Notes on Orthogonal Frequency Division Multiplexing (OFDM)ece587.cankaya.edu.tr/uploads/files/Notes...

18
Notes OFDM – HTE March 2013 Sayfa 1 Notes on Orthogonal Frequency Division Multiplexing (OFDM) 1. Discrete Fourier Transform As a reminder, the analytic forms of Fourier and inverse Fourier transforms are , exp 2 (1.1) X f xt tf dt xt j ft dt where we have assumed that Fourier transform direction is from time axis (domain) to frequency axis (domain). In this sense xt is our time signal, while X f is the frequency signal or rather the frequency spectrum of xt . Finally , exp 2 tf j ft is the operator that enables the act of transformation from time axis to frequency axis. The inverse Fourier transform will be given by , exp 2 (1.2) xt X f ft df X f j ft df (1.2) allows the return to the time axis from the frequency axis, where the operator is now * , , exp 2 ft tf j ft , i.e. the conjugate of the operator in (1.1). Symbolically (1.1) and (1.2) can be represented as below 1 , (1.3) X f Fxt xt F X f Some comments are in order regarding Fourier transform and its inverse In general Fourier transform is applicable to transformations between any axes and domains, for instance to go from spatial axis to spatial frequency axis, again Fourier transform and its inverse can be used. In these lecture notes, our axes will continue to be time and frequency as given in (1.1) and (1.2). Fourier transform can be multidimensional. It means that the single integrals in (1.1) and (1.2) can be extended to double or triple or N number of integrals. As shown by (1.1) and (1.2), to include all behaviours along one axis, the integral limits range from to . In general these limits will be finite. The transformation operators have the property that , , 1 ft tf . Thus no scaling is applied during transformations. This way, we can expect that the energies or the powers (of the same signal) in the two domains to be exactly equivalent. We now introduce the discrete Fourier transform by associating the time and frequency functions, variables and the operators as follows , , , , , , , , (1.4) t n f k xt xn X f Xk tf nk ft kn

Transcript of Notes on Orthogonal Frequency Division Multiplexing (OFDM)ece587.cankaya.edu.tr/uploads/files/Notes...

Page 1: Notes on Orthogonal Frequency Division Multiplexing (OFDM)ece587.cankaya.edu.tr/uploads/files/Notes on OFDM_20131.pdf · Notes OFDM HTE March 2013 Sayfa 1 Notes on Orthogonal Frequency

Notes OFDM – HTE March 2013 Sayfa 1

Notes on Orthogonal Frequency Division Multiplexing(OFDM)

1. Discrete Fourier Transform

As a reminder, the analytic forms of Fourier and inverse Fourier transforms are

, exp 2 (1.1)X f x t t f dt x t j ft dt

where we have assumed that Fourier transform direction is from time axis (domain) to frequencyaxis (domain). In this sense x t is our time signal, while X f is the frequency signal or rather

the frequency spectrum of x t . Finally , exp 2t f j ft is the operator that enables theact of transformation from time axis to frequency axis. The inverse Fourier transform will begiven by

, exp 2 (1.2)x t X f f t df X f j ft df

(1.2) allows the return to the time axis from the frequency axis, where the operator is now *, , exp 2f t t f j ft , i.e. the conjugate of the operator in (1.1). Symbolically (1.1)

and (1.2) can be represented as below

1 , (1.3)X f F x t x t F X f

Some comments are in order regarding Fourier transform and its inverse

In general Fourier transform is applicable to transformations between any axes and domains,for instance to go from spatial axis to spatial frequency axis, again Fourier transform and itsinverse can be used. In these lecture notes, our axes will continue to be time and frequency asgiven in (1.1) and (1.2).

Fourier transform can be multidimensional. It means that the single integrals in (1.1) and (1.2)can be extended to double or triple or N number of integrals.

As shown by (1.1) and (1.2), to include all behaviours along one axis, the integral limits rangefrom to . In general these limits will be finite.

The transformation operators have the property that , , 1f t t f . Thus no scalingis applied during transformations. This way, we can expect that the energies or the powers (ofthe same signal) in the two domains to be exactly equivalent.

