On the Reverse Link Capacity of cdma2000 High Rate Packet Data Systems

Transcription

1 On the Reverse Link Capacity of cda High Rate Packet Data Systes Eduardo Esteves QUALCOMM Incorporated 5775 Morehouse Drive San Diego, CA Abstract cda high rate packet data standard, also known as IS-856, is a next generation solution to high-speed wireless internet access. The IS-856 air interface incorporates several advanced techniques such as adaptive odulation and coding, turbo codes and increental redundancy, Hybrid-ARQ, fast channel feedback inforation and ultiuser diversity just to nae a few. In this paper, we present an overview of the reverse link characteristics and its perforance. In particular, we evaluate the capacity of the reverse link in both a single cell and ulti-cell, 3-sectored networks. Analytical expressions are derived to include the effect of practical antenna patterns on the reuse factor, f. Also, a novel approach to odel sector load is considered in order to account for the effects of high data rate transissions. Index ters Reverse link capacity, cda 1xEV, IS-856, Wireless internet access, high-data rate cellular systes, IS-95, HDR. I. INTRODUCTION Third generation cellular systes are designed to provide enhanced voice capacity and the support of high data rate packet data services. These data services are typically characterized by asyetric traffic requireents subjected to the adverse effects of the obile channel. Such conditions require the use of advanced techniques such as fast feedback channel inforation, adaptive odulation and coding, increental redundancy, ultiuser diversity, receive diversity, efficient handoff algoriths, adaptive data rate control, etc. Many of these features are present in the cda high rate packet data air syste (a.k.a IS-856 [1], which is a spectralefficient air interface that has been recently aproved as part of IMT-. A detailed overview of the IS-856 basic concepts can be found in []. In this paper, we focus on the perforance and capacity of the reverse link channel. One of the ain differences fro IS-95 reverse link is that IS-856 uses a pilot signal to coherently deodulate the reverse traffic channel. In addition, the reverse link of IS-856 introduces new low-rate subchannels that are used to support advanced ediu access control (MAC on both forward and reverse links. For low-rate voice systes, reverse link capacity analysis has been extensively analyzed in the literature, e.g.[3][4]. In this paper we extend such results to account for high data rate transissions as well as the effects of practical antenna gain patterns on sectorization gains. II. THE REVERSE LINK PHYSICAL LAYER The IS-856 reverse channel structure, as described in Figure 1, consists of the Access channel and the Reverse Traffic channel. The Access channel, which further consists of Pilot and Data channels, is used by the access terinal (AT when in the idle state to send signaling essages to the access network. In connected state, the AT transits on the Reverse Traffic channel, which contains a Pilot channel, a Reverse Rate Indicator (RRI channel, a Data Rate Control (DRC channel, an Acknowledgeent (ACK channel, and a Data channel. The RRI indicates whether or not the Data channel is being transitted on the Reverse Traffic channel and its associated data rate. Explicit rate inforation allows deodulation without coplex rate deterination algoriths. The DRC channel indicates to the access network the supportable data rate on the Forward Traffic channel and the best serving sector for the forward link. The DRC channel provides a priori channel state inforation used for forward link adaptation and ultiuser diversity scheduling. The ACK channel infors the access network whether a data packet transitted on the Forward Traffic channel has been received successfully. The ACK channel provides a echanis to perfor fine link adaptation based on a posteriori channel state feedback inforation []. Access Reverse Traffic Mediu Pilot Data Pilot Access Ack Control Reverse Rate Indicator Data Rate Control Figure 1 Reverse HDR Structure A. Reverse Link Wavefor Figures and 3 show the Reverse Traffic channel block diagra of the IS-856 standard. There are four orthogonal code-division ultiplexed channels. The Pilot/RRI channel is tie ultiplexed so that the RRI channel is transitted during 56 chips at the beginning of every slot (1.66 s. The 3-bit RRI sybol transitted every frae (16 slots, is encoded using a 7-bit siplex codeword [1]. Each codeword is repeated 37 ties over the duration of the frae, while the last 3 code sybols are not transitted. The DRC sybols (4 bits indicating the desired data rate are encoded using 16-ary biorthogonal code. Each code sybol is further spread by one of the 8-ary Walsh functions indicating the desired transitting sector on the forward link. The DRC essage is transitted Data

