CDMA2000 1 spl times EV-DO revision a a physical layer and MAC layer overview

EV-DO物理层和mac层介绍

EVOLVING TECHNOLOGIES FOR 3G CELLULAR WIRELESS COMMUNICATIONS SYSTEMS

CDMA2000 1xEV-DO Revision A: A Physical Layer and MAC Layer OverviewNaga Bhushan, Chris Lott, Peter Black, Rashid Attar, Yu-Cheun Jou, Mingxi Fan, Donna Ghosh, and Jean Au, QUALCOMM, Inc.

ABSTRACTThis article presents key enhancements to CDMA2000 1xEV-DO systems embodied in 1xEV-DO Revision A. These enhancements provide significant gains in spectral efficiency and substantial improvements in QoS support relative to 1xEV-DO Revision 0. In particular, 1xEV-DO Revision A approximately doubles the uplink spectral efficiency and doubles the number of terminals with delay-sensitive applications that can be simultaneously supported on the system. It provides substantial reduction in latencies (approximately 50 percent) during both connection setup and the connected state. It offers comprehensive network control over terminal and application performance to enable the desired trade-offs between capacity and latency/ fairness, thereby providing full QoS support and enhanced user experience. It also provides coverage improvement (approximately 1.5 dB) relative to 1xEV-DO Revision 0. This enables operators to offer services such as VoIP, video telephony, mobile network gaming, push-to-talk, Web browsing, file transfer, and video on demand to a larger number of simultaneous users. The 1xEV-DO Revision A network can provide downlink sector capacity of 1500 kb/s and uplink capacity of 500 kb/s (two-way receive diversity) or 1200 kb/s (four-way receive diversity) with 16 active users per sector, using just 1.25 MHz of the spectrum.

®

CDMA2000 is a registered trademark of the Third Generation Partnership Project 2 (3GPP2).1

A flow is a source with transmission requirements associated with an application such as video telephony, voice over IP (VoIP), gaming, Web browsing, and file transfer.2

INTRODUCTIONCDMA2000® 1xEV-DO Revision 0 (also referred to as DO Rev 0) was driven by the design vision of a“wide-area-mobile wireless Ethernet” as described in[1]. The result was a high-rate wireless packet data system with substantial improvement in downlink capacity[2] and coverage over traditional CDMA2000 systems such as IS-95 and IS-2000. In addition to high throughput, DO Rev 0 provides quality of service (QoS) support to enable operators to offer a variety of applications with different

throughput and latency requirements. These improvements were accomplished through the use of large packet sizes encoded with low-rate turbo codes, transmitted using adaptive modulation and coding, downlink physical layer hybrid automatic repeat request (H-ARQ), and downlink multi-user diversity, together with antenna diversity at the receiver. DO Rev 0 systems support per flow1 QoS on the downlink and per terminal QoS on the uplink. Increasing demand for high-speed wireless Internet access has resulted in rapid growth of the number of CDMA2000 1xEV-DO users worldwide. 2 Operators have observed a

strong demand for applications such as VoIP, video telephony, wireless gaming, and push-to-talk (PTT), along with demand for downlink-intensive applications such as Web browsing and file transfer. These applications demand a system that can support large numbers of simultaneous users while meeting their desired latency requirements. In order to meet this demand, 3GPP2 approved enhancements to CDMA2000 1xEVDO Revision 0 (TIA-856). The CDMA2000 1xEV-DO Revision A system (TIA-856-A) was therefore standardized in March 2004 by 3GPP2 and TIA.3 We provide a summary of 1xEV-DO Revision 0 and highlights of 1xEV-DO Revision A. A detailed discussion of the enhancements to connection setup is presented. We provide a detailed discussion of the enhancements to the uplink and downlink, respectively. We provide a performance summary followed by conclusions.

1XEV-DO SALIENT FEATURES1XEV-DO REVISION 0As in IS-95 and IS-2000 systems, the 1xEV-DO Revision 0 carriers are allocated 1.25 MHz bandwidth and use a direct sequence (DS) spread waveform at 1.2288 Mchips/s. The fundamental timing unit for downlink transmissions is a 1.666... ms slot that contains the pilot and medium access control (MAC) channels, and a data

There are over 16 million 1xEV-DO Revision 0 users as of September 2005.3

TIA is the Telecommunications Industry Association of North America.

IEEE Communications Magazine February 2006

0163-6804/06/$20.00© 2006 IEEE

75

EV-DO物理层和mac层介绍

1/2 slot 1024 chips Data 400 chips MAC 64 chips Pilot 96 chips MAC 64 chips Data 400 chips Active slot Data 400 chips MAC 64 chips

1/2 slot 1024 chips Pilot 96 chips MAC 64 chips Data 400 chips

n Figure 1. 1xEV-DO downlink slot structure.portion that may contain the traffic or control channel as shown in Fig. 1. Unlike IS-2000, where a frame is 20 ms, a frame in 1xEV-DO Revision 0 is 26.66... ms. The pilot channel is transmitted at full power for 96 chips every half-slot (each slot is composed of 2048 chips), providing not only a reference for coherent demodulation of traffic and MAC channels but also a 1200 Hz sampling of the channel state. These samples are used to estimate and predict the received signal-to-interference-and-noise ratio (SINR) at the access terminal (AT) in the near future, which aids the terminal in determining the maximum data rate that can be supported on its downlink. This provides the system with a mechanism for fast adaptation of modulation and coding schemes to different mobile channel environments. The MAC channel consists of a reverse activity (RA) channel and reverse power control (RPC) channel. The RA channel from a particular sector provides a 1-bit feedback to all terminals that can receive that sector’s forward link indicating whether or not its uplink load exceeds a threshold. The RPC channels from a particular sector carry a unique 1-bit closed loop power control command (update rate of 600 Hz) for each of the access terminals that include that particular sector in their active set.

The data rate control (DRC) lock channel is punctured into the RPC channel and is used to indicate the channel state from the access terminal to the access network. 1xEV-DO Revision 0 uses a time-division multiplexed (TDM) downlink (transmit to one user at a time). The traffic channel data rate used by the access network for transmission to an access terminal is determined by the DRC message previously sent by the access terminal on the uplink. The DRC indicates not only the data rate but also the modulation, code rate, preamble length, and maximum number of slots required to achieve the desired physical layer error rate. 1xEV-DO Revision 0 introduced physical layer H-ARQ on the downlink. The access network transmits packets to an access terminal over multiple slots staggered in time.4 At higher SINR, significant coding gains are achieved by incremental transmission of parity bits, and at lower SINR, powerful coding is achieved by simple repetition of a rate 1/5 turbo encoded packet. The 1xEV-DO Revision 0 downlink traffic channel is a shared medium that provides highpeak-rate transmissions to active access terminals. Addressing on the shared channel is achieved by a MAC index that is used to identify data transmissions from a sector to a particular access terminal. The packet preamble is covered with a bi-orthogonal sequence 5 determined by the MAC index assigned to the access terminal. In order to maximize performance in a variety of channel conditions, three basic mechanisms exist to control access to the downlink traffic channel. Open-loop rate control: A DRC message is sent by all ATs containing a requested data rate and a transmitting sector indication. The transmitting sector chosen by the AT is the one that provides the best downlink channel and can receive the a priori downlink channel state information (CSI) with acceptable reliability. Adaptive data scheduler: This takes into account fairness, queue sizes, and the most recent a priori downlink CSI provided by the DRCs. While the specification does not specify the details of the scheduler, some form of proportional-fairness scheduler[2] is typically used, so as to exploit multi-user diversity on the downlink. Closed-loop rate control: A fast feedback acknowledgment (ACK) channel allows the data rate of a packet to be effectively increased beyond the data rate corresponding to the requested DRC if the channel conditions experienced by the transmission to the AT improve relative to the channel estimate used to generate the DRC. A combination of H-ARQ and multi-user diversity improves performance in a variety of channel conditions; the former provides capacity gains in fast-fading channels, the latter in slowfading channels[3]. The uplink in 1xEV-DO Revision 0 is similar to that in IS-2000 with some key differences: Stochastic distributed rate control with direct measurement of sector loading via rise-over-thermal (RoT) measurement defined as Io/No, where Io is the total received po

