Link Budget Limitations -- Because of the large path loss encountered in serving wide areas, the link budget is challenging for these data rates and the desired performance level. Modeling path loss accurately is a difficult task requiring different, often empirical, models for different environments. For the purposes of this article we consider a simple model for path loss,
where
is the local mean received signal power, Pt is the transmitted power, and d is the distance between the transmitter and receiver. The path loss exponent 
= 2 in free space, and is between 2 and 4 for typical environments. The average received signal-to-noise ratio (SNR) is
where No is the one-sided noise power spectral density, B is the signal bandwidth, and K is a constant independent of distance, power, and bandwidth. Given the performance requirement SNR >= SNRo, it is clear that the path loss imposes limits on the bit rate,
and the signal coverage,
For example, consider the target data rate of 1 Mbaud, which is about 30 times that of an AMPS cellular voice circuit. Since the SNR is inversely proportional to the baud, this corresponds to a 15 dB increase in the required transmitted power to achieve the same bit error performance and cover the same area as a typical cellular voice circuit. Clearly, the coverage and performance of such systems will be severely limited without the introduction of new techniques. This is especially true for the uplink, where a mobile terminal cannot overcome the link budget limitations and still maintain a reasonable level of complexity and power consumption.
Multipath Fading Limitations -- In addition to the link budget limitations, the bit rate is also limited by the multipath nature of the radio environment. Physically, the received signal can be considered as the summation of a large number of signals which have traveled different paths to the receiver and, therefore, have different time delays, amplitudes, and phases. Depending on the extent of the channel impulse response, called the delay spread, and the resulting intersymbol interference (ISI), the maximum data rate can be severely limited.
This delay spread in a macrocellular environment could be as large as 40 µs, limiting the data rate to about 5–10 kbaud if no measures are taken to counteract the resulting ISI. In addition, the channel is time-varying, with Doppler rates as high as 200 Hz if operating in the 2 GHz PCS band.
A Physical Layer Solution
Asymmetric Service -- Since portable terminals must be powered by batteries and their transmit power is limited to about 1 W, achieving transmission rates beyond 10–100 kb/s in large-cell environments is impractical. On the other hand, base stations are usually powered from commercial main power systems and can transmit with higher power; subsequently, bit rates of over 1 Mb/s may be possible. Therefore, we propose an asymmetric service: a high-speed downlink with about 1–2 Mb/s peak data rates and a lower-speed 50–100 kb/s uplink. This alleviates the severe power problem at the mobile terminal and should be suitable for the most attractive new applications which would be supported by ACIS. In particular, Web browsing and information access, which have caused the recent explosion in Internet usage, are highly asymmetrical in transmission requirements. Only the transmission path to the subscriber needs to be high-speed. Many other services provided over the Internet can be provided with low to moderate bit rates transmitted from subscribers. Only video telephony and large file transfers in the direction from a terminal toward the network may require high-speed transmission from terminals.
Multicarrier Modulation -- One possibility for overcoming the delay spread limitations on the downlink is to use a single-carrier system modulated at 1 Mbaud with equalization and coding (e.g., [6]). This equalizer could require 20–40 taps and must be updated at the highest Doppler rate. In addition, the extensive period required to train the equalizer could be a major source of inefficiency in a packet-based system.
An alternative approach, and the one taken here, is to use a multicarrier system. The basic concept is to divide the total bandwidth into many narrowband subchannels which are transmitted in parallel. The subchannels are chosen narrow enough so that the effects of multipath delay spread are minimized. The particular multicarrier technique used here is called orthogonal frequency-division multiplexing (OFDM) [7–8] and is the standard for digital audio broadcasting (DAB) in Europe [9] and for asymmetric digital subscriber line (ADSL) in the United States [10], and has been proposed for many other applications.
To ensure a flat frequency response and achieve the desired bit rate, 100–200 subchannels are required, each modulated at 5–10 kbaud. With 5–10 kbaud subchannels and guard periods of 20–40 µs, delay spreads as large as 40 µs can be accommodated with little or no intersymbol interference (ISI). Since no equalization is required, OFDM alleviates the need for a long training period.
