1Capacity Estimation of IEEE 802.16e Mobile WiMAX Networks
Chakchai So-In, Raj Jain, Abdel-Karim Al Tamimi
Department of Computer Science and Engineering, Washington University in St. Louis, St. Louis, MO 63130 USA
We present a simple analytical method for capacity estimation of IEEE 802.16e Mobile WiMAX? networks. Various overheads
that impact the capacity are explained and methods to reduce these overheads are also presented. The advantage of a simple model
is that the effect of each decision and sensitivity to various parameters can be seen easily. We illustrate the model by estimating the
capacity for three sample applications – Mobile TV, VoIP, and data. The analysis process helps explain various features of Mobile
WiMAX. It is shown that proper use of overhead reducing mechanisms and proper scheduling can make an order of magnitude
difference in performance. This capacity estimation method can also be used for validation of simulation models.
Index Terms— WiMAX, IEEE 802.16e, Capacity Planning, Capacity Estimation, Application Performance, Overhead, Mobile TV,
This paper is organized as follows. In Section II, we
I. INTRODUCTION present an overview of Mobile WiMAX physical layer.
Understanding this is important for performance modeling. Mobile WiMAX? based on IEEE 802.16e standard is now
The key input to any capacity planning and estimation a reality. Equipment is becoming available from a number of
exercise is the workload. We present thee sample workloads vendors and WiMAX Forum has developed profiles for
consisting of Mobile TV, VoIP, and data applications in interoperability testing of these equipments. A number of
Section III. Our analysis is general and can be used for any service providers have started planning deployments based on
other application workload. Section IV explains the upper the Mobile WiMAX. The key concern of these providers is
layer overheads and ways to reduce those overheads. Section how many users they can support for various types of
V presents parameters of a sample WiMAX system that we applications in a given environment or what value should be
use to illustrate the capacity estimation procedure. Section VI used for a various parameters. This often requires detailed
explain overheads in physical and MAC layers. The number simulations and can be time consuming. Also, studying
of users supported for the three workloads are finally sensitivity of the results to various input values requires
presented in Section VII. It is shown that with proper multiple runs of the simulation further increasing the cost and
scheduling capacity can be improved significantly. Both complexity of the analysis. Therefore, in this paper we
error-free perfect channel and imperfect channel results are present a simple analytical method of estimating the number
presented. Finally conclusions are drawn in Section VIII. of users on a Mobile WiMAX system.
There are four goals of this paper. First, we want to present II. AN OVERVIEW OF MOBILE WIMAX PHY a simple way to compute the number of users supported for
various applications. The input parameters can be easily be One of the key development of the last decade in the field changed allowing service providers and users to see the effect of wireless broadband is the practical adoption and cost of parameter change and to study the sensitivity to various effective implementation of orthogonal frequency division parameters. Second, we explain all the factors that affect the multiple access (OFDMA). Today, almost all upcoming performance. In particular, there are several overheads. broadband access technologies including Mobile WiMAX and Unless steps are taken to avoid these, the performance results its competitors use OFDMA. For performance modeling of can be very misleading. Although, the standard specifies WiMAX, it is important to understand OFDMA and hence these overhead reduction methods, they are not often modeled. we provide a very brief explanation that helps us introduce Third, proper scheduling can make an order of magnitude the terms that are used later in our analysis. For further difference in the capacity since it can change the number of details, we refer the reader to one of several good books on bursts and the associated overhead significantly. Fourth, the WiMAX [8, 9, 10]. method can also be used to validate simulation models that Unlike WiFi and many cellular technologies which use can handle more sophisticated configurations. fixed width channels, WiMAX allows almost any available
