Panasonic Plasma - great picture produced by heavy dithering

VierraFan

Established Member
Joined
Feb 17, 2012
Messages
213
Reaction score
74
Points
41
Although this topic was discussed here and elsewhere, it turned out to be impossible to get an accurate information how modern plasma displays actually display color. What we have on the internet are some rare descriptions and several patents, mostly from 10 years ago, and some marketing stuff, claiming 600 Hz subfield drive and billions of colors displayable by PDP. Classical description is that each frame consists of 8 subfields, so we have 256 levels of each color. Somewhere we can read about more subfields and more levels, somewhere we can read that not all levels are used because of false contouring effect etc ... dithering is frequently mentioned, but not how much dithering we actually have ... etc.

So, in the absence of accurate information, I decided to make a timewarp of one PDP to find out. I want to point out that I made it just out of curiosity, because I'm a kind of technical guy, and not to start a war LCD against PDP. I'm quite satisfied with the TV, I just wanted to find out how it works. In this case, it was Panasonic TX-P42UT30E. It's an European model, placed below GT30, lacking recording and color management options (compared to GT30), but featuring G14 panel, 3D etc.

A Timewarp was performed using quite cheap equipment, on old Canon 400D DSLR, using its mechanical rolling shutter feature to record individual subframes on red, green and blue gradient, displayed on the TV at 24fps using built-in media player. In that case, it is displayed at refresh rate 60 Hz, using 3:2 pulldown, as it seems that TV can't display 24p video using 48 or 96 Hz from built-in player (so 24p smooth movie feature is not available). Made about a hundred photos in 3fps burst mode, using the shortest exposure of 1/4000 (for blue) or 1/2000 (for red and green) and stitched them all together in one big image (hence some imperfections) illustrating how a single frame is actually displayed.

On attached picture, x-axis is duration of one 60 Hz frame. y axis is intensity of displayed color, with black on the top and maximum intensity at the bottom, so it's evident which subfields are used to display various intensities of red, green and blue. Contrast was set to maximum and Eco mode was off, so TV produced maximum available brightness

Well ... it turned out that engineers in Panasonic did an excellent good job, because a really good picture coming from this TV is achieved by only several thousand possible colors PER ONE 60 Hz FRAME, as nobody noticed that. Also, I can't find expected ten subframes. There are 8 subframes for red and green, and 7 for blue - at least to display a full range of individual red, green and blue color at 60 Hz refresh rate. Also, it's evident that only a small part of combinations of that subfields are actually used, so number of total colors within a single frame is several thousand at best. Everything else is achieved by strong usage of temporal and spatial dithering, which is clearly visible on the photos (evident from the enlarged photo of one green subfield)

It's interesting that usage of subrafmes is quite different when reproducing 25 or 50 Hz PAL video, with even less possible colors. I didn't test 3D mode, but according to Panasonic brochure, in 3D modes subframes are reversed, with longest first and shortest last, to reduce ghosting caused by phosphor decay

Also, it's visible that there is still a big difference in decay time between red, green and blue phosphor, with green being the slowest.


Alex
 

Attachments

  • T50_24_RGB_small.jpg
    T50_24_RGB_small.jpg
    49.3 KB · Views: 2,053
  • Dithering.jpg
    Dithering.jpg
    82.2 KB · Views: 2,249
I will admit I didn't understand most of that. But you may well be best placed to answer a question I have about how these current models produce their picture compared to older models.

I have already posted here about my dissatisfaction about the screen noise seen at the top portion of the screen. One of the reviewers here mentioned that they noticed it at the top ten percent or so of every Panasonic they reviewed.

But what I've noticed is - compared to my old PZ80 - is that that screen noise appears anywhere on the screen where there are sections of bright white. You can see it clearly as you get closer to the screen and it follows the white sections of the picture as it moves around.

One other thing I have noticed as well compared to my PZ80. Is that up close to the screen when it is black (on an HDMI input with no signal for example) is that you can see the scan lines making up the picture and can see red/green pixels in with the black.

On my PZ80, a back screen looked black close up, whereas this current model Pansonic it looks "coloured" by comparison.

Sorry if this is off-topic somewhat, but you seem like you may now what the differences are between older and newer models. For what its worth, I think my older set had a far superior picture. Thanks
 
Thx for your experiment.

If you got time to go further you should see that the number of subfields/subframes will differ depending on :
- input frequency (you already witnessed it @60Hz vs 50Hz (actually 100Hz));
- picture mode (vivid vs normal vs cinema/pro/isf) ;
- Average Picture Level (contrast set at maximum is not enough here, you could use 10 to 100% APL greyscale slides) ;
- motion (photo picture mode vs the other modes) and motion speed (good luck shooting it).
Same goes for the usage and precision of spatial and temporal dithering.

