FORWARD LINK CODING AND MODULATION FOR CDMA2000 1XEV-DO (IS-856)

Transcription

1 FORWARD LINK CODING AND MODLATION FOR CDMA2000 XEV-DO (IS-856) Nagabhushana T. Sindhushayana, Peter J. Black 2 Qualcomm. Inc., 5775 Morehouse Drive, San Diego, CA 922, SA, nbhush@qualcomm.com 2 Qualcomm. Inc., 5775 Morehouse Drive, San Diego, CA 922, SA, pblack@qualcomm.com Abstract - The IS-856 system delivers high spectral efficiency on the forward link by using dynamic link adaptation, hybrid ARQ and multi-user diversity scheduling. The channel coding/modulation scheme used on the forward link is designed to support incremental redundancy for hybrid ARQ, while maximizing the coding gain over static channels. This paper presents the design of the channelcoding scheme used on the IS-856 forward link. It highlights a novel interleaver design that provides near-optimal code puncturing and modulation symbol mapping, in addition to randomizing burst errors on the channel. Keywords - cdma2000 xev-do, IS-856, punctured turbo codes, interleaver design, incremental redundancy, link adaptation I. INTRODCTION TIA/EIA standard IS-856 [2,3] defines the air interface of an enhanced version of CDMA2000 [], which is optimized for high rate packet data services, such as wireless Internet for static and mobile users. The IS-856 forward-link offers a peak data rate of 2.4 Mbps and an average throughput of.5 Mbps per sector, over a signal bandwidth of.23 MHz [5] (with dual antenna terminals, under IT Pedestrian-A model with slow fading). The application layer throughput is shown to be about 85% of the physical layer throughput [6], consistent with the higher layer protocol overhead. In this paper, we describe the channel coding/modulation scheme used on the IS-856 forward link, and discuss various design considerations underlying the proposed scheme. In particular, we propose a novel interleaver design, that not only randomizes bursty channel errors, but also provides near-optimal puncturing and symbol-mapping for a wide range of code rates and modulation schemes. We validate the design choices with simulation results, showing packet error rate and system throughput characteristics for different choices of coding/modulation parameters. In order to achieve high bandwidth efficiency in the presence of rapid, unpredictable channel variations (shadowing, fast fading) seen in a mobile wireless environment, IS-856 optimizes the utilization of radio resources, using a combination of dynamic link adaptation, hybrid ARQ and (multi-user) diversity scheduling. The coding and modulation scheme is therefore designed to support open loop rate control (link adaptation) and as well as closed-loop rate control (incremental redundancy), while providing significant coding/diversity gains on typical wireless channels. The coding/modulation scheme incorporated in IS-856 has formed the basis for other 3G wireless standards currently being developed, aimed at providing high-speed data access to mobile users. II. AN OVERVIEW OF IS-856 OATION The IS-856 system consists of a network of servers (base stations) that provide high-rate data access on a wireless CDMA channel, to a collection of static or mobile user terminals. During a IS-856 connection, each (user) terminal maintains a radio link with one or more servers, which constitute the active set of the user. If the active set of a terminal consists of more than one server, the terminal is said to be in soft handoff. The terminal receives data from a server in its active set on the forward link, and transmits (broadcasts) data to all the servers in its active set on the reverse link of the IS-856 channel. Link adaptation refers to the process of allocating/changing the transmission data rate (and possibly other resources, such as power, bandwidth, code channels) in an efficient manner, in response to channel variations. The IS-856 link adaptation consists of an open loop rate control procedure, which determines the data rate of a packet prior to its transmission [4]. In order to support link adaptation, IS-856 defines a set of physical layer packet types, for data transmission on the forward link. Each packet type offers a certain nominal data rate and nominal packet length (duration), and requires a certain minimum signal quality, for reliable reception by the user terminal. IS-856 uses incremental redundancy (hybrid ARQ), in order to refine the actual data rate of the packet, while it is being transmitted. If the user terminal succeeds in decoding a packet even before it has been transmitted in its entirety, it informs the server not to transmit the remainder of the packet. Early termination of a packet increases its effective data rate above the nominal data rate associated with the packet type. This constitutes a closed loop rate control procedure, which enables the terminal to compensate for the margin built into the open loop rate requests. This margin in open loop rate control is needed to account for unpredictable channel variations, during the transmission of a packet in the future. In addition to the link adaptation and incremental redundancy techniques that seek to optimize the throughput /02/$ IEEE PIMRC 2002

