Orthogonal Frequency Division Multiple Access in ... - IEEE Xplore

3 downloads 0 Views 238KB Size Report
The first versions of WiMAX (WiMAXv1) and LTE (LTE R8) are not classified as 4G wireless systems as they do not meet the requirements drafted by the Inter-.
FERNANDO LAYOUT_Layout 1 8/23/12 12:55 PM Page 153

TOPICS IN RADIO COMMUNICATIONS

Orthogonal Frequency Division Multiple Access in WiMAX and LTE: A Comparison S. Srikanth and P. A. Murugesa Pandian, Anna University Xavier Fernando, Ryerson University

ABSTRACT Orthogonal frequency-division multiple access has been adopted by both the WiMAX and Long Term Evolution standards, which are the two main contenders in the 4G wireless scenario. In this article, we do a detailed comparison of the implementation of OFDMA in LTE and WiMAX. The multiuser diversity advantage of OFDMA is well used in LTE, whereas the frequency diversity advantage is nicely exploited in WiMAX. The physical layer overhead in LTE is significantly better than in WiMAX. The network entry process of an LTE mobile is simpler than in a WiMAX mobile. In addition, LTE systems are expected to react faster to changes in the traffic and channel conditions due to the smaller frame times. However, the OFDMA design in WiMAX will cater for more extreme channel conditions, which is perhaps more than what is expected in practice as our investigation reveals.

INTRODUCTION IEEE 802.16e-based WiMAX and Third Generation Partnership Project (3GPP)-based Long Term Evolution (LTE) are the two standards that are likely to dominate the fourth generation (4G) wireless landscape [1, 2]. The first versions of WiMAX (WiMAXv1) and LTE (LTE R8) are not classified as 4G wireless systems as they do not meet the requirements drafted by the International Telecommunication Union (ITU). However, the evolutions of WiMAX (based on IEEE 802.16m and henceforth called WiMAXv2) and the evolution of LTE called LTE–Advanced (LTE-A) are the main contenders for 4G wireless systems. Both evolutions are expected to be backward compatible with their preceding versions; that is, WiMAXv2 will be compatible with WiMAX, and similarly, LTE-A will interoperate with LTE. Consequently, operators are likely to deploy the first versions of WiMAX and LTE, and then upgrade to the 4G versions as appropriate. However, recent news from the industry indicates that LTE and its evolutions are likely to dominate over WiMAX in 4G wireless deployments.

IEEE Communications Magazine • September 2012

Both LTE and WiMAX standards use several common technologies with subtle differences [3]. The main common technology is orthogonal frequency-division multiple access (OFDMA) [4]. In WiMAX, OFDMA is used on both the downlink and the uplink, whereas in LTE it is used only on the downlink. However, the technology used in the uplink of LTE, single-carrier frequency division multiple access (SCFDMA), is nothing but a simple modification of OFDMA [2]. Therefore, many of the OFDMA related points are valid for SCFDMA as well, and we shall point out the minor differences as we go. There are many good reasons for choosing OFDMA such as multipath handling capability, scalability of operation in different bandwidths, the ability to handle different data rates, and the ability to easily combine with multiple antenna techniques [4]. Mainly, OFDMA enables relatively simple channel compensation techniques in frequency selective fading channels, which makes it popular [4]. In addition, frequency diversity and channel feedback can be used effectively to improve robustness and throughput [4]. Therefore, orthogonal frequency-division multiplexing (OFDM) is preferred in most modern communication systems like digital subscriber line (DSL), power line communications, and wireless local area networks (WLANs). Now 4G cellular has also adopted OFDM as the base technology due to these advantages. Next-generation wireless systems need to handle both data and voice communications in a seamless manner. Second-generation (2G) and 3G technologies like GSM, IS-95, and UMTS (3G) mainly cater to voice with some support for low to medium data rates. However, advanced technologies like HSPA and EVDO, which are backward compatible with UMTS and IS-95, respectively, do support high data rates [3]. Such an approach needs investments in multiple radio technologies and core networks leading to high cost. For 4G wireless systems, an integrated radio and core network catering to various services is envisaged [1, 2]. OFDMA technology helps to achieve this objective as resources can be split into smaller granular units and allocated for var-

0163-6804/12/$25.00 © 2012 IEEE

153

FERNANDO LAYOUT_Layout 1 8/23/12 12:55 PM Page 154

ious services as required. The same radio frequency (RF) spectrum and bandwidth may not be allocated for 4G networks in different countries due to various regional and national constraints. Hence, there is uncertainty about the bandwidth availability for 4G wireless systems. The OFDMA technology fits this bill as it can be designed to operate under different bandwidth conditions while still keeping essential basic parameters like the subcarrier spacing constant, leading to reuse of baseband algorithms in different implementation scenarios. This is the reason for the use of the term scalable OFDMA in many documents on 4G wireless systems [1]. The ability of OFDM systems to integrate well with multiple-input multiple-output (MIMO-OFDM) technology is an important factor in the choice of OFDMA [1, 2]. It is well known that MIMO technology helps achieve high spectral efficiencies as MIMO-OFDM has successfully proven in IEEE 802.11n. Certain disadvantages like high peak-to-average power ratio (PAPR) have been the reason for not using OFDMA for the uplink in LTE. However, in WiMAX, intelligent use of the time-frequency resources can reduce the PAPR burden on the mobile and make it practical [1]. The focus of this article is on the comparative use of OFDMA in LTE and WiMAX systems. We shall compare the usage and relate it to various concepts like frequency diversity and multiuser diversity (MUD). For example, the MUD advantage of OFDMA is well used in LTE, whereas the frequency diversity advantage is

