Professional Documents
Culture Documents
020303
2
1 Abstract ................................................................................................. 1
2 Theory. .................................................................................................. 1
2.1 OFDM Principle ........................................................................... 1
2.2 Fourier Transform ....................................................................... 1
2.2.1 Complex Discrete Fourier Transform ..................................... 2
2.2.2 The inverse DFT..................................................................... 4
2.2.3 The Fast Fourier Transform .................................................... 5
2.3 IEEE 802.11a standard (frequency domain frame composition)
5
2.4 Sampling vs. signal spectrum....................................................... 6
3 Block descriptions................................................................................. 8
3.1.1 Transmitter ........................................................................... 10
3.1.1.1 Discrete constant ........................................................... 10
3.1.1.2 IFFT.............................................................................. 10
3.1.1.3 Unbuffer........................................................................ 11
3.1.2 Receiver................................................................................ 11
3.1.2.1 Buffer............................................................................ 11
3.1.2.2 FFT ............................................................................... 12
3.1.3 Measurement Tools .............................................................. 12
3.1.3.1 Short-Time FFT ............................................................ 12
3.1.3.2 Frequency Frame Scope ................................................ 13
3.1.3.3 Upsample ...................................................................... 13
4 The system basics................................................................................ 13
4.1 The system setup parameters .................................................... 14
4.1.1 Initial commands .................................................................. 15
4.2 The signals flow .......................................................................... 16
4.2.1 Measurement section ............................................................ 18
4.3 Transmission analyses................................................................ 18
4.3.1 The frequency domain frame composition ............................ 18
4.3.2 Received signal spectrum...................................................... 20
4.3.3 Transmission statistics .......................................................... 21
5 Bibliography: ...................................................................................... 22
2
OFDM by IFFT modulation
1 Abstract
2 Theory.
1
2
If one or more of the terms: DFT, FFT, or IFFT are not known, it is
absolutely necessary to study references and other litterature before continuing
with this manual and attached Simulink model.
2.2.1 Complex Discrete Fourier Transform
If you are already familiar with Fourier transform, you should know that
complex DFT can be viewed as a way of determining the amplitudes and phases
of sine and cosine waves forming an analyzed signal. The equation (1) called
rectangular form of forward complex DFT, explains such signal decomposition
from a mathematical point of view:
1 N −1 kn kn
X [k ] = ∑
N n =0
x[n] cos(2π ) − j sin( 2π )
N N
(1) ,
The kn/N signifies the frequency of the cosine/sine wave (for each k∈[0,N-
1], n varies between 0 and the total number of time domain samples). The
parameter k defines the number of complete cycles of sine/cosine wave that
2
3
occur over the N points of the time domain signal stored in the x[n] array. The
parameter n signifies the number of collected time domain sample.
Since (1) defines a complex Fourier transform, both: the time domain and
the frequency domain array store complex values.
The X[k] array includes both: positive and negative frequencies, where an
index between k=0 and k=N/2 defines positive frequencies, and an index
between k=N/2+1 and k=N-1 defines negative frequencies (Fig.2).
There are two main ways of applying complex DFT in electronic systems:
• the time domain signal is assumed to be totally real,
o the real part of frequency domain signal has an even symmetry and
imaginary part has an odd symmetry (fig.3);
• the time domain is assumed to be complex:
o positive and negative frequencies are independent of eachother.
The system of totally real time domain signal is for e.g., applied in ADSL
(Asynchronous Digital Subscriber Line) technology, and will not be discussed in
this manual.
The second above mentioned system is recommended by 802.11a IEEE
Standard in W-LAN (Wireless LAN) applications.
The last thing to note about the DFT is the frequency distance between each
sample in the frequency domain (the resolution). It depends on the sampling
frequency fS and FFT length N (2):
fS
∆F = (2)
N
3
4
X [ k ] = Re X [ k ] + j Im X [ k ] (4).
4
5
N −1
kn kn
x[ n ] = ∑ (Re
k =0
X [ k ] + j Im X [ k ] ) cos( 2π
N
) + j sin( 2π ) =
N
N −1
kn kn
= ∑ Re X [ k ] cos( 2π ) + j Re X [ k ] sin( 2π ) +
k =0 N N
N −1
kn kn
+ ∑ j Im X [ k ] cos( 2π ) − Im X [ k ] sin( 2π ) =
k =0 N N
N −1
kn kn
= ∑ Re X [ k ] cos( 2π ) − Im X [ k ] sin( 2π ) +
k =0 N N
N −1
k k
+ ∑ j Im X [ k ] cos( 2π n ) + Re X [ k ] sin( 2π n)
k =0 N N
In other words, each frequency domain value produces both the real
sinusoid and the imaginary sinusoid in the time domain.
Adding all those signals reconstructs the output wave.
Cosine and sine waves in (1) and (3) can be understood as real signals
generated by the physical circuits
5
6
There is nothing to say about transmitting with use of all of the 64 carriers
modulated by applying the IFFT. But, the problems are encountered when it
comes to the signal reception.