2 of a given terminal, the IS-8586 network employs multi-user diversity scheduling, which seeks to optimize the overall system throughput. In an IS-856 network, each server provides data access to a large number of terminals using a time-division-multiplex (TDM) scheme, in which the server transmits data to at most one user at any given time. The server uses an opportunistic scheduling algorithm [7], such as the proportional-fair scheduler, to select the recipient of the next data packet, based on the packet type requested by the different terminals in the system. The scheduler takes advantage of uncorrelated time-variations of the channel seen by the different terminals in the system, and seeks to serve each user at a local peak of his channel fading process. In other words, the IS-856 forward link scheduler exploits multi-user diversity, providing enhanced system throughput (relative to a static TDM scheme), while maintaining fairness among users. III. IS-856 FORWARD PACKET TYPES In IS-856, system time is measured in frames, representing time intervals of length msec. Each frame is divided into 6 slots, spanning a duration of.66.. msec. A packet spans an integral number of slots, and may be earlyterminated, at the end of any slot. Table IS-856 Forward Link Complete Packet Types Pkt Type Nominal Data (kbps) Nominal Duration (slots) Pkt Size (bits) 38k k k k/2S k/4S k/S k/2S k M/S, M/2S, M, M 2, In order to support link adaptation, the IS-856 forward link defines multiple packet types for data transmission. Each packet type has several attributes, including its nominal data rate, nominal duration, packet size, nominal coding rate, modulation scheme and nominal repetition factor. IS-856 defines 2 different packet types, with data rates from 38.4 kbps to 2.4 Mbps. Table shows the nominal data rate, nominal packet duration and packet size (number of data bits) associated with each forward link packet type. The packet types in Table, used for open loop rate control, are referred to as complete packet types. The incremental redundancy (hybrid ARQ) scheme used in IS-856 allows physical layer packets to be terminated early, at the end of any slot. Early termination of a packet leads to a decrease in the packet duration and consequently an increase in its data rate (relative to their nominal values in Table ). In other words, incremental redundancy gives rise to a much larger set of truncated packet types, shown in Table 2. Note that many of the truncated packet types have an equivalent, complete packet type, of a higher data rate. Table 2 IS-856 Forward Link Truncated Packet Types Pkt Type(s) Early Termination Duration Data (kbps) 38k 9 n /n Equivalent Pkt Type 38k k 38k, 76k 5 n /n 38k, 76k k 38k, 76k, 53k k, 76k, 53k 07.07k/2S 38k, 76k, 53k, 307k/2S k/S 307k/4S k/4S k/2S 307k/4S, M/S 64k/2S, M/S 92k, M.2M/2S 2, M As seen from Table, IS-856 defines multiple packet types at certain data rates (307.2 kbps, 64.4 kbps,.2288 Mbps). In each case, the longer packet type is twice as large (in payload size and transmit duration) than the shorter packet type. This is motivated by the fact that packet types with longer duration provide better time diversity (on a fading channel) than the shorter packet type with the same data rate. The introduction of longer packets provides throughput gains from a combination of (open loop) link adaptation,

3 and (closed loop) incremental redundancy. The link adaptation algorithm tends to request shorter packets on nearly static channels, but prefers longer packet types on more dynamic channels. Moreover, on very dynamic channels, many of the long packets are terminated early (by hybrid ARQ), thereby increasing the effective data rate provided by the longer packet types. We provide simulation results that demonstrate the throughput gains, due to these longer packet types. IV. CODING AND MODLATION PARAMETERS Data 5.2 kcps Turbo Encoder Pilot 53.6 kcps MAC Channel Interleaver Packet Preamble 38.4 kcps 92.6 kcps Time Division Multiplex QPSK, 8-PSK, 6-QAM Modulator Sequence Truncation or Repetition Hadamard Transform (Walsh Covering) subjected to PN spreading. The resulting spread-spectrum sequence is pulse-shaped at chip-rate, generating the baseband waveform as shown in Fig.. Out of the total chip rate of.2288 Mbps, a fraction 92.6 kcps is allocated to the data channel, while the remaining chip rate is distributed