EEB317 Principles of Telecommunications

Angle Modulation

Version: Last update: Date: Lecturer:

1.0 November 2010 November 2010 Dr. Ephraim Gower

University of Botswana 2010

1. Introduction Consider the sinusoid

y ( t ) = B cos ( t ) , where B is the maximum amplitude and ( t ) is the instantaneous phase information of the signal. The instantaneous frequency of y (t ) is defines as the derivative of (t ) with respect to time, that is

i ( t ) =
Meaning that

d (t ) , dt

(1.1)

( t ) = i ( t ) dt .

(1.2)

Take for example the signal y (t ) = B cos[c t + k p m(t )] where the instantaneous phase is given by

( t ) = c t + k p m ( t ) .
From this the instantaneous frequency of the signal is given by

i ( t ) =

d d c t + k p m ( t ) = c + k p m ( t ) . dt dt

It is possible to vary the phase information (t ) in accordance with the message signal m(t ) in which case the modulation scheme is called Phase-Modulation (PM). If instead of the phase (t ) we vary the instantaneous frequency i (t ) the modulation variant is called Frequency-Modulation (FM). These are collectively known as Angle-Modulation techniques because they vary the phase/angle of the carrier in one way or another.

2. Angle Modulation In FM, it is the instantaneous frequency that is modulated or varied with respect to the message signal m(t ) . That is

i ( t ) = c + k f m ( t ) ,
with c as the carrier frequency and k f as the FM modulation constant. The phase information from this modulation is given by

( t ) = c + k f m ( ) d

= c t + k f m ( ) d

Therefore the FM modulated signal is

(1.3)

On the other hand, PM modulation is a result of varying the phase information directly. Suppose the initial carrier is y (t ) = B cos(c t ) , the PM result in
PM ( t ) = B cos c t + k p m ( t ) ,

(1.4)

for the PM modulation constant k p . It is clear that these two types of modulation are inseparable. PM always results in FM and FM always results in PM and the relation is illustrated in fig. 1.1.
m (t ) m (t )

d dt

m ( ) d

PM

FM ( t ) PM ( t )

d m (t ) dt

FM

Fig. 1.1. FM can be seen as the PM of an integrated message signal and PM can be seen as the FM of a differentiated signal.

2.1. Bandwidth of Angle Modulated Signals For convenience, let

(1.5)

Using complex notation, we can represent the FM signal as FM ( t ) = Be j c t + k f a ( t ) .

CHECK:

By Euler, FM ( t ) = B cos c t + k f a ( t ) + j sin c t + k f a ( t ) , which means that FM ( t ) = B cos ct + k f a ( t ) . FM ( t ) = FM ( t ) . where (.) refers to the real part of an argument, in this case FM ( t ) as We can write FM ( t ) = Be jct e jk f a( t ) = B cos (c t ) + j sin (c t ) e jk f a( t ) . Taylors expansion formula is such that
ex = 1 + x + x2 + 2! + xN + N!

(1.6)

For our case, let x = jk f a ( t ) ,

+ jN

k fN a N ( t ) + N!

As a result,
2 k2 f a (t ) FM ( t ) = B cos (c t ) + j sin (ct ) 1 + jk a t f ( ) 2!

+ jN

k fN a N ( t ) N!

(1.7)

Using (1.6) and (1.7), or taking the real value of (1.7) results in
k f a (t ) FM ( t ) = FM ( t ) = B cos (c t ) Bk f a ( t ) sin (c t ) 2! cos (ct ) +
2 2

(1.8)

The FM signal is comprised of exponentials of the message signal a ( t ) . Consider the multiple a ( t ) a ( t ) , which in the frequency domain is implemented as A ( ) A ( ) . Fig. 1.2 illustrates the expansion of the spectrum of multiples of a (t ) .

B ( )

A ( )

W Hz
B ( ) A ( )

W Hz

W Hz

W Hz

W Hz

Fig. 1.2. the convolution of two spectra of bandwidth W Hz results in a bandwidth of 4W Hz.

Lets start by considering the convolution of A( ) with a Dirac impulse ( r ) . The result is the spectrum of A( ) at = r where the net bandwidth is still 2W -Hz (i.e. bandwidth of A( ) ). For two spectra of bandwidth 2W Hz each the convolution result is given by

A ( ) B ( ) d .