spectrum width to be used. Allowed channel bandwidths vary
1 This work was sponsored in part by a grant from Application Working Group of WiMAX Forum.
from 1.25 MHz to 28 MHz. The channel is divided into many each tile consists of 4 subcarriers over 3 symbol times. Of the equally spaced subcarriers. For example, a 10 MHz channel is 12 subcarrier-symbol combinations in a tile, 4 are used for divided into 1024 subcarriers some of which are used for data pilot and 8 are used for data. The slot, therefore, consists of transmission while others are reserved for monitoring the 24 subcarriers over 3 symbol times. The 24 subcarriers form a quality of the channel (pilot subcarriers), for providing safety subchannel and thus at 10 MHz, 1024 subcarriers form 35 UL zone (guard subcarriers) between the channels, or for use as a subchannels. The slot formation in downlink is different and reference frequency (DC subcarrier). is shown in Fig 2b. In the downlink, a slot consists of 2 The data and pilot subcarriers are modulated using one of clusters where each cluster consists of 14 subcarriers over 2 several available MCS (modulation and coding schemes). symbol times. Thus, a slot consists of 28 subcarriers over two Quadrature Phase Shift Keying (QPSK) and Quadrature symbol times. The group of 28 subcarriers is called a Amplitude Modulation (QAM) are examples of modulation subchannel resulting in 30 DL subchannels from 1024 methods. Coding refers to the forward error correction (FEC) subcarriers at 10 MHz. bits. Thus, QAM-64 1/3 indicates an MCS with 8-bit (64 The WiMAX DL subframe, as shown in Fig. 1, starts with combinations) QAM modulated symbols and the error one symbol-column of preamble. Other than preamble, all corrections bits take up ? of the bits leaving only 1/3 for data. other transmissions use slots as discussed above. The first In traditional cellular networks, the downlink - Base field in DL subframe after the preamble is a 24-bit Frame station (BS) to Mobile Station (MS) - and uplink (MS to BS) Control Header (FCH). For high reliability, FCH is use different frequencies. This is called frequency division transmitted with the most robust MCS (QPSK ?) and is duplexing (FDD). WiMAX allows FDD but also allows time repeated 4 times. Next field is DL-MAP which specifies the division duplexing (TDD) in which the downlink (DL) and burst profile of all user bursts in the DL subframe. DL-MAP uplink (UL) share the same frequency but alternate in time. has a fixed part which is always transmitted and a variable The transmission consists of frames as shown in Fig. 1. The part which depends upon the number of bursts in DL DL subframe and UL subframe are separated by a TTG subframe. This is followed by UL-MAP which specifies the (transmit to transmit gap) and RTG (receive to transmit gap). burst profile for all bursts in the UL subframe. It also consists The frames are shown in two dimensions with frequency of a fixed part and a variable part. Both DL and UL MAPs
along the vertical axis and time along the horizontal axis. are transmitted using QPSK ? MCS.
(a) (b) Fig. 2. Symbols, Tiles, Clusters, and Slots
The key parameters of Mobile WiMAX PHY are
summarized in Table I through III. Fig. 1. A sample OFDMA TDD frame structure  Table I: OFDMA Parameters for Mobile WiMAX  Parameters Values In OFDMA, each MS is allocated only a subset of the System bandwidth 1.25 5 10 20 3.5 7 8.75 (MHz) subcarriers. The available subcarriers are grouped in to a few Sampling factor 28/25 8/7 subchannels and the MS is allocated one or more subchannels Sampling frequency (Ffor a specified number of symbols. There are a number of 1.4 5.6 11.2 22.4 4 8 10 ) s,MHzways to group subcarriers in subchannels of these Partially Sample time 25714.3 178.6 89.3 44.6 125 100 (1/Fnsec) 0 s,Used Subchannelization (PUSC) is the most common. In 51FFT size (N) 128 512 1024 2048 1024 1024 PUSC, subcarriers forming a subchannel are selected FFT2 Subcarrier spacing randomly from all available subcarriers. Thus, the subcarriers 10.93 7.81 9.76 (Δƒ, kHz) forming a subchannel may not be adjacent in frequency. Useful symbol time (T102.Users are allocated variable number of “slots” in the =1/Δƒ, 91.4 128 b4 µs) downlink and uplink. The exact definition of slots depends Guard time (T= g upon the subchannelization method and on the direction of 11.4 16 12.8 T/8, µs) btransmission (DL or UL). Fig. 2 shows slot formation for OFDMA symbol 115.time (T=T+T, 102.8 144 PUSC. In uplink (Fig. 2a), a slot consists of 6 “tiles” where sbg2 µs)
