3 Link Adaptation for IEEE 802.11a Systems 27
3.6 Computer Simulations
In this section, we evaluate the effectiveness of the proposed LA algorithm using the NS-2 network simulator from LBNL [27]. We present modifications to this simulation environment to support 802.11a. We have modified NS-2 PHY and MAC layer parameters from 802.11b to 802.11a standard specification. These parameters include MAC and PHY header formats, data rates, and SIFS, DIFS, and EIFS time spacing as listed in Table 2.2. For the calculation of BER, the modulation curves shown in Figure 3.5 were implemented in the simulation model. The signal reception model looks up BER for a given SNR and uses this probability to determine whether the received frame with or without errors.
After constructing 802.11a model in NS-2, to verify the proposed algorithm, we simplify the simulation environment as a 2-node topology. The testing schemes under consideration are: six single-mode schemes using the PHY mode 1 (SM-1), mode 5 (SM-5) and mode 8 (SM-8) for both the basic and RTS/CTS access methods, respectively, and the proposed link adaptation scheme (LA). For each of the testing schemes, an experiment is repeated 10 times to estimate the average goodput for each
SNR value. At each experiment, the receiver requests a data delivery service of 10,000 MSDUs from the transmitter and the length of each MSDU is 2,000 octets using constant bit rate traffic generator.
Figure 3.12 shows the average goodputs obtained with SM-1, SM-5, SM-8, and LA. We can see that the simulation result confirms the theoretical analysis. The proposed algorithm switches to the higher PHY mode with increasing SNR and choose the proper access method to achieve the best goodput among other single mode schemes.
3.7 Summary
In this chapter, the goodput performance is derived analytically for peer-to-peer communication with numbers of contending stations in the IEEE 802.11a DCF system.
The effective goodput is expressed as a closed-form function of the data payload length, wireless channel condition, number of contending stations in the same network, and selected transmission rate for both the basic and RTS/CTS access methods.
Assuming the availability of the wireless channel conditions and number of contending stations, we propose an integrated LA algorithm, which selects the best combination of the PHY mode and MAC mechanism based on the theoretical analysis. We compare the performance of the proposed algorithm against three single-mode schemes by network simulator (NS-2). Simulation results show that the proposed algorithm works as expected.
0, 0 0, 1 0, 2 0,W
0-2 0,W
0-1
1 1 1
... ... ... ... ... ... ...
i, 0 i, 1 i, 2 i,W
i-2 i,W
i-1
1 1 1 1 1
... ... ... ... ... ... ...
i-1, 0
m, 0 m, 1 m, 2 m,W
m-2 m,W
m-1
1 1 1 ... 1 1
1/W
0p/W
1p/W
ip/W
m1
-p 1-p
-p 1 1
1
1
Figure 3.1: Markov chain model for the backoff window size
0 0.2 0.4 0.6 0.8 1 0
0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
Collision probability
Probability of frame transmission
Increasing number of contending stations
Figure 3.2: Solving for collision probability of IEEE 802.11a DCF system as the number of contending stations increases
0 10 20 30 40 50 60 70
0.1 0.2 0.3 0.4 0.5 0.6
Number of contending stations
Collision probability
Figure 3.3: Collision probability of IEEE 802.11a DCF system versus the number of
contending stations
(a)
Figure 3.4: Frame formats of IEEE 802.11 MAC (a) Data frame (MPDU) (b) RTS frame (c) CTS frame (d) ACK frame
0 5 10 15 20 25 30 10-20
10-15 10-10 10-5 100
Average received SNR per symbol (dB)
BER
Mode 1 Mode 2 Mode 3 Mode 4 Mode 5 Mode 6 Mode 7 Mode 8
Channel Reservation Psucc,ch Data Transmission 1-Ps,xmit,data
1-Psucc,ch
Success
Fail
SLRC(1~nl -1)
1-Ps,xmit,data
SLRC (nl) Ps,xmit,data
Figure 3.5: The upper bound BER performance of eight PHY modes of IEEE 802.11a versus the average received SNR per symbol
Figure 3.6: Two stages of RTS/CTS access method
0 5 10 15 20 25
Average received SNR per symbol (dB)
Effective goodput (Mbps)
Average received SNR per symbol (dB) Ecteoodpuffe gt (iv
