Mitsubishi 2010 3D DLPs

[cybereality]

http://www.engadget.com/2010/04/08/new- ... terious-s/" onclick="window.open(this.href);return false;

The cheapest model sounds very affordable, I might want to get one.

[DmitryKo]

It looks like the TV sets will not support HDMI 3D formats, so a Mitsubishi 3DC-1000 converted is required for Blu-ray 3D and PS3 stereo gaming. I wonder why Mitsubishi didn't include the functionality in the TV, are they trying to make some additional money by selling you say an $300 accessory box? I'm just making assumption though because the price for the converter box has not been revealed so far.

That specific thing aside, it really good news that Mitsubishi is not abandoning real projection TVs, unlike Sony, Samsung and JVC. If only they were present on my local market... sigh. I already have a SONY KDF-E42A11 and the real contrast ratio beats anything currently on the market including best plasma displays I've ever seen. I don't know how DLP TVs fare in this regard, but even in the current series, $1700 for a 73" 1080p screen is a terrific offer which is very hard to resist. Similar stereoscopic front projection setup would cost you at least three times as much, because you would need to buy dual 1080p projectors, which are not exactly cheap, and a silver screen. I remember when 70" RPTVs would cost around $7000, and I don't even want to talk about one of these professional 1080p stereoscopic projectors which start at like $30 K.

[cybereality]

At first I was upset that HDMI 1.4 wasn't supported but I think the checkerboard might be a better format at this point in time. With checkerboard it should be possible to use the Nvidia, IZ3D or DDD drivers. 3D BluRay will also be supported on PowerDVD 10. PS3 will be supported with the add on box which I heard would be around $100. And 1080P half-res is still more pixels than 720P full-res. So in terms of support I might rather go with checkerboard. The price doesn't hurt either.

[DmitryKo]

Well, Samsung C7000 series support PC formats like checkerboard, but they support HDMI 3D formats as well. If Mitsubishi can do a converter box, why couldn't they update the firmware to support HDMI 3D? It just doesn't make sense.

[Fredz]

And 1080P half-res is still more pixels than 720P full-res. So in terms of support I might rather go with checkerboard.

This article explains why 720p is better than 1080i : http://www.bluesky-web.com/numbers-mean-little.htm" onclick="window.open(this.href);return false;
I don't know if this analysis can be extended to 720p vs 1080checkerboard but it's a very interesting article anyway.

[cybereality]

Well, Samsung C7000 series support PC formats like checkerboard, but they support HDMI 3D formats as well. If Mitsubishi can do a converter box, why couldn't they update the firmware to support HDMI 3D? It just doesn't make sense.

I agree. It is a strange move. Maybe they want to make the extra money off the converter box?

[DmitryKo]

I don't know if this analysis can be extended to 720p vs 1080checkerboard

It can't. Checkerboard picture is essentially 1080p60 per eye, so it's free of any interlacing artefacts; effects of lossy video compression such as low efficiency do not apply as well.

As a side note, EBU has been recommending 720p50 over 1080i50 since at least 2003, in a view that the future is 1080p50 production and broadcasting; most European HDTV channels have ignored this and went for 1080i broadcasting though. This is however less relevant to movies, because they are effectively scanned at 25 PsF (progressively segmented frames) for broadcasting.

Maybe they want to make the extra money off the converter box?

Well, if they want more money, include HDMI 3D and raise the price accordingly. A 70" screen only costs $1700 and that can only buy you a 50-52" plasma or LCD TV, so another $50-100 for HDMI 3D capabilities are not going to burn the pocket.

[Fredz]

I don't know if this analysis can be extended to 720p vs 1080checkerboard

It can't. Checkerboard picture is essentially 1080p60 per eye, so it's free of any interlacing artefacts; effects of lossy video compression such as low efficiency do not apply as well.

Well, it seems you didn't really read the article...

It's wrong to say checkerboard is 1080p60 per eye, since only half the pixels are displayed on each refresh for each eye. So, the pixel rate is quite equivalent to 1080i60 and 720p60, at about 60M pixels per second as explained in the article.

