December 18, 1997

WHAT'S HOT IN THE DESIGN COMMUNITY

Schottky diode cuts 100 mV from forward drop

A family of 20V Schottky diodes from International Rectifier targets 3.3V switch-mode power supplies. The company claims the diodes spec 100-mV lower forward voltage than existing state-of-the-art Schottkys. The 87CNQ020, an 80A diode that specs 300-mV V_F at 40A, is available in a surface- or clip-mount D-61 package. The 40A 47CTQ020 specs 300-mV V_F at 20A and comes in a standard TO-220, surface-mount D²Pak, or I²Pak (TO-263) package. The 87CNQ020 costs $5.85 (10,000).

--by Bill Travis

International Rectifier, El Segundo, CA. 1-310-252-7105, fax 1-310-252-7100, www.irf.com.

Engineer sleuths crack Mystery Circuit on EDN's Web site

The mystery is solved. If you read our story about connecting embedded systems to the Internet, you might recall that EDN Contributing Editor David Shear put an embedded Web server on the Net (see "Going global with an embedded system on the Internet," EDN, Sept 12, 1997, pg 36). You might also recall that Shear's Web server connects, via an analog I/O board, to EDN's Mystery Circuit, a puzzle that he designed especially for the amusement of engineers. As Shear described it in his article, the Mystery Circuit is "essentially a nerdy game of solitaire for engineers who need a distraction from serious work." He challenged readers to exercise the Mystery Circuit over the Internet to try and determine what components are inside the circuit's "black box."

Well, lots of engineers did exercise the Mystery Circuit, but only one figured out what it is. David Kanceruk, a principal technologist at Lockheed Martin Canada (Winnipeg, MB, Canada), correctly deduced that the circuit comprises four Schottky diodes connected in series. OK, so that's hardly a complex or even useful circuit, but it's not a trivial task to identify when all you can do is apply a couple of voltages to a black box and read a few other voltages at internal--and unseen--test points. We congratulate Kanceruk for his sleuthing abilities. To reward his efforts, we're giving him his choice of $200 worth of Tech Toys (see "Tech Toys," EDN, Dec 4, 1997, pg 99).

A couple of other readers came close to solving the Mystery Circuit. Chris Vagnini, a senior hardware engineer at RadiSys Corp (Newton, MA), concluded that the circuit comprised four diodes, but he was unsure whether they were germanium or Schottky. Vagnini was on the right track, though. Although he thought the diodes probably were germanium, he wondered, in his contest entry, "Where in the world did you find germanium diodes? Do people still make them?" Later, when we informed Vagnini that he almost won our contest, he responded, "Hmm, so they were Schottky after all. I guess it makes sense, since you probably cobbled this circuit up from parts lying around in a lab." Yup. For Vagnini's insight and good effort, we're awarding him a consolation prize, his choice of $100 worth of Tech Toys.

The most detailed analysis of the circuit that we received in a contest entry came from Ronald Reder, who worked at Eastman Software Inc (Billerica, MA) when he submitted his entry and has since moved to ColorAge Inc (Billerica, MA), where he's a software engineer. Reder determined that the circuit contained four diodes, and he measured the on-resistance (2 kiloohms), the knee voltage   (0.38V), and the reverse leakage (less than 0.1 µA at 1V). "I think they're germanium diodes," Reder said in his entry, "but I can't remember if they have that low a leakage current. I'd look it up, but that would be cheating." (What a high-minded guy!) Reder later said that his choice of germanium diodes probably reflects how long it's been since he did hardware design for a living (20 years). We're impressed that a software guy came so close, so we're awarding Reder $100 of Tech Toys. We trust that the toys he selects will be hardware-oriented.

If you haven't yet exercised EDN's Mystery Circuit, you still can. You'll find it on our Web site at www.ednmag.com.  For now, it's still the four-diode circuit, but we'll soon implement a new circuit to challenge you. In the meantime, try to forget that you know the puzzle's answer and see how you might go about solving it.

To exercise the Mystery Circuit (Figure 1), you apply two voltages, and you measure six. Two of the voltages you measure are simply a confirmation of the voltages you apply. Because they're actual measurements, they probably differ somewhat from your nominal applied voltages. The other four voltage measurements provide clues to what the circuit is doing and, therefore, what the circuit is. Bear in mind that Figure 1's schematic shows voltages on the embedded Web server's analog-I/O board. Thus, the output voltages are outputs on the I/O board but inputs to the Mystery Circuit. Similarly, input voltages are the Mystery Circuit's outputs.

To apply voltages to the Mystery Circuit, you simply specify their values (Figure 2). The example illustrates the application of 5V (Voltage 1) and 0V (Voltage 2) to the circuit. The form that you fill out to specify voltages also contains boxes for digital values, which don't apply to our diode circuit. You need not specify these values, and the circuit ignores them if you do.

After you specify voltages, EDN's embedded Web server applies the voltages to the Mystery Circuit and measures resulting voltages at the circuit's test points. The results of applying 5 and 0V to our diode circuit appear in Table 1. For this circuit, you can ignore the columns for digital inputs and outputs.