2 half-slot offset with respect to a slot boundary. The reason is to iniize prediction delay while providing enough tie for processing at the desired sector before transission on the forward link starts on the next slot. The ACK channel is BPSK odulated in the first half-slot (14 chips of an active slot. Transissions on the ACK channel only occur if the access terinal detects a data packet directed to it on the Forward Traffic channel. For a forward data packet transitted in slot n, a '' bit is transitted on the ACK channel in slot n+3 if a data packet has been successfully, otherwise a '1' bit is transitted. The 3 slots of delay allow the terinal to deodulate and decode the received packet before transitting on the ACK channel. Pilot (All 's RRI One 3-Bit Sybol per 16-Slot Physical DRC One 4-Bit Sybol DRCCover One 3-Bit Sybol ACK 1 Bit per Slot Data Physical Layer Packets Siplex Encoder Bi- Orthogonal Encoder Encoder 7Binary per Physical 8Binary Codeword (Factor = 37 Codeword (Factor = Bit (Factor = 18 Interleaver 59 Binary per Physical 16 Binary Puncture Last 3 Interleaved Packet Physical s Code Bits Rate (kbps Rate Rate (ksps Rate (ksps /4 1, /4, , /4 4, , /4 8, , / 8, A C B D TDM 7:1 56 Binary 18 Binary per Physical per Slot 18 Binary per Slot (Transitted in 1/ Slot 16 W = ( W 8 Walsh Cover 8 Wi, i =,..., 7 A 1.88 Mcps = ( W 4 B 1.88 Mcps = ( W = ( + + Figure Reverse Link Chanel Structure (1 of ACK Relative DRC Relative Data Relative P I I- Short U I I- User Long-Code I I Baseband Filter Q Quadrature Spreading (Coplex Multipl I=I PN I Q PN Q Q=I PN Q +Q PN I PN I PN Q Deciator by Factor of Q Baseband Filter Walsh Cover (+ P Q Q- Short U Q Q- User Long-Code cos(πf C t sin(πf C t Figure 3 Reverse Link Structure ( of C 1.88 Mcps D 1.88 Mcps The Data channel supports data rates fro 9.6 to kbps with 16-slot packets (6.66 s. The packet is encoded using either rate 1/ or rate 1/4 parallel turbo code as specified in IS The paraeters of the reverse link encoder for different s(t data rates are suarized in Figure. The code sybols are bit-reversal interleaved and block repeated achieving a fixed 37. ksps odulation sybol rate. The Pilot/RRI, DRC, ACK and Data channel odulation sybols are each spread by an appropriate orthogonal Walsh function as shown in Figure. Before quadrature spreading (see Figure 3, the Pilot/RRI and ACK channels are scaled and cobined to for the in-phase coponent. Siilarly, the Data and DRC channels are scaled and cobined to for the quadrature coponent of the baseband signal. Reverse link power control (both open and closed loops is applied to the Pilot/RRI channel only. The power allocated to the DRC, ACK and Data channels are adjusted by a fixed gain relative to the Pilot/RRI channel in order to guarantee the desired perforance of these channels. For exaple, the relative gain of the Data increases with the data rate so that the received Eb/Nt is adjusted to achieve the required packet error rate (PER. In this paper, we concentrate on the reverse Data channel perforance and sector capacity estiates. For reverse link capacity purposes, the DRC and ACK channels can be considered overhead channels. While the addition of Pilot and RRI channels iprove the deodualtor perforance at the base station, thus enhancing reverse link capacity, DRC and ACK channels ostly contribute to iprove forward link throughput. Moreover, for a terinal that has an active connection, Pilot/RRI, DRC and ACK channels are transitted on the reverse link even when the reverse Data channel is not being used. As a consequence, reverse link capacity decreases as the nuber of active terinals increase. III. DATA CHANNEL PERFORMANCE The perforance of the reverse Data channel was siulated under different channel scenarios and with power control. We consider independent and equal-strength paths, each being affected by Rayleigh fading channel with a classic Doppler spectru at vehicular speeds of, 3, 3 and 1 k/h (1.9GHz carrier frequenc. The k/h speed corresponds to a siple AWGN channel. The Data gains used in these siulations were the default values provided in [1] and reproduced in Table I. The required per path average Pilot Ec/Nt for 1% PER is shown in Figure 4 for all data rates as a function of the vehicular speed. The perforance is very siilar for all code rate ¼ packets, requiring a per path Pilot Ec/Nt between 4.8 and.8 db to achieve 1% PER. Also in this case, we can conclude that there is a sall difference (.db in perforance between stationary versus obile users, which is a desirable feature for adjusting power control setpoint as the obile environent changes [4]. For the code rate ½ packet (153.6kbps, the spread in required Pilot Ec/Nt is about 3.5dB. For kbps data rate, the default Data gain of 18.5dB was chosen in order to equalize the required Pilot Ec/Nt across the range of speeds. At low speeds the selected gain akes the occasional transitions to 153.6kbps to be conservative. On the other hand, such transitions will achieve a slightly higher PER at high