5 ms frame Downlink subframe

Uplink subframe

Control

Control

Preamble

Subchannels

User data

Slot 1 Slot 2

Slot 1

Slot N

Slot N

Symbol 1

10 ms frame 1 ms

Slot 1

Slot 0

Subframe 0

......

Subframe 1

Subframe 9 (b)

Figure 1. Frame structures in a) WiMAX; b) LTE.

154

USE OF OFDMA IN WIMAX AND LTE FRAME STRUCTURE In WiMAX, time-division duplexing (TDD) is more popular. Typical frame duration is 5 ms with common deployment bandwidths of 5 and 10 MHz [1, 5]. The frame is divided into number of OFDM symbols (e.g., 48), some of which are allocated for downlink and the rest for uplink transmissions as desired by the system operator. The first symbol in the frame is used for preamble transmission. Control and data transmissions are sent using subchannels, which are formed out of a group of subcarriers. A typical allocation spans the subchannel and symbol axes, and typically a 2D region is assigned for both downlink and uplink transmissions, as shown in Fig. 1a. The base station (BS) announces a schedule every frame period (e.g., 5 ms) to convey the downlink and uplink allocations [1, 5]. In LTE, the frame duration is fixed at 10 ms, which is divided into subframes of 1 ms duration. Two slots of 0.5 ms duration are formed out of a subframe, as shown in Fig. 1b. The BS schedules transmissions every 1 ms, which is also known as the transmission time interval (TTI), and resource blocks are formed from the subcarriers for allocation on the downlink. Currently, frequency-division duplexing (FDD) is the popular deployment scenario for LTE. However, a strong time-division duplex (TDD)-based LTE deployment called TD-LTE is also likely to be deployed in countries like China and India [2].

RESOURCE MAPPING FROM SUBCARRIERS

(a)

0.5 ms

nicely exploited in WiMAX systems in most deployment scenarios. In addition, LTE systems are expected to react faster to changes in traffic and channel conditions due to the small frame times compared to WiMAX. To quantify the differences in the use of OFDMA in WiMAX and LTE, we analyze the physical layer overhead in both systems. We also calculate the physical layer related channel parameter bounds that can be tolerated and highlight the differences between downlink and uplink usage. Finally, we review the signals that have to be observed by a mobile before it performs the network entry to give an idea of the complexity involved in these systems with respect to this important step.

Subcarriers (also referred to as resource elements [REs] in LTE) are the smallest granular units in the frequency domain, and OFDM symbol duration is the smallest granular unit in the time domain in OFDMA systems. However, because subcarriers are too large in number, groups of subcarriers are considered together in an OFDM symbol. On the time axis, a group of OFDM symbols are handled together to minimize the signaling overhead while still achieving granularity in the achievable rates to support various services. In WiMAX, subchannels are formed from a group of subcarriers in an OFDM symbol. There are two important subchannelization methodologies; the first mandatory one is based on a distributed subcarrier grouping, called partially used

IEEE Communications Magazine • September 2012

FERNANDO LAYOUT_Layout 1 8/23/12 12:55 PM Page 155

subcarriers (PUSCs), and the other optional method is based on adjacent subcarrier grouping, called band adaptive modulation and coding (BAMC) [1, 5]. The number of subcarriers in a subchannel is not the same in both methods. A slot is formed by combining a subchannel with different numbers of OFDM symbols. Fortyeight data subcarriers are available in a slot in all WiMAX subchannelization methods. For example, in a PUSC, 24 subcarriers over 2 consecutive OFDM symbols make up a slot, while in the popular BAMC implementation, the slot is made up of 16 subcarriers spread out over 3 OFDM symbols [1, 5]. In PUSC, the subchannel bandwidth is about 250 kHz, while in BAMC it is about 170 kHz. The BS allocates an integer number of slots for transmitting data and control information. Hence, the minimum possible physical (PHY) layer data rate on the downlink assuming a 5 ms frame is 9.6 kb/s, assuming that this minimum allocation is granted in every WiMAX frame. The number of useful subcarriers in a slot is the same in the uplink of WiMAX. The differences between the downlink and uplink lie in the use of pilots. In LTE, 12 adjacent subcarriers are grouped as a unit in the frequency axis, and 7 OFDM symbols (or 6 OFDM symbols in special cases) are considered as a unit in the time axis [2]. The 84 (72) subcarriers thus grouped is called as a resource block (RB). Two RBs are the minimum unit that can be allocated in a frame and this implies that a minimum of 168 subcarriers can be allocated in the case of LTE. The minimum PHY layer rate that can be supported on the downlink is 16.8 kb/s assuming that resources are granted in every subframe. For transmitting certain control messages, resource element groups consisting of four contiguous subcarriers in an OFDM symbol can also be used. In LTE, the uplink uses SCFDMA, which can be viewed as discrete Fourier transform (DFT)-spread OFDMA. The data is passed through a DFT block before being input to the OFDMA module. The concept of RBs remains the same as in the downlink. The main advantage of using SCFDMA is the potential reduction in the peakto-average-power ratio (PAPR) of the transmitted signal. In WiMAX, to counter the PAPR problem, the uplink allocations for a mobile are given from left to right in a frame, as shown in Fig. 1a, so the number of subcarriers used per OFDM symbol is reduced.

