Microsoft to Use DVDs & HTPCs for High Definition ...

[James Morrow]

Following recent improvements in the performance of WM9 with high definition 1920 by 1080 pixels, 24 frames per second (1080p24) at low bitrates (below 10Mbps) - see the H.264/WM9? posting for details - Microsoft (you may have heard of them) are now pushing to have HD content agreed as a DVD standard, and for it to be released on a large scale at 8Mbps on standard DVDs within months, using DVD9/DVD18. An 8.5GB DVD9 would support around two hours of 8Mbps HD Video and compressed DD/DTS audio, whilst a 17GB DVD18 would support nearly four hours. A standard DVD player can't support the required data-rate, but modern PCs with multi-speed drives can, so expect a push for "high-definition compatible" home theatre PCs.

Apart from seriously denting the rationale for typical consumers behind next generation discs such as Blu-ray and HD-DVD, whilst 8Mbps WM9 appears to work very well with man-made objects, natural objects such as trees look artificial, colour gradients are obviously stepped, there’s a “cogging” effect on trees and similar as the camera moves through the scene – a bit like insufficiently accurate sub-pixel processing, etc.. We need 24Mbps high definition demos using modern codecs soon if the mass-market isn't going to be waylaid by limited quality 8Mbps HD on DVDs. :shock:

Of course, using 8Mbps WM9 or similar on BD as an extended play mode (around ten hours of HD recording on a 50GB disc), with 16, 24 and 32Mbps available too, would be a different matter. Also, MPeg4 will have a hard time on its hands if WM9 becomes established on PCs as the de facto standard codec for high definition. Then again, it does do an exceptionally good job at 8Mbps - but there's room for improvement - which higher rates such as 24Mbps would hopefully allow. If Microsoft would demonstrate at 24Mbps as well we could see what should be available on BD, possibly further strengthening Microsoft's case, of course, but the main objective would be to avoid developing a high definition market at 8Mbps on DVD shortly followed by higher quality 24Mbps (or whatever) on BD. :?

[phloyd]

I don't feel that I can vote without seeing the examples.

In any case, don't think of WM9 as a recording codec. Anyone who has tried to compress HD on their 3+ GHz P4 wll no that this is far from a real time codec. Integration onto ASIC might help some, but I believe it would be prohibitively expensive to encode WM9 in real time well.

So, we are talking about stamped DVD's, encoded at 8Mbps in a lab by the gurus with the best tweaks for the encoder, taking a lot of CPU cycles...

If they don't look as good as the D5 to the untrained eye, then no way. (Sadly, this might appeal to the studios, who seem to hate the idea that anything they give us is the perfect image.)

Once the high capacity discs are out, the issue becomes moot since MPEG2 is good enough. (though HD-DVD, which is a little thin on capacity, may use MPEG4.etc. to get the run time).

Cheers!
DAve.

[Blu-Wave]

Phloyd,

you can download samples of 1080p24 and 720p24 using WM9 at around 8Mbps from Microsoft - see the H.264/WM9? posting for details. Note that the 720p24 samples need 1280 by 1024 resolution to work best and the 1080p24 versions 1920 by 1440. Sony and Apple have 23" TFTs suitable for displaying 1080p24, whilst if you've a bit more money, Samsung have a 54" widescreen TFT that can cope with 1920 by 1080 p60, along with a 57" widescreen TFT that goes up to 1920 by 1080 p100 - but they're not exactly cheap! :?

Note that as each sample is around 100MB you really need a broadband connection in order to download them. For me, the quality is very good, considering the bitrate - and very stable. However, although man-made objects look very good, natural object (trees, bushes, etc.) rendition was very unconvincing - talking from a high definition and film (and reality) based viewpoint. Compared to DVD it probably isn't that bad, but compared to film and uncompressed HD it is terrible in this regard. I haven't seen WM9 or MPeg4 at rates such as 20Mbps or higher, but I would hope that the obvious failings of 8Mbps are addressed. :?

[phloyd]

I have seen some of the Microsoft examples.

We can display them natively (ie, 1280x720 and 1920x1080) here and sometimes use them as test material.

In general, I think that they look pretty decent. Though I haven't really scrutenized them.