among the Pilot/MAC channels and packet preamble. The spectral efficiency of a packet type is determined by its data rate, and may be expressed either as bps per Hz, or as bits per chip. The two measures are related by the bandwidth of the pulse-shaping filter, normalized by the chip-rate. The spectral efficiency (bits per chip) of a packet type may also be expressed as the product of its modulation rate (code symbols per modulation symbol) and code rate (bits per code symbol), divided by the repetition factor. For each packet, we need to choose the code rate, modulation rate and repetition factor, so as to achieve the required spectral efficiency. Increasing the repetition factor of a packet type improves its processing gain (diversity and interference rejection), without providing any coding gain (Eb/No reduction on static channels). On the other hand, lowering code rate and modulation rate in a spread-spectrum system provides both processing gain and coding gain. The above argument suggests that we should decrease the coding and modulation rates to the maximum extent possible, consistent with the required spectral efficiency. However, decreasing the code rate (and modulation rate) beyond a certain point yields marginal improvement in coding gain, while significantly increasing the encoding/decoding complexity. In IS-856, we limit the lowest code rate to /5, in view of the following facts Mcps PN Spreading & Pulse Shaping Fwd Link Baseband Waveform Fig.. Model of the IS-856 Forward link waveform. As described in the previous section, each packet type uses a specific code rate and modulation scheme. The data payload is encoded with a turbo code, which generates a sequence of code symbols. The output of the turbo encoder is permuted by a channel interleaver, and applied to a QPSK/8-PSK/6-QAM modulator. Adjacent code symbols at the interleaver output are aggregated to form a complex modulation symbol. The sequence of modulation symbols is truncated or repeated, to the extent needed to fill the available transmit duration. The extent of truncation determines the actual code rate of the packet. The punctured/repeated sequence is subjected to Hadamard transformation (Walsh-covering), time-multiplexed with the Pilot/MAC channels on the forward link waveform, and Pkt Error Comparison of Low Turbo Codes Eb/No (db) Fig. 2. Comparison of low-rate turbo codes R/5 R/4 R/3 Canonical turbo code constructions provide unpunctured code rates of /3 and /5. The Shannon capacity of a rate- /5 code (with QPSK modulation) is about 0.4 db lower than that of a rate-/3 code, and about 0.5 db lower than the capacity of a rate-/4 code. As shown in Fig. 2, practical codes (such as the cdma2000 turbo codes with a packet length of 024 bits) at rate-/5 offer similar coding gains,

4 relative to rate-/3 and rate-/4 codes. Moreover, the rate- /5 turbo code is already used as the fundamental encoding structure in cdma2000 systems. As a result, we use rate-/5 codes for most packet types below 64.4 kbps. A rate-/5 code with QPSK modulation achieves a spectral efficiency of 2/5 = 0.4 bits per modulation symbol. Packet types that require lower spectral efficiency employ partial or complete sequence repetition to achieve the desired spectral efficiency. Packet types that require higher spectral efficiency use code rate higher than /5. The higher rate codes may be obtained by puncturing basic rate-/5 turbo code. In IS-856 we puncture the turbo code indirectly, by first interleaving the output of the rate-/5 code, and then truncating the interleaver output sequence. This facilitates the hybrid ARQ process, as early termination of a packet has the same effect as puncturing the underlying code. Moreover, the technique of indirect code puncturing through interleaving and sequence termination provides a streamlined procedure for constructing a wide range of code rates, including non-standard code rates like 6/49 /3. Table 3 Coding, Modulaton and Repetition parameters Pkt Type Bits / chip Modulation Code Repetition 38k /24 2 (QPSK) /5 48/5 = k /2 2 (QPSK) /5 24/5 = k /6 2 (QPSK) /5 2/5 = k/2S /3 2 (QPSK) /5 6/5 =.2 307k/4S 6/49 2 (QPSK) 6/ k/S 2/3 2 (QPSK) /3 64k/2S 32/49 2 (QPSK) 6/49 92k 48/49 3 (8-PSK) 6/49.2M/S 4/3 2 (QPSK) 2/3.2M/2S 64/49 4 (6-QAM) 6/49.8M (8-QAM) 2/3 2.4M 8/3 4 (6-QAM) 2/3 Based on this criterion, we distribute the spectral efficiency of each packet type over its code rate, modulation rate and repetition factor as shown in Table 3. Most of the IS-856 packet types are constructed with the lowest possible code rate and modulation rate, consistent with the spectral density requirement. This rule is violated in some cases, where certain packet types are realized as code-extensions or symbol-repetitions of certain other high-rate packet types. For example, the packet type 304k/4S is obtained by a symbol-repetition of the packet type 64k/2S, which in turn is obtained as a code-extension of.2m/s. Similarly, the packet type 92k is a code-extension of packet type.8m, while the packet type.2m/2s is a code-extension of the packet type 2.4M. This feature has two advantages:. Deriving certain packet types as a simple modification of other pre-defined packet types reduces the number of distinct encoder/decoder configurations, resulting in a streamlined design, with lower modem complexity. 