wer and No is the thermal noise floor. This is enabled by the introduction of a silence interval when all terminals in the system do not transmit, enabling the access network to measure no. Explicit uplink rate indication using the reverse rate indication (RRI) channel. Closed-loop power control of the pilot channel, with gains for the data and overhead channels specified relative to the pilot. The traffic-to-pilot power ratios are selected such that the pilot SINR is not a function of the data rates and therefore does not require adjustment at rate transitions.

4

A three-slot separation between subpacket transmissions allows the AT time to demodulate and decode the packet, and indicate to the access network whether or not the packet was successfully decoded.5

Bi-orthogonal sequences are chosen for the preamble due to better distance properties than orthogonal codes. In addition, shorter Walsh sequences better preserve orthogonality in fading.

76

IEEE Communications Magazine February 2006

EV-DO物理层和mac层介绍

The uplink MAC channel protocol employs a distributed algorithm subject to feedback control and defines the rules used by each AT for data transmissions. The AT receives a reverse activity (RA) bit from each sector in its active set indicating whether or not the RoT exceeds a predetermined threshold. This information determines whether the ATs increase or decrease their data rates. The AT also receives an explicit rate limit message indicating the maximum data rate at which it may transmit. The access network provides each AT with two rate transition probability vectors the ATs use to increase or decrease data rate, respectively, depending on the effective RA bit (logical OR of RA bits from all sectors in the active set), available data, and power amplifier (PA) headroom. The rate transition vectors specify the probability that the AT will increase or decrease its uplink data rate based on the current data rate and effective sector loading.

1XEV-DO REVISION AThe salient enhancements offered by CDMA2000 1xEV-DO Revision A are: An uplink physical layer with H-ARQ, higher-order modulation (quadrature phase shift keying[QPSK] and 8-PSK), higher peak rate (1.8 Mb/s), and finer rate quantization. An uplink multiflow MAC with QoS support, comprehensive network control of spectral efficiency and latency trade-off for each application flow, and a more robust interference control mechanism that permits system operation at higher load. A downlink physical layer with higher peak rate (3.1 Mb/s), finer rate quantization, and short packets for transmit delay reduction and improved link utilization. A downlink MAC layer that permits the access network to serve multiple users with the same physical layer packet, improving not only transmission latency but also packing efficiency. The transmission latencies are further improved by seamless adaptive server selection and service to users reporting null-rate DRCs. Rapid connection setup fo

r applications that require“instant connect” via use of shorter interpacket intervals and a higherrate access channel.

to send an acknowledgment to the access terminal. On the other hand, using SC to send pages incurs large delays, because SC transmissions are rather infrequent. Moreover, SC and AC are transmitted at 38.4 kb/s (16-slot duration) or 76.8 kb/s (8-slot duration) using a 1024-bit payload to ensure high coverage for the control channel. However, this results in poor packing efficiency and inefficient utilization of downlink resources when transmitting a page 6 (typically 128 bits of data) to an AT using the SC. In order to allow the ATs to get quick access to the system while maintaining long terminal standby time and ensuring efficient use of downlink resources, the following enhancements were introduced for the control channel in 1xEV-DO Revision A: Subsynchronous control channel (SSC) transmitted synchronously with shorter intertransmit duration relative to the SC Use of short physical layer packets (128, 256, or 512 bits) for transmitting SSC and AC to ATs that do not have an active connection with the access network The minimum intertransmit interval for the SSC packets may be as low as four slots. A page arriving at the access network between two SC packet transmissions is not transmitted until the next SC in DO Rev 0 but is transmitted using the next SSC (64-slot SSC interval) in DO Rev A. In addition, the page transmission and access ACK take a maximum of just four slots each in DO Rev A, due to the use of smaller payload sizes (128, 256, or 512 bits). DO Rev A also allows grouping of users that listen for SSC at certain time instants, thereby further improving terminal standby time.

The control channel in 1xEV-DO Revision 0 is used for transmission of control and signaling information on the downlink. Two types of control channels, synchronous control (SC) and asynchronous control (AC), are supported.

ACCESS CHANNELKey access channel enhancements in DO Rev A are: Reduced access channel connection setup time–Higher-rate access channel (up to 38.4 kb/s)–Shorter access channel preamble (4 instead of 16 slots)–Reduced intrasequence interprobe latency Dynamic adjustment of access channel transmit power level Initial probe power adjusted based on forward link pilot strength to achieve desired probe success rate and minimize network interference Network control over access channel performance–Access probe success rate adaptation via network adaptive OpenLoopAdjust–Network control over the maximum access channel data rate and maximum payload size per terminal DO Rev A access terminals may transmit (access network controlled) on the access channel at up to 38.4 kb/s subject to data and power constraints, while in DO Rev 0 the access channel data rate is restricted to 9.6 kb/s. The amount of data that the access terminal is permitted to transmit on the access channel is also access networ

k controlled.

CONNECTION SETUPCONTROL CHANNELThe control channel in 1xEV-DO Revision 0 is used for transmission of control and signaling information on the downlink. Two types of control channels, synchronous control (SC) and asynchronous control (AC), are supported. The former is transmitted once every 256 slots (426.66... ms) and the latter whenever needed but not overlapping with the SC. The AC may be used to transmit delay-sensitive signaling information to ATs that either have an active connection with the access network already, or are trying to establish an active connection. For example, once the access network decodes an access probe from an AT, AC may be used

6

A Message sent by an access network to an AT instructing the AT to set up a traffic channel.

IEEE Communications Magazine February 2006

77

EV-DO物理层和mac层介绍

Since the open loop turnaround parameter, OpenLoopAdjust, is generated based on some assumptions of channel conditions and sector loading, the access probe success rate may be different than desired, resulting in either unacceptable Access Channel latencies or undesired uplink interference.