We now introduce the discrete Fourier transform by associating the time and frequencyfunctions, variables and the operators as follows

, , ,

, , , , , (1.4)

t n f k x t x n X f X k

t f n k f t k n

Page 2: Notes on Orthogonal Frequency Division Multiplexing (OFDM)ece587.cankaya.edu.tr/uploads/files/Notes on OFDM_20131.pdf · Notes OFDM HTE March 2013 Sayfa 1 Notes on Orthogonal Frequency

Notes OFDM – HTE March 2013 Sayfa 2

n and k will be integer variables, limited to the range

0 1 , 0 1 (1.5)n N k N

Therefore N may be regarded to represent the infinity of the discrete world. Bearing in mind thatintegration of the analytic world will correspond to summation in discrete domain, the discreteFourier transform (DFT) of x n into X k will be

1

0exp 2 / (1.6)

N

nX k x n j kn N

1/ N in the argument of the exponential operator in (1.6) acts as normalization factor. Theinverse discrete Fourier transform (IDFT) of X k into x n will be

1

0

1 exp 2 / (1.7)N

nx n X k j kn N

N

Then symbolically, (1.6) and (1.7) can be combined as

(1.8)DFT

IDFTx n X k

In analytic world, (1.1) and (1.2) work without any problems. When switching to discrete case,care must be taken, because there are three main points to observe there

a) Infinity of analytic world must suitably be converted to the appropriate lengths in discreteworld.

b) Grid spacing of the two domains of the discrete world must be taken into account. In analyticcase this is properly covered since dt and df correspond to the same infinitesimally smallintervals, thus creating no scaling problems.

c) The rules of sampling theorem should be properly adhered to. This is because n and k areeffectively the sampling rates in time and frequency domains. Hence the smallest timeresolution that can be achieved is 2n and correspondingly the smallest frequency resolutionthat we can attain is 2k .

Example 1.1 : Let x n be a delta (impulse) function such that

1 0

(1.9)0 1 1

nx n

n N

Find X k for this discrete time delta function.

Solution : Graphically, (1.9) will be as shown in Fig. 1.1. Using (1.6), we get

Page 3: Notes on Orthogonal Frequency Division Multiplexing (OFDM)ece587.cankaya.edu.tr/uploads/files/Notes on OFDM_20131.pdf · Notes OFDM HTE March 2013 Sayfa 1 Notes on Orthogonal Frequency

Notes OFDM – HTE March 2013 Sayfa 3

1

0

0 01

0

exp 2 /

0 exp 0 1 exp 2 / 2 exp 4 /

1 exp 1

N

nX k x n j kn N

x n x n j k N x n j k N

x n N j k

(1.10)

So the result is 1X k which agrees with the analytic result of

1 (1.11)X f F x t t

0

x ( n )

1

0 0

nn = N - 1

Fig. 1.1 The graph of discrete (time) delta function given by (1.9).

Fig. 1.2 displays the graph of 1X k .

X ( k )

k

1

k = N - 1

0

1 1

Fig 1.2 Discrete Fourier transform of discrete time delta function of (1.9).

Example 1.2 : Let x n be time shifted a delta function such that

Page 4: Notes on Orthogonal Frequency Division Multiplexing (OFDM)ece587.cankaya.edu.tr/uploads/files/Notes on OFDM_20131.pdf · Notes OFDM HTE March 2013 Sayfa 1 Notes on Orthogonal Frequency

Notes OFDM – HTE March 2013 Sayfa 4

1

(1.12)0 elsewhere

n mx n

Find X k for this discrete time delta function.

Solution 1.2 : Graphically (1.12) will be as shown in Fig. 1.3. Again using the generalformulation given in (1.6), we get

1

0

0 0 0

0 01

0

exp 2 /

0 exp 0 1 exp 2 / 2 exp 4 /

1 exp 2 1 / exp 2 / 1 exp 2 1 /

1 exp exp 2 /

N

nX k x n j kn N

x x j k N x j k N

x m j k m N x m j km N x m j k m N

x N j k j km N

(1.13)

x ( n )

1

0 0 0 00

n = m

n

n = N - 1

Fig. 1.3 The graph of discrete time shifted delta function given by (1.12).

Again the finding of (1.13) agrees with the analytic result of

1 1exp 2 (1.14)X f F x t t t j ft

Since (1.13) is complex, it can be plotted in two parts as split below

1exp 2 / (1.15)

2 /

X kX k j km N

k km N

The plot associated with (1.15) is given in Fig. 1.4.

Page 5: Notes on Orthogonal Frequency Division Multiplexing (OFDM)ece587.cankaya.edu.tr/uploads/files/Notes on OFDM_20131.pdf · Notes OFDM HTE March 2013 Sayfa 1 Notes on Orthogonal Frequency

Notes OFDM – HTE March 2013 Sayfa 5

( k )k = N - 1

0

| X ( k ) |

1

k

k = N - 1k

- 2 m ( N - 1 ) / N

Fig. 1.4 Discrete Fourier transform of time shifted delta function (from (1.15)).

