OakLeaf Systems: DV Central’s Area 1394: Roger Jennings Reporting

Background: In the late 1990s, I became interested in digital videography and purchased for about $2,500 a Sony DCR-VX1000 Mini-DV camcorder, which was the first video device to offer a FireWire (later IEEE-1394) digital output:

image_thumb9

I was a consultant to Adaptec Corp. for its AHA-8940 the first commercial consumer-oriented FireWire adapter for Wintel PCs and was a contributor to Microsoft’s PC98 System Design Guide for DV-to-disk recording and editing under Windows 98. I also was a beta tester for early versions of Adobe Premiere video editing software for Windows.

I intended to write a Special Edition Using Windows Desktop Video title for QUE Publishing about professional/consumer (“prosumer”) videography using the Mini-DV (DigitalVideo) tape format and the IEEE-1394 bus, but updating my Special Edition Using Microsoft Access and Microsoft Access In Depth series (now in its 14th edition) consumed my available spare time.

This archive is the third of my three lengthy documents about Sony’s implementation of the IEEE-1394 standard and Mini-DV recording format. IEEE-1394 has been superseded by the lower-cost Universal Serial Bus 2.0 and 3.0. The DV data and Mini-DV recording formats were made obsolete by HDTV broadcasting and consumer camcorders recording to high-capacity flash memory in the MPEG-4 AVC format. Other related articles are:

• Fire on the Wire: The IEEE 1394 High Performance Serial Bus

• DV vs. Betacam SP: 4:1:1 vs. 4:2:2, Artifacts and Other Controversies

This introduction is by Bertel Schmitt, the founder of DV Central:
Roger Jennings is a resident of Oakland, CA, where he cranks out books like Special Edition Using Windows NT Server 4, Special Edition Using Access 97, Database Developer's Guide with Visual Basic 4, and many more.

For years, he's been writing his seminal oeuvre Special Edition Using Digital Video, which is tentatively scheduled to be published by Que Books, an imprint of Macmillan Computer Publishing, in Spring 1998 - if the book is not (as is usually the case) overtaken by events. In Area 1394, Roger writes about all things Firewire. Opinions expressed herein are those of the author, who also holds the copyright to what he writes in Area 1394.

PC 98 and IEEE-1394: Déjà Vu All Over Again

Remember Microsoft's PC 97 specification? July 1, 1997 was the official deadline for Wintel PCs to meet the requirements imposed by Microsoft's PC 97 Hardware Design Guide (ISBN 1-57231-381-1 for history buffs with a spare $45.) Amongst other consumer-oriented enhancements, PC 97's Entertainment PC category required an IEEE-1394 connector.

Well, July 1, 1997 has come and gone, and there aren't any Entertainment PC 97s at CircuitCity, CompUSA, or anywhere else. Even Sony's latest desktop PC offerings and the Gateway Destination PCTV lack a FireWire connector. The problem? Microsoft bit off more technology than it could chew. Much of the PC97 spec required common Win32 Driver Model (WDM) drivers for Windows 9x (Memphis) and Windows NT 5.0. Not only is Memphis missing and WinNT 5.0 in the "Not There" category, but even WDM remains a member of the Microsoft Most Wanted list. PC 97 and the Simply Interactive PC (SIPC), the highlights of 1996's Windows Hardware Engineering Conference (WinHEC), apparently fell victim to Microsoft's allocation of most of its development resources to "embracing and extending the Internet." It's more than ironic that WinHEC 96 opened on April 1: "April Fool. We meant 1998, not 1997."

It's just as ironic that on July 1, 1997 a link to a downloadable version 0.9 of the PC 98 spec appeared on Microsoft's Hardware Development Web page. The PC 98 spec is a joint effort of Microsoft and Intel that's presently scheduled to go into effect on July 1, 1998. (Hopefully, Microsoft will finish Memphis and Windows NT 5.0 before the new enforcement date.) PC 98 defines four configurations: Consumer, Office, Workstation, and Entertainment. (The nascent NetPC has it's own spec.) MPEG-2 replaces PC 97's MPEG-1 as the primary digital video decoder, and DVD drives with downward compatibility replace last year's CD-ROM drives. The Entertainment PC is the only PC 98 version to require IEEE-1394 connectors, so it gets the full treatment in Area 1394. (The spec recommends IEEE-1394 for the Consumer and Office variants.)

The spec's Chapter 5, "Entertainment PC 98," outlines the specific requirements for a PCTV combination, preferably with a 27-inch or larger SVGA (800x600) monitor to replace a conventional TV set in consumers' living rooms or dens. The alternative is baseband NTSC or PAL composite and Y/C video outputs. The Entertainment PC also requires "analog video capture" capability, implying the requirement for baseband NTSC or PAL video inputs. Nowhere in the specification is there a minimum standard for image size, frame rate, or audio synchronization. Even stranger, there's no definitive statement in the spec that motion video capture be accommodated or a Y/C input be included. The way version 0.9 is written, still-frame capture appears to meet the requirements. Video capture via the required IEEE-1394 port appears to be optional.

Note: The Glossary defines analog video as "A video signal that represents an infinite number of smooth gradations between given video levels. Compare with digital video." This definition appears to cement the requirement for baseband NTSC and/or PAL inputs for the Entertainment PC.

The Entertainment PC requires an analog TV tuner covering VHF and UHF frequencies, with analog CATV channels optional, plus an "Advanced Television Systems Committee (ATSC) tuner-demodulator." Elsewhere in the spec, a DTV tuner is optional for Entertainment PCs or is an alternative to a DBS satellite tuner.