P-ARQ bit or ARQ channel1.5 subframe ACK

L-ARQ bit or ARQ channel1.5 subframe 3 slots ACK

H-ARQ bit or ARQ channelNAK NAK NAK First subpacket for the next physical layer packet transmission Transmit subpacket 1 Transmit sub-packet2 Transmit subpacket 3 Transmit subpacket 4 Transmit subpacket 1

Reverse traffic channel physical layer packet transmissions Subframes

n

n+ 1 n+ 2 n+ 3 n+ 4 n+ 5 n+ 6 n+ 7 n+ 8 n+ 9 n+ 10 n+ 11 n+ 12

n Figure 2. Example of normal packet termination (16-slot latency target7).In addition to a higher data rate on the initial access probe, DO Rev A allows the access network to specify the interprobe latency (within an access probe sequence). Therefore, DO Rev A can achieve significant reduction in reverse link access latency. The access channel transmit power is a function of the total received power on the forward link and the open loop turnaround parameter. If the OpenLoopAdjust is designed based on cell edge assumptions, ATs closer to the cell center would transmit more power than necessary (the access probes directed to the sector with the best forward link). DO Rev A therefore provides the ATs a mechanism to adjust the transmit power level as a function of the forward link pilot strength of the sector to which the access probe is directed. Since the open loop turnaround parameter, OpenLoopAdjust, is generated based on some assumptions of channel conditions and sector loading, the access probe success rate may be different than desired, resulting in either unacceptable access channel latencies or undesired uplink interference. Access terminals in DO Rev A include the probe number (within a sequence) in the probe transmission that permits the access network to determine the access probe success probability distribution and adapt OpenLoopAdjust to achieve the desired latencies on the access channel

. Hybrid ARQ— Substantial gains in downlink capacity were achieved in DO Rev 0 due to the use of physical layer H-ARQ[4]. DO Rev A introduces H-ARQ on the uplink in order to exploit the excess E b/N 0 due to power control imperfections and channel variations. With H-ARQ, the uplink packet transmissions are staggered in time to allow the access network to demodulate and decode the packets and then transmit an ACK to the AT indicating whether or not the transmitted packet was decoded. As shown in Fig. 2, each physical layer packet is sent using one or more four-slot subpackets (maximum of four subpackets). There are eight slots between successive transmissions of the subpackets of the same physical packet, which can be used for transmissions of other packets (total of three interlaces). Figure 2 shows the ARQ mechanism for a packet transmitted over all four subpackets. In this example the access network transmits negative ACK (NAK) responses on the ARQ channel using the H-ARQ bit after the first three subpackets are transmitted by the access terminal. After the fourth subpacket, the sector transmits an ACK using the last ARQ (L-ARQ) and packet ARQ (P-ARQ) bits indicating that the sector successfully received the packet. The L-ARQ and P-ARQ bits are used to support MAC layer ARQ which will be described later in this section. The ARQ channel (to convey ACKs for uplink transmissions) is carried on the forward MAC channel, which is limited to 256 chips/slot. The additional overhead of the ARQ channels is accommodated by reducing the power control overhead, achieved by lowering the closed loop power control update rate to 150 Hz for DO Rev A terminals (from 600 Hz for DO Rev 0). This enables transmission of ARQ, RPC, and DRC lock channels for large numbers of simultaneous users without increasing the total MAC overhead.

UPLINKPHYSICAL LAYER7

The latency target is defined as the number of slots of transmission required to achieve the desired packet error rate (PER, typically 1 percent) regardless of channel conditions.

Key physical layer enhancements to the DO Rev A uplink are: Physical layer H-ARQ and support for MAC layer ARQ Higher data rates (peak data rate of 1.8 Mb/s) and finer rate quantization Higher order modulation that allows the use of low-rate turbo codes at high data rates and large payload sizes

78

IEEE Communications Magazine February 2006

EV-DO物理层和mac层介绍

Figure 3 shows the impact of the reduced closed power control update rate on the pilot channel signal-to-noise ratio (E c/N t). With the exception of channel model A[5], we see that the Pilot E c/N t with 150 Hz closed loop power control and interlacing is lower than that with 600 Hz power control. More important, the combination of 150 Hz closed loop power control and H-ARQ results in total required Eb/N0 (after early termination) that is substantially lower (by 2 to 3 dB) than DO Rev 0, across all channel conditions. These gains are corroborated by the early termination

statistics under various channel conditions and latency targets shown in Fig. 4. Figure 4 shows the early termination probabilities for the different channel models for 16-slot and 8-slot latency targets after 4, 8, and 12 slots. The 8-slot early termination probabilities include the termination after 4 slots, and the 12-slot early termination probabilities include termination after 4 and 8 slots. Physical layer ARQ also results in an uplink coverage improvement of 1.5 dB for a nominal data rate of 9.6 kb/s. Physical Layer Structure— Figure 5 shows physical layer channels in DO Rev A with the channels new to DO Rev A and the modified channels shown in addition to the uplink physical layer subframe structure. The ACK and DSC channels are TDM. The pilot, RRI, ACK/DSC, DRC, data, and auxiliary pilot channels are orthogonally spread by Walsh functions of length 2, 4, 8, 16, or 32. Each AT uses the DSC channel to provide the access network with an early indication of the exact instant of time at which the change in downlink server takes place. The reverse link capacity impact due to the DSC channel is minimized as it is transmitted at a low gain relative to reverse link pilot (–15.5 dB for active cell size of one and–9 dB for active cell size larger than one) as it is transmitted over several slots (typically 64). The auxiliary pilot channel is used to provide reverse link channel estimation for the large reverse link physical layer packets. Physical Layer Modulation— The DO Rev A uplink supports higher-order modulation8 (i.e., QPSK and 8-PSK), and each modulated data stream is covered by either a quaternary or binary Walsh function. The modulation types supported by the DO Rev A uplink are B4 (BPSK modulation with quaternary Walsh cover), Q4 (QPSK modulation with quaternary Walsh cover), Q2 (QPSK modulation with binary Walsh cover), Q4Q2 (sum of the Q4 and Q2 modulated symbols), and E4E2 (sum of E4[8-PSK modulation with quaternary Walsh covers] and E2[8PSK modulation with binary Walsh covers] modulated symbols). The use of higher order modulation (QPSK and 8-PSK), and introduction of the binary Walsh cover and Q4Q2 and E4E2 modulation types result in more efficient use of the bandwidth, which significantly improves the code rates for medium and high data rates. In DO Rev A data rates up to 307.2 kb/s can all be transmitted with rate 1/5 turbo code[6], whereas in DO Rev 0 the 153.6 kb/s data rate could only

Impact of decimated power control+ H-ARQ on Ecp/Nt–19–20–32–22 1xEV-DO revision 0 1xEV-DO revision A 1 2 3 Channel model 4 5

Ecp/Nt (dB)

Impact of decimated power control+ H-ARQ on Eb/Nt 5 Eb/Nt (dB) 4 3 2 1 1 2 3 Channel model 4 5 1xEV-DO revision 0 1xEV-DO revision A

n Figure 3. Received pilot Ecp/Nt and total Eb/Nt vs. channel condition for a256-bit packet transmission (DO Rev 0 vs. DO Rev A).

Early termination probabilities, 256-bit payload, 16-slot LT 100 4-slot 8-slot 12-slot

Percentage

50

0 A B C D E Chan

nel models Mix

Early termination probabilities, 256-bit payload, 8-slot LT 100 4-slot

Percentage

50

0 A B C D Channel models E Mix

n Figure 4. Uplink physical layer packet early termination probability.be sent using rate 1/2 code due to the use of BPSK modulation. The DO Rev A uplink supports data rates from 4.8 kb/s to 1.8432 Mb/s and payload sizes ranging from 256 to 12,288 bits. It allows the access network to set two latency targets for each payload size for the AT. The latency target is defined by one of two transmission modes, low latency (LoLat) and high capacity (HiCap). The former uses less than four subpackets (16 slots) to transmit the packet, and the latter typically uses four subpackets. The payload sizes and latency targets from which the access network

8

DO Rev 0 uplink supports BPSK modulation with quaternary Walsh covering for data rates up to 153.6 kb/s.

