Mac companies fudge facts in ’spec wars’

The quest for full-screen, full-motion digital video on the Macintosh rivals the search for the Holy Grail. It is a frenzied and sometimes downright nasty competition. To date, three desktop hardware companies actually claim to have captured the Grail: New Video — the only one shipping a product — SuperMac Technology and RasterOps.

Each has made claims that its product has the capability to capture, display and output 30 frames of video per second at “full screen.” This is true on the surface, since all three products can display a full screen of video at full speed.

However, for most people in the computer industry full screen means 640×480 pixels, with no cheating. None of these companies’ products — at least in their current versions — actually achieve capture at this resolution during capture and/or output.

WHAT’S SO HARD ABOUT FULL-SCREEN, FULL-MOTION VIDEO?

Although computers use display screens that look an awful lot like television screens, one of the hardest things to do on a computer is to process and display full-frame, full-motion video images. The reason is that, at present, analog television and digital computers have remarkably little in common.

The desired result is to digitize, compress and decompress full-frame (640 pixels by 480 scan lines for NTSC video), full-motion (30 frames per second, or fps) video, and full CD-quality sound.

Most companies, especially those developing desktop products, use techniques such as interpolation and doubling the video scan lines to expand a lower-resolution image to fill a 640×480 window and achieve the illusion of 640×480 video.

What’s the big deal? The first problem is that in order to avoid the appearance of flicker at the 30-fps refresh rate used for NTSC (the North American broadcast television standard), TV images are “interlaced”: a TV first paints a “field” containing all the odd-numbered scan lines, then a second field containing all the even-numbered scan lines.

Computer monitors, which are no longer interlaced, have three choices for dealing with this problem.

First, they could use a memory buffer to interleave both the fields, then paint them both onto the screen at once. This is very expensive, because two memory buffers would be required: one for the frame in progress, and one for the frame right behind it. A control mechanism would have to ping-pong between the two buffers to continually refresh the monitor.

The second choice is to pick up one field of scan lines and simply paint every one of those scan lines twice, essentially throwing away half of the video information.

The third, which not many do, is to interpolate the missing scan lines, or take an average between Scan Line 1 and Scan Line 3, for example, and paint that as Scan Line 2.

Choosing samples. Those solutions deal with the interlacing problem. The other decision to make is the number of samples across each scan line. To maintain the 4:3 aspect ratio of a 480-line television image on a computer screen with square pixels requires 640 samples per scan line. This is more than twice the resolution of VHS videotape and one-and-a-half times the resolution of a laserdisc.

In the interest of reducing the amount of digital data, most vendors actually take 320 samples per scan line. They have to interpolate the missing data to blow this up to 640 pixels for a full 640×480 image display.

A true 640×480 digital television image displayed on a computer-quality monitor would be stunning — far better than anything you see on your television set. An enlarged representation of 320×240 image — especially one that has been compressed and decompressed — will not look nearly as good. However, if the technology used is really clever, it may look as good as many TV images — certainly good enough for a great many applications.

Taking 640 samples per scan line (as SuperMac appears to do) should yield a distinctly better picture. Ultimately, as the technology improves we will get “true” 640×480 digital video on computer screens, but in the meantime, purveyors of good-looking desktop video ought to be proud of themselves for the progress they’ve made to date.

Come on, come clean. New Video and RasterOps admit to doubling the scan line and interpolating the “square pixels” from interlaced NTSC video to produce full-screen video on a display device.

The only difference between these two and SuperMac, according to many sources outside of the company, is that SuperMac won’t admit it’s fudging. Steve Blank, SuperMac’s vice president of marketing, maintains that the company’s DigitalFilm board is powerful enough to sample 640×480’s worth of pixels — though others who’ve used the product say it is actually doing only 640×240. Blank refused any other comment on this article, and SuperMac did not return repeated telephone calls by press time.