Like you, I think that I want to see the examples of high bandwidth encoding to ensure that the problems that exist are no longer an issue with higher bandwidth.

When I have tried using the encoder to make my own high bandwidth streams, I have found that playback is not smooth with the higher data rate. But I am not sure if that is due to the bandwidth needed to move the information or if the decoder has stream bandwidth issues...

I have also noticed that if you try to encode slowly changing greys, you get banding.

Cheers!
DAve.

[Blu-Wave]

There's plenty of banding on the 8Mbps WM9 examples - if you look at the dolphin demo, for example, you can see plenty of contours in the water, along with blockiness - although as the blocks appear smaller at HD than on a normal DVD, and as they're fainter than MPeg2's efforts, they're probably less objectionable ...

On the subject of (en)coding and decoding video, DVD players finally came out in 1997, after a bit of a format war of words. DVD recorders didn’t start to arrive until five years later. DVDs are based around the Mpeg2 codec which, like most high compression codecs, is asymmetric – much more processing power is required for encoding than decoding. A DVD player only has to decode the Mpeg2 video – a relatively easy task. In contrast, a DVD recorder has to be able to digitise and encode analogue video, and Mpeg2 encoding is much more computationally intensive than decoding. Also, when studios encode films they have experts who tweak the compression parameters to maximize the quality of the video. An expert wouldn’t fit in a sleek modern DVD recorder – plus you have to feed them, etc.. – so the machine has to tune its parameters automatically. As a result, the quality available from early recorders in one hour mode was little better than that of many two hour films, although companies such as Phillips seem to be getting better results these days. :?

In high definition, with six times the amount of data, Mpeg2 encoding again becomes expensive, and more modern codecs such as WM9 and Mpeg4 require even more processing power. Luckily, typical high definition machines will not have to encode high definition, as it will be either playing back pre-recorded firms or decoding encoded video received from broadcast, cable or satellite. The vast majority of current commercial high definition material is in Mpeg2 format. If an early mass-market BD machine could encode video, it would in preference choose Mpeg4 or WM9 or similar, but it is a slave to the encoders used by those who make pre-recorded material – such as the film industry. :shock:

High definition 1080p24 camcorders are probably several years away, and whilst Microsoft can employ experts to carefully tune its very low bitrate 8Mbps high definition video, automatic encoders don’t work well when pushed to their limit, so expect them to operate at at least 16Mbps and probably 24Mbps in practical high definition camcorders. Whilst a DVD9 could hold around two hours of material at 8Mbps, only forty minutes or so would be available at 24Mbps – but that would be no problem for home movies… The only problem then for portable use is the size of the disc. Higher bitrate encoding such as 24Mbps or even 32Mbps with modern codecs improves the potential quality available in the home. As modern codecs are capable of significantly better picture quality than Mpeg2, and at even lower bitrates, and as BD can easily handle to bitrates involved, Sony etc. would do well to demonstrate modern codecs at their optimum bitrates – much as Microsoft is demonstrating WM9 at a very low bitrate.

… and if the market does go for high definition at 8Mbps on DVD9, a more compact all-purpose 12 or 24GB high-density minidisc for pre-recorded films, music concerts, high definition camcorders, standard definition timeshifting, audio only, etc. might make sense…

[phloyd]

Speaking of Hi density 'minidisc", you are no doubt aware of the HDMD which Sony showed at CES (and no doubt other places).

I also saw mention of a 3"/8cm version of the PDD at NAB. I forget the expected capacity - I recall it being a little less than I expected - perhaps less than 10GB.

But now we are getting way off topic... did anyone vote?

Cheers!
DAve.

[Blu-Wave]

Yes, Hi-MD is a single-layer, single-sided development of conventional 780nm-based MD technology, using a technique called magnetic super-resolution [MSR] and more advanced coding and reading techniques to push the capacity up from MD's 140MB to 1GB. Because it uses the same optical head as MD, it is possible to fit the new format into compact walkmen such as the five models recently launched.

High Density Minidisc, in contrast, uses a 405nm blue-violet laser to achieve 6GB per layer - allowing 12GB dual-layer, 24GB quad-layer and 48GB quad-layer double-sided products. Unfortunately, optical heads for blue lasers are currently a lot bigger than the heads used in CD/MD walkmen - which have had twenty years to develop, so HDMD walkmen are unlikely in the short term - although the heads are small enough to allow a separates unit for use in a home Hi-Fi/AV setup.