Chapter 8, "IEEE-1394," defines the implementation standards for any PC 98-compliant computer that acts as a IEEE-1394 host (bus) controller. The most important change from the PC 97 spec is the requirement for two standard 6-conductor IEEE-1394 ports, one on the front panel for easy accessibility and another on the back panel. (PC 97 required only a single front-panel connector, which would make the host controller a leaf node.) A third, internal 1394 port is recommended for connecting other devices, such as internal fixed-disk drives or the 5-1/4-inch Sony OEM DVTR that's presently being resold by FAST Multimedia. PC 98 requires S400 (400 Mbps) host capability and strongly discourages S100 (100 Mpbs) devices, such as the three million or so Sony DCR-VX1000s shipped to date. (Devices connect to the PC's IEEE-1394 host controller ports.) 400-Mbps PHY chips are just coming to the market; most of today's LINK chips can handle the S400 speed.

The going gets tougher in Chapter 15, "Video and Broadcast Components," which includes sections on "Video Input and Capture Requirements" and "Digital Broadcast Television Requirements." The "Video Input and Capture" section doesn't include requirements for the PC, only a recommendation that timecode be accommodated and a requirement that digital video cameras (not camcorders) use the USB or IEEE-1394 bus. The "Digital Broadcast" section is subject to a well-deserved "early stage of development" caveat. Chapter 15 appears to muddy the Microsoft Broadcast Architecture waters by requiring erstwhile suppliers of DBS adapter cards for DirecTV or EchoStar to add terrestrial DTV and even digital CATV tuning capabilities. (Imagine three type-F connectors for DBS, DTV, and CATV sprouting from a single adapter.) DTV reception implies a requirement for a 16:9 monitor for HDTV, yet Chapter 5 recommends a large-screen 800x600 4:3 display. You can expect plenty of changes to Chapter 15 between versions 0.9 and 1.0, as well as conforming modifications to related chapters.

Chapter 18, "Storage and Related Peripherals," requires that EIDE drives implement Ultra DMA/33 (a.k.a. Ultra ATA) and recommends that "... IDE should be supported on IEEE 1394. An interface device or dongle that allows all IDE devices to be connected to IEEE 1394 and the operating system will support booting from this device." IDE-over-1394 stems from the Device Bay specification, which was still under NDA wraps on July 1, 1997. Device Bay uses IEEE-1394 and USB to let users plug in storage and other peripherals without opening an otherwise sealed-case PC. The "interface device," called Tailgate, is a single chip with core logic and a self-contained PHY to make the transition from a drive's IDE (ATA/ATAPI) connector to an IEEE-1394 socket. Silicon Systems, Inc. (SSI, a subsidiary of Texas Instruments) and Symbios Logic make the $10 chips, which mount on small circuit board behind the drive. Dataquest estimates that 64 million IEEE-1394 fixed-disk drives will ship in 2000, versus 92 million conventional IDE drives. By 2000, the Tailgate chip will become an integral component of the drive electronics.

Establishing a set of basic requirements for video-enabled PCs is an admirable objective. Using an industry specification to force-feed consumers with technology of dubious short-term benefit, such as DTV and digital CATV reception, however, is far less admirable. The problem with version 0.9 of the PC 98 spec is that the video capture requirements are ambiguous, especially as to IEEE-1394 capture and playback. There will be a mighty unhappy bunch of consumers with FireWire-enabled DV camcorders who find their newly-purchased Entertainment PC 98 can't capture, play, or edit their home movies. Adding a requirement that the Entertainment PC demonstrate DV recording and playback capability with DirectShow 2.0 (ActiveMovie's new moniker) would solve this potential technical service nightmare. It behooves the Microsoft and Intel spec writers and software developers to concentrate on making sure consumers receive the full benefit of today's IEEE-1394 technology and not on pie-in-the-sky gimmicks like making a PC into a DTV/CATV set-top box.

Note: For an overview of Microsoft's plans to bring Windows CE to DTV/CATV set-top boxes, read "Bill Gates, the Cable Guy" in the July 14, 1997 issue of Business Week.

According to an article by Joel Brinkley in the July 7, 1997 issue of the New York Times, IBM, Dell Computer, Packard Bell, Hewlett-Packard, Gateway 2000 and Sony have no plans to build DTV receivers into their PCs. The article quotes HP's Lawrence Sennett: "We collect enormous amounts of consumer data, and in recent months we have asked people, flat out basically, if watching TV on a PC was a big thing for them. And the answer we got back was no." Although Microsoft and Intel continue to promote computer-friendly progressive scan formats for DTV, it appears that the DTV broadcast tempest brewed by the computer industry at NAB '97 has subsided. CBS has decided on 1080I (interlaced) for HDTV, and the other three networks are likely to do the same. Brinkley says, "The cable television industry, too, is showing little interest in the computer companies' views. Wendell Bailey, vice president for science and technology at the National Cable Television Association, said: 'I think we are going to stick with interlace.'"

The Entertainment PC 98 is sure to suffer the same fate as the Entertainment PC 97 if its defining specification includes costly features that folks don't want or can't use. What camcorder-owning consumers want today is an easy way to edit and title analog or DV home videos with a PC, but without having buy 15G of additional fixed-disk storage for a 30-minute production. A simple, reasonably fool-proof hybrid (combined linear and non-linear) video editing system for both analog and DV source content will sell many more Entertainment PCs in 1998 than DBS, CATV, and especially DTV tuners.

JVC’s D-VHS Gets FireWired

It's been more than two years since JVC announced its consumer-oriented D-VHS digital bitstream recording format in April, 1995, but no D-VHS gear has yet to hit the shelves of the consumer electronics chains. That's about to change as major and minor players in the direct-broadcast satellite (DBS) receiver business release new D-VHS time-shift recorders in time for the 1997 Christmas selling season. All of the forthcoming digital decks sport IEEE-1394 connectors, but only one of the new models, JVC's HM-DSR100, claims to implement the full IEEE-1394 protocol. The RCA and Hitachi receivers use an incompatible protocol (commonly known as "brain-dead 1394") to connect the DBS set-top box (called an IRD for integrated receiver-decoder) to a separate consumer D-VHS deck.

