From HDV to DVD
Posted Nov 19, 2004

In October 2002, I made my one and only trip to Japan, courtesy of Panasonic, to check out the latest and greatest in consumer electronics technology at CEATEC 2002. Of course, what I was really seeing was technology that was a year or so ahead of the curve for the Japanese, which meant North Americans might not get their hands on it for several years, if it ever made it to these shores at all. The overarching theme of the show was the "ubiquitous network." As far as I could tell, that nebulous term bespoke a notion of convergence going far beyond the union of the office and the television, centered around the PC; this convergence meant the integration of every electric device throughout the household, with the cell phone/PDA/portable video player at the center of it all. This of course meant some robots too, which have always been more popular over there than over here (except in movies)—all of which made me wonder how much of this stuff would ever play in Peoria.

The defining announcement of the show, however, was a new consumer camcorder from JVC that captured 720p, high-definition video to a standard MiniDV tape, in a 19 or 25Mbps MPEG-2-based format called HDV. Exciting stuff, to be sure, and the best news of all was that the camera and the format would debut in the States only six months later at NAB 2003, with Canon, Sony, and Sharp joining JVC in backing the format.

Did that mean the world was ready for HDV? Not exactly. The concept was nothing short of genius: not only was HDV retrofitted to the ubiquitous MiniDV tape format and available in sub-$4,000 camcorders, but it was also compressed enough to fit into the familiar post-production workflow of mainstream PCs. In terms of processing, any PC that could handle DV editing could theoretically handle HDV. Of course, it would take a little tweaking for popular NLEs like Premiere and Final Cut Pro to get into the HDV swing; for one thing, they'd need plug-ins to capture the video and convert it into a format suitable for editing. The format also had its naysayers, who insisted that 720p-only HDV wasn't true HD. The fact was, HDV brought hi-def video tantalizingly close to videographers and other prosumer-level editors attuned to the magic of MiniDV and the brave new world of software-only video editing—that is, preview and processing without special hardware.

At present, HDV represents HD for an SD world. In recent months, as Sony Electronics has become the second manufacturer to throw its hat into the HDV ring, and a slew of software NLE makers—Adobe, Ulead, Pinnacle, Canopus—have announced "native editing" support for the format, the SD world's leading vendors have rushed to embrace it. (Apple, at this writing, remains a holdout.) But what place does HDV input have in a desktop post-production world that's predominantly wed to SD output?

Wed to SD
The event videography world has emphatically gone digital for both editing and output—and for output, the digital format of choice is DVD. Adobe Group Product Manager Richard Townhill is most emphatic about the HDV-DVD connection for those implementing HDV now: "When I saw the FX1 demo in Tokyo, Sony said the best distribution format for HDV content was Blu-ray. But that's not realistic at this point. The realistic workflow is HDV to widescreen DVD." 

By definition, DVD means 720x480 video. That amounts to considerably fewer pixels than the JVC HD-1 and HD-10's 1280x720 and the Sony HDR-FX1's 1280x720 or 1440x1080. That's a significant reduction in the number of pixels available to deliver an image—and that pixel differential is arguably what defines high-definition video as a thing apart from its standard-definition counterpart. 

So what, if any, reason does that leave videographers who deliver their content on DVD to invest in the JVC or Sony camcorders (remarkably cheap for HD, but far more expensive than SD cameras with similar features), and cast their lot with HDV? 

The Capture Game
September 7 was a heady day for HDV. Sony announced its HDR-FX1 HDV camcorder, which, if anything, generated even more hype than the JVC announcements at NAB 2003. And not without reason. For one thing, the JVC models were 1-CCD units, and the Sony is a three-chip camera, which has become the standard for prosumer-level shooting. More than a few pundits would argue that with the effectiveness of today's CCDs the number of chips doesn't matter as much as it used to, but be that as it may, many commercial videographers wouldn't even consider a single-chip SD camcorder, so why would they bite on a 1-CCD HDV model? The other highlight of the Sony announcement was the FX1's vaunted 1080i support. This isn't quite 1080i as we've come to know it; 1080i means 1920x1080 in high-end HD cameras, and the Sony does only 1440x1080. That said, 1080 lines of horizontal resolution is still a remarkably high number for a $3,700 camcorder using MiniDV tapes. 

What really took the buzz to deafening levels on September 7 was the spate of announcements that followed the news from Sony, beginning with Adobe's unveiling of a new plug-in that would allow Premiere Pro editors to capture HDV streams and edit them "natively" in the timeline. [See Eric Schumacher-Rasmussen's "HDV Goes Mainstream," October, pp. 60-65.] What "native" means in this context is naturally a bit complex—if it weren't, promising "native" support wouldn't be such a big deal. 