Example 1.3 : Now take a single exponential at a specific frequency k m , thus we may write

So we try to find the discrete Fourier transform of x n as given in (1.16).

Solution 1.3 : Again by using (1.6), we get

1 1

0 0

1

0

exp 2 / exp 2 / exp 2 /

exp 2 / (1.17)

N N

n n

N

n

X k x n j kn N j mn N j kn N

j n m k N

By using the identity

1

0

if , integerexp 2 / (1.18)

0 if , integer , integer

N

n

N m k pN pj n m k N

m k m k

X k given by (1.17) becomes

if

(1.19)0 otherwiseN k m

X k

(1.19) means a delta function at k m . The corresponding relation of the analytic world is

1 1exp 2 (1.20)X f F x t j f t f f

Fig. 1.5 displays the graph of (1.19).

exp 2 / (1.16)x n j mn N

Page 6: Notes on Orthogonal Frequency Division Multiplexing (OFDM)ece587.cankaya.edu.tr/uploads/files/Notes on OFDM_20131.pdf · Notes OFDM HTE March 2013 Sayfa 1 Notes on Orthogonal Frequency

Notes OFDM – HTE March 2013 Sayfa 6

0 0 0 00

X ( k )

k

N

k = N - 1k = m

Fig. 1.5 Discrete Fourier transform of a single exponential as given by (1.16).

Exercise 1.1 : Prove the identity in (1.18). Note that if andm k are limited to a single N range,so that 0 1m N and 0 1k N , then (1.18) can be converted into

1

0

if , 0 1exp 2 / (1.21)

0 if , 0 1

N

n

N m k m Nj n m k N

m k k N

Exercise 1.2 : Find the discrete Fourier Transform of the sum of two exponentials as givenbelow.

1 1 2 2exp 2 / exp 2 / (1.22)x n A j m n N A j m n N

Check that the discrete Fourier Transform that you have found for (1.22) resembles thecorresponding analytic expression.

2. Basis for Orthogonal Frequency Division Multiplexing (OFDM)

Conventionally we place a message signal on a single carrier. For instance, to have double sideband AM type of modulation, if s t is the message signal, we simply multiply by a sinusoidal atthe frequency of cf to obtain the modulated signal as

exp 2 or cos 2 or sin 2 (2.1)c c cy t s t j f t y t s t f t y t s t f t

Due to the frequency spectrum of s t , i.e., S f occupying a finite bandwidth, we have to takeinto account, the frequency response of the channel for the transmitted signal to arrive at thereceiver without distortions.

Page 7: Notes on Orthogonal Frequency Division Multiplexing (OFDM)ece587.cankaya.edu.tr/uploads/files/Notes on OFDM_20131.pdf · Notes OFDM HTE March 2013 Sayfa 1 Notes on Orthogonal Frequency

Notes OFDM – HTE March 2013 Sayfa 7

Suppose that S f has a bandwidth of 2W , then Y f F y t has the two sided spectrum of

2W each centered around cf f as shown in Fig. 2.1. Assuming that the communicationchannel that our modulated signal is going to pass through has the following frequency response

1 or any constant 1 if

(2.1)arbitrary elsewhere

cf f WC f

ff = fcf = - fc

1

Y ( f )

1

C ( f )2 W 2 W

Fig. 2.1 A communication channel that introduces no distortion.

then confined to the frequency range of cf f W at the output of the channel, we will have

1 1 1 (2.2)r t F C f Y f F Y f y t

The implication of (2.2) is that we have been able to receive the exact copy of the transmittedsignal and no distortions due to channel impairments have been experienced. Note (2.2) is as aresult of the channel response being frequency independent within the band of our transmittedsignal.

Next, let’s assume that our transmitted signal remains the same, but the channel response turnsinto the one shown in Fig. 2.2.

Page 8: Notes on Orthogonal Frequency Division Multiplexing (OFDM)ece587.cankaya.edu.tr/uploads/files/Notes on OFDM_20131.pdf · Notes OFDM HTE March 2013 Sayfa 1 Notes on Orthogonal Frequency

Notes OFDM – HTE March 2013 Sayfa 8

ff = fcf = - fc

Y ( f )

> 2 W

C ( f )> 2 W

Flattened bandwidth slices Flattened bandwidth slices

Fig. 2.2 A communication channel that will introduce distortion.