According to Nyquist sampling theorem, the signal can be properly
sampled, only if it does not contain frequencies above one-half of the sampling
rate. If such requirement would not be fulfilled, frequency domain aliasing
would occur (ch.3 of [1]).
Thus, it is necessary to filter the part of the signal spectrum that could
destroy the information after aliasing (all frequencies above 0,5Fs). Such
filtering is performed using an analog filter (anitialias filter).
Antialias filter is a low-pass circuit designed to block all frequencies above
the cutoff frequency, while passing all frequencies below. The key parameters of
such circuit are: the stopband attenuation and of course the filter roll-off.
The best circuits offer roll-off of about 0,1 of sampling frequency, and
hundreds of decibels of stopband attenuation, what is of course far from ideal.
The consequences are obvious: the frequency band between about 0,4 and
0,5 of the sampling rate is simply a wasted resource due to the filters slow roll-
off and non ideal stopband attenuation.
Taking all into consideration, the frequency frame samples defining
amplitudes of signals between 0,4 and 0,5 of the sampling rate are not to
transmit any data since it would be lost anyway. Thus, the 802.11a standard
recommends only 521 of 64 carriers to be used for data transmission.
The other case is the DC component, which according to 802.11a is not to
be used either to avoid the degradation from carrier leakage or a DC offset
caused by the analog circuits.
1
Cumulated around the DC component.
6
7
Thus, both sampling methods: with impulse train2 and zero-order hold,
generate cyclic but not identical spectrums. The frequency spectrum of unity
impulse train is also a unity amplitude impulse train, with the spikes occurring at
multiples of the sampling frequency, fs, 2fs, 3fs, 4fs, etc., (fig.4). Since the time
domain signal is a multiplication of the data and the impulse train, in the
frequency domain – due to the convolution – original spectrum is copied to the
location of each spike in the impulse train spectrum.
In the stairs signal case, the spectrum is additionally multiplied by the sinc
function (5).
2
Unity amplitude impulse train
7
8
sin( πf S )
f
H( f ) = πf
(5).
fS
Equation (5) describes high frequency amplitude reduction due to the zero-
order hold. The sampling frequency represents fs. For f=0 H(f)=1 (fig.5).
Fig.5. The comparison of the impulse train and zero-order hold sampled signal
spectrum.
3 Block descriptions
8
Fig.7. The system introducing the OFDM transmission by IFFT modulation and FFT demodulation.
9
All of above mentioned sections introduce some new blocks, listed below:
• FFT,
• IFFT,
• Buffer,
• Unbuffer,
• Discrete constant,
• Baseband S-QASK demod,
• Frequency Frame Scope,
• Short-Time FFT,
• Upsample.
3.1.1 Transmitter
Fig.6. The discrete constant block dialog box (on the left) and Real-Imag to
complex block in single subsystem(on the right).
3.1.1.2 IFFT
The IFFT block (fig.8) computes the Inverse Fast Fourier Transform of
each Y complex input channels. The block assumes that the input is an X-by-Y
frame matrix and that each of the Y frames contains X sequential time samples
from an independent signal.
10
11
3.1.1.3 Unbuffer
Unbuffer block is used in the system as parallel to serial converter (fig.9).
3.1.2 Receiver
3.1.2.1 Buffer
The buffer block is used as serial to parallel converter in this system. It fills
each buffer with 64 samples incoming form the channel and passes the frame to
farther processing. The number of samples forming the time domain frame is
determined by the 802.11a IEEE standard.
The block introduces a nonzero latency, which can be calculated by typing
the command rebuffer_delay(f,n,m) in the Matlab command window, where
f defines the input frame size – in this case 1, n is the Buffer size (fig.10)
parameter setting – in this case 64, and m is the Buffer overlap parameter setting
which in this case is 0. When executed, this command returns 64 – single, full
buffer latency (frame size).
11
12
3.1.2.2 FFT
The FFT block computes the fast Fourier transform (FFT) of each input
frame, independently at each sample time (fig.11). Each input frame is called a
channel in the Simulink environment.
FFT length
The received signal is first buffered. Each frame consists of 256 time
domain samples, thus includes four OFDM symbols. Since the input frame does
not need to be zero padded, the frame size can be inherited from input frame
12
13
width. Input frame size was determined experimentally, so the center frequency
could be clearly identified as the not modulated part of signal spectrum.
3.1.3.3 Upsample
The Upsample block resamples the discrete input at a rate L times higher
than the input sample rate by inserting L-1 zeros between consecutive input
samples.
13
14
The following description explains the signals flow and tasks that are
realized by different system stages.
14
15
15
16
Fig.15. The data source generates cyclically repeatable 1000 frames of M-ary
symbols. Each frame is Nst=52 samples wide, which corresponds to
52 simultaneous data substreams.
16
17
17
18
18
19
19
20
20
21
21
22
5 Bibliography:
22