AVS Forum banner
1 - 20 of 34 Posts

·
Registered
Joined
·
40 Posts
Discussion Starter · #1 ·
I've been reading about (or trying to anyway) dithering on plasma televisions but I'm not really clear on some things. For example, when someone says dithering are they including the "flickering" of the individual sub pixels to vary the intensity, or are they truly talking about old-school 16 color image style dithering where you mix the colors into, say, a grid (of course the real algorithms would be much more complex) to get a color in between?


More importantly, exactly how many intensities can a single sub pixel produce on a modern plasma, and does this vary significantly from model to model? Panasonic claims the V10 series can do 6,144 "Shades of Gradation" but then it also says "equivalent" .... what the heck does that mean?? And where does this number come from it seems way to high to be real...


To me, anything less than 256 (ie 8 bits per channel) would be completely unacceptable. Granted this is based more with experiences with dithering in other places (eg printers, pre-24 bit color displays, etc) but if you ask me even 256 shades is not enough as I can see the banding clearly on a single color ramp, so if the dithering plasma does is to defeat this, then that's a great feature... if its to overcome the inability of a single sub pixel to produce a sufficient number of intensity levels ... that's not so good.


I need to drag some test patterns into best buy and convince them to let me look at them regardless of the answer to this question, to see just how good/bad it looks, but I have a hard time believing what would essentially be a less than 8 bit display could look good in situations with smooth color ramps...


As a side note why can't the companies just release the real information for things like this. I mean I can understand things that can be measured in different ways eg contrast (though even that has a standard they SHOULD be using) but there is no real fudge factor in how many shades a sub pixel can produce so why not just come out and say it?!? Also display input lag, that REALLY needs to be a listed spec its not hard to measure, and in fact most of it they know about before they even begin building the displays since its based largely on how many frames the various processing routines they are doing need in advance... argh picking a new tv is turning into a major headache.
 

·
Registered
Joined
·
3,186 Posts
Unfortunately the answer to your question is rather complicated. The following information is what I've gathered from company patents and literature. There are two main driving methods for Plasma displays.


1 - Binary (Panasonic) - Pulse widths are weighted in Binary (8subfields = 1,2,4,8,16,32,64,128) and combinations of these weights make up 256 possible levels per subpixel. Newer Panasonic models claim 10 subfields which adds up to 1024 levels per subpixel.


2 - Contiguous (Pioneer) - Pulses can only be adjacent to each other (one after another) and therefore the number of subfields equals the number of possible gray levels per subpixel. Pioneer uses 14 subfields and thus only 14 gray levels per subpixel.


The amount of "equivilent" gray levels is generally the total amount of gray levels the panel is capable of creating through a combination of true sub-pixel gray level, error diffusion, spacial dithering, and temporal dithering.


Pioneer chose the path of Contiguous driving to completely eliminate the artifact known as DFC (Dynamic False Contouring) even though it meant a higher black level and the need for sophisticated Halftoning. Ironically, this decision ultimately led to the revolutionary KURO blacks and ECC.


Cheers
 

·
Registered
Joined
·
40 Posts
Discussion Starter · #3 ·
So panasonic can do 256 or more true intensity differences on each sub pixel and the pioneers that seemingly everyone considers the best displays in the world can only do 14? something doesn't sound right.


I know manufacturers try to make it complicated, but it really isn't - spatial, temporal, voodoo or any other kind of dithering is nonsense if you ask me. The manufacturers need to say plainly how many intensities each sub pixel can produce.


If you start including dithering in your "shades of gradation" or however you want to specify it then even a black and white panel where each pixel is either fully on or fully off can go ahead and claim 1920*1080= over 2 million shades of gray! (1 pixel on, 2 pixels on, etc) Cleary this is complete nonsense, the correct answer for such a panel would be 2 because thats what each pixel can do, but my point is including dithering makes these numbers even more worthless than contrast ratios.


