Abstract

In this thesis we investigate the capacity of multi-user wireless communication channels. We first establish a duality relationship between the Gaussian broadcast and multipleaccess channels. This relationship allows us to characterize the capacity region of the Gaussian broadcast channel in terms of the dual Gaussian multiple-access channel, and vice versa. This relationship holds for constant, fading, and multiple antenna channels. We also extend this duality to a class of deterministic, discrete memoryless channels. We then consider multiple-antenna multi-user channels, for which the capacity region is known only for the multiple-access channel. A duality is established between the dirty paper coding achievable region of the broadcast channel, which is a lower bound to the capacity region, and the true capacity region of the multiple-access channel. We use duality to prove that dirty paper coding achieves the sum rate capacity of the multipleantenna broadcast channel. Next, we provide an efficient and provably convergent algorithm that computes the sum rate capacity. Finally, we analytically upper bound the advantage that dirty paper coding provides over the sub-optimal technique of time division. Next we consider power allocation for time-varying wireless channels. We propose the concept of minimum rate capacity, in which average rates are maximized subject to the constraint that a minimum rate must be maintained for each user in all fading states. We explicitly characterize the minimum rate capacity and the corresponding optimal power allocation policies for the fading broadcast channel. We then use duality to characterize the minimum rate capacity of the dual fading multiple-access channel. Finally, we find the optimal power allocation policies that maximize ergodic capacity of a fading broadcast channel with both independent messages and a common, or multicast, message. In the final chapter, we study the benefits of cooperative communication in an ad-hoc network. In such a network there are multiple transmitters and multiple receivers. We consider cooperation schemes within the transmitting and receiving clusters. Using our earlier work on multiple antenna multi-user channels, we compare achievable rates to those achievable without cooperation.

MULTIPLE-ACCESS TECHNIQUES

Cellular systems divide a geographic region into cells where a mobile unit in each cell communicates with a base station. The goal in the design of cellular systems is to be able to handle as many calls as possible (this is called capacity in cellular terminology) in a given bandwidth with some reliability. There are several different ways to allow access to the channel. These include the following.

frequency division multiple-access (FDMA) time division multiple-access (TDMA) time/frequency multiple-access random access code division multiple-access (CDMA) o frequency-hop CDMA o direct-sequence CDMA o multi-carrier CDMA (FH or DS)

As mentioned earlier, FDMA was the initial multiple-access technique for cellular systems. In this technique a user is assigned a pair of frequencies when placing or receiving a call. One frequency is used for downlink (base station to mobile) and one pair for uplink (mobile to base). This is called frequency division duplexing. That frequency pair is not used in the same cell or adjacent cells during the call. Even though the user may not be talking, the spectrum cannot be reassigned as long as a call is in place. Two second generation cellular systems (IS-54, GSM) use time/frequency multiple-access whereby the available spectrum is divided into frequency slots (e.g., 30 kHz bands) but then each frequency slot is divided into time slots. Each user is then given a pair of frequencies (uplink and downlink) and a time slot during a frame. Different users can use the same frequency in the same cell except that they must transmit at different times. This technique is also being used in third generation wireless systems (e.g., EDGE). Code division multiple-access techniques allow many users to simultaneously access a given frequency allocation. User separation at the receiver is possible because each user spreads the modulated waveform over a wide bandwidth using unique spreading codes. There are two basic types of CDMA. Direct-sequence CDMA (DS-CDMA) spreads the signal directly by multiplying the data waveform with a user-unique high bandwidth pseudo-noise binary sequence. The resulting signal is then mixed up to a carrier frequency and transmitted. The receiver mixes down to baseband and then re-multiplies with the binary { 1} pseudo-noise sequence. This effectively (assuming perfect synchronization) removes the pseudo-noise signal and what remains (of the desired signal) is just the transmitted data waveform. After removing the pseudo-noise signal, a filter with bandwidth proportional to the data rate is applied to the signal. Because other users do not use completely orthogonal spreading codes, there is residual multiple-access interference present at the filter output. This multiple-access interference can present a significant problem if the power level of the desired signal is significantly lower (due to distance) than the power level of the interfering user. This is called the near-far problem. Over the last 15 years there has been considerable theoretical research on solutions to the near-far problem beginning with the derivation of the optimal multiuser receiver and now with many companies (e.g., Fujitsu, NTT DoCoMo, NEC) building suboptimal reduced complexity multiuser receivers. The approach being considered by companies is either successive interference cancellation or parallel interference cancellation. One advantage of these techniques is that they generally do not require spreading codes with period equal to the bit duration. Another advantage is that they do not require significant complexity (compared to a minimum mean square error-MMSE-detector or a decorrelating detector). These interference cancellation detectors can also easily be improved by cascading several stages together.