IEEE Communications Magazine February 2006

79

EV-DO物理层和mac层介绍

The DO Rev A uplink supports data rates from 4.8 kb/s to 1.8432 Mb/s and payload sizes ranging from 256 to 12,288 bits. It allows the access network to set two latency targets for each payload size for the access terminal. The latency target is defined by one of two transmission modes, low latency (LoLat) and high capacity (HiCap).ACK DSC ACK Pilot Data Primary pilot Access

Reverse

Traffic

Auxiliary pilot

Medium access control

ACK

Data

Reverse rate indicator RRI Data channel DRC channel DSC ACK

Data rate control

Data source control

DSC

ACK

DSC

Auxiliary pilot channel Pilot channel 1 subframe 1 slot

n Figure 5. 1xEV-DO Revision A uplink physical layer channels and subframe structure.and AT can choose are listed in[6], with packet transmission latencies in the range of 6.66–66.66 ms, as desired by the application. Capacity and Latency Trade-Off— Due to physical layer H-ARQ in DO Rev A and latency control, nominal9 data rate is no longer an accurate indicator of the system resources used (or the interference caused to the system) by an AT or a flow of an AT. Unlike DO Rev 0 where all data rates have a fixed latency target (i.e., 16 slots), each data rate in DO Rev A is specified by payload size and latency target. For example, 9.6 kb/s can be achieved by a 256-bit packet transmitted over 16 slots or a 128-bit packet transmitted over 8 slots. Packet transmissions with shorter latency targets have less time diversity and fewer ARQ rounds, and are hence more susceptible to channel variations. The lowerlatency transmissions therefore require a larger traffic-to-pilot power ratio (T2P, i.e., system resource) to achieve the same error performance at the latency target as higher-latency transmissions of the same data rate. The AT can transmit each physical layer packet using one of two transmission modes, LoLat or HiCap. Packets transmitted in LoLat mode are power-boosted (transmitted with a higher T2P) to ensure termination within the latency target. The access network assigns an AT-specific T2P profile10

for each payload size for both transmission modes. If packets of a LoLat transmission mode fail to decode by the latency target, the AT continues transmission until the packet is successfully decoded or a total of four subpackets have been transmitted. Given a target packet error rate (PER) of 1 percent at the latency target, this achieves a sub-1 percent physical layer PER, which improves the performance of higher-layer protocols. For a given T2P allocation, the AT can trade off between capacity and latency (i.e., achieving high capacity using a large payload or low latency with a smaller payload). System Operation at High Rise-over-Thermal— System operation at high RoT may lead to power control instability in code-division multiple access (CDMA) systems, as small changes in load at high operating points result in large variations in RoT. For example, with a load of 0.7, the RoT equals 5 dB with other cell interference factor of 0.2. However, if the other-cell interference is changed to 0.4, the RoT exceeds 20 dB. The changes in load or other cell interference may be triggered by changes in transmission data rate or power control imperfections. Use of direct RoT measurement allows operation of DO Rev 0 systems at high RoT without compromising system stability. System stability at higher RoT can be further improved by reducing the delay in the RoT control loop. This is achieved in DO Rev A by the use of a quick reverse activity bit (QRAB). The Reverse Activity (RA) bit is updated every slot (1.66... ms) in DO Rev A instead of per-frame (26.66... ms) as in DO Rev 0. Due to the reduced delay of the RoT control loop the DO Rev A system can

9

The nominal data rate is the data rate achieved after transmission of the number of subframes as specified by the latency target.10

The T2P profile is set per AT for each payload size and transmission mode. A T2P profile is an ordered set of T2P values that the AT is required to use for each subpacket transmission.

80

IEEE Communications Magazine February 2006

EV-DO物理层和mac层介绍

operate at the same average RoT (5 dB in this case) as that of a DO Rev 0 system with significantly increased stability. Alternatively, it can operate at a notably higher RoT (higher RoT implies higher sector capacity) with the same stability as that of a DO Rev 0 system. Note that the overhead channel performance is unaffected as long as the AT can close the uplink, since the desired pilot SINR is maintained by power control, and all overhead channel gains are fixed relative to the pilot. MAC Layer ARQ— In order to reduce the PER seen by the upper layers, it is desirable to recover physical layer packet errors through a MAC layer ARQ mechanism with quick turnaround. DO Rev A allows the AT to retransmit the MAC layer payload contained in a packet that fails to decode even after 16 slots of transmission. MAC layer retransmission is triggered by the L-ARQ and/or P-ARQ bits that are received within 12 subframes of the start of packet transm

ission.

MEDIUM ACCESS CONTROL LAYERKey enhancements to the uplink traffic channel MAC in DO Rev A are: Comprehensive centralized control with minimal signaling overhead Unified approach to intra-AT QoS and inter-AT QoS Adaptive token bucket access control Multiflow MAC with per flow QoS control with explicit capacity, latency, and fairness trade-off per flow RoT-sensitive per flow T2P allocation Comprehensive Centralized Control— 1xEV-DO Revision A provides the access network several mechanisms for centralized control in addition to those provided in 1xEV-DO Revision 0. 1xEV-DO Revision A adopts a flow-oriented QoS approach. It achieves precise control of access terminal resource allocation with centralized resource allocation by the access network and distributed rate selection at the access terminal. The access network determines the long-term resource allocation for each flow in the network, while the AT controls the time-critical allocation (for each physical layer packet) based on the rules specified by the access network, in conjunction with local information about queue buildup and packet delay. This twosided control philosophy minimizes signaling overhead as well as delays in physical layer allocations. The AT executes the multiflow MAC algorithm using parameters downloaded from the access network, thereby synthesizing the resource allocation desired by the access network. In this framework each AT may be visualized as a state machine whose behavior is controlled by the access network. DO Rev A also defines request and grant messages for resource allocation, which may be regarded as an explicit mechanism to force the AT state machine to a desired state from time to time. In DO Rev A, the T2P is treated as a sector resource, and the uplink MAC specifies rules for mapping the continuous-valued allocation of the T2P resource into discrete-valued transmit T2P

(TxT2P) corresponding to a physical layer packet transmission. The centralized control mechanisms in DO Rev A are: Per-flow QoS control: Each flow is assigned a LoLat or HiCap attribute, which indicates the preferred transmission mode for data associated with that flow. Fast RA bit control: Per-slot update of RA bit. Max per AT TxT2P control: The access network can control the maximum TxT2P that each AT can use for uplink traffic channel transmissions. Latency shaping: By appropriate choice of TxT2P value for transmission of each payload in each transmission mode, the access network can control the latency characteristics of each AT. Flow adaptive PER: Assignment of a transmission mode (i.e., LoLat or HiCap) to a flow determines the physical layer PER for packet transmissions of each flow. Flow-specific interpretation of sector loading: While ATs typically use an“OR of the BUSY” rule to determine the effective RA bit, the access network can allow selected flows to modify the rule. HiCap flow to LoLat conversion: The access networ

k specifies rules using which a HiCap flow can be transmitted in LoLat mode based on network loading and flow QoS requirements. Explicit request and grant: In addition to the centralized resource allocation with distributed rate selection, DO Rev A supports changing MAC flow parameters and state using an explicit request and grant mechanism. Explicit capacity/fairness/interference control: DO Rev A allows the access network to explicitly trade off capacity, fairness, and latency among active flows in addition to limiting interference among flows. We next discuss the explicit request and grant mechanisms. The access network can impose conditions on whether or not the AT is required to transmit the request messages and the frequency at which it is required to transmit them. The request messages indicate the PA headroom estimate and the queue length for MAC flows for which the AT requests an explicit resource assignment. The grant messages are optionally transmitted (i.e., not coupled with the request messages) by the access network and contain the T2PInflow, duration for maintaining the T2PInflow (resource allocation) constant, and the token bucket level per flow. The T2P grants are resource allocations based on network loading dynamics, and not per packet allocations. The time-critical events are handled directly at the AT. The use of the explicit request and grant messages allows a seamless transition between the allocation based on a distributed rate selection and centralized resource allocation, and a direct centralized allocation for the desired flows. Each flow seamlessly converges to its long-term allocation as specified by the access network once the duration for which the T2PInflow allocation is frozen expires. Use of the explicit request and grant messages allows faster

DO Rev A allows the AT to retransmit the MAC layer payload contained in a packet that fails to decode even after 16 slots of transmission. MAC layer retransmission is triggered by the L-ARQ and/or P-ARQ bits that are received within 12 subframes of the start of packet transmission.