Since the lines are not complete for each eye, you can't consider the vertical resolution to be 1080 lines either. It still has to be demonstrated if this disposition of pixels does give a better or lower perceived vertical resolution than 720p.

And as mentionned in the study cited in the article, resolution is one of the smaller factors in the determination of the subjective quality of a television picture (contrast : 64%, resolution : 21%, brightness : 15%). Checkerboard suffers from the same problem than interlaced mode, since it halves the perceived brightness and contrast for each eye compared to a progressive mode.

[cybereality]

Checkboard DLP refreshes at 120Hz even though the input is 60Hz. The source has both views encoded into one image coming in at 60Hz. The DLP TV then breaks this image into two but doubles the refresh rate for the glasses. So you still get the same number of pixels per second as a full 1080P source at 60Hz.

[Fredz]

Sure, but you compare 1080checkerboard at 120Hz with 1080p at 60Hz, which is not the point since non checkerboard screens refresh at 120Hz too. At this frequency, pixel rate for each eye is still equivalent wether you use 1080checkerboard or 720p and it's the double for 1080p, at 124M pixels per second.

[DmitryKo]

It's wrong to say checkerboard is 1080p60 per eye, since only half the pixels are displayed on each refresh for each eye.

Wrong. The picture is up-sampled to full 1080p60 per each eye using the bilinear interpolation. If you think the resolution drops as much as in interlacing, feel free to experiment with Photoshop masks and blur filters.

If marketing materials say that checkerboard pattern is related to wobulation in DLP displays, this can't be true since the wobulation is horizontal only, and wobulating in a checkerboard pattern requires diagonal wobulation, that is both vertical and horizontal directions. Truely native format for horizontal wobulation would be column interleaved format.

Checkerboard suffers from the same problem than interlaced mode, since it halves the perceived brightness and contrast for each eye compared to a progressive mode.

Stereo images are not sampled in checkerboard mode, they are sampled at full resolution progressive 1080p per eye. Checkerboard is only a transmission method to the display, where checkerboard image is decoded to two full resolution pictures.

[pixel67]

I previously used my Sammy HLT 60" 1080p HDTV as my primary 3D display and it used checkerboard. I was quite happy with its performance and quality however i have to disagree with those that state there aren't any interlacing artifacts associated with Checkerboard. While overall the 3D quality is very crisp, i noticed issues displaying the fine details of objects that consist of only a few pixels. Examples would be the chain link fence that I know we have all seen in HL2, gamer tag names, trees in the distance, etc... It wasn't horrible by any means, but there nonetheless and made those objects look like stair steps.

I now own a 720P DLP Projector that does full page flip and while the aliasing from 720p vs 1080p is definitely noticeable, a healthy dose of AA improves image quality significantly but with a MUCH larger screen (which was my sole reason for going with a projector). I haven't run HL2 to compare the fence but again, this wasn't a huge problem in the first place. It is a real shame that the HDTV industry is moving away from DLP because it is clearly the best when it comes to 3D. Near zero, if not zero ghosting is inherent in the technology and just can't be beat. Now if I could just get my hands on a 1080P or higher projector that does full page flip then I would be in 3D heaven!

Long live DLP!!

[mickeyjaw]

If marketing materials say that checkerboard pattern is related to wobulation in DLP displays, this can't be true since the wobulation is horizontal only

No, if the DMD consists of a sparse checkerboard array, then the wobulation can be either horizontal or vertical and still interleave the two images in a checkerboard format, but you end up with raggedy edges.

[DmitryKo]

i noticed issues displaying the fine details of objects that consist of only a few pixels. Examples would be the chain link fence that I know we have all seen in HL2, gamer tag names, trees in the distance, etc... It wasn't horrible by any means, but there nonetheless and made those objects look like stair steps

The fact that you've seen these artefacts in stereo 3D does not mean it's related to checkerboard pattern. It could be antialiasing, texture fitering, LOD, or Z-buffer issues within the game.

if the DMD consists of a sparse checkerboard array, then the wobulation can be either horizontal or vertical and still interleave

Interesting.

http://www.informationdisplay.org/issue ... 5/art5.pdf" onclick="window.open(this.href);return false;
http://whatsonhdtv.blogspot.com/2005/07 ... -no-3.html" onclick="window.open(this.href);return false;

So in TI's "offset-diamond pixel layout", the mirrors are rotated 45 degrees and are arranged in a pattern where each mirror matches only pixels situated in a checkerboard pattern in the standard rectangular pixel grid.

