DV Cassette: The New Wave of Digital Camcorders

The DV cassette standard is finally set in stone, and five companies are shipping digital camcorders that will make Hi8 and S-VHS cameras obsolete and change the way multimedia developers approach video. An exclusive 7-part report from video expert Bob Doyle.

DV Cassette: Here Come the Digital Video Camcorders

By Bob Doyle

While the audio world has largely fallen to the rising tide of digital media, video still remains a bastion of analog recording technology. Videotape still far overshadows digital video, and there are some 300 million analog VCRs and 50 million analog camcorders worldwide. But the age of analog camcorders appears to be up. The major camcorder vendors have settled on a digital video cassette format dubbed DV that offers image quality superior to that of Betacam SP, the current high-quality standard.

The new DV cassette videotape format uses 5:1 data compression to store one hour of digital video in component (Y, R-Y, B-Y) format on a cassette smaller than a digital audio tape (DAT). Five DV camcorders from Sony and Panasonic will be on the market in October, with several more due early next year. DV VCRs will be available after a copy-protection scheme is in place to protect copyright video materials.

The new standard has been two years in the making. In July 1993 Hitachi, JVC, Mitsubishi, Sanyo, Sharp and Toshiba endorsed a proposal by Matsushita, Philips, Sony and Thomson to bring digital video recording technology to consumer VCRs. These 10 companies have since been joined by 45 more, and have created a specification for the DV cassette format and the digital VCR. If fully implemented, this historic agreement may signal the end of 20th-century format wars of VHS vs. Betamax and S-VHS vs. Hi8.

Like the digital video disc (DVD), whose worldwide standardization has just been settled by the very same consumer electronics powers, the new standard for digital videotape uses a form of MPEG 2 data compression. Within a year of introduction, there will probably be more digital media recorded in these new standards than all the current digital video in nonlinear editors, multimedia CD-ROMs, and video servers combined. And the camcorders can also do double duty as digital still cameras, competing in that fledgling market as well.

Whereas the DVD uses interframe compression (I, B and P frames), looking ahead to compress those parts of the screen that don't change from frame to frame, the DV cassette's compression, dubbed I-MPEG, is intraframe (I frames) only. This means that like today's de facto standard of Motion-JPEG (M-JPEG), it compresses each frame individually. Whereas MPEG has had a hard time moving into video editing due to the difficulty in isolating frames for precise editing, the new I-MPEG is born to edit--in time it should replace M-JPEG for professional nonlinear editing systems.

Panasonic's Broadcast Co. has developed a proprietary version of the DV format called DVCPRO for a complete line of electronic newsgathering (ENG) and editing equipment. The complete system includes an airline carry-on-size laptop editing system for globetrotting news producers.

Although they are expensive compared to other consumer camcorders, the DV cassette's image quality, freedom from analog tape artifacts and dropouts, digital (PCM) audio, and lack of generation loss, will very likely push aside the Hi8 and S-VHS analog formats. Compared to Betacam SP, on the other hand, new DV camcorders are an amazing bargain. MiniDV cassettes are 1/10 the size, 1/5 the weight, half the cost and twice the capacity of Betacam camcorder tapes. The standard DV cassette records 270 minutes of video compared to the larger Betacam tape's 90 minutes. Like Hi8 and S-VHS, DV consumer camcorders have Y/C and composite video outputs. Panasonic's non-DV-standard DVCPRO industrial version has component video outputs and inputs, making it a viable contender to Betacam SP in broadcast television.

Relative sizes of DV and MiniDV compared to other standard sizes (approx 75% scale).

The DV cassette features 500 lines of resolution, component recording, signal-to-noise ratios (54dB) better than Betacam, three times the chroma bandwidth of Hi8 and S-VHS (1.5MHz, the same as Betacam), time code, 16-bit DAT-quality audio (or four tracks of 12-bit, 32kHz), separate video and audio insert editing, and direct digital input and output based on P1394 serial SCSI (Apple's FireWire) for multiple video-stream transfers and editing with no generation loss. DV-compatible cards, in development for Macs and PCs, will support nonlinear editing, broadcast-quality video servers and LAN videoconferencing. (For more technical details on the DV spec, click here.)

The DV standard makes provision for both standard-definition TV (SDTV) and high-definition TV (HDTV). It anticipates direct digital recording of HDTV signals like those defined by the U.S. Grand Alliance (the HDTV standards-making group), as well as precompressed video services like the Digital Satellite System (DSS) video used by services like DirecTV. Recording quality is so good that no VCRs will be sold until a copy management system has been adopted that satisfies the nervous motion picture industry. This follows the example set by the music recording industry when it insisted digital audio tape be equipped with a serial copy management system before it reached the market.