2. It facilitates incremental redundancy (hybrid ARQ), since the early termination of these packet types result in other packet types, whose parameters have been optimized for good error performance. V. CONSTRCTION OF IS-856 PACKET TYPES In this section, we present the details of channel coding, interleaving and modulation, used to construct the various packet types. The IS-856 modem needs to support a large number of packet types, each with its own code rate, modulation type, and repetition factor. Some of the packet types use non-standard code rates like 6/49 /3. A naïve realization of these packet types would involve choosing a basic turbo encoder rate (/3 or /5), a suitable puncture pattern, and modulation symbol mapping. Each puncture pattern needs to be optimized with respect to the modulation-symbol mapping, and further, account for the possibility of early termination. In other words, the puncture pattern needs to ensure good error performance not only on the complete packet type, but also on its earlyterminated versions (i.e., the truncated packet types derived from the complete packet type). In addition, the proposed solution should minimize the implementation complexity of the modem. This motivates the need for a simple, elegant solution that meets all the above design goals. In this section, we present a novel technique, in which the whole family of IS-856 packet types may be realized from a single turbo code configuration, simply by truncating or repeating the output of a channel interleaver. The interleaver is carefully constructed, so as to ensure nearoptimal performance on all early terminated versions of any packet type. With this approach, we puncture the turbo code indirectly, by first interleaving the output of the rate-/5 code, and then truncating the interleaver output sequence. IS-856 channel coding is based on the basic rate-/5 turbo code, used in cdma2000 systems [], as depicted in Fig. 3. It consists of two identical 8-state, rate-/3 systematic, recursive convolutional codes, separated (at the input) by a code interleaver. As is customary, the systematic part of the second convolutional code is discarded, resulting in a rate- /5 code. Let the polynomial (D) denote the binary sequence at the input to the encoder, and let (D) denote the interleaved

5 input sequence. The parity outputs of the two constituent encoders are given by 3 [ V ( D) V ( D) ] ( D) 0 = 3 [ V ( D) V ( D) ] ( D) 0 = We ignore trellis termination issues (tails bits), as it is unrelated to the main theme of this paper. (D) Code Interleaver (D) + D 3 3 (D) V (D) (D) V (D) Fig. 3. Schematic Diagram of the Turbo Encoder The output of the encoder is permuted by the channel interleaver, which serves four important objectives:. To transform burst errors at the channel interleaver output to random errors at the encoder output, or equivalently, at the decoder input. Most decoders are better equipped to deal with random errors than burst errors. 2. To ensure that the code symbols that go into a given modulation symbol are well separated, at the encoder output. This transforms a modulation symbol errors into random error in the code symbol sequence. 3. To ensure that truncation of the interleaver output sequence leads to good puncture patterns at the encoder output. This improves the effectiveness of the incremental redundancy (hybrid ARQ) used as a part of dynamic link adaptation. 4. To ensure that consecutive (unpunctured) code symbols at the encoder output are associated with different bit-positions of the modulation symbol. This is significant for 8-PSK and 6-QAM modulation with gray code mapping, wherein code symbols contained in a modulation symbol enjoy different levels of protection from channel noise. The action of the channel interleaver is described schematically in Fig. 4. The output of the turbo encoder is written row-wise, into three rectangular buffers, labeled, ( / ) and (V /V ). The buffer is filled with the (D) sequence from the turbo encoder. The ( / ) buffer is filled with the (D) followed by the (D) sequence. Similarly, The (V /V ) buffer is filled with the V (D) followed by the V (D) sequence. The number of rows in each buffer is determined by the modulation scheme used for the