Wanna play with a new Mystery Circuit and try to win some Tech Toys for yourself? Soon you can. On Jan 15, we'll put a new circuit online to give you a fresh challenge. In the meantime, check out our Dec 4 issue or our Web site (www.ednmag.com) for nifty Tech Toy prizes you can win. Who says engineering isn't fun?

--by Gary Legg

Figure 1

EDN's Mystery Circuit connects to an embedded Web server on the Internet. Access the circuit through our home page at www.ednmag.com to figure out its components and configuration.

Figure 2

You can set inputs to the Mystery Circuit by filling out an online form with your Web browser.

IEEE 488 speed-up proposal divides test-and-measurement community

Controversy over proposed enhancements to the 25-year-old IEEE 488 instrument-control and communication standard is rocking the usually staid test-and-measurement field. A working group chaired by a National Instruments (Austin, TX) R&D manager wants to add a high-speed mode to IEEE 488.1, a standard that is also known as IEC 625-1. At the heart of the proposal is a proprietary protocol, HS488, which National Instruments (NI) announced in 1993 and has implemented in IEEE 488 controller ICs and controller boards that use those ICs. Critics hope to delay or prevent an early-1998 ballot on the proposal. They say that the working group hasn't adequately demonstrated the enhancements' compatibility with the existing standard.

The working group claims that the 8-Mbyte/sec high-speed mode would in no way interfere with the existing IEEE 488.1 protocol, which limits data transfers to about 1 Mbyte/sec. According to a group spokesman, the proposed standard would supplement but not replace IEEE 488.1; no manufacturer would have to cease supplying devices that support only the 1-Mbyte/sec protocol. Second, the spokesman says, any device that implements the new protocol would still have to support the old one. Indeed, such devices would default to the 1-Mbyte/sec mode.

These assurances and NI's four-year track record with the essentially identical HS488 have not persuaded the critics. They say they want to see documentation of the compatibility-test protocols and more complete documentation of the results of compatibility testing than the working group has so far provided.

One company that opposes the adoption of the working group proposal is Hewlett-Packard (Palo Alto, CA) (Reference 1). HP began in 1972 to market instruments that support the protocol that has evolved into IEEE 488.1 and IEC 625-1. The company dominates the benchtop-instrument market and probably offers more products that implement IEEE 488.1 and has a larger installed base of such instruments than all other suppliers combined.

HP is not the only critic, however. You can find an extensive analysis at http://ourworld.compuserve.com/homepages/acea/HS_CONCL.htm.   This Web page not only criticizes the proposed changes, but also suggests that so few instrumentation applications use the full speed of the standard that a higher speed version is unnecessary. The page's most negative comments, which appear near the beginning, can leave a false impression: If any device connected to the bus does not support the high-speed protocol, high-speed transactions are impossible. Later, the writer provides a somewhat different picture: If cable lengths are within the proposal's limits, transactions involving only high-speed devices can proceed at high speed, even if other devices in the system don't support the high-speed protocol.

HP suggests that those who need greater speed than the current standard provides should adopt newer standards, such as IEEE 1394 (Firewire). Tom DeSantis, president of IOtech (Cleveland), whose representatives have attended some of the working-group meetings, echoes this sentiment. DeSantis remarks that he supports no modification of IEEE 488.1 because the current version works and the proposed modifications could confuse prospective users. However, as NI representatives point out, IEEE 488 is so popular that its continued use well into the next century is certain. Although NI supports newer standards and is working on applying them to instrumentation, the company is certain that many users want IEEE 488 extended to higher speeds. Indeed, some users routinely achieve faster operation by modifying off-the-shelf products in ways that violate the standard.

Officially, no company can support or oppose a change in an IEEE standard because the IEEE is an organization of individuals, and it accepts no corporate memberships. Those who vote on IEEE standards are individuals qualified in the area of the standard who request ballots and have the time to review and comment on the proposals. Nevertheless, most of those members work for corporations, many of which have a great deal to gain or lose from the proposal's approval or rejection.

IEEE 488.1's dual nature as both an IEEE and an IEC standard adds complexity to approving a change. IEEE approval by no means assures IEC approval, even though IEC members strongly favor maintaining consistency between the IEEE and IEC standards. The IEC version carries the greatest weight in Europe--especially Germany--where sentiment against a change reportedly runs high. European companies worry about having to ship different versions of products to different parts of the world--a possibility if the IEEE but not the IEC approved a change. Therefore, several European companies are expected to do whatever they can to ensure that the IEEE rejects the proposed standard.

By press time, the IEEE had received but had not yet mailed an unusually large number (estimated at several hundred) of requests for ballots. Those who oppose the proposal hope that this intense interest will persuade the working group to abandon its proposal or at least to provide additional documentation of its compatibility tests. Others hope that the working group will add IEEE members who work for companies that make IEEE 488 controller ICs. Philip Fleming, PhD, a Colorado Springs, CO consultant, who is not a member of the working group, says that greater participation by such individuals could result in a simpler controller design. Such a design would be easier to understand yet could achieve the same objectives.

Meanwhile, Robert Canik, the working group chairman hopes for prompt balloting and looks forward to early adoption of the proposal. Bob Rassa, chairman of technical standards activities for the IEEE Instrumentation and Measurement Society, is also eager to proceed with the ballot and is critical of what he characterizes as inappropriate lobbying by companies interested in IEEE 488.1.