Note that in the case of Fig. 2.2, the spectrum of the transmitted signal is still accommodatedwithin the channel response, but since C f no longer satisfies (2.1), the transmitted signal willdefinitely be received definitely with some distortion. But if we imagine that we slice thechannel into finer frequency intervals, then it is possible to obtain nearly flat channel (frequencyindependent) responses like the one given in (2.1). For each slice, we can assign a differentcarrier rather than using a one single carrier centered at cf . Slicing into the source signal bandinto narrower frequency slots means that we can use extended symbol durations.

3. Construction of Orthogonal Frequency Division Multiplexing (OFDM)

OFDM is historically based on frequency shift keying (FSK). In OFDM, the number of carrierscalled subcarriers, are too many and they are arranged to be orthogonal over one symbol interval.

For the derivation of the basic relation for orthogonality, we first consider the following integralcovering two sinusoidal signals

0

cos 2 cos 2 (3.1)T

p p m mI f t f t dt

It is easy by hand derivation that the integral in (3.1) will evaluate to zero under the followingconditions

, , integers

, arbitrary (3.2)

p m

p m

p mf f p mT T

(3.2) can also be interpreted as

Page 9: Notes on Orthogonal Frequency Division Multiplexing (OFDM)ece587.cankaya.edu.tr/uploads/files/Notes on OFDM_20131.pdf · Notes OFDM HTE March 2013 Sayfa 1 Notes on Orthogonal Frequency

Notes OFDM – HTE March 2013 Sayfa 9

1 1 1 1 , , (3.3)

p m

p m m pp m

p mf fT

T T T T Tf f m p

The proof of (3.1) becoming zero with the setting of (3.2) is given below

0 0

0 0

using cos cos 0.5 cos cos

0.5 cos 2 0.5 cos 2

sin 2 sin 2 0.5 0.5

2 2

sin 2 0.5

T T

p m p m p m p m

T T

p m p m p m p m

p m p m

p m

A B A B A B

I f f t dt f f t dt

f f t f f t

f f f f

p m

sin sin 2 sin0.5

2 2

0 0 (3.4)

p m p m p m

p m p m

p mf f f f

Note that the validity of (3.4) can also be shown in Matlab as follows

syms Fip Fim tT = 0.1;p = 5;m = 2;fp = p/T;fm = m/T;I = int(cos(2*pi*fp*t + Fip)*cos(2*pi*fm*t + Fim),t,0,T)

After running this piece of Matlab code, we get the result of I = 0.

(3.4) can also be established using exponentials in a shorter way.

0

0

0

exp if

exp 2 exp exp 2 exp

exp exp 2

exp 2 exp 2 exp ex

or

p2

T

p p m m

T

p m p m

T

p m p m

p m p mp

m p m

m

pT j f f

I j f t j j f t j dt

j j f f t dt

j f f t j f f Tj j

j f f

p m

1 for integer

1

2

exp 2 1 exp 0 if (3.5)

2

p m

p m

p mp m

j f f

j p mj p m

j f f

Note that for the complex case, orthogonality integral is written by taking the complex conjugateof the second function.

Page 10: Notes on Orthogonal Frequency Division Multiplexing (OFDM)ece587.cankaya.edu.tr/uploads/files/Notes on OFDM_20131.pdf · Notes OFDM HTE March 2013 Sayfa 1 Notes on Orthogonal Frequency

Notes OFDM – HTE March 2013 Sayfa 10

The implications of (3.2) to (3.5) are that if cos 2 pf t and cos 2 mf t are to be selected asOFDM subcarriers, then the orthogonality condition stated in (3.2) or (3.3) must hold with therelative phase difference between the subcarriers being unimportant.

Example 3.1 : In Fig. 3.1, we illustrate the simple case of four OFDM subcarriers

0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1-4

-3

-2

-1

0

1

2

3

4

t - Time (seconds)

Rel

ativ

e am

plitu

de

Sample OFDM subcarriers

T = 0.08 sec

c4 ( t ) = A cos ( 2 f

4t +

4 )

c3 ( t ) = A cos ( 2 f

3t +

3 )

c2 ( t ) = A cos ( 2 f

2t +

2 )

c1 ( t ) = A cos ( 2 f

1t +

1 )

Fig. 3.1 Four sample OFDM subcarriers.

