HIGH DEFINITION, OR HD, FOR THE MASSES.
Compact and affordable, like Mini DV 10 years ago. Still feels like a pipe dream.
Or maybe a fever dream. Discussing the new HDV format stirs confusion because there are two flavors, three manufacturers and four camcorders. A major player, JVC, launched a popular 24p version of HDV that a year later none of the major nonlinear editing systems can readily capture. Plus there’s Panasonic’s compact P2 camcorder, the HVX200. It’s not HDV — as Panasonic reminds us at every turn — but belongs to this discussion because of its Handycam dimensions and autofocus zoom lens.
One thing is certain, however. Don’t let anyone tell you that HDV isn’t true HD. While both of Sony’s HDV camcorders are professional versions of consumer products, each camcorder mentioned below is capable of superb 720p or 1080i images. I’ve shot documentary projects with all four HDV camcorders: Sony’s Z1 and A1, JVC’s HD100 and Canon’s XL H1. I’ve test-driven Panasonic’s newly arrived HVX200. I’m not talking out of my hat.
WHAT HDV BRINGS
Listservs, user groups, forums, blogs and podcasts set the arrival of HDV apart from the introduction of any previous format, broadcast or consumer.
In much the same grass-roots manner as the Internet overtook the last presidential election, myriad sources of HDV information and peer support for camcorders and editing systems have sprung up, unbidden by manufacturers. Well, with one exception. To set the record straight about P2, Panasonic foisted a blog by a fictitious young metrosexual named Tosh Bilowski (toshpit.blogs.com/). Tosh’s photo is tagged “Thinking HD-licious thoughts.” (Your guess is as good as mine.)
Non-fictitious sources of information about these cameras and this format include the incomparable Adam Wilt (adamwilt.com/HDV/), camcorderinfo.com, dv.com, videosystems.com, abcdv.com, hdforindies.com, studiodaily.com, hdvcafe.com, even Wikipedia. You’ll also find energetic posts and threads in user forums at 2-pop (2-pop.com), Creative Cow (forums.creativecow.net/index.html) and Digital Producer Magazine (digitalproducer.com).
As Google will corroborate, this is hardly a complete list. What’s striking to me about this populist democratization of digital technology is the raw passion with which initiates and pros alike choose sides over engineering arcana like bit rates, CCD architecture, temporal compression and chrominance encoding. As if moral precepts were at stake.
Digital-video wonkism is a horrible/wonderful feature of our time. We free ourselves from a priesthood of broadcast engineers and video technicians more concerned with signal levels than artistic expression and from secretive manufacturers with market agendas not our own. We venture down paths of desktop color matching, conforming, compositing, sound mixing and duplication, domains previously controlled by commercial facilities at commercial rates.
At the same time, we must now sift through torrents of FAQs and Web site entries, often amateur or incomplete. How often have self-appointed Internet experts gone off the deep end and lost sight of the forest (how remarkable HDV is) for the trees (the MPEG2 artifacts which are slight and not fatal)?
My criterion is: how much real-world, hard experience — from shoot to edit to post to projection — do the dogmatic bring to their free advice? Experience with any long-form project earns points in my book.
Low-cost HD is now a practical reality whose methodology is evolving before our eyes. Perhaps the price of the ticket for this unique ride is that we all become, like it or not, digital video wonks.
Load up your pocket protectors.
WHAT HDV IS
JVC began the HDV era in April 2003 at NAB with its three-pound single-CCD 720-line/30p consumer HD1 and its pro counterpart, the HD10. Both were notorious for flat color, excessive enhancement, highlight clipping and pronounced MPEG-2 artifacting. Needless to say, initial impressions of HDV were poor. But JVC’s coup got the world’s attention. In July 2003, JVC, Canon, Sharp and Sony jointly announced HDV standards. A new format was born.
Their idea was to exploit the established economies of Mini DV to jump-start consumer HD. Specs would stay the same — tape dimensions, head-to-tape speed, track pitch, metal-evaporated layers — but instead of DV compression, a superefficient “long-GOP” MPEG-2 compression (see below), the same used in ATSC digital broadcasting and DVDs, would squeeze a 720p or 1080i HD signal onto a Mini DV tape at the same bit rate as DV. As with DV, inexpensive IEEE-1394 (a.k.a. FireWire, iLink) would be used for file transfer.
