As technology continues its exponential curve into cyberspace, keeping abreast of developments can be a nightmare. Rest easy as Mixed Media takes you by the hand and tiptoes through the tulip field of sound and vision that is Multimedia

Digital double-feature

During its formative years in the 1980s, the CD-ROM industry had a volatile popularity cycle, with people like Robert Maxwell becoming embroiled in the whole business. He bought a large portion of what was then Nimbus Records; the people responsible for developing the CD mastering lathe, and then took the plunge, leaving behind a bit of a mess. It was left to an American Merchant Bank to pick up the threads. Whilst Nimbus Records went back to selling classical records, and also making a bit on the side selling mastering lathes to the rest of the world, the bank set up two new CD pressing plants; one in Wales and one in Virginia, USA, and re-invented themselves as Nimbus Information Systems.

The chief architects of Nimbus's CD-ROM and video CD catalogue are Jim Orr and Russell Curvengen. Their work with MPEG began with experimentation about a year ago, when they were also writing other software titles for Nimbus, including the widely distributed Anecdote series of CD-ROMs, the complete works of Lewis Carroll and Jane Austen, and The European Multimedia Yearbook. At this stage, there was no real media involved, only text and the occasional graphic file. With the introduction of a new digital video editing suite (see kit list), the emphasis has shifted to the huge growth market of CD video.

CD video uses a compression system developed by the Motion Picture Expert Group (hence the name MPEG), for motion video and animation on computers. Developed specifically as a consumer format, it has suffered from some rather poor commercial releases in the past, although in terms of possible quality, it's broadly similar to VHS. As a demo whizzed by on the CD-I machine, Jim explained the principles of this technology:

"It's difficult to compare VHS and Video CD, because they're different artefacts altogether. In terms of access, Video CD is superior to VHS, and you don't get the colour smearing on Video CD that you do on VHS. But against that, you might get pixelation with Video CD, particularly with a very poor source. Sound quality is better on Video CD. It is digital, but it's not identical to audio CD, because you're now squeezing video into the same space that's normally occupied by the CD quality sound only. Giving an A to B comparison; I know you can hear a difference, but I wouldn't swear which one is which, they're that close. They used some well researched techniques to get around the fact that there's so much information on the CD, whilst still getting a very acceptable sound. The quality is sort of similar to MiniDisc or DCC. It uses the same Lossy compression algorithms, after all."

Despite having more than 600Mb of space on the average disc, compressing video data onto a CD is certainly not an easy task. Whereas with audio, the standard sampling rate is 16-bits at 44.1 KHz, broadcast standard video uses a much higher sampling rate, but only has four bits to a monochrome signal, and two bits to each of the colour differences.

"You don't need so much information for video in a resolution sense, you just need to sample it at an awfully high rate. So you have this high data rate, and you've got to squash 70 minutes' worth down onto a CD, and still have space for the music. So with MPEG 1, you first of all throw away half the lines, leaving you with a quarter of the screen. This is actually almost the same as VHS, which also has a very low resolution system. You then have to decide how long you're going to take before the next DCT intercoded frame. This is then fully squashed down, and used as a reference."

DCT (Discrete Cosine Transform) is a standard compression algorithm, used by MPEG 1 to squash files down to a manageable size. JPEG picture files, and even Digital Betacam film also use the same algorithm to halve the data rate, and save physical space. The process of compression begins by capturing one frame from a video stream (the I-frame), and applying a number of filters to it.

These I-frames are used as a reference frame; one of these is consistently taken, in the case of PAL video, usually every twelve frames. You then store the differences between the I-frames, rather than the actual frames themselves, which saves inordinate amounts of space. It's a similar system to the Delta compression used by computer animation software, except with this method, sometimes only the first image is used as a reference frame.

Problems only really occur when there's a noisy source, or where there are excessively sharp movements (especially cuts). When this happens, there are no real references any more, because the two juxtaposed frames are very different, and so it causes a shimmering effect.

"MPEG doesn't cope with motion too well, but this is all being cured as time goes by", explains Jim. "We've just got a new piece of equipment [MPEG Lab Suite], with which we can just insert an I-frame where we want. When we first tried it, it cleaned up a scene we were working on really well. It was like you'd gone up two levels of quality instantly."