We can compile the following numeric data from Fig. 3.1

1 1 1 1 1

2 2 2 2 2

3 3 1

0.08 sec 80 msec First subcarrier : cos 2 , 12.5 Hz 1/ , / 2

Second subcarrier : cos 2 , 25 Hz 2 / ,

Third subcarrier : cos 2 ,

Tc t A f t f T

c t A f t f T

c t A f t

3 3

4 4 2 4 4

37.5 Hz 3 / , / 2

Fourth subcarrier : cos 2 , 50 Hz 4 / , 0 (3.6)

f T

c t A f t f T

As seen from (3.6) and Fig. 3.1, the OFDM subcarriers despite their differing (initial) phases, 1 2 3 4, , ,c t c t c t c t have complete (integer) number of cycles within one whole symbol

duration, T . Similarly the differences between the frequencies of the subcarriers are integermultiples of the inverse of the symbol duration, T . This way, orthogonality conditions of (3.2)and (3.3) are perfectly satisfied by 1 2 3 4, , ,c t c t c t c t .

Exercise 3.1 : Using the Matlab file orthogonality_syms.m or otherwise, confirm that thesubcarriers 1 2 3 4, , ,c t c t c t c t do indeed satisfy the orthogonality, when any of the twosubcarriers are integrated over the interval T , that is

0

0 , 1 4 , 1 4 , (3.7)T

p mc t c t dt p m p m

Page 11: Notes on Orthogonal Frequency Division Multiplexing (OFDM)ece587.cankaya.edu.tr/uploads/files/Notes on OFDM_20131.pdf · Notes OFDM HTE March 2013 Sayfa 1 Notes on Orthogonal Frequency

Notes OFDM – HTE March 2013 Sayfa 11

At the transmitter side OFDM signal will be obtained by modulating the OFDM subcarriers by aPSK or QAM signals. The simplified block of an OFDM transmitter is given in Fig. 3.2.

Fig. 3.2 Simplified block diagram of an OFDM transmitter.

As seen in Fig. 3.2 for an M ary PSK or QAM, K number of subcarriers is reserved. It could bethat K M , which will be the setting in these notes. To exemplify the operations carried out inFig. 3.2, we take the constellation diagram of 4 PSK and a typical flow of the first four PSKsymbols along time axis as illustrated in Fig. 3.3.

2( t )

1

4 PSKConstellation

Time flow of 4 PSK waveforms

s2 ( t ) s3 ( t )

s3

s2j

s4 - j

s3- 1

2Ts

s1 1

1( t )

2

s2

3Ts

4TsT

s

sm ( t )

0t - Time

s2 ( t )

s2

s1 ( t )

s1

Fig. 3.3 4 PSK constellation and related first four symbols.

Note that on the constellation diagram of Fig. 3.3, we also show the Matlab notation of numericvalues for the 4 PSK signals, 1 4s s . These are

1 2 3 41, , 1, (3.8)j j s s s s

Page 12: Notes on Orthogonal Frequency Division Multiplexing (OFDM)ece587.cankaya.edu.tr/uploads/files/Notes on OFDM_20131.pdf · Notes OFDM HTE March 2013 Sayfa 1 Notes on Orthogonal Frequency

Notes OFDM – HTE March 2013 Sayfa 12

As indicated in Fig. 3.2, modulation of subcarriers will be established by the multiplication ofthe PSK symbols of (3.8) by the subcarriers. This way, k th modulated subcarrier will beobtained from

cos 2 , 1 (3.9)k k m k k m ky t c t f t k K s s

where K is the total number of subcarriers. We show in Fig. 3.4 how the four PSK signals of Fig.3.3 are assigned to the four OFDM subcarriers of 1 4c t c t , to constitute 1 4y t y t withthe initial phases of the subcarriers given in (3.6) all being zero. As demonstrated by Fig. 3.4, theassignment operation also allows the extension of symbol duration from sT to 4 sT T . This isequivalent to dividing the signal bandwidth into finer slices so that the response of thecommunication channel will appear nearly flat, i.e. the act illustrated in Fig. 2.1.

Ts 2T

s3T

s

sm ( t )

0

0

0

0

0

4Ts = T

t

t

t

t

t

s2 ( t ) s2 ( t ) s3 ( t )

s2 s3 s1

s1 ( t )

s2

| y2 ( t ) | = | s2 | c2 ( t ) = | s2 | cos ( 2 f2t )

| y3 ( t ) | = | s3 | c3 ( t ) = | s3 | cos ( 2 f3t )

y4 ( t ) = s1 c4 ( t ) = s1 cos ( 2 f4t )

| y1 ( t ) | = | s2 | c1 ( t ) = | s2 | cos ( 2 f1t )

Fig. 3.4 The assignment of the four PSK symbols to the four OFDM subcarriers.

