Two-way multichannel, multipoint distribution service (MMDS) is a cost-effective, wireless alternative to digital subscriber lines (DSL) and cable modem service for delivering broadband wireless access (BWA) to the last mile.

In the United States, MMDS uses microwave frequencies in the 2.1 GHz to 2.7 GHz band. Compared to higher frequency wireless access methods, such as local multipoint distribution service (LMDS), MMDS systems are less affected by weather and can deliver reliable service up to a 25 mile radius under line-of-sight (LOS) conditions. Since two-way MMDS is often configured as a multipoint-to-point network, the broadband capacity delivered to individual users depends on how the system is engineered for parameters such as network concentration, and percent users of users on-line.

Today's deployed MMDS equipment represents a snapshot of the technology available several years ago. To leverage the network, security, and operations support systems (OSSes) from the data over cable service interface specification (DOCSIS), a number of equipment providers have chosen to base their MMDS wireless systems on this cable industry standard. DOCSIS-based systems have evolved to include physical layer improvements such as multiple modulations, diversity, and improved equalization. Looking ahead, next generation BWA systems will incorporate elements from 3G mobile efforts, as well as technology unique to the fixed wireless application.

This article offers a survey of physical layer technologies for next generation broadband wireless access. The intent is not to go into great depth on any particular technology. Rather, the various technologies are compared and contrasted with respect to their utility, availability and cost.

The core service requirements of a BWA system are capacity, coverage and cost. These three interdependent elements are strongly aligned with the physical layer. System capacity is a function of the spectral efficiency of the modulation scheme and the frequency reuse factor. The conventional definition of spectral efficiency is simply the delivered bits per second per Hz (bps/Hz) of occupied bandwidth. Capacity can be improved by increasing the modulation order (such as from 16-QAM to 64-QAM) but this will typically reduce coverage and frequency reuse. Coverage, of course, can be expanded at the expense of capacity.

Finally, capacity and coverage can be simultaneously increased, but at a cost. The most successful BWA systems will be those that most effectively balance capacity, coverage and cost.

One of the assumptions within the MMDS industry is that the initial supercell systems will be followed by smaller cellular deployments using frequency reuse. As such, the majority of next generation wireless access systems are targeting minicells with cell radii of 3 miles to 7 miles and relatively low antenna heights of 50 feet to 150 feet at the base station and ˜15 feet at the customer premise equipment (CPE). At these antenna heights, support for non-LOS operation is essential, as many potential subscribers will not have an unobstructed visual path to the base antenna.

The non-LOS MMDS channel is typically modeled as Rayleigh fading, with RMS delay spread between 1 µsec and 5 µsec, and Doppler of 0.5 Hz to 2 Hz. This is a slowly varying channel with significant time dispersion.

There are a number of technologies under development for next generation broadband wireless access systems.

• Adaptive modulation and coding

  Adaptive modulation and coding (AMC) refers to the ability to vary the constellation order and code rate as environmental conditions permit. AMC is a standard element of most next generation mobile wireless specifications¹ that offer significant advantages for BWA, as well. In the limit, a system with AMC will vary the modulation and coding on a packet-by-packet basis. This will significantly increase the throughput available to a particular subscriber in a fading environment. In addition, it allows higher burst rates for those subscribers with an excess of fade margin. Essentially, it obviates the need to design a cell plan based on the subscriber with the weakest signal (figure 1 represents the advantages of adaptive modulation in a Rayleigh fading environment²).

  Overlaid on the conventional fixed constellation bit error rate (BER) versus the signal-to-noise ratio (SNR) curves are two curves associated with adaptive modulation. One is the delivered bits per symbol and the second is the associated BER — both as a function of SNR. The BER tracks below that of binary phase-shift keying (BPSK) and the bits per symbol curve rises steeply over the working SNR range of 15 dB to 25 dB. This is a clear depiction of the value of adaptive modulation — the spectral efficiency and BER can be simultaneously optimized as channel conditions permit.

  Obviously, AMC requires the transmitter to have knowledge of the channel characteristics for a particular subscriber. This is not a huge drawback because the requested combination of modulation and coding can be efficiently appended to packets sent from the receiver. AMC will improve the capacity of a given coverage area with no significant increase in cost. The relatively low Doppler (0.5 Hz to 2 Hz) of the MMDS fixed wireless channel reduces the implementation complexity of AMC, relative to mobile systems.

• Dynamic channel allocation

  Dynamic channel allocation (DCA) is a means of rapidly switching channels to avoid fades or other impairments. By efficiently switching to another channel when the current channel is severely impaired, DCA can improve overall availability.