Figure 3.7: Effective goodputs of different PHY modes using RTS/CTS access method versus average received SNR per symbol. Assume there are five contending stations in IEEE 802.11a WLAN. (a) MSDU size: 200 octets (b) MSDU size: 2,000 octets
0 5 10 15 20 25
Average received SNR per symbol (dB)
Effecitive goodput (Mbps)
Average received SNR per symbol (dB)
PHY mode
Average received SNR per symbol (dB)
PHY mode
Average received SNR per symbol (dB)
Effective goodput (Mbps)
Figure 3.8: Adaptive PHY mode selection for improving the effective goodput using RTS/CTS access method in IEEE 802.11a WLAN with five contending stations. (a) MSDU size: 200 octets (b) MSDU size: 2,000 octets
0 5 10 15 20 25
MSDU size: 200 octets(a)
Average received SNR per symbol (dB)
Effective goodput (Mbps)
Average received SNR per symbol (dB)
PHY mode
Average received SNR per symbol (dB)
PHY mode
MSDU size: 2,000 octets(b)
Average received SNR per symbol (dB)
Effective goodput (Mbps)
Figure 3.9: Adaptive PHY mode selection for improving the effective goodput using basic access method in IEEE 802.11a WLAN with five contending stations. (a) MSDU size: 200 octets (b) MSDU size: 2,000 octets
0 10 20 30 40 50
Figure 3.10: Maximum effective goodput of different PHY modes using basic access method versus number of contending stations in IEEE 802.11a WLAN.
MSDU size: 2,000 octets.
Wireless Medium
Figure 3.11: Proposed system architecture for link adaptation
0 5 10 15 20 25 0
5 10 15 20 25 30
Average received SNR per symbol (dB)
Averaged goodput (Mbps)
Link Adaptation SM-1 with RTS/CTS SM-1 with Basic SM-5 with RTS/CTS SM-5 with Basic SM-8 with RTS/CTS SM-8 with Basic
Figure 3.12: Performance evaluation for proposed link adaptation by NS-2 with 2-node topology. 2,000-octet MSDU is generated with CBR traffic.
Table 3.1: Adaptive PHY mode and MAC mechanism selection corresponding to given channel condition with five contending stations. Deep color represents RTS/CTS access method while light one stands for basic access method. (a) MSDU size: 200 octets (b) MSDU size: 500 octets (c) MSDU size: 1,000 octets (d) MSDU size: 2,000 octets
(a)
Mode 1 Mode 3 Mode 4 Mode 5 Mode 6 Mode 7 Mode 8 SNR (dB) ~6 6~9 9~13 13~16 16~21 21~23 23~
(b)
Mode 1 Mode 3 Mode 3 Mode 4 Mode 5 Mode 6 Mode 7 Mode 8 SNR (dB) ~6 6~7 7~9 9~13 13~16 16~21 21~23 23~
(c)
Mode 1 Mode 3 Mode 4 Mode 5 Mode 5 Mode 6 Mode 7 Mode 8 SNR (dB) ~6 6~9 9~13 13~14 14~16 16~21 21~23 23~
(d)
Mode 1 Mode 3 Mode 4 Mode 5 Mode 6 Mode 7 Mode 8 SNR (dB) ~6 6~9 9~13 13~16 16~21 21~23 23~
Table 3.2: Adaptive PHY mode and MAC mechanism selection corresponding to given channel condition with 20 contending stations. Deep color represents RTS/CTS access method while light one stands for basic access method. (a) MSDU size: 200 octets (b) MSDU size: 500 octets (c) MSDU size: 1,000 octets (d) MSDU size: 2,000 octets
(a)
Mode 1 Mode 3 Mode 3 Mode 4 Mode 5 Mode 6 Mode 7 Mode 8 SNR (dB) ~6 6~7 7~9 9~13 13~16 16~21 21~23 23~
(b)
Mode 1 Mode 3 Mode 4 Mode 5 Mode 6 Mode 7 Mode 8 SNR (dB) ~6 6~9 9~13 13~16 16~21 21~23 23~
(c)
Mode 1 Mode 3 Mode 4 Mode 5 Mode 6 Mode 7 Mode 7 Mode 8 SNR (dB) ~6 6~9 9~13 13~16 16~21 21~22 22~23 23~
(d)
Mode 1 Mode 3 Mode 4 Mode 5 Mode 6 Mode 7 Mode 8 SNR (dB) ~6 6~9 9~13 13~16 16~21 21~23 23~
Chapter 4
Link Adaptation for MIMO-Enhanced 802.11a Systems
The main goal in developing next generation wireless communication systems is to increase the link throughput. Significant improvements in throughput can be achieved when multiple antennas are applied at both the transmitter and receiver side, i.e. multiple-input multiple-output (MIMO) systems, especially in a rich scattering environment. This has been shown for wireless communication links in both flat-fading [12] as well as frequency-selective channels [16]. A potential application of the MIMO systems is next generation WLANs. An advantage of WLAN systems is that they are mainly deployed in indoor environments. These environments are typically characterized by richly scattered multipaths. As explained in [12], this is a good condition for having a high MIMO capacity.