Diversity and Coding -- To reduce the link budget shortfall in the downlink, techniques for reducing the required SNR must be incorporated. To eliminate some of the 15 dB link budget shortfall, multiple base station transmit antennas are used, with each antenna transmitting a subset of the total number of subchannels, and up to two-branch antenna diversity is used at the mobile. In the scheme proposed here, the cluster of subchannels on each antenna is composed of a subset of widely spaced subchannels covering the entire transmission bandwidth. Alternatively, each subset can include widely spaced clusters of subchannels (e.g., two or three adjacent subchannels). Spreading the subchannels over the transmit antennas randomizes the fading across the OFDM bandwidth. To realize the full potential of the diversity, this is combined with Reed-Solomon coding across subchannels, using a combination of erasure correction, based on signal strength, and error correction. As shown in the next subsection, with four transmit antennas at the base and two receive antennas at the mobile, the required SNR can be reduced by 10 dB.
In addition, since a wider bandwidth and multiple transmit antennas are being used, the total transmitted power can be increased to make up the rest of the link budget shortfall. By transmitting at a power level about equal to that of an ordinary analog channel on each of four antennas, cellular link budgets can be achieved. Switched-beam smart antennas could also be used; in particular, using electronically switched beams, with four beams per antenna, another 6 dB of gain could be obtained without increasing transmitter power. Which beam is used for any given packet will be part of the channel assignment scheme.
Simulation Model -- The transmitter and receiver architectures are shown in Fig. 2. In the results presented here, QPSK modulation is considered with differential detection on each branch of the mobile receiver. Differential detection avoids the additional training required to recover the carrier phase; however, the error performance is inferior to ideal coherent detection. The differentially detected signals are then added together to provide an approximation to maximal-ratio combining. Ideal coherent detection is also considered for comparison. However, coherent detection requires a reliable and accurate method of channel estimation. Finally, the performance of this technique in the presence of a single co-channel interferer is presented.
In the radio link simulations, Rayleigh fading is assumed with Doppler rates as high as 200 Hz, along with a two-ray, equal-amplitude, delay spread profile, with impulse separations as large as 40 µs. In the example here, the OFDM signal is composed of 120 subchannels. Thus, data modulates each tone with a 160 µs symbol period. An additional 40 µs guard interval is used to eliminate any ISI due to the channel delay spread. This results in subchannels which are spaced by 6.25 kHz, block rates of 5 kbaud, and a total data rate of 600 kbaud, or, equivalently, channel bit rates of 1.2 Mb/s for QPSK. Reed-Solomon coding with 6-bit code symbols, corresponding to grouping three QPSK symbols in time, is used for error correction. The Reed-Solomon codeword is then formed across subchannels. For QPSK, a (40, 20) code, which corrects 10 erasures, based on signal strength, and five randomly errored symbols, is used. With the reduction in the delivered bit rate due to the 1/2-rate code, the peak rate for the sample QPSK system is 600 kb/s.
Results -- The corrected word error rate (WER) versus the average SNR is shown in Figs. 3–5 for the sample system described above. Here, word refers to one Reed-Solomon codeword, or a block of 240 bits for QPSK.
In Fig. 3 the performance of QPSK modulation using four transmit antennas at the base and two receive antennas at the mobile is shown in a channel with a 10 Hz Doppler rate (i.e., very slowly varying, as might be expected for pedestrian traffic) and a 200 Hz Doppler rate (i.e., around 100 km/hr at 2 GHz) and varying amounts of delay spread, including flat fading. For a target WER of 1 percent, less than 8.5 dB SNR is required, rather than the 17–20 dB typically needed for cellular systems. This represents about a 10 dB reduction in the link budget shortfall discussed earlier. The additional 2 dB needed in a flat fading environment is due to the reduced diversity effect. The penalty due to the increased time variation for a 200 Hz Doppler is only about 1 dB. Results are also shown using ideal coherent detection, with 40 µs delay spread and 10 Hz Doppler. In this case, the required SNR is reduced by another 3 dB. This would, of course, require a very reliable and accurate method of channel estimation.