USE OF FREQUENCY DIVERSITY Both WiMAX and LTE are likely to experience frequency-selective fading due to the wide bandwidths used [1, 2]. This means that the FD in the channel can be leveraged by suitable mapping of subcarriers to subchannels, and by coding and interleaving. It has been shown that by leveraging the FD, the performance of OFDMA can be much better than that of single-carrier FDMA systems that are used in the LTE uplink due to power efficiency reasons [2]. In WiMAX, in the mandatory subchannelization method called PUSC, subchannels are formed by grouping 24 subcarriers that are present in different parts of the spectrum. This pseudorandom selection of the positions of the

IEEE Communications Magazine • September 2012

Frequency diversity in WiMax 5 MHz system, cellId 0, segment 0, subchannels 1 and 5

Subchannel 1 Subchannel 5

Guard carriers 50

100

Guard carriers 150

200

250 300 350 Subcarrier numbers

400

450

500

Figure 2. Frequency diversity in action. subcarriers is dependent on the cell identifier (called downlink perm base [DLPB]) and is clearly specified in the standard. A quick look shows that the subcarrier positions are distributed over the entire band, as seen in Fig. 2. All the basic control messages are also sent using this diversity-based subchannelization method so that it can benefit from the frequency diversity gain. Average signal-to-noise-plus-interference ratio (SINR) over the entire band is used to choose the appropriate modulation and coding scheme (MCS). In LTE, an RB contains the same 12 contiguous subcarriers for 7 OFDM symbols. However, to leverage FD, instead of using the same RB in the second part of the subframe, another RB can be used in the second slot of the subframe, as illustrated in Fig 3. If multiple RBs are allocated, the physical position of the allocated RBs need to be contiguous to obtain frequency diversity. These methods, which are used in leveraging the frequency diversity in LTE, result in the formation of a virtual RB. The difference from WiMAX is that within an OFDM symbol, there is no FD. However, since the coded and interleaved data is distributed over a subframe, the FD can be leveraged in LTE as well. As mentioned earlier, certain control messages in LTE use small bunches of contiguously placed subcarriers spread out over the entire bandwidth to leverage the FD. Based on the implementation trends, one can comment that WiMAX has placed more emphasis on frequency diversity than has LTE. Since the feedback rate in WiMAX is slower than that in LTE (due to the smaller frame durations in LTE), one can expect that channel information will be leveraged strongly in LTE at the cost of detailed feedback. WiMAX will tend to use frequency diversity in more situations and will not critically depend on the feedback. In high-speed environments, the feedback is not expected to be accurate, and consequently, usage of frequency diversity might be more suitable. However, retransmission schemes like hybrid automatic repeat request (HARQ) are planned to be used to improve performance in such scenarios. Both LTE and WiMAX have a strong emphasis on

155

FERNANDO LAYOUT_Layout 1 8/23/12 12:55 PM Page 156

account. For instance, MUD is already leveraged in 3.5G systems like EVDO and HSPA by using only the time-varying nature of the channel. However, in OFDMA systems, there is a potential for using both the time and frequency variation of the channel, leading to higher gains [4]. This feature of OFDMA is considered an important advantage over the techniques used in the 3G evolutions like HSPA and EVDO. In WiMAX, the BAMC method uses groups of contiguous subcarriers spread out over a few OFDM symbols to achieve MUD. The subcarriers are organized into groups of nine contiguous subcarriers, which are called bins, as shown in Fig 5. Each bin has eight data and one pilot subcarriers, and four such contiguous bins are grouped together in a band. The user feeds back the best four bands as seen by the user and also updates this information using certain messages defined in the standard. Based on this, the base station chooses two bins in one of these bands and allocates the same bin over three consecutive OFDM symbols, resulting in 48 data subcarriers for a BAMC slot. The above BAMC subchannelization is the most popular method required for WiMAX certification [1, 5]. In WiMAX, the channel feedback is used to determine the best bands for a user. The feedback is typically the received signal strength indicator (RSSI) and the SINR. The statistics of these quantities are reported regularly using medium access control (MAC) messages. In addition, there can be messages that can report changes in conditions as well as respond to requests from the BTS [1, 5]. For BAMC, a single feedback value is used for a contiguous set of 36 subcarriers. In LTE, RBs are formed from a contiguous

1 subframe = 1 msec 1 slot = 0.5 msec

12 subcarriers User 1 1 resource block 1 resource block User 2

7 OFDM symbols

Figure 3. Frequency distributed data mapping in LTE downlink. the use of HARQ techniques to counter the wireless channel conditions [1, 2].

