While the term 'modulation' is commonly used to describe OFDM, it may be more appropriate to call it a communications system like direct sequence or frequency hopping. It can also be referred to as a transport technology or scheme. Comparisons are often made between OFDM and narrowband systems. However, since OFDM can be either narrowband or wideband, OFDM should only be compared to single-carrier systems.

Basically, OFDM is the concept of multicarrier communications, where the different carriers are orthogonal to each other. Orthogonal in this respect means that the signals are totally independent; it is achieved by ensuring that the carriers are placed exactly at the nulls in the modulation spectra of each other. In OFDM the data on each carrier overlaps the data in adjacent carriers. That overlap is an extra source of spectral efficiency in OFDM.

Another source of OFDM spectral efficiency is the fact that the drop-off of the signal at the band is primarily due to a single carrier that is carrying a low data rate, allowing for sharp edges that correspond closer to the desired rectangular shape of the spectral power density of the signal.

OFDM provides several advantages. It leads to less intersymbol interference than if the overall throughput was attempted on a single-carrier system. Intersymbol interference occurs when a symbol echoes over the propagation channel and distorted replicas of it are present for a number (n) of future symbols. In single-carrier systems intersymbol interference is dealt with through the use of equalizers. Conventional equalizers require n squared processing (order of magnitude). It is generally desirable to have the fourier transform be of size 'n.'

The achievable throughput may not be realized using single-carrier systems. The average group delay, which is readily available in the receiver, is a measure of the separation between the transmitter and the receiver. The almost-rectangular shape of the OFDM spectrum allows for close packing of adjacent OFDM systems.

OFDM is generally implemented through the application of a Fourier transform to frames of modulated, coded data just before it would have been ready for upconversion in a conventional wireless communications system.

The presence of an inverse discrete Fourier transform (IDTF) suggests that if the data entering the IDTF is correlated in certain ways, the output of the IDTF could be a pulse of magnitude n, the size of the IDTF, or a sinusoid of peak 1. That has been demonstrated to increase the peak-to-average ratio of OFDM transmission. The peak- to-average power problem has been mitigated in various ways. For example, the use of highly linear amplifiers, although expensive, can be used.

Wi-LAN uses a patented signal whitening process; certain symbol patterns cause higher peak powers and, conversely, certain symbol patterns cause lower peak powers. A known symbol pattern is multiplied with the transmitted data to reduce the peak power and the same symbol pattern is multiplied to the received data to recover the original data stream. In this way, security is also added to the system. That process has been referred to as prewhitening.

Alternatively, the pseudorandom symbol pattern can be sent as a training symbol for channel estimation. The receiver correcting for channel effects also reverses the whitening process, thus eliminating the need for a priori knowledge of the symbol pattern.

The IEEE802.11a standard relies on retransmission with scrambling of the data to ensure that a data packet will eventually make it across even if the peak is clipped during the first transmission.

It is not typical to use uncoded OFDM. In fact, almost all current OFDM implementations use some form of forward error correction (FEC). Early use of FEC with OFDM became known as coded-OFDM or COFDM. It is, however, inappropriate to distinguish the standard through the use of prefix "C" since all OFDM in use today is coded.

Early OFDM systems featured a narrow frequency band as part of their design criterion. That was to ensure that all carriers suffered the same attenuation due to the channel and was over come by introducing a channel estimator in the receiver. That estimation could be achieved by using training OFDM frames with no data on them. In the case of these training OFDM frames, the signal Y(k) would be an estimate of the channel's frequency response. Data frames transmitted in the same packet of OFDM frames can then be divided by or multiplied by the reciprocal of this frequency response to remove the effects of the channel.

Often the estimate of the channel response contains nulls, resulting from multiple rays' destructively interfering with each other. Those nulls could be so deep that dividing by the channel response leads either to an increase in the noise level or to a division by zero. That is often circumvented through truncating the response to a certain value if it drops lower than a particular value. A more appropriate solution is either to use an FEC that allows erasures of suspect symbols like Reed Solomon coding or to feed the entire channel response into the FEC in the receiver and use it for some form of weighting.

Wi-LAN has developed a number of products based on its patented W-OFDM technology. These are the I.WiLL 300-24, providing 30 Mbits/second and operating in the 2.4-GHz band, and the POP 25 and CPE 25, operating in the 2.5-GHz MMDS/MCS bands. The current Wi-LAN system uses a direct sequence header to achieve initial synchronization using 256 subcarriers. Six training symbols are used to estimate the channel, thereby aiding in the error correction. If a subcarrier is predicted to be poor, the data on the subcarrier can be assumed to be in error-known as an erasure-and conserves the error-correcting capability of the code. The W-OFDM system uses (216, 200, 8) Reed Solomon (RS) encoding. There are 8 bits per symbol, and a RS code word consists of 216 bytes formed from 200 input bytes. Hence, the coding rate is 200/216. The modulation scheme is 16-ary quadrature amplitude modulation. The W-OFDM channels are 6-MHz wide and can support a raw data rate up to 19 Mbits/s.