The use of four antennas at the base station and two at the mobile may be undesirable in some applications. In Fig. 4 the trade-offs in using fewer antennas at either the transmitter or receiver are evaluated. WER is plotted versus SNR for one particular case of delay spread (40 µs) and Doppler (10 Hz) and for QPSK with differential detection. At a 1 percent WER with no diversity, an SNR of about 18.5 dB is required compared to 8.5 dB for four transmit and two receive antennas. Clearly, diversity is needed at the mobile; even with four transmit antennas, an SNR of about 14.5 dB is still needed. However, the number of transmit antennas can be reduced with a much smaller resulting penalty. For example, for two transmit antennas less than a 2 dB penalty is encountered.
In Fig. 5 the performance of OFDM with diversity and coding is shown in the presence of a single co-channel interferer. WER is plotted versus the average signal-to-interference ratio (SIR) for 40 µs delay spread, 10 Hz Doppler, and QPSK with differential detection and 1/2-rate Reed-Solomon coding. In this example it is assumed that there is no noise. Results are shown when there is no diversity and for the case of four transmit antennas and two receive antennas. The performance is very similar to the noise-only situation; in particular, a 1 percent WER can be achieved with an average SIR of only 9 dB -- roughly a 10 dB improvement over the no diversity case. By optimally combining the signals at the two antennas to maximize the signal-to-(interference+noise) ratio (SINR), a significant improvement beyond these results should be possible. In the next section, the results in Fig. 5 are used to determine an access threshold for the proposed dynamic packet assignment algorithm.
Other Issues
Uplink Transmission -- The uplink, or terminal-to-base transmission direction, could also use OFDM by dividing the wideband channel into clusters of 10–20 tones for use by individual terminals to achieve bit rates of about 100 kb/s while requiring acceptable transmit power levels. An alternative to OFDM is the use of multiple narrow carriers, which must then be demodulated using a filter bank at the receiver.
Synchronization -- Any realistic burst or packet transmission system must achieve reliable frame synchronization. Several techniques have been proposed for OFDM systems [11]. In addition, OFDM is more sensitive to frequency offset and timing mismatch than single-carrier systems. One goal of our current research is a synchronization technique which can achieve fast frame synchronization while simultaneously providing reliable estimates of both the frequency offset and the multipath channel response.
Sensitivity to Transmitter Nonlinearities -- Any multicarrier signal, such as an OFDM signal, can have a large peak-to-mean envelope power ratio (PMEPR). For example, a baseband OFDM signal with N subchannels has a PMEPR = N. For N = 128, PMEPR is approximately 21 dB. When passed through a nonlinear device, such as a transmit power amplifier, the signal may suffer significant spectral spreading and in-band distortion. The conventional solutions to this problem are to use a linear amplifier or to back off the operating point of a nonlinear amplifier; both result in a significant power efficiency penalty. Two alternatives have been proposed to reduce the PMEPR of the signal input to the amplifier:
-
Nonlinear block coding [12], which avoids the transmission of large amplitude sequences
-
Deliberately clipping the OFDM signal before amplification
In [13] it is shown that digital clipping and filtering can reduce the PMEPR to 6 dB with only a slight impact on performance. This is comparable to single-carrier QPSK with raised-cosine pulse shaping.
Implementation -- If ACIS is to be a viable service, a low-cost, low-power implementation of these receivers must be developed. One important advantage of OFDM, as shown in Fig. 2, is that both the transmitter and receiver modulation can be implemented as fast Fourier transforms (FFTs), allowing for very efficient digital signal processing (DSP) implementations.