--by Dan Strassberg

Reference

1. Richey, Dave, "Will HS488 change the 25-year-old IEEE 488.1 standard for better or worse?" Personal Engineering, November 1997, pg 9.

Flash memory hits a new low

AMD's AM29SL800, also available from Fujitsu (San Jose, CA), is the first NOR flash memory that supports both reads and writes at voltages as low as 1.8V. The device's hardware-, software, and architectural backward compatibility with previous 2.2, 2.7, and 5V versions simplifies design migrations. Automatic power-down circuitry lowers typical sleep-mode current draw to 65 nA between accesses. Typical read and program/erase currents are only 5 and 10 mA, respectively.

The currently available AM29SL800 speed bin, primarily targeting system prototyping, delivers 170-nsec read performance across the ­40 to +85ºC operating-temperature range. AMD hopes to reduce the read-access time to 100 nsec by the end of the first quarter of next year via design optimizations. The company also plans to sample a 120-nsec, 16-Mbit version in the third quarter of next year and to follow that with a 120-nsec, 32-Mbit version. AMD also plans to offer 1.8V, simultaneous-read/ write flash memories.

The AM29SL800 comes in a 48-lead TSOP and a 48-bump, 0.8-mm-pitch BGA, similar to Sharp's (Camas, WA) CSP package. AMD, once a strong µBGA-package advocate, now believes that encapsulated FBGA offers several significant advantages (see "Flash memory goes smaller with BGAs," EDN, Jan 16, 1997, pg 18). These advantages include easier handling and board manufacturing; a more mature and cost-effective packaging technology; and a pitch and pinout that are insensitive to die density, design, and lithography changes.

FBGA might even deliver a smaller board footprint than other vendor's µBGA alternatives, depending on die size, although µBGA gives the lowest package height. The extended-temperature, 170-nsec, TSOP AM29SL800-170EI in both top- and bottom-boot versions costs $8.70 (10,000). The commercial-temperature version costs $8.35 (10,000). Contact the company for the prices of higher speed and FBGA versions.

--by Brian Dipert

AMD Corp, Sunnyvale, CA. 1-408-732-2400, fax 1-408-749-3240, www.amd.com.

Design-contest victor wins BMW Z3 roadster

With his design for a small, stabilized antenna for direct TV reception in mobile vehicles, Vincent J Tarricone of Middleton, RI, has won the XA Performance Challenge and will drive home in a $30,000 BMW Z3 roadster. Tarricone's winning submission, the TracVision II, uses four Philips XA microcontrollers and two 80C320s to control two stepper motors, to perform the axis transforms from the earth's coordinate frame to the antenna frame, to control the gyros, and to provide overall system control.

Respective second-, third-, and fourth-prize winners of $2500 IBM Thinkpad notebook computers are Greg Ingalsbe of Innovation Racing Products (Lanexa, VA), Sui Lam of Lucent Technologies (Woodcliff Lake, NJ), and Jack and Mark Nowinski (Waterloo, ON, Canada). Ingalsbe used the XA to design a real-time digital data-logging tachometer that measures 500 to 20,000 rpm with 10-rpm resolution and accuracy. Lam's design, a low-cost RS-232C networked controller for a hospital airflow/particulate-filtering control system, uses the dual-UART on the XA to daisy-chain the RS-232C channel. Fourth prize goes to the Nowinskis for a design that measures colors and verbally communicates the measured values.

The design contest, which Philips Semiconductors creates and EDN co-sponsors, challenged engineers to create an application that highlights the features and capabilities of the Philips' XA 16-bit microcontroller. The judges--a panel of Philips' engineers and EDN editors--based their decision on how well the entries demonstrated the XA's performance, by the innovation of the design, and by how effectively the design integrated peripheral functions into the XA and lowered part counts. The judges awarded additional points on how well the design demonstrated the upward compatibility of 8051 code on the XA and on how effectively the design demonstrated the XA's multitasking capabilities.

--by Michael Markowitz

Conference tackles testing of system-level ICs

"BIST will be tomorrow's technology forever," said the CEO of one ATE manufacturer.

As the number of gates on complex ICs climbs without apparent limit, so does the number of analog functions on those ICs. As they have done for several years, manufacturers of ICs, IC testers, and EDA tools are expressing growing concern over testing of ever-more-complex devices. The manufacturers' level of consternation shrinks, however, when you compare it with that of the digital designers and test engineers who must deal for the first time with analog functions.

Those issues formed the backdrop and major focus of the 28th annual International Test Conference (ITC), which convened from Nov 2 though Nov 5 at the Sheraton Washington Hotel and Conference Center in Washington. Despite financial turmoil in East Asia, the home of most IC-fabrication facilities, the boom-or-bust order rate for IC automatic-test equipment (ATE) remains strong. Because these favorable conditions have loosened company purse strings, conference attendance exceeded 4000, a jump of more than 25% from the levels of recent years.

