Notes On AM
Notes On AM
Amplitude modulation
The most salient feature of information signals is that they are generally low frequency.
Sometimes this is due to the nature of data itself such as human voice which has
frequency components from 300 Hz to app. 20 KHz. Other times, such as data from a
digital circuit inside a computer, the low rates are due to hardware limitations.
Due to their low frequency content, the information signals have a spectrum such
as that in the figure below. There are a lot of low frequency components and the one-
sided spectrum is located near the zero frequency.
0 1 2 3 4 Frequency
Bandwidth
In the time domain this 4 component signal may looks as shown in Figure 2.
Now let’s modulate this signal, which means we are going to transfer it to a
higher (usually much higher) frequency. Just as information signals are characterized by
their low frequency, the transmission medium, or carriers are characterized by their high
frequency.
The simplest type of modulator for nearly all modulation schemes is called the
Product Modulator consisting of a multiplier or a mixer and a band-pass filter. Let’s
modulate the above signal using the Product Modulator, where m(t) is the low frequency
message signal and c(t) is the high frequency carrier signal. The modulator takes these
two signals and multiplies them.
f (t ) m(t ) c(t )
c(t)
In Figure 4a, we see the two sided spectrum of the message signal. After mixing,
modulating or heterodyning (all of these terms refer to the same thing), we get a spectrum
such as in Figure 4b. The spectrum is now shifted up to the carrier frequency and we see
that it is replicated on both sides of the y-axis.
-fm fm Frequency
Passband Bandwidth
Another way to describe the process is that multiplication by a sinusoid, shifts one
copy of the spectrum to fc and an another to -fc. Why does this happen? The reason is
explained by the Fourier Transform of this signal which is a product of two signals, one
of them a sinusoid.
f (t ) m(t ) cos wc t
The Fourier transform of f(t) is just the Fourier Transform of the signal m(t), half
of it shifted up and half of it down.
1
F( f ) [ M( f fc ) M ( f f c )]
2
This half is
down shifted.
Half is up
shifted.
In Figure 4b the two-sided spectrum of the signal is shifted up to the plus and
minus carrier frequency. The negative frequency twin on the other side of y-axis is
usually no problem and can be easily filtered out by a real passband filter. And now we
just work with half of the spectrum, usually the positive half recognizing that it has one-
half the magnitude of the actual signal.
In time domain, we see that this signal has much higher frequency. But its
envelope is still the original low frequency signal of Figure 2.
Passband
Spectrum
Baseband
Spectrum Lower Sideband Upper Sideband
Baseband Signal - The information signal is called the baseband signal. The
bandwidth is always a positive quantity so the bandwidth of this signal is fm.
Passband Signal - The multiplication of this signal with a sinusoid carrier signal
translates the whole thing up to fc. This signal is now called the passband signal. This
signal extends in range from (-fc - fm ) to (fc + fm.). The new signal has doubled in
bandwidth. The passband signal bandwidth is double that of the baseband signal.
The fact that the same signal has double the bandwidth in passband is often
confusing. We think of bandwidth as something physical so how can it just double? The
answer is imbedded in the question itself. In keeping with our concept of bandwidth as
something real, we do not allow it to cross from the positive to the negative domain. It
exists as a separate quantity on each side of the y-axis and does not cross it. There is no
free lunch even in signal processing, so another simplistic way of considering this fact is
that the passband signal contains not just the message signal but the carrier as well, so
wouldn’t you expect it to have a larger bandwidth?
Sidebands
Now note that in Figure 6, the passband spectrum has two parts (on each side of fc) that
are identical.
The upper part of the passband spectrum above the carrier is called the upper
sideband and the one below is called the lower sideband.
We notice that since the passband spectrum is symmetrical (not only about the y-
axis but also about the carrier frequency) the upper sideband is the mirror image of the
lower sideband. Do we need the whole spectrum to recover the baseband signal? Perhaps
we can get by with only half.
Passband
Spectrum
Figure 7 - Filter passband to for the upper and lower side-band as separate
signals.
So can we just transmit only half of the signal? Can we figure out some way of
transmitting an another signal in the rejected half? Then we can transmit two signals for
the price of one!
This realization leads to the single and double side-band modulation techniques.
In double side-band, we use the whole spectrum just as we show above. Both halves are
used. In single sideband modulation, we filter out the lower or the upper band to separate
out these signals as if they were two independent signals. Each half is enough to recover
the signal.
Filter 1 and Filter 2 in Figure 7 do just that and show how we could transmit two
signals in the place of one. Use F1 before transmitting, and you get only the lower side
band, and use F2 and you get only the upper side band. We get two channels in place of
one. Where ever bandwidth limitations exist, SSB is used. Most notable application is in
telephony. Telephony signals have ideal characteristics for the use of SSB. There is very
little signal content below 300 Hz so the SSB signal does not suffer much distortion. Also
telephone signals are bandwidth-limited, and SSB maximizes bandwidth usage. HAM
radio and HF communications is one area where the Single Side Band (SSB) modulation
is used to this advantage.