USE OF MULTIUSER DIVERSITY In a typical scenario, each user can experience different frequency selectivity depending on their location as shown Fig. 4. Since some users will experience better channel conditions in certain bands there is a potential for using MUD by smart scheduling [4]. However, for leveraging the MUD, one needs accurate feedback of the channel response, and the scheduler has to take the time-frequency characteristics of the channel into

Frequency response of channel to user 1

Frequency response of channel to user 2

Power

Power

BTS Frequency Frequency Good bands for user 2

Good bands for user 1

Figure 4. Frequency distributed data mapping in LTE downlink.

156

IEEE Communications Magazine • September 2012

FERNANDO LAYOUT_Layout 1 8/23/12 12:55 PM Page 157

Reference subchannel/ neighbor subchannel

Number of subcarriers contributing interference to reference subchannel

To operate under

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

1

5

0

1

1

1

3

0

3

1

2

3

1

1

0

2

2

2

2

0

2

2

2

5

1

1

0

2

2

1

2

0

3

1

2

4

0

1

1

2

4

1

1

0

2

1

1

3

tight frequency reuse conditions,

4

1

1

1

3

0

1

1

1

5

2

3

0

3

2

0

5

0

0

0

1

5

1

4

2

2

2

1

0

0

2

4

interference coordination is achieved among LTE base stations by assigning suitable RBs based on information exchanged between

Table 1. Olympic summer games statistics.

neighboring base stations bunch of subcarriers over several OFDM symbols, and the same RBs can be assigned for the subframe duration, which is the scheduling interval in LTE. The RB to be used for sending data to a user is chosen by the base transceiver station (BTS), and it uses the channel feedback from the mobile to schedule an RB for the user in a frame. The channel feedback in LTE, rather than explicitly sending the downlink channel status, sends a recommendation on the transmission configuration for the BS for its scheduled downlink. The mobile recommendations are based on its estimates of the instantaneous downlink channel conditions. The feedback can be periodic, where the mobile sends channel feedback in a separate control channel at predefined regular time intervals. The maximum and minimum periods of feedback transmission are sent in a separate control channel. Typically, the minimum duration between feedback messages is 2 ms, and the maximum gap between feedback messages is 160 ms. In aperiodic feedback, the BTS requests the mobile to send the channel status report. There are different modes of channel feedback that are differentiated in their frequency resolution. Depending on the mode of feedback, the channel feedback can be sent as one value for the entire operating bandwidth or as a sequence of values for a sequence of subbands covering the entire bandwidth. The subbands are basically groups of RBs. The minimal bandwidth resolution of feedback possible in LTE is 2 RBs. Hence, both standards have provisions for leveraging MUD. In WiMAX, it is currently an option, but it is expected to be supported by all vendors since it is required for certification. In LTE, the default resource formation is based on MUD, and this was a strong factor in the selection of OFDMA for the downlink in LTE. Special sounding reference signals are sent by the users in the designated portions of the subframe to help in this process. In these standards, the FD and MUD-based subchannelization can be present in the same frame [1, 2]. In WiMAX, the concept of zones is used to separate the two (or more) types of transmission; typically, for cell edge and/or high-speed mobile users, the frequency-diversity-based method is preferred. For low-mobility users close to the BTS, BAMC can be used to improve throughput. In WiMAX,

IEEE Communications Magazine • September 2012

frequency diversity and MUD-based transmission cannot coexist in time, whereas in LTE both can be used simultaneously for different users.

USE OF INTERFERENCE DIVERSITY In WiMAX, the subchannel formation depends on the value of the downlink perm base. To enable tight frequency reuse, the subcarriers used in a reference subchannel in a reference cell are distributed in different subchannels in the reuse cells. This means that the subcarriers used in a subchannel in a certain cell are unlikely to be repeated in a particular subchannel in another cell provided the downlink perm base values are different. Hopefully, the different subchannels will be allocated to different users, leading to frequency-selective interference power [1]. Hence, the user is likely to experience interference diversity (or interference averaging), which would give better performance than having a dominant interferer without good feedback about the interference. Note that interference diversity can be leveraged only in the case of PUSC transmissions. In Table 1, we show how the subcarriers in a reference subchannel get distributed across multiple subchannels in a neighboring cell. For BAMC transmissions, interference diversity cannot be used, but MUD with channel feedback can be used to obtain significant performance benefits. In LTE, the RB definition remains the same; hence, two co-channel cells can face interference from one another as there can be a clash in the positions of the data subcarriers that can be used in neighboring cells. This implies that there is no interference diversity in LTE. However, to operate under tight frequency reuse conditions, interference coordination is achieved among LTE BSs by assigning suitable RBs based on information exchanged between neighboring BSs [2].