But stuff tends to get even more complicated. The max number of sf, their order (partially inverted in 3D etc) and the weight (length) of each subfield seem to be also dynamically managed depending on the content displayed. Oh and btw, make sure you've waited long enough before shooting your screen after a cold start if your room temperature is low because this will also come into play. Yeah, crazy I know...

Welcome to the insane world of pdp tech.
 
MrBungle2005,

Yes, it is quite high technical stuff, intended to those who already have some basic knowledge how plasma display works, so I didn't expect that everybody would understand

In short, picture on the TV is usually refreshed (painted) 60 times per second (although there are other possibilities 50, 48, 96 ..., depending on video source). If you are watching a movie, it has 24 frames per second, so even frames are painted 3 times and odd frames 2 times (unless you have 24p input via HDMI). It is called 3:2 pulldown. I made a test video with 24 frames / second and displayed it on the TV (see attached file)

Each pixel on the TV consists of red, green and blue subpixel. Unlike CRT and LCD, in plasma display each pixel can be only turned on or off. So various levels of red, green and blue are produced by pulsating those pixels. In this case, within each refresh of the screen (lasting 1/60s) we can have 1 to 8 pulses for red and green and 1 to 7 for blue. First is the shortest, last is the longest. Duration of those pulses are the same for all pixels, just various number of pulses is used to lit the. Each pulse is usually called a subfield, so here we have 8 subfields during one refresh of the screen

If you look at picture attached in this post, there are three arrows. Top arrow is for dark blue. It is produced only by the first pulse. Middle arrow is for medium blue. It is produiced by 1st, 2nd, 3rd and 5th pulse. Bottom arrow is for bright blue. It is produced by all 7 pulses.

However, we can see that from black to bright blue only about 10 combinations is used (out of 127):

1
1+2
1+2+3
1+2+3+4
1+2+3+5
1+2+3+4+5
1+2+3+4+6
1+2+3+4+5+6
1+2+3+4+5+7
1+2+3+4+5+6+7

In digital video input there are about 240 possible levels of blue, but each time plasma display screen is refreshed, it can produce only 10 levels of blue. That's where dithering comes is. Pixels are alternated among those 10 levels to produce all levels. Dithering can be temporal (during successive refreshes of the screen) and spatial (among neighboring pixels)

It is understandable that plasma makers don't want to explain how this works. People won't be happy about buying a TV that can, at each refresh, display only a small number of colors. However, in real life it works quite good. Dithering is actually visible only if you are observing the screen from close distance, mostly in dark areas, where you can see individual pixels dancing around, because even one (first) subfield too bright to reproduce darkest possible color. And yes, actual implementation of these subfields and dithering is different from model to model. Also, as I said, in this TV, refresh rate of 50 times / second (for PAL video material we have here in Europe) is implemented considerably different than 60 times per second

If someone has a DSLR with video mode (which my camera doesn't have) , it should be much more easy to record similar thing like I displayed here, by reproducing attached video clip on the TV. Those cameras have a sensor with "rolling shutter", which is scanned from top to bottom about 30-60 times per second (mechanical shutter in my camera is much faster, so I had to stich many photos together to get the full picture). However, you have to manually set very short video exposure - if the camera has that possibility (1/1000 or shorter) and ISO about 1600 or higher. I'm surprised than nobody on various sites testing plasma TVs didn't come to that idea
 

Attachments

  • T50_24_blue2.jpg
    T50_24_blue2.jpg
    50 KB · Views: 530
  • HDtest31_24.rar
    9.9 KB · Views: 95
Orso,

unfortunately, experimenting with a camera without video mode is quite time consuming. I could only observe 60 fps mode, as I was lucky that my camera has almost precisely 3 fps, so successive photos in the burst are coming at similar position in the screen refresh (which is not the case with 50 fps, which can actually be 100, as you mentioned), just slightly shifted. Shutter is very fast, so I can record only about 1/10 of the whole refresh on one photo. Should be much easier to experiment with a camera with CMOS sensor and video mode, as I mentioned in previous post (compact camera with CCD won't show this, as it has a global shutter - whole sensor is scanned at once). So I attached a test video, if someone is willing to experiment further

I attached here another picture, showing how gray gradation is displayed during 1 screen refresh. Shutter 1/1000 was used here, so blue pulses look much wider than in previous picture (where they are also considerably wider than they really are, as they are shorter than 1/4000s)
 