I haven't seen details of the 3" PDD you mention, but on a related topic, Sony and Sharp talked of a 2" 1GB MO drive almost four years ago. Definitely something to look into... Of course, the current Universal Media Disc [UMD] specification is based upon dual-layer DVD technology and achieves 1.8GB on a 6cm disc, so Sony are looking at a lot of options.

[James Morrow]

Several people have voted Phloyd; but you're not one of them ops:

[Blu-Wave]

No details of the 3"/8cm PDD yet - if it was single layer, single-sided using 23GB technology it would be no more than 10GB (with 27GB, 12) - though 20GB dual-layer would be a bit more useful. Still a little big for portables though - Sony invested a lot of research in coming up with the smaller Minidisc form factor, and UMD is slightly smaller again at 6cm.

Talking of 3" PDDs, Phillips has been demonstrating single-layer, single-sided 1GB 3cm discs using Blu-ray technology for quite some time - although I would have thought that they could have got around 1.2GB using 23GB technology (1.5GB with 27GB technology), where quad-layer products could have a similar capacity to current DVDs but in a disc just over an inch across.

[phloyd]

Originally Posted by James Morrow

Several people have voted Phloyd; but you're not one of them ops:

Indeed.

I said in my first post that without examples of what 24Mbps looks like, I don't feel that I can vote.

Can I just vote against the 8Mbps one?

Ok, I'll run with the crowd and say "give me 24Mbps for all the codec's"...!

Cheers!
DAve.

[phloyd]

Originally Posted by Blu-Wave

Phillips has been demonstrating single-layer, single-sided 1GB 3cm discs using Blu-ray technology for quite some time

At 3cm they are getting a bit small really - I guess it could be cool for portable applications.

12cm is good for home but sucks for portable.

8cm is pretty reasonable for portable...?

Cheers!
DAve.

[Blu-Wave]

Yes, 3cm seems a bit small to me too - but I guess it might find its niche. 8cm also doesn't seem that big to me but modern minidisc players are smaller than this and solid-state portables are usually even smaller - and as the market is used to things getting smaller or staying the same size but becoming more capable, a bigger portable might not go down well. After all, even dual-layer HDMD would offer 12GB, which for most applications - even home high definition camcorders - is quite enough...

Talking of Phloyd, a certain group of roughly that name were selling well when they were called in to see their managers and be congratulated. However, they were not so happy with the introductory question: "ok, so which one of you is Pink?"

[phloyd]

I guess that I was specifically thinking camcorder.

12cm is too big for camcorder, IMHO. Maybe 8cm is too big too...? Well, maybe for the smallest camcorder. But then again, a fairly complex camcorder becomes hard to manage if it is too small...

WRT the Floyd, you gotta wonder if they would have been as popular if they were called Anderson Council. 8)

Cheers!
DAve.

PS. They were all pink...

[Blu-Wave]

On the camcorder front, there are a few based on 1.4GB 8cm DVDs - and nobody's complained about their size yet in the reviews I've seen, but as the capacity using blue lasers is so much higher, they could afford to go for the caddied (minidisc form factor) HDMD and still have a useful high definition camcorder - using MPeg4/WM9 at its "optimum" bitrate even a 6GB single-layer disc is enough for nearly an hour of 1080p24 high definition video.

... so if you wanted to spend an afternoon listening to the Council, would you be going for a Counciling session?

[phloyd]

Using WM9/MPEG4 is still dodgy in the "real time encoder" space.

HDV uses 25 Mbps or so, and should offer 1440x1080 at 24p/30p. This is more of a now thing than MPEG4 (I have a distrust of WM9 since I have an inherent distrust of Microsoft). since JVC is effectively akready doing this at 1280x720...

None the less, you would get probably 30 mniutes or so in 6GB, which is not bad. Dual layer would get us to more reasonable record times...

Regardless, I am sure that there will be an XDCAM with HDV variant as well as the smal;ler form factor XDCAM at some stage, since Sony seem to be very gung ho about XDCAM...