The keynote speaker, Jim Healy, is president and CEO of Genus Corp (Sunnyvale, CA), which makes thin-film-deposition and ion-implantation equipment. Healy, a veteran of more than three decades in ATE, is now a test-industry alumnus. He echoed a familiar theme: Testing higher clock-rate, gate-count, pin-count devices at costs tolerable to semiconductor manufacturers requires test-system architectures that differ radically from those of current testers. Nevertheless, Healy largely dismissed the only such architecture that has so far emerged--built-in self test (BIST). Indeed, the CEO of one ATE manufacturer proclaimed that we will never see BIST implementations except in a few limited niches. He also stated that neither our children nor their children would see BIST implemented in their lifetimes.

"BIST will be next year's technology forever," he said. When reminded of the many years that CMOS had been next year's technology before it emerged as the uncontested leader in IC fabrication, this CEO merely asked the questioner whether he was looking for a fight.

Meanwhile, although their BIST sales remain relatively modest, EDA companies that promote BIST continued to announce advances. Some companies, such as HPL Inc (Milpitas, CA) and OpMaxx (Beaverton, OR), do most of their work behind the scenes. Others try to attract the maximum possible attention. In the second category are Mentor Graphics (Wilsonville, OR) and LogicVision (San Jose, CA). LogicVision demonstrated its mixed-signal BIST capabilities, which, when embedded in a chip, allow analog and digital PLLs and successive-approximation and flash ADCs to test themselves with minimal help from purely digital ATE (see "Digital BIST cells boost mixed-signal chip testing," EDN, Nov 20, 1997, pg 16).

Activity in two other areas was particularly noteworthy. Nearly all IC ATE companies--even those that don't make testers for commodity-RAM ICs--announced test capabilities for RAMbus parts. According to one vendor, Intel's (Santa Clara, CA) endorsement of RAMbus caught the ATE industry off guard, and the test-equipment manufacturers have been scrambling ever since to show that they can test 800-Mbyte/sec RAMbus parts. Companies that don't make RAM testers feel that they still must have RAMbus-test capabilities. These companies believe that their customers will need to test many non-RAM ICs that either generate or interface with the RAMbus. In fact, some ITC attendees speculated that RAMbus might succeed PCI as the general-purpose interconnect bus in PCs. Credence Systems (Fremont, CA) boasted that it was one of the first companies--not just in ATE but in any industry--to appreciate the potential of RAMbus. CEO Bill Bottoms, PhD, remarked that Credence's early support of RAMbus enables the company to use the technology royalty-free. According to Bottoms, even Intel must pay royalties on its RAMbus parts.

A spirited panel discussion dealt with the analog boundary-scan standard still known as IEEE P1149.4. (The P will disappear when the standard is ratified, most likely in the spring.) Although the panelists who favor the standard emphasized that it is no panacea, an opposing panelist presented a list of problems that he had uncovered. For example, he observed that ICs that implement the standard can latch up. You can induce this destructive condition by connecting voltages only slightly greater than the supply voltages to the pins of the devices' analog transmission gates. Although another panelist demolished these arguments one by one, his solution to the latchup problem was to avoid applying excessive voltages. Because the standard supposedly enables testing at the device, board, and system levels and allows system-level testing in the field as well as the factory, that caveat may be unrealistic.

Testing of embedded µP cores and the ICs that contain them was another hot topic and will be the theme of next year's ITC. Until recently, suppliers have referred to these cores as "intellectual property" (IP), a name whose vagueness makes some industry observers uneasy. A new name, "virtual components" (VCs), seems more specific and more appropriate. A panel discussion brought out some potential problems with the approaches that most test experts thought would form the basis for embedded-core test. One problem stems from the likelihood that cores will be embedded several levels deep--cores within cores within cores. In chips that use such architectures, tests based on IEEE-1149.1 boundary-scan collars (rings of scannable latches surrounding embedded cores) appear to be too slow for production-IC testing.

Although ATE has become an industry populated almost exclusively by large companies, smaller suppliers still exist. Usually, their strategy is to exploit market niches that larger competitors find unattractive. One such company is TMT (Sunnyvale, CA), which announced a low-cost tester for RF and wireless-communication ICs. TMT says its prices are about 40% those of other systems that can test similar ICs. According to TMT, it achieved the low costs by tailoring its system to production test and sacrificing features that would suit its testers to use in quality assurance and R&D.

The list of ITC exhibitors is not limited to suppliers of ATE and EDA tools. Some vendors showed instruments that can become part of larger systems. Wavecrest (Edina, MN) is in that category. The company's DTS-2075 measures jitter on digital waveforms at frequencies as high as 800 MHz. Resolution is 0.8 psec. The unit displays its results in a variety of ways. Among the displays are statistics and period, frequency, and phase vs time. At least one ATE vendor is integrating Wavecrest's units into its digital-IC testers.

--by Dan Strassberg

Custom, standard ICs attack RF challenges

As you consider meeting your RF-design requirements using standard ICs, chip sets, or custom ICs, the choices in each category continue to proliferate ("RF-component diversity complicates design," EDN, Oct 23, 1997, pg 109). VLSI Technology's OneC IC is an IC platform with associated DSP-code packages that incorporates most of the functions for GSM (Global System for Mobile communications) cellular phones and similar RF-intensive products. Occupying a 12×12-mm board space in a 144- or 176-lead fine-pitch BGA, the modular device integrates a voiceband analog front end, an RF analog interface, keyboard- and display-I/O and peripheral-device drivers, and two processor cores. You can add other functional blocks, such as a UART, an IrDA interface, or an analog interface to customize the IC. By building on previous generations of ICs with Final Type Approval, this silicon product should minimize regulatory issues.