DV camcorders should be a huge market success, but it will take some time to reach the vast consumer market that Sony and Panasonic are after. After all, there are 25 million camcorders in the United States, and most of those are standard VHS camcorders, not the pricier Hi8 and S-VHS. The DV camcorders are pricy, and the tape costs two to four times more than Hi8 and S-VHS. But at less than half the cost of Betacam, and better than Betacam quality, adoption of DV by prosumers, corporate and event videographers, and by multimedia and digital video producers, is likely to be rapid.

Bob Doyle, NewMedia magazine's digital video guru, is director of the Desktop Video Group in Cambridge, Massachusetts.

Inside the DV Spec

By Bob Doyle

When looking into the details behind the DV cassette specification, it's hard not to be impressed with the technology. From internal drum speeds to track layout, DV breaks new ground.

The DV cassette's MPEG 2-like compression format is DCT-based YUV 4:1:1 with 13.5 MHz luminance sampling. The video data rate is 3MBps, delivering 5:1 compression, out of an overall 5MBps data rate. DV contains very sophisticated error-correction code to eliminate the video dropouts of small-format tapes. MiniDV cassettes, smaller than a DAT tape, come in 30- and 60-minute sizes. Standard DV cassettes, slightly larger than Hi8 tapes, record 4.5 hours in DV digital VCRs that offer two more channels of audio dubbing. The tiny two-head DV drum (about the diameter of a nickel) spins at an astonishing 9000rpm, five times the standard speed of NTSC recorders and faster than Barracuda hard drives (7,200rpm). DV cassette tape backup systems will store 11GB on DVM60 cassettes and over 50GB on DV270 cassettes. What's more, they could operate at near hard disk speeds.

Unlike D1, DV does not digitize the horizontal and vertical blanking and sync intervals. It only digitizes the 720-by-480-pixel area of "active video." This cuts the uncompressed data rate back from 20.25MBps to 15.875MBps (or, when using megabits per second, from 162Mbps to 127Mbps). DV samples the luminance at the D1 rate (13.5MHz) and samples color difference signals at half the D1 rate (3.375MHz). D1 is 4:2:2 YUV, and DV is 4:1:1. It then compresses the video at a ratio of 5:1 down to 3.1 MBps.

Track layout on the DV cassette is quite extraordinary. Most VCRs record a frame of video as two tracks in a helical scan pattern made by two rotating heads in the drum mechanism, one track for each field of video. The drum spins at 1,800rpm for NTSC (30 frames per second (fps) times 60 seconds). When the tape is in pause, the spinning heads continue to read the same two fields of video. The DV drum spins at 9,000rpm. This means it turns five times during one frame time, and thus makes 10 helical tracks. When DV is in pause, the current frame comes from a frame memory buffer.

Each track has an Indexing and Tracking Information (ITI) sector, an Audio sector, a Video Sector and a Sub Code sector where time code and data codes are stored. Between each sector are Edit Gaps, so that the data for an individual sector may be overwritten independently. This allows insert editing of video, audio or both separately.

The 10 DV video tracks do not contain successive sections of the picture. As with a CD-ROM or a hard disk, the digital data for a single frame is "shuffled" and distributed with error-correction codes over all ten tracks. And the information is made highly redundant, so that a dropout in a portion of one track will not prevent the error correction from restoring all the data for the frame. (Such dropouts are very common in small formats like Hi8, and DV is much smaller.) DV designers claim that two whole tracks of the ten may be lost to dropouts and DV will still be able to construct the picture. DV cassette error correction is like RAID Level 5 redundancy in hard drives, in which one of the five drives can be disconnected and all the data is still available. The 10 DV audio sectors can divide into two sets of five (shown as red and blue), each of which can record 12-bit stereo audio. Sixteen-bit stereo uses all 10 sectors.

The DV cassette has a unique memory IC in the cassette shell that saves a table of contents with record dates and times. Sony cassettes have the memory IC, which enables an on-screen table of contents with record dates of each scene and rapid random access to a scene. . Panasonic does not have cassette memory. This, and perhaps somewhat higher-quality media, make the Sony DV cassettes twice the cost of Panasonic's ($10 for 30 minutes, $12.50 for 60 minutes). Panasonic tapes can be used in Sony camcorders, and Sony will no doubt introduce tapes without cassette memory, if consumers demand them.