To use the known result of the convolution between an impulse and a given function A( ) , we visualize B( ) as a sum of impulses as illustrated in the first diagram of fig. 1.2. For each impulse, the spectra of A( ) is shifted to its location. This means that the impulses right at the edge of B( ) will produce an overlapping bandwidth of W Hz at both ends of B( ) as illustrated in the B ( )* A( ) spectrum. Therefore, it is clear that the convolution result bandwidth will be somewhere around 4W Hz. Going back to a 2 (t ) or A( )* A( ) , the resulting bandwidth is twice that of a (t ) . Using the same reasoning, the bandwidth for higher order terms a N (t ) will be N -times that of a (t ) and these terms go up to infinity. This means that FM and PM have an infinite bandwidth!!!!
2.2. Narrow-band FM Let k f a ( t ) << 1 , then (1.8) simplifies to

NB-FM ( t ) = B cos (c t ) k f a ( t ) sin (c t ) ,

(1.9)

because the high-order terms k fN a N ( t ) are very small. For PM, we simply substitute a (t ) with m(t ) because the message signal is not integrated. That is NB-PM ( t ) = B cos (c t ) k p m ( t ) sin (ct ) . (1.10)

The equations (1.9) and (1.10) resemble DSB with carrier and the relation between them is illustrated in fig. 1.3.
B m ( t ) sin (c t )

m(t )

DSB-SC

NB-FM ( t )

B sin (c t )

/2

B cos (c t ) B m ( t ) sin (c t )
DSB-SC

m(t )

m ( ) d

NB-FM ( t )

B sin (c t )

/2

B cos (c t )

Fig. 1.3. Both narrow-band FM and PM can be generated using DSB-SC.

From now on we will only consider FM since the result can be easily evaluated to PM following the processing relations of fig. 1.1.
2.3. Wide-band FM The instantaneous frequency of an FM signal is given by

i ( t ) = c + k f m ( t ) .
From this it is clear that the maximum frequency shift occurs when m(t ) is at its maximum value m p , that is i ( t ) max = i + k f m p . The minimum frequency occurs when m(t ) is at its minimum value m p , (1.11)

 i ( t ) min = i k f m p .

(1.12)

Therefore due to the frequency modulation effect of m(t ) , the angular frequency deviation is given by = i ( t ) max i ( t ) min = 2k f m p The frequency bandwidth is given by f = 2k f m p = . 2 2 (1.14) (1.13)

It would seem that the bandwidth of a FM signal is just f , and indeed when FM was being developed this is what the pioneers thought was the case! In wide-band FM, k f a ( t ) is no sufficiently small to ignore the higher order terms. To get a better understanding of the bandwidth of an FM signal, let us approximate the message signal m(t ) with a series of pulses sampled with a period T . A m(t )

t
Fig. 1.4. The message signal m(t ) can be approximated by a series of pulses sampled at intervals of period T .

If the bandwidth of the signal m(t ) is W Hz, then by Nyquist the sampling period is at least T= 1 1 = . f s 2W

Since we are looking at the pulses of fig. 1.4, I follows that the spectrum of each pulse is
P( ) =
T /2

T / 2

e jt dt ,

where the amplitude of each pulse is assumed to be B = 1 (which does not change the bandwidth of the signal). Simplifying we have, P ( ) = T sin (T / 2 ) . T / 2 7

The significant bandwidth of this spectrum goes to zero when sin(T / 2) = 0 and this happens when

T
2

1 = 2 2W

= 4 W

(1.15)

Therefore the significant bandwidth of the pulse is = 4 W radians. The bandwidth of P( ) extends to infinity, which agrees with the analysis of Section 2.1. However, here we are only considering the significant bandwidth of the signal. This bandwidth is called the spectral leakage of m(t ) and it increases on [i (t )]max of (1.11) such that now
i ( t ) max = c + k f m p + 4 W ,

(1.16)

and there is also some spectral leakage at the minimum angular frequency such that
i ( t ) min = c k f m p 4 W .

(1.17)

Hence a more accurate estimation of the SIGNIFICANT bandwidth of FM BFM is

= 2 ( f + 2W ) Hz

(1.18)

However, recall that we are estimating the FM bandwidth using the estimation signal of m(t ) as in fig. 1.4. A more realistic estimation for the real signal m(t ) is
BFM = 2 ( f + W ) ,

(1.19)

and this is the equation to use when estimating the bandwidth of an FM signal. That is, in FM we take the difference between the maximum and minimum frequency deviations (i.e. f ) and then add the bandwidth of the signal W then multiply the sum by a factor of two.