Because it is MPEG-2-based, HDV is an IBP Long GOP format, which means that each frame in the transport stream actually consists of a group of pictures (GOP): an Index (I) Frame, a Bi-directional (B) Frame, and a Prediction (P) Frame. This makes HDV difficult to edit. If your scene change or cut happens to fall on an I-Frame, you're in luck; if it doesn't, you're stuck trying to pin an edit point on a frame that doesn't exist functionally outside that group of pictures. The fact that there's only one editable frame out of three means the decoder is constantly processing multiple frames for one from which it can generate an image for preview and editing. This means a drain on processor resources and slowed performance.  

Long GOP MPEG-2 is nothing new; DVD-Video itself is a Long GOP MPEG-2 format, as is digital cable content. But it has never been a desktop editing format. As HDV has begun to make inroads into broadcast and news production workflow, Pinnacle Systems has come up with a scheme for making Long GOP editing more manageable by using multiple decoders simultaneously to minimize the system burden that slows performance. Pinnacle Liquid Edition 6, debuting in Q4 2004, takes this approach, according to Pinnacle senior product manager Andrew Baum. "We can support up to five HDV streams in real time" on a dual-processor Windows PC, Baum says. "We can also mix and match native HDV with DV, SD MPEG-2, and PAL video in the same timeline. With the ability to edit in HDV format"—leveraging the compression efficiency of Long GOP MPEG-2 as an interframe codec—"you're getting the best return for the bit," Baum adds.

Other vendors such as HEURIS, Ulead, and most recently Adobe have launched utilities that convert HDV to a more easily edited format. Pinnacle argues that any approach that converts HDV into a non-IBP format (such as all-I-Frame) adds an unnecessary transcode and sacrifices compression efficiency; by Pinnacle's calculation, it takes 60Mbps I-Frame MPEG-2 to deliver the same quality as 25Mbps IBP MPEG-2. Ulead product marketing manager for video products Ian Andes says, however, that Ulead's HD plug-in does nothing more to the video stream than any DV capture utility does when it applies the "AVI wrapper." He adds, "We don't care which camera you use. We take exactly what the camera captures, and get it into the timeline in a format you can edit."

Similarly, Adobe's Townhill, who oversees such applications as Premiere Pro, After Effects, and Encore DVD, says Adobe's new HDV plug-in for Premiere Pro 1.5 simply "transcodes to a wavelet format that stores frame data independently" and maintains all the salient elements of the source video. "Picture resolution, frame size, frame rate, and image quality are all equivalent to the camera signal."

HEURIS, which developed a product called XtractorHDV as a part of its Indie HD Toolkit to enable Final Cut Pro users to capture HDV streams, has "suspended direct sales of XtractorHDV" because Apple hasn't followed up on its initial announcement of support for the format. That said, the application works in much the same way as the Ulead and Adobe plug-ins. "We talk to the camera via 1394, copy the MPEG-2 HD transport stream to the computer, demux it, and then offer choices for converting it to some other format," says HEURIS CEO Brian Quandt. "Frankly, none of us differ in any technical regard in this, or in how long it takes to do the conversions." Regarding the quality of the converted file, Quandt adds: "From a format perspective, the camera is 1280x720 at 30 frames per second. Converted to DVCPRO HD"—the choice HEURIS recommends—"you'd end up doubling the frame rate. But bottom line is, this is an HD format."

Ins and Outs
The usual maxim for compressed and converted digital video is "the better the source, the better the output." That comes with the usual "all other things being equal" caveats, referring to the quality of capture and encode and the like. With HDV, that comes down to "a higher pixel count and better granularity," says Pinnacle's Baum. But does "better in/better out" really apply when the source is better because of pixels that are destined to be discarded? Some argue yes, but those are the least concrete arguments for HDV-to-DVD.

If you want to argue based on other characteristics of the camera—say, comparing the three-chip Sony HDR-FX1 to a three-chip SD camera, comparing low-light shooting, stabilizing features, manual audio and exposure controls, etc.—that's reasonable, and relevant to the why-buy question for existing HDV models. But it's not an argument for HDV itself as a superior shooting format for DVD-destined content.

One key advantage of HDV fully demonstrated in its initial implementation is 16:9 support: all HDV cameras do 16:9, whether in 720:30 or 1080:60 mode. "Buying an HDV camera gets you 16:9 production equipment," says JVC national marketing communications director Dave Walton, and for videographers, this may make a big difference with higher-end clients. "Anyone watching on a plasma, LCD, or rear-projection screen" is optimized for 16:9, widescreen playback. "If you produce in 4:3, the sides will be cropped or the image will be stretched."