D-VHS is a 1/2-inch consumer format that records digital data at a variety of bitrates on standard S-VHS ferric-oxide cassettes. D-VHS uses a conventional VHS transport mechanism and head design, so D-VHS decks also can record and playback analog VHS content. (Some of JVC's high-end Digital-S decks can play, but not record, S-VHS tapes.) Standard-format D-VHS records at a fixed 14.1 Mbps data rate with a tape speed of 16.67 mm/sec and VHS's 1,800-RPM drum rotation speed. The overall recording rate is 19.14 Mbps, which includes about 5 Mbps of auxiliary data and inner/outer error correction code (ECC.) Standard length (DF-300) tapes deliver five-hour hour recording time (31.7G); longer (DF-420) tapes hold seven hours (44.4G) of data. Currently, the only application for D-VHS is time-shift recording of DBS programming, although JVC envisions use of D-VHS decks as "data refrigerators," capable of storing bits from a variety of unidentified sources. DBS uses bitrates in the 6-Mbps range for SDTV and audio, so much of the D-VHS tape is likely to be filled with empty bits. At a January 1997 press conference, JVC also discussed recording data rates of 7 and 28 Mbps; the latter is sufficient to handle DV-format data in real time.

Hitachi and Thomson Consumer Electronics, the U.S. subsidiary of France's money-losing Thomson Multimedia that owns the RCA and GE brands, were the first to detail their D-VHS time-shift offerings. Hitachi showed a prototype $599 D-VHS deck with an IEEE-1394 connection to its IRD at the Winter Consumer Electronic Show (CES) in January, 1997. Thomson chose the Satellite Broadcasting & Communications Association (SBCA) show in March to outline its plans for the DSS-3 product line, which includes an as-yet-unpriced D-VHS deck. Much to the dismay of FireWire purists, both decks use brain-dead 1394. TWICE (This Week in Consumer Electronics) quoted Randy Staggs, Thomson VCR product management manager as saying: "The simplified digital A/V bus doesn't have the simultaneous two-way communication of the full IEEE-1394 spec, which allows us to use a less-expensive microprocessor." The same article attributes to Hitachi TV product manager Jim Abrahamsen the somewhat misleading observation that "the interface is 'a simplified digital A/V bus,' meaning that while the IEEE-1394 connection is used, only four pins out of the six specified in the standard are used to carry bitstream data." It's true that four conductors carry the IEEE-1394 data; the two missing conductors in the four-pin IEEE-1394.1 consumer connector carry power. But the IEEE-1394.1 connectors used by today's DV and DVCAM camcorders and decks implement the full IEEE-1394 protocol, not some proprietary "simplified digital A/V bus."

The U.S. DBS market has three major programmers: DirecTV/USSB, PrimeStar, and EchoStar. DBS Digest figures for May 20, 1997, show DirecTV/USSB leading the pack with 2,644,274 subscribers, followed by PrimeStar at 1,899,724, and EchoStar with 549,691. A fourth entrant, AlphaStar, gained only 54,913 subscribers before a bankruptcy filing in early June. DirecTV/USSB and EchoStar transmit MPEG-2 encoded video and audio from high-power satellites to 18-inch dishes purchased by consumers; PrimeStar uses medium-power Ku-band transponders and leases three-foot dishes to users. Rupert Murdoch's News Corp., which acquired the DBS licenses that MCI gained in last year's FCC auctions, originally planned to join forces with EchoStar in a $1 billion venture. News Corp. abruptly cancelled the EchoStar marriage and made a deal to exchange its DBS slots, plus two under-construction DBS satellites, for $1.1 billion worth of non-voting PrimeStar stock. The high-power DBS satellites will give PrimeStar added channel capacity and the ability to use the less-obtrusive 18-inch antenna. So far, there haven't been any reports of digital VTRs for PrimeStar IRDs.

Thomson, Sony, Toshiba, Hitachi, Hughes Network Systems, and Uniden produce DirecTV/USSB IRDs, which carry the DSS (Direct Satellite System) trademark. Thomson's RCA and Sony branded IRDs dominate the consumer electronics superstore channel. Hughes Network Systems (HNS) is best known for its commercial satellite systems (VSAT) and DirecPC Internet service. Microsoft and DirectTV have a long-standing agreement that forms the backbone of the forthcoming Broadcast Architecture for Memphis (AKA Windows 9x.) DirectTV has dedicated a 30-Mbps transponder to broadcasting Web pages and other HTML-encoded content to PCs with an IRD adapter card. (DirecTV recently complained that Microsoft wasn't progressing fast enough with its support for satellite-delivered data, which originally was intended as an add-on to Windows 95.) EchoStar, which markets IRDs under its DISH (DIgital Sky Highway) network brand, has experienced difficulty getting major chains to carry its products. EchoStar relies primarily on TVRO (C/Ku-band "big ugly dish") dealers to distribute its systems. In April, 1997 EchoStar and JVC entered into a agreement whereby JVC would market the DISH IRDs under the JVC label, giving EchoStar the chance to gain shelf space in the superstores. EchoStar says it plans to deliver Internet content to PCs, but the details remain missing. Adaptec supplies PC OEMs with PCI-bus IRD adapter cards for both DirectTV/USSB DSS and DVB-based DISH network systems, which let you view DBS programming and cache downloaded data on your (very large) hard drive.

Note: Current DSS receivers have low-speed and high-speed data ports on the rear panel, but DirecTV, which licenses the design of DSS IRDs, didn't enable them or even explain what use they might be to consumers in the future. Thus Thomson and Sony spent untold millions of dollars installing what amounted to dummy connectors on more than 2.5 million IRDs. Installing dumbed-down IEEE-1394 connectors that can't interconnect with industry-standard IEEE-1394 devices isn't much of an improvement.