NETWORK ENTRY STEPS IN WIMAX AND LTE In this section, we compare the steps in the radio network entry process with respect to OFDMA. This will delineate various control signals/messages sent in the OFDMA framework. Both WiMAX and LTE use frame-based transmissions, and the first task for any mobile is to

157

FERNANDO LAYOUT_Layout 1 8/23/12 12:55 PM Page 158

Pilot subcarriers to help in channel tracking

Data subcarriers

Subcarriers in a bin DC subcarrier Frequency Bin 1 Bin 2 Bin 3Bin 4

Bin 1 Bin 2 Bin 3Bin 4

Physical band 1

Physical band N

Figure 5. AMC subchannels in 802.16e. identify the start of the frame in the presence of multipath channel, frequency offsets, Doppler shifts, and other impairments. In addition, both LTE and WiMAX systems are expected to work under different channel bandwidths. In this section, we review the various signals used to achieve synchronization in the two standards. Then the network entry related steps and messages are discussed.

SYNCHRONISATION In WiMAX, a preamble is transmitted at the start of every frame [1]. The frame duration is not fixed, but the current popular implementation value is 5 ms. The preamble signal is generated by using every third subcarrier in the allowed bandwidth. For example, the current popular bandwidth for WiMAX is 10 MHz, which implies 1024 subcarriers. One third of these (without the guard subcarriers) are used for preamble transmission. The pseudorandom sequence to be sent on the subcarriers is specified in the IEEE 802.16e standard. Once the cell identifier is selected by the operator, the sequence to be sent on the subcarriers for that particular preamble is determined. To achieve frame synchronization, a mobile uses the timedomain properties of the preamble sequence (based on the use of every third subcarrier) along with the structure of the cyclic prefix (CP). Since bandwidth of the system is variable, correct detection of bandwidth is part of the above operation [6]. Since the duration of the CP is not fixed, a search procedure might have to be performed to obtain the CP duration. Frequency synchronization can be achieved once the starting point of the frame is detected using the various properties of the preamble. The identification of the sequence sent on the preamble subcarriers helps the mobile determine the DLPB, which is important for reading further control messages in the frame. The preamble typically spans the entire bandwidth (except the guard band); hence, channel estimation can also be performed after the detection of the pseudo-random sequence [7]. Since the preamble is sent at the start of every frame, corrections to the channel estimate, frequency, and time can be done using the preamble. Typically, a WiMAX mobile locks to a frame

158

after observing several repetitions of the preamble from the strongest BS. Mobiles that have completed network entry use the preamble to resynchronize and select handoff candidates. In LTE, the frame synchronization is obtained by detecting the primary synchronization sequence (PSS), which is sent twice in a frame the duration of which is fixed at 10 ms. Irrespective of the bandwidth of operation, the sequence is sent in the center 64 subcarriers occupying 0.96 MHz bandwidth. Hence, unlike the preamble in WiMAX, the number of subcarriers used for PSS is fixed. Moreover, consecutive subcarriers in the middle part of the bandwidth are used for sending the sequence because no information about the bandwidth is needed for detecting the PSS. The Zadoff-Chu sequence loaded onto the subcarriers is specified in the standard. Identification of the PSS in the received signal gives two potential starting points in the frame as there are two PSS transmissions in the frame. The ambiguity is resolved by the secondary synchronization sequence (SSS), which is sent one OFDM symbol ahead in the same set of subcarriers as the PSS. The detection of the SSS gives information about the CP duration and the cell ID. Hence, irrespective of bandwidth, a common processing can be evolved to achieve frame synchronization, CP duration detection, and cell identifier detection [8]. These signals are also used for achieving frequency synchronization. In summary, the frame synchronization in LTE appears to be simpler considering operation in different bandwidths and different CP durations. This is an important advantage for LTE mobiles when different channel bandwidths are used in different networks.

NETWORK ENTRY In WiMAX, a mobile has to search for a valid preamble to acquire frame synchronization. The uncertainties are the radio frequency (RF), bandwidth, CP duration, number of subcarriers, the segment, and the cell identity (DLPB). Once synchronized, the mobile reads the frame control header (FCH) message, which points to the length of the DL-MAP message that contains the various allocations in the frame. Both FCH and DL-MAP messages are sent using repetition coding to ensure reliable detection. In WiMAX, the location of the FCH and the start of the downlink MAP are fixed once the segment is identified in the preamble processing. Using the DL-MAP, further messages like the downlink channel descriptor (DCD) and (periodic) uplink channel descriptor (UCD) have to be read along with the UL-MAP to initiate the network entry process. All these control messages are sent with the lowest modulation and coding rate to ensure reliable detection even at the cell edge [1]. In LTE, irrespective of bandwidth and the number of subcarriers, the first step remains the same for all mobiles: locating the primary synchronization signal (PSS) and secondary synchronization signal (SSS), and obtaining the cell identifier. This is an important difference from WiMAX. However, to read complete system information, one needs to read the physical broadcast channel (PBCH), which contains information about the bandwidth of the LTE sig-