  On the other hand, different channels could be configured for different constellation sizes to achieve a crude version of AMC. In a system with DCA, the subscriber could switch from one channel to another based on a predetermined metric such as received codeword error rate. The overall result is not as efficient as true AMC but it is a means of improving coverage and capacity. One drawback is that the partitioning of bandwidth required by DCA may not be practical in all applications. DCA would not add significant intrinsic cost but it is an incremental burden to system complexity and overhead.

• Automatic repeat requests

  Automatic repeat request (ARQ) protocols have been a part of data communications from the very earliest days of the industry. Strictly speaking, ARQ is normally not part of the physical layer, but it is implemented in direct response to PHY impairments. It is essentially a form of time diversity where the packet received in error is retransmitted at a later time. The randomly varying channel is assumed to change rapidly enough that the retransmittal is error-free.

  For data flows that are not real-time, ARQ can significantly improve error-free throughput with minimal complexity and very little added cost. The changes are in the protocol rathrer than requiring significant increases in computational complexity. To be sure, computational complexity is often preferred to protocol and overhead complexity from the perspective of those involved in product development.

• Forward error correction

  Advanced forward error correction (FEC) is critical to any BWA system. All currently deployed BWA systems use some form of FEC to reduce sensitivity to channel impairments and improve link margins. Recent advances in coding/decoding technology, such as turbo codes, promise a few more dB of gain from FEC in next generation systems. The cost of this advanced FEC is greater computational complexity and more latency. The cost impact of computational complexity for advanced FEC is lightened by the fact that FEC technology can completely exploit Moore's Law. The latency hit can be easily addressed by support for several coding schemes of varying degree of latency.

  Advanced FEC will improve BWA performance in an incremental fashion, but it will not dramatically improve capacity or coverage by itself. To be most effective, advanced FEC must be included in the implementation of adaptive modulation and coding (AMC). To reduce FEC overhead, the coding rate should be optimized for each user in a point-to-multipoint system.

• OFDM — Multicarrier modulation

  Orthogonal frequency division multiplexing is a transmission scheme that has been deployed on a wide scale in commercial wireless applications in recent years. OFDM technology is used today in the IEEE 802.11 wireless local area network (LAN) standard as well as the terrestrial digital video broadcast standard (DVB-T) in Europe³. The fundamental utility of OFDM is its inherent ability to compensate for frequency-selective fading caused by multipath propagation.

  The degree to which multipath is frequency-selective depends on the relative length of the multipath time delay spread compared to the symbol period. When the symbol period is shorter than the delay spread, the multipath will introduce intersymbol interference (ISI) which must be mitigated by an equalizer. OFDM lengthens the effective symbol length by dividing the signal bandwidth amongst multiple carriers and adding guard intervals to eliminate intersymbol interference.

  Although it is widely held that OFDM is the key technology for enabling transmission over non-LOS channels, this is not entirely accurate; non-LOS channels may also experience flat fades. Relative to LOS channels, non-LOS systems will also experience significantly increased mean path loss and shadowing. OFDM does not address flat fades or otherwise improve link margin in shadowed environments. As such, OFDM in and of itself is not sufficient to enable transmission over non-LOS channels.

  The cost of wireless OFDM technology can be divided into two categories: the cost of the OFDM modem and the cost of the associated RF components. A good baseline for the cost of a wireless OFDM modem for MMDS environments is the current cost of integrated DVB-T receivers. DVB-T is a good comparison because it operates over 6 MHz channels using 2,048 carriers with constellations up to 64-QAM. Today, these chips can be purchased in volume for less than $15. DVB-T demonstrates that OFDM receivers can be implemented in a cost effective manner.

  The cost of the associated RF components is another matter. A poorly designed OFDM system can burden the RF circuitry with considerable cost and complexity. Compensation for phase noise, frequency offsets and high peak-to-average ratios can make it difficult to achieve low-cost subscriber equipment. Properly designed systems, however, can cost-effectively overcome these problems and the full potential of OFDM for multipath mitigation can be realized.

• Spatial processing

  Spatial processing is a broad topic in radio communications that encompasses many technologies from the very old to those “freshly escaped from the lab.” The common element of all spatial processing approaches is the use of multiple antennas. For instance, sectorization is actually a mature form of spatial processing that uses multiple directional antennas to improve capacity and spectral efficiency.

Typically, however, the term spatial processing is used in the context of improving link quality across a flat fading channel. A slow, flat fade is very difficult to mitigate using any combination of modulation and coding. Various forms of spatial processing are the key tools for solving the flat fade problem.

