Strange 'dithering' effect seen on DLP (1 Viewer)

[Lee Shankman]

Hello all

Today, as I was checking out the Toshiba DLP rear projection TV I want to buy. I understand the 'rainbow' effect caused by the color wheel, but what I was thinking about was the idea that a pixel can be only 'on' or 'off' in DLP, not partially on for fractional brightness. This led me to wonder how DLPs resolve different levels of brightness -- a smooth grayscale, for example. I suspected that they must use a dithering algorithm or perhaps a pulsating that is quicker than the framerate.

On close examination and comparison with LCD units nearby, I feel that LCD is far superior in its ability to resolve subtle differences in levels, especially in shadows. I saw annoying dithering in dark areas, and the same dithering in bright areas when panning across the frame. None of this artifacting was apparent on the LCD units. I am not wondering if I'll be happy with the DLP unit. What is actually the benefit of DLP? I don't like the rainbow effect, and I really don't like the dithering phenomenon I'm seeing. I know those DLP chips must be doing a heck of a lot of work, changing so quickly (it must have to change at least 3x per frame, right?), but it seems to me that the picture still needs some work.

Thoughts?

Thanks...

 

[Allan Jayne]

DLP's achieve gray scale by varing the amount of time each pixel's mirror is in the "on" position.

Still, there may be discrete steps in terms of microseconds.

DLP's advantage over LCD is a better contrast ratio, liquid crystals are not all that opaque in the "off" condition.

Video hints:
http://members.aol.com/ajaynejr/video.htm

 

[Philip Hamm]

I've never noticed nor have I ever heard of this "dithering" effect. My DLP projector looks stunning, guests can't believe how good it looks.

 

[Leo Kerr]

Think of it as 'pulse width modulation.'

In a well designed system, the mirror will never be fully 'on' or fully 'off,' and will 'flicker' at some obscenely high rate. I've heard the clock speed, but I can't remember it now; it's in excess of 16million flickers/30th second.

The reason why you don't want the mirrors to 'stick' on or off for full brightness or full black is for the stair-step phenomonon: the difference between mostly black and all black is tremendous, while the difference between mostly black and nearly black isn't anywhere near as obvious. If they can get mostly black close enough to all black, then you don't really have a big step down to all-black. The same holds true at the white end of the scale.

How the flickers are broken up at the extremes is different depending on the manufacturer's drive algorithms. A 50% gray done by holding the mirror on for half a cycle and then off for the other half is horrid, compared to flickering it much faster than that.

Leo

 

[ChrisWiggles]

Philip: then immediately stop reading this thread if you want to continue your happy viewing of DLP products as you are now.

Dithering is definitely viewable, and especially so in dark areas. Near-black shadow details are difficult for DLP to reproduce because of the inability to be "gray" but only approximate gray by flashing on and off quickly. If you look in dark areas of pictures especially, even still ones, by walking right next to the screen, you can see this artifact. It is usually not that visible when viewing. Another difficulty, is that MPEG compression contains TONS of mosquito noise, which is probably MORE what you are seeing, and you will see this at normal viewing distances if resolved. This is very apparent usually in things like gradiations like clouds, skies, etc, and is NOt the fault of DLP.

So be very careful to differentiate DLP dithering (which will can be seen even during STILL pictures, like paused frames, etc, and MPEG noise (which is FAR more prevalent).

If you were indeed seeing MPEG noise instead, then this is not the fault of the DLP in resolving this. It may not have been resolved as well with slower response times of the cells in an LCD, that tend to smooth out these source details.

 

[Lee Shankman]

Thanks, guys! Those are great answers, and they are what I suspected. I guess I had never considered that the mirrors could flip that fast. It seems to be somewhat slower than 16 million cycles per 1/30 sec., because I can't imagine we could notice the effect if it were that fast.

I am aware of the MPEG artifacting (why is it called "mosquito noise"? and, incidentally, does anyone other than me find it interesting that it is tolerated when the technology is so good? I think I will start a post on that!) and can say that what I noticed was definitely not that -- it was the dithering. I noticed significant dithering as images panned across the field -- I guess that gives your eye the ability to see the "trail" of on/off cycles.

 

[Philip Hamm]

How do LCD projectors handle this? Seems to me they'd need to do the same thing, would they not? As I understand it, an LCD cell is like a DLP mirror in that it has two states: on or off. Does the fact that there are three of them help mask this effect?Lee, you have to remember that DVD was introduced as a format for public consumtion, a finished product, in mid- 1997. That means that all the development for the system had to be complete and set in stone probably by mid-1996. I remember when DVD came out it -barely- worked. Old first gen DVD players (I still have one) were quite large with many chips inside and very advanced video chips for the time. This kind of MPEG artifact is a product primarily of bandwidth limitations imposed in 1996. (and the number of colors displayable by DVD, but that again is a product of bandwidth limitations in 1996).

I don't see these artifacts nearly as often when watching HDTV broadcasts, but they are still there occasionally. Newer standard, better handling of the data.

MPEG style encoded video is always going to have this effect I think.

 

[ChrisWiggles]

Philip: That's not the case, LCD pixels are variable, there is no on and off limitation like DLP mirrors.

 

[Allan Jayne]

Did you say steady state dithering where one pixel is always a little lighter and the one next to it always a little darker, or does each pixel rapidly change from lighter to darker as if to shimmer? The latter could be mosquito noise if the changes are 1/60'th a second apart corresponding with the incoming video frames.

Does non-moving subject matter (such as a seemingly solid color patch) have the dithering?

 

[ChrisWiggles]

You can differentiate the two, by pausing your video, and this eliminates any effects of mosquito noise that people might confuse with dithering, as they can seem similar. Even with paused frames, you may be able to see the DLP dithering, the best frames are dark ones with lots of near black, you will see attempts to create very-near black as dithering, EVEN with the frame paused, there will be a sort of dance of pixels going on. You can use this dithering to set your black level *extremely* accurately on DLP devices, BTW.