JVC announced on June 20, 1997 it's $999 combination IRD and D-VHS deck for the DISH network, expected to ship in October. According to a same-day report in TWICE, the JVC HM-DSR100Du (with a dual-output dish) and HM-DSR100Ru ($949 without antenna), both of which also include a conventional cable-ready VCR tuner, will record and play D-VHS and conventional VHS tapes. You can record DBS content on D-VHS or VHS tapes, but terrestrial NTSC TV and cable content currently is VHS-only. (Initial specs for the Thomson and Hitachi decks only provide for VHS playback.) TWICE quotes Hiroki Shimizu, a JVC (Japan) managing director and general manager of the audio/video sector, as promising "full compatibility with the MPEG-2 video and IEEE-1394 FireWire digital PC interface and standards." What makes D-VHS's use of IEEE-1394 interesting is that it isn't used to make the IRD-to-deck connection; it's designed for external connectivity to MPEG-2 programming content and/or the D-VHS recorder. If JVC writes the necessary Win32 Driver Model (WDM) IEEE-1394 miniport drivers, you'll be able to watch DISH network programming on your 266-MHz or faster Pentium II PC or back up your hard drive to D-VHS tape with a Spark card. (Software MPEG-2 decoders need plenty of CPU horsepower to handle both video and audio streams.) JVC also claims the HM-DSR100 boxes are "forward-compatible" as set-top converter/recorders for forthcoming terrestrial DTV transmission. Shimazu says the D-VHS IRD is ready to handle Dolby AC-3 when EchoStar begins 5.1 audio transmission "sometime next year." EchoStar will distribute the JVC units to their brave band of TVRO dealers.

Note: Shimizu envisions D-VHS as a prerecorded video format, apparently competing directly with DVD. TWICE says: "JVC has already been in talks with major movie studios about their releasing movies in D-VHS, and there have been expressions of interest," and that "JVC was envisioning a 'DigiAna' cassette, that would carry a film in both digital and conventional VHS formats, eliminating the need for studios to issue, and dealers to carry, double inventory." Shimizu appears to be optimistic that JVC can sell the millions of D-VHS decks necessary to encourage the major studios to release movies recorded in yet another format (YAF.) The DVC Consortium envisioned DV as a prerecorded video medium and Sony blamed "copy-protection issues" for the long delay in releasing its DHR-1000 DVTR. I haven't heard even a rumor of entertainment content coming in the DV format.

The three D-VHS entrants in the DBS time-shift market put Sony between a rock and a hard place. Industry reports say Thomson has garnered the lion's share of the DSS IRD market with its second-generation RCA units and the low-price, first-generation GE brand for the discount clubs. Third-generation RCA and new Hitachi IRDs, about to appear on retailers' shelves, promise upward compatibility with separate $599 D-VHS decks coming by Christmas. Sony's higher prices and muted marketing for its DSS systems have appealed primarily to Sony loyalists. With its commitment to the consumer DV format and full IEEE-1394 connectivity, it doesn't behoove Sony Electronics to install a brain-dead 1394 connector on its DSS IRD to connect competitors' D-VHS decks. It's not very likely, either, that Sony will rebrand someone else's D-VHS drive. But Sony must answer the digital recording challenge at its high end of the DBS market without killing the DV golden goose that's laying 24-carat ($3,500+ consumer camcorder and DV deck) eggs. So what's in store for DBS time-shift recording from Sony? A $799 6.35-mm deck that's IEEE-1394 compliant, but won't handle DV? A $999 VTR that will record and play back DSS content but only play your DV tapes? Who knows what evil lurks in the minds of Sony's consumer marketing folks? The Shadow may know, but I sure don't. When I find out, DV-L readers will hear it first.

DV Routs Beta SX for ENG

Hot on the heels of its $24 million deal to sell DVCPRO gear to CBS's 13 owned and operated (O&O) stations, Panasonic Broadcast and Digital Systems Company (PB&DSC) took a $26 million NBC order out from under Sony's nose. According to an article in the June 16, 1997 edition of Broadcasting & Cable, NBC will transition its 11 O&Os to DVCPRO gear for local ENG operations. With a $50 million commitment by two of the three major networks, Panasonic's variation on the DV theme appears to be on its way to becoming what Broadcasting & Cable calls "the digital field format of choice."

PB&DSC claims 17,000 DVCPRO components have shipped worldwide as of June, 1997 and projects U.S. sales for 1997 of 10,000 units. These numbers pale in comparison with the 3 million plus "consumer" DV camcorders shipped by Sony, Panasonic, Sharp, and other members of the DV Consortium since the format's commercial introduction in fall 1995. It's likely that a substantial percentage of Sony DCR-VX1000s and Panasonic EZ-DV1s are in the hands of professional news shooters and video paparazzi throughout the world. On the other hand, 17,000 DVCPRO units represents 5 percent of today's installed base of Betacam SP gear, gained in less than a year of DVCPRO availability. The real significance of the CBS and NBC buy-in to the DV-based format, however, is that DV's 5:1 compression of 4:1:1 sampled video satisfies the networks' quality standards for locally-produced news programs and network feeds of local content.

The Broadcasting & Cable article's author says: "The deal is a blow to Sony, which leads the analog ENG market with its Betacam format and has been trying to compete in the digital world with Betacam SX." Sony's marketing blunder was the assumption that the dominance of Betacam SP would make the transition to Betacam SX a shoo-in at the major networks. Caught without a "professional" version of DV to compete with DVCPRO, Sony rushed DVCAM to the market by adding 5 microns to consumer DV's track width and locking the audio track. But Sony positioned DVCAM as an "industrial" product line for corporate and institutional video, while continuing to push Betacam SX (which some TV folks deride as a "Sonyism") for broadcast ENG. Like Apple and FireWire, Sony converted the heathen with the VX1000, but PB&DSC's selling the bibles. You can buy DVCPRO gear from PB&DSC, Philips, and Hitachi; Sony remains the sole (and erstwhile) supplier of gear using the very proprietary Betacam SX format.