Panasonic Goes Proprietary with DVCPRO

By Bob Doyle

Designed for the emerging electronic newsgathering market, Panasonic's proprietary DVCPRO format increases the DV tape speed from 18.8mm per second to 33.8mm per second for a lower bit-error rate, so track pitch is almost double the DV standard. The DVCPRO drum incorporates two extra heads for "confidence" playback of material as it's being recorded, and two flying erase heads for insert edits.

DVCPRO also adds stationary audio cue and control signal heads which record on tracks at the tape edges. The cue track provides a third audio channel and a control track to improve editing accuracy and slow-motion playback. The DVCPRO cassette, with six-minute to one-hour tape capacity, is a nonstandard size about twice the MiniDV package, and uses slightly thicker (9 micron) metal particle rather than more fragile (7 micron) metal evaporated tape.

While DVCPRO differs mechanically from the DV standard, the digital data rates and the 5:1 intraframe DCT-based compression scheme are identical, so digital files are interchangeable. Unfortunately, DVCPRO does not include support for P1394 (FireWire), nor does it have provision for any other digital video file input/output. Panasonic has expressed concern about P1394's 4.5-meter limit on cables and its current maximum of 200Mbps, compared to the 300-meter cables and 300Mbps data rate of the SMPTE 259 serial digital interface (SDI). But desktop video configurations are much smaller than traditional studios, and DV's uncompressed 127Mbps data rate (25Mbps compressed and 40Mbps with audio, error correction, and other overhead) fits into P1394 more easily than D1's 162Mbps fits into the SDI. Substantial pressure is being exerted upon Panasonic to offer interface cards for both SDI and P1394 (FireWire), but it's unclear whether they will give in on this point.

Sony and Panasonic First to Ship DV Camcorders

By Bob Doyle

The first two consumer electronics companies to release digital video camcorders based on the DV cassette standard are Sony and Panasonic. Five models will be available by the end of October. (See table.) Sony and Panasonic DV camcorders all have 180,000-pixel color viewfinders and image stabilization (Sony is optical, Panasonic electronic). Both have a 2X digital zoom to expand optical 10:1 to effective 20:1. Many other DV manufacturers are expected in 1996. JVC has shown an ultra-miniature DV camcorder, and GE and Hitachi models are expected soon.

Sony is now shipping two DV camcorders in Japan and is reported to have an enormous backlog of orders. They are now available in some parts of the United States (for example, at Frye's in the San Francisco Bay Area). Sony is expected to ship in quantity in the United States by the end of October.

The Sony DCR-VX1000 is a three-chip model ($4,199), and the DCR-VX700 ($2,999) has a single CCD chip. Sony CCDs each have 410,000 pixels (NTSC needs only 350,000) for 500 lines, the limiting resolution of the DV cassette. The Sony DV camcorders have S-video (Y/C) and composite outputs but no inputs. This significant drawback is intentional, to prevent users from making high-quality dubs of copyrighted material.

The VX1000 camera section is an improved version of the top-of-the-line consumer Hi8 camcorder, the CCD-VX3, with new Super SteadyShot optical stabilization. Optical zoom is 10:1 (20:1 digital) compared to the VX3's 12:1. Unlike the VX3, the VX1000 can manually override the audio gain, an important plus for professional use.

Both Sony models record 12-bit variable bit size audio, with a dynamic range better than 16-bit linear recording, but a smaller (72dB) signal-to-noise ratio. They can play back 16-bit audio, at 44.1kHz CD or 48kHz DAT sampling rates (recorded in a Panasonic model, for example). Both camcorders record a wide range of extremely valuable data with every scene, including date, time, shutter speed, gain, iris and f-stop, which can be called to the screen on playback. If you have to redo a shot, you can reproduce your original camera settings. They both have Sony Control-L (LANC) connectors for editing.

Both Sony models implement the P1394 (FireWire) digital video interface, for transfers in and out of a computer. No P1394 boards are currently available, but several manufacturers (including Apple and Adaptec, to start with the A's) are expected to show boards at Fall Comdex. TI representatives say they expect to sell 1 million P1394 chips in 1996, with the market driven by multimedia camcorders. Sony's P1394 implementation also sends time-code and edit-control information over the DV interface, so a computer could serve as an edit controller for multiple DV camcorders and VCRs connected by P1394.

For more on the Sony DV camcorders, check out www.sel.sony.com/SEL/consumer/camcorder/digital.html.

The chief difference between the Panasonic DV camcorders and the Sony models is that none of the Panasonic models includes the DV-standard P1394 FireWire interface. Panasonic's PV-DV1000 (about $3,500) is a three-chip consumer model with 270,000 pixel CCDs, suggesting camera resolution below 450 lines. Panasonic claims resolution is enhanced to the 500 lines of the DV tape by displacing green pixels half a pixel relative to red and blue. The DV1000 records either 16-bit or 12-bit audio. If you record 12-bit stereo, you can later dub two more 12-bit tracks in the camcorder. On playback, you can select audio1, audio2 or a mix.