So the mirrors

Code: Select all

 /\/\/\
/\/\/\/\
\/\/\/\/
 \/\/\/

only match the "black" pixels

 *O*O*
 O*O*O*
 *O*O*

and horizontal wobulation makes them match "white" pixels.
I need to think about it.

[pixel67]

i noticed issues displaying the fine details of objects that consist of only a few pixels. Examples would be the chain link fence that I know we have all seen in HL2, gamer tag names, trees in the distance, etc... It wasn't horrible by any means, but there nonetheless and made those objects look like stair steps

The fact that you've seen these artefacts in stereo 3D does not mean it's related to checkerboard pattern. It could be antialiasing, texture fitering, LOD, or Z-buffer issues within the game.

It wasn't just HL2. It was in every game where there were objects that are just a few pixels or, worse case, a single pixel in width. I tried all kinds of different filtering and antialiasing and it never got any better. This makes sense since each eye doesn't see the full image but only the checkerboard sample. In single pixel objects it was clearly visible when they were missing. Now, if the sampling of the image actually occurred 4 times (240hz) and the sample pattern were inverted every 120hz then the missing spatial information would truly be replaced in the time domain. At 120hz and the source only being sampled once for each eye, there is definite spatial information loss in moving images.

[wnielsenbb]

If you had 120Hz input you could do full 1080p page flipping. The checkerboard is a compromise for getting two different images (left and right eye) in one frame of a 60Hz input. You are basically taking a 1080p image, compressing it to 720p sending it and then taking that 720p image and displaying it at 1080p and you should lose about that much information. It won't be horrible, but it won't be as clean as straight 1080p.

[GoldChain]

If you think the resolution drops as much as in interlacing, feel free to experiment with Photoshop masks and blur filters.

I took your advice. I took a photo, cropped it to 1920x1080, then used the mask to make a selection (of the checkerboard pattern), then did a simple filter matrix that made the checkerboard pixels the average of the 4 pixels around it (i.e. 4 of the surrounding pixels that weren't also part of the checkerboard). At a 1:1 pixel size (i.e. image zoom of 100%), I couldn't see any difference between the original 1920x1080 crop and the checkerboard-interpolated version. Even at 4:1 (i.e. 400%) there was only slight indication that the checkerboard-interpolation was used. Very impressive.

At first, upon learning about the checkerboard format, I was bummed because 1/2 of the resolution is 1/2 of the resolution (i.e. in theory it's no different than interlacing stereo views together). However, using the checkerboard and averaging the 4 pixels around to fill in the "holes" of the board seems to be far superior than (say) interlacing stereo images, as trying to use interpolation (averaging) to fill in every other line wouldn't give as good of a representation of the original.

Now, I just need to find the answer to the question: is the stereo checkerboard frame (a full 1080p frame) sent 60 times a sec? or 30?

GC

EDIT: I removed the original TIFs of the processed and difference images and posted below.

[cybereality]

Ok cool, that is very good news.

Also, I believe the input to the DLP TV is at full 1080P@60Hz then the TV takes that checkerboard image, does its magic, and displays at 120Hz (60Hz/eye).

Oh, and the forum doesn't support tifs. Pngs should work though.

[GoldChain]

Ok cool, that is very good news.

Also, I believe the input to the DLP TV is at full 1080P@60Hz then the TV takes that checkerboard image, does its magic, and displays at 120Hz (60Hz/eye).

Oh, and the forum doesn't support tifs. Pngs should work though.

Yah........ it appears the TIFs didn't come thru right =) I wanted to use a lossless image format so that the checkerboard image could be compared visual (even at a pixel level) compared to the original. Same with the difference images. Maintaining a pixel-level accuracy is a must for the true results to be seen. I always believed that PNG was lossy, but I guess that's an option you can control. So I've resaved as PNG and am reposting (I won't repost the original as it was in the above post).