ACIS: Downlink Medium Access Control
Medium Access Control Challenges
ACIS is targeted for applications such as Web browsing with a peak downlink rate on the order of 1–2 Mb/s using the wide-area cellular infrastructure. In order to provide bandwidth on demand using scarce spectrum, the downlink medium access control (MAC) protocol must:
-
Handle dynamic and diverse traffic with high throughput
-
Efficiently reuse limited spectrum with high peak rates and good quality
It is clear that wireline-based MAC protocols cannot work well without addressing spectrum reuse issues, and circuit-based wireless access techniques are insufficient for handling highly dynamic and asymmetric packet traffic. Therefore, an efficient packet channel assignment algorithm is required and must be carefully designed.
Channel Assignment Methodology
As discussed, very high efficiency will be required for 1–2 Mb/s macrocellular Internet access. In [14] a comparison is presented of various techniques to improve spectrum efficiency. The results show that interference averaging techniques, such as provided by CDMA systems, can perform better than fixed channel assignment techniques, and that interference avoidance techniques can perform even better. In particular, it is shown that dynamic channel assignment (DCA) with power control can provide a capacity that is two to three times higher than synchronous spread-spectrum and achieve efficiencies (measured in bits per second per Hertz per sector) as high as 50 percent [14, Figs. 3 and 5]. DCA combined with circuit-based technology (which has generally been the approach to date) can provide some benefits, but it cannot provide the large capacity gains predicted in [14] because of the dynamics of interference in a mobile system. Channel reassignments must take place at high speed to avoid rapidly changing interference. With certain assumptions, measurement-based DCA schemes can provide very high performance, particularly if channel selection is based on a combination of both mobile and base station measurements.
Based on the arguments above, DCA, for which any carrier is allowed to be used in any time slot, is considered here. This method provides significantly higher spectrum efficiency than fixed assignment except when the traffic load is heavy everywhere. Even though this method imposes higher complexity, such as base station time synchronization and a rapid frequency switching requirement, it appears to be a good approach to efficiently reuse spectrum for ACIS without requiring higher total bandwidth. In the following, we refer to the DCA algorithm proposed here for packet access as dynamic packet assignment (DPA).
For channel selection in the DPA algorithm, we combine attributes of two existing methods:
-
Interference sensing, or measurement-based DCA, which avoids selecting channels already in use in the neighborhood of a given base station, thus adapting to the interference environment
-
Channel segregation (CS) [15], which provides an adaptive learning process to form good reuse patterns in the neighborhood of a given base station, thus making assignment more robust against possible measurement error in the rapidly changing environment
Clearly, CS and measurement-based DCA are complementary. To introduce memory for preferred channels, a priority value is associated with each available channel for all radios at all base stations. For example, consider a system having three downlink frequencies, three sectors per base station, two radios per sector, and nine time slots per radio. There are six channel lists per base station, each associated with a radio, in which each of the 27 (time slot, frequency) choices are assigned different priority values (from 1 to 27). Initially each radio is given a randomly shuffled channel priority order. An adaptive learning process is then used to continuously update the channel priority order following each channel assignment process. This is easily performed by swapping the priority positions of the first channel having acceptable quality with the first channel failing the quality test (which is the highest-ranking unused channel before updating priority ranking). Compared with the original CS algorithm, this method learns preferred channels while employing interference look-ahead, thus avoiding using high-ranking bad channels.
Downlink Dynamic Packet Allocation Algorithm
Basic Algorithm -- The DPA algorithm, shown in Fig. 6, gives good downlink performance with low implementation complexity. This method requires mobile stations (MSs) to perform interference measurements, based on the channel priority order, once informed of pending packets at the base station. The first unused channel with acceptable quality (here, this corresponds to an SIR greater than a threshold) is selected. To reduce measurement delay, narrowband pilot tones corresponding to the downlink traffic channels could be transmitted simultaneously by individual base stations. Mobiles need only scan the pilots, similar to spectrum analysis using techniques (such as FFT) which can be completed in a short time; this is convenient to implement when OFDM modulation is considered. Figure 7 shows an example of pilots sent by different base stations. These pilots can also be sent during a short time slot in each frame.