Flying erase heads (not available in Sony DV camcorders) let the DV1000 insert audio and video, although new material must be live through the lens, since there are no video inputs. The DV1000 has a 5-pin Panasonic edit control jack. You can also record in a widescreen 16:9 aspect ratio. The DV1000 has a Top Scan feature that moves from one scene start-point to the next. This could be valuable for logging shots in the field. There is no digital video interface.

Panasonic's model AJ-D700 (about $17,000) adheres to the Panasonic proprietary DVCPRO standard and is aimed at electronic newsgathering applications. The AJ-D700 is part of a family that includes studio and field recorders, a four-times high-speed transfer player, and an airline carry-on laptop editor.

The most notable of the recorders is the Field Edit Package (FEP), a mobile unit which offers two tape transports, two color LCD monitor screens and two jog/shuttle controllers. Although it won't be available until late summer 1996 (for about $25,000), the FEP has caused a minor sensation among travelling news producers, who can get more edit control over their productions. DVCPRO recorders, including FEP, will play standard MiniDV cassettes (with an adapter), but not vice versa. DVCPRO camcorders record on a special ENG version DV-incompatible two-hour medium-size M cassette. Panasonic calls the MiniDV cassette S (small) and standard DV cassettes L (large).

Panasonic also lists a DVD player as part of the DVCPRO family, indicating their intention to integrate diverse MPEG 2 video sources in the production environment. The Matsushita-owned company will also market a special version of the consumer PV-DV1000 camcorder, with different accessories, called the AG-EZ1.

FireWire and I-MPEG: DV Cassette Reinvents Editing

By Bob Doyle

The DV cassette's combination of the P1394 FireWire interface and a standard I-MPEG codec could completely change the compression process for digital editing. Consider how FireWire could change the rules for compression: Typically, once you transfer DV material to a hard disk and disconnect the camcorder, you would need a DV-compatible codec card to play back the material on your computer. Similarly, for very-high quality LAN videoconferencing, each computer would need a DV codec card to decompress files coming over the network, or from a video server for store-and-forward DV video messaging.

However, you may not need a codec card if you keep the DV camcorder connected, and use it for decompressing the hard disk files. During preview in Adobe Premiere, for example, Premiere could send the files out the P1394 connection to the camcorder. In this scenario, the DV camcorder would decompress the files and show video (and audio) on a monitor connected to the DV camcorder outputs, not on the computer display screen. How would Premiere do transitions and special effects, which require uncompressed frames? Although it would be slow, Premiere could use a software codec to decompress the necessary frames for the transition, then recompress the mixed images.

The DV codec card, probably a PCI card with built-in P1394 FireWire interface, could be much less expensive than today's top-quality Motion-JPEG codec cards from Avid, Data Translation, Radius, Truevision, Fast and Matrox. The reason is that the card will not need high-quality A/D converters and the encoding and decoding chips that digitize the video. Many computers now come with AV interfaces. Apple's Macintosh 7500 and 8500 have a Philips video digitizing chipset on the motherboard. Panasonic Broadcast will soon be shipping a developer's kit with its DV codec chipset to a number of companies. Sony claims it will not be making such a card, and will not comment on whether its codec chips are available to developers. At the press conference introducing its DV Camcorder, a Sony spokesman said, "Sony does not comment on things Apple Computer is doing."

Hybrid (disk and tape) editing will also be possible. Multiple DV video stream editing with real-time special effects and transitions would be possible with multiple machine control for camcorders or digital VCRs, and a real-time compositing and blending board like the Fast Video Machine or Matrox Studio.

I-MPEG will also have a major impact. There are three basic reasons intraframe MPEG 2 is likely to replace Motion-JPEG as the de facto standard for nonlinear editing systems: media interchange, audio compression and clean post-compression to full both MPEG 1 and MPEG 2.

Motion-JPEG compression may be in wide use on leading nonlinear editors from Avid, Data Translation, D/Vision, Fast, Matrox, Radius, Truevision, etc., but every company has implemented its Motion-JPEG files differently. Despite media interchange efforts like Avid's Open Media Framework (OMF) and the Open Digital Media (Open DML) consortium of Windows developers, it's not possible to use a digital video archive of Avid media files on a Data Translation or D/Vision editing system.

This will change when DV codec cards are built for the Macintosh and PC. DV media files can be transferred digitally from the camcorder to computer, between computers and servers over local area networks (for video teleconferencing), and can be assembled into the MPEG 2 System Transport Layer for travel wherever MPEG can go.