IEEE Communications Magazine • September 2012

FERNANDO LAYOUT_Layout 1 8/23/12 12:55 PM Page 159

nal. The PBCH message does not leverage frequency diversity as it contiguously occupies the 1.08 MHz wide central portion of the transmission bandwidth. To enable reliable decoding of the system messages in the BCH, heavy repetition coding is used [2]. In WiMAX, the preamble and the initial control messages are always sent using a single antenna. In LTE systems, the mobile does not know whether the BS had used multiple antennas for the transmission of PBCH. Consequently, it has to determine the number of LTE BTS antennas. This can be done by identifying the cyclic redundancy check (CRC) mask, as each antenna configuration uses a different mask as defined in the standard. Then one needs to read physical control format indicator channel (PCFICH), which gives information about the number of OFDM symbols allocated for the resource allocation messages. Finally, the last step before initiating network entry is to read the physical downlink control channel (PDCCH) information, which conveys the allocation information to the mobile. A loose link between the various messages, and signals in WiMAX and LTE is presented in Table 2.

PERFORMANCE BOUNDS FOR SYNCHRONIZATION AND CHANNEL ESTIMATION Typically , the channel estimation in WiMAX is done using the preamble. However, taking mobility into account, channel estimation can also be done using the pilots in the data symbols. The frequency domain spacing of the subcarriers when the preamble sequence is transmitted is approximately 33 kHz. The maximum amount of delay spread that can be tolerated for channel estimation is approximately equal to the inverse of pilot spacing in the frequency domain [7]. Hence, the maximum delay spread that can be tolerated for channel estimation is about 30 ms. In the mandatory PUSC mode of transmission, the average subcarrrier spacing between pilots is 43.75 kHz. Hence, the maximum delay spread that can be estimated for WiMAX using the data symbols is approximately 22.85 ms. This value is higher than the maximum supported CP duration supported in WiMAX, which means that the pilot spacing is more than sufficient for the designed OFDM parameters. On the uplink, pilots are placed close to each other in the tiles used by the individual users. Hence, in terms of estimating the channel response in the frequency domain there are no issues in the uplink. However, if time-domain truncation based methods are to be used in the uplink [7], there can be challenges as the number of pilots is restricted to the number of tiles occupied by the user. Since the preamble PN sequence modulates every third subcarrier, the maximum frequency offset that can be estimated by correlating the quasi periodic portions of the preamble is 16.4 KHz [6]. The maximum Doppler spread that can be taken care by the pilots for channel estimation is the inverse of the time duration between OFDM symbols carrying pilots. Since the pilots

IEEE Communications Magazine • September 2012

Control messages/signals Information WiMAX

LTE

Frame and symbol timing, Frequency synchronization, cyclic prefix information, CELL_ID

Cell-specific preamble

PSS & SSS

Bandwidth

Preamble

PBCH

Resource allocation messages; location, MCS, number details

FCH

PCFICH

Physical layer resource allocation info for both downlink and uplink transmissions

DL-MAP and UL-MAP

PDCCH

Type of MCS used in various parts of the frame or in various subframes for both downlink and uplink

DCD and UCD

PDCC

Table 2. Control messages/signals in Wimax and LTE.

are transmitted in each OFDM symbol, which is approximately 100 ms in duration, WiMAX systems can tolerate a maximum Doppler spread of 10 kHz. This Doppler spread is much higher than that likely to occur in practice for most WiMAX deployments. In LTE, the channel estimation is not done using the PSS and SSS signals. Hence, channel estimation is entirely dependent on the pilots in the data OFDM symbols. Consequently, a maximum of 11.11 µs delay spread can be estimated since the pilot spacing in the frequency domain is close to 90 kHz. In LTE the pilots are not transmitted in all OFDM symbols. The symbols that carry pilots are spaced in time by 4*71.42 ms, which implies that the maximum Doppler spread that can be tolerated in channel estimation is around 3.5 kHz. Hence, considering both the pilots carrying symbols, where the pilots are staggered in the subcarrier positions, the maximum delay spread that can be estimated is around 22 ms. The pilot distribution is different in the uplink. Users who have been allocated respective RBs transmit pilots in them. However, as in WiMAX uplink, time domain truncation methods cannot be applied that easily as it would depend on the number of RBs allocated to a user. For LTE, the maximum tolerable frequency offset is ±13.45 kHz.

PHYSICAL LAYER OVERHEAD As seen above, known information is transmitted by the BTS in both standards to help with various physical layer aspects. In WiMAX, a preamble is sent in the first symbol in every frame, and pilots are sent in every symbol. Considering a 10 MHz system with just the downlink operation, we can calculate the overhead due to these signals as follows. In a 5 ms frame there are 48 OFDM symbols, and each symbol has 840 usable subcarriers. The overhead due to the use of preambles and pilots in every OFDM symbol is given by the ratio of the subcarriers taken up for preamble plus pilots divided by the total number of available subcarriers in a frame. For the above considered WiMAX system, the overhead is around 16 percent. When multiple antennas are