With the new system, reference frames can be placed where they're needed; for example, more I-frames during the parts with greater motion, as opposed to a set space apart. Unfortunately, it isn't really viable to make every frame a reference (although this is the way some editors work during the editing process), because then no space is saved.

Noise is eliminated, as the video is captured using an NRS30 (Snell & Wilcox) noise reducer. It's all done before MPEG compression, else this too would be coded alongside the actual images, making the data larger than it needs be. For the most part, it works rather well, but nothing improves the quality more than using either a digital source, or an analogue tape in excellent condition. Jim told a harrowing tale of trying to transfer several episodes of The Wombles from a crumbling 1" C-format tape onto CD. The outcome was very good, with shimmering and noise only noticeable when the viewer was actively looking out for them.

Like trying to get four elephants into a Mini, there are casualties with this form of compression. Aliasing effects occur when the video is replayed; high frequencies appearing as low frequencies on the picture. But this isn't as disastrous as it is with audio. Indeed, there are a lot of things you oan do to video, and get away with, according to Jim.

"And then there's always a quality-versus-time trade-off, depending on how fast you need to get stuff back to a client, because it all has to be done in real-time. We'll probably invest in an upgrade now we have the MPEG suite, which will allow us to set a 'quality level'. Whenever the quality falls below that level, it'll flag that automatically as it goes through. When it finishes, it'll go back over those sections, and store them properly as continuous I-frames. This is followed by a software compression on them, then the I-frames are slotted back in, so that everything is up to the quality level again."

The potential for MPEG 1, and MPEG 2 if it's released as a consumer format, is astounding. Applications not only include films and television programmes, but interactivity too, in the form of computer gaming, educational and entertainment films. One company are already using these video CDs instead of VHS for video DJing. The quality is much better, as is the controllability.

"One of the things you can do, not so much with video CD but with CD-I, is overlays of graphics or moving images, and things like Chroma Keys. For example, you could ChromaKey things onto an MPEG background. Both CD-I and CD-ROM cards support that. We intend to exploit that, because it's not been done so far. It's almost the reverse of what you see on one of those new games consoles, where they have static backgrounds, and software decompression of people Chroma-Keyed onto it. We could do the reverse of that, and have a very high quality background, moving continuously, and generate the characters, which you can control more easily."

MPEG 2 would be more suitable for games which require a variable playback speed, like car games. Jim predicts that it could very well be possible with MPEG 1, but there's enough work just trying to get the basics working effectively. More constructive applications can be applied by taking advantage of CD video's audio capabilities. There are three options available with MPEG audio: mono, stereo, and dual mode.

"Dual mode is where you can actually record different languages on each channel, and switch between them. CD video allows you to have up to 16 audio channels with an MPEG stream. It does mean you get less video, and there is this exponential fall in quality as you take the video data rate down, but there are a lot of cases when audio like this is useful. A training disc, for example, which will play all over the world. Not only will you have numerous languages, but it'll also play in a CD-I, a PC and Mac (with an MPEG card), and the 3DO, whereas a CD-I will only play in a CD-I machine.

"Because you can have-these different audio streams, there's no reason why you couldn't actually record on the master recording say, the rhythm guitar, bass guitar or whatever, and have them come out of the CD-I on each channel, for a spot of interactive mixing. This is more 'what could possibly be done', than 'what can be done', but the option is there for really clever things in the future."

Currently, Nimbus's CD-ROM and Video CD pressings run in the thousands, rather than tens of thousands, but this is already comparable to some audio CDs. This includes a lot of promotional material, particularly for Philips. With the cost of CD-I and MPEG cards for the Mac and PC becoming more affordable (the new Performas are fitted as standard with these cards, as are some new PCs), the format is increasing in popularity. Nimbus are also High Density CD-ready, which the CD-ROM Division are testing now, although like the CD-R, it will be a couple of years before there is a consumer-priced version. Emil Dudek, the division's director, concluded the discussion:

"With the new building and all, we'll be bringing our pressing capacity up to 70,000,000 CDs a month. Now that's capacity, and we may not always reach that because of seasonal peaks and troughs, so effectively, it will be lower. CD-ROMs now account for between 25-30% of those figures, and it's growing very well. But the edges are also fuzzing between CD-ROM clients and classically audio CD clients, as they advance their products more towards interactivity. It's pressure from consumers, and especially the artists; people like Peter Gabriel, but also the independent acts too. The demand is certainly there."

More from: Nimbus Information Systems, (Contact Details).