. (We've had 1600x1200 resolution images for years.)


- Since CRTs automatically make use of temporal aliasing, banding is less of an issue.


- Bits Per Pixel only really becomes a factor for still images. You will be hard pressed to notice banding on moving images.


- Moving pictures usually don't emphasize color gradients.


- Computer graphics generate motion in real-time, composited one frame at a time, hence BPP plays a bigger role.

Quote:
Trying to prove theory in real world just does not work out these days. Trying to explain why the theory does not hold true to an end user is asking for punishment. I agree with your answer, what I don't agree with is the manufacturer's results. They tend to spec the chip under ideal conditions instead of the real condition it is under. Technically, a manufacturer can say 10 bits per color but do they tell you what the bit error is. There can be a 1-2 bit error or even higher depending on their circuit. Even bandwidth can be stifled with a poor layout. The semi can call out 300MHz but the effects of the connectors capacitance, vias in the board, trace pairing and widths can take 300MHz down to 50MHz if you are not careful. I would like to see a spec that shows input to output that is measured in the circuit not just ideal chip specs. This applies for everything - bits, bandwidth, etc.
Yeap, that is what I call extra "noise", which together with temporal aliasing, blurs the images, making the specs less important.


The BIG problem is that there are few standardized tests. And it cetainly doesn't help that our eye is non-linear towards brightness. ANY lighting difference almost completely invalidates any testing. :(

Quote:


I stick to the same old line - "See it for yourself and don't get caught in the spec game." :)
I agree.


Specs are fine and good, but unless you're a professional that has been trained specifically what to look for (I'm not, but I've been "lucky" since I keep noticing these artifacts :)), you're not going to notice the extra precission. (No offense intended to amatuers, like myself.) Basically we could summrize: "Decreasing returns", and all that jazz.


Cheers
 

·
Banned
Joined
·
3,772 Posts
Things are somewhat limited by the bit depth of the source material. HD and DVD source material already has limited bit depth for chroma, so what real advantage will a "12 bit display" give over a "10 bit display"?
 

·
Registered
Joined
·
822 Posts
Quote:
Bits Per Pixel only really becomes a factor for still images. You will be hard pressed to notice banding on moving images.
My observation may not be specifically relevant to your analysis but it is relevant to plasma displays, or at least for the 50" Fujitsu displays: Smooth pans showcase banding more prominently than still images. I have speculated that this could be because of a difference in decay rates of the various colored pixels, as suggested by the observation that the bands follow the motion and are of specific colors. I'm pretty sure I've seen this phenomenon on every plasma display I've studied. It is not present on a CRT display.


I would say that this artifact, plus the poor fill factor which limits how close one might like to sit to the screen, are the significant remaining drawbacks of plasma display technology.


But back on subject, what I'd like to hear are some opinions as to what would be a reasonably adequate bit depth to basically completely eliminate banding. 8 bits aren't enough, by a long shot. And I'll stipulate in advance that smearing "cheats", like what th Fujitsu 50" displays do to simulate better bit depth and black levels, don't count. Nor does the natural anti-banding degradation you get from using non-digital video.
 

·
Registered
Joined
·
518 Posts
Quote:
Originally posted by Colmino
My observation may not be specifically relevant to your analysis but it is relevant to plasma displays, or at least for the 50" Fujitsu displays: Smooth pans showcase banding more prominently than still images. I have speculated that this could be because of a difference in decay rates of the various colored pixels, as suggested by the observation that the bands follow the motion and are of specific colors. I'm pretty sure I've seen this phenomenon on every plasma display I've studied. It is not present on a CRT display.
I agree that some pans can hilight the problem precisely. But it really depends on the pan -- namely the colors shown in it. Banding tends to show up when there is a small (color) gradient over a "long" pixel distance. You will most likely notice banding in pans when you see "purer" gradients ala white, such as fog, dust, or haze scenes.


For most pans where there is a good mix of color, banding is not normally an issue, due to small distances.


The Matrix * has now become my torture test for white quantization and false contours.


Here are the timestamps that everyone can use to check for themselves. Format is Chapter, Hour:Min:Sec.Frame. (Sorry, no pictures -- yet.)


Banding Pan Torture Test

DVD: The Matrix

Timestamp: 1, 0:01:28-40s


Banding Still-Frame Torture Test

DVD: The Matrix

Timestamp: 1, 0:01:34.8


I've been *very* critical towards white gradients on my new plasma since I noticed the banding artifacts *immediately* via S-Video cable. Interestly enough, after switching over to component cables (and progressive scan), banding is almost gone! What a *noticable* difference.


Now, I didn't realize that Plasma's "dither" colors, and use time-cycling (decay rates) to show colors. (Ironically, I didn't pay any attention to the technical specs when I bought my Plasma - just the PQ. Which goes to show specs aren't everything, even for us guys that do care about them :)) Colmino, I completely agree with your conclusion that decay rates, and seating distance can also make a difference in perceived banding.


* Mini-rant: While The Matrix is my favorite movie, I think the encoding "was just a little too much for it" - especially while watching it with S-Video and seeing the banding artifacts. Also, the decreased PQ might be due due to the green tint applied to it -- which our eyes happen to be the most sensitive too, of all the colors. If anyone has a good "green test", please share!

Quote:


I would say that this artifact, plus the poor fill factor which limits how close one might like to sit to the screen, are the significant remaining drawbacks of plasma display technology.


But back on subject, what I'd like to hear are some opinions as to what would be a reasonably adequate bit depth to basically completely eliminate banding. 8 bits aren't enough, by a long shot. And I'll stipulate in advance that smearing "cheats", like what th Fujitsu 50" displays do to simulate better bit depth and black levels, don't count. Nor does the natural anti-banding degradation you get from using non-digital video.
It's going to be real hard to say exactly how many BPP we need (10, 12, ? without seeing it, other then, yeah, we need more then 8 BPP. Interestly enough, it looks like the Panasonic ED Plasma is around ~ 7.5 bits/pixel.


Finally, we can't ignore the DVD specs. i.e. 4:4:4, etc., but I'm not well-versed with the "hardware" specs.


Cheers
 

·
Banned
Joined
·
17,607 Posts
In Plasma grey scale levels are created with pulses. There is a limit as to how many pulses can be packed in a 60th of a second.
 

·
Banned
Joined
·
3,772 Posts
I think the biggest test is when there is a pale blue sky with sun and clouds. If you don't have enough bpp then there are visible bands (often in a graceful arc)
 
1 - 16 of 16 Posts
Status
Not open for further replies.
Top