Cheers!
DAve.

PS. I guess I have had a lot of Counciling in my time

[Blu-Wave]

Michael Rowe is being sought in connection with attempts to pass off red laser products as capable of high quality, high definition video. Practices include pushing 1280 by 720 pixel material as high rather than intermediate definition material when high definition (as defined by Sony in the 1960s) requires around 2,000 by 1,000 pixels (currently set at 1920 by 1080 pixels), advertising scaled 1440 by 816 pixel material as full 1920 by 1080 pixel material, so that the datarates achieved by the WMV codec are not necessarily fully representative of the quality that could be expected of full 1920 by 1080 pixel material at those rates (i.e. rates sustainable on a red-laser product). :?

Of course, Mr Rowe is not the only one to exploit the confusion over what is high definition and the current lack of full resolution displays to demonstrate the difference, and WMV may provide superb performance at the data rates allowed by Blu-ray (24 to 28Mbps), but if people buy into 720p displays claiming to be "high definition" they will have good reason to be angry when they can't display pre-recorded 1080p films at full resolution on their expensive screens. :x

[Rob]

Are there any lcd/plasma displays that have full hi def resolutrion? just that I've seen a Panasonic 26"/32" lcd that claims hi def but is only 1280 by 768!? why's that :?:

http://www.panasonic.co.uk/lcd-tv/tx26lxd1/index.htm

[Blu-Wave]

Obviously 720p capable displays are easier to make than 1080p ones, so can be sold at a lower price. As long as people keep on calling anything above PAL/NTSC resolution "high definition" sales teams will happily call everything they can high definition if it helps to sell product. I think it would make more sense to call 720p (and all the other resolutions above 720p but below 1080p) intermediate resolution, giving standard definition, intermediate definition and high definition, but sales teams love ambiguity. A lot of sub 1080p displays are defined as "high definition capable" - i.e. if you give them a 1080p signal they will scale it to whatever resolution they work at - but this totally misses the point of high definition in the first place - a one to one mapping between the pixels captured by the camera and those displayed on the screen using a 2Megapixel (or so) sensor.

Plasma technology doesn't lend itself to resolutions much above 720p unless the panel size goes up to around 80", but now that mass-market high definition is on its way and a minimum spatial and temporal resolution has been defined (1920 by 1080 p24), many display manufacturers are working on 1080p displays. Samsung make two LCD displays (a 54" and a 57") that are 1080p60 capable, but they currently cost around $25,000 - although they are also planning to bring out lower cost mass-market ones at around 42", both Sony and Apple make 23" LCD widescreen computer monitors at around $2,000 which are 1080p24 capable, and Apple is introducing a 30" 2560 by 1600 pixel LCD display - primarily for computer use. OLED displays up to 40" and 1600 by 1200 pixels have been demonstrated, but commercial high definition ones are realistically several years away. JVC, Sony and TI have all launched 1080p24 capable display modules in a range of technologies for use in home cinema projectors, and JVC and Sony also have higher resolution 2160p24 capable modules available for use in digital cinemas.

So now probably isn't a very good time to buy a high definition display - unless you're either very keen or very wealthy (or both). Broadcasters are transmitting 720p24 as high definition but Blu-ray discs are designed to support true 1080p24 high definition (sometimes called "full" high definition) and so to get the full experience 1080p24 or better is required. Keep looking at the actual panel resolution and as long as it supports 1920 by 1080 pixels with progressive scan 24 frames per second it'll be reasonably future-proof - although p60 or higher would be better. For home cinemas, the plan is to stick with 1920 by 1080 for a long time but increase the available frame rate over time towards 100Hz. A p24 display will give the full spatial resolution but not benefit from the added temporal resolution over time. Then again, this might not make much difference in the shorter term, as the vast majority of films are shot at 24 frames per second - apart from IMAX which runs at 72 fps.

Note that a 720p display is optimum for 720p material, but 720p material scaled to display on a 1080p display is less of a compromise than 1080p scaled to fit on a 720p display - plus of course 1080p on a 1080p display should give you a high quality cinema-like experience in the home. Hope that this helps, Rob.

By the way, note that the expensive 54" and 57" Samsung LCD displays claim an 8 millisecond response time - which allows them to support p60.