3 vehicular speeds. Of course for long periods of 153.6kbps transissions, the outer loop power control will eventually converge to the correct setpoint [4]. Table I Default Data s Data Rate ( r i Data (db (kbps Null - Pilot Ec/Nt per Path, db Pilot Ec/Nt, db (per antenna Required Pilot Ec/Nt per path for 1% PER - Single Cell, Equal-Strength Paths 9.6 kbps 19. kbps 38.4 kbps 76.8 kbps kbps Vehicular Speed, k/h Figure 4 Required Pilot Ec/Nt for 1% PER Required Pilot Ec/Nt per Antenna for 1% PER RX Antennas 4 RX Antennas Vehicular Speed, k/h Figure 5 The Effect of antenna Diversity on Required Pilot Ec/Nt Next, we consider the effect of diversity reception at the base station. Diversity reception is a relatively siple technique to iprove reverse link capacity and power consuption of obile terinals. In Figure 5, we present a coparison between the per antenna Pilot Ec/Nt required to achieve 1% PER. We consider 9.6kbps packets only with a 1 path per antenna Rayleigh channel odel. Except for very low speeds, where coherence loss due to weaker pilots is observed, the perforance is iproved by at least 3dB. At 1k/h the gain is in the order of 3.4dB because of the increased diversity provided by the 4-branch receiver. No attept to optiize the default Data channel gain or the channel estiator was perfored for the four-antenna case. IV. CAPACITY ANALYSIS Several analyses of capacity of CDMA cellular systes have appeared in the literature [3][4]. For the reverse link, capacity is derived by characterizing the statistical distribution of the total received power at the desired base station antennas. In this paper, we consider an idealized hexagonal placeent of base stations, each with 3 sectors as depicted in Figure 6 (Just one tier of interference cells is shown for siplicit. For each cell, the 3 ideal 1-degree boundaries are shown with the corresponding labels, β and γ. As in [4], we consider soft handoff between ultiple sectors since it provides a significant reduction on the interference levels seen at the desired sector. In our odel, soft handoff is allowed to occur only with the Nc nearest cells ( 3Nc sectors. In addition, we assue soft cobining of the received signal across the 3 sectors within a cell and that each sector is equipped with an N a antenna diversity receiver. It is assued that a single lognoral shadowing rando variable affects the link fro the terinal to a cell site and that the terinal s transit power is deterined by the cell with iniu haronic su of the propagation losses to all of its three sectors. In IS-856, the power controlling base station (BS attepts to control each terinal s transit power so that the pilot SINR received at each antenna of the BS is averages γ c,thetarget pilot SINR for a particular user. If data is available at the users terinal (AT for transission, the AT sends a packet at a given rate r i deterined by the reverse-link MAC algorith. The data packet transission causes the total transit power to be increased with respect to the pilot power by an aount proportional to the data rate. As a result, we can define the total received power fro the access terinal i (AT i ata given sector as T P Si = Si ( µ i + ηi (1 where S P i is the received Pilot power fro AT i and DataDB( r /1 η ( 1 i i = η r i = ( µ 1 1 DRCDB /1 i = + (3. where ηi and µ i are the relative gains of the Data channel and the cobined effect of Pilot and DRC channels, respectively. Note that, the Data channel gains in ( are given in Table I. Clearly, if a data packet is not being transitted, η i =. Moreover, in (3, we have neglected the effects of the ACK channel. Given that the ACK channel is transitted only

4 when a packet is received on the forward link and that transissions on the forward link are tie division ultiplexed, the effect of ACK channel on the total power received at a given sector is neglible. In this paper we will assue that the noinal values for the DRC channel gain and length are adjusted as a function of the handoff state of the access terinal. The values are shown in Table II and represent a good tradeoff between forward and reverse link capacities while providing a reliable channel state feedback channel when the terinal is in soft-handoff with ultiple base stations [7]. The probabilities shown represent the relative occurrence of softer (handoff between sectors of a given cell and soft handoff