Design Features

Choices and confusion spread wider as spread spectrum goes mainstream

Bill Schweber, Technical Editor

ICs, subsystems, and modules for spread-spectrum communications are now widely available. However, before committing to the technology, you should resolve issues of system performance, complexity, compatibility, and interoperability.

 
Spread-spectrum-communications designs are finding use in wireless applications ranging from cordless phones, through handheld roaming inventory units, to high-speed LANs. The technology lets you build a reliable system that maximizes use of total bandwidth, lets you easily add users with only minor performance degradation (within a reasonable limit), and work despite noise and interference. The systems also provide data security: To an intended receiver, the desired signal stands out; to unintentional receivers without the proper code for correlation, the spread-spectrum signals appear as noise.

  What makes spread spectrum practical now is the integration of its many complex but necessary functions—including node management and control—into a few ICs and the resulting cost reduction. Vendors are offering building-block ICs, chip sets, modules, subsystems, and complete systems. You can also get spread-spectrum PC plug-in or PCMCIA cards and dedicated boxes.

  Yet, despite these features and its wide acceptance, spread spectrum violates some traditional communication-system-design rules. For example, it complicates design by flagrantly using precious bandwidth, and it blurs limited transmitter power across the spectrum. It's also difficult to quantify spread spectrum's throughput. Receiving and decoding a spread signal requires use of autocorrelation techniques typically used in wideband analog and high-speed digital design.

  Further adding to the confusion, the technique has two distinct variations: direct-sequence spread spectrum (DSSS) and frequency-hopping spread spectrum (FHSS). Most vendors that support just one of the techniques proclaim that the other has serious flaws (see box, "How do you like to hop?"). You should carefully look at vendors' claims and understand the virtues and shortcomings of the systems and architectures.

Standards set the pace?

  Don't be deceived: Spread spectrum has complicated hardware and algorithms, and a reliable design is not trivial. There's no doubt that spread spectrum can solve some difficult communications problems, but it comes with additional circuitry cost and potential frustration due to its inherent complexity when compared to traditional, non-spread-spectrum designs.

  The good news is that the industry realized early that a complex, algorithm-based architecture needs a standard defining frequencies, format, protocol, and other issues. The not- so-good news is that the nearly completed spread-spectrum standard for wireless LANs, IEEE 802.11, allows for many variations (see box, "It's a very inclusive standard"). You can comply with the 802.11 standard and yet be incompatible with another system that is also in accord with the standard.

  IEEE 802.11 does provide a start, however. Because the standard offers you choices, you must evaluate your communications burden and critical operating parameters as you pick the best combinations and manage latency, power, bit error rate (BER), and similar factors. For example, packet size, synchronization time, and format overhead affect overall throughput, error tolerance, security, and how the system behaves as additional users become active.

  Each choice brings trade-offs. Consider the pseudorandom-noise (PN) code pattern used by the system to uniquely identify specific users, as well as to establish data synchronization via autocorrelation and provide security against unauthorized listeners. A longer pattern supports more users and provides greater security. However, it also requires a longer time to establish the correlation, which adds to overhead and decreases effective data throughput. The 802.11 standard allows for PN codes that range to as many as 11 bits long.

  Not all vendors, especially those that provide complete subsystems and modules, are following the 802.11 specification. There are a variety of reasons for this. Many vendors feel that the standard is too late, is too complex, has too many options, or is technically excessive for lower speed applications, such as industrial data links.

  Digital Wireless Corp, for example, offers both DSSS and FHSS modules and subsystems. The company feels that DSSS is better suited for the 915-MHz band and that FHSS offers technical advantages at 2.4-GHz, where the spread bandwidth is larger and, therefore, it is harder to avoid interference sources. (Again, many 802.11-compatible vendors feel that DSSS is actually better than FHSS at handling such interference.) For DSSS, the company employs a proprietary modulation technique called "recombinant spread spectrum" to minimize the impact of multipath, using a chipping sequence that causes continual amplitude transitions in the transmitted sequence; therefore, the multipath-delayed signal can't completely cancel the direct signal. In the IC world, Micron Communications has a system on a chip targeted for applications of smart RFID-tag applications, where low power and limited but well-defined messages and protocols are dominant factors.

Assessing actual performance

Like all communications techniques, spread-spectrum systems with various modulation schemes and algorithms have been analyzed in great detail by researchers. Figure 1 shows theoretical BER vs S/N ratio for a variety of modulation schemes and data rates.

