|Side-by-side: DV & Hi8||
This article was originally published 15 October 1995 on rec.video.professional when the Sony DCR-VX1000 first became available. Some 1998-era comments have been added in [brackets]. Sample images from the A/B tests will be added in the near future. Copyright (c) 1998 Adam J. Wilt
See also the comparison between DV and BetacamSP.
[Most of the following commentary could also apply to a comparison between DVC and SVHS. In my experience, SVHS has a lesser (but not nonexistent) dropout problem, but more color spreading and lower color purity, than Hi8 does. Of course, the usual disclaimers apply -- I haven't done a direct A/B with SVHS and DVC; your mileage may vary; and the views expressed below are mine alone and not necessarily those of my employer, my online service, Sony, etc.]
I hooked up the S-Video out from a Sony DCR-VX1000 3-chip DVC camcorder to the record side of an EVO-9700 Hi8 editor, put a tape in both machines, and shot a variety of indoor and outdoor subjects for ten minutes. I then played back both tapes (moving the Hi8 tape to the TBC'ed player side of the 9700) through a Panasonic WJ-MX50 mixer, allowing me to wipe back and forth between the Hi8 and DVC images for direct comparison.
As has been noted previously, the DVC images are essentially noise free. The color purity is superb, and dropouts are not a problem -- I have been thrashing the first ten minutes of my DVC tape repeatedly (say, about 30-50 passes over it, with multiple backup-replay and eject-rethread cycles in that section of tape), and have seen no mosaic artifacts, "sticky blocks," or dispersed pixels that could indicate a dropout condition. Traditional analog dropout artifacts are completely absent, as one would hope.
Color spread, Y/C delay, and color loss on fine details are also non-issues. In a slow zoom back from electronics gear with small red and green LEDs illuminated, the Hi8 colors dim and drop out entirely, leaving a small luminance "dot" where an LED is present. Even zoomed in, so that the LED covers ten or more scanlines, the DVC colors are purer, and an LED retains its color, co-located with the luminance, as the zoom out progresses. Even when the LED is covered by one or two scanlines only (it's very small), it is still distinctly colored, essentially indistinguishable from the same LED seen live through the camera.
Chroma streaking and noise on large, saturated areas -- for me, Hi8's biggest problem even considering the Russian-roulette dropout issue -- is absent on DVC. I shot a close-up of a deep blue lampshade (the one on the halogen Luxo lamp, if you want to repeat this experiment); blue fish in an aquarium, and several close-ups of saturated greens, reds, and warm woods, and the DVC playback was as clean as the live image.
Luminance detail is much finer on DVC than on Hi8. Small text, such as the button labels on the MX50, were still readable on DVC at 2/3 the size at which they blurred out on Hi8. Amusingly, wide shots of leaves in Hi8 initially looked better (sharper, more contrasty) than on DVC; the edge enhancement circuitry in the 9700 playback "punched up" the appearance of the leaves a bit. However, really looking at detail on individual twigs and leaves showed the DVC image to be less "cartoonish" and more accurate; again, comparison with the live camera image showed the DVC recording to be truer to life.
Also of note is the stability of the detail; on Hi8, vertical edges, even when sharply defined, "quiver" or "wiggle" slightly, artifacts of both time-base error and the noisy Hi8 format (although in fairness, compared to DVC even Betacam vertical detail is noticeably "wiggly"). DVC doesn't look like an off-tape image; its quiescent details always look "live."
The final differentiator was DVC's immunity to DC level errors. Shooting a bright object against a flat background is always problematic on low-end formats; the portion of the scanline following the highlight (the image to the right of it) typically shows a level shift, either for a short time or for the rest of the scan. The result is a "ghosting", darkening, exponentially decaying level shift, or some such artifact; some recorders handle this better than others, but most if not all show this problem to some degree. It's most noticeable shooting a ceiling fluorescent lighting grid at a slight angle, and looking at the ceiling tiles to the right of the lights. Try it; you'll see what I mean.
In my test, I shot the control panel of the WJ-MX50 vision mixer, which consists of small white buttons on a gray background. The Hi8 of this looked fine -- until I wiped to the DVC playback. Further examination showed that the Hi8 image showed a slight "ghosting" or luminance echo of the buttons, faintly mottling the gray plastic area surrounding them. I didn't notice this at first; it took repeated A/B wipes for me to discern why the Hi8 image looked different. The distortion on Hi8 was so subtle that it was almost invisible -- but the DVC image was so clean that the Hi8 suffered by comparison.