In a MIMO system, the benefit of transmitting from multiple antennas may be utilized to improve either the diversity order or information rate of the system. These two transmission strategies are commonly denoted as spatial diversity and spatial multiplexing, respectively. Therefore, MIMO coding techniques can basically be split into two groups:
Spatial diversity [13], e.g. space-time block code (STBC) [14][15], increases
►
the performance of the communication system by coding over the different transmit branches.
Spatial multiplexing [12][16][17], e.g. Bell laboratories layered space-time (BLAST) [19], achieves a higher throughput by transmitting independent data streams on the different transmit branches simultaneously and at the same carrier frequency.
►
While STBC improves the link quality of the system by exploiting the spatial and temporal diversity, BLAST significantly improves the spectral efficiency of the system while suffering from poor error performance. In this chapter, we aim to compare these two techniques using the performance metric of effective goodput defined in Chapter 3 and select the most suitable MIMO technique under the given channel condition to increase the link throughput in an OFDM-based WLAN system with MIMO coding, while focusing on the IEEE 802.11a standard.
This chapter is organized as follows: Section 4.1 introduces the system models and performance analyses of STBC and VBLAST. According to the analysis results, we have to adopt the RTS/CTS exchange to obtain the MIMO channel information in the proposed LA algorithm in Section 4.2. For the purpose of carrying adaptive selection information based on the proposed LA algorithm, we modify the IEEE 802.11a MAC/PHY layer overheads in Section 4.3. Finally, we evaluate the proposed LA algorithm in Section 4.4.
4.1 MIMO Systems
4.1.1 MIMO Channel Model
The most commonly used channel model for MIMO systems involves independent quasi-static flat Rayleigh fading at all antenna elements [15][28][29]. The basic assumptions behind this channel model are:
1. A large number of scatters are present in the wireless channel so that the signal at any receive antenna of the MIMO systems is the sum of several multipath components. In this case, the distribution of the received signal at each antenna will be complex Gaussian. The amplitude of such complex Gaussian distributed signals is Rayleigh distributed.
2. The channel delay spread is less than the symbol rate (i.e. narrowband transmission). This assumption guarantees flat fading.
3. The channel characteristics remain constant at least for the transmission period of an entire frame. This assumption accounts for quasi-static fading.
4. The antenna elements at the transmitter and receiver of the MIMO system are placed far enough (spatially) such that the effect of the channel at a particular antenna element is different from the effect at all other antenna elements. This supports the assumption of independent or spatially uncorrelated fading.
Using all these assumptions, the independent quasi-static flat Rayleigh fading MIMO channel for a system with nT transmit and nR receive antennas can be represented as
(4.1)
where hi,j is the path gain. This is the effect of the channel on signals transmitted from jth transmit antenna and received at the ith receive antenna. The path gains are
modeled as zero mean independent complex Gaussian random variables with variance 0.5 per real dimension. Therefore, given that the signal xj is transmitted from the jth transmit antenna, the signal received at the ith receive antenna is given by
(4.2)
The additive noise ni at each receive antenna i is assumed to be white and Gaussian
with spectral density N0.
However, broadband wireless systems encounter a large delay spread, and therefore have to cope with frequency-selective fading. In this chapter, we focus on OFDM-based WLAN systems with MIMO coding, i.e. MIMO-OFDM systems. A K-subcarrier MIMO-OFDM system divides the large bandwidth into K narrow
subbands, hence it decouples the frequency-selective MIMO channel into a set of K parallel MIMO channels [30]. Thus, in each subband, the model given above can be used.