Bearing in mind that the OFDM subcarriers modulated by the PSK symbols are going to beadded to each other within one OFDM symbol duration, i.e., within sT MT as shown in Fig.3.2, then we will have the following mathematical expression for the OFDM signal, y t

1 1 1

cos 2 , 1 , 0 (3.10)K K K

k k m k k m kk k k

y t y t c t f t m M t T

s s

where k ms is one of the m th symbol of PSK , i.e., ms , placed onto the k subcarrier. Consideringthe numeric values of ms in (3.8), we see that y t will be complex. Then y t can berepresented as

0.52 2 1

real part of , imaginary part of

tan (3.11)

r i r i

ir i y

r

y t y t y t y t y t y t y t

y ty t y t y t t

y t

Page 13: Notes on Orthogonal Frequency Division Multiplexing (OFDM)ece587.cankaya.edu.tr/uploads/files/Notes on OFDM_20131.pdf · Notes OFDM HTE March 2013 Sayfa 1 Notes on Orthogonal Frequency

Notes OFDM – HTE March 2013 Sayfa 13

Thus, , yy t t and their combination can be plotted as shown in Fig. 3.5, where some

numeric values of the magnitude and the phase are given as well. Thus we can visualize y t as avector whose length (magnitude) and phase are changing with the progress of time.

))

t

t

| y ( t ) |

4Ts = T

0

0

y ( t )

0

yi ( t )2

2

yr ( t )

2

t

2

/ 3

0.3

0.3

yr ( t ) yr ( t )

/ 3

/ 2

/ 2

Fig. 3.5 The magnitude and phase plots of OFDM signal.

By converting the OFDM subcarriers into exponentials, (3.10) will become

1

exp 2 , 1 , 0 (3.12)K

k m kk

y t j f t m M t T

s

Written in discrete Fourier transform notation, (3.12) will be

1

1 exp 2 / (3.13)K

k m kk

y n j f n NN

s

where we have let /t n N . Comparing (3.13) to (1.7), we conclude that (3.12) is in the form ofinverse discrete Fourier transform (IDFT). This means that the transmitter of Fig. 3.2 canimplement the OFDM modulation via IDFT in a fast manner even if the number of subcarriersreaches numbers like 512. We know that the job of the receiver will be to demodulate andrecover the symbols k ms placed on the subcarriers. From the relations given in (1.6) and (1.7),we expect that the demodulation of the message symbols can be performed by the discreteFourier transform (DFT) operation, that is

1

0exp 2 / (3.14)

N

k k k mn

d y n j f n N

s

We shall prove later that the operation in (3.14) will indeed deliver the k ms symbols.

Page 14: Notes on Orthogonal Frequency Division Multiplexing (OFDM)ece587.cankaya.edu.tr/uploads/files/Notes on OFDM_20131.pdf · Notes OFDM HTE March 2013 Sayfa 1 Notes on Orthogonal Frequency

Notes OFDM – HTE March 2013 Sayfa 14

It is instructive to examine the frequency spectrums of OFDM signal. For this, we select thesimple case of 4 PSK and four OFDM subcarriers given in (3.6). As apparent from Fig. 3.4 and(3.8), the sinusoidal OFDM subcarriers are going to be modulated by rectangular shaped timelimited (PSK) symbols. Therefore it is natural to anticipate that, the frequency spectrums of thusmodulated OFDM subcarriers will be in the form of sinc functions as shown in Fig. 3.6. Here therelative phases introduced by PSK symbols are ignored. As seen from Fig. 3.6, the orthogonalitycondition embedded into these subcarriers ensures that at the spectral peak of one modulatedsubcarrier, the spectrums of all other subcarriers pass through zero. This point is important fromthe point of view of successful demodulation and detection of the transmitted PSK symbols.

f

Zero crossings of three other spectrums

Peaks of individual spectrums

f = f2f = f1 f = f3 f = f4

Y1 ( f ) Y2

( f ) Y3 ( f ) Y4

( f )

Fig. 3.6 Spectrums of modulated OFDM subcarriers.

We now move onto OFDM receiver. Fig. 3.7 gives the simplified block diagram of such areceiver.

Fig. 3.7 Simplified block diagram of an OFDM receiver.