IEEE Communications Magazine February 2006

81

EV-DO物理层和mac层介绍

T2PInflow= new resource inflow (mean T2P resource based on AN assigned flow priority)

BucketLevelSat= maximum allowed bucket size

BucketLevel= unused accumulated resource

T2P= uplink MAC resource

PotentialOutflow= maximum allowed resource withdrawal Data Physical layer packet TxT2P Outflow= actual T2P withdrawal from the bucket

n Figure 6. Token-bucket-based access control (DO Rev A).convergence to the desired resource allocation and aids in QoS enforcement and policing. Unified Approach to QoS— The DO Rev 0 MAC supports only a single uplink MAC flow per AT. In other words, there is no differentiation of packets from different applications with distinct QoS requirements. The DO Rev A uplink MAC enables intra-user QoS. Packets from latency-sensitive flows that arrive at the access terminal following a large packet for d

elay-tolerant flow are transmitted using the first MAC packet by an intra-AT QoS aware (multiflow) MAC in contrast to the operation of an intra-AT QoS unaware (single-flow) MAC where packets are transmitted in order of arrival, which results in higher and unpredictable latency and jitter— both undesirable for real-time multimedia applications. The DO Rev A uplink MAC algorithm specifies the behavior of each MAC flow in addition to the AT behavior as specified in DO Rev 0. The DO Rev 0 approach relies on AT implementations to appropriately allocate resources among the applications, whereas the DO Rev A approach allows access network direct control over behavior of each flow. The DO Rev 0 approach requires renegotiation of assigned rate transition probabilities each time the nature or composition of the constituent flows at the AT changes. The DO Rev A MAC uses a unified approach to inter-AT QoS and intra-AT QoS, and decouples flow behavior from its location, so flows with the same MAC parameters achieve the same performance irrespective of whether residing at a given AT or across multiple ATs (subject to power and system load constraints). Adaptive Token Bucket Access Control— The DO Rev A uplink MAC uses an adaptive token bucket access control per flow with T2PInflow as the adaptation parameter, as shown in Fig. 6. This allows a restriction to be imposed on the source behavior and the contribution of each flow to the sector RoT to be controlled. With this mechanism the AT appears as a token bucket regulator to the rest of the network. A token

bucket is defined for each uplink traffic channel MAC flow and represents the stored T2P resource. Control of the token bucket and other MAC parameters enables access network control of the transmitted T2PInflow. The token bucket allows smoothing of T2PInflow allocation when interarrival time between higher layer packets is short. With appropriate parameter settings, the access network can also grant an AT the ability to handle bursty sources (i.e., high peak-to-average throughput) with well-defined restrictions on the peak-to-average ratio of the TxT2P. The token bucket is filled each subframe with T2PInflow and emptied by the actual TxT2P allocation for a physical layer packet transmission. Figure 7 shows a comparison of the throughput and bit delay between DO Rev 0 and DO Rev A using a temporal trace. For the example presented here, both systems use the default settings— the default rate transition probabilities for DO Rev 0 and the default MAC parameters for DO Rev A. It can be seen that the flow throughput for DO Rev A is nearly constant, while in DO Rev 0 it suffers from startup effects due to the rate transition probabilities and a startup data rate of 9.6 kb/s. The bit delay for DO Rev A is low and constant, leading to significantly lower jitter than DO Rev 0. DO Rev A terminals can achieve instantaneous higher-rate transmission without sacrificing transmit power efficiency with interferenc

e limited by the token bucket access control mechanism. The adaptive token bucket mechanism allows not only improved support for applications with latency-sensitive data bursts, but also uniform allocation independent of the burstiness of a flow. While the use of rate transition probabilities in DO Rev 0 controls how quickly the access terminal can transmit data from a given flow, the use of the adaptive token bucket mechanism in DO Rev A allows a quicker transition to a higher rate and also permits initial transmissions at a higher data rate. Furthermore, the deterministic token bucket mechanism reduces variation in delay (jitter) compared to the probabilistic algorithm in DO Rev 0. A source with a short data burst continues to receive the same allocation over time and uses the stored resources to transmit the data bursts. This permits self-tuning for each flow. For example, a video flow generates short data bursts periodically to convey scene changes. Since the T2PInflow increases at a steady rate, it fills up the token bucket, and this stored resource is used to transmit the scene change information at a high data rate, as needed, to achieve the desired latency. This peak-to-average ratio is controlled by the access network. The adaptive token bucket mechanism also specifies a saturation level for each flow, and a bucket level in excess of the saturation level indicates overallocation that may be due to either a data limitation or a power amplifier headroom limitation. In this event the T2PInflow reduces to the access network granted minimum value and starts increasing only when the flow is no longer overallocated. The adaptive token bucket mechanism also imposes a restriction on the T2PInflow increase based on the difference between the average T2PInflow and

82

IEEE Communications Magazine February 2006

EV-DO物理层和mac层介绍

T2POutflow. That is, the resource allocation to a flow is a function of the resource utilization by that flow. This enables self-tuning of the flow’s resource allocation based on the arrival characteristics of the data source and the power amplifier headroom at the AT. Flows with a fixed T2PInflow allocation (e.g., a low-rate delay-sensitive flow) stop T2PInflow increase once adequate T2PInflow has been accumulated for the next burst of data. Multiflow Uplink MAC— The delay-sensitive low-rate flow (e.g., VoIP) and delay-sensitive high-rate flow (e.g., fixed-rate video telephony) are assigned a fixed allocation that permits the flow to use the desired system resources (up to a certain T2PInflow) except under extreme system loading. A fixed allocation attempts to provide a circuit-like performance to the flow, while taking advantage of statistical multiplexing when the flow does not utilize its allocation. The delaysensitive elastic flow (e.g., adaptive-rate video telephony) is assigned an elastic allocation with a fixed component that permits this flow to utilize resources in addition to a fixed allocation if the sector is lightly

loaded. The delay-tolerant best effort source (e.g., file transfer) utilizes the resources not currently used by the higher-priority flows in the system. We now illustrate the impact of power amplifier headroom limitations on flow throughput and latency using the flows for which the priority functions are as shown in Fig. 8. This effect is illustrated in Fig. 8 for two ATs with the same delay-sensitive low-rate, delay-sensitive highrate, and delay-tolerant best effort flows but with different PA headroom limitations (due to ATs in different locations experiencing different path loss). We see that the delay-sensitive low-rate flows achieve the desired allocation in both cases, but the throughput achieved for the delaysensitive high-rate and best effort flows is different due to different PA headroom limits at the two ATs. The relative priority of the flows at the two ATs— higher priority for the delay-sensitive high-rate flows relative to the best effort flows— is maintained as the path loss experienced by the terminals increases. Load-Sensitive Resource Allocation— The uplink rate transition probabilities in DO Rev 0 are designed conservatively to ensure stable system performance when the uplink is heavily loaded, leading to inefficient resource utilization when the uplink is lightly loaded. DO Rev A improves on the dynamics of resource utilization by introducing load-sensitive resource allocation that allows the ATs to use a greedy resource allocation strategy subject to feedback control from the access network. In other words, the rate of increase in T2P allocation for a flow is dependent on the system load level and allows a trade-off between convergence rate and variation in resource allocation as a function of load. If the DO Rev 0 transition probabilities were designed assuming a lightly loaded sector, the RoT CCDF tails would be excessive when the sector is heavily loaded, resulting in loss of coverage and higher potential for instability. The load-sensitive resource allocation in DO Rev A

103

Delay and throughput comparison between DO Rev. 0 and DO Rev. A with bursty source DO Rev 0 DO Rev A DO rev 0: avg. rate= 55.9 kb/s (16-slot avg.) DO rev A: avg. rate= 53.5 kb/s (16-slot avg.)