The OneC contains Vector ARMThumb and Vector Oak cores, along with various GSM algorithms, including enhanced-full-rate, half-rate, and full-rate voice coding; V.42bis data compression; and other data-service formats. Development and debugging tools that VLSI developed as part of its Communication Standard Platform allow the OEM to combine blocks as needed with a typical three- to six-month design cycle. Nominal battery life for the 3.6V IC when operating from a 1200-mAhr battery pack is 7 hours (talk time) and 500 hours (standby). Representative price is less than $15 (1 million).

If your RF-design preference for personal communications services and digital cordless-phone applications leans toward off-the-shelf ICs, consider a subset of eight silicon ICs from Texas Instruments that covers the receiver front end, synthesizer, modulator, and driver/power-amplifier functions. The group of devices (all prices in 10,000-unit quantities) comprises:

• TRF1015DC, a 20-pin, SSOP, 900-MHz, $1.35 (10,000) receiver front end, with a mixer, low-noise amplifier (LNA), and VCO;

• TRF1500PFB, a 48-pin TQFP, $4.17 dual-mode receiver for 900-MHz analog/digital and 1900-MHz digital systems, with an LNA, RF mixer, and IF amplifier for each band;

• TRF2020PW, a 24-pin TSSOP, $2.54, 1.2-GHz synthesizer for GSM, with one main and two auxiliary synthesizers and a serial-data-bus interface;

• TRF2050PW, a 20-pin TSSOP, $3.64, 1.1-GHz synthesizer for RF and IF PLLs that operates in fractional- or integer-N modes;

• TRF2052PW, a 20-pin TSSOP, $386, DECT synthesizer for RF and IF PLLs;

• TRF7003PK, an SOT-89, $1.24, 1-GHz discrete MOSFET power amplifier operating from a 3.6 to 4.8V supply, with 32.5-dBm output at 900 MHz; and

• TRF8010PWP and TRF8011PWP, a pair of thermal-pad TSSOP, 900-MHz driver/amplifiers for AMPS or GSM. The $1.43 TRF8010PWP has transmit enable/disable, 24-dBm output at 4.8V, and linear ramp control of power; the TRF8011PWP is similar but differs in some specifications and features.

Most of the devices operate from 3.3 to 3.6V supplies; some also can operate from lower voltages to minimize power consumption or from higher voltages to increase power output and efficiency.

--by Bill Schweber

VLSI Technology Inc, San Jose, CA. 1-408-929-5250, fax 1-408-922-5155, www.vlsi.com

Texas Instruments Inc, Dallas, TX. 1-800-477-8924 ext 4500, www.ti.com.

Figure 1

You can use the appropriate RF-function building blocks from Texas Instruments to build PCS and cordless-phone applications.

Autoranging-rectifier power takes twofold leap

VI-ARM autoranging-rectifier modules from Vicor offer twice the output power of existing models in the same micro-size package. Output-power ratings are 500 and 1000W. The modules house the entire front end of a switching power supply in a package measuring 36.8×57.9×12.7 mm. You simply add a filter and a holdup capacitor whose values are appropriate for the application. The modules include a controller that continuously tracks the ac line to ensure correct strapping of the 110/220V doubler. The VI-ARM family has UL, CSA, TüV, VDE, IEC950, CE, and BABT approvals. Single-piece prices are $49 for the 500W VI-ARM-C12 and $75 for the 1000W VI-ARM-C22.

--by Bill Travis

Vicor Corp, Andover, MA. 1-800-735-6200, fax 1-508-475-6715, vicorexp@vicr.com.

Comdex '97 delivers, despite long days, longer nights, and even longer lines

The annual Comdex trade show held last month in Las Vegas yielded the usual bounty of exciting new technologies. Unfortunately, only the future will tell just which companies will deliver on their aggressive promises. Serial interfaces continue to garner headlines but remain a $0 billion market. New products are threatening to overheat the battle for leadership in the removable-storage market. Flat-panel displays appear finally ready to challenge CRTs. And a new de facto market standard has emerged in handheld computing--a technology that's gaining popularity so fast that it's reminiscent of the original IBM PC.

For the third consecutive year, trade-association pavilions prominently featured two desktop-serial-bus technologies--the 12-Mbps USB and the 400-Mbps IEEE 1394 Firewire interface. Unfortunately, just like last year, manufacturers will not widely deploy either technology during this year's holiday season. Several things have conspired to delay that deployment.

For starters, Microsoft has delayed shipment of the next major Windows release until mid-1998. That Windows 98 release presumably will include drivers for popular USB and 1394 host-interface ICs and generic device-class drivers for still cameras, audio devices, streaming videocameras, and other peripherals. Although some peripheral vendors have the budgets and manpower to quickly develop drivers for these new interfaces, many others are waiting for Microsoft. Rumors among USB and 1394 IC vendors suggest, however, that Microsoft may be unable to deliver working, generic, device-class drivers. The drivers for USB can be especially difficult to develop, because the interface relies on the host CPU for all operations, whereas 1394 peripherals operate as intelligent peer nodes with the host.