4.1.2 Performance Analysis of STBC-Enhanced 802.11a PHY
A simple case of STBC proposed by Alamouti consists of two antennas at the transmitter and any number of receive antennas [14]. The case of two antennas at both the transmitter and receiver is considered in this chapter. For a given subband k, the Alamouti’s scheme for two transmit antennas is determined by the complex orthogonal design
where and are transmitted from antennas 1 and 2 respectively at a given time t. In the following OFDM symbol at time t+T, is transmitted from antenna 1 while antenna 2 transmits , where and represent the complex conjugates of and x . The mean energy per symbol is given by
The kth-subband channels between the transmit and receive antennas can be expressed as follows:
The signal arriving at the receive antenna is a noisy superposition of the faded version of the transmitted OFDM symbols. After passing through the FFT process, the following received signals can be generated:
At time t:
Left-multiplying (4.8) by the transposed and conjugate channel matrix ikH
H , i.e.
(4.12)
The variance per real dimension of the additive white Gaussian noise nki is given by
2 2 2
where is the variance per real dimension of the noise at each of the receive antennas, i.e.
σ2
2 0
2 σ = N .
Thus, due to the orthogonal structure of STBC, we obtain decoupled equations for and . The equation (4.11) corresponds to the maximum ration combining.
Figure 4.1 and 4.2 illustrates the linear combining and the resulting equivalent maximum ratio combining model. Furthermore, We can see that STBC together with a linear combiner at the receiver transform the fading MIMO channel towards an SISO channel with a lower probability of deep fades compared to the channel from a certain transmit antenna to a certain receive antenna. Consequently, STBC, the fading MIMO channel, and linear combiner can be described by an equivalent scaled AWGN channel model as depicted in Figure 4.3, which is determined by
1k
The resulting channel gain is given by
2 2
which is the squared Frobenius norm of Hk. Therefore, the effective instantaneous received SNR in subband k at the receiver is
2
From this result, we can obtain the instantaneous error performances of different PHY modes with the STBC-enhanced 802.11a system using the formulas in Section 3.3. The results are depicted in Figure 4.4 with the flat Rayleigh fading channel
matrix . Compared with Figure 3.5,
we can see that the effect of STBC provides an SNR gain in the STBC-enhanced 802.11a system. It clearly improves the error performance of each PHY mode, i.e.
making the data transmission more reliable.
0.0079513 0.69987 0.56974 0.20471
4.1.3 Performance Analysis of VBLAST-Enhanced 802.11a PHY
The main idea of the BLAST architecture is to split the information bit stream into several sub-streams of equal length (called layers) and transmit them in parallel using a set of transmit antennas (the number of transmit antennas equals the number of sub-streams) at the same time and frequency. VBLAST (vertical BLAST) is known as a simplified version of the BLAST algorithm [31]. It is capable of achieving high spectral efficiency while being relatively simple to implement [32]. The layers are vertically arranged and each of them is transmitted over one particular antenna as shown in Figure 4.5. We use the same MIMO-OFDM system as in Section 4.1.2. The received signal vector in subband k can be written as
(4.17)
superposition of all transmitted symbols scaled by the channel gain and corrupted by AWGN. Assume the receiver uses a ML detection criterion based on perfect channel knowledge. The estimated symbol vector in the subband k is
2
where the minimization is performed over all admissible vectors Sk. An error occurs when the receiver mistakes a transmitted vector for another vector from the set of possible vectors. The probability that the receiver mistakes the transmitted vector for another vector
i
Sk j
S , given knowledge of the channel realization at the receiver (also k
referred to as the pairwise error probability), is [33][34]
( ) ( )
2is the squared minimum distance of the separation of the vector constellation points at the receiver. Using the Rayleigh-Rits criterion, we can bound by squared minimum distance of the transmit scalar constellation. Therefore, we can get the upper bound of the instantaneous SER for VBLAST transmission
2
With a Gray coding, BER can be approximated by
,
1
b VBLAST s VBLAST
P P
≈ N ⋅ , (4.22)
where N is the bit numbers modulated into one symbol. Figure 4.6 shows the performance of the QPSK modulated VBLAST system compared with the derived upper bound over the flat Rayleigh fading channel
matrix . The result confirms the
performance analysis in this section. From (4.22), we can obtain the instantaneous error performance of different PHY modes with the VBLAST-enhanced 802.11a system using the formulas in Section 3.3. Figure 4.7 illustrates the obtained results with the same channel matrix as before.