Clearly, monitoring pilots at mobiles yields a direct downlink interference measurement, but mobiles must inform base stations of acceptable channels without significant delay to make these channels useful. Narrowband feedback channels are sufficient to carry the necessary information without taking excessive uplink bandwidth; however, these feedback channels require higher frequency reuse to reduce feedback errors. To reduce the effect of setup latency due to the feedback process, a staggered frame assignment procedure is introduced in the next subsection.
Staggered Frame DPA -- Three key questions must be addressed:
-
How do adjacent base stations with concurrent traffic avoid selecting the same channel?
-
Is it possible to reduce channel quality degradation as time progresses when a selected channel is assigned to new mobiles by nearby base stations?
-
How do mobiles and base stations exchange information about channel selection and packet delivery?
We introduce a staggered frame DPA method to address these issues.
The frame structure, shown in Fig. 8, uses a staggered schedule for DPA among neighboring base stations. The purpose of a staggered frame is similar to the concept of token passing (but without specifically passing tokens since centralized coordination of base stations is not required; the only requirement is time synchronization among base stations) to allow only a single base station to perform the assignment procedure in a small geographical area at one time. Similar to the concept of frequency reuse, different base stations can perform assignments simultaneously as long as they are sufficiently separated in distance. This ensures that any base station performing interference sensing can detect all used channels in its neighborhood without any blind spots, thereby avoiding any co-channel interference (inter-base collision) once the assigned channel is used to deliver pending packets. Furthermore, since the assignment is performed once every
superframe,1 the degradation as time progresses is thus bounded if the superframe duration is reasonably short.
In Fig. 8 other information needed to be exchanged between base stations and mobiles is also specified:
-
Base stations must inform all mobiles which are running applications, such as Web browsing, of pending traffic.
-
Once paged, a mobile must perform a scanning procedure using an FFT to determine a list of acceptable channels and feed the information back to its base station. A list rather than only one channel is sent so that the base station can assign different channels to different mobiles in case multiple mobiles select the same channel.
-
Upon getting lists from all mobiles having pending packets, a base station must assign channels to deliver packets and inform the corresponding mobiles of its decision before the start of the next superframe.
Notice that this approach also gives the additional flexibility of assigning more than a single server for a given mobile according to its quality-of-service requirement. A designated control channel with a higher reuse factor is probably required to perform these information exchanges. It is important to note that, during the assignment frame, a base station must turn off the downlink pilots so that its mobiles can sense all interference from other base stations by monitoring the pilots.
Simulation Model -- To characterize the DPA performance, a system of 36 base stations arranged in a hexagonal pattern is assumed, each having three sectors using idealized antennas with 120° beamwidths and a 20 dB front-to-back ratio. The mobile antennas are assumed to be omnidirectional. In each sector, one radio provides nine 2 ms traffic slots, each of which can be used to deliver a downlink traffic packet using any of the two available carrier frequencies. The same channel can be used in different sectors of the same base station as long as the SIR measured at the DPA admission process exceeds 10 dB.
The co-channel-interference-limited case is considered; that is, noise is ignored in the simulation. In the propagation model, the average received power decreases with distance as d–4 and the large-scale shadow-fading distribution is log-normal with a standard deviation of 10 dB. Rayleigh fading is ignored here, which approximates the case where sufficient antenna diversity is employed. Automatic repeat request (ARQ) is employed, assuming perfect feedback, to request base stations for retransmissions. If a packet cannot be successfully delivered in 3 s, it is dropped from the queue.
Results -- Figure 9 and Fig. 10 show the probability of packet retransmission as a function of spectrum utilization, defined as the percentage of total available spectrum used in each sector. The results of Fig. 9 and Fig. 10 correspond to the cases where each mobile is assigned one or multiple channels to deliver packets, respectively, resulting in different trade-offs of quality versus delay. For utilization higher than 33.33 percent, the same spectrum is, on average, used more than once at the same base station. Retransmission is performed if a packet is received in error, which is simulated based on the WER curve obtained in Fig. 5. A 3–4 percent retransmission probability generally results in successful packet delivery after a few ARQ attempts without incurring excessive (greater than 3 s) packet delay.