At Comdex, the USB Implementers forum (www.usb.org) claimed that manufacturers would ship 20 products for the holiday season. A post-Comdex search through Fry's Electronics and Computer City in San Diego, however, turned up only one shipping USB peripheral. Much of the delay is the result of software-driver issues. USB-IC vendors at Comdex privately conceded that volume shipments of the peripherals that use their ICs were most likely delayed until the '98 holiday buying season.

Drivers aren't the only problem in the USB camp. The interface has a maximum data rate of 12 Mbps, but tests show that realizable sustained data rates will be less than half the maximum. EDN Technical Editor Markus Levy found the maximum rate to be about 750 kbytes/sec (see "Unveiling the hidden secrets of PC-bus architectures," EDN, Dec 4, 1997, pg 112). According to some peripheral and IC vendors that have tested USB printers and scanners, the realizable rate may be less than 600 kbytes/sec in typical systems with multiple peripherals. On a heavily loaded bus, the data rate could be much lower. Meanwhile, the enhanced parallel port on most new PCs delivers more than 700 kbytes/sec in a point-to-point fashion. The plug-and-play aspects of USB may be a welcome feature for peripheral vendors, but the performance will in many cases be a step backward and may prove insufficient for office-quality printers and scanners. USB certainly won't suffice for most video and CD-quality-audio applications.

Although the 1394 camp shares the software-driver concerns of USB, the much faster interface appears poised to move ahead in applications ranging from scanners and printers to videocameras to disk drives. Popular perception would have you believe that 1394 lags behind USB by one or two years, but some vendors have been shipping 1394-based video-editing systems for almost a year. Moreover, many vendors have just announced third-generation, physical-layer (PHY) 1394 ICs delivering full 400-Mbps data rates--a critical milestone for high-speed applications, such as disk drives.

Symbios Logic (Fort Collins, CO) led off with the SYM13FW403 PHY for host applications, which implements three 400-Mbps ports. The company also demonstrated target-side, 400-Mbps ICs, including a 48× CD-ROM decoder and an ATAPI/1394 disk bridge. Adaptec (Milpitas, CA) announced a disk-drive chip set, including DSP, servo control, and a 400-Mbps interface. Lucent Technologies (Berkeley Heights, NJ), meanwhile, had perhaps the most compelling host-side announcement with a six-port, 400-Mbps PHY IC. The Lucent FW810 targets applications such as remote-access servers, in which designers see the hot-plug 1394 interface as superior to PCI or ISA for connecting multiport modem boards or digital communication boards. Lucent also announced the three-port FW813 for more typical desktop applications. Texas Instruments (Denver) joined with a 400-Mbps PHY IC and a 1394 link-layer controller designed to Microsoft's Open Host Controller Interface spec. Fujitsu Microelectronics (San Jose, CA) showed not only 400-Mbps ICs, but also prototypes of an 800-Mbps IC that it plans to deliver next year.

One of the arguments against using 1394 in scanners and other cost-sensitive applications centers on price. These IC announcements, however, prove that 1394 ICs can meet low-price requirements. Symbios claims it will sell its SYM13FW403 for less than $6 in high volume, and even the six-port Lucent offering costs only $20 (1000). New peripheral and system products were also on display in the 1394 Trade Association pavilion. These products included cameras and FireNet from Unibrain (Athens, Greece) that routes Ethernet traffic over 200-Mbps or faster 1394 links. For more information, see www.1394ta.org.

Fibre Channel is finally viable

While we await widespread USB and 1394 deployment, one serial interface appears ready to deliver on its promise for high-speed data-storage subsystems. The Fibre Channel Loop Community (FCLC) pavilion was full of shipping products ranging from ICs to boards to disk drives to switch fabrics to bridges. Seagate (Scotts Valley, CA) has been shipping Fibre Channel Arbitrated Loop (FC-AL) disk drives for some time and now claims equal prices on FC-AL and SCSI drives. In reality, SCSI volumes will surely result in slightly lower prices, but the FC-AL premium is shrinking. Today, you can buy virtually identical SCSI and FC-AL controllers from companies such as Qlogic (Costa Mesa, CA) and virtually identical SCSI and FC-AL disk drives from Seagate. Build a RAID subsystem with the SCSI and FC-AL products, and you will likely pay a 20 to 30% premium for the FC-AL subsystem. And, although performance depends on more than the interface data rate, the FC-AL supports maximum rates more than double those of available SCSI drives. Hitachi (Brisbane, CA) and Quantum (Milpitas, CA) stand ready to join Seagate supplying FC-AL drives in the first quarter of '98. To learn the full scope of Fibre-Channel-based products, check the trade association's Web site at www.symbios.com/fclf.

Not all I/O subsystem news focuses on hardware interfaces this year, however, as the Intelligent I/O (I₂O) initiative, which Intel (Santa Clara, CA) instigated, starts to mature. I₂O presumably will add an OS-independent, industry-standard abstraction layer in µP-based I/O subsystems. The I/O subsystem would then work with any OS that has a standard I₂O driver. A number of companies announced I₂O boards, software, and systems, but no interoperability tests exist. Moreover, the I₂O community cannot agree on a marketing theme. Intel is promoting I₂O as a way to boost system performance as if intelligent-I/O systems didn't exist before I₂O. Most other I₂O vendors are promoting the standard as a way to speed intelligent-I/O-based systems to market while making them  easier to support and maintain. Watch for the March 2, 1998, issue of EDN for a detailed report on I₂O, and see www.i2osig.org.