0.0079513 0.69987 0.56974 0.20471
4.2 Link Adaptation Scheme
In (4.16), the instantaneous received SNR of the STBC-enhanced 802.11a system is governed by the squared Frobenius norm of the channel matrix Hk, that is to say the
squared Frobenius norm of channel matrix Hk determines the performance of the STBC-enhanced 802.11a system. On the other hand, (4.21) shows that a larger value of
guarantees a smaller error probability of the VBLAST-enhanced 802.11a system. The performance of the VBLAST-enhanced 802.11a is strongly linked to the smallest singular value of the channel matrix H
,min
λk
k. Therefore, the complete channel information is required to evaluate the performances for both the STBC and VBALST systems for selecting these two schemes. We adopt the RTS/CTS exchange to obtain the channel information and perform LA. The receiver estimates the channel
information while receiving the RTS frame, then it uses this information to computes the effective goodput defined in Chapter 3, the metric for selecting the appropriate transmission scheme, and feeds the decision back to the transmitter via the CTS frame.
For this purpose, we modified the algorithm proposed by [8]. In [8], instead of carrying the duration of the reservation, the frames carry the data rate and data frame length. We add into one more parameter, the MIMO scheme. This modification serves the dual purpose of providing a mechanism by which the receiver can communicate the chosen transmission strategy to the transmitter, while still providing neighboring nodes with enough information to calculate the duration of the requested reservation.
The detailed algorithm is as follows.
At the first, the transmitter chooses the transmission strategy as the lowest rate PHY mode, i.e. STBC with the BPSK modulation and rate-1/2 convolutional coding, and then stores the corresponding parameters into the RTS frame. This ensures the reservation period is the most robust. The neighboring nodes hearing the RTS frame calculate the duration of the requested reservation, Drts, using the information carried in the RTS frame and update its NAV. The receiver uses the RTS frame to estimate the channel information and selects the appropriate transmission strategy based on this estimation. Then, it transmits the CTS frame with the selected MIMO scheme, data rate, and data frame length back to the transmitter. The neighboring nodes hearing the CTS frame calculate the duration of the requested reservation, Dcts, using the information carried in the CTS frame and update its NAV to account for the difference between Drts and Dcts. Finally, the transmitter responds to the receipt of the CTS by transmitting the data frame with the transmission strategy chosen by the receiver.
In the instance that the transmission strategies chosen by the transmitter and
receiver are different, the reservation duration, Drts, calculated by the information
carried in the RTS frame is no longer valid. Thus, we refer to Drts as a tentative reservation. A tentative reservation serves only to inform neighboring nodes that a reservation has been requested but the duration of the final reservation may differ. The tentative reservation effectively serves as a placeholder, denying any later requests that would conflict with it, until either a new reservation is received (Dcts) or it is confirmed as the final reservation. Final reservations are confirmed by the presence of a special subheader, called the reservation subheader (RSH), in the MAC header of the data frame. RSH consists of a subset of the header fields that are already present in the IEEE 802.11 data frame, plus a check sequence that serves to protect the subheader.
The fields in RSH consist of those needed to update the NAV, and essentially preserve the same fields present in an RTS frame. Furthermore, the fields (minus the check sequence) still retain the same functionality that they have in a standard IEEE 802.11 header. The functionality of RSH is as follows: when transmitter sends the data frame with the special MAC header containing RSH. The nodes out of the CTS transmission range can use the information carried in RSH to calculate the final reservation, Drsh,
to update the NAV to account for the difference between Drts and Drsh.
Note that, for the neighboring nodes to update their NAV correctly, they must know what contribution Drts has made to their NAV. This can be done by maintaining a list of the end times of each tentative reservation, indexed according to the (transmitter, receiver) pair. Thus, when an update is required, a node can use the list to
Note that, for the neighboring nodes to update their NAV correctly, they must know what contribution Drts has made to their NAV. This can be done by maintaining a list of the end times of each tentative reservation, indexed according to the (transmitter, receiver) pair. Thus, when an update is required, a node can use the list to