given packet type. Specifically, the number of rows is equal to the number of code symbol per modulation symbol (2 for QPSK, 3 for 8-PSK, 4 for 6-QAM). The number of columns of each buffer is determined by the length of the sequences (D), (D) etc. X /5 Turbo Encoder V V ( Row-wise Write into Rectangular Buffers ) / V / V Column End-Around Shift / V / V Bit Reversal / V /V / Interleave Symbol Modulation / V / V R = 2/3 R = /2 R = /3 R = /4 R = /5 ( First, then ) ( First V, then V ) V / V Fig. 4. Forward Link Channel Interleaver Within each buffer, the symbols in the j th column are cyclically shifted by an amount floor[j/4]. This is referred to as the end-around-shift operation. (The cyclic shift parameter floor[j/4] is chosen to account for rate-2/3 codes, where 3 out of 4 parity symbols are punctured.) The symbols in a given column are eventually aggregated to form a modulation symbol. The column-wise end-around shift operation is intended to randomize the unequal error protection offered to different bit-positions, by the 8-PSK and 6-QAM constellations with gray code mapping. The symbols in each row are permuted using a bit reverse interleaver. As a result, code symbols that are adjacent to

6 one another at the encoder output are moved to columns that are as far away as possible. QPSK Constellation (s Q s 0 ) = 2 = 2 I The code symbols in each column are aggregated to form modulation symbols, which are read out sequentially, first from the buffer, then from the ( / ) buffer, and finally from the (V /V ) buffer. The figure also shows the extent of sequence truncation needed to generate code rates of /4, /3, /2, 2/3 etc. Bit reversal interleaving ensures that sequence truncation after interleaving results in a regular puncture pattern at the encoder output, which provides near-optimal performance for all packet types defined in IS-856. Thus, bit reversal interleaving facilitates objectives () and (3) of the interleaver design. Row-column interleaving facilitates objective (2) of the interleaver design M 2.4M PSK Constellation 0 Q 0 0 (s 2 s s 0 ) I QAM Constellation Q (s 3 s 2 s s 0 ) Fig. 5. Signal Constellations and Labeling for QPSK, 8- PSK and 6-QAM modulation I 0.0.8M/NoCyc Shift 2.4M/NoCyc Shift SNR (db) Fig. 6. performance with and without cyclic (endaround) shift in the channel interleaver Suppose the columns of the channel interleaver buffers are labeled s 0, s, s 2, s 3. Fig. 5 shows how the groups of 2,3 or 4 code symbols in each column of the interleaver buffer are used to form QPSK/8-PSK/6-QAM modulation symbols. From the 8-PSK signal constellation, it is apparent that the code symbols in the s 0 bit-position are not as well-protected from channel errors, as the s and s 2 bit positions. Similarly, in the 6-QAM constellation, the bit positions s 0 and s 2 are not as well-protected as s and s 3. The end around shift in the channel interleaver prevents the clustering of less protected (reliable) bit positions in any contiguous sub-block of the encoder output. As shown in Fig. 6, such a clustering would be detrimental to the error performance of the turbo code. To summarize, the end around shift operation facilitates objective (4) of the channel interleaver design. VI. ERROR FORMANCE AND THROGHPT RESLTS Fig. 7 shows the simulated error performance of several primary packet types, on the AWGN channel. The packet error rate is plotted against the SNR at the antenna, which is generally higher than the base-band SINR, estimated by the modem. The base-band SINR includes the impact of linear and non-linear distortion of the signal at the transmitter and receiver, as well as the receiver noise floor. On a typical IS- 856 forward link receiver, the antenna SNR of 9.6 db may result in a base-band SINR of 8 db.

7 Recall that one of the objectives of coding and modulation is to provide near optimal performance for early-terminated packets. At nominal data rates greater than or equal to 64.4 kbps, every truncated packet type is equivalent to a complete packet type, in all respects including the preamble length, including performance. At data rates below 64.4 kbps, the truncated packet types have a longer preamble than their equivalent, complete packet types, and hence a somewhat worse performance. For instance, the performance of packet type 64k/2S terminated after one slot is identical to the performance of the complete packet type.2m/s. On the other hand, the performance of the packet type 307k/2S, terminated after slot, is slightly worse than the performance of the complete packet type 64k/S. This is because the packet type 307k/2S uses a longer packet preamble (28 chips) than the packet type 64k/S (64 chips). Table 4 shows the required (antenna) SNR for % packet error rate, for several truncated packet types. Packet