Antenna diversity is a practical and elegant form of spatial processing that has its roots in the earliest days of radio communication. The basic concept is to exploit the random nature of the propagation channel by using several statistically uncorrelated signal paths. In a way, antenna diversity actually depends on the very same impairment that it is trying to overcome. That is, the problem with fading is its random and unpredictable nature, and it is this very same property that is exploited in antenna diversity. To achieve decorrelated paths, the antenna separation is of the order of 10 wavelengths (˜4 feet) at the base station and the order of 1 wavelength (˜5 inches) at the CPE. These distances are practical and even convenient amounts in the MMDS band.

Antenna or spatial diversity can take many forms, the simplest of which is called switched/selection receive diversity. In this approach, the signal quality of two or more antennas is used to determine which antenna output is physically connected to the receiver. The measurement of signal quality typically takes place in a post-detection mode with the switching done only when the present path is taking errors.

This is an inexpensive technique well suited to burst communication systems. Its practical advantage of this scheme is that multiple RF receive chains are not required. In fact, switched diversity with two antennas can be implemented at the CPE in the $10 to $20 range with link budget gains in the range of 5 dB to 10 dB over a single antenna. This increase in link budget will directly increase a cellular coverage area.

Higher performance receive diversity is possible when all signal paths are available simultaneously. That is, each antenna has an associated RF chain and the multiple demodulated signals are combined in an optimal fashion. Maximal ratio receiver combining (MRRC) is such an approach where the SNR is maximized via the weighted sum of all the received signals.

Typically, this approach can yield dual path diversity gain in the range of 10 dB to 15 dB depending on the degree of FEC. However, MRRC requires a full RF receive path and demodulator for each antenna, which substantially increases CPE cost. For a two-antenna diversity scheme the cost increase is of the order of $100 to $150 at the CPE. The higher cost of MRRC is not a burden at the base station receiver because the cost can be distributed across the entire subscriber base.

The unappealing economics of implementing MRRC diversity at the CPE have led to recent work on transmit diversity at the base station, where higher cost is more tolerable. Transmit diversity is an extremely active research area has already led to practical techniques that have been adopted by the various 3G mobile efforts.

The basic concept is to replicate the diversity gain achieved with multiple receivers through the use of multiple transmitters and some additional signal processing. Typically, identical information is transmitted from the multiple antennas, but the encoding for each signal path is different. The diversity gain is then achieved via conventional combination, estimation and detection techniques.

Remarkably, transmit diversity techniques⁴ can be implemented in a manner that requires no bandwidth expansion, achieving performance within 3 dB of MRRC diversity (see figure 2). This 3 dB penalty is due to the division of power between two transmit antennas to equal the power of a single transmit antenna. However, the diversity gain is still in the range of 7 dB to 12 dB depending on the amount of FEC. Transmit diversity is not a complete replacement for receive diversity since MRRC diversity schemes have some ability to suppress co-channel interference. Nevertheless, the economic advantage of transmit diversity is compelling.

Spatial multiplexing and space-division multiple access (SDMA) are two developing antenna technologies that promise to dramatically increase capacity and spectral efficiency⁵. Spatial multiplexing uses multiple antennas at the transmit and receive sites to transport multiplexed bit streams over the same frequency. Assuming adequate decorrelation of the wavefront at the CPE, it is possible to de-multiplex the spectrally overlapping bitstreams. This feat of signal processing magic is possible with spatially distinct receivers operating in highly scattering environments.

In indoor laboratory experiments, spatial multiplexing has produced remarkable spectral efficiencies of 20 bps/Hz to 40 bps/Hz. The suitability of spatial multiplexing to the outdoor environment is currently under investigation, but there are some issues that are already understood. For one, multiple antennas and RF receive chains at the CPE add cost. Assuming that the RF CPE cost is split approximately in half between transmit and receive, a three-antenna CPE corresponds to nearly a 100 percent premium over a single antenna CPE. In addition, there is a considerable increase in baseband complexity at the receiver, but this cost will diminish over time as semiconductor integration improves. Clearly, the overall premium may be acceptable to some subscribers due to the increase in available downlink capacity.

Even if the cost is acceptable with spatial multiplexing, there are other drawbacks related to the propagation environment. Spatial multiplexing requires a highly scattering environment to spatially decorrelate the signals. In practice, not all subscriber signals will be suitably scattered, or the degree of scattering will change over time in a significant fashion. This may create problems for maintaining committed information rates to subscribers.

Another issue is that, typically, there is much less scattering at the base station. Therefore, receive antennas would require wide spacing if they were employed. However, the even higher cost of adding additional transmit RF chains at the CPE may preclude uplink spatial multiplexing altogether. If data traffic demands become more symmetric over time, this may present a problem.