In other data-storage news, the digital-versatile-disk (DVD) industry seems to have weathered the recent publicity that occurred when Sony (San Jose, CA) and Philips (Sunnyvale, CA) announced a proprietary, rewritable DVD format. Hitachi is already shipping rewritable DVD drives that meet the format that the DVD Forum published (see "DVD: breathtaking sight and sound, significant challenges," EDN Aug 15, 1997, pg 47). Moreover, Toshiba (Irvine, CA) and Panasonic (Secaucus, NJ) plan to ship such drives in the first quarter. Sony and Philips appear to be behind in DVD and had possibly hoped to slow the industry with news of the new format. Moreover, Sony and Philips are accustomed to generating significant royalties because of the use of their intellectual property in CD-ROM drives. The companies have desperately tried to gain the same advantage in DVD technology, but it appears that the published DVD standards will persevere. As it stands, about 15 companies have contributed intellectual property to DVD-drive standards.

Rewritable DVD may not make it into every PC, but some 100-Mbyte or larger device with removable media will. Iomega (Roy, UT) and its Zip drive still appears to be winning the battle to replace the floppy over rivals such as the LS-120 drive championed by Imation (Oakdale, MN), Mitsubishi (Sunnyvale, CA), and OR Technology (Campbell, CA). OR Technology is in dire financial straits, and Mitsubishi has stepped up to supply the 120-Mbyte drives that are backward-compatible with 1.4-Mbyte floppies. Still, Iomega has the more impressive list of OEMs shipping Zip drives. Zip is much faster and has retained a cost advantage. With rebates, Zip drives can sell for $85, meaning that you can buy a Zip drive and a floppy drive for less than you would pay for an LS-120 drive. Moreover, LS-120 drives still have problems as the boot device in some systems, meaning that you still need a floppy. NEC (Melville, NY) has also begun manufacturing, shipping, and marketing Zip drives. (For more information on the battle, see "The final showdown begins for a floppy replacement," EDN, Jan 16, 1997, pg 63 and "Removable storage drives shake floppy foundation," EDN, Jan 18, 1996, pg 40).

Floppy-replacement battles

The real challenges to Zip are much higher capacity drives because, like the US dollar in times of inflation, a megabyte seems to stores less every day. Sony and Fujitsu (San Jose, CA) have announced a 200-Mbyte drive that's much faster than the LS-120 yet still backward-compatible with the 1.4-Mbyte floppy. Working drives, however, weren't in evidence on the Comdex show floor. It's unclear whether drives can reliably store much more than 100 Mbytes on relatively low-cost media.

The higher capacity removable drives that are garnering attention use rigid media and hard-disk technologies, and the action in these drives at Comdex was fast and furious. For several months, Iomega and Syquest (Fremont, CA) have fought a battle on the store shelves with the 1-Gbyte Jaz and 1.5-Gbyte Syjet, respectively. Syquest has matched Iomega's Jaz price at approximately $300 in computer superstores and offered a second free 1.5-Gbyte cartridge. At Comdex, Iomega showed a 2-Gbyte Jaz drive that will retail for approximately $600, with disks selling for $150 each in a three pack. Syquest, meanwhile, countered Iomega's offering both high and low. The company announced SparQ, a scaled-down, 1-Gbyte version of the Syjet, which will sell for $200, and Quest, a 4.7-Gbyte, high-end drive, which will sell for $599.

It was a 40-Mbyte drive, however, that literally made the most noise at Comdex. Iomega handed out thousands of annoying clickers to celebrate the Clik! drive that uses media measuring approximately 2 in. sq and that makes an audible click when you load the media. The drive mechanism occupies the footprint of a PC Card but measures 6.5 mm high, so it can't fit into a Type II PC Card slot. Iomega claims that a subsequent design will meet the 5-mm Type II height requirement. The company plans to offer OEMs both the bare-drive mechanism and a battery-powered, handheld unit that looks much like a cellular phone. The external unit can connect to a host via a docking station or a direct-cable connection. Iomega claims the drive will sell for less than $200 at retail, and the disks will cost less than $10. The company is promoting the product for use with digital cameras, handheld computers, and other mobile applications. Remember, however, Clik! is a revamped version of n-hand, a similar, 20-Mbyte product that Iomega introduced a year ago but never shipped. It remains to be seen whether Iomega has the expertise to master the miniature manufacturing process. The company is using Citizen Electronics (Japan) as a partner, and Citizen successfully built Hewlett-Packard (Boise, ID) 1.3-in. hard drives that never found market success.

AGP shines in 3-D applications

Iomega also stole the show with a cast of Broadway dancers and actors performing eight skits per day to standing-room-only audiences, although I'm not sure those performances sold any disk drives. More visually appealing to the technologist, 3-D graphics appear to have significantly improved over last year. The graphics-IC companies have continued to improve hardware rendering engines, and Intel's Accelerated Graphics Port (AGP) initiative has proved an unqualified success. A number of the most compelling games that will ship in the next quarter require AGP. The technology allows programmers to store 3-D textures in main memory rather than the frame buffer, providing access to more textures and in turn yielding more compelling images.