HDV is therefore HD compressed as conventional MPEG-2 and recorded to familiar Mini DV tape. Note that for reasons of branding, HDV is defined by its manufacturers as a tape-only format. In other words, a tapeless camcorder sending HDV-compressed HD “direct to disc” instead of tape is not considered HDV, at least at the big NAB trade show. A trivial distinction, to say the least, but there it is.
The two flavors of HDV are HDV-1 (JVC) and HDV-2 (Sony, Canon).
HDV-1 is 720 lines progressively scanned (1280 x 720 pixels). Frame rates (in the U.S.) are 30p, 60p and JVC’s variant, 24p. (This may surprise, but 24p was excluded in the agreed-upon HDV specification. As always, follow the money.) Bit rate is 19.7 Mbps (in comparison, DV is 25 Mbps). MPEG-2 video compression is 17:1. Group-of-pictures length is six frames. For purposes of appreciating what a serious tape dropout might accomplish, each GOP is .2 seconds. (Explanation of GOPs follows.)
HDV-2 is 1080 interlaced lines (1440 x 1080 pixels). Frame rate is 60i. Bit rate is 25 Mbps (same as DV). MPEG-2 compression is 22.5:1. GOP length is a whopping 15 frames, a half second.
Compression ratios of 17:1 and 22.5:1 are key to how HDV can squeeze an HD signal onto to the same tape used by a standard-definition DV signal (itself squeezed 5:1). Also key is the fact that MPEG-2 compression at the same bit rate is about 2.5 times more efficient than DV.
This is due to the fact that DV is intraframe compression, also called spatial compression. Each frame is individually compressed and can be individually uncompressed. This makes editing DV relatively simple. MPEG-2 is interframe compression, in which (1) individual frames are compressed and then (2) a group of frames (GOP, for group of pictures) is further compressed over time, called temporal compression. Only one frame in the GOP, called an I-frame, remains intact (like a key frame in animation), while the other frames in the group are preserved as a set of their differences from the I-frame. Redundant picture information, like a blue sky that exists across all frames, is thereby not repeatedly saved as full detail.
JVC says the 6-frame GOP of HDV-1 is milder than the 15-frame GOP of HDV-2 favored by Sony and Canon, yielding fewer motion artifacts during, for instance, a swish pan. Sony and Canon would disagree. Bottom line: both look great. I’ve shot over 100 hours with Sony HDV camcorders and have yet to encounter an objectionable motion artifact.
GOPs are unkind to editing, however. How to make an edit in the middle of a GOP where no real frames exist? Must you make an edit only every 15 frames where an I-frame exists? That would work but is hardly the answer. Apple’s Final Cut Pro, for one, solved the problem by uncompressing/recompressing HDV on the fly in the viewer as the editor makes decisions.
When you render an effect, however, or output an edited sequence, FCP must analyze the resulting image stream in order to re-impose a smooth, uninterrupted 15-frame GOP cadence (in the case of HDV-2). A time sinkhole is putting it mildly. You might as well go out to dinner and a movie, and then rent a hotel room in town as any intrepid, beleaguered HDV editor can tell you. This rendering penalty has contributed to a small trend of transcoding HDV original tapes to HDCam or DVCPRO HD, formats with discrete frames instead of GOP structures, to edit and finish. This also obtains the benefits of far milder compression. Of course working in these formats isn’t cheap.
Surprisingly, HDV’s GOP structure confers a major benefit when it comes to recording. HDV is more robust, less prone to dropouts, than Mini DV. During recording, each frame of Mini DV is written sequentially to tape on 10 slanted tracks. Mini DV error correction is pretty good, but when a large dropout occurs in recording or playback, parts of one or more tracks are not recovered and unsightly macroblocks appear to disfigure the image.
Because of HDV-2’s 15-frame GOP structure, 15 frames must be gathered and compressed before writing to tape (which means there’s always a half second delay in picture during HDV-2 playback). HDV turns this into a virtue by interleaving data from each frame across all 150 tracks (15 frames x 10 tracks), significantly reducing the exposure of a single frame to catastrophic dropout. I’ve yet to experience a dropout in HDV, although I know of others who have.
Panasonic’s P2 flash memory cards (P2 for Professional Plug-in) and DVCPRO HD format avoid both tape and interframe MPEG-2 compression entirely. No tape, no dropouts! DVCPRO HD is 25 Mbps DV multiplied by four to a bit rate of 100 Mbps. Intraframe compression means no GOP structure, which Panasonic touts as editing-friendly. P2 frames are further wrapped in “IT-friendly” MXF (Material Exchange Format) files that contain extended metadata. In coming months you’ll see MXF file formats used in most tapeless recording, from XDCAM HD optical discs to removable hard disks in Grass Valley’s Infinity camcorder.