Also are you sure the pioneer can only do 14 shades per sub pixel? that seems crazy low really, I can't imagine any amount of dithering making that not look terrible. I suppose if thats before the "flickering" style dithering which is the only dithering that I could see legitimately being included that might be right... By flickering I mean that for a given frame, say the sub pixel is at half brightness half the time making it 1/4 brightness, etc. If the sub pixels are driven at 600 hz then for a 60 fps display then thats up to 10 different states the pixel can be in per frame. if it can be in any of the 14 states for each of the 10 slots then thats 14^10 shades (though the distribution of them would be... a bit odd) so I could see it working out...



Anyway I suppose the really simple question is if you display a nice static spectrum image like this one (attached) on a plasma display, is it nice and smooth or does it show the usually obvious artifacts caused by dithering?
 

·
Registered
Joined
·
3,186 Posts

Quote:
Originally Posted by veridiac /forum/post/16982120


So panasonic can do 256 or more true intensity differences on each sub pixel and the pioneers that seemingly everyone considers the best displays in the world can only do 14? something doesn't sound right.

I agree. And when I first read it I posted the same thoughts. However, after further reading I realize the dramatic benefits that this driving method allows for which includes much better low level gradation. The drawback is the requirement for very sophisticated halftoning. It is very complicated IMO.

Quote:
Originally Posted by veridiac /forum/post/16982120


If you start including dithering in your "shades of gradation" or however you want to specify it then even a black and white panel where each pixel is either fully on or fully off can go ahead and claim 1920*1080= over 2 million shades of gray! (1 pixel on, 2 pixels on, etc) Cleary this is complete nonsense, the correct answer for such a panel would be 2 because thats what each pixel can do, but my point is including dithering makes these numbers even more worthless than contrast ratios.[

Plasma display panels do not use TFTs to control individual pixel levels like LCDs do. They use common electrodes and therefore (for the most part) there are only two possible "states" for every subpixel on the panel. That being "on" or "off". The true gray level per subpixel is determined by the time "on". The equivilent gray level you see is determined by halftoning.


Quote:
Originally Posted by veridiac /forum/post/16982120


Also are you sure the pioneer can only do 14 shades per sub pixel? that seems crazy low really, I can't imagine any amount of dithering making that not look terrible.

That is what the patent and literature state. Here is an example of Panasonic vs Pioneer type driving.

Quote:
Originally Posted by veridiac /forum/post/16982120


I suppose if thats before the "flickering" style dithering which is the only dithering that I could see legitimately being included that might be right... By flickering I mean that for a given frame, say the sub pixel is at half brightness half the time making it 1/4 brightness, etc. If the sub pixels are driven at 600 hz then for a 60 fps display then thats up to 10 different states the pixel can be in per frame. if it can be in any of the 14 states for each of the 10 slots then thats 14^10 shades (though the distribution of them would be... a bit odd) so I could see it working out...

You have this wrong. There are only 2 possible "states". The flickering pixels you speak of is due to the temporal dithering (rotation of "on" state across adjacent pixels). And as I said the weighted subfields are binary for Panasonic. Therefore at 600Hz (60Hz) the panels has 10 subfields with the following weights (1,2,4,8,16,32,64,128,256,512). Only combinations of those subfields can happen. That is 2^10 and that add up to 1024 possible shades.
 

·
Registered
Joined
·
40 Posts
Discussion Starter · #5 ·
Hmm I think I see... though really it seems more like ALL plasmas are stuck with 10-14 levels of intensity per sub pixel which is scary... I need to see that test pattern on a plasma...


It looks to me that, given x subfields, any plasma regardless of drive method can have a sub pixel turned on for 1/x, 2/x ... x/x percent of the time correct? So for 600 hz subfield drive you get 10 subfields on 60 hz video, etc.