As a typical example, Fujitsu has a multistage parallel interference canceler with full parallel structure that allows for short processing delay. Accurate channel estimation is possible using pilot and data symbols. Soft decision information is passed between stages, which improves the performance. Fujitsu's system uses 1-2 stages giving fairly low complexity. Fujitsu claims that the number of users per cell increases by about a factor of 2 (100%) compared to conventional receivers and 1.3 times if intercell interference is considered.

MULTIPLE ACCESS A limited amount of bandwidth is allocated for wireless services. A wireless system is required to accommodate as many users as possible by effectively sharing the limited bandwidth. Therefore, in the field of communications, the term multiple access could be defined as a means of allowing multiple users to simultaneously share the finite bandwidth with least possible degradation in the performance of the system. There are several techniques how multiple accessing can be achieved. The are four basic schemes [1], [5]: 1. Frequency Division Multiple Access (FDMA) 2. Time Division Multiple Access (TDMA) 3. Code Division Multiple Access (CDMA) 4. Space Division Multiple Access (SDMA) 2.1 Frequency Division Multiple Access (FDMA) FDMA is one of the earliest multiple-access techniques for cellular systems when continuous transmission is required for analog services. In this technique the bandwidth is divided into a number of channels and distributed among users with a finite portion of bandwidth for permanent use as illustrated in figure 2.1. The vertical axis that represents the code is shown here just to make a clear comparison with CDMA (discussed later in this chapter). The channels are assigned only when demanded by the users. Therefore when a channel is not in use it becomes a wasted resource. FDMA channels have narrow bandwidth (30Khz) and therefore they are usually implemented in narrowband systems. Since the user has his portion of the bandwidth all the time, FDMA does not require synchronization or timing control, which makes it algorithmically simple. Even though no two users use the same frequency band at the same time, guard bands are introduced between frequency bands to minimize adjacent channel interference. Guard bands are unused frequency slots that separate neighboring channels. This leads to a waste of bandwidth. When continuous transmission is not required, bandwidth goes wasted since it is not being utilized for a portion of the time. In wireless communications, FDMA achieves simultaneous transmission and reception by using Frequency division duplexing (FDD). In order for both the transmitter and the receiver to operate at the same time, FDD requires duplexers. The requirement of duplexers in the FDMA system makes it expensive. Figure 2.1 Channel usage by FDMA (courtesy: www.ant.unibremen.de/research/cdma/) 2.2 Time Division Multiple Access (TDMA) In digital systems, continuous transmission is not required because users do not use the allotted bandwidth all the time. In such systems, TDMA is a complimentary access technique to FDMA. Global Systems for Mobile

communications (GSM) uses the TDMA technique. In TDMA, the entire bandwidth is available to the user but only for a finite period of time. In most cases the available bandwidth is divided into fewer channels compared to FDMA and the users are allotted time slots during which they have the entire channel bandwidth at their disposal. This is illustrated in figure 2.2. TDMA requires careful time synchronization since users share the bandwidth in the frequency domain. Since the number of channels are less, inter channel interference is almost negligible, hence the guard time between the channels is considerably smaller. Guard time is a spacing in time between the TDMA bursts. In cellular communications, when a user moves from one cell to another there is a chance that user could experience a call loss if there are no free time slots available. TDMA uses different time slots for transmission and reception. This type of duplexing is referred to as Time division duplexing (TDD). TDD does not require duplexers. Figure 2.2 Channel usage by TDMA (courtesy: www.ant.unibremen.de/research/cdma/) 2.3 Code Division Multiple Access In CDMA, all the users occupy the same bandwidth, however they are all assigned separate codes, which differentiates them from each other as shown in figure 2.3a. Figure 2.3b shows typical Walsh codes used for this purpose. CDMA systems utilize a spread spectrum technique in which a spreading signal, which is uncorrelated to the signal and has a large bandwidth, is used to spread the narrow band message signal. Direct Sequence Spread Spectrum (DS-SS) is most commonly used for CDMA. In DS-SS, the message signal is multiplied by a Pseudo Random Noise Code (PN code), which has noise-like properties. Each user has his own codeword which is orthogonal to the codes of other users. In order to detect the user, the receiver is required to know the codeword used by the transmitter. Unlike TDMA, CDMA does not require time synchronization between the users. A CDMA system experiences a problem called self-jamming which arises when the spreading codes used for different users are not exactly orthogonal. While dispreading, this leads to a significant contribution from other users to the receiver decision statistic. If the power of the multiple users in a CDMA system is unequal, then the user with the strongest signal power will be demodulated at the receiver. The strength of the received signal raises the noise floor for the weaker signals at the demodulators. This reduces the probability that weaker signals will be received. This problem, known as the near-far problem can be taken care of by using power control. This ensures that all the signals within the coverage of the base station arrive with same power at the receiver. Figure 2.3a Channel usage by CDMA (courtesy: www.ant.unibremen.de/research/cdma/) Figure 2.3b Walsh Codes 2.4 Space Division Multiple Access (SDMA) SDMA utilizes the spatial separation of the users in order to optimize the use of the frequency spectrum. A primitive form of SDMA is when the same frequency is re-used in different cells in a cellular wireless network. However for limited cochannel interference it is required that the cells be sufficiently separated. This limits the number of cells a region can be divided into and hence limits the frequency re-use factor. A more advanced approach can further increase the capacity of the network. This technique would enable frequency re-use within the cell. It uses a Smart Antenna technique that employs antenna arrays backed by some intelligent signal processing to steer the antenna pattern in the direction of the desired user and places nulls in the direction of the interfering signals. Since these arrays can produce narrow spot beams, the frequency can be re-used within the cell as long as the spatial separation between the users is sufficient. Figure 2.4