There are also good arguments based specifically on the number of pixels in the original source material. One key element of the case for HDV, regardless of the resolution or frame size of the output format, relates to pan and zoom. If you've ever worked used pan-and-zoom utility for still images that works with images at their native resolution (for example, images come off my consumer-level digital camera at a relatively modest 2160x1440) and compared it to one that automatically converts the image to 720x480 before editing, the difference is tremendous. Say your zooming destination is a portion of an image that's 720 pixels by 480 in the original, and can be encoded for a DVD-legal stream without any pixel loss; compare that to the same portion of a 720x480 source image that must then be either displayed at less than full screen with black bars around it, or expanded to fill the frame.

The same principle applies to panned, zoomed, and cropped video: "Because you were working with more data in the beginning," says Adobe's Townhill, "there's a higher data threshold. Any time you're rendering part of an image with effects or filters, there's less blockiness" in the resulting image. Ulead's Andes concurs. "You've got a lot more latitude in how far you can zoom in. It's always better to have it and downscale than to fake it and upscale."

Sony Pictures Digital director of engineering Dave Hill advocates the use of HD and HDV sources—even for eventual SD delivery—for similar reasons. "If you need to re-frame a slightly off-kilter shot, or do an ultra-smooth push-in, for instance, you can do any of these to a significant degree with no image degradation," Hill says. "This would not be true if you did pan and scan-type moves, in software, using SD footage for SD delivery; at some point, the image is going to fall apart, and that point comes much sooner than it does with high-resolution source-material footage." Hill adds, in summary, "High-resolution sources open up many possibilities for impossible-with-SD ‘fix-it-in-post' scenarios."

On the Outs
But not everyone agrees that it's all that advantageous to create a DVD from an HDV source. One question Andes raises is the age-old, "Will anyone notice?" In many cases, the impact of a high-resolution source on lower-resolution output—especially when delivered on middling equipment (like a standard CRT television)—may be so slight that the visual impact will not be felt by the majority of viewers.

But there's another issue, Andes argues, that applies to all HDV authored to DVD. Andes frames it in an analogy to still images. "Here's the really big problem in the industry," he says. "Once we started getting into 3-5 megapixel cameras, things started looking a little weird on the computer. Pixels had to get replaced, remapped, reorganized. Some companies had good algorithms and some didn't. It's the same problem with HDV." When you go from HDV to DVD, you're removing pixels—specifically, however many pixels you have in your HD image (1440x1080 or 1280x720) minus the 720x480 of the DVD. "How the system decides what pixels to take away and where to map them is a complex algorithm. Mathematically, 1:1 is always better than removing pixels." 

In Andes' view, that also means, in effect, an additional encode, something we always try to avoid (like, say, editing in MPEG-2, applying effects, and then deploying an encoder that re-renders the entire MPEG-2 stream along with the effects). "Technically, you're going through an extra encode from the rasterized HDV signal to SD to MPEG-2, changing pixel dimensions and raster size. In this case, you're unpacking a digital signal, resizing and reshaping pixels. It's nowhere near as bad as analog generation loss. But it's a much bigger issue than people realize." 

Sony Pictures' Hill says another potential problem comes with HD post workflow, and particularly the HD-to-SD downconversion. "Probably the biggest negative to using an HD source for SD delivery is the amount of data you'll need to process in your editing application," he says. "Modifying more pixels takes more time, so your render times could and probably would be significantly longer if you used HD as opposed to SD source material." He also suggests a solution. "Some of the HDV cameras will allow you an in-camera SD conversion," he says. "If render time is your biggest concern, you could just capture as [SD] DV, edit with that, maybe grab a few shots as HD for re-framing, and then output your MPEG-2 file for DVD." 

Sharp Shooting
HEURIS' Quandt cites problems with the HDV format itself, and the camcorders that shoot in the format, in terms of offering benefits for DVD authors equivalent to "true" HD. "If you asked this question regarding Panasonic Varicam and DVC PRO HD, for conversion to DVD/SD I would definitely answer that better quality is the result. Some high-end DVD authors are beginning to use this process." But not so with HDV. One problem, he says, is that HDV uses a smaller palette than traditional HD. The camera shoots in 4:2:0 colorspace, contrary to the 4:2:2 of DVCPRO that he says "really makes those images look better." What's more, 720p HDV is by definition a 30fps shooting format. Among the three current offerings, only the Sony HDR-FX1 with its 1080/60 mode can get you 60fps interlaced like you get with DV or the high-end HD models. 30fps, Quandt says, "is not fast enough for the average videographer to shoot material cleanly." Another problem, he says, is "glass": "You need to have $20K+ lenses to shoot good stuff. Bottom line, resolution alone is not what makes HD video, even though that is how we categorize things today." 

Of course, for most videographers—and just about anyone who would consider HDV—high-end HD is out of the question, so the issue is framed more in terms of HDV vs. comparably priced SD cameras, rather than whether to make the jump to a $50,000+ HD camcorder. Ulead's Andes sees comparison-shopping as the best reason to make the HDV plunge. With HD-DVD and Blu-ray "right around the corner" (though that may be a pretty long corner), he says, there's a "big advantage" to investing in HDV now for videographers currently considering new camera purchases. "Why should I buy a $2,000 SD camcorder now if I'm going to have to buy a $3,000 HDV camcorder when Blu-ray comes out and I need HD? It's not worth the $2,000 in opportunity costs to have to replace a new SD camera in two years." 