DV, DVCAM, and DVCPRO tape formats are equally suitable for ENG acquisition, despite Panasonic's claim that metal particle tape is more robust than DV and DVCAM's evaporated metal media. Whether the added track width of DVCAM and DVCPRO is a significant factor in linear editing accuracy remains to be seen. What's important is that PB&DSC's decks play all three formats, so O&Os can handle footage from VX1000s, EZ-DV1s, PC-7s, or any other mini-DV source without an format conversion step. (DVCPRO decks use an adapter, much like a VHS-C adapter, to handle mini-DV cassettes; full-size DV or DVCAM cassettes aren't supported.) DVCPRO gear still lacks the long-promised IEEE-1394 adapter cards, but Panasonic's recent announcement of FireWire-enabled consumer DV camcorders may accelerate IEEE-1394 development at the pro level. Apparently, IEEE-1394 connectivity isn't yet a major issue in the broadcast market, where analog component output to Betacam SP for on-air operations and, to a lesser extent, SMPTE 259M SDI and AES/EBU audio are the favored interchange methods.

A surprising factor in CBS's and NBC's decision to go with DVCPRO was the availability of PB&DSC's pricey (about $35,000 list) laptop editor. Broadcasting & Cable quotes Larry Kaplan, senior vide president of Sony's broadcast systems operation: "Sony's inability to deliver some key pieces of its SX line, such as a laptop editor, have hurt SX's chances." You'd need a big lap to hold a portable Betacam SX editor with a couple of Betacam SP-size transports and the batteries to overcome the inertia of 1/2-inch tape mechanisms. Sony's first pass at countering PB&DSC's editor, shown at NAB '97, is a laptop PC and a docking station connected to two rack-mount Betacam SX decks. Sony didn't include a portable generator in the demo setup. Is there a compact Betacam SX cassette in the works?

Momentum is the key to success of a new video format in the broadcast market. Panasonic's M-II format never made serious inroads in the U.S, but it looks like PB&DSC's fortunes are on the mend with DVCPRO. Sony created its own demon by not promoting DVCAM for ENG. A few more DVCPRO buys in the $10 million and up range will permanently relegate Betacam SX to the niche status it so well deserves.

Ultra DMA/33 Promises Low-Cost E-IDE Drives for DV

The First Commandment of DV editing, "Thou shalt use A/V-rated Ultra-Wide SCSI-3 drives with DV," is on its way to repeal as upgraded motherboards and faster fixed-disk drives join the PC mainstream. Today, it's a given that recording and playing back lengthy DV content at 3.5 MBps requires a 1394-to-PCI card, an Ultra-Wide SCSI PCI adapter, several thousand dollars worth of 7,200-rpm Seagate Barracuda or 10,000-rpm Cheetah drives, and multiple blowers to prevent the drives from setting off your smoke alarm. DPS as an example, certifies only Cheetahs and the newest Barracudas for use with the Spark card, and embargoes the older (and very widely-used) 4.3G Barracuda ST15150W. Video professionals buy into this high-priced scenario, but it's not likely to appeal to consumers who plan to edit DV home movies via the built-in 1394 connector on the front panel of their newly-purchased "Entertainment PC."

Enter Ultra DMA/33, a new busmastering protocol to improve the performance of E-IDE (Enhanced Integrated Device Electronics) drives. Called Ultra ATA by Quantum, who developed and patented the technology, Ultra DMA/33 increases the standard Fast-ATA burst data rate of 16.7 MBps to 33 MBps in PCs that use the recently-available Intel 430TX PCIset. (You can identify motherboards that use the 430TX PCIset by checking for a pair of Intel 82439TX and 82371AB chips.) Quantum's 2.1G, 4.2G, and 6.3G Fireball ST drives, which are now appearing in the channels, are the first to support Ultra DMA/33. I paid less than $400 each for my three 5,400-rpm 4.2G STs. By the end of the year, the Fireball SE series will extend drive capacity to 8.4G at an reported street price of $659. (8.4G corresponds to about 40 minutes of DV content.) Most folks have an existing E-IDE system drive and an IDE CD-ROM drive connected as primary master and slave, so two Ultra ATA drives on the secondary cable max out the IDE bus.

Continuous sequential read and write speed, not burst speed, is what's important to video capture and playback. The following table shows the speeds reported for my 4.3G Fireball ST by Michael's Disk Benchmark for Windows 95 (Mdb95.exe from Mdb95.zip), a drive performance test that's widely used by the digital video community. The sequential read and write performance figures for Windows 95 (OSR2) FAT16 compare quite favorably with tests run on my 133-MHz Pentium DV test platform, which uses Seagate ST15150Ws connected to an Adaptec AHA-8940UW. (Despite failing DPS's apparently over-stringent qualification test, FAT16-formatted ST15150Ws work fine for me with the Spark card under Windows NT 4.0 Server, even without disabling the network.) The Fireball ST's very high areal density (2.1G per platter), achieved through the use of magnetorestrictive (MR) read-write heads, is a major contributor to its high read-write speed. Increasing areal density has the same affect on a drive's potential throughput as increasing the RPM. It only takes 33% greater areal density for a 5,400-rpm drive to match the basic data rate of a 7,200-rpm device.