subscribers supported assume a certain workload for the Table I lists the OFDMA parameters for various channel subscriber. The main problem is that workload varies widely widths. Note that the product of subcarrier spacing and FFT with types of users, types of applications, and time of the day. size is equal to the product of channel bandwidth and One advantage of the simple analytical approach presented in sampling factor. For example, for 10 MHz channel, this paper is that the workload can be easily changed and the 10.93kHz×1024 = 10MHz×(28/25). This table shows that at effect of various parameters can be seen almost 10 MHz the OFDMA symbol time is 102.8 µs and so there instantaneously. With simulation models, every change would are 48.6 symbols in a 5 ms frame. Of these, 1.6 symbols are require several hours simulation reruns. In this section we used for TTG and RTG leaving 47 symbols. If n of these are present 3 sample workloads consisting of Mobile TV, VoIP, used for DL then 47-n are available for uplink. Since DL slots and data applications. We use these workloads to demonstrate occupy 2 symbols and UL slots occupy 3 symbols, it is best to various steps in capacity estimation. divide these 47 symbols such that 47-n is a multiple of 3 and The VoIP workload is symmetric in the sense that DL data n is of the form 2k+1. For a DL:UL ratio of 2:1, these rate is equal to the UL data rate. It consists of very small considerations would result in a DL subframe of 29 symbols packets that are generated periodically. The packet size and and UL subframe of 18 symbols. In this case, the DL the period depend upon the vocoder used. We will use G723.1 subframe will consists of a total of 14×30 or 420 slots. The in our analysis. It results in a data rate of 5.3 kbps and UL subframe will consist of 6×35 or 210 slots. generates packets every 30 ms. Table II lists the number data, pilot, and guard subcarriers The Mobile TV workload depends upon the quality and for various channel widths. A PUSC subchannelization is size of the display. A sample measurement on a small screen assumed, which is the most common subchannelization. Mobile TV device produced an average packet size of 984 bytes every 30 ms resulting in an average data rate of 350.4 Table II: Number of Subcarriers in PUSC  Parameters Values kbps. Note that Mobile TV workload is highly asymmetric (a) DL with almost all of the traffic going downlink. System bandwidth (MHz) 1.25 2.5 5. 10 20 For data workload, we selected the Hypertext Transfer FFT size 128 N/A 512 1024 2084 rd# of guard subcarriers 43 N/A 91 183 367 Protocol (HTTP) workload recommended by the 3# of used subcarriers 85 N/A 421 841 1681 Generation Partnership Project (3GPP) . # of pilot subcarriers 12 N/A 60 120 240 The characteristics of the three workloads are presented in # of data subcarriers 72 N/A 360 720 140 (b) UL Table IV. System bandwidth (MHz) 1.25 2.5 5. 10 20 FFT size 128 N/A 512 1024 2084 Table IV: Workload Characteristics # of guard subcarriers 31 N/A 103 183 367 Parameters Mobile VoIP Data # of used subcarriers 97 N/A 409 841 1681 TV Type of transport layer RTP RTP TCP Table III: MCS Configurations Average packet Size (bytes) 983.5 20.0 1200.2 MCS Bits per Coding DL Bytes UL bytes Average data rate (kbps) w/o headers 350.0 5.3 14.5 symbol Rate per slot per slot UL:DL traffic ratio 0 1 0.006 QPSK ? 2 0.125 1.5 1.5 Silence suppression (VOIP only) N/A Yes N/A QPSK ? 2 0.25 3 3 Fraction of time user is active 0.5 QPSK ? 2 0.5 6 6 ROHC packet type 1 1 TCP QPSK ? 2 0.75 9 9 Overhead with ROHC (bytes) 1 1 8 QAM-16 ? 4 0.5 12 12 Payload Header Suppression (PHS) No No No QAM-16 ? 4 0.67 16 16 MAC SDU size with header 984.5 21.0 1208.2 QAM-16 ? 4 0.75 18 16 QAM-64 ? 6 0.6 18 16 QAM-64 ? 6 0.67 24 16 QAM-64 ? 6 0.75 27 IV. UPPER LAYER OVERHEAD QAM-64 5/6 6 0.83 30 Table IV which lists the characteristics of our Mobile TV,
VoIP, and data workloads includes the type of transport layer
used: Real Time Transport (RTP) or TCP. This affects the
upper layer protocol overhead. RTP over UDP over IP Table III lists the number of bytes per slot for various MCS
(12+8+20) or TCP over IP (20+20), both can results in a per values. For each MCS, the number of bytes is equal to (#bits
packet header overhead of 40 bytes. This is significant and per symbols × Coding Rate × 48 data subcarriers and symbols
can severely reduce the capacity of any wireless system. per slot / 8 bits). Note that for UL, the maximum MCS level
There are two ways to reduce upper layer overheads and to is QAM-16 ? .