Spatial division multiple access uses adaptive beamforming technology to separate co-channel subscribers in both the uplink and downlink directions. SDMA has been tried with only marginal success in the high-speed mobile cellular environment, but it has offered some value in limited mobility applications. The problem is that the local scattering at the subscriber makes downlink signal separation difficult. The scattering that is the friend of spatial multiplexing is actually the enemy of SDMA. An advantage of SDMA is that it can be configured to have multiple antennas only at the base station, thus saving cost. However, this configuration requires that the base station have channel knowledge to work properly. For this reason, successful SDMA schemes are usually based on time-division-duplexing (TDD) where the channel knowledge is gained via the duality property of the propagation channel. Like spatial multiplexing, SDMA offers the promise of dramatic improvements in spectral efficiency.

ACM, ARQ and advanced FEC are relatively uncontroversial features that will almost certainly be a part of next generation wireless systems. Their intrinsic costs are low and any relative cost differential will diminish over time as semiconductor integration improves. They will become available in cost-effective systems over the next year.

Properly designed, OFDM systems can be made cost-effective while retaining their advantages in frequency-selective multipath environments. OFDM is an enabling and incremental technology, however, and it will not directly improve capacity and coverage.

The most significant advances in BWA coverage and capacity will come from the application of spatial processing. The key question is which spatial processing techniques offer the best value. The spatial processing techniques summarized above are not mutually exclusive. Theoretically diversity, spatial multiplexing and SDMA could all be combined in one implementation. In that way, all propagation scenarios would be covered with the maximum potential for improved capacity and coverage. This is somewhat unrealistic since there is a practical limit to the number of antennas a base station or CPE can physically support. The economics of such an approach are unfavorable as well.

Spatial multiplexing and SDMA promise great gains in capacity and coverage but as yet neither technology has been proven in wide deployment. And even if the technology is proven in the field, the performance enhancement must be commensurate with the additional cost. Spatial multiplexing and SDMA take a different approach to cost distribution. Spatial multiplexing as a supported feature requires multiple receive antennas at the CPE.

Nevertheless, a stripped-down, lower cost CPE might not support spatial multiplexing, but still could interoperate in the network. That is, a subscriber need only pay for spatial multiplexing on an incremental basis. SDMA places all the cost burden at the base station where it is then distributed over all subscribers. Both spatial multiplexing and SDMA depend on largely uncommercialized aspects of radio propagation. For these techniques, the proof and payoff is in the propagation.

Diversity, in contrast, has been commercialized with great success. MRRC diversity does add RF chain cost to every subscriber, but the new advances in transmit diversity promise to shift the cost from the CPE to the base station. The coverage improvement offered by diversity nearly assures that next generation wireless systems will employ it in some form. The cost advantage of transmit diversity makes it perhaps the highest value feature in next generation wireless access.

1. S. Nanda, K. Balachandran, and S. Kumar, “Adaptation Techniques in Wireless Packet Data Services,” IEEE Communications. Magazine, Jan. 2000, pp. 54-64.

2. C.H. Wong and L. Hanzo, “Upper Bound Performance of a Wide-Band Adaptive Modem,” IEEE Transaction Communications, March 2000, pp. 367-369.

3. R. Van Nee and R. Prasad, OFDM for Wireless Multimedia Communications (Artech House: 2000).

4. S. Alamouti, “A Simple Transmitter Diversity Technique for Wireless Communications,” IEEE JSAC special Issue on Signal Processing for Wireless Communications, Oct. 1998, pp. 1451-58.

5. K. Sheikh, et al., “Smart Antennas for Broadband Wireless Access Networks,” IEEE Communications Magazine, Nov. 1999, pp. 100-105.

1. Figure 1 represents the advantages of adaptive modulation in a Rayleigh fading environment.

2. Overlaid on the conventional fixed constellation BER versus SNR curves are two curves associated with adaptive modulation. One is the delivered bits per symbol and the second is the associated BER — both as a function of SNR. The BER tracks below that of BPSK and the bits per symbol curve rises steeply over the working SNR range of 15 dB to 25 dB. This is a clear depiction of the value of adaptive modulation — the spectral efficiency and BER can be simultaneously optimized as channel conditions permit.

Mike Rude, Ph.D. is currently a principal engineer in ADC Inc. (www.adc.com) Office of Technology where he has responsibility for broadband access technology. Rude holds a Ph.D. in Electrical Engineering from the University of Southern California. He can be contacted at: mike.rude@adc.com.