Rate (kb/s)

102

101 100 0 1000 2000 3000 (a) 0.5 0.4 DO Rev 0: mean delay= 129.7 ms, max delay= 331.7 ms DO Rev A: mean delay= 36.8 ms, max delay= 75.0 ms 4000 5000 6000

Bit delay (s)

0.3 0.2 0.1 0 0 1000 2000 3000 Time (slot) (b) 4000 5000 6000

n Figure 7. Throughput and delay comparison between 1xEV-DO Revision 0and 1xEV-DO Revision A for a bursty data source.

103

Delay-sensitive, low-rate 1 Delay-sensitive, high-rate 1 Best effort 1 Delay-sensitive, low-rate 2 Delay-sensitive, high-rate 2 Best effort 2

102 Throughput (kb/s)

101

100 0 5 10 15 20 Increase in AT path loss (dB) 25 30

n Figure 8. Impact of PA headroom limitation on flow throughput at differentaccess terminals with the same flow priorities. results in the desired beh

avior without increasing the likelihood of system instability. Explicit Capacity, Latency, and Fairness Control— DO Rev A provides several mechanisms for explicit interference control: Maximum transmit T2P control as a function of serving sector pilot strength Short-term sector loading as a function of downlink sector pilot strength, pilot strength of other sectors in the AT’s active set, and the corresponding DRCLock bit

IEEE Communications Magazine February 2006

83

EV-DO物理层和mac层介绍

1 0.9 0.8

Biased MAC-case1 Default MAC Biased MAC-case2

with the rest of the bits made up of padding. Improvement in packing efficiency for terminals in poor channel conditions can be achieved by the use of short packets with payload sizes of 128, 256, or 512 bits.

MEDIUM ACCESS CONTROL LAYER0.7 0.6 cdf 0.5 0.4 0.3 0.2 0.1 0 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 AT relative throughput

Key enhancements to DO Rev A downlink MAC layer are: Packet-division multiple access via the use of multi-user packets One-to-many mapping of DRC index to transmission formats Seamless adaptive server selection Application-adaptive physical layer PER Reduction in transmission delay via nullDRC to non-null rate mapping Packet-Division Multiple Access— Substantial improvement in link and packing efficiency can be achieved by the use of multi-user packets (MUPs), transmitting data to multiple ATs using the same physical layer packet. This technique is called packet-division multiple access (PDMA). It enables DO Rev A to support a large number of low-rate delay-sensitive applications. A MUP is a single physical layer packet containing data for multiple ATs (maximum of eight ATs per packet). The downlink scheduler continues to serve single-user packets (SUPs) using opportunistic scheduling to exploit multi-user diversity where possible. DRC Index to Transmission Format Mapping— A one-to-one mapping between requested DRC and data rates/packet sizes is used in DO Rev 0. A more flexible mapping allows for improved packing efficiency as well as better matching of the requirements of the flow to the physical layer transport. Each MAC packet may be transmitted using several different transmission formats that are consistent with the MAC packet size. As an example, the transmission formats (1024, 16, 1024), (1024, 8, 512), (1024, 4, 256), (1024, 2, 128), (1024, 1, 64), (1024, 2, 64), and (1024, 4, 128) are consistent with a 994-bit MAC packet. Each DRC index in DO Rev A has a set of associated transmission formats for single- user packet and multi-user packet. For example, DRC index 0x3 is associated with transmission formats (128, 4, 256), (256, 4, 256), (512, 4, 256), and (1024, 4, 256), where (1024, 4, 256) is defined as the canonical transmission format and is the transmission format associated with DRC index 0x3 in DO Rev 0. All the other consistent transmission formats above are called non-canonical transmission formats. A detailed listing of DRC indices and

their associated transmission formats is provided in[6]. Since ATs always attempt to decode MUPs with packet transmission formats of (128, 4, 256), (256, 4, 256), (512, 4, 256), and (1024, 4, 256), the access network can serve packets to an AT whose DRC is erased using these transmission formats. This allows for improved performance of delay-sensitive applications. Given a nominal PER target of 1 percent, use of short packets (includes packets with payload sizes less than 1024 bits and non-canonical SUPs) and MUPs may result in a sub-1 percent physical

n Figure 9. Reverse link rate shaping in 1xEV-DO Revision A. T2PInflow scaling as a function of downlink serving sector pilot strength Figure 9 shows examples of rate shaping for the uplink, which is a form of explicit interference control. With appropriate choice of DO Rev A uplink MAC channel parameters, any desired normalized AT throughput and consequently interference distribution can be achieved. For the examples shown in Fig. 9, the fairness among users can be either decreased (biased MAC case 1) or increased (biased MAC case 2) as desired. DO Rev 0 MAC does not provide mechanisms for uplink rate shaping.

DOWNLINKPHYSICAL LAYERKey enhancements to DO Rev A downlink physical layer are: Short packets (i.e., 128-, 256-, and 512-bit packets) Higher peak data rates (3.1 Mb/s) and finer rate quantization DO Rev A downlink physical layer packets are defined by their transmission formats. The transmission format is an ordered triple defined by the physical layer packet size (bits), nominal packet duration (slots), and the preamble length (chips). For instance, (128, 16, 1024) indicates that the packet has a 128-bit payload, nominal duration of 16 slots, and a 1024-chip preamble. Transmission formats and their corresponding code rate, modulation, and nominal data rates are listed in[6]. DO Rev A introduces new packet sizes of 128, 256, 512, and 5120 bits in addition to the 1024-, 2048-, 3072-, and 4096-bit packet sizes in DO Rev 0. In addition, DO Rev A permits nominal spans of one through 16 slots, resulting in data rates ranging from 4.8 kb/s to 3.072 Mb/s. The smallest physical layer packet size on DO Rev 0 downlink is 1024 bits, and the packet may contain only a few bits of delay-sensitive traffic