improve the number of supported users. These are Payload
Header Suppression (PHS) and Robust Header Compression III. TRAFFIC MODELS AND WORKLOAD CHARACTERISTICS (ROHC). PHS is a WiMAX feature. It allows the sender to
not send fixed portions of the headers and can reduce the 40 The key input to any capacity planning exercise is the
byte header overhead down to 3 bytes. ROHC, specified by workload. In particular, all statements about number of
the Internet Engineering Task Force (IETF), is another , which reduces the DL-MAP entry overhead to 11 bytes
including 4 bytes for Cyclic Redundancy Check (CRC) . higher layer compression scheme. It can reduce the higher
The fixed UL-MAP is 6 bytes long with an optional 4-byte layer overhead to 1 to 3 bytes. In our analysis, we use ROHC-
CRC. With a repetition code of 4 and QPSK?, both fixed RTP packet type 0 with R-0 mode. In this mode, all RTP
DL-MAP and UL-MAP take up 16 slots. sequence numbers functions are known to the decompressor. The variable part of DL-MAP consists of one entry per This results in a net higher layer overhead of just 1 byte [6, 7]. bursts and requires 60 bits per entry. Similarly, the variable For small packet size workloads, such as VoIP, header part of UL-MAP consists of one entry per bursts and requires suppression and compression can make a significant impact 52 bits per entry. These are all repeated 4 times and use only on the capacity. We have seen several published studies that QPSK ? MCS. It should be pointed out that repetition use uncompressed headers resulting in significantly reduced consists of repeating slots (and not bytes). Thus, both DL and performance which would not be the case in practice. UL MAPs entries also take up 16 slots each per burst.
PHS or ROHC can significantly improve the capacity and B. Uplink Overhead should be used in any capacity planning or estimation. The UL subframe also has fixed and variable parts (See Fig. 1). Ranging and contention are in the fixed portion. Their One option with VoIP traffic is that of silence suppression size is defined by the network administrator. These regions which if implemented can increase the VoIP capacity by the are allocated not in units of slots but in units of “transmission inverse of fraction of time the user is active (not silent). opportunities.” For example, in CDMA initial ranging, one
opportunity is 6 subchannels and 2 symbol times. V. WIMAX SYSTEM CHARACTERISTICS The other fixed portion is channel quality indication (CQI) The analysis method presented in this paper can be used and acknowledgements (ACK). These regions are also for any allowed channel width, any frame duration, or any defined by the network administrator. Obviously, more fixed subchannelization. For our examples, we assume a 10 MHz portions are allocated; less number of slots is available for the Mobile WiMAX TDD system with 5 ms frame duration, user workloads. In our analysis, we allocated three OFDM PUSC subchannelization mode, and a DL:UL ratio of 2:1. symbol columns for all fixed regions. These are the default values recommended by WiMAX forum Each UL burst begins with a UL preamble. One OFDM system evaluation methodology and are also common values symbol is used for short preamble and two for long preamble. used in practice. We allocate one slot for the UL preamble. The number of DL and UL slots for this configuration can C. MAC Overhead be computed as shown in Table V.