3. Demodulation of Angle Modulated Signals Fig. 1.5 illustrates a simple demodulation technique for detecting FM and PM signals using an envelope detector.
FM demodulation

FM ( t )

d dt

d FM ( t ) dt

Envelope Detector

B c + k f m ( t )

dcblocking

Bk f m ( t )

PM demodulation

PM ( t )

Bk p
FM demodulation

d m (t ) dt

Bk p m ( t )

Fig. 1.5. In FM demodulation, the FM signal is differentiated to expose the frequency information as the signal envelope. The envelope detector extracts the envelope signal and dcblocking is used to remove the Bc term thus remaining with the scaled message Bk f m ( t ) . PM demodulation is FM demodulation followed by integration of the FM output.

The FM signal is differentiated to give

ct + k f t m ( ) d = B c + k f m ( t ) sin

From this we see that the message signal is contained in the envelope B[c + k f m(t )] . Therefore an envelope detector extracts this envelope signal assuming c k f m p 0 to avoid the envelope going negative in which case the diode in the detector would not be able to trace negative values as t would be reverse biased. However, this is never a worry since c is always much larger than the maximum deviation k f m p . The envelope signal is given by e (t ) = B c + k f m ( t ) = Bc + Bk f m ( t ) , which contains the dc-signal (non-time varying) Bc and this is removed via dc-blocking to end up with the scaled message signal Bk f m p . Since PM and FM are inseparable, we see in fig. 1.5 that the same FM demodulator is used to extract the envelope of the PM signal. However, this time

d PM ( t ) = B cos ct + k p m ( t ) dt d = B c + k p m ( t ) sin c t + k p m ( t ) dt

so the envelope is in dm(t ) / dt unlike in FM. The get the message signal m(t ) , the FM demodulator is followed by an integrator and we end up with Bk p m(t ) . A significant advantage of Angle modulation over Amplitude modulation is that it is immune to amplitude noise. If amplitude noise is added AM (t ) , then the message signal is corrupted since it is conveyed by amplitude variations. However, in FM (t ) the message is not carried as amplitude variations but as frequency variations. If the amplitude noise has no effect on the frequency information then the FM signal can be cleaned prior to demodulation. Fig. 1.6 shows how to eliminate amplitude noise in an FM system prior to demodulation.
FM ( t )
Hard limiter

FM ( t )

Band pass filter

FM ( t )

d dt

Envelope detector

dc blocking

Bk f m ( t )

Fig. 1.6. A hard limiter converts the noisy FM signal into a square wave of varying period but with a fixed amplitude for positive values and fixed amplitude for negative values. A band pass filter extracts the noise free FM signal from the Fourier series of the non-periodic square wave. The signal is then demodulated.

Due to amplitude noise, we can model the noisy FM signal as

where B(t ) is used instead of B to indicate that it is time varying due to the channel noise. The function of the hard limiter can be implemented using a signal comparator like in fig. 1.7. The process of eliminating the amplitude noise using a hard-limiter is known as hardlimiting.

10

+Vs Noisy FM signal + Vs +Vs Hard-limited signal

Noisy FM signal Hard-limited signal

Vs
Fig. 1.7. The hard-limiter can be implemented as a signal comparator. When the noisy FM signal is positive the output is forced the +Vs and when negative it is forced to Vs . The output is a non-periodic square wave but at least the variations in B due to amplitude noise have been eliminated.

To better understand how the cascade network of the hard-limiter and the band-pass filter help to extract a noise free FM signal, consider the Fourier series of a period square wave

F ( ) =

4 1 1 cos cos 3 cos ( 5 ) + ( ) ( ) 3 5

1 cos ( N ) + N

(1.20)

where it is clear that is a constant and not time-varying like the phase information of an FM signal which is given by

( t ) = c t + k f m ( ) d .

To obtain the Fourier series of the hard-limited noisy FM signal, we simply substitute in (1.20) with the FM phase information (t ) and multiply the amplitude by Vs . Therefore,
t t 4Vs 1 FFM ( t ) = cos c t + k f m ( ) d cos 3c t + 3k f m ( ) d + 3

(1.21)

The Fourier series of the hard-limited signal contains our FM ( t ) with no amplitude noise in the form of
FM ( t ) = 4Vs

t cos c t + k f m ( ) d .

The other sinusoids of the Fourier series in (1.21) are harmonics of FM (t ) and these can be removed by band-pass filtering, hence the cascade filter in fig. 1.7 immediately after the

11

hard-limiting process. The centre frequency of the filter must be at = c and have a bandwidth of BFM so that the whole of FM (t ) passes through undisturbed while all the other harmonics are blocked. After the extraction of FM (t ) , demodulation is the same as in fig. 1.5 for both PM and FM signals.

12