84

IEEE Communications Magazine February 2006

EV-DO物理层和mac层介绍

layer PER as the network may transmit packets using a compatible transmission format that provides additional SINR margin. For example, a terminal with a requested DRC corresponding to the transmission format (5120, 1, 64) may be served using a MUP with transmission format (1024, 4, 1024). Adaptive Server Selection— The DO Rev 0 downlink supports virtual soft handoff via adaptive server (base station) selection. Server changes in DO Rev 0 may result in packet transmission delay that does not impact best effort traffic, but may limit the performance of delaysensitive applications. In DO Rev 0 ATs indicate downlink server changes vi

a the DRC channel by changing the serving sector information (also called the DRC cover). When the access network detects a consistent change in the DRC cover, it reroutes the data queue for the AT to the new server. During this transition phase, neither the old nor the new server is capable of transmitting packets to the AT. DO Rev A eliminates this transmission delay via the use of the data source control (DSC) channel. Each AT uses the DSC channel to provide the access network early indication of the exact instant of time at which the change in downlink server takes place, as shown in Fig. 10. As a result, the data queue is already set up at the new server by the time the AT points its DRC to the new server. Application-Adaptive Physical Layer PER— DO Rev A systems permit access network control of the downlink physical layer PER based on the composition of the flows to an AT. This is accomplished by providing the AT with a mapping from the tentative DRC to a transmitted DRC. With an appropriate mapping, additional rounds of H-ARQ are added to the physical layer without affecting the overall downlink capacity. For example, a mapping from a tentative DRC corresponding to the transmission format of (4096, 1, 64) to a transmitted DRC corresponding to the transmission format of (4096, 2, 64) results in an additional slot being transmitted for this packet 1 percent of the time (assuming a target PER of 1 percent after one slot) and permits achieving approximately 0.01 percent PER after two slots. Null-Rate Conversion— The DO Rev A downlink supports null-rate conversion where the null-rate DRC received from the AT can be mapped to transmission formats (1024, 16, 1024), (512, 16, 1024), and (256, 16, 1024). Since ATs that support MUP reception attempt to detect the above transmission formats, null-rate conversion provides the access network a mechanism to serve ATs with null-rate requested DRC. In a vast majority of cases, a null-rate DRC converted to the transmission format (1024, 16, 1024) results in an effective data rate exceeding 38.4 kb/s. Similar gains are seen for smaller payload sizes transmitted in response to a null-rate DRC. In addition, the use of short packets and non-canonical SUPs for low-rate delay-sensitive traffic improves downlink transmit efficiency (transmission of 100 bits of data using transmis-

Transmit delay Forward link serving cell BS1 t1

t2

Forward link serving cell BS2

DRC cover change DRC cover change detection at BS1 and BS2

Qtransfer to BS2

Time

IS-856 Revision 0

Forward link serving cell BS1 t1 t2

Forward link serving cell BS2

DSC cover change DRC cover change detection at BS1 and BS2

Qtransfer to BS2 DRC cover change

Time IS-856 Revision A

n Figure 10. Adaptive server selection (DO Rev 0 vs. DO Rev A).sion format (128, 4, 1024) results in a worst-case usage of four slots on the downlink).

PERFORMANCESIMULATION MODELS AND PARAMETERSThe capacity simulations are based on a network of 57 sectors ar

ranged in a hexagonal grid as specified in[5]. The channel models, A through E, are also specified in[5]. The system simulation is complete in the sense that all algorithms at each layer within the access network and ATs are modeled. More specifically, the simulations include a model for the downlink equalizer along with the thermal noise set by the receiver noise figure, post-AGC receiver self noise, baseband pulse shape, and the effect of base station waveform quality (rho factor). In addition, the signaling on both the downlink channels (ARQ and RPC) and uplink channels (DRC, DSC, RRI, and ACK) is modeled. The uplink outer loop power control and MAC algorithm with direct RoT measurement are also modeled. The downlink scheduler with support for MUPs and short packets is simulated along with the MAC channel power allocation algorithm.

SYSTEM PERFORMANCETable 1 summarizes the results using the fullbuffer model for both DO Rev 0 and DO Rev A systems. The forward link capacity numbers for DO Rev A assume the use of an MMSE equalizer. Table 2 shows the comparison of uplink capacity and latency between DO Rev 0 and DO Rev A. The sector capacity for DO Rev A with 10 ATs per sector, all with 8-slot termination target, is roughly the same as that for DO Rev 0, but with over 50 percent reduction in mean delay (including queuing and transmission delay). On the other hand, when using a 16-slot

IEEE Communications Magazine February 2006

85

EV-DO物理层和mac层介绍

Criteria Forward link sector capacity (full buffer) 16 users/sector dual antenna receiver 16 users/sector dual antenna receiver 16 users/sector fourantenna receiver

1xEV-DO Rev 0 1200 kb/s

1xEV-DO Rev A 1500 kb/s

310 kb/s

500 kb/s

Reverse link sector capacity (full buffer)

600 kb/s

1200 kb/s

n Table 1. DO Rev A and DO Rev 0 performance comparisons.DO Rev 0 Number antenna 2 Number of ATs 10 Sector capacity (kb/s) 320 Mean RoT (dB) 5 Mean delay (ms) 40

cy, support large numbers of mobile users, sub30 ms round-trip latencies to support interactive applications, provide performance comparable to toll-quality voice applications, support end-toend IP QoS that allows operators to maximize revenue through tiered services, and provide comprehensive network control over terminal behavior.

DO Rev A 8-slot termination target Number antenna 2 Number of ATs 10 Sector capacity (kb/s) 318 Mean RoT (dB) 5.2 Mean delay (ms) 15.2

ACKNOWLEDGMENTSThe authors thank Matt Grob and Rajesh Pankaj for their constructive comments that led to many improvements to this article, and Mehmet Gurelli for simulation data used for several illustrations and the applications capacity.

DO Rev A 16-slot termination target Number antenna 2 4 Number of ATs 10 10 Sector capacity (kb/s) 628 1325 Mean RoT (dB) 5.2 5.6 Mean delay (ms) 39.5 40

REFERENCES[1] P. Bender et al.,“CDMA/HDR: A Bandwidth Efficient High-Speed Data Service for Nomadic Users,” IEEE Commun. Mag., vol. 38, July 2000, pp. 70–77.[2] P. J. Black and M. I. Gurelli,

“Capacity Simulation of CDMA2000 1xEV Wireless Internet Access System,” 3rd IEEE Int’l. Conf. Mobile and Wireless Commun. Networks, Recife, Brazil, Aug. 2001.[3] Y. Jou,“Developments in Third Generation (3G) CDMA Technology,” Proc. IEEE 6th Symp. Spread-Spectrum Tech. and App., Newark, NJ, Sept. 2000.[4] Q. Wu and E. Esteves, Advances in 3G Enhanced Technologies for Wireless Communications J. Wang and T. Ng, Eds., Ch. 4, 2002.[5] 3GPP2 TSG-C WG3 Contribution C30-20031002-004,“1xEV-DO Evaluation Methodology,” Oct. 2004.[6] 3GPP2 C.S20024-A v1.0,“CDMA2000 High Rate Packet Data Air Interface Specification,” Mar. 2004.

n Table 2. DO Rev 0 and DO Rev A (capacity and latency trade-off).termination target, the sector capacity for DO Rev A is almost doubled relative to that of DO Rev 0 while achieving roughly the same mean delay. A 32-byte ping transmission over an experimental 1xEV-DO Revision A system showed that 90 percent of the pings experience an endto-end delay of less than 30 ms. and nearly all packets experience an end-to-end delay of less than 32 ms.