There is a guard band of 3.3 microseconds between OFDM symbols. Delay spreads of over 3.3 microseconds caused by multipath propagation will cause OFDM symbols to interfere with one another. But typical values of root mean square (rms) delay spreads generally do not exceed 3 microseconds. Though statistically unlikely, longer delay spreads-due to reflections from faraway mountains, for instance-can be dealt with using traditional cellular systems planning. Using sectored antennas will cut the amplitude of reflected signals and hence the rms delay spread.

Another feature of Wi-LAN's W-OFDM system is the use of signal whitening. With that technique, security can also be implemented into the system.

Wi-LAN had implemented simple antenna diversity in its prototype system in 1994 and removed it from later prototypes since it added unreasonable complexity at the time. Wi-LAN plans to launch systems utilizing multiple input multiple output antenna systems in 2003.

We have demonstrated that its implementation of W-OFDM can operate in a vehicle at speeds exceeding 70 mph. That opened the door for OFDM to become a standard for high-speed data cellular telephones.

Last month, Cisco Systems launched a family of products in the 5-GHz band and 2.5-GHz band that it described as utilizing its vector OFDM (VOFDM) coding scheme. VOFDM is OFDM with antenna manipulation whereby multiple OFDM signals are transmitted from different antennas in an antenna array and the different signals are separated by some time interval. This is the concept of spatiotemporal diversity.

VOFDM exploits multipath by introducing multiple antennas into a multipath environment, thereby increasing the S/N and increasing the data rate. Ironically, multipath is needed to increase the data rate. If multipath is not present, the complexity added by the introduction of the vector system is wasted.

Limited interoperability

The IEEE-802.11 standard describes the media-access control (MAC) and physical (PHY) layers for wireless LAN systems that support data rates of 1 or 2 Mbits/s. The PHY layer supports signaling via infrared or radiofrequency transmission in the 900-MHz or the 2.4-GHz ISM bands. In the RF bands, either direct-sequence or frequency-hopping spread-spectrum modulation is employed. Those few options have limited the interoperability of devices made by different vendors. Further, the delay in reaching a standard has meant that it is not representative of modern advances in spread-spectrum technologies. The IEEE-802.11 committee members recognized those issues and addressed them through a request for proposals, proceeding at a very fast pace from the request to accepting some proposals as draft standards.

The IEEE-802.11a standard is based on OFDM and the result of a joint submission by Lucent and Nippon Telephone and Telegraph (NTT). It is a high-speed standard (9 to 54 Mbits/s) to be implemented in the 5.x-GHz ISM frequency band.

The broadband radio access networks (Bran) group of the European Telecommunications Standards Institute (ETSI) has developed a standard, HiperLAN/2, which is very similar to the IEEE 802.11a. The basic difference between the two standards is that the IEEE standard's MAC layer is for ad-hoc networks, whereas HiperLAN/2's has the ability to provide some quality-of-service guarantees. There is currently an industry forum led by Microsoft, Intel and Compaq encouraging the harmonization of the two standards.

It has been speculated that OFDM will be the ideal technology for a fourth-generation cellular network. AT&T has entered into a contract with Nortel to develop a 4G standard based on its Angel product. The downlink in such a system would be OFDM and would be capable of transmitting data to the phones at a speed of 10 Mbits/s; the uplink back to the basestations would be a higher-speed time-division multiple-access link. Similar developments have taken place in Japan. It is believed that the mobile multimedia network, Magic, which is part of the NTT Docomo vision 2010, will provide 10 Mbits/s to the user through OFDM technology.

The IEEE 802.16 has invited interested parties to submit proposals for a high-speed fixed wireless-access standard to operate below 11 GHz. That submission includes possible applications in the 2.5-GHz multipoint, multichannel distribution systems band. As of the end of January, there are six proposals under discussion: Three are OFDM, two are single-carrier and one is a turbo-coding proposal. A PHY was not specified.

There are no assurances as to the outcome of future voting. However, if the number of applications is any indication, OFDM is the most likely physical-layer protocol.

ETSI's Bran group is also developing a fixed wireless-access standard below 11 GHz.

Deploying Angel

It is important to note that the fixed wireless-access system Angel, developed by AT&T and currently deployed in a number of markets in the United States, utilizes OFDM technology (www.att.com/technology/features/0005fixedwireless.html). The author is unaware of attempts to make the Angel technology an industry standard.

Meanwhile, Dedicated Short-Range Communications (DSRC) standards are being developed for intelligent transportation systems applications. These applications include distribution of real-time broadband traveler services.