Attachments

  • T50_24_white.jpg
    T50_24_white.jpg
    89.2 KB · Views: 792
I finally managed to stitch together photos of 50 Hz mode. Because of limited number of them, resulting image is somewhat incomplete, but I got what I wanted. You can see it in attached picture - for green color only

I would classify this as 50 Hz refresh rate and not 100 as Orso suggested, because whole pattern repeats 50 times per second. It looks different than 60 Hz refresh pattern. And it looks different than in slides provided by Panasonic. There can be 12 subfields within 1/50 of second (600/50 = 12), but only 10 of them are used to display all gradations of green and red and only 9 for blue. And out of 1024 possible combinations of 10 subfields = 1024 gradations of green (2^10=1024), only 11 (ELEVEN) are used in this setting (normal mode, maximum contrast). Also, slide by Panasonic suggest that width of pulses is rising continuously from first to last, like in 60 Hz mode (although middle subfields are not shown), but they are actually in 2 groups here. A big difference between marketing stuff and reality.

In 60 Hz mode, there are about 14 gradations of green (not quite sure because picture is blurred by usage of dithering, so there could be 1 or 2 more - not a big difference though). There was a remark in some reviews that 50 Hz mode exhibits more banding than 60 Hz and this could be a part of explanation. On the other side, conclusion in other reviews was that no significant difference in color reproduction was observed. But one thing is sure - method to display color in 50 and 60 Hz mode is different.

Unfortunately, can't test 24p mode to see what it looks like, because I don't have HDMI source and built-in media player displays 24fps video at 60 Hz, so with IFC off, there is similar pulldown judder like on my 60 Hz LCD monitor, just little more obvious because of faster response time of PDP.
 

Attachments

  • T30_50_green.jpg
    T30_50_green.jpg
    45.2 KB · Views: 447
  • Panasonic PDP 2500Hz focused field drive.jpg
    Panasonic PDP 2500Hz focused field drive.jpg
    295.2 KB · Views: 536
So is the first pulse in the series the brightest one or the faintest?
Would make sense to have brightest to achieve minimum lag.
Leo
 
but they are actually in 2 groups here.

50Hz @100Hz : same picture displayed twice within 20ms. 10ms per frame compared to 16,7ms@60Hz, hence less sf and/or shorter (narrower) pulses. In order to maintain a decent light output, pulse length cannot be cut too much, hence the need to reduce the number of sf. Depending on picture mode, APL (how bright the frame needs to be) and motion, the number of sf and their length will differ.

If you could test 24p@96Hz, you would prolly see 4 groups of sf.
 
Last edited:
Hello,

that's just subfield grouping. The frame rate is still 50hz and 24p is most probably displayed at 48hz. Please check this pdf.
 
So is the first pulse in the series the brightest one or the faintest?
Would make sense to have brightest to achieve minimum lag.
Leo

In both 60 and 50 Hz, first pulse is the shortest, so it is recorded as faintest by the camera. In case of 60 Hz, width of pulses is rising towards the end of 1/60 s interval (which is also recorded by the camera, as they are wider, although photo doesn't show real width of the pulse, because all except last pulses are much shorter than exposure of the photo). In case of 50 Hz, there are two groups within 1/50 s interval. First group consists of 6 pulses, second group consists of 4 pulses. Pulses sorted by their lengths would be in this order 1st, 2nd, 3rd, 7th, 8th, 4th, 9th, 5th, 10th, 6th, as visible from the photo in my last post

In brochure by Panasonic, it's stated that subfields are indeed reversed (longest first) in 3D modes, which I didn't test so far (I made a test video in SBS format also, but TV refused to display it 3D; will try to fool it somehow when time permits)

I attached a picture describing Clear driving of Pioneer plasmas, which, by coincidence, is showing usage of subfields similar way as I recorded it by the camera (time on x axis, brightness of image on reversed y axis). Maybe it will be easier to understand what it is all about comparing it with my photos. It is visible that methods used in this TV are a kind of hybrid between conventional and clear driving method, but closer to clear driving, as number of available gradations is not much higher than number of subfields
 

Attachments

  • Pioneer Kuro - CLEAR driving.png
    Pioneer Kuro - CLEAR driving.png
    60.7 KB · Views: 894
50Hz @100Hz : same picture displayed twice within 20ms. 10ms per frame compared to 16,7ms@60Hz, hence less sf and/or shorter (narrower) pulses. In order to maintain a decent light output, pulse length cannot be cut too much, hence the need to reduce the number of sf. Depending on picture mode, APL (how bright the frame needs to be) and motion, the number of sf and their length will differ.

If you could test 24p@96Hz, you would prolly see 4 groups of sf.