ADDITIONAL READING[1] 3GPP2 C.S20024 v2.0,“CDMA2000 High Rate Packet Data Air Interface Specification,” Oct. 2000.[2] E. Esteves, P. J. Black, and M. I. Gurelli.“Link Adaptation Techniques for High-Speed Packet Data in Third Generation Cellular Systems,” Euro. Wireless Conf., 2002.[3] N. Bhushan and P. J. Black,“Forward Link Coding and Modulation Design for CDMA2000 1xEV (IS-856),” PIMRC 2002, Lisbon, Portugal, Sept. 2002.[4] S. Chakravarty, R. Pankaj, and E. Esteves,“An Algorithm for Reverse Traffic Channel Rate Control for CDMA2000 High Rate Packet Data Systems,” GLOBECOM 2001, San Antonio, TX, Nov. 2001.[5] 3GPP2 TSG-C WG3 Contribution C30-20031013-209R4,“Link Budget for 1xEV-DO Revision A,” Oct. 2004.[6] M. Fan et al.,“On the Reverse Link Performance of CDMA2000 1xEV-DO Revision A System,” ICC 2005.[7] C. Lott et al.,“Reverse Traffic Channel MAC Design of CDMA2000 1xEV-DO Revision A System,” VTC 2005.

SUMMARY AND CONCLUSIONS1xEV-DO Revision A is designed to offer efficient support for both delay-sensitive and delaytolerant applications. Salient features added to 1xEV-DO Revision A are short packets and multi-user packets on the downlink, and physical layer ARQ and multiflow reverse link MAC layer on the uplink. The inclusion of these features provides substantial improvement in the performance of delay-sensitive applications such as VoIP, gaming, and video telephony. 1xEV-DO Revision A is fully backward compatible with 1xEV-DO Revision 0 networks, and an upgrade involves only a change to the mobile station and base station ASICs with no other change to the base station hardware. 1xEV-DO Revision A systems deliver high spectral efficien-

BIOGRAPHIESN AGA B HUSHAN (nbhush@) obtained his B.Tech. degree in electronics from the Indian Institute of Technology, Chennai, in 1989. He pursued his graduate study at Corne

ll University, where he secured his M.S. and Ph.D. degrees in 1992 and 1994, respectively, both in electrical engineering. He has been working as a systems engi-

86

IEEE Communications Magazine February 2006

EV-DO物理层和mac层介绍

neer at QUALCOMM since 1994, where he is now vice president of technology in the Corporate R&D group. During the course of his graduate study and professional career at QUALCOMM, he has been involved in the analysis, design, and development of high-speed wireless communication systems, with emphasis on channel coding, link adaptation techniques, modem design, advanced transmission and receiver techniques, MAC design, and performance optimization for high-speed wireless packet data systems. CHRISTOPHER LOTT received his B.S.E.E. from Massachusetts Institute of Technology (MIT), his M.S.E.E. from Stanford, and his Ph.D. from the University of Michigan, where he received the Lucent Distinguished Dissertation award. Prior to joining QUALCOMM he worked at Hewlett-Packard on applied statistical signal processing projects, at Trimble Navigation on GPS and Inmarsat Std-C system development, and as a technical consultant in Trimble's Asian market. Since he joined QUALCOMM in 2001 he has been working on 1xEV-DO system design and standardization, with an emphasis on resource allocation problems, QoS, and MAC design. His research interests include stochastic systems, resource allocation, distributed algorithms, communication theory, dynamical systems, and wireless networks. PETER J. BLACK is senior vice president of technology for Corporate Research and Development, QUALCOMM Incorporated. He joined QUALCOMM in April 1993, where he was first engaged in the system design and development of dual mode CDMA/AMPS mobile station ASICs. In 1997 he co-led the system design and prototype development of a high-speed cellular packet data system known as HDR. This system design was the framework for the cdma2000 highrate packet data standard more commonly known as 1xEVDO, published in 2000. He also co-led the subsequent commercialization of 1xEVDO, which has now achieved large-scale deployments in all major markets. Since 2001 he has continued to contribute to the evolution and enhancements of the EVDO standard and products. Most recent initiatives include hybrid OFDM broadcast, VOIP, and multicarrier EVDO. He received his B.E. degree in electrical engineering from the University of Queensland, Australia, in 1985. He received his M.S.E.E. and Ph.D. degrees from Stanford University, California, in 1990 and 1993, respectively. He is a Fulbright Scholar and was awarded the University Medal by the University of Queensland in 1985. R ASHID A TTAR (rattar@) obtained his B.E. degree in electronics from Bombay University in 1994 and M.S.E.E from Syracuse University in 1996. He joined QUALCOMM in June 1996, where he was first engaged in the integration of IS-95-based cellular systems. Since 1998 he has been working in the Corporate Research and Development

group on 1xEV-DO system design, development, standardization, and optimization as a senior staff engineer/manager. He is currently working on 1xEV-DO Revision B and future technologies, and pursuing his Ph.D. at the University of California at San Diego, under the guidance of Prof. Larry Milstein. His research interests include topics in multi-access techniques and multihop cellular networks.

Additionally, he has been awarded several patents throughout his career. Y U -C HEUN J OU received his B.S.E.E. degree from National Taiwan University, and his M.S.E.E. and Ph.D. degrees from the University of Southern California. He has been with QUALCOMM since 1989 and is currently vice president of technology. He served as chief technology officer of QUALCOMM China from November 2003 to June 2005. At QUALCOMM he has been involved in designing, developing, testing, and standardizing CDMA cellular technologies. He was also involved in the design of the Globalstar LEO satellite system. He is a key contributor to the design and standardization of the third-generation (3G) wireless system for voice and data services, especially to the development of the IS-2000 (cdma2000 1X) and IS-856 (cdma2000 1X-EV-DO) series of standards. He has a wide area of interests in wireless communications, recently focusing on designing wireless data communication systems to support applications with different QoS requirements as well as designing advanced receivers to maximize the performance of existing standards. He is the author or co-author of approximately 50 U.S. patents. M INGXI F AN received his Bachelor’s, Master’s, and Ph.D degrees in electrical engineering in 1999 and 2002, respectively, from MIT. From 1996 to 2001 he was a research intern with the systems group at Hughes Network Systems, San Diego, California, where he designed and implemented various signal processing algorithms for several wireless and satellite communication systems. In summer 2000 he was a DSP consultant and project leader at Vanu, Inc., Cambridge, Massachusetts, where he implemented the physical layer processing of an IS-95B system on a software radio platform. He joined QUALCOMM, Inc. in July 2002 and is currently working on cdma20001xEV-DO related research, implementation, and standards development. His research interests include topics in spread-spectrum modulation, multiuser detection, and adaptive antenna array. He received the Ernst A. Guillemin EE Master's thesis award from MIT in June 1999. DONNA GHOSH received a Ph.D. degree in computer science and engineering from the Pennsylvania State University, University Park, in June 2003. Her Ph.D. research work was awarded the NSF-ITR grant in 2002 in the area of highspeed networking. She joined the Corporate Research and Development Group at QUALCOMM Incorporated, San Diego in July 2003. Her research interests are in the areas of pricing and QoS for wired and wireless networks, and stochastic modeling and analysis for wired and wireless ne

tworks. JEAN AU received his B.A. (economic), B.Sc and M.Sc (both in electrical engineering) from Queen's University, Kingston, Canada in 1995, 1996, and 1999 respectively. He joined QUALCOMM in 1999 as a systems engineer with the Corporate Research and Development Department. He was involved in the design, verification, optimization, and deployment of the CDMA2000 1xEV-DO system. Several patents have been filed by him throughout his career.

1xEV-DO Revision A is designed to offer efficient support for both delay-sensitive and delay-tolerant applications. Salient features added to 1xEV-DO Revision A are short-packets and multi-user packets on the downlink, and physical layer ARQ and multi-flow reverse link MAC layer on the uplink.

IEEE Communications Magazine February 2006

87

本文来源：https://www.bwwdw.com/article/slx4.html