Additional performance enhancement is achieved by employing downlink beamforming and power control. Both techniques are idealized for the results shown here. For beamforming, each 120° sector is simply divided into four 30° beams (with the same 20 dB front-to-back ratio and idealized antenna pattern), and a packet is delivered using the beam that covers the desired mobile station. For power control we assume that there is sufficient time for each mobile station, upon each new packet assignment and before packet delivery, to adjust five times with a 1 dB step size and 30 dB dynamic range, based on a 10 dB target SIR. The iterative power control process is performed independently by each mobile assuming perfect SIR measurements. This provides an upper bound on the performance of the algorithm.
With a 3–4 percent target retransmission probability, this DPA algorithm can achieve between 25 and 50 percent spectrum utilization with only two frequencies if beamforming is applied. If beamforming is not employed, a similar retransmission probability is achieved using three frequencies, at the cost of reducing spectrum efficiency. Integration of the power control algorithm with the DPA process to achieve the significant improvement shown here requires further study. The results shown here are consistent with the idealized performance predicted in [14], which is significantly superior to the efficiency provided by current cellular systems.
Summary -- In this section, an access protocol has been described, based on a dynamic packet assignment scheme, which allocates spectrum on demand with no collisions and low interference to provide high downlink throughput. To reduce the probability that adjacent bases assign the same channel at the same time, all radios have independent channel priority orders. However, simultaneous assignment in a local neighborhood results in blind spots. This latter problem is handled by a staggered frame assignment. In addition, the quality degrades as new channels are assigned at other bases. This is handled by reassigning channels every superframe. Finally, rapid measurements are made by using pilots and the existing FFT structure (from the OFDM receiver).
Additional Issues
In the second and third sections a high-bit-rate wide-area packet data service is described, and physical and MAC layer techniques have been proposed for providing this service. However, before such a service can become viable, several issues must be addressed, including:
-
Spectrum requirements
-
Economic viability
-
Compatibility with voice services
-
Synergy with second-generation systems
Spectrum Requirements
IS-136 TDMA provides a spectrum efficiency of about 4 percent measured in bits per second per Hertz per sector (3 x 8 kb/30 kHz x 1/21 reuse); GSM also provides a spectrum efficiency of about 4 percent (8 x 13 kb/200 kHz x 1/12 reuse); and IS-95 CDMA provides a spectrum efficiency of 4–7 percent (12 to 20 x 8 kb/1.25 MHz x 1 reuse x 1/2 voice activity). Service launch in as little as 1 MHz x 2 is desirable for a high-speed packet data system to enable refarming of spectrum that is currently in use for cellular and PCS access. Therefore, much higher spectrum efficiency will be required.
A typical Web object has a median size in the range of 1.5–3 kbytes with a mean size of 10–20 kbytes. Assuming a Web object of 3 kbytes, a transmission speed of 240 kb/s is required to achieve a transmission time of 100 ms. Suppose an Internet user is modeled as demanding an average of 20 kb/s, with a peak of 400 kb/s, on the downlink and perhaps an order of magnitude lower average rate on the uplink. This corresponds to about one Web object/s at the median size, but only about one Web object/10 s at the mean size. If service were launched with 1 MHz x 2 of spectrum, an efficiency of 4 percent per sector would support only 40 kb/s per sector or two active users. In order to support 10–20 active users per sector, a spectrum efficiency of 20–40 percent is required, which is possible using DPA, as indicated above.
One of the advantages of the approach described in this article is that service can be started with limited spectrum. An initial introduction could use one to three carriers with 1–3 MHz for the downlink. This could be accommodated in the cellular or PCS bands. Full deployment with a reasonable level of interference would require a few channels per base. For future growth a block of 10–30 MHz may be needed.
Economic Considerations
For fixed access, Internet users have shown a willingness to pay on the order of US$3/hr of service or 5 cents/min, although many Internet providers have switched to a flat rate service at US$15–20/month while providing 10–20 hr/month of usage, which is a rate of only US$1–2/hr or 1.7–3.5 cents/min. This again suggests that spectrum efficiency for high-speed packet service must be very high in order to provide low-cost service that will be attractive to end users.