  Under the 802.11 standard, FHSS supports bit rates as high as 1 Mbps, and DSSS allows rates as high as 2 Mbps. These raw numbers, however, are the ideal upper bound on actual performance. Even under favorable conditions with low noise and few users, you should expect to achieve no more than one-half to two-thirds of those rates. FHSS, for example, requires a guard time between hopping time slots to account for clock errors, as well as a settling time for both sides to switch to and stabilize at the new carrier frequency. Improvements in fast switching, including the use of direct-digital-synthesized carriers, are helping reduce these time-wasting settling periods.

  Theoretical analysis can't account for less-than-perfect circuitry, either. A static tuning error between transmitter and receiver appears as a system offset, which, in turn, can decrease S/N ratio and affect bit threshold levels. Dynamic tuning errors due to temperature drift, supply-rail variations, and other hard-to-define causes further aggravate the situation.

  When choosing between FHSS and DSSS techniques, the most important thing you can do is analyze your needs and overall environment, a guideline that applies to any communications-system design. Quantify factors such as the type of data you are sending (continuous vs burst), your required data rate and effective throughput, typical transmission length, acceptable latency (due to system retransmission when the protocol resends after errors are detected), typical and maximum number of users, and an acceptable type of performance degradation with additional users.

  Even more important, look closely at the operating environment. How well- characterized is the setting? Electrically, is it relatively benign or hostile? Are users roaming continuously or in relatively fixed locations? Are noise and interference sources constant and well-understood, or are they changing and uncontrolled? What control do you have over the number of users? For example, a wireless LAN for an office or campus setting is very different in user patterns and accessibility needs—as well as in your control over them—from a cordless phone for neighborhood residential use.

  All spread-spectrum systems are affected by multipath, which is inherent in any system where frequency modulates a carrier. A signal from transmitter to receiver can simultaneously travel along more than one path, and these paths have differing lengths. Therefore, time-delayed (phase-shifted) versions of the signal arrive and confuse the receiver, causing deep fades of 40 dB or more. This fade increases interference, reduces S/N ratio, and increases BER or the need for retransmission.

  Both FHSS and DSSS claim to be better at resisting multipath: FHSS because the nominal carrier frequency—and, thus, the critical multipath lengths—repeatedly changes and DHSS because it spreads the modulated signal spectrum and, thus, has no single frequency on which it is overdependent. The reality is that protocol, along with error detection and correction techniques, also plays a very significant role in the final performance you achieve.

 If you're looking at implementing a spread-spectrum system, you have many choices for deciding whether to buy or make a system. Unless you have specific requirements for very high volumes, sophisticated implementation, debugging, test skills, and equipment or need the ultimate in performance, you are probably better served by starting with matched ICs (Figure 2), a chip set, ICs with a reference design, or a complete module or assembly. If you're doing a wireless LAN implementation, consider how you will develop both the roving units and the corresponding base station or access point.

You should decide whether compatibility with IEEE 802.11 is critical to your application and whether the standard's potential for interoperability is relevant or worthwhile. Finally, some small but critical details of allowed operation vary from country to country. If you need to install your system worldwide, be sure to study the regulations carefully, engage a regulatory specialist, or use modules or systems that have already been approved in the region of interest.

References

1. Kahn, David, "Cryptology and the origins of spread spectrum," IEEE Spectrum, September 1984.
2. 1996 Wireless Communications Design Seminar Handbook, Harris Semiconductor Corp.
3. Strassberg, Dan, "Powerline communication: wireless technology," EDN, June 6, 1996, pg 71.
4. Strassberg, Dan, "Spread-spectrum communication rises from military roots to star in wireless world," EDN, Dec 22, 1994, pg 59.
5. Gallant, John, "Digital wireless networks," EDN, March 4, 1993, pg 78.
6. Wickelgren, Ingrid J, "Local-area networks go wireless," IEEE Spectrum, September 1996.

Acknowledgments

Thanks to Doug Grant of Analog Devices Inc, Tom Tombler of Harris Semiconductor Corp, and Angela Champness and Jan Haagh of Lucent Technologies for their insight into spread-spectrum systems.

You can reach Technical Editor Bill Schweber at (617) 558-4484, fax (617) 558-4470, bill.schweber@ cahners.com

| EDN Access | feedback | subscribe to EDN! |
| Table of Contents |