GC

NOTE: The difference image below is NOT just a black image. You need to download/open it and zoom in to the pixel level to really see what this file is telling you is the difference between the original and checkerboard.

[GoldChain]

Just for another comparison, I did this same process with another image. I'm posting for consideration.

GC

[DmitryKo]

Well, it does prove that checkerboard works very well as a transmission format for non DLP displays.

I may even reconsider my concerns that recent 3D TVs like Samsung C7000/C8000 series only support checkerboard 1080p format and similar "half-resolution" formats for the PC connection, and do not support any full-resolution 60 Hz format.

[cybereality]

Well if you open both images in different tabs and then quickly switch between them you can see a difference. It is very small but its there. The filtered image looks a bit softer than the original, as you would expect. This is most apparent on fine details like the general's hair, for example. Even so, it doesn't look anywhere close to "half-resolution" and is seriously convincing me I need one of these Mitsu displays. I am probably going to get last years model cause its cheaper though. So I guess checkerboard is a win.

[GoldChain]

You're correct, that if you switch between them you *can* detect the difference. Given that it averages pixels around the pixel that was left out, it would make sense that some places this would be virtually no different (i.e. a pixel in the middle of a gradient in both directions would be recomputed and almost identical (if not the same) as the original) but in some places would more obvious, like along high-contrast edges. Either way, as DmitryKo pointed out, this interpolation has the same effect as a blur operation, so at the end of the day, the checkerboard/interpolated image won't (can't ever be) as sharp as the original.

Having said that, these frame are in motion, so the difference we see by flipping between the original and modified images becomes lost in a sea of motion, putting me squarely in the boat with cybereality, which is seriously checking out last years Mits DLP models. Given the price you can pickup a 73" Mits DLP, and given that I already have the PS3, I'm thinking I can live with the checkerboard 3D to slide into major 3D viewage

[GoldChain]

Well, it does prove that checkerboard works very well as a transmission format for non DLP displays.

I'm confused. I thought that checkerboard was what was required primarily only by DLP displays.? (i.e which is why you would wedge that $100 Mits adapter between a 3D Blu-Ray/PS3 and a Mits or Samsung DLP set)?

GC

[DmitryKo]

The extra softness can probably be worked around by using linear interpolation instead of bilinear, i.e. only avarage among any 2 neighbouring pixels... not to mention the infamous "unsharp mask" filter which is the core of all the stupid "image enhancers" featured in the consumer TV's processing engines.

[wnielsenbb]

GC, they came up with the checkerboard method as a way of transmitting two seperate images at once at 60Hz. Checkerboard leaves you with a much better way to recreate the image than trying to compress vertically or horizontally. It has nothing to do with the display method, just the transimssion method.
Current 3D does this by sending the images one after the other, but bumping the speed up to 120Hz. 60 Hz for each eye.
The only place I think checkerboard would be a problem is with chat text while gaming, and it would probably not be any worse than playing your game at a different resolution than the display.

I think the new HDMI 1.3 3D spec uses some kind of funky left eye and right difference method or something. Haven't researched that yet.

[tritosine5G]

One problem with checkerboard seems to be that it wont give twice framerate in games right?

EDIT: I mean the GFX cards have twice less job to do in theory, so +100% framerate . Ok ok Im dreaming.

[wnielsenbb]

No, the tv's cannot accept a 120Hz input for anything, at least not at 1080.

[DmitryKo]

Checkerboard leaves you with a much better way to recreate the image than trying to compress vertically or horizontally. It has nothing to do with the display method, just the transimssion method.

Well, as mikejaw pointed above, it seems like checkerboard actually IS the display method for recent DLP 1080p DMD imagers, which use rotated mirrors arranged in a checkerboard position and wobulate horisontally to present the full-resoltion picture. So in DLP Tvs, checkerboard image is actually a quick hack around their hardware setup to enable 120 Hz stereo 3D at half resolution.

However checkerboard has since been established as a transmission format for Samsung plasmas and LED TVs, whose display engines are do not use the checkerboard pattern unlike DLP TVs, and similar format that use "quincunx sampling" have been patented by Sensio and RealD.