Circuit Voice Access
The service described so far, with, as an example, 80-ms superframes, would be unsuitable for voice applications because of the delay associated with DPA. One solution to this problem is simply to use second-generation systems and their evolutionary systems to provide voice services. An alternative is to have OFDM as the fundamental physical layer and then adapt the network layer protocol for packet DCA or for circuit voice and data with low delay. The spectrum would be split with different bands for packet data and circuit voice and data. One proposal for using OFDM for voice traffic employs tone-hopping and interference averaging [5].
Synergy with Second-Generation Systems
A significant advantage of ACIS is its synergies with second-generation systems. For example, since it is designed to be used with the same coverage (i.e., link budget) as second-generation systems, the base station infrastructure can be shared. In addition, common RF circuits and frame and clock rates can be employed in the base stations, with terminals supporting multiple modes of operation.
Summary
In this article, we show that a wireless Internet service to support high-speed Web browsing that performs comparably to emerging fixed access network solutions is possible for wide-area cellular systems. We propose an asymmetric wireless packet data service for mobile users in macrocells, with peak downlink bit rates of 1–2 Mb/s, while requiring channelization of about 1 MHz. Thus, service can be deployed with a limited amount of spectrum, and hierarchical cell structures are feasible. Such a wireless service, optimized to meet the needs of a client-server model for information retrieval and Web browsing and combined with evolutionary enhancements in second-generation TDMA technologies, can provide an attractive option for third-generation systems. In addition, there is a possible synergy with second-generation systems, sharing base station and network facilities and circuits and frame clock rates in terminals.
The proposed scheme combines OFDM with transmit and receive antenna diversity and Reed-Solomon coding to overcome the link budget and dispersive fading limitations of the cellular mobile radio environment. In addition, a dynamic packet assignment process based on rapid interference measurements was proposed which can provide spectrum utilization efficiencies of two to four times those existing approaches and which would support initial deployment with as little as 1 MHz x 2 of spectrum.
References
[1] D. C. Cox, "Wireless Personal Communications: What Is It?" IEEE Pers. Commun., vol. 2, no. 2, Apr. 1995, pp. 20–35.
[2] ETSI TC-RES 06921, "High Performance Radio Local Area Network (HIPERLAN); Functional Specification," draft prETS 300 652, Sophia Antipolis, France, July 1995.
[3] D. Raychaudhuri, "Wireless ATM Networks: Architecture, System Design and Prototyping," IEEE Pers. Commun., vol. 3, no. 4, Aug. 1996, pp. 42–49.
[4] J. S. DaSilva, D. Ikonomou, and H. Erben, "European R&D Programs on Third-Generation Mobile Communication Systems," IEEE Pers. Commun., vol. 4, no. 1, Feb. 1997, pp. 46–52.
[5] ARIB, "Report on FPLMTS Radio Transmission Technology Special Group," v. E1.2, Jan. 1997.
[6] J. G. Proakis, "Adaptive Equalization for TDMA Digital Mobile Radio," IEEE Trans. Vehic. Tech., vol. 40, no. 2, May 1991, pp. 333–41.
[7] S. B. Weinstein and P. M. Ebert, "Data Transmission by Frequency-Division Multiplexing Using the Discrete Fourier Transform," IEEE Trans. Commun. Tech., vol. COM-19, no. 5, Oct. 1971, pp. 628–34.
[8] J. A. C. Bingham, "Multicarrier Modulation for Data Transmission: An Idea Whose Time Has Come," IEEE Commun. Mag., vol. 28, no. 5, May 1990, pp. 5–14.
[9] M. Alard and R. Lassalle, "Principles of Modulation and Coding for Digital Broadcasting for Mobile Receivers," EBU Tech. Rev., no. 224, Aug. 1987, pp. 168–90.
[10] P. S. Chow, J. C. Tu, and J. M. Cioffi, "A Multichannel Transceiver System for Asymmetric Digital Subscriber Line Service," Proc. GLOBECOM '91, pp. 1992–96.