4.3G Fireball ST	ASUS Pentium 200		SuperMicro MMX Pentium 200
512-k Random File	Cached	Non-Cached	Cached	Non-Cached
Sequential Write:	5398 kb/sec	6232 kb/sec	6152 kb/sec	7792 kb/sec
Sequential Read:	6663 kb/sec	6425 kb/sec	6976 kb/sec	7976 kb/sec
Random Write:	1865 kb/sec	1757 kb/sec	2134 kb/sec	2069 kb/sec
Random Read:	2055 kb/sec	1950 kb/sec	2169 kb/sec	2091 kb/sec

I was impressed by the benchmark figures, so I installed and tested the Spark card in both the ASUS (32M S-DRAM) and SuperMicro (32M EDO DRAM) machines. Both had no problem recording DV to the dedicated Fireball ST drive under Windows NT 4.0 Workstation, but I had to disable networking to achieve reliable results with files larger than 500M using the ASUS motherboard. I found no difference between using the system drive and a separate, dedicated drive when recording with either system. Attempts to transmit the .AVI files back to my VX1000 from either the system or the dedicated drive failed, although the ASUS machine occasionally would transmit for a minute or so from the dedicated drive before breaking up into superblocks and halting. I used Premiere 4.2 to make a 9-minute movie on the second partition of the ASUS C: drive, then sent the resulting .AVI file over the network to a Barracuda. Transmitting the movie file from the ST15150W to the VX1000 worked fine.

E-IDE drives consume more processor cycles than SCSI-3 drives, just as software codecs need more CPU horsepower than hardware versions. SCSI adapters and SCSI-3 drives have on-board smarts to minimize CPU overhead. Adaptec's ThreadMark 2.0 benchmarking program shows that my SCSI drives consume only about 40% of available CPU cycles during the overall test, while the Ultra ATA drive took around 56%. (I used ThreadMark's default weighting profile, which isn't tuned for DV's sequential read-write operations, so the comparison might not represent a real-life situation.) I attribute the transmit problem with Ultra ATA drives to the AHA-8940 device driver's need for more CPU cycles during transmission than while recording, along with the drive's slightly slower write than read rate. DV codecs, whether of the hardware or software variety, only come into play during the creation of titles, transitions, animations, and special effects; the codec plays no part in the record or transmit process. Having to write to the archaic Video for Windows file format doesn't help matters, either. For now, it looks to me as if Ultra ATA drives are limited to recording DV clips, and the current version of the Spark card needs high-end Wide or Ultra-Wide SCSI drives to handle data transmission to your camcorder or DVCR. Adaptec's new AHA-8945, an amalgam of the AHA-8940 FireWire card and the AHA-2940UW Ultra-Wide SCSI host adapter, will save PCI slots and dollars for those who haven't already invested in Spark and SCSI cards.

I'm optimistic that developments during the latter half of 1997 will lead to the ability of Ultra ATA drives to reliably handle DV transmit operations. Intel's scheduled to release new PCIsets for Pentium Pro and Pentium II systems that include Ultra DMA/33; Intel says all PCIsets, commencing with the 430TX, will support Ultra DMA/33. (The current 440FX "Natoma" chipset gives Pentium Pro and Pentium II systems bus-master IDE, called BMIDE, but not Ultra DMA/33.) Microsoft claims Memphis boosts disk drive performance, but an NDA prevents further discussion at this writing. Microsoft's scheduled an open beta of Memphis (aka Windows 98) for mid-summer, so stay tuned for the numbers. The Win32 Driver Model (WDM) streaming architecture and ActiveMovie 2.0 are certain to improve both record and transmit performance under Windows 98 and Windows NT 5.0. By the time Quantum's Fireball SE drives hit the channels, Memphis is likely to be in the final beta stage, and suppliers of DV-over-1394 systems should have new WDM 1394 miniport drivers and ActiveMovie 2.0 DV filter graphs available for testing. Hopefully, by then you'll also be able to buy Pentium Pro or Pentium II motherboards with Ultra DMA/33. Swapping a motherboard to reduce DV storage costs by a factor of two or three looks like a good investment to me.

Intel Wakes Up: 1394 Shares New York Stage with USB

New York, June 2, 1997, 7:30 a.m. PDT - Intel Corp., the champion of the Universal Serial Bus (USB), admitted today that IEEE-1394 is a contender for high-speed peripheral connectivity. Speaking at Intel's "Visually Connected PC" extravaganza in New York City, Intel's senior vice-president of its Microprocessor Group, Dr. Albert Yu, said that USB and 1394 are the high-bandwidth peripheral connectivity technologies needed for graphics-intensive business applications. Dr. Yu emphasized that 1394 will come "later" than USB, but it's interesting to find 1394 sharing the spotlight at Millenium Broadway's Hudson Theater with Intel's home-grown and heavily-promoted USB.

Dr. Yu's timing is correct, at least on the PC side of the equation. Virtually all PC motherboards produced in 1997 include at least one USB port. A USB port is required to meet Microsoft's PC '97 specs and thus gain the right to apply the "Designed for Windows Whatever" logo to a PC. Windows 95 OEM Service Release (OSR) 2.1, which is available only for pre-installation by computer assemblers, is required to recognize the USB port(s). If you upgrade your mobo, the USB port is useless unless you have OSR 2.1 (it's included on the MSDN CD-ROMs) and are willing to reinstall Win95 from scratch.

The next most vexing problem is the lack of USB peripherals in the channels. Try to find a USB mouse, keyboard, or joystick at your local computer superstore. A clerk at my local CompUSA in Emeryville, CA responded to product inquiries with "What's a USB?" But the biggest hurdle facing USB proponents is the "ho-hum" factor: USB peripherals simply replace existing high-volume, low-cost components; keyboards, mice, and joysticks dominate the list of forthcoming USB products. It's hard to imagine that USB versions of these devices will have significantly better performance or lower prices than the traditional items they replace. The only reason I was interested in buying a USB component was to check whether the ports on my new SuperMicro and ASUS motherboards were live. (The USB and IrDA ports on my ASUS mobo are on a riser that prevents use of one PCI slot. The single USB port on my ATX-style SuperMicro mobo has its own cutout on the back panel.)