Amplitude Modulation
Let’s take the information signal m(t). The output of the mixer gives us
Now instead of transmitting just the signal times the carrier, we add the carrier to
the to the product. The block diagram of this, called the AM product modulator, would
look like this.
s(t)=m(t) Ac cos(wt)
m(t) s(t)=m(t) Ac cos(wt) + Ac cos(wt)
c(t)=Ac cos(wt)
Figure 8 - A basic AM modulator, its output contains the modulated signal and the
carrier
1
C( f ) [ Ac M ( f f c ) Ac ( f f c ) Ac M ( f f c ) Ac ( f f c )
2 ]
Spectrum on the Spectrum on the
positive negative
x-axis x-axis
(We are using properties of the Fourier Transform here; the first term comes from the fact that
the FT of a signal, multiplied by a cosine is just the same spectrum shifted, and the second term is
just a delta function times the amplitude of the original carrier. Fourier Analysis is the absolute
fundamental of all signal processing and I suggest reading tutorials 6 and 7 so you are clear on
the main concepts. You are welcome to email me your questions.)
The
Carrier
Baseband
Spectrum
Now you see the carrier signal pop up in the middle of the spectrum. We can put a
filter around this signal and recover the carrier at the receiver. This is then fed to the
demodulation circuitry later.
This modulation is called Double Side Band (DSB) modulation. It is the most
basic form of the AM modulation. From here on, we can do a variety of things such as
suppress the carrier, use one band or the other etc.. All of these are variations of the
Double Side Band (DSB) Amplitude Modulation.
By varying the amplitude of the carrier vs. the amplitude of the information
signal, we can create different looking waveforms. As long as certain parameters are not
exceeded, the envelope of this signal would look like the information signal and using an
Envelope Detector (demodulation) we can recover this signal.
In above equation, quantity Ac represents the power of the modulated signal. Both
the carrier and the message signal are assumed to have normalized amplitude. The
quantity km( t ) is called the modulation index of the signal. The index effects how the
received signal looks. Modulation index larger that 100% distorts the signal so an
envelope detector can not be used to demodulate it any longer.
The carrier
Ac cos wct
Modulation
Index, k
Ac (1+ k m(t))cos wct
m(t)
Message signal
x x +
k m(t) Ac cos wct
The following two figures show the effect of the modulation index on the
received AM signal.
As long as the modulation index is less than 100%, the envelope of the signal can
be used to remove the information signal. For index greater than 100% as shown in figure
above, the envelope detector will no longer be able to correctly detect the signal. We see
that the envelope in the lower figure is no longer a copy of the original signal in Figure 2.
We just added the carrier, but now we realize that it actually takes a lot of power
to include the carrier and perhaps it makes no sense to do that after all. But we want to
somehow include the carrier information but without actually doing so. And we want to
use the envelope detector as the receiver. How can we do that?
s(t)=m(t) Ac cos(wt)
m(t) s(t)=m(t) Ac cos(wt) + Ac cos(wt)
This balanced modulator is basically two product modulators added together. The
input to one is a negative information signal and a negative carrier. The product of this
modulator when added to its positive counterpart results in canceling the carrier as we
can see in the output. (The math above is quite straightforward and worth checking for
that wonderful feeling that comes when you really understand something.)
Envelope detector can not be used with DSB-SC carrier because the envelope of
the DSB-SC signal is not the same as the baseband signal. A more sophisticated
modulator is needed with this signal.
The DSB-SC modulation is identical to BPSK, which we will discuss later.
How do we create a SSB signal? There are two main ways that SSB signals can
be generated.
The simplest solution would be to just take the DSB-SC signal and filter the
unwanted band before transmission so that the unwanted side is not sent at all as shown
in the figure below. By keeping only the part shown, we have gotten rid of all the other
images, all of the negative components and the upper side-band.
Figure 12 - A passband filter after DSB-SC modulation results in getting rid all but
one band.
Problem with this method is that it is hard to build practical filters with steep
enough cut-offs at high frequencies. Such a filter ends up distorting the desired signal as
well as including some of the unwanted side-band anyway.
The second method involves the use of Hilbert Transform and the Analytic signal
we talked about in the last Tutorial. As a way of review, the figure below shows the
baseband spectrum of our signal. The second part shows the Hilbert Transform of the
same signal. (Recall that the Hilbert Transform rotates the positive frequency
components.)
-fm fm Frequency
fm
-fm Frequency
m( t ) cos( c t )
Now let’s take the Hilbert transform of this signal and modulate it by a sine wave, so we
get
^
m( t ) sin( c t )
^
c( t ) m( t ) cos( c t ) m( t )sin( c t )
The signal Its Hilbert
Transform
m(t) Ac cos(t)
X
Ac cos(wt)
Ac sin(wt)
Hilbert
Trans-
former X m(t) Ac sin(t)
Figure 14 - SSB Modulator using the Phasing method
The SSB signal created in this way is essentially two signals in quadrature. The
combination gives us the equation for the SSB signal. By changing the sign of the
analytic signal, we can create either the upper sideband or the lower.
