Vous êtes sur la page 1sur 8

Simulation of Simplified IEEE 802.

11a/g OFDM

John W. Thomas
University of Texas at Dallas - EE6390 - Spring 2009

1 Introduction
Orthogonal frequency division multiplexing (OFDM) has been adopted in sev-
eral wireless LAN standards, including IEEE 802.11a, IEEE 802.11g as well
as WiMAX (IEEE 802.16e) and digital audio/video broadcast systems. In
this project, we implement a simplied version of IEEE 802.11a/g and WiMAX
OFDM modulation and demodulation techniques in the MATLAB software.
This paper is organized as follows: Section 2 will cover the parameters for
the simulation, Section 3 will cover the performance of our simulation under
Additive White Gaussian Noise (AWGN), Section 4 will cover the performance
of our simulation under a complex Gaussian multi-path channel, Section 5 will
introduce spectral efficiency improvements of an adaptive modulation scheme,
and Section 6 will conclude the paper

2 Simulation Parameters
2.1 Symbol Parameters
Below are a list of parameters for describing the allocation of symbol time used
in this simulation:

• Data Duration (Td ) : 64 samples (3.2 µsec)


• Used Data Duration (Tu ) : 48 samples (2.4 µsec)
• Left Guard Interval Duration (Tl ) : 8 samples (0.4 µsec)

• Right Guard Interval Duration (Tr ) : 8 samples (0.4 µsec)


• Cyclic Prex (µ) : 16 samples (0.8 µsec)
• Symbol Duration (Ts ) : 80 samples (4 µsec)
With these parameters, the symbol duration, Ts , is displayed in the below
figure in terms of samples. Each sample is 50 x 10−9 seconds.

1
Figure 1: Symbol Duration Format - 4 µsec

2.2 OFDM Parameters


Below are a list of parameters for describing how the total bandwidth is divided
into orthogonal subcarriers and subchannels.

• Bandwidth (B) : 20 MHz

• FFT Size (N) : 64


• Subcarrier Spacing (fs ) : 312.5 kHz

The below figure displays the 64 subchannels spread across the entire band-
width of 20 MHz.

2
Figure 2: OFDM Configuration - 64 subchannels, 20 MHz

In this simulation, the OFDM modulator and demodulator are implemented


in the MATLAB module, ”OFDM DFT.m” attached to this report. This mod-
ule includes a 64-point discrete inverse Fourier transform (IFFT) at the trans-
mitter and√a 64-point discrete Fourier transform (FFT) at the receiver. A scaling
factor of 64 was added to ensure that each function doesn’t influence symbol
energy.

3 AWGN
The performance of our OFDM simulation has a best-case scenario under Ad-
ditive White Gaussian Noise being the only source of noise in the system. The
reason for this is because the transmitted signal only experiences a small change
in amplitude which is uniform throughout the all subchannels.
A simple block diagram below displays how AWGN noise, n, is added at
the receiver after the transmitted signal, x, has exited the channel, h. In the
diagram, y displays the received signal and is simply expressed in the equation
below:

y =x+N (1)
In this simulation we express noise as a complex Guassian random variable
which is normalized to 1 with the below MATLAB equation.

noise = (randn(1) + i ∗ randn(1))/sqrt(2) (2)


With a constant symbol energy, Es , of 1 and variable signal-to-noise ratio,
SNR, the variance of noise is also variable.
Es
V ar(noise) = = 10(−0.1)SN R (3)
SN R

3
Therefore the final equation for the received signal is below:

y = x + sqrt(V ar(noise)) ∗ noise (4)


Our results below are calculated using almost 500,000 symbols to find an
average bit error rate (BER). Performance of our OFDM implementation over
an AWGN channel using QPSK is displayed below:

−1
10

−2
10

−3
10
BER

−4
10

−5
10

−6
10
5 6 7 8 9 10 11 12 13
Avg. SNR per RX (dB)

Figure 3: Constellation Diagram - QPSK

4 Multi-path Complex Gaussian Channel


In this simulation we introduce complex Gaussian three-ray multi-path channel,
h. This multipath channel is described with the below channel impulse response
with each ”c” being a complex Guassian random variable, same as described
above, but with different variances:

h[n] = c1 δ[n] + c2 δ[n − n2 ] + c3 δ[n − n3 ] (5)


In this simulation n2 = 3 samples, n3 = 10 samples, Var(c1 ) = 0.6, Var(c2 )
= 0.3, and Var(c3 ) = 0.1. So the discrete impulse response, h[n], has a length
of 11 samples, so this length is also our maximum delay spread, Tm .
Below are performance plots for QPSK, 8-PSK and 16 QAM averaged over
almost 500,000 symbols. An assumption is made that the receiver has perfect
channel knowledge and uses this to eliminate phase shifts.

4
0
10

BER

−1
10

−2
10
5 6 7 8 9 10 11 12 13
Avg. SNR per RX (dB)

Figure 4: Multi-Path Channel Performance- QPSK

0
10
BER

−1
10

−2
10
5 6 7 8 9 10 11 12 13
Avg. SNR per RX (dB)

Figure 5: Multi-Path Channel Performance - 8PSK

5
−1.2
10

−1.3
10

−1.4
10

−1.5
10
BER

−1.6
10

−1.7
10

−1.8
10

5 6 7 8 9 10 11 12 13
Avg. SNR per RX (dB)

Figure 6: Multi-Path Channel Performance - 16QAM

5 Adaptive Modulation
At higher receive SNR points it would advantageous to use a higher scale mod-
ulation scheme in order to increase the amount of bits transmitted over time.
This can be done with adaptive modulation using some form of feedback of the
receive SNR at the transmitter. In this paper we will use a metric known as
spectral efficiency to quantify this gain.
Given a target probability of error, Pb , a range of SNR points from 10 dB
to 30 dB, plus the below equation:

(−1.5)SN R
M =1+ (6)
ln(5Pb )
The below modulation table with respect to receive SNR is contructed at
the transmitter.

6
Figure 7: Constellation Diagram - QPSK

Given this table, the below spectral efficiency plot is contructed. Just for
comparison, a constant modulation scheme using QPSK, as above, has a con-
stant spectral efficiency of 2 across all SNR points.

6
Spectral Efficiency (bits/sec/Hz)

1
10 15 20 25 30 35 40
SNR (dB)

Figure 8: Spectral Efficiency

The empirical spectral efficiency includes a randomness in receive SNR and


the modulation scheme used. We do this by implementing the random three-
ray channel simulation mentioned above and the definition of empirical spectral

7
efficiency below:
R
= E[log2 M ] (7)
B
An attempt of displaying the empirical spectral efficiency is given below.
SNR points were taken from 10 dB to 40dB.

5.5

5
Empirical Spectral Efficiency

4.5

3.5

2.5

1.5

1
10 15 20 25 30 35 40
Avg. SNR per RX (dB)

Figure 9: Empirical Spectral Efficiency

6 Conclusion
One can observe from this simulation that a simplified version OFDM modula-
tion and demodulation can be constructed with MATLAB software. We have
shown that under AWGN performance increases exponentially with an increase
in receive SNR. Under a multi-path complex Gaussian one can observe that the
performance is very low. Perhaps using a higher SNR or cyclic prefix sample
length would help combat the multi-path channel shown in this report.

Vous aimerez peut-être aussi