At MAC layer, the smallest unit is MAC protocol data unit Table V: Mobile WiMAX System Configuration (PDU). As shown in Fig. 3, each MAC PDU has at least 6-Configurations Downlink Uplink bytes of MAC header and a variable length payload consisting DL/UL Symbols excluding preamble 28 18 Ranging, CQI and ACK (column symbols) N/A 3 of a number of optional subheaders, data, and an optional 4-# of symbol columns per Cluster12/ Tile 2 3 byte CRC. The optional subheaders include fragmentation, 12# of subcarriers per Cluster/ Tile 14 4 packing, mesh and general subheaders. Each of these is 2 12Symbols × Subcarriers per Cluster/ Tile 28 12 12bytes long. Symbols × Data Subcarriers per Cluster/ Tile 24 8 12# of pilots per Cluster/ Tile 4 4 In addition to generic MAC PDUs, there are bandwidth 12# of clusters/ #Tiles per Slot 2 6 request PDUs. These are 6 bytes in length. Bandwidth Subcarriers × Symbols per Slot 56 72 requests can also be piggybacked on data PDUs as a 2-byte Data Subcarriers × Symbols per Slot 48 48 Data Subcarriers × Symbols per DL/UL Subframe 23,520 12,600 subheader. Number of Slots 420 175 12Cluster for DL and Tile for UL UL preambleMAC/BW-REQOtherDataCRC HeaderSubheaders(optional) VI. OVERHEAD ANALYSIS Fig. 3. UL burst preamble and MAC frame (MPDU) In this section, we consider WiMAX PHY and MAC
overheads. The PHY overhead can be divided into DL VII. PITFALLS
overhead and UL overhead. Each of these three overheads is Many WiMAX analyses ignore the overheads described in discussed next. Section VI, namely, UL-MAP, DL-MAP, and MAC
overheads. In this section, we show that these overheads have A. Downlink Overhead
a significant impact on the number of users supported. Since In DL subframe, overhead consist of preamble, FCH, DL-
some of these overheads depend upon the number of users, MAP and UL-MAP. The MAP entries can result in a
the scheduler needs to be aware of this additional need while significant amount of overhead since they are repeated 4
admitting and scheduling the users. We present two case times. WiMAX Forum recommends using compressed MAP
studies. The first one assumes an error-free channel while the assumes a dumb scheduler. A smarter scheduler will try to
second extends the results to a case in which different users aggregate payloads for each user and thus minimizing the
have different error rates due to channel conditions. number of bursts. We call this enhanced scheduler. It works
as follows. Given n users with any particular workload, we A. Case Study 1: Error-Free Channel divide the users in k groups of n/k users each. The first group Given the user workload characteristics and the overheads is scheduled in the first frame; the second group is scheduled discussed so far, it is straightforward to compute the system in the second frame, and so on. The cycle is repeated every k capacity for any given workload. Using the slot capacity frames. Of course, k should be selected to match the delay indicated in Table III, for various MCS, we can compute the requirements of the workload. For example, with VoIP users, number of users supported. a VoIP packet is generated every 30 ms but assuming 60 ms thOne way to compute the number of users is simply to divide is an acceptable delay, we can schedule a VoIP user every 12the channel capacity by the bytes required by the user payload WiMAX frame (recall that each WiMAX frame is 5 ms) and
and overhead . This is shown in Table VI. The table send two VoIP packets in one frame as compared to the thassumes QPSK ? MCS for all users. This can be repeated for previous scheduler which would send 1/6 of the VoIP packet
other MCS. The final results are as shown in Fig. 4. The in every frame and thereby aggravating the problem of small
number of users supported varies from 2 to 46 depending payloads. A 2-byte packing overhead has to be added in the
upon the workload and the MCS. MAC payload along with the two SDUs.