and will be used to characterize the average received interference level as a function of N s, the nuber of connected terinals per sector. Table II DRC Paraeters Handoff Scenario No- Handoff Probability DRC Length (slots DRC DB 1 psofter -1.5 µ softer = psoft µ i terinals power controlled by other cells to the average received power fro terinals in the desired sector. In [3], this is coputed for equaly loaded, oni cells for a variety of propagation loss exponents and lognoral shadowing standard deviations. In this paper, we extend the derivations in [3] to include the effect of sectorization with a coonly used 65- degree antenna pattern, G a (θ, as the one given in Figure 7. The analysis is described in Appendix A and the results are suarized in Table III. Antenna, db Softer Handoff Softhandoff p = 1/ µ softer = softer p = 1/ µ soft = 1. 5 soft Angle, degree Figure 7 Typical Antenna (Noralized for a Sectorized Cell γ β γ β Figure 6 Three Sector Hexagonal Cell Model A. Sector Load and Rise-Over-Theral Using (1, the total reveived power at the target sector can be odeled as N s P IoW = NoW + ( 1+ f Si ( µ i + ηi (4 i= 1 The other-cell interference factor, referred to as f, corresponds to the ratio of average interference generated by Table III Average Interference Factor f for 3-Sector Cells LogNoral N c = 4, = 4 Standard Deviation, σ The per user pilot power, S P i, is controlled by the target cell such that the pilot SINR per receiving antenna is given by γ c,i, which is typically assued to be a lognoral rando variable with ean c (db and standard deviation given by σ c (db. Moreover, due to the sector cobining assuption, the target SINR at the desired sector, say, is reduced by an antenna cobining gain factor G a. This can be approxiated by averaging (A.1 over the idealized hexagonal region defining the desired sector (see Figure 6. For the pattern showninfigure7, G a = Next we follow the approach in [], where the rise-over-theral, Z = I o / N o, is calculated based on (4 and the target pilot SINR. Then, it can be shown that 1 Z = (5 1 Y where the sector load Y is given by

5 N s Ga γ c, i ( µ i + ηi Y = (1 + f (6 i= 11+ Ga γ c, i ( µ i + ηi The significance of the denoinators in (6 is to convert each terinal s required E c / Nt toe c / Io. In [3][4] this effect is neglected since for voice signals the nuerator in (6 is typically uch less than 1. For high data rates, however, the denoinator in (6 ay take on values in the order of 1.5. In this paper, each ter under the suation in (6 is approxiated by the first two ters of its Taylor expansion about the expected value of the nuerator. That is define, X i = G a γ c, i ( µ i + ηi (7 ( βσc β c + X = Ε{ X i } = Ga e ( µ + ηi (8 i where µ = p soft µ soft + ( 1 psoft µ softer and β = ln(1 /1. Then, the sector load is approxiated by Ns X i Y (1 + f + (9. 1+ i= 1 (1 + X i It can be show that the average error in (9 is sall for reasonable values of c and σ c. B. Capacity of a Genie-Aided MAC Algorith The capacity analysis of the decentralized reverse link MAC algorith in IS-856 is beyond the scope of this paper. In [5], a detailed description of the algorith and network siulation results is presented. In this paper, we consider the liiting capacity scenario where a genie-aided MAC algorith is assued. This iplies that all terinals know what data rate allocation ( { r 1, r,..., r Ns } axiizes total throughput and satisfies an outage criterion. The outage criterion is specified in ters of the probability that the sector s rise over theral (ROT, or equivalently its load, exceeds a certain level. Under these assuptions capacity is achieved by axiizing N s r i such that i= 1 Prob( Y > Lax p out (1 We use a Gaussian approxiation to evaluate (1 so that only the first and second order oents of (9 are necessary 1. In this case, defining β + ( βσ ( βσ σ = G c c c a e e ( η + µη + µ i i (11 ( ηi + µηi + µ where µ = p soft µ soft + ( 1 psoft µ softer, it can be easily shown that, 1 The Gaussian approxiation is justified by assuing a sufficiently large nuber of terinals per sector and the fact that the genie-aided MAC algorith tends to distribute capacity as equally as possible aong all terinals. Y N s X = ( 1+ f i (11 i= N s σ X i σ Y = (1 + f (1 4 i= 1 (1 + X i Thus, we rewrite (1 as L Y ax Q pout σ (13 Y The genie-aided capacity can be evaluated using non-linear constrained optiization techniques. This is carried out for the perforance of a antenna per sector receiver as discussed in Section III. In this case we assue, =.75 db, σ c = 1.3 db, p soft = 1/ 3, L ax =. 