Error.E+00.E-0.E-02.E-03.E-04 38k 76k 53k 307k/2S 307k/4S 64k/S 64k/2S Antenna SNR (db) terminal is assumed to have a dual antenna receiver. It may be seen that the error performance of the complete and truncated packet types is comparable to that of the best turbo codes, at the given spectral efficiency. The packet error performance on fading channels shown in Fig. 8 reflects upon the system performance only in the absence of link adaptation and incremental redundancy. The effective system performance of the IS-856 forward link is considerably enhanced by link adaptation on slow fading channels, and by incremental redundancy on fast fading channels. Nevertheless, the fixed rate results on fading channels provide some useful insight on the design of link adaptation and incremental redundancy. Let us refer to packet types 307k/4S, 64k/2S and.2mk/2s as long packets. Let the corresponding packet types 307k/2S, 64k/S and.2mk/s be referred to as short packets. It is clear from Fig. 8 that the long packets significantly outperform the short packets (of the same data rate), on fast fading channels (20 km/h). The long packets outperform the short packets even at 3 km/h, although the margin is somewhat smaller. Furthermore, on fast fading channels, the long packets are early terminated after the first slot, leading to a doubling of data throughput. Hence, the link adaptation algorithm may be designed to choose shorter packets on slow-fading channels (thereby reducing the prediction margin and improving packing efficiency for short payloads), while choosing longer packets on fast-fading channels (thereby enhancing time diversity and hybrid ARQ gain). The throughput gains resulting from the use of long packets may also be seen from Fig. 9, which shows the system throughput (per sector) with and without the use of long packets. Table 4 Required SNR for %, for some Truncated Packet Types Packet Error.E+00.E-0.E-02.E-03 64k/S 64k/2S 920k.2M/S.2M/2S.8M 2.4M Pkt Type Num Slots % SNR (db) Pkt Type Num Slots % SNR (db) 38k k.8 38k k k k E Antenna SNR (db) Fig. 7. AWGN Performance of IS-856 Packet Types Fig. 8 shows the performance of some of the packet types on a -path Rayleigh fading channel, at 3 km/h and 20 km/h, referred to the S PCS band (f c =.8 GHz). The user 38k k k k k/2S k k/4S k k/4S k k/4S 3-2.3

8 0. 76k, 20km/h 76k, 3km/h 307k/2S, 20km/h 307k/4S, 20 kmh 307k/2S, 3km/h 307k/4S, 3kmh 64k/S, 20km/h 64k/S, 3km/h 64k/2S, 20kmh 64k/2S, 3km/h exploit the coding and diversity gains offered the packet types and their truncated versions. 400 Ped.A, with long packets Ped.A, w/o long packets Veh.A, with long packets Veh.A, w/o long packets SNR (db) 64k/S, 20km/h 64k/S, 3km/h 64k/2S, 20kmh 64k/2S, 3km/h 92k, 20kmh 92k, 3km/h 0..2M/S, 20 km/h.2m/s, 3 km/h.2m/2s, 20 km/h.2m/2s, 3 km/h Sector Throughput (kbps) Number of sers Fig. 9. System throughput with and without long packets REFERENCES SNR (db) Fig. 8. Fading Performance of IS-856 Packet Types VII. CONCLSIONS In this paper, we have provided an overview of the main techniques used to enhance the spectral efficiency of the IS- 856 forward link. We have also presented a detailed discussion of the coding and modulation design for the forward link physical packet types, and the rationale behind several design-choices. In particular, we have presented a novel technique of generating a large number of packet types in a unified framework, while ensuring near-optimal error performance for the original packet types, as well as their truncated versions. Finally, we have provided some insights on the design of link adaptation and ARQ techniques, which [] 3GPP2 C.S002-A, Physical Layer Standard for cdma2000 Spread Spectrum Systems [2] 3GPP2 C.0024 Version 2.0, cdma2000 High Packet Data Air Interface Specification [3] E. Esteves, High Data Evolution of cdma2000 cellular system, Multi-access, Mobility and Teletraffic in Wireless Communications: Volume 5, Ed. G. Stuber and B. Jabbari, Kluwer Academic Publishers. [4] E.Esteves, P.J.Black and M.I.Gurelli, Link adaptation techniques for high-speed packet data in third generation cellular systems, Proc. of European Wireless, Florence, February [5] Peter J. Black and Mehmet I. Gurelli, "Capacity Simulation of cdma2000 xev Wireless Internet Access System", The 3rd IEEE International Conference on Mobile and Wireless Communications Networks (MWCN 200), Recife, Brazil, August 200,. [6] B. Mohanty, R. Rezaiifar and R. Pankaj, "Application Layer Capacity of the cdma2000 xev Wireless Internet Access System", World Wireless Congress 2002, San Francisco. [7] X. Liu, E. K. P. Chong, N. B. Shroff, Opportunistic transmission scheduling with resource-sharing constraints in wireless networks, IEEE Journal on Selected Areas in Communications, vol. 9, Issue: 0, Oct 200.