I think the new HDMI 3D spec uses some kind of funky left eye and right difference method or something

You confuse it with MPEG-4 AVC extension called Multiview Video Coding which defines the Stereo High profile used in the Blu-ray 3D specification. This is not related to the video link format defined in HDMI 3D spec, which uses either full-resolution "frame packing" and various half-resolution formats like side-by-side and top/bottom.

[spartanshu]

I'm curious what Sony's announcement about the 720p resolution for PS3 3D games means to current and prior mits DLP tv owners. From what I can tell, the 3DC-1000 adapter will convert top/down & side by side 3D formats and convert them to checkerboard at 1080p. Will this mean that mits DLP owners won't be able to play PS3 games in 3d?

[cybereality]

I'm curious what Sony's announcement about the 720p resolution for PS3 3D games means to current and prior mits DLP tv owners. From what I can tell, the 3DC-1000 adapter will convert top/down & side by side 3D formats and convert them to checkerboard at 1080p. Will this mean that mits DLP owners won't be able to play PS3 games in 3d?

The 720P/60 3D mode is part of the required spec for HDMI 1.4 3D so I am almost certain Mitsubishi would have to support it if they wanted their converter to be complaint. The converter would then need to upscale the image to 1080P before doing the checkerboarding, but that should be simple enough.

[GoldChain]

No, the tv's cannot accept a 120Hz input for anything, at least not at 1080.

So what's the point of a 1080p 120Hz display if it cannot accept 1080p @ 120Hz input? I mean seriously. I've been skeptical of the 120Hz (and now 240Hz) displays being more marketing than technology. This only bolsters that belief. I mean, for 15+ years, "digital displays" (i.e. computer monitors/CRTs) that supported 120Hz refresh could actually handle a signal that refreshed 120 times per second!!! No gimmicks, no BS. I'm sick of this age of "marketecture" where all info is simply to sell product and nothing to do with true performance.

</rant>

Sorry for dumping on you guys. I just get fed up with this crap sometimes.

GC

[DmitryKo]

for 15+ years, "digital displays" (i.e. computer monitors/CRTs) that supported 120Hz refresh could actually handle a signal that refreshed 120 times per second

Not at higher-end resolutions like 1600x1200 or 1920x1080.

[GoldChain]

Not at higher-end resolutions like 1600x1200 or 1920x1080.

My point was, when you bought a monitor, it's specs would tell you EXACTLY what resolution/refresh rates were supported. The fact that people are out there in stores selling displays labeled "1080p 120Hz" clearly implies 120 distinct frames of 1920x1080 video. If the display cannot accept a signal that is 1920x1080@120Hz, then the manufacturer's labeling of the product is misleading at best. As I said in my previous post, what's the point of 120Hz if you can only accept a 60Hz signal? To display each frame twice? What's the point of that? (now in the case of stereo, if the display has the "smarts" to take a 60Hz "checkerboard" signal and split it into 120Hz L/R frames, then there's SOME validity to the 120Hz labeling). However, since 99.9999% of people are NOT doing 3D on their 120/240Hz sets, it's a complete waste..... i.e. more marketing to get people who buy sets based on specs more than image quality.....

(for what it's worth, I *do* understand the value of taking a 24fps source video and playing at 120fps so that there's no need for 3:2 pulldown, but unless the display itself is doing the 24->120fps conversion, you'd still need the 3:2 pulldown, as the display itself isn't capable of accepting 120Hz input (i.e. if the player were doing the conversion from 24->120fps))

GC

[DmitryKo]

what's the point of 120Hz if you can only accept a 60Hz signal? To display each frame twice?

The point is, the DLP displays actually support 120 Hz, though they only accept a half-resolution signal transmitted over 1080p60, meaning you can use your 10-year old computer and a DVI-HDMI cable to actually see stereo 3D.

If they went on and implement full 120 Hz input at 1080p, then far less people would able to use the feature, since new it would require a HDMI 1.3 transmission which is not compatible with DVI anymore.

FYI, even the latest 3DTVs do not allow 120 Hz input in 1080p mode.

if the display has the "smarts" to take a 60Hz "checkerboard" signal and split it into 120Hz L/R frames, then there's SOME validity to the 120Hz labeling

I don't think anyone labels DLP 3DTVs as "120 Hz", they label them "3D ready".