In addition, SuperMac is publicly bad-mouthing RasterOps as an inferior solution. Not only is this bad karma and in bad taste, but these kinds of spitting matches are enormously confusing to the customer. It might be more productive for them all to address the real issues surrounding widespread adoption of digital video products: storage requirements, which to date are enormous, even with compression, and image quality, which in all cases is somewhat south of perfect.

WHAT DO THEY DO, OTHER THAN KVETCH?

Products by all three companies do an excellent job of capturing analog video, converting it to digital, compressing it for storage or transmission, then decompressing it with a minimal number of artifacts so it can be displayed on a computer screen or output to videotape or television. When you consider the technical heroics that must be accomplished to do this using a desktop computer — taking the analog standards of NTSC or PAL and actually figuring out a way to get things like scan lines and vertical blanking intervals sampled into digital format — all three companies have earned great applause.

If they would stop mudslinging long enough to look at the desktop video market, they would realize that right now not one of the three has a competitor in its chosen segment.

NEW VIDEO GOES AFTER PRESENTATION MARKET

To be fair, New Video has not been much of a participant in the mud baths. Its EyeQ product — actually the first and only product to come to market at this point — was shipped in July without fanfare, hissing or spitting. The company is young and does not yet have the marketing and sales power of either SuperMac or RasterOps.

Its product, designed for the Macintosh in either a two-board set for digitizing and playing back video ($4,495) or with one board for playback only ($2,495), is targeted toward the multimedia presentation market, including networked video for education and training. Like the other two products, EyeQ sports the standard specs: It can capture, compress and display simultaneous video and audio. It can capture and display 30 frames of video per second and make it appear at full screen size without noticeable artifacts. It supports composite, S-Video or RGB input, as well as NTSC and PAL broadcast standards.

But EyeQ has some obviously superior design features that have snagged the interest of such companies as Apple (it was the only board Apple used for digital video demonstrations during its Worldwide Developers Conference) and other large multimedia developers.

For one, EyeQ has a DSP chip on board, which can sample 16-bit sound at CD-quality levels without requiring participation by the computer’s central processor — a real plus considering the sheer mass of audio and video files. It can also play back full-motion video even from slow data rate devices, including CD-ROM and networks, a feat none of its competitors has mastered.

In addition, EyeQ uses Intel’s vastly improved new DVI programmable video processing technology, which, in turn, provides the product with the capability to support a variety of compression schemes and hardware platforms.

A prescient move. In what turns out was a prescient move, New Video decided to buck the past trend toward JPEG compression and built Intel’s i750 video codec chip set into EyeQ. The i750 supports a variety of algorithms, including RTV (real-time video) and PLV (production-level video), JPEG for high-resolution still images and AD-PCM for audio. It can easily be reprogrammed to support new algorithms as they’re developed. Cases in point: New Video showcased a proprietary near-broadcast quality (60 fields per second) video compression algorithm during Macworld Boston this year, and Microsoft is using the DVI chip set in its Video for Windows product (see p. 19).

EyeQ also offers the allure of cross-platform playback capabilities, since it supports RTV 2.1, a cross-platform, scalable video algorithm supported under DVI. This technology enables video output from the Macintosh in the same file format supported by IBM’s ActionMedia board, which in turn allows playback of EyeQ-created video on any computer equipped with ActionMedia.

One drawback to the existing EyeQ board set for many consumers is that it does not provide support for Adobe Premiere 2.0 and Diva VideoShop, both popular with desktop video users. A free software upgrade that supports these products is on the way for those who already own the EyeQ hardware. EyeQ does support QuickTime as well as applications such as HyperCard, IMC’s Special Delivery and MacroMedia Director.

RASTEROPS TAKES A MODULAR APPROACH

The modular approach taken by RasterOps to digital video technology makes it particularly attractive to the professional video production market.

The company has created a family of products with the expectation that professional video editors will want to customize their digital video studios. This approach allows producers to take advantage of certain components they already own, which in turn helps them configure a system that suits their particular needs.