Microsoft's promoting USB for connecting self-powered speakers; Altec Lansing Multimedia and Philips currently offer beta versions of USB speakers, but you need Memphis to listen to them. The USB-enabled Kodak DVC 300 Digital Video Camera and a similar offering from Intel compete with other low-cost videoconferencing cameras, but you need a sound card for the audio channel. Telex just announced its NexVoice.USB stand-mounted digital microphone at Spring Comdex. Telex plans shipments to PC OEMs in 1997Q3, but last time I checked, Windows 9x isn't expected to arrive until 1997Q4 or, more likely, 1998Q1. Prices for the NexVoice.USB weren't available at press time, but I bet that the first USB mics will cost more than today's low-end sound cards. Bear in mind that you also need a USB hub (if you can find one) to connect multiple peripherals to PCs with only a single USB port. And if you want to route USB audio through your home theater system, where's the USB audio input port on your A/V control center? And finally how about MIDI? I haven't seen any proposals by Roland, Korg, E-mu and other musical heavy-hitters to replace MIDI IN, MIDI OUT, and MIDI THRU ports on their tone generators, samplers, and keyboards with a USB connector.

It might be the existence of three million camcorders with consumer 1394 connectors that's suddenly piqued Intel's interest in a USB competitor. Both Intel and Microsoft position 1394 as a supplemental bus technology to be added only when you need more bandwidth than USB can deliver. In reality, USB is the supplemental bus, replacing conventional mini-DIN and D-sub PC peripheral connectors but not enabling any interesting new technologies. IEEE-1394, on the other hand, has its roots in the consumer electronics industry, with Digital Video components here today and multi-channel audio and MIDI systems based on Yamaha's mLAN (music local area network) protocol planned for 1998. What's more, IEEE-1394 doesn't depend on a PC to handle bus management; you simply plug one smart consumer electronics component or musical instrument into another. If you have a PC to act as a 1394 bus controller, so much the better.

I'm not in the market research business, but I bet by the end of 1997 that there will be more camcorders connected to PCs via 1394 than the total number of USB peripherals of every proposed species.

POF to Extend 1394's Maximum Hop Distance to 50 (or 70?) Meters

Plastic optical fiber (POF) appears to be the most practical method, at least in the near term, for extending the reach of IEEE-1394 beyond today's 4.5-meter, living-room-size limit. One of the lesser-known applications for 1394 is the Video Electronic Systems Association (VESA) home network (HomeNet) that's proposed to tie together PCs, TV sets, set-top boxes, audio gear, thermostats, and other household appliances by a combination of room-to-room, in-wall wiring and local conventional 1394 cabling. POF is a primary candidate for the in-wall segments of HomeNet. 4.5-meter hops are likely to be sufficient for most video editing applications, assuming you can find a 4.5-meter 1394 cable, but longer distance transmission is needed to fend off the champions of status-quo SMPTE 259M video and AES/EBU audio, and proponents of proprietary variations on the SDI theme.

Note: Andy Carter of Apple Computer wrote in August, 1994 a white paper that proposed use of repeaters and thicker in-wall cables to reach 35 meters with 100-Mbps FireWire systems. POF, which presumably supports at least 200 Mbps, is likely to be substantially less expensive than a repeater-based approach.

POF got its start for remote sensing in the automotive industry after duPont sold its technology to the Japanese in the 1960s. In 1992, Boston Optical Fiber (the sole U.S. producer of POF) proved that graded-index POF could transmit 3 Gbps at distances up to 100 meters. Less-expensive data-grade POF, called step-index, can handle 300 Mbps over the same distance. The ATM Forum's specification for 155 Mbps over Plastic Optical Fiber was scheduled for approval by the end of May, 1997, and a slower 50-Mbps version is expected by August. Commercial tests of data-grade POF for 100-Mbps Fast Ethernet are taking place with eight-port hubs having seven POF ports and a single copper or glass-fiber interface. Boston Optical Fiber claims that the cost of POF is substantially lower than glass fiber, at least for distances to 100 meters, and rivals that of Cat-5 cabling in 100-Mbps networks. The one-millimeter diameter of POF makes terminations much less critical. The alignment tolerance is 100 microns, greater than the diameter of conventional glass fiber, so you can easily make the connections yourself.

NEC announced in a September, 1996 press release a "multimedia information socket" that mated the company's High Speed Plastic Optical Fiber Transceiver (HiSPOT) with translation circuitry for the 1394 copper protocol. NEC's press release claimed transmission distances of up to 70 meters with the HiSPOT system, but didn't identify the maximum signalling speed. NEC's Patrick Yu mentioned a shorter, 50-meter limit at WinHEC '97; the shorter distance may be the result of moving from 100 Mbps to 200 Mbps, the current standard for 1394 PCI adapter cards. The 50-meter hop length is likely to be sufficient for most home networks. Folks with mansions (like Bertel Schmitt's new Long Island residence) can use repeaters to reach a theoretical 709 meters at 1394's maximum 16 cable hops.

One of the advantages of POF is total common-mode isolation between in-room networks and the elimination of the power management issues that potentially plague large-scale 1394 implementations. Presumably, the HiSPOT and its associate circuitry can be powered at each end of the POF run by a 6-pin in-room node, such as a 1394-to-PCI adapter card in a PC. Where only self-powered 4-pin consumer devices connect to a POF link, a 4-pin to 6-pin converter with a wall-wart power supply could provide current to the HiSPOT.

So far, I haven't seen evidence of an off-the-shelf implementation of 1394-over-POF. After all, 1394-to-PCI adapters have only recently become available. As soon as I get my hands on a POF interconnect, I'll give it a try, assuming there aren't any NDA restrictions at the time. Stay tuned for later reports.

Serial Warfare.

Not content to settle for a Digital Video tape format war between consumer DV, DVCPRO, and DVCAM, Panasonic and Sony have opted for a second front: A battle to establish hegemony in the serial audio/video transport and switching business. Like the recent coups d'etat in Africa, everyone is a loser in the forthcoming Serial War.