Granted the first method has 2^10 number of states for 10 subfields, but if the pixel is only on or off, I don't see where there is any difference between say, on off on off on off... and off on off on off on, etc. all that should matte ris the % of time the pixel is on. Also what do they do about different refresh rates? to get any intensity variation at all the speed of the phosphor is going to need to be matched to the refresh rate so that a 100% drive for 1 frame puts the phospher at 100% brightness... either that or the phosphor is super fast and you let peoples retina smooth out the flicker? That would explain why plasmas in stores always looked a little flickery to me....


I had no idea that a plasma cell couldn't be driven to different intensities by varying the voltage... I suppose it makes sense now that I think about it, I mean it probably can (probably thats what the contrast adjustment does) but not without some threshold between 0 brightness and minimum brightness where it activates at. I suppose my question now is more "How do plasma displays not look like utter crap?" I suspect after I look at some test patterns I'm going to be back for searching in vain for a large low input lag LCD with a half decent black level



ARGH


and correct me if I'm wrong on any of that.
 

·
Registered
Joined
·
40 Posts
Discussion Starter · #6 ·
As a side note what happens when the V10 starts displaying 24p content at 96hz... does it get even fewer levels of intensity? It would have to wouldn't it, the sub fields are still only driven at 600 hz right? though for that matter, 600 hz doesn't divide evenly into 96, comes out to 6.25... are they really driving subfields more often than that? 6 levels shouldn't be enough no matter what tricks you play...
 

·
Registered
Joined
·
3,186 Posts

Quote:
Originally Posted by veridiac /forum/post/16982518


Hmm I think I see... though really it seems more like ALL plasmas are stuck with 10-14 levels of intensity per sub pixel which is scary... I need to see that test pattern on a plasma...

Only in contiguous. In Binary, 10 subfields will yeild 1024 distinct levels. For example:


10 binary subfields weight will be (1,2,4,8,16,32,64,128,256,512)


In order to get level 483 the panel would use the following combination sequence


(1,2,-,-,-,32,64,128,256,-)

Quote:
Originally Posted by veridiac /forum/post/16982518


It looks to me that, given x subfields, any plasma regardless of drive method can have a sub pixel turned on for 1/x, 2/x ... x/x percent of the time correct? So for 600 hz subfield drive you get 10 subfields on 60 hz video, etc.

That is incorrect for any drive method. All plasma displays use weighted subfields.

Quote:
Originally Posted by veridiac /forum/post/16982518


I don't see where there is any difference between say, on off on off on off... and off on off on off on, etc.

Again, you have to take into account that each subfield is weighted differently.


Quote:
Originally Posted by veridiac /forum/post/16982518


either that or the phosphor is super fast and you let peoples retina smooth out the flicker? That would explain why plasmas in stores always looked a little flickery to me....

Each subfield is weighted in increasing order and that is why Plasma displays produce flicker. Essentially there is an effective duty cycle around 35-50%.

Quote:
Originally Posted by veridiac /forum/post/16982518


I had no idea that a plasma cell couldn't be driven to different intensities by varying the voltage... I suppose it makes sense now that I think about it, I mean it probably can (probably thats what the contrast adjustment does) but not without some threshold between 0 brightness and minimum brightness where it activates at. I suppose my question now is more "How do plasma displays not look like utter crap?" I suspect after I look at some test patterns I'm going to be back for searching in vain for a large low input lag LCD with a half decent black level

Sophisticated halftoning is the answer. Even high end (1M$) digital color printers use halftoning to create stunning images


Also, the contrast adjustment does not adjust the voltage. It adjusts the width of the subfields (number of sustain pulses)


Cheers
 

·
Registered
Joined
·
958 Posts

Quote:
Originally Posted by veridiac /forum/post/16982518


I suppose my question now is more "How do plasma displays not look like utter crap?" I suspect after I look at some test patterns I'm going to be back for searching in vain for a large low input lag LCD with a half decent black level



ARGH


and correct me if I'm wrong on any of that.