shows three users served by SDMA using the same channel within the cell. In a practical cellular environment it is improbable to have just one transmitter fall within the receiver beam width. Therefore it becomes imperative to use other multiple access techniques in conjunction with SDMA. When different areas are covered by the antenna beam, frequency can be re-used, in which case TDMA or CDMA is employed, for different frequencies FDMA can be used. Figure 2.4 Intra-cell SDMA

What is the Difference Between GSM and CDMA?

The GSM Association is an international organization founded in 1987, dedicated to providing, developing, and overseeing the worldwide wireless standard of GSM. CDMA, a proprietary standard designed by Qualcomm in the United States, has been the dominant network standard for North America and parts of Asia. However, GSM networks continue to make inroads in the United States, as CDMA networks make progress in other parts of the world. There are camps on both sides that firmly believe either GSM or CDMA architecture is superior to the other. That said, to the non-invested consumer who simply wants bottom line information to make a choice, the following considerations may be helpful. Coverage: The most important factor is getting service in the areas you will be using your phone. Upon viewing competitors' coverage maps you may discover that only GSM or CDMA carriers offer cellular service in your area. If so, there is no decision to be made, but most people will find that they do have a choice. Data Transfer Speed: With the advent of cellular phones doing double and triple duty as streaming video devices, podcast receivers and email devices, speed is important to those who use the phone for more than making calls. CDMA has been traditionally faster than GSM, though both technologies continue to rapidly leapfrog along this path. Both boast "3G" standards, or 3rd generation technologies. EVDO, also known as CDMA2000, is CDMA's answer to the need for speed with a downstream rate of about 2 megabits per second, though some reports suggest real world speeds are closer to 300-700 kilobits per second (kbps). This is comparable to basic DSL. As of fall 2005, EVDO is in the process of being deployed. It is not available everywhere and requires a phone that is CDMA2000 ready. GSM's answer is EDGE (Enhanced Data Rates for GSM Evolution), which boasts data rates of up to 384 kbps with real world speeds reported closer to 70-140 kbps. With added technologies still in the works that include UMTS (Universal Mobile Telephone Standard) and HSDPA (High Speed Downlink Packet Access), speeds reportedly increase to about 275380 kbps. This technology is also known as W-CDMA, but is incompatible with CDMA networks. An EDGE-ready phone is required.

In the case of EVDO, theoretical high traffic can degrade speed and performance, while the EDGE network is more susceptible to interference. Both require being within close range of a cell to get the best speeds, while performance decreases with distance. Subscriber Identity Module (SIM) cards: In the United States only GSM phones use SIM cards. The removable SIM card allows phones to be instantly activated, interchanged, swapped out and upgraded, all without carrier intervention. The SIM itself is tied to the network, rather than the actual phone. Phones that are card-enabled can be used with any GSM carrier. The CDMA equivalent, a R-UIM card, is only available in parts of Asia but remains on the horizon for the U.S. market. CDMA carriers in the U.S. require proprietary handsets that are linked to one carrier only and are not card-enabled. To upgrade a CDMA phone, the carrier must deactivate the old phone then activate the new one. The old phone becomes useless. Roaming: For the most part, both networks have fairly concentrated coverage in major cities and along major highways. GSM carriers, however, have roaming contracts with other GSM carriers, allowing wider coverage of more rural areas, generally speaking, often without roaming charges to the customer. CDMA networks may not cover rural areas as well as GSM carriers, and though they may contract with GSM cells for roaming in more rural areas, the charge to the customer will generally be significantly higher. International Roaming: If you need to make calls to other countries, a GSM carrier can offer international roaming, as GSM networks dominate the world market. If you travel to other countries you can even use your GSM cell phone abroad, providing it is a quad-band phone (850/900/1800/1900 MHz). By purchasing a SIM card with minutes and a local number in the country you are visiting, you can make calls against the card to save yourself international roaming charges from your carrier back home. CDMA phones that are not card-enabled do not have this capability, however there are several countries that use CDMA networks. Check with your CDMA provider for your specific requirements.