As understood from Fig. 3.7, the communication channel that exists between the transmitter ofFig. 3.2 and receiver of Fig. 3.7 is assumed to be band unlimited. The way, at the input of the,

Page 15: Notes on Orthogonal Frequency Division Multiplexing (OFDM)ece587.cankaya.edu.tr/uploads/files/Notes on OFDM_20131.pdf · Notes OFDM HTE March 2013 Sayfa 1 Notes on Orthogonal Frequency

Notes OFDM – HTE March 2013 Sayfa 15

receiver, there is an exact copy of the transmitted OFDM signal. On each arm of the receiver ofFig. 3.7, multiplication of the received signal by one of the subcarriers corresponds to the act ofcorrelation. Thus the output of k th arm of the correlator is

0

cos 2 (3.15)T

k kd y t f t dt It is easy to prove that the mathematical operation of (3.15) will deliver the individual PSKsymbols placed on the subcarriers sequentially at the transmitter. Below we do this for thesample case of 4 PSK of Fig. 3.4.

Initially we write the complete expression of the transmitted signal with the PSK symbolsassigned to the subcarriers as shown in Fig. 3.4.

4 4 4

1 1 1

2 1 2 2 3 3 1 4

1 2 3 4

cos 2

cos 2 cos 2 cos 2 cos 2

cos 2 cos 2 cos 2 cos 2 (3.16)

k k m k k m kk k n

y t y t c t f t

f t f t f t f t

j f t j f t f t f t

s s

s s s s

Now based on (3.15), we evaluate the correlator output for the first arm of the receiver in Fig.3.7.

1 10

1 1 2 10 0

3 1 4 10 0

1 2

3 4

cos 2

cos 2 cos 2 cos 2 cos 2

cos 2 cos 2 cos 2 cos 2

T

T T

T T

I I

I I

d y t f t dt

j f t f t dt j f t f t dt

f t f t dt f t f t dt

(3.16)

Due to the selection of frequencies 1 2 3 4, , ,f f f f , the orthogonality of (3.7) is satisfied, thus

0

0

cos 2 cos 2 0 , if and

cos 2 cos 2 0.5 , if (3.17)

T

p m p m

T

p m

p mf t f t dt p m f fT

f t f t dt T p m

then the individual integrals in (3.16) will be

Page 16: Notes on Orthogonal Frequency Division Multiplexing (OFDM)ece587.cankaya.edu.tr/uploads/files/Notes on OFDM_20131.pdf · Notes OFDM HTE March 2013 Sayfa 1 Notes on Orthogonal Frequency

Notes OFDM – HTE March 2013 Sayfa 16

11 1 1 0

0 0

2 2 1 3 3 10 0

4 4 10

0, due to orthogonality

sin 4cos 2 cos 2 0.5 0.5

8

cos 2 cos 2 0 , cos 2 cos 2 0

cos 2 cos 2 0

TT

T

T T

T

f tI j f t f t dt jt jT

f

I j f t f t dt I f t f t dt

I f t f t dt

(3.18)

So, eventually the output of the first arm of the correlator will become 1 20.5 0.5d jT T swhich means that we have been able to demodulate successfully the PSK symbol (the scaledversion of) 2s placed on the subcarrier 1c t .

It is easy to demonstrate that, the output of the remaining arms, i.e., 2 3 4, ,d d d , in the 4 PSKequivalent of Fig. 3.7 will deliver the other symbols placed onto subcarriers 2 3 4, ,c t c t c t .

Exercise 3.2 : Bearing in mind the assignment of 4 PSK symbols in Fig. 3.4, prove that thecorrelation operation of Fig. 3.7 and (3.15) will correctly demodulate the symbols on thesubcarriers of 2 3 4, ,c t c t c t .

4. Orthogonal Frequency Division Multiplexing (OFDM) Analysis in DiscreteDomain

Now we describe modulation and demodulation processes of OFDM in terms of discrete Fourierand inverse Fourier transforms. For this, we again take the simplified case of 4 PSK and write forthe discrete equivalent of subcarriers 1 2 3 4, , ,c t c t c t c t

1 1 2 2

3 3 4 4

exp 2 / 4 , exp 2 / 4

exp 2 / 4 , exp 2 / 4 (4.1)

c n j f n c n j f n

c n j f n c n j f n

In (4.1) we have intentionally changed the time range into 4N , since the subcarriers arelimited to that time range, which can also be denoted by 4T . Keep in mind that in discreteapplications, the frequencies can only take integer values, hence the appropriate choice is

1 2 3 41, 2, 3, 4 (4.2)f f f f

As seen from (4.1), the subcarriers are complex, hence, we revise the orthogonality test of (3.7)as follows

*

0

0 , 1 4 , 1 4 , (4.3)T

p mc t c t dt p m p m

where * takes into account the possibility of subcarriers being complex. The discrete equivalenceof (4.3) is