Which models did you see that were unacceptable to you? Large low input lag LCDs with half decent black level exist... Have you even gone out and experimented with a V10 yet? Phosphor lag is what stopped me from purchasing a PRO-101 or TC-P50V10... but I'm told that most folks out there are oblivious to the effect (other magic associated with PDPs took a back seat). I think it has something to do with my eyes picking up on the largest sub field!
 

·
Registered
Joined
·
40 Posts
Discussion Starter · #9 ·
What do you mean by weighting? if the pixel can be only on or off, then are the subfields of different length in time? If they are all the same duration, and the pixel has no intensity variation, then whether its on for the first half of the frame or the second would be the same.


And by input lag I mean the lag time from when the image is sent over the HDMI cable (or whatever the source is) and when its displayed, pretty much all PVA displays need a minimum of 1 frame look-ahead to reduce pixel response time, though most LCD tvs take FAR longer than that, sometimes up to 100 ms. Game mode is less but from what I gather, even game mode is frequently in the 50-75 ms range.


And the thing about input lag and games is even if you don't notice it, it is affecting your reaction time, literally adding whatever the lag is to your reaction time since thats how much later you see the image, so a display with 100 ms of lag would give a teenager the reaction times of a 90 year old. (typical reaction time is ~200-220 ms)


As far as phosphor lag goes... I'm still not sure if I can see it or not, haven't noticed it in stores but I haven't seen much high speed black and white stuff either.


Also nobody around here even has a V10 for me to look at that I've found yet
been looking at the S1s in best buy




Anyway thats off topic - has anyone displayed some nice smooth static RGB spectrums on plasmas? how do they look? I'm starting to think maybe they really do only have 14 shades, and that they just dither the various nearby pixels at 600 hz.... I'm thinking maybe thats the flicker I see? Not really sure what rapidly changing dither at 600 hz would look like, I'm thinking either flicker shimmer or noise or some combination of those... How do still images look on plasmas in general?
 

·
Registered
Joined
·
40 Posts
Discussion Starter · #10 ·
Another random but possibly related question - why do they list "moving lines of resolution" and why is it less than 1080 on some displays? if its a true 1080p display shouldn't there always bee 1080 lines of resolution? (whether or not you can see them is another matter, but they should be there...)
 

·
Registered
Joined
·
3,186 Posts

Quote:
Originally Posted by veridiac /forum/post/16983683


What do you mean by weighting? if the pixel can be only on or off, then are the subfields of different length in time? If they are all the same duration, and the pixel has no intensity variation, then whether its on for the first half of the frame or the second would be the same.

A subfield is an alotted time period to emit light. 10 subfields means there are 10 time periods to emit light. Weight is the length of time.


In my example above, 10 subfields have the following weight profile


sub1 = 1

sub2 = 2

sub3 = 4

sub4 = 8

.

.

.

sub10 = 512


You can take this to mean that sub1 is 1/512th as bright as sub10. In other words sub1 is extremely dim while sub10 is extremely bright.


Or you can look at it as sub1 is 1/512th as long as sub10. In other words sub1 is extremely short while sub10 is 512 times as long.


Another way to look at it is that sub1 only has 1 pulse of light while sub10 has 512 pulses of light.
 

·
AVS Forum Special Member
Joined
·
11,139 Posts

Quote:
Originally Posted by veridiac /forum/post/16983696


Another random but possibly related question - why do they list "moving lines of resolution" and why is it less than 1080 on some displays? if its a true 1080p display shouldn't there always bee 1080 lines of resolution? (whether or not you can see them is another matter, but they should be there...)

Fixed-pixel 1080p displays should display 1920 alternating static B&W test-pattern lines vertically across a 16X9 screen and 1080 alternating static B&W horizontal lines from top to bottom. This may require a known 1080p source from a test-pattern generator since other sources such as HD test discs/players or broadcast patterns may involve higher frequency/resolution filtering.