The fact that people are out there in stores selling displays labeled "1080p 120Hz" clearly implies 120 distinct frames of 1920x1080 video

No, it implies some stupid and unnatural "motion interpolation" processing.

http://en.wikipedia.org/wiki/File:Motio ... xample.jpg" onclick="window.open(this.href);return false;
(slow motion)

[wnielsenbb]

my nec 21" CRT, which cost $1400 back then, advertised 120Hz, but it only did that at 640x480. Not to mention the fact that it could only diaplay a 19" image. Marketing crap has always been the same. The more bogus of the new stuff is that some of the sets advertised at 240Hz are really only 120Hz, but they flash the back light at 120Hz so they call that 240Hz.
Refreshing at 120Hz even without the input is still good at smooting the image. Some of these sets do frame interpolation to guess at an image between the two frames. If they do it right some people like it.

[GoldChain]

my nec 21" CRT, which cost $1400 back then, advertised 120Hz, but it only did that at 640x480. Not to mention the fact that it could only diaplay a 19" image. Marketing crap has always been the same. The more bogus of the new stuff is that some of the sets advertised at 240Hz are really only 120Hz, but they flash the back light at 120Hz so they call that 240Hz.
Refreshing at 120Hz even without the input is still good at smooting the image. Some of these sets do frame interpolation to guess at an image between the two frames. If they do it right some people like it.

You're correct that there's always been a marketing spin on products. Fair enough. To my point earlier, though, CRTs never just stated a refresh rate without coupling that with a resolution. By the specs on a monitor you, as an informed consumer, knew exactly what a display was capable of. Now with 1080p 120Hz, it's a mixed bag as the 120Hz frame-interpolation is a bunch of crap, some TVs saying 120Hz only accept 60Hz input, etc. Without MOUNDS of research, it's hard to know exactly what you're buying (as opposed to simply looking at a single page spec sheet for a CRT monitor). For something like scrolling text, I can see how an interpolation might work well, but anything organic (like the horse galloping as shown in that example link in a previous post), it's udder nonsense to think that any chip would have the intelligence to really "know" what position a horses leg, leaves on a tree swaying in the wind, or a moving hockey stick, etc would be in the "in-between" frames.

That's been one of my biggest pet peeves with the whole 120/240Hz sets is the notion that they can take 24/30fps content and magically make it into 120Hz. I love the commercials where they're like "With 120Hz displays, it makes fast motion sports crisper without the blur." Nonsense! The source is still 30fps! And the "interpolation" they might need to employ to generate the "in-between" frames would only serve to muddle/blur the motion as it interpolates! Talk about total "marketecture".

This is all good discussion. I appreciate all of the comments, and I sincerely hope that folks don't take my posts as being completely negative or argumentative. It's actually nice to discuss and go back and forth with people who have more knowledge and understanding about this stuff.

GC

[wnielsenbb]

Actually 24fps content is where 120Hz display is really desired, since it can show every frame 5 times, where with a 60Hz display it shows one frame 2 times, and the next frame 3 times, the next 2.. etc, thus giving motion judder (or whatever the term is) on panning scenes. When they started making 120Hz display's there was no bluray player that could output 120Hz so it probably didn't make any sense to accept a non standard input.

[cybereality]

The fact of the matter is that there is nothing to be confused about. Its all bogus. There are no HDTVs that I know of that accept higher than 60Hz. And even if there was there is no content outside a PC that will support those frame rates. Broadcast is 60hz max, BluRay is 24hz, etc. The only place where 120Hz actually plays a factor is the Nvidia 3D Vision ready LCDs which actually do refresh at 120Hz. But as far as I know, that is it. So we are talking about a half dozen displays at most. I hate marketing BS. Over in the Nvidia forums I swear every day there is at least one person with a "120/240Hz" HDTV wondering if 3D Vision will work. So much confusion for nothing. And I've seen this "motion interpolation" first hand and its garbage. Probably about as good as the 3D conversion on the new Samsung sets.

[wnielsenbb]

I assumed the new samsung 3D tv's had 120Hz input. That will suck if they don't work with nVidia 3D vision. I can't find anything on searches.