In none of my testing was "mosquito noise" -- the quantization artifacts within a DCT block containing high contrast edges -- or "quilting" -- discontinuities between adjacent DCT blocks -- visible in the DVC images (see Craig Birkmaier's article "Wrestling in Wonderland; Compressing the Future of Video, part III" in the May 1994 issue of Videography for examples of these artifacts). I have been able to generate "pathological" images containing mosquito noise, typically by shooting a TV showing white text against a blue background or a fluorescent light angled against a white ceiling, but it takes a very sharp, large monitor viewed close up to see these artifacts [shots of leaves will typically show mosquito noise as well]. Nonetheless, my experience is that DVC compression artifacts are of little concern in the vast majority of shooting situations.
Having said all this, the Hi8 showed remarkably well against the DVC. With most subject material, it took a sharp eye and close observation to see the differences. Without the A/B wipe back and forth, the eye rapidly adjusts to the Hi8 image, and its flaws are not readily apparent. Only when shooting large patches of intense color; small colored LEDs against a colorless background; or when dropout appeared, did the Hi8 call attention to itself.
Based on this, I wouldn't recommend dumping Hi8 for DVC in a blind panic. Hi8 is still a very viable format, although its place as the high quality low-end format of choice has perhaps been lost [indeed, it has been].
To Hi8's advantage are its low shooting cost ($6/hr using Fuji M221MP, vs $12-24/hr with the various DVC tapes available); wide availability of cheap shooting gear (from single-chip throwaways to 3-chip professional camcorders) and editing gear (again spanning the range from control-L consumer gear, to the 9720 dual-transport editor, to the 9850 standalone pro deck); a two-hour tape length (very useful for long meetings, interviews, and presentations); and the finest image available in a color-under format (IMHO!).
DVC today [October 1995] can't offer the economies of shooting, the frame-accurate editing capability [now available in 1998 via DVCAM or DVCPRO decks, and more often than not using a Control-L connection between a DHR-1000 and a VX1000], or the wide range of available gear (this will also change [it did!]). DVC pricing, even as more gear becomes available, will likely stay a notch above Hi8, probably more in line with UVW and PVW Betacam lines (although with some sort of consumer/prosumer lineup as well) [as of early 1998, it's halfway between high-end Hi8 and low-end UVW Betacam gear].
DVC will probably take over as the format of choice for:
1) Any shooting involving large areas of saturated color, such as underwater videography;
2) Shooting requiring the rendition of fine luminance and chrominance detail, such as product demos, advertising, night shooting, or other image-critical low-end applications;
3) Any corporate/event/industrial/low-end-pro work where the market will support the higher equipment cost and per-minute shooting cost (on a par with Betacam shooting costs);
4) Applications requiring Betacam quality but desirous of bulk, weight, and equipment cost reductions (i.e., the ENG market that DVCPRO is targeted at).
Hi8 (and, okay, SVHS, too) will probably remain the format(s) of choice for:
1) Cash-strapped schools, colleges, wedding/event videographers, and other cost-sensitive markets that still require the ability to edit without excess quality loss and then go another generation down for distribution [until they amortize their investments in Hi8 and SVHS gear];
2) Producers and production companies with a hefty investment in Hi8/SVHS gear [but DV is creeping in as replacement cycles occur];
3) Those who need a full featured pro camera like the Sony EVW-300, the Toshiba TSC-100/100G/200, and/or Panasonic Supercam, which offer true professional camera quality, accessories, and operational control: the consumer camcorders don't have the S/N performance, the high-quality glass, or the infinitely-variable zoom of a real camcorder, and the DVCPRO ENG and EFP camcorders are predicted to be more than twice the price of the EVW-300 or the Supercam [no longer true; the Canon XL-1 is excellent, and the DVCAM and DVCPRO lines have affordable pro camcorders available or soon to be available];
4) Most prosumers and hobbyists who are not insanely wealthy [this is less true in 1998, but decent Hi8 cams can still be had for half the price of a roughly comparable DV cam].
Copyright (c) 1998 by Adam J. Wilt.
You are granted a nonexclusive right to duplicate, print, link to, frame, or otherwise repurpose this material,
as long as all authorship, ownership and copyright information is preserved, and a link to this site is displayed.
|Home||SW Engineering||Film & Video Production||Video Tidbits||>DV<|
Contact me via email
Last updated 30 March 1998.