159

FERNANDO LAYOUT_Layout 1 8/23/12 12:55 PM Page 160

Category

WiMAX

LTE

Downlink/uplink

OFDMA/OFDMA

OFDMA/SCFDMA

Resource formation

Subchannels; predominantly frequency diversity but MUD allowed

Resource blocks; predominantly MUD but frequency diversity allowed

Interference handling

Interference diversity; use of different DLPB values in neighboring co-channel cells

Interference management by exchanging information and scheduling

TTI

5ms TTI evolving to 1ms TTI

1ms TTI

Sync/channel estimation

Preamble plus pilots

PSS, SSS plus reference signals

Overhead

High due to separate preamble symbol, denser pilots in the time axis

More efficient PSS and SS design, efficient reference signal design

Scalable bandwidth operation

Same OFDM parameters but preamble bandwidth varies

Same OFDM parameters and same sync signals for different bandwidth

MIMO aspects

Good integration of MIMO but separate MIMO preambles needed

MIMO support is better and reference signals scale smoothly with MIMO

Duplexing/peak data rate

TDD deployments only, peak downlink data rate in a given bandwidth constrained by a minimum uplink allocation

FDD and TDD deployments likely;peak rates in FDD can utilize full channel bandwidth

Future Prospects

Very few deployments, existing operators planning to switch to TD-LTE

Bright prospects for being a dominant 4G standard; different RF bands in different countries are an issue; HSPA evolutions also have to be taken into account.

Table 3. WiMAX vs. LTE – summary. used, the overhead varies depending on the type of transmission mode. For example, if spacetime coding is used, the number of useful data sent over two antennas is the same as the oneantenna case. The number of pilots is also scaled correspondingly to retain the same overhead. If spatial multiplexing is used, the pilots are used in a similar fashion, but the number of data symbols is increased, leading to better overhead. However, WiMAX allows the optional use of a MIMO mid-amble, which is an exclusive preamble to enable channel estimation in the MIMO scenario. In the WiMAX uplink, there are no preambles, but the pilot overhead is high so as to enable accurate channel estimation. Consequently, the overhead due to pilots is around 33 percent. For an LTE system, the subcarriers used for physical layer processing purposes are the ones that carry reference symbols, plus primary and secondary synchronization sequences. For a single antenna transmission, there are 50 RBs in a 10 ms frame; and 4000 subcarriers are used for sending reference symbols. Counting the number of subcarriers for primary and secondary synchronization sequences, the overhead in a singleantenna LTE transmission is about 5 percent. If two antennas are used (likely in most LTE deployments) at the BTS, the number of reference symbols are doubled, and the maximum overhead is around 10 percent as in the pure spatial multiplexing case, the corresponding number of data symbols is also doubled. In the LTE uplink, the number of reference signals per

160

RB is increased to enable accurate channel estimation. In each RB, 12 subcarriers are assigned for reference signals; thus, the overhead is around 15 percent for single-antenna transmission. As the number of antennas increases for uplink transmission, we can expect more overhead, but despite this, the spectral efficiency improvement will make it attractive. Therefore, LTE has a definite advantage in the physical layer overhead compared to WiMAX. The reasons are that WiMAX standardization preceded LTE work, and consequently, the approach in WiMAX was more conservative compared to LTE; there were more pilots, and a whole symbol was devoted to the preamble. In addition, the integration of MIMO into the WiMAX standard was not as seamless as in LTE even though WiMAX pioneered the use of MIMO-OFDM in cellular environments. The same is applicable for the use of scalable OFDMA, but LTE went one step further and made the initial synchronization step independent of the bandwidth.

EVOLUTION OF WIMAX AND LTE As mentioned earlier, theWiMAX and LTE camps have submitted candidate proposals to satisfy the official requirements of the ITU IMTAdvanced criterion for 4G wireless systems [9, 10]. The evolution of WiMAX to WiMAXv2 is based on the IEEE 802.16m standard, which has backward compatibility with 802.16e WiMAX systems as an important criterion [9]. With

IEEE Communications Magazine • September 2012

FERNANDO LAYOUT_Layout 1 8/23/12 12:55 PM Page 161

respect to OFDMA features, 802.16m does not use subchannel notation or the concept of PUSC and other similar subchannel formation methods. The resource formation mechanism is handled by forming fixed size physical resource units (PRUs) from the subcarriers, which are then used to form distributed and contiguous localized resource units (DRUs and LRUs). Flexibility is given in choosing the numbers of DRUs and LRUs within a given bandwidth. All messages (data and control) are assigned either a DRU or an LRU and can coexist in the same OFDM symbol. Thus, the whole resource formation methodology is simplified, and efficiencies are higher as both types are allowed to coexist in the same OFDM symbol. The TTI is also brought down to 1 ms duration from the 5 ms TTI in WiMAXv1. Improvements in pilot overheads are also achieved, and the ability to aggregate multiple frequency channels is also proposed along with adjustments to the frame structure to enable lower latencies. In LTE-A no major changes have been proposed with respect to the OFDMA aspects except for the aggregation of multiple carriers to obtain wider channel bandwidths [10]. Thus, the OFDMA related features are very similar in both WiMAXv2 and LTE-A, as is the case when other aspects are also looked at. Hence, there might be little to choose among them from a technical perspective. However, legacy networks and product volumes could play a major role in the choice of 4G wireless system by a certain operator. As mentioned earlier, most major operators are favoring LTE over WiMAX. The existing WiMAX deployments, like that of Sprint/Clearwire in the United States, are also planning to migrate to TDLTE. Thus, the future looks to be more promising for the LTE camp than for the WiMAX camp, even though the latter championed the use of many technologies that are likely to dominate the 4G wireless world. A comparative summary of the features between WiMAX and LTE is given in Table 3.