[11] L. Hazy and M. El-Tanany, "Synchronization of OFDM Systems over Frequency Selective Fading Channels," Proc. VTC '97, pp. 2094–8.
[12] T. A. Wilkinson and A. E. Jones, "Minimisation of the Peak to Mean Envelope Power Ratio of Multicarrier Transmission Schemes by Block Coding," Proc. VTC '95, pp. 825–29.
[13] X. Li and L. J. Cimini, Jr., "Effects of Clipping and Filtering on the Performance of OFDM," IEEE Commun. Letts., vol. 2, no. 5, May 1998, pp. 131–33.
[14] G. J. Pottie, "System Design Issues in Personal Communications," IEEE Pers. Commun., vol. 2, no. 5, Oct. 1995, pp. 50–67.
[15] F. Furuya and Y. Akaiwa, "Channel Segregation, A Distributed Adaptive Channel Allocation Scheme for Mobile Communications Systems," Trans. IEICE, vol. E74, June 1991, pp. 1531–37.
Biographies
Leonard J. Cimini, Jr. [SM '89] received B.S.E. (summa cum laude), M.S.E.. and Ph.D. degrees in electrical engineering from the University of Pennsylvania in 1978, 1979, and 1982, respectively. During his graduate work he was supported by a National Science Foundation Fellowship. Since 1982, he has been employed at AT&T, where his current research interests are in wireless communications systems. He is a member of Tau Beta Pi and Eta Kappa Nu. He has been very active in the IEEE Communications Society and is currently serving as Area Editor for Wireless Communications for the IEEE Transactions on Communications and Editor-in-Chief of the IEEE JSAC Wireless Communications Series. He is also an adjunct professor at the University of Pennsylvania.
Justin C.-I. Chuang [F '97] received his B.S. degree (1977) from National Taiwan University, and his M.S. (1980) and Ph.D. (1983) degrees from Michigan State University, all in electrical engineering. From 1979 to 1982 he conducted thesis research on transient electromagnetics for radar target discrimination. From 1982 to 1984 he was with GE Corporate Research and Development, Schenectady, New York, where he studied personal and mobile communications. From 1984 to 1993 he was with Bellcore, Red Bank, New Jersey, as a member of the Radio Research Department. At Bellcore, his work on delay-spread effects, low-complexity modem design, and radio resource management led to key system parameter specifications in the Personal Access Communications System (PACS) which is now a personal communications system standard. From 1993 to 1996 he was with the Electrical and Electronic Engineering Department of the Hong Kong University of Science and Technology (HKUST), where he established the teaching and research program in wireless communications. In June 1996 he returned to the United States and joined AT&T Laboratories -- Research in New Jersey, where he is now a technology consultant in the Wireless Systems Research Department, involved in creating technologies to provide reliable services on wireless platforms. He continues to serve as an adjunct professor of HKUST. He has published widely on various aspects of wireless communications, including radio techniques, system architecture, resource management, and prototype implementation. He is a member of Phi Kappa Phi and editor of Wireless Access Techniques for IEEE JSAC's Wireless Communications Series. He was an editor for IEEE Transactions on Communications and a guest editor for two special JSAC issues on wireless personal communications. Currently, he is chair of the Technical Committee on Personal Communications of the IEEE Communications Society.
Nelson R. Sollenberger [F] heads the Wireless Systems Research Department at AT&T. His department performs research on next-generation wireless systems concepts and technologies, including high-speed transmission methods, smart antennas and adaptive signal processing, system architectures, and radio link techniques to support wireless multimedia and advanced voice services. He received his Bachelor's degree from Messiah College (1979) and his Master's degree from Cornell University (1981), both in electrical engineering. From 1979 through 1986 he was a member of the cellular radio development organization at Bell Laboratories, where he investigated spectrally efficient analog and digital technologies for second-generation cellular radio systems. In 1987 he joined the radio research department at Bellcore, and was head of that department from 1993 to 1995. At Bellcore he investigated concepts for PACS. In 1995 he joined AT&T.