Page 17: Notes on Orthogonal Frequency Division Multiplexing (OFDM)ece587.cankaya.edu.tr/uploads/files/Notes on OFDM_20131.pdf · Notes OFDM HTE March 2013 Sayfa 1 Notes on Orthogonal Frequency

Notes OFDM – HTE March 2013 Sayfa 17

1 1

*

0 0

0 , , orthogonal exp 2 / 4 exp 2 / 4

4 , , otherwise1 4 , 1 4 , 0 1 3

N N

p m p mn n

p mc n c n j f n j f n

p mp m n N

(4.4)

For instance after setting, 1, 2p m , (4.4) evaluates to

1 3

1 20 0

0 1 2 33

0

exp 2 / 4 exp 2 / 4 exp 2 / 4 exp 4 / 4

exp 2 / 4 exp 0 exp / 2 exp exp 3 / 2

1 1 0

N

n n

n n n n

n

j f n j f n j n j n

j n j j j

j j

(4.5)

Hence we conclude that with the choice of numeric settings of (4.1) and (4.2), 1c n and 2c nare indeed orthogonal.

Exercise 4.1 : Prove that the validity of the first line of (4.4) by inserting for the subcarriers from(4.1 ) and (4.2) for the case other than the one evaluated in (4.5).

Now we consider the operations on the side of the receiver. We recollect from (3.13) and (3.14),in OFDM modulation corresponds to inverse discrete Fourier transform operation, while OFDMdemodulation corresponds to discrete Fourier transform operation. Then after modulation, for thesimple 4 PSK case, we will have

4

1 1

1 1 2 2 3 3 4 4

2 1 2 2 3 3 1 4

1 2

1 1exp 2 / exp 2 /

1 exp 2 / exp 2 / exp 2 / exp 2 /

1 exp 2 / exp 2 / exp 2 / exp 2 /

1 exp 2 / exp 2

K

k m k k m kk k

m m m m

y n j f n N j f n NN N

j f n N j f n N j f n N j f n NN

j f n N j f n N j f n N j f n NN

j j f n N j j f nN

s s

s s s s

s s s s

3 4/ exp 2 / exp 2 / (4.6)N j f n N j f n N

On the receiver side, performing the correlation on the first arm of Fig. 3.7, we will get

1

1 10

1

1 1 1 2 2 10

3 3 1 4 4 1

1 2

exp 2 /

1 exp 2 / exp 2 / exp 2 / exp 2 /

exp 2 / exp 2 / exp 2 / exp 2 /

0 0 0

N

n

N

m mn

m m

m

d y n j f n N

j f n N j f n N j f n N j f n NN

j f n N j f n N j f n N j f n N

j

s s

s s

s s (4.7)

As seen from the result of (4.7), demodulation is successful, hence the PSK the symbol placed onthe first subcarrier 1c n at the transmitter, is correctly recovered as 2 or js at the receiver.

Page 18: Notes on Orthogonal Frequency Division Multiplexing (OFDM)ece587.cankaya.edu.tr/uploads/files/Notes on OFDM_20131.pdf · Notes OFDM HTE March 2013 Sayfa 1 Notes on Orthogonal Frequency

Notes OFDM – HTE March 2013 Sayfa 18

Exercise 4.2 : Prove that the symbols placed onto other subcarriers, i.e. 2 3 4, ,c n c n c n arealso demodulated successfully, as it is done for the first symbol in (4.7).

Exercise 4.3 : Available on the course webpage, the Matlab file OFDM_exp_simple.mimplements the modulation and demodulation for an OFDM system whose symbols are obtainedfrom an 8 rectangular QAM constellation. To increase the number of subcarriers to 8, four higherfrequencies are selected in addition to those listed in (3.6). Run and observe the outputs ofOFDM_exp_simple.m. Find the additional subcarriers, observe the orthogonality of allsubcarriers both from the time waveforms and the Ctmat matrix written onto the workspace.Comment on the spectrums of modulated subcarriers. Make several runs to see if demodulationis successful. Repeat this by changing the targeted symbol of demodulation on line 44 of thecode. By modifying the symbol duration T on line 4 of the code, observe that the orthogonalityis destroyed and the transmitted symbol is not demodulated correctly.

References

1. John G. Proakis, Masoud Salehi, “Communication Systems Engineering” 2nd Ed., PrenticeHall 2002, ISBN : 0-13-061793-8, Chapter 9.

2. S. K. Mitra, “Digital Signal Processing”, 3rd Ed. McGraw Hill 2006, ISBN : 0-13-084788-7,pp 233 - 244.

3. Matlab help - DFT4. My own lecture notes.