Table VII shows the capacity analysis for the three Table VI: Capacity Estimation using a Simple Scheduler workloads with QPSK ? MCS and the enhanced scheduler. Parameters Mobile VoIP Data TV The results for other MCS can be similarly computed. These MAC SDU size with header (bytes) 984.5 21.0 1208.2 results are plotted in Fig. 5. Note that the number of users Data rate (kbps) with upper layer headers 350.4 5.6 14.6 (a) DL supported has gone up 2 to 600. Compared to Fig. 4, there is Bytes/5 ms frame per user (DL) 219.0 3.5 9.1 an capacity improvement by a factor of 1 to 30 depending Number of fragmentation subheaders 1 1 1 Number of packing subheaders 0 0 0 upon the workload and MCS. DL data slots per user with MAC header + packing and fragmentation subheaders 38 2 3 Total slots per user Proper scheduling can change the capacity by an order of (Data + DL-MAP IE + UL-MAP IE) 46 18 19 magnitude. Making less frequent but bigger allocations can Number of users (DL) 8 22 21 (b) UL reduce the overhead significantly. Bytes/5ms Frame per user (UL) 0.0 3.5 0.1 # of fragmentation subheaders 0 1 1 TABLE VII: Capacity Estimation using an Enhanced Scheduler # of packing subheaders 0 0 0 Parameters Mobile VoIP Data UL data slots per user with MAC header + TV packing and fragmentation subheaders 0 2 2 MAC SDU size with header (bytes) 985.5 21.0 1208.2 Total slots per user (Data + UL preamble) 0 3 3 Data rate (kbps) with upper layer headers 350.4 2.8 14.6 Number of users (UL) ? 58 58 Deadline (ms) 10 60 25 Number of users (min of UL and DL) 8 22 21 (a) DL Number of users with silence suppression 8 44 21 Bytes/5 ms frame per user (DL) 437.9 42.0 454.9 Number of fragmentation subheaders 1 0 1 Number of packing subheaders 0 1 0 DL data slots per user with MAC header + packing and fragmentation subheaders 75 9 78 Total slots per user WiMAX Capacity (Data + DL-MAP IE + UL-MAP IE) 83 25 94 5046464646464646Number of users (DL) 8 192 200 444445(b) UL 40Mobile40Bytes/5 ms frame per user (UL) 0.0 42.0 2.9 TV 353210MHzNumber of fragmentation subheaders 0 0 1 30VoIP 25232323232323Number of packing subheaders 0 1 0 252222222210MHz211919182017UL data slots per user with MAC header + 1414Data Number of users15packing and fragmentation subheaders 0 9 2 1110MHz810Total slots per user (Data+UL preamble) 0 10 3 452Number of users (UL) ? 204 2900 0Net number of users (min of UL and DL) 8 192 200 QPSK 1/8QPSK 1/4QPSK 1/2QPSK 3/4QAM16QAM16QAM16QAM64QAM64QAM64QAM641/22/33/41/22/33/45/6Number of users with silence suppression 8 384 200 Modulation and Coding Schemes Fig. 4. Number of users supported in lossless channel (Simple scheduler)
The main problem with the analysis presented above is that Note that the per user overheads impact the downlink it assumes that every user is scheduled in every frame. Since capacity more than the uplink capacity. The downlink there is a significant per burst overhead, this type of subframe has DL-MAP and UL-MAP entries for all DL and allocation will result in too much overhead and too little UL bursts, and these entries can take up a significant part of capacity. Also, since every packet (SDU) is fragmented, a 2-the capacity and so minimizing the number of bursts byte fragmentation subheader is added to each MAC PDU. increases the capacity. What we discussed above is a common pitfall. The analysis
TABLE IX: PERCENT MCS FOR 1X1 AND 2X2 ANTENNAS  1 Antenna 2 Antenna Average MCS WiMAX Capacity %DL %UL %DL %UL 1000528528504504504480456432FADE 4.75 1.92 3.03 1.21 384450550600400450250350550216200QPSK ? 7.06 3.54 4.06 1.68 Mobile TV 120100QPSK ? 16.34 12.46 14.64 8.65 10MHz1005034QPSK ? 15.30 20.01 13.15 14.05 32282424VoIP 2216QPSK ? 12.14 21.23 10.28 15.3 10MHz128Number of users10QAM16 ? 20.99 34.33 16.12 29.97 Data 4QAM16 ? 0.00 0.00 0.00 0.00 10MHz2QAM16 ? 9.31 5.91 14.18 22.86 1QAM64 ? 0.00 0.00 0.00 0.00 QPSK 1/8QPSK 1/4QPSK 1/2QPSK 3/4QAM16QAM16QAM16QAM64QAM64QAM64QAM641/22/33/41/22/33/45/6QAM64 ? 14.11 0.59 24.53 6.27 Modulation and Coding SchemesTotal 100.00 100.00 100.00 100.00 Fig. 5. Number of users supported in lossless channel (Enhanced Scheduler)
Average bytes for in each direction can be calculated by There is a limit to aggregation of payloads and
summing the product (percentage users with an MCS × minimization of bursts. First, the delay requirements for the
number of bytes per slot for that MCS). For 1 antenna payload should be met, and so a burst may have to be
systems this gives 10.19 bytes for the downlink and 8.86 bytes scheduled even if the payload size is small. In these cases,
for the uplink. For 2 antenna systems, we get 12.59 bytes for multi-user bursts in which the payload for multiple users is
the downlink and 11.73 bytes for the uplink. aggregated in one DL burst can help reduce the number of
Table X shows the number of users supported for both bursts. This is allowed by the IEEE 802.16e standards and
simple and enhanced scheduler. The results show that the applies only to the downlink bursts.