With its AGP-enabled Rage Pro IC, ATI Technologies (Thornhill, ON, Canada) demonstrated by far the most compelling 3-D graphics for mainstream systems--approaching the images generated on high-end, 3-D-only cards. ATI was first to discover how to effectively use an optional-AGP capability to execute textures directly from main memory. Most other early implementations use DMA to transfer textures from system memory to the frame buffer as needed.

With 3-D and video-enabled graphics cards maturing, you might wonder how best to display those images. For the first time at Comdex, I saw TFT-based LCD monitors that were superior to CRT-based monitors, even when displaying full-motion video. These new panels have incredible viewing angles and work flawlessly in the less-than-ideal Comdex lighting. Samsung Electronics (San Jose, CA) panels driven by the LCD-version of ATI's Rage Pro yield arguably the best images. Nokia Display Product's (Sausalito, CA) initial foray into LCD was a close second. These panels measure a little more than 13 in. diagonally but deliver an image size equal to that of 15-in. CRTs. At 1024×768-pixel resolution and 16 million or more colors, the images are stunning. Moreover, the panels take next to no room on a desktop. Unfortunately, the units still cost more than $2000, so they won't replace the CRT this year or next.

Other than Iomega's stage show, the biggest crowds gathered at the ubiquitous handheld-computer exhibits that were seemingly in every booth. Only a couple, however, consistently drew crowds: the PalmPilot from 3Com (Mountain View, CA) and the Rolodex Electronics (REX) PC Companion from Franklin Electronic Publishers (Burlington, NJ). The PalmPilot was featured in the company's main booth and in a dedicated PalmPilot booth in the Las Vegas Hilton. Both exhibits remained jammed throughout the show.

The PalmPilot has generated more enthusiasm among users and developers than has any system since the PC. According to Dataquest (San Jose, CA), the PalmPilot owns 66% market share, and Windows CE-based systems have garnered only 20% market share. Even more amazing, 3Com has 3500 registered developers working on PalmPilot applications and has sold more than 5000 software-development kits. Applications for the PalmPilot start with the typical, such as address books, and range to the more unusual, such as book readers; Web browsers; accounting packages; and vertical applications in medical, legal, and other fields.

Just before Comdex, 3Com announced the first of what should be many OEM partners. IBM (White Plains, NY) will sell Workpad PC Companion, a customized version of the PalmPilot. Symbol Technologies (Holtsville, NY), which specializes in scanning, wireless-data transactions and point-of-sale applications, has agreed to develop custom vertical products based on a PalmPilot core. Finally, 3Com, Motorola (Phoenix), and PageMart Wireless (Dallas) agreed to partner on a offering a $169 wireless messaging card for the PalmPilot and offer standard paging as well as services, such as stock quotes. It was interesting to meet with conglomerates pushing other handheld devices and µPs for other devices and almost invariably to witness their executives using PalmPilots.

It's hard for me to imagine that anything could derail the PalmPilot, but, if anything can, it will likely be something smaller than the 4.7-in.-tall, 6-oz PalmPilot rather than something larger with a keyboard. The other Comdex exhibit that drew exceptional crowds featured just such a miniature device. Franklin Electronic Publishers' REX PC Companion is the size of a Type II PC Card. An LCD with only five buttons for controlling the device occupies most of the upper side of the card. Unlike the PalmPilot, which allows pen input, the REX PC Companion is read-only. And, despite the limited number of controls and the lack of a pointing device, the REX makes it surprisingly easy to look up names and addresses and to view the calendar.

Both the PalmPilot and the REX PC Companion allow you to automatically synchronize your notebook- or desktop-based database. Therefore, you perform data-entry chores with a full-sized keyboard. I prefer the pen-input PalmPilot but could certainly get used to the shirt-pocket size of the REX PC Companion. The REX also slips directly into a PC Card slot for synchronization. You can buy a direct-cable connection now for PalmPilot synchronization as an alternative to the standard desktop cradle that doesn't fit easily into briefcases. The REX PC Companion ranges in price from $129 to $179, and prices for the PalmPilot start at $299, but the PalmPilot stores more data entries. Moreover, you can buy an optional modem for the PalmPilot. The deciding factor, however, will prove to be software, and, for now, 3Com has the developers waiting in line.

Speaking of lines, they were everywhere--from restrooms to food stands to transportation centers. Without a chauffeured vehicle, you couldn't leave the main show area or the adjacent Hilton from 4 to 7 pm unless you walked or waited an hour for a bus or cab. The show organizers fell short in other areas, as well. Someone decided that the press room needed no modems because the computers provided there were connected directly to the Internet. Unfortunately, most of the press there couldn't check their e-mail without a modem. Furthermore, the show organizers moved the press facilities and some exhibits into parking-lot tents, which were fine except for the fact that the nearest restrooms were more than a quarter-mile away. To continue to thrive, Comdex simply must conceive a better plan for the decidedly difficult task of handling a quarter-million people.

--by Maury Wright

| EDN Access | Feedback | Table of Contents |