JVC's Walton frames the "future-proofing" argument in somewhat different terms. "If you're producing something that is going to have some archival value, it's worth considering that future TV systems will be HDTV," he says. If you want to repurpose your video effectively and profitably in the future, it will need to look as sharp as what your clients are seeing on their TVs. His recommendation is to edit in HD, master for DVD as needed for current projects, but retain the edited footage for remastering to HD-capable media. "You will be creating material that will have a longer shelf life. For now, it all depends on where your audience is and what kind of set they're watching on." 

Best of Both Worlds Of course, the ideal solution is to get HDV at its original high resolution directly onto DVD. In the absence of an accepted and widely circulated high-density DVD format (which may turn out to be HD-DVD, Blu-ray, or something else), what's an HDV-DVD author to do? "Currently, Windows Media Video 9 represents the most realistic alternative to DVD[-Video] for HD on a shiny disc," Townhill says. "You can distribute it on disc or stream it. It's a good way of delivering your movies, but not an interactive format like DVD-Video." Townhill cites the new Terminator 2 DVD—which includes both SD and HD versions of the film—as an example of WM9-format HD content delivered on DVD. "You can start to see the beginning of the potential" with early-entry titles like T2.  

But we're still in the very early stages of HD WMV encoding for DVD. Although there's been talk of the technology dating back at least as far as NAB 2004, Sonic only debuted the encoder at IBC in September, with plans to ship the product sometime in Q4 2004. What's more, the Sonic HD-Series system is a high-end hardware encoder designed as an add-on to Sonic's SD series encoders—not exactly the kind of thing most videographers can casually add to their workflow or slip into their post-production budget. For now at least, the convergence of WMV, HD, and DVD is happening in Hollywood and high-end DVD authoring studios exclusively.

Which once again leaves standard-definition DVD as the most realistic delivery medium for HDV-source videography today, with all the pros, cons, and caveats that entails. Of course, for anyone considering new investments in HDV, that might be looking at the question backwards—wondering this early in the game what you can do realistically with HDV once you shoot it, and what you'll lose in the process. So maybe as a videographer producing content for SD-DVD, you should ask it this way: Is HDV the most realistic choice for you?

See Sidebar 1 for more information about JVC's launch of HDV and Sidebar 2 for companies mentioned in this article. 

Sidebar 1: JVC launch of HDV
What's been odd about HDV's launch is the rather curious positioning of the first products. Though the spec is clearly meant to encompass not only prosumer and pro desktop video use but also broadcast, the initial camcorders have been consumer-minded devices with consumer feature sets at prices higher than most prosumer SD models, like the Panasonic DVX-100, the Sony DCR-2100, and the Canon GL and XL series. 3-CCD is widely considered the dividing line between consumer and prosumer camcorders, and the first two JVC models were 1-CCD. Three-chip configurations help considerably with low-light performance (a common knock on the JVC HDVs); the first HDV cameras have also lacked standard prosumer-level operational features like manual audio level and exposure controls.

One explanation for launching "consumer" cameras that cost close to $4,000 into a market where SD cameras with more of the on-camera capabilities that pros expect go for $2,000 is that these are ahead-of-curve Japanese products and Japanese consumers tend to pay more for gadgets and hobby items than Americans. JVC national marketing communications manager Dave Walton explains it differently. "It's expensive to develop these products," he says, and only through targeting a mass market can manufacturers hope to recoup those development costs. "There aren't enough prosumer videographers to justify the expense," he explains, so you have to hunt potentially bigger game—consumers who "are impressed by HDTV" and "don't care what they have to spend" to shoot it.

Future HDV will see products in more categories with a broader range of feature sets; for example, the high-end ENG HDV camcorder demo'd at NAB 2004 won't necessarily be the next HDV product released by JVC, Walton says. JVC showed that camera, he said, to deliver a message to NAB attendees: "Why invest in $20,000 SD camcorders when we'll have HDV products in that price range? There will be other models developed," he continues. "Some of those might come sooner." But in the meantime, videographers and other digital studio pros will have to weigh the value of the HDV format itself against the cameras' lack of standard prosumer features. Which makes it only fitting, perhaps, that the strongest argument for producing DVDs from HDV sources today is all the great "fix it in post" possibilities the higher resolution presents.

Sidebar 2: Companies Mentioned in this Article
Adobe Systems, Inc.,
Apple Computer, Inc.,
Pinnacle Systems, Inc.,
Sony Electronics,
Sony Pictures Digital Networks,
Ulead Systems, Inc.,