[Rob]

cheers. That does make it all a bit clearer. The reason I was interested was because I'm considering replacing my 21" bog standard Sony CRT with a 26" lcd display. Either Sony's KLV26HG2 or Panasonics TX26LXD1. Both have a resolution of 1280:768. Both have a component input with pal/ntsc prog scan. Which I would like to use in conjunction with the my Pan rec dvd/hdd dmr85 which has pal/ntsc prog out via comp. The Pan has one more rgb scart input, as well as a digital freeview tuner. Both are the same price. Sound is irrelevent as I have a seperate surround sound amp etc...both will accept hi-def via prog. So the decision is purely down to picture quality.

By the way, what happens if you've recorded a tv show to dvd and try to watch it via a compoment lead? will it result in flicker as the original signal is interlaced!? nto really sure about this aspect. Does anyone know?

[Blu-Wave]

Good choice on a 26" 720p capable display for now - so you can get a 1080p one later when there's more choice at lower prices. Note also that 1080i capable displays have to roughly half the vertical resolution in order to avoid interlace artifacts (down to around 540 lines), so "full high definition" displayed as interlaced fields has lower vertical resolution than 720p - if you can't afford 1080p yet, progressive 1080 scaled to 720p is much preferable over using an interlaced display mode. :shock:

On the subject of your DMR85 displaying interlaced data, it depends how good the player/recorder is or, more precisely, which deinterlacing algorithms it supports. Obviously the DMR85 is a high quality machine, so hopefully it'll have a suitable display mode.

When playing films shot progressively at 24fps, the best deinterlacing mode is film mode (not surprisingly). With true progressive source material this gives the highest quality (note however, that even though source material might be progressive, it is stored in an interlaced structure on the DVD - because DVD is an inherently interlaced medium from the days of almost exclusively CRT-type displays - in film mode the player recombines the appropriate fields into true frames for progressive display).

Progressive display of interlaced source material - such as typical current television broadcasts and concerts on DVD, etc., will not produce as good a picture as progressive source material (typically film), as each field is captured at a different point in time, so any motion between pairs of (odd and even) fields has the potential to produce artifacts. Also, as the vertical resolution has to be almost halved in displaying interlaced data (to avoid "twittering", where an object appears to jump up and down between lines as it is displayed first in one field and then in the other), unless the vertical filtering is performed just before display the interlaced data will have half the vertical resolution of progressive data. :x

At the most basic level, if the two fields are simply both displayed at the same time, still images might look great, but any motion such as panning will produce strange "comb" effects protruding out of the edges of objects, and line structure will be obvious in any areas of movement. "Bob" mode is almost an basic, and causes a loss of resolution and objects appears to "bob" up and down a little between fields, but the comb effect is softened. "Weave" mode is a pretty basic weaving together of pairs of fields, with the usual comb, line structure and resolution problems. :roll:

Probably the best deinterlacing mode currently available at reasonable prices is motion adaptive deinterlacing - particularly when it is performed at the pixel level (per pixel) rather than at the field level. This detects stationary and moving parts of the image and treats them appropriately, in non-moving areas, that part of the fields can simply be displayed together, in moving parts, some careful blurring and perhaps a little simple motion compensation can minimse the effects of "combs" etc.. Motion adaptive deinterlacing can look very good in general with interlaced data - but again, often the actual vertical resolution of interlaced source data is half what you might expect, so the jump to true progressive data quality will still be obvious even if the interlacing artifacts generally aren't. :?

A more powerful form of deinterlacing of interlaced source material is to use algorithms which were originally designed to track objects across frames for the purpose of compression in MPeg4 and WMV, for example. Motion vector estimation and hence the associated motion compensation deinterlacing has the potential to get the best out of interlaced material displayed progressively, but at the moment it's both computationally quite intensive and not cheap, and I don't believe that any DVD machines currently use it. However, as progressive displays become more common, along with progressive high definition films on BD, for example, and as the cost of MPeg4/WMV processing falls, there's a lot of archive interlaced video material that somebody will probably want to watch on modern displays, and motion compensation is the way to do it - so expect to see it becoming a high end feature on DVD machines fairly soon.