enhanced scheduler still increases the number of users by an The second consideration is that the payload cannot be
order of magnitude, especially for VoIP and data users. aggregated beyond the frame size. For example, with QPSK
?, a Mobile TV application will generate enough load to fill TABLE X: NUMBER OF SUPPORTED USERS IN LOSSY CHANNEL Workload 1 Antenna 2 Antenna the entire DL subframe every 10 ms or every 2 frames. This is Simple Enhanced Simple Enhanced much smaller than the required delay of 30 ms between the Scheduler Scheduler Scheduler Scheduler Mobile TV 13 14 14 18 frames. VoIP 44 456 46 480 Data 22 300 22 350 B. Case Study 2: Imperfect Channel
In section A, we saw that the aggregation had more impact
on performance with higher MCSs (which allow higher VIII. CONCLUSIONS
capacity and hence more aggregation). However, it is not In this paper, we explained how to compute the capacity of always possible to use these higher MCSs. The MCS is a Mobile WiMAX system and account for various overheads. limited by the quality of the channel. In this section, we We illustrated the methodology using three sample workloads present a capacity analysis assuming a mix of channels with consisting of Mobile TV, VoIP, and data users. varying quality resulting in different levels of MCS for Analysis such as the one presented in this paper can be different users. easily programmed in a simple program or a spread sheet and
effect of various parameters can be analyzed instantaneously. Table VIII: Simulation Parameters  This can be used to study the sensitivity to various parameters Parameter Value so that parameters that have significant impact can be Channel Model ITU Veh-B (6 taps) 120 km/hr Channel Bandwidth 10 MHz analyzed in detail by simulation. This analysis can also be Frequency Band 2.35 GHz used to validate simulations. Forward Error Correction Convolution Turbo Coding We showed that proper accounting of overheads is Bit Error Rate threshold 10-5 important in capacity estimation. A number of methods are MS Receiver noise figure 6.5 dB available to reduce these overheads and these should be used BS Antenna Transmit Power 35 dBm BS Receiver noise figure 4.5 dB in all deployments. In particular, robust header compression Path loss PL(distance) = 37×log(distance) + or payload header suppression, compressed MAPs are 1020×log(frequency) + 43.58 10examples of methods for reducing the overhead. Shadowing Log normal with σ =10 Proper scheduling of user payloads can change the capacity # of sectors per cell 3 by an order of magnitude. The users should be scheduled so Frequency reuse 1/3
that their number of bursts is minimized while still meeting Table VIII lists the channel parameters used in a their delay constraint. This reduces the overhead significantly simulation by Leiba et al . They showed that under these particularly for small packet traffic such as VoIP.
We showed that our analysis can be used for loss-free conditions, the number of users in a cell which were able to
channel as well as for noisy channels with loss. achieve any particular MCS was as listed in Table IX. Two
cases are listed: single antenna systems and 2 antenna
 IEEE P802.16Rev2/D2, “DRAFT Standard for Local and metropolitan area networks, Part 16: Air Interface for Broadband Wireless Access Systems,” 2094 pp, December 2007.  WiMAX FORUM, “WiMAX System Evaluation Methodology V2.0,” 230 pp, December 2007.  So-in, C., Jain, R., and Al-Tamimi, A., “Scheduling in IEEE 802.16e Mobile WiMAX Networks: Key Issues and a Survey,” Submitted to IEEE Journal on Selected Areas in Communications (JSAC), January 2008.  IEEE C802.16d-03/78, “Coverage/ Capacity simulations for OFDMA PHY in with ITU-T channel model,” 24 pp, November 2003.  3GPP2-TSGC5, HTTP and FTP Traffic Model for 1xEV-DV Simulations, 3GPP2-C50-EVAL-2001022-0xx, 2001.  Jonsson, L-E., Pelletier, G., and Sandlund, K., “Framework and four profiles: RTP, UDP, ESP, and uncompressed,” RFC 3095, July 2001.  Pelletier, G., Sandlund, K., Jonsson, L-E., and West, M., “RObust Header Compression (ROHC): A Profile for TCP/IP (ROHC-TCP),” RFC 4996, January 2006.  Eklund, C., Marks, R-B., Ponnuswamy, S., Stanwood, K-L., and Waes, N-V., “WirelessMAN Inside the IEEE 802.16 Standard for Wireless Metropolitan Networks,” 400 pp, May 2006.  Jeffrey G. Andrews, J., Arunabha Ghosh, A., Muhamed, R., “Fundamentals of WiMAX Understanding Broadband Wireless Networking,” 496 pp, March 2007.  Nuaymi, L., “WiMAX: Technology for Broadband Wireless Access,” 310 pp, March 2007.  Yaghoobi, H., “Scalable OFDMA Physical Layer in IEEE 802.16 WirelessMAN,” Intel Technology Journal, vol. 8, no. 3, August 2004.