Jon Leland's sidebar, "Beyond FireWire," in the May 1997 issue of Videography magazine, illustrates digital I/O FUD applied to DV formats. Panasonic Broadcast & Digital Systems Company is armed with a proprietary CSDI protocol for 4X transfer of DVCPRO footage. Sony Corp., not content to strike with a single non-standard protocol, is attempting flanking attacks with QSDI and SDDI for DVCAM and Betacam-SX. Leland says, "Hopefully, the industry will find a way to create a universal standard soon." That's like saying, "Hopefully, Microsoft and Netscape will resolve their differences and co-develop a common Microscape browser."

Fat chance.

Without question, SMPTE 259M (SDI) and AES/EBU are the current "industry standards" for digital video and audio transport, respectively, at TV networks, O&Os, large-market affiliates, and post houses. The problem is that SDI's 270-Mbps data rate translates to 27 MBps with 8-bit video data. You need a full-blown RAID 0 system to handle 27 MBps, which consumes 1G of disk space for every 37 seconds of video.

While this isn't a problem for folks editing 15- or 30-second commercials, it does create a serious economic issue when using NLE for one-hour documentaries.Leland contends, "[I]f you want your nonlinear system to accept digital input from digital tape formats other than DV, you definitely want a serial digital port, not a FireWire port on your nonlinear system." Leland didn't say which serial digital port, but presumably intended SDI, not CSDI, QSDI, or SDDI, the latter three of which are specific to a single manufacturer. One such SDI offering is Truevision's $11,995 TARGA 2000 SDX adapter card that compresses the SDI data with an M-JPEG codec. For an additional $6,995, you (someday) can buy the company's MADRAS box to convert DV-over-1394 to uncompressed SDI, then recompress it with M-JPEG. (Affluent folks can buy a combination of the TARGA 2000 SDX and MADRAS for a bargain-basement $15,995.) I'm unconvinced that transcoding DV, Digital-S, or DVCPRO-50 content from its compressed state to SDI, recompressing with M-JPEG, then decompressing M-JPEG to SDI, and finally recompressing to the original back to DV, Digital-S, or DVCPRO-50 format is the way to go for high-end NLE.

Note: Digital Processing Systems' three-board Hollywood Video Recorder captures and plays back uncompressed 8-bit or 10-bit SDI, but doesn't have audio I/O. In 10-bit mode, four 4.3G drives store three minutes of video content.

Another enlightening quotation appears in the sidebar from Mark Lemmons, identified as a training consultant for Scitex and VP of DigiTrek: "I don't see any reason why I would run out and embrace FireWire. It's not an industry standard." This statement poses the question of what constitutes an "industry standard." IEEE-1394, like Ethernet, certainly is an electronics industry standard, and it's one of Microsoft's two recommended I/O subsystems for PCs running Windows 9x and NT 5.0. 1394's problem seems to be that it's tarred by the consumer brush--FireWire is used in consumer products, so it can't possibly be a "professional" standard. 1394 is convenient and economical; it lets you multiplex video, audio, SMPTE drop-frame timecode, and device control over a single, low-cost cable. There's no technical reason that 50-Mbps Digital-S or DVCPRO-50, or even 4X DV, can't travel over a 200-Mbps FireWire cable. All you need is a RAID 0 array that can swallow the bits. SDI requires coax, AES/EBU needs an XLR-terminated cable, and device control adds an RS-422 connection. If DV's quality works for you, why complicate your life and max out your credit cards? If you must mix analog footage with DV, buy a $3,500 Sony DSR-30 or DHR-1000 DVTR. I'll bet most professional videomakers won't be able to tell the difference between S-video and YCrCb as the analog source for the DV format. (For comparison shoppers, Sony's SDI interface for Betacam SP decks has an MSRP of $5,400.)

Note: If you aren't satisfied with 1394's 4.5-meter maximum hop length, NEC's forthcoming plastic optical fibre (POF) link gives you a maximum of 70 meters without having to worry about equalization.

The sidebar also contains an observation that you need a need a DV codec to gain access to DV content with QuickTime or AVI-based programs, and that Apple's QuickTime 3.0 will include a software DV codec "sometime this summer."

Apparently, Leland missed the Adaptec/DPS, FAST Multimedia, miro, and Radius announcements of DV adapter cards for PCs and PowerMacs. Waiting for QuickTime 3.0, let alone a DV codec, from Apple infers an outcome similar to waiting for Copeland. Remember Apple's promise of "FireWire on the motherboard" by the end of 1996? Apple converted the heathen, but Adaptec/DPS are selling the bibles to the three million or so folks who've bought camcorders with FireWire connectors as of Spring 1997. It's a pretty safe bet that there will be 10 million or more 1394-compliant DV camcorders and DVTRs in use by 2000. It's an equally safe bet that 10% or more of these units will be connected to a PC via 1394.
The initial acceptance of consumer-based DV products in the corporate and broadcast video markets blind-sided the "professional" video marketing types used to the bigtime gross margins of high-end video gear. What worries the marketers is not that "[s]mall-volume producers may be served by the early entry of ... FireWire," but that the mainstream corporate and broadcast markets for $20,000-and-up VTRs and $5,000-plus circuit cards will dry up. Prospective purchasers, anticipating the cost of upgrading facilities for DTV, want to minimize investment in new NTSC gear. If today's 25-Mbps 4:1:1 DV format satisfies CBS News, it's likely that FireWire will satisfy them, too, once Panasonic gets around to delivering their promised 1394 adapter cards for DVCPRO gear.Leland concludes his sidebar with statement with which it's hard to argue: "In the long run, however, the flexibility that's provided by one of the various forms of SDI may give you more flexibility." John Maynard Keynes put it another way: "In the long run, we are all dead." The victims of the Serial War will be those who bet against IEEE-1394 as a full-fledged "industry standard" and buy into proprietary serial video offerings