The main drawback to the new DV format is its fixed 5:1 compression rate. Digital video editors would like to reduce the compression to 2:1 or ultimately to uncompressed video. So we may see DV enhancements to the basic format with increased data rates in the future.

As for audio compression, every nonlinear editor has its own implementation of audio-video file interleave or separate audio file structures, and few of them get the storage benefit of audio compression. At Avid-resolution AVR1, the audio files are much bigger than the compressed video files. With MPEG audio compression, a hard disk could store almost twice the amount of AVR1 media files.

MPEG will also have an advantage in post-compression. When today's M-JPEG files are compressed to MPEG 1 for CD-ROM, or to MPEG 2 for the new DVD, the most minute Motion-JPEG artifacts show up to the MPEG compressor as noise. This is why MPEG encoding systems now offer expensive filtering preprocessors, which reduce the noise, but invariably also throw away image details.

The Moving Pictures Experts Group explicitly states in the standards documentation that MPEG was not designed as an editing format. That's true, but JPEG was not designed as a moving video format, either. Already, you can edit with an intraframe MPEG format that has been developed by C-Cube Microsystems and is being marketed successfully by Sigma Designs--Sigma's Real Magic Producer compresses an I-MPEG 1 stream in real time. You then edit in Adobe Premiere, and when your sequences are complete, and all transitions and special effects in place, you post-compress to full I, B, P syntax. Moreover, Avid held a technology demonstration of MPEG editing at the 1995 NAB show.

For more on IEEE 1394, point to www.skipstone.com.)

DV Camcorders as Digital Still Cameras

By Bob Doyle

DV camcorders using the DV cassette standard will probably find wide use as a 640-by-480 digital still camera, like Apple's QuickTake and Kodak Digital Science cameras. Sony and Panasonic DV camcorders have a high-quality still-picture capture mode, which record about 6 seconds of audio with each still photo. Sony describes their still image processing as Adaptive Frame Interpolation. It extracts a high-quality still from several successive video frames. Unfortunately, the output is only S-video, which is lower resolution than 640 by 480.

As Play's low-cost Snappy ($199) has shown, about four frames (8 fields) can produce a 1,024-by-968 resolution picture in a computer file format like TIFF or BMP. When in photo still mode, the DV camcorders also record about six seconds of audio and video, using up 180 frames, and limiting the number of photo stills to about 300 on a 30-minute cassette. You're better off using Snappy to process groups of four frames. The Sony cameras have an intervalometer (time-lapse) capability that can fire off as few as four frames with a single click of the button. Processing the four-frame sequences through Snappy, you can record 13,500 high-quality stills on a 30-minute DV cassette.

DV Camcorders

Sony DCR-VX7 00 Sony DCR-VX1 000 Panasoni c PV-DV10 00 Panasoni c AG-EZ1U Panasoni c AJ-D700
Price $2,999 $4,199 Under $4000 Under $5000 $16,900
Available Now Now October October November
# CCDs 1 3 3 3 3
Pixels per CCD 410,000 410,000 270,000 270,000 410,000
Zoom /Digital 10X/20X 10X/20X 10X/20X 10X/20X Options /none
Image Stabilization Optical Optical Electronic Electronic None
Min. Illumination 3 lux 4 lux 5 lux 5 lux 2 lux
Photo Still Yes Yes Yes Yes No
Video inputs None None None None Composite
Video outputs Y/C, composite Y/C, composite Y/C, composite Y/C, composite Composite
Digital input /output DV standard DV standard None None None
Audio 12-bit 12-bit 12-bit, 16-bit 12-bit, 16-bit 16-bit
Flying Erase Insert Edit No No Yes Yes Yes
Dub 2nd audio No No Yes Yes No
Track Pitch 10 microns 10 microns 10 microns 10 microns 18 microns
Tape thickness 7 microns 7 microns 7 microns 7 microns 9 microns
Tape speed 18.8 mm/sec 18.8 mm/sec 18.8 mm/sec 18.8 mm/sec 33.8 mm/sec
Edit control LANC LANC 5-pin 5-pin n/a
Weight 2.6 lbs. 3.2 lbs. 2.4 lbs. 2.4 lbs. 11 lbs.
Cassette size mm 66x48x12 66x48x12 66x48x12 66x48x12 98x64x15
Tape speed 18.8mm /sec 18.8mm /sec 18.8mm /sec 18.8mm /sec 33.8mm /sec
Special Features SFX - dissolves, etc. Zebra 16:9 wide 16:9 wide Interchangeable lenses