Motion of test patterns involves so-called Kell factor reduction of vertical resolutions. Kell factor relates to vertical sampling and typically reduces vertical resolution by 0.7 X line count. So 1080 becomes 756 resolvable lines of effective resolution. (Pattern-generator and computer test-pattern sources aren't sampled.) Kell factor applies to both progressive sources, such as 1280X720/60p HD, and 1080/60i, as outlined by SMPTE-fellow/consultant Michael Robin in separate articles sublinked here . Also, vertical resolution varies directly with the rate of vertical motion. As the table of test results for approval of ATSC HD show, the static resolution of 1080/720 formats varies greatly from a 15-rpm test pattern (dynamic resolution results shown), and resolutions would be different for other test pattern rpms. Table 2.3 shows the original test results from the FCC advisory committee; the sublink in the 'table link' just above doesn't work now.


In addition, interlaced SD/HD has morel vertical filtering applied, vertically smearing lines together, to minimize 'twittering' of fine high-contrast details (see Joe Kane link below). Oversampling/down-scaling motion-video program material, like using 1080/24p master tapes to produce 720p recordings for broadcast, will boost both vertical and horizontal effective resolutions (resolvable details) compared to capturing images at only 720p. And as consultant Joe Kane pointed out (see last quote here ), filmed material telecined to 1080/24p master tapes (~270 Mbps HD-D5 tape) is typically only 800--1300 lines maximum effective horizontal 16X9 resolution. An updated study involving newer recorders or non-telecined recordings boosting these effective resolutions for movie dramas hasn't been published AFAIK.


Comparisons of displays using special motion test patterns, like those at hdguru.com, are valuable since the same signal is used for each display. But IMO the numerical figures shown (e.g., 900, 1080 lines) need to combined with other test patterns and methods for complete evaluation. And, as outlined above, non-sampled test patterns differ from sampled motion-video program material, although the ATSC-approval tests linked above certainly appear to be with sampled test patterns, not computer or hardware-generated.. Static non-sampled patterns (guru's 'bandwidth' test) usually show full resolution. Interlaced patterns and standard sampled programming fed to fixed-pixel displays involve deinterlacing and/or scaling by the video processing section, which vary widely in effectiveness. -- John
 

·
Registered
Joined
·
18 Posts
Wow, just found this thread.


This covers the area I am currently investigating on my Pioneer plasma. i.e. the way it drives the pixels using CLEAR.


The CLEAR method seems rather limited in terms of colour palette. I'm talking about absolute colour palette PER FRAME rather than extended palette over several frames using time/spatial dithering techniques.


I'm not entirely happy with my kuro especially on HD material so I found some articles about how they drive the panel using 14 contiguous sub fields etc.


I drew out my interpretation of what this means for the picture on a frame by frame basis and it appears the kuro has 14 chances (subfields) within a frame to edit the status of the sub pixel. As the system is contiguous the pixel gets turned on once and off once PER FRAME. (or it can stay OFF or ON throughout the frame)


This makes for a very crude PWM based system with only 15 possible duty cycle settings for the sub pixel per frame? (assuming I'm reading this correctly)


Therefore if you grabbed ONE FRAME from a kuro display it would show a limited colour palette and would also show signs of spatial dithering to try and extend the apparent palette. It must take a LOT of processing to do this dithering adequately!


Over the course of several frames the eye would integrate all this to 'see' a much richer colour palette. (time dithered and also spatial)


On a still image this would produce really good results (with a small loss in ultimate resolution due to spatial dithering)


Also, I would guess there would be wide dynamic range (subpixel always off across all subfields gives deep black) (pixel ON across all subfields gives very bright image)


Also, the (essential) spatial dithering will mask colour banding due to its peppering effect over several frames.


However, I don't see how this system can cope with fast motion?


Doesn't the time dithering fall on its face? the spatial dithering would struggle as well.


In contrast, the conventional non contiguous drive method can produce a rich colour palette on every frame. eg the panasonic method.


The blacks might not be as good but the image must surely look more natural and have better resolution?


That's what I see when I compare the two TVs side by side....
 