DVCPRO HD’s codec creates either 720 progressively scanned lines (960 x 720 pixels) or 1080 interlaced lines (1280 x 1080 pixels). (Note the reduced horizontal pixel counts compared to HDV-1 and HDV-2.) In either case, whether output is 720p or 1080i, the HVX200’s powerful digital signal processing (DSP) renders HD internally as a full progressive scan of 1920 x 1080 pixels. This provides versatility: 1080-line frame rates of 60i, 30p and 24p, and 720-line frame rates of 60p, 30p and 24p. (A version for PAL countries offers 50i, 50p and 25p.) P2 tapelessness in turn enables recording and instant playback in 720p mode of 12 Varicam-like variable frame rates from 12 fps to 60 fps for “undercranking” effects.
One more significant (or not?) difference between DVCPRO HD and HDV is what’s called subsampling of color. DVCPRO HD is a 4:2:2 signal, which means it retains half the color samples that the CCDs capture. The other half is discarded. Why? It’s not considered detectable to the eye, thereby reducing signal size. D-1, the gold standard of standard-def digital video, is also 4:2:2. HDV is a 4:2:0 signal, retaining only a quarter of the original color samples, as does DV. Is Mini DV hobbled as a result? Some post experts claim reduced color sampling creates chroma jaggles in compositing.
I dunno. An HDV feature I’m producing, Miguel Coyula’s Memories of Overdevelopment, shot with a Sony Z1, is chock-full of greenscreen and matte shots done in HDV with Final Cut Pro. They look terrific on the big screen. I couldn’t be more pleased.
Bottom line: both HDV flavors obtain great-looking results. As with all things digital video, your mileage will vary depending upon how much you know about what you’re doing.
HDV audio is another matter. Like all previous digital video — from Mini DV to DigiBeta to HDCam — DVCPRO HD records uncompressed 48 kHz, 16-bit PCM audio. HDV instead records lossy, compressed, MPEG-1 Audio Layer II audio, which, while 48 kHz/16 bits, relies on perceptual masking to eliminate frequencies deemed non-critical to the human ear. Audio pros might beg to differ. While HDV audio sounds fine in general, some of those filming in HDV are electing to record double-system sound, as in film. One company, Edirol, makes a compact four-track flash-memory PCM audio recorder that connects via the camcorder’s LANC jack to automatically start/stop with the camcorder.
WHAT HDV CAMCORDERS ARE
ini DV connotes 1/3-inch CCDs, though there’s no reason a larger camcorder with 2/3-inch CCDs can’t record to Mini DV too. In fact several did, first from JVC (if memory serves), then Sony, Panasonic and others. This happened as consumer DV proved so effective, it gave rapid rise in the late 1990s to Sony’s DVCam and Panasonic’s DVCPRO.
Similarly, there’s no reason HDV camcorders must limit themselves to 1/3-inch sensors. Sony’s new XDCAM HD, with a 1/2-inch CCD, records 1080i MPEG-2 to Blu-ray disc at 25 Mbps, identical to HDV. Grass Valley’s Infinity, with a 2/3-inch CCD, will record 1080i MPEG-2 at 25 Mbps to compact flash and Iomega REV drives. Larger pro HDV tape camcorders are on the way too. But to date, probably for reasons of market segmentation, 1/3-inch sensors are the coin of the realm in HDV.
All camcorders below feature native 16:9, 1/3-inch sensors as well as high-res 16:9 LCD screens and color viewfinders, or at least 4:3 screens with adequate resolution to display letterboxed HD. Use of color viewfinders makes possible color peaking for focus detection, newly found in most HDV camcorders. Another new focusing aid is expanded focus, which, at the press of a button on a Sony Z1 or A1, momentarily enlarges the image four times. Notably, only Panasonic’s HVX200 allows use of this feature during actual recording, when of course it’s needed most.