CONCLUSION This article has compared the use of OFDMA in WiMAX and LTE standards in detail. Both systems leverage many facets of OFDMA, including frequency diversity, MUD, and frequency and time axes granularity. Subtle differences in exploiting different advantages of OFDMA in both systems are highlighted. Note that the physical layer overhead is higher for WiMAX systems than for LTE. However, the LTE-A and WiMAXv2 systems are very similar from a physical layer angle; hence, non-technical factors like backward compatibility, regulation, and product costs might have more of a bearing on the choice of a 4G wireless system.

IEEE Communications Magazine • September 2012

ACKNOWLEDGEMENT We would like to thank Dr. Arun Pachai Kannu (now at IIT Madras) for introducing us to various nuances of LTE PHY layer processing and for reviewing the article.

The LTE-A and WiMAXv2 systems are very similar from a physical layer angle

REFERENCES

and hence

[1] J. G. Andrews, A. Ghosh, and R. Muhamed, Fundamentals of WiMAX: Understanding Broadband Wireless Networking, Prentice Hall, 2007. [2] S. Sesia, I. Toufik, and M. Baker, LTE – The UMTS Long Term Evolution: From Theory to Practice, Wiley, 2nd ed., 2011. [3] M. L. Roberts and M. A. Temple, “Evolution of the Air Interface of Cellular Communications Systems Towards 4G Realization,“ IEEE Surveys and Tutorials, no 1, vol. 8, 1st qtr., 2006, pp. 2–23. [4] M. Sternad et al., “Towards Sytems Beyond 3G Based on Adaptive OFDMA Transmission,” Proc. IEEE, vol. 95, no. 12, Dec. 2007, pp. 2432–55. [5] K. Etemad, “Overview of Mobile WiMAX Technology and Evolution,” IEEE Commun. Mag., vol. 46, no. 10, Oct. 2008, pp. 31–40. [6] M. Morelli, C.-C. Jay Kuo, and M.O. Pun, “Synchronization Techniques for OFDMA: a tutorial Review,” Proc. IEEE, vol. 95, no. 7, July 2007, pp. 1394–27. [7] H. Arslan et al., “Channel Estimation for Wireless OFDM Systems,” IEEE Surveys and Tutorials, vol. 9, no. 2, 2nd qtr. 2007, pp. 18–48. [8] Y. Tsai et al., “Cell Search in 3GPP Long Term Evolution Systems,” IEEE Vehic. Tech. Mag., June 2007, pp. 23–29. [9] S. Ahmadi, “An Overview of Next-Generation Mobile WiMAX Technology,” IEEE Commun. Mag., vol. 47, no. 6, June 2009, pp. 84–98. [10] E. Dahlman and S. Parkvall, LTE/LTE-Advanced for Mobile Broadband, Academic Press, 2011.

non-technical factors like backward compatibility, regulation, product costs etc. might have more of a bearing on the choice of the 4G wireless system.

BIOGRAPHIES S. SRIKANTH ([email protected]) is a research scientist at the AU-KBC Research Centre, Anna University, Chennai, India. He obtained his Ph.D. from the University of Victoria, Canada. He works on physical layer issues in wireless systems. He has been awarded the young scientist fellowship by the government of India. He consults and conducts training programs for various organizations around the world. His areas of interest include OFDM, MIMO, and their application to wireless standards like LTE, WiMAX, and WLANs. P. A. MURUGESA PANDIAN is a research engineer at the AUKBC Research Centre, Anna University. He obtained his M.E. from Anna University. His interests are in physical layer issues in wireless communications XAVIER FERNANDO [SM] (http://www.ee.ryerson.ca/~fernando) is a professor at Ryerson University, Toronto, Canada. He is a ComSoc Distinguished Lecturer. He finished his Ph.D. at the University of Calgary, Alberta, Canada, in 2001. He has worked for AT&T for three years. He is the author of over 100 research articles and a patent. He has delivered invited talks worldwide, including at Cambridge University and Princeton University. He is in the IEEE ComSoc Education Board Working Group on Wireless Communications. He is an author and editor of the IEEE ComSoc Wireless Communication Body of Knowledge (WEBOK, http://www.comsoc.org/cert/index.html). He was a visiting scholar at the Institute of Advanced Telecommunications, Wales, United Kingdom, in 2008. His work has won several awards and prizes including Canadian best paper award (2001); Opto-Canada best poster award (2003); Sarnoff Symposium second prize (2009); and Microwave Theory and Techniques Society prize in 2010. He is the Chair of the IEEE Toronto Section.

161