·
Registered
Joined
·
3,186 Posts

Quote:
Originally Posted by Chelsea_Fan /forum/post/17018373


Wow, just found this thread.


This covers the area I am currently investigating on my Pioneer plasma. i.e. the way it drives the pixels using CLEAR.

I had the exact same questions when I first learned how CLEAR driving works. However, after studying this method for years now I came to realize the benefits outweigh the drawbacks IMO.


Benefits of CLEAR (contiguous):


1 - Complete elimination of DFC (Dynamic Fasle Contour)

- binary drive like Panasonic requires sophisticated subfield orders to just reduce DFC


2 - Only 1 Reset per frame

- this enabled drastic improvements to black level, brightness, and low level grayscale


3 - Variable sustain pulse #

- using contiguous subfields enables many more gray levels at low APL than does Binary


As for the supposed drawbacks of too few subpixel gray levels. Well, Pioneer uses sophisticated halftoning to render a frame. Combinations of error diffusion, spacial dither, and rotational dither along with variable sustain pulse weights enable billions of colors possible. Add to that the possibility that Pioneer uses a scan sustain method (They have in the past - "Advanced CLEAR")


But you may be right about comparing individual frame renders since rotational dither is definitely used in CLEAR driving.


However, Panasonic also claims billions of colors even though the binary method cannot come close to rendering this amount. I assume they also use plenty of halftoning.
 

·
Registered
Joined
·
885 Posts
I admit I have not delved into all the math yet to fully compare these dithering methods. However, can I make the blanket statement that an LCD screen, with its variable crystal states (vs. the limited PWM pixel states of a plasma) can forego dithering, and therefore, would be a better display device for static computer generated images?


Michael
 

·
Registered
Joined
·
2,263 Posts
In my experience, eliminating the ill effects of dithering is largely a matter of finding the right viewing distance.


Since most people sit fairly close to their computer screens, LCD is generally the best choice.
 

·
Registered
Joined
·
1,071 Posts
xrox, you claim Pioneer's method reduces DFC but I see it all the time when I watch CGI animated films like Kung Fu Panda or Cars! It's most visible in bright pictures with slow panning movements from the camera.


I even see it in real world material occasionally. It bugs the hell out of me!


Ironically, on really detailed scenes (bright and dark, various colors), I can't see it at all.


I've been running my PRO-111FD in Pure Mode since the day I got it and use the video orbiter as well as the 1 hour IR fixer (as I thought this was a result of uneven break-in thus this method to fix that). Is there any way to eliminate this effect?
 

·
Registered
Joined
·
3,186 Posts

Quote:
Originally Posted by ChuckZ /forum/post/17277646


xrox, you claim Pioneer's method reduces DFC but I see it all the time when I watch CGI animated films like Kung Fu Panda or Cars! It's most visible in bright pictures with slow panning movements from the camera.


I even see it in real world material occasionally. It bugs the hell out of me!


Ironically, on really detailed scenes (bright and dark, various colors), I can't see it at all.


I've been running my PRO-111FD in Pure Mode since the day I got it and use the video orbiter as well as the 1 hour IR fixer (as I thought this was a result of uneven break-in thus this method to fix that). Is there any way to eliminate this effect?

Not sure what you are seeing is DFC as it is technically impossible with continuous driving. This is assuming of course that the 111FD is actually using continuous driving (aka - CLEAR, super CLEAR, or advanced super CLEAR).


What does the DFC look like?
 

·
Registered
Joined
·
1,071 Posts

Quote:
Originally Posted by xrox /forum/post/17279060


Not sure what you are seeing is DFC as it is technically impossible with continuous driving. This is assuming of course that the 111FD is actually using continuous driving (aka - CLEAR, super CLEAR, or advanced super CLEAR).


What does the DFC look like?

The DFC looks like banding--isn't that the very definition?


Now granted, I do sit very close (5 feet away) so it's very easy to see.
 
1 - 20 of 34 Posts
Top