Bob Currier, Synthetic Aperture
In a world that seems increasingly dominated by capture cards that all use the same Motion-JPEG approach to compression, the Wakeboard Multimedia Pro and its unique approach to compression is a welcome change, if only to break the monotony.
For non-linear editing, motion-JPEG is a good compression solution. But for multimedia authors whose final work is more likely to be a Cinepak- or Indeo-compressed file on a CD-ROM or web page than a tape, M-JPEG capture introduces a source of errors. M-JPEG capture introduces one set of compression artifacts which is then reproduced and amplified by the compression artifacts of the final compression codec. A single generation of recompression is taken for granted in most situations, but eliminating it has the potential to improve the final product.
The Wakeboard from Digital Video Arts Ltd. eliminates the M-JPEG step by offering capture and compression directly into Intel's Indeo Video Interactive (IVI), as well as uncompressed YVU9 and Digital Video Arts' own Wake codec, using a digital signal processor (DSP) rather than a dedicated M-JPEG chip set. Not only does compressing directly to IVI eliminate a compression step--IVI can be used as the final distribution codec--but it saves the time that would normally be required to do the final compression.
The Wakeboard is a PCI-bus board supplied with Windows 95 Plug and Play compatible drivers. It has connectors for both composite and S-video input and output and supports both NTSC and PAL video. Hardware installation was straightforward, with no problems encountered beyond having to remove another board from my crowded system to free up an IRQ for the Wakeboard to use. Software installation also went smoothly, with Plug and Play actually doing its job correctly for a change.
The board comes bundled with Adobe Premiere LE, a feature-limited version of Premiere, that I used for my testing. As an AVI-compatible board, however, the Wakeboard also works with other video capture and editing software such as the venerable VIDCAP.
I started my testing by connecting a Sony DCR-VX1000 digital video camcorder to the Wakeboard's S-video input. My goal in using the VX1000 was to take advantage of the Wakeboard to eliminate one source of compression artifacts. A common operation in my work is to capture S-video from the VX1000 on a Radius VVS and to then do final compression with Indeo. This means going from the DV-compressed tape to analog to M-JPEG compression to Indeo compression; a path that clearly can build up a lot of recompression artifacts.
With the Wakeboard, I was going from DV-compressed tape to analog, then directly to Indeo Video Interactive. Theory said I should get better results by eliminating the M-JPEG step, and my tests showed that I did. Using some footage of waves crashing against rocks--notoriously hard to compress and therefore artifact-prone--I examined the results of both compression paths. At high data rates (1000 K/second) there was little if any visible difference. But when I reduced the data rate to 250 K/second there was a distinct reduction in compression artifacts when using the Wakeboard.
More surprising was that when I performed the same comparison using "talking head" footage and a 1x CD-ROM data rate, artifacts were also reduced. Even the static background area looked better, with less blockiness and "shifting" of the background at key frames.
Similar tests using composite video from a 3/4" source deck also showed improvement using the Wakeboard. The differences between the two capture/compression methods were not as pronounced, but they were still visible.
While this single-step capture and compression yields high quality results, it does assume that no additional processing is required before final compression. In most cases, however, I need to capture video, edit it, add transitions and titles, and then compress the final result for final distribution. With these intermediate steps the advantage of the single-step Wakeboard method is lost.
For such cases, the Wakeboard offers capture and compression using the uncompressed YVU9 codec. This yielded excellent results but, lacking any compression, it limited the size and frame rate I could successfully capture. However, 320x240 at 30 frames/second was readily achievable, and having an uncompressed capture method is also handy for grabbing still images from tape, a common operation for me when I'm creating QuickTime VR object movies.
For more traditional NLE applications, where the final result will be going back to tape, the Wakeboard offers its own Wake codec, a wavelet-based codec that fills the role of the more familiar M-JPEG. In this mode the board is limited to 320x480 when capturing a full 60 fields, yielding lower resolution than some other boards. I expected that the compression artifacts from the Wake wavelet compression would be markedly different from the "blockies" that M-JPEG produces, but I found them to be similar: no worse, no better.