RasterOps’s digital video is based on MoviePak, a product that is still in beta testing and not expected to be shipped until mid-November. MoviePak is a $1,999 add-on card that must be attached to one of the company’s 24-bit video display boards.

MediaTime, one of RasterOps’s two 24-bit cards that can drive a 13-inch monitor, also provides CD quality sound capture and playback. The card uses DigiDesign’s AudioMedia chip set to sample 16-bit audio. The other RasterOps configurations require the use of an audio-digitizing card from a third party.

For output to videotape or to a TV monitor, RasterOps opted for an off-board encoding/decoding solution, thus offering professionals the chance to use high-end encoders/decoders.

What’s required is either RasterOps’s Video Expander, a $699 external controller box that enables video editors to output encoded video to a VCR or TV monitor with RGB pass-through, S-Video, and/or composite video connections, or a RasterOps-compatible encoder/decoder box. Video Expander has genlocking capabilities as well.

MoviePak uses a compression chip set from LSI Logic Corp. that supports a variety of ratios from 2:1 to 160:1. When it’s first released, it will be bundled with Adobe Premiere 2.0 video editing software.

SUPERMAC OFFERS A SINGLE-SLOT SOLUTION

The high-end video professionals moving down to the Macintosh and the multimedia presenters market are only part of the digital video puzzle. There is yet another segment left in the market, one that SuperMac’s technology comfortably fills.

SuperMac’s DigitalFilm product is a double-card, single-slot configuration and an external box that makes it an easy choice for computer users who have begun experimenting with digital video. It is what is known as a plug-and-play solution — albeit an expensive one.

DigitalFilm, now in beta form, is expected to retail at $5,999. It’s a NuBus card set that handles simultaneous video and audio, video compression and decompression, and video encoding and decoding, in a package that only requires a single Macintosh slot. SuperMac also bundles Adobe Premiere 2.0 video editing software with the DigitalFilm board set.

DigitalFilm fits into a Mac IIfx or, preferably, into one of Apple’s Quadra series of computers. (Both RasterOps and New Video have similar minimum configuration requirements.) Although it can capture, playback and output video at the full-motion standard of 30 frames per second, DigitalFilm does not yet support true CD-quality (16-bit) audio on board, though the company says it will.

Compression and decompression is on board via C-Cube Microsystems’ JPEG hardware for still images. SuperMac calls this “motion JPEG,” though there is actually no such standard. JPEG was not designed for time-based media, either audio or video. That said, the on-screen image quality produced by SuperMac’s products is excellent, even in beta. The card can handle a compression ratio of up to 70:1.

A PIECE OF THE PIE FOR EVERYONE

The irony of the spec wars, where vendors fudge the facts on paper to get a “perfect 10,” is that these details ultimately don’t matter. Desktop video customers look to the personal computer as an inexpensive tool for creating compelling prototypes for clients and storyboards for film and video projects. They do not yet use desktop technologies to produce final broadcast output.

Customers, in particular the professional video editors evaluating digital video products, look for image quality. They don’t really care about the technical details of “full-motion video running at a resolution of 640×480,” as long as the quality of the video is good and the storage requirements are somewhat reasonable, at least in comparison to other solutions.

They won’t even notice. Hobbyists dabbling in desktop video aren’t likely to challenge vendor claims. Even if they notice the difference, they aren’t likely to care. The same disregard for specs holds true when you get to the presenters or business communicators. For the most part, they’re seeking something that’s easy to use and makes them look good in front of their bosses and peers. They don’t want or need to know about square pixels and doubling scan lines. But this doesn’t mean they shouldn’t be told the truth.

Desktop video is not a war, nor should it be seen as an exercise in ego gratification. It is an industry and it appears that there are plenty of customers to go around — at least there had better be, since there are many powerful and well-regarded companies like Radius that are just a few months behind in announcing these types of products themselves. It’s a good bet that customers would like to see companies stop the irrelevant competitor-bashing and start figuring out how to fix the storage and image-quality problems that are really holding back digital video from widespread adoption.

Janice Maloney