In this case, same picture is displayed all over again anyway, because it was static image with color gradient, although encoded as video at 24p, 25p, 50p and 60p

I wrote I would classify it as 50 Hz because whole subfield pattern repeats in 1/50 s. There are two groups, but number of pulses within each is different. However, flickering is reduced this way, so in that sense, it is a kind of 100 Hz (for marketing purpose)

Wondering how this is made in 96 Hz mode, because 600 isn't dividable by 96, so we can't have 1/96 or even 1/48 s interval with integer number of subfields - if subfield frequency is still 600 Hz
 
Last edited:
I wrote I would classify it as 50 Hz because whole subfield pattern repeats in 1/50 s. There are two groups, but number of pulses within each is different. However, flickering is reduced this way, so in that sense, it is a kind of 100 Hz (for marketing purpose)

2 groups of sf using a slightly different sequence instead of 2 exactly similar groups is imho a clever way to display 50@100Hz without killing the number of available gradations/colors. It prolly has many cons motion-wise thou.
 
Agreed.

In my opinion, it's irrelevant if we call it 100 or 50 Hz mode, although for me it's definitively 50, just with 4th, 5th, 7th and 9th subfield moved to the end of the group, to reduce flickering. Also, shadows that are displayed using only first three subfields will be displayed only 50 times per second (clever, because flickering on that parts usually won't be visible). It is, so to say, a hybrid 50/100 Hz mode

However, in this topic, my intention was actually to see how many gradations we have on this TV. And it turned out that this number is considerably lower than I thought, which was much lower than number claimed by the manufacturer at the first place ... However, I can believe that TV is capable of optimizing their usage, depending on various parameters like displayed video (dynamic range and number of actual colors), picture mode etc, because small number of gradations used per one refresh cycle is rarely visible (from normal viewing distance)
 
If you are in the mood for further experimentation : if you dig in AVSHD709 DVD/BR, you should find greyscale patterns at 10% APL increments.
 
I already checked those videos, but they were not intended for this kind of experiment, so I had to make my own to get what I wanted. Not a big problem - a short piece of delphi code to make a bmp + short AviSynth script + x264 encoder

However, I'm quite surprised that nobody expressed any doubts about my conclusion that during one refresh rate (1/50 or 1/60 s), we only have 10-15 actual gradations of each color (without taking dithering into account), while manufacturer specified "6144 equivalent" whatever that means

Another thing came to my mind after reading lengthy green blob threads. Take a look at attached two pictures. First one is image from the video, that was displayed at left half of the screen. It is uniform green gradient, with green level 0 at the left edge, rising to level 255 towards the middle of the screen. Second image is a photo of the screen, taken with DSLR using exposure 1/2000 s. It is one of images I used to make previously posted images of whole refresh cycle (by stitching many images together and rotating them 90 degree clockwise). Since shutter in the camera travels in vertical direction, bottom of the TV is recorded first, and top of TV about 1/400s later, so photo shows firing of one of subfields in the middle of the screen and phosphor decay towards the top. Below, we see decay from previous subfield. However, traces of decay should be straight vertical - and they are straight for red and blue. But for green, they are rounded (and not always the same way). Wondering if this is actually an explanation of green blob issue? It looks like TV is intentionally displaying higher intensity of green at the center than off-center, possibly to compensate for non-uniform light emission of the PDP caused by some reason (heat for instance). So, if it is compensating for it, it could also be over-compensating it, adding too much green around the middle of the screen ...
 

Attachments

  • GreenGradient.jpg
    GreenGradient.jpg
    24.6 KB · Views: 461
  • GreenGradientSubfield.jpg
    GreenGradientSubfield.jpg
    114.3 KB · Views: 430
Last edited:
However, I'm quite surprised that nobody expressed any doubts about my conclusion that during one refresh rate (1/50 or 1/60 s), we only have 10-15 actual gradations of each color (without taking dithering into account), while manufacturer specified "6144 equivalent" whatever that means

Hi,

I believe that the amount of native gradations is higher than that. Each subfield has a predefined weight, for example 1 32 64, and Panasonic should be able to fire any combination of these subfields during one field.

0(reset)
0,1 -> 1 weight
0,32 -> 32 weight
0,1,32 ->33 weight
0,64 ->64 weight
0,1,64 ->65 weight
0,32,64 ->96 weight
0,1,32,64 ->97 weight

Even special combinations should be possible for dark scenes like 1, 1, 1, but this example ,of course, limits maximum brightness to 3.

6144 value might come from the following equation: max brightness/first subfield brightness.
 