Unlike many other boards--such as the Radius VVS and Targa boards--the Wakeboard does not contain circuitry to drive a computer monitor. That means that while it can capture high frame rates and output them back to tape, it cannot display them at full speed on the computer's VGA screen. Any previewing needs to be done with Premiere's Print to Video function and an NTSC or PAL video monitor connected to the Wakeboard's video output. It also means that many operations with the board--such as selecting the video source and adjusting video settings--are a tedious process where you wait many seconds for data to move from the Wakeboard to your VGA card and for the screen to update itself. I found that the Wakeboard's video settings (brightness, contrast, hue, etc.) did not always stay set unless I opened and closed the settings dialog before each capture.
I experienced infrequent DMA timeouts and GPFs during capture sessions. I updated to the latest (1.07) drivers, but the problems persisted.
The Wakeboard is a video capture board only. To capture audio you need to use a separate sound card. Not having audio and video locked to a single clock is always worrying, not only because of the possibility of multiple clocks drifting in relation to each other, but because capturing audio and video as independent entities provides no guarantees that there will be a consistent number of audio samples for each video frame.
Capturing short segments did not show any loss of lip sync. However, I have a "sync torture test" that I use to check for such problems. This is a 30 minute tape consisting of a slate at the beginning and every 5 minutes thereafter. By capturing this and checking the sync at each slate, one can check for drift between the two. Sure enough, audio and video were out of sync by several frames at the 10 minute mark, and almost one-half second out after 30 minutes. Whether this is the fault of the Wakeboard or my audio card doesn't really matter; it is having no choice but to use an external sound card that is the root of the problem.
Even some capture cards that include audio capture have sync problems, since they don't rigorously match audio sample rates to video sample rates, which are in a non-integer ratio to each other: 29.97 Hz does not divide very conveniently into 44.1 KHz. For many short form applications the sync on the Wakeboard will be no worse than what you'd find on those other cards and will be quite acceptable. But I did find the statement in the manual that you should just capture on separate boards and "Video for Windows then synchronizes the audio and video" to be a trifle disingenuous, considering that Video for Windows--unlike QuickTime--does nothing at all to maintain sync, simply letting the audio and video clocks freewheel and hoping for the best. I guess it is a measure of the improving overall performance of PCs that you can actually get away with this now.
Despite my reservations about the lack of audio capability, the Wakeboard should be looked at seriously by anyone capturing video for distribution as IVI files. Re-purposing existing training videos, for example, would be a perfect application for the Wakeboard, which should produce superior results and save time by eliminating the extra compression step. And unlike video cards that use dedicated M-JPEG hardware, the Wakeboard's use of a "soft" DSP gives it the potential to go beyond the three codecs it currently supports, protecting your investment.
The Wakeboard Multimedia Pro has a suggested retail price of $1095, which includes the board, drivers, cables, and Adobe Premiere LE. Digital Video Arts can be contacted at (215) 576-7920, or via their web site at <http://www.dval.com/>.
| TABLE 1 | ||
|---|---|---|
| Supported Formats, Maximum Frame Rates and Data Rates | ||
| Video Format | Resolution | Average Data Rate |
| Indeo Video Interactive | 320x240x30fps | 2-22 KB/frame |
| 160x120x30fps | 1-11 KB/frame | |
| Wake Video | 640x480x15fps | 9-94 KB/frame |
| 320x480x60fps | 8-44 KB/frame | |
| 640x240x30fps | 8-44 KB/frame | |
| 320x240x30fps | 8-39 KB/frame | |
| YVU9 Uncompressed Video | 640x480x5fps | 345 KB/frame |
| 640x240x10fps | 172 KB/frame | |
| 320x480x10fps | 190 KB/frame | |
| 320x240x30fps | 86 KB/frame | |
Bob Currier is President of Synthetic Aperture, a multimedia production company specializing in digital video and QuickTime VR. He also serves as Sysop of the Macintosh Multimedia Forum on CompuServe.
He can be reached at rcurrier@synthetic-ap.com. Be sure to visit the Synthetic Aperture web site at <http://www.synthetic-ap.com/> for more tutorial information, sample content, and information on new media services.
This article orignally appeared in a slightly different form in the March 27, 1997 issue of TV Technology magazine.
|
Tips & Articles | Press Room | What's New |