Tiny 1/3-inch sensors give rise to compact HDV and DVCPRO HD camcorders. Sony’s Z1 and Panasonic’s HVX200 follow the classic Handycam profile of the original Sony VX1000, while JVC’s HD100 and Canon’s XL H1 are shoulder-supported. Sony’s A1 is a diminutive palmcorder. In my experience, camcorder ergonomics dictates the style of handheld shooting. For me it is the key factor in choosing a camcorder. I’m partial to the Handycam paradigm of hand-wielding a compact camcorder, although many camerapersons I admire adamantly prefer a shoulder-mount camcorder.
Similarly, there are strong feelings about fixed vs. removable lenses. In these small HDV camcorders I prefer fixed, stubby, auto-focus lenses (Sony Z1, A1, Panasonic HVX200), while others are adamant about removable lenses (Canon XL H1, JVC HD100) with mechanical focus. The Z1 and HVX200 feature, respectively, Zeiss and Leica zooms of amazing compactness and sophistication (multiple aspherical elements!). On the other hand, removable lenses provide the opportunity to use mechanical zooms, like the 16x provided with JVC’s HD100. I admit to the joys of mechanically zooming and focusing when using the HD100.
What must be kept in perspective about these lenses: as someone on the Web quipped recently, what can you expect from the lens of a camcorder that costs less than a professional matte box? Under no circumstances will an HDV lens approach the performance of a $20K HD zoom or Zeiss DigiPrime. In fact, given the miniaturization of these camcorders and sensors, subtle errors induced by off-axis lens elements are of far greater consequence. Noticeable color fringing caused by chromatic aberration is not infrequently apparent in HDV images upon close inspection.
As was the case with Mini DV, 1/3-inch sensors limit the ability of creative cinematographers to create a shallow depth-of-field, and PL-mount adapters for 35mm cine lenses answer this need. Alongside the classic P+S Technic adaptor, with its battery-powered oscillating ground glass, Germany’s Kinomatik is introducing a fixed ground glass alternative, MOVIEtube. A docking system that also accepts a wide range of Nikon and Canon still lenses, MOVIEtube accommodates the Sony Z1 and Panasonic HVX200 (www.movietube.com).
In addition to HD, each camcorder discussed here records standard definition to Mini DV cassettes. This feature was necessary to encourage those working in SD to make the transition sooner to HDV and DVCPRO HD. I’ve shot two standard-def DV projects with HDV camcorders, and they do indeed create noticeably better Mini DV images than Mini DV camcorders. Because HDV camcorders oversample image detail (compared to Mini DV), then process the images as HD before internally downconverting them to SD, images are clearer, sharper and free of stair-step aliasing, often seen in images of slanted roofs and phone lines. Camcorders that internally process 24p signals, like JVC’s HD100 and Canon’s XL H1, subsequently add 3:2 pulldown to achieve conventional 60i SD output.
For those shooting HDV for film-out to 35mm, only JVC’s HD100 offers the ability to turn off the detail-enhancing circuit. Similar to the sharpening filter in Photoshop, excessive detailing adds a plastic-looking outline around fine detail. It exists to enhance image detail on video displays, but when transferred to film, the same images can look excessively video-ish.
Filmmaker Andy Young, VP of special projects at DuArt Film & Video in New York, has made, and continues to make, blow-up tests of HDV to 35mm that demonstrate this. To my eye, his HD100 tests that look most like film are those with the detail circuit turned off completely. While Sony’s Z1 and A1 and Canon’s XL H1 allow adjustment of “sharpness” level, it’s never clear what the sliding scale means, and there is no off setting.
It’s worth remembering that low-budget alternatives to HDV exist. At Slamdance this year I saw a 35mm print of Larry Clark’s Wassup Rockers, handsomely photographed by Steve Gainer, ASC. I almost fell out of my seat upon learning it had been shot with a Canon XL2 (standard definition, native 16:9, 24p).
Down at Sundance, Maria Maggenti’s Puccini for Beginners, photographed by Mauricio Rubinstein, also looked terrific on the big screen. This InDigEnt romantic comedy was shot with a Sony XDCAM (standard definition, 50 Mbps MPEG-2, native 16:9, 24p) and projected as HDCam.
We’ve made great strides since Mini DV arrived 10 years ago. Low-budget digital origination if done right no longer inflicts grunge quality on the big screen. Sweet music to distributors’ ears indeed.
But once you’ve tasted HD, it’s nearly impossible to revert to standard definition. I first filmed with an HDV camcorder over a year ago. It was a Sony Z1, the only HDV camcorder available at the time. I loved it. I haven’t picked up a Mini DV camcorder since.