I already checked those videos, but they were not intended for this kind of experiment, so I had to make my own to get what I wanted. Not a big problem - a short piece of delphi code to make a bmp + short AviSynth script + x264 encoder

However, I'm quite surprised that nobody expressed any doubts about my conclusion that during one refresh rate (1/50 or 1/60 s), we only have 10-15 actual gradations of each color (without taking dithering into account), while manufacturer specified "6144 equivalent" whatever that means

Another thing came to my mind after reading lengthy green blob threads. Take a look at attached two pictures. First one is image from the video, that was displayed at left half of the screen. It is uniform green gradient, with green level 0 at the left edge, rising to level 255 towards the middle of the screen. Second image is a photo of the screen, taken with DSLR using exposure 1/2000 s. It is one of images I used to make previously posted images of whole refresh cycle (by stitching many images together and rotating them 90 degree clockwise). Since shutter in the camera travels in vertical direction, bottom of the TV is recorded first, and top of TV about 1/400s later, so photo shows firing of one of subfields in the middle of the screen and phosphor decay towards the top. Below, we see decay from previous subfield. However, traces of decay should be straight vertical - and they are straight for red and blue. But for green, they are rounded (and not always the same way). Wondering if this is actually an explanation of green blob issue? It looks like TV is intentionally displaying higher intensity of green at the center than off-center, possibly to compensate for non-uniform light emission of the PDP caused by some reason (heat for instance). So, if it is compensating for it, it could also be over-compensating it, adding too much green around the middle of the screen ...
Great information there. I was wondering what could Panasonic do to reduce the effect of the pushing a higher intensity of green on the screen? Is it a panel drive issue that could be sorted by a firmware update/eeprom update do you think?

This thread is like a fountain of information and Ive learnt quite alot more about PDP tech cheers gang. :thumbsup:
 
I must say I agree with Sampo in that I assumed that Panasonic plasmas use a derivative of the binary weighted system. But I don't own a Panny plasma so I can't do these tests to find out if it's true.

However, I do own a Pioneer 5090 and a year or so ago I did do some tests with a fast shutter camera to try and evaluate the CLEAR driving and I found that the colour palette per frame period was quite poor and this was most noticeable on skin tone tests and also smooth greyscale plots. Presumably the TV uses various temporal/spatial dither algorithms to fool our eyes into seeing a richer palette across numerous frames and this is at the expense of dither noise (something I think my eyes are very sensitive to?)

I did also loosely couple up an RF spectrum analyser (HP8566B) to the TV by means of a pickup antenna close to the screen and looked at the RFI emitted from the TV and this showed some interesting responses related to changes in drive mode and panel refresh rate when the analyser was set to zero span at the frequency of the main peak in the RF spectrum emanating from the TV.

I did wonder about testing with an opto diode to the analyser with the TV set to display a dynamic test pattern appearing on a single line on the TV but never got around to doing it...
 
Last edited:
Principles of PDP are usually explained with binary weighted system of subframes (length of pulses 1:2:4:8:16 ...), for easier understanding I suppose, but even 10 year old papers suggest that different schemes were actually used by various manufacturers at that time. So it's hard to believe that last generation of PDP is using binary weighted system, which has problems with false contouring and flickering

Close look to one of photos of the screen reveals that spatial dithering is obviously used, as you can see in attached photo. If there were no dithering, there wouldn't be different layers of pixel patterns, as TV was displaying uniform color gradient. So in this test, TV obviously wasn't able to display all 240 gradations of each color during one refresh cycle, but significantly less

I believe that in different situations TV can combine those subfields differently and possibly use more gradations. However, it's hard to believe that it would use all 240 gradations in some cases, and only 10-15 in some other cases (like here)

Maybe I'll make a more realistic test in the future
 

Attachments

  • Dithering2.jpg
    Dithering2.jpg
    95.7 KB · Views: 394
When I get to a thread like this one, I know I've gone too far in trying to buy a new telly.:D
 
When I get to a thread like this one, I know I've gone too far in trying to buy a new telly.:D

hehe
It could be worse, you could have get this far, trying to decypher stuff like that. :)
 
Last edited:
hehe
It could be worse, you could have get this far, trying to decypher stuff like that. :)

Cool stuff, albeit too complicated for me. Btw, I searched the term "green blob" and nothing came up ... I was certain they'd patent it.:D
 
Cool stuff, albeit too complicated for me. Btw, I searched the term "green blob" and nothing came up ... I was certain they'd patent it.:D

Ghostbusters already did.
ghostbusters.jpg
 

The latest video from AVForums

Is 4K Blu-ray Worth It?
Subscribe to our YouTube channel
Back
Top Bottom