8 (Z<7dB, p out = % and the three lognoral shadowing scenarios in Table III. The results are shown in Figure 8. The axiu capacity in the order of 4-8kbps is achieved when the nuber of terinals per sector is about 1. Beyond this point, the sector capacity and consequently the average user throughput decrease alost linearly with the nuber of terinals in the sector. Note the nuber of users refer to the ones that are actively oving data over the air as opposed to a uch larger nuber of users that can be dorant due to the burtiness of data traffic. Also, note the rugged appearance of the curves is due to the discrete set of available data rates. Sector Capacity (kbps Genie-aided MAC Capacity, % Outage at ROT>7dB Terinals per Sector c lognoral siga=6db lognoral siga=8db lognoral siga=1db Figure 8 Sector and User Throughput for Antennas per Sector C. Special Cases In this Section, we evaluate the ipact of certain scenarios on reverse link capacity. In particular, we evaluate the capacity of a single 3-sector cell since it would be of interest wherever there is a high data deand concentrated in a sall geographical area. This is obtained fro the previous equations if we set f = and p soft = (other paraeters reain unchanged. In addition, we consider two special ulti-cell cases with lognoral shadowing standard deviation

6 of 8 db: 1 A very slow obility case ( Rx antennas with c = 4. db, σ c = 1.1dB and µ softer = 1. 5, A fourantenna diversity receiver (obile with = 6 db, σ = 1. db. The results are shown in Figure 9. c Sector Capacity (kbps c Genie-aided MAC Capacity, % Outage at ROT>7dB low obility, antennas per sector obile, 4 antennas per sector single cell, antennas per sector Terinals per Sector Figure 9 Single Cell and User Throughput V. CONCLUSIONS In this paper we provided an overview of the IS-856 reverse link and its Data channel packet error perforance. In addition, we extended the typical CDMA capacity analysis to include the effect of practical antenna pattern on sectorization gains and high data rate transission on the sector load. A genie-aided MAC algorith was used to establish the liiting capacity that can be achieved. It is shown that the capacity decreases as the nuber of connected terinals increase due to the overhead of Pilot and DRC channels. However, a uch larger nuber of active users can be present due to the burstiness of data traffic. In addition, peak capacities in the order of 5 to 6 kbps can be achieved depending on the nuber of receiving antennas, interference environent and user obility. APPENDIX A The ratio of the received interference at the target sector generated by terinals power-controlled by other cells to the received signal at the desired sector (say is proportional to ζi /1 r i 1 Gi ζ r o /1 o 1 Ga with Ga ( θo, Gi =, Ga β γ (A.1 and Ga = Ε{ Go } assuing a unifor distribution of terinals in the sector of cell. r i is the distance between a terinal located at the (x, coordinates and cell i, is the path loss exponent, θi, is the incident angle fro theterinalat(x, tothe sector of cell i, and ζ i lognoral shadowing process affecting the link between the terinal and cell i. Following the notation in [4] but redefining M i = 1 log( ri 1 log( Ga ( θ i, β γ for i =,, Nc. we can write the average noralized interference factor as f = ( I S + I /( S N s Ga (A. where N s is the nuber of connected terinals per sector and N c 1 z M M bβσ / ( e j o I S = e R j Q z bβσ + + bσ S j π = 1 N c 1 M j M i Q z + bβσ + dz G j ρ dxdy bσ i= 1 i j and N z b c / ( βσ e IS = e R j S j π = 1 Nc M j Mi Q z bβσ + + dz G j ρ dxdy i 1 bσ = i j with 1 b being the correlation coefficient between any pair of lognoral rando variables, β = ln(1 / 1, S and S are the regions of points (x, for which the nearest Nc cells contain and not contain the center cell (cell, respectively. The noralized interference factor given in (A. has been nuerically evaluated and the results are suarized in Table III. ACKNOWLEDGEMENTS The author would like to thank Peter Black and Rajesh Pankaj for coents that helped iprove the quality of this paper. REFERENCES [1] 3 rd Generation Partership Project (3GPP cda High Rate Packet Data Air Interface Specification, Technical Report C.S4 v., Oct. [] E. Esteves, The High Data Rate Evolution of the cda Cellular Syste, in Multiaccess, Mobility and Teletraffic for Wireless Counications: Volue 5, pp 61-7, Kluwer Acadeic Publishers,. [3] R. Padovani, Reverse Link Perforance of IS-95 Based Cellular Systes, IEEE Personal Counications, 3 rd Quarter, 94. [4] A. J. Viterbi, CDMA Principles of Spread Spectru Counication, Addison-Wesley, Boston, MA, [5] R.A.Attar and E. Esteves, A Reverse Link Outer-Loop power Control Algorith for cda 1xEv Systes, subitted to ICC. [6] S.Chakravarty, R.Pankaj and E.Esteves, An Algorith for Reverse Traffic Rate Control for cda High Rate Packet Data Systes, to appear Globeco 1, Phoenix, AZ. [7] IS-856 Reverse Perforance, Qualco Technical Meo, 1.