^
c( t ) m( t ) cos( c t ) m( t )sin( c t )
Now let’s take the Fourier Transform of each part. The Fourier Transform of the first part
is
1
2 ( M ( c ) M ( c ))
1
2j ( M ( c )( j ) M ( c )( j )
Figure below shows the two spectrums and we see at once that adding these two
representations give us a nice clean signal with only one side band, upper or lower as we
desire.
Thanks to Dr. Hilbert and his analytic signal there is nothing to filter, just a clean
single band.
Another interesting fact is that the sum of the two side bands give us the DSB-SC
waveform.
Frequency
Figure 15 - a. Spectrum of part one, b. spectrum of part two, d. the sum of these two
gives us the lower side band, the difference would give the upper side band.
A variation of DSB is used for broadcast TV. Under the FCC requirements, the standard
video signal occupies a bandwidth of 4.5 MHz. The sound signal is separate and is
transmitted at the upper edge of this signal. When carrier is shifted to bandpass, this one
sided bandwidth becomes 9 MHz. This is nearly ten times as large as the total bandwidth
occupied by all the channels of the AM radio. Use of SSB modulation would cut this in
half but SSB is not used for video signals because of the complexity of the SSB receivers.
TV manufacturers particularly American companies were instrumental in setting these
standards like to keep the cost of the TV’s as low as possible so SSB receivers are not
used.
A modulation technique used for commercial video broadcasting which lies some
where in the middle of SSB and DSB is called theVestigial sideband Modulation (VSB).
In figure below we show a hypothetical bandpass video signal. The sound signal
which is sent separately is at the upper edge of the spectrum.
Carrier Sound signal
4.5 MHz
a.Video signal DSB Spectrum
In Figure b, we show a peculiar kind of filtering of this video signal that takes place after
modulation with a carrier but before transmission.
Sound signal
4.5 MHz
This filter takes in a small part of the upper edge of the lower sideband, starting
from -1.25 MHz. The signal is attenuated in this range from -1.25 MHz to -.75 MHz.
From here on to 4 MHz, the signal is transmitted full strength. At 4 MHz it is once again
attenuated down to 4.5 MHz so as not to interfere with the sound carrier which is
demodulated separately. The shaded portion is what is transmitted.
The term vestigial is used since a tiny trace part of the lower sideband is also
included in the transmission. The net result is that instead of transmitting a 9 MHz signal,
we transmit only 6 MHz, the standard video signal today.
Unlike voice signals which have no components near the zero frequency, Video
signals are very sensitive to their low frequency content. Distortion in these components
degrades the picture. So extra care has to be taken to make sure that all the low frequency
components (which are located in the center) are transmitted without distortion. VSB
modulation transmits these low frequencies at the twice level. The motivation for filtering
the signal in this way also comes from the desire to use a diode demodulator which
requires an explicit carrier. But to recover the carrier we need to go a little to the other
side of the carrier frequency and take in an attenuated part of the signal because of the
limitations of practical filters. The development of this filter was a function of a
compromise between bandwidth and the TV receiver complexity.
Product Demodulator
All AM signals discussed here, DSB, DSB-SC, SSB and VSB can be demodulated using
a product demodulator. In principle it is the reverse of the modulation process. We take
the incoming signal, which now also includes noise and we multiply it by a known
carrier. The product obtained is then low pass filtered and what remains then is the
information signal.
The main problem with the product demodulator is that the carrier phase is not
known. We do not know if the starting phase was 30 or 45 or 90 or some other
number. For some signals this is not such a big problem. An audio signal can be
demodulated incoherently which means that the phase of the carrier at the receiving end
is not synchronized with the transmitter. In radio AM broadcasting we can get away with
ignoring the phase because our ears are not very sensitive to phase deviations of the
signals. We can hear and understand the signal just fine. In such cases, an incoherent
product demodulation makes sense and would be the cheapest solution.
Now if we are sending data, this is indeed a big problem and we need to exactly
recover the phase of the transmitted carrier. Even video signals are not forgiving of phase
errors. Phase information for nearly all signals except, telephone and radio signals is
considered very important.
There are two methods of making sure that we know the phase of the incoming
signal; 1. The Costas loop and 2. The phase locked loop. Both are variations of a
technique to find and lock on to the phase (we will discuss these in another tutorial in
detail.) This variation of the product demodulation where we make a special effort to
determine the phase of the transmitted carrier is called coherent demodulation.
Square-Law Demodulator
y kx 2
Now let’s take an amplitude modulated signal
Putting this through the above non-lienearity, after some manipulations and clever
trigonometric substitutions, we get
k k 2
y( t ) g 2 ( t ) kAc [kAc m( t ) Ac m2 ( t )] cos 2 c t[kAc Ac m2 ( t )]
2
2 2
Now throw away the DC term, filter out the terms at two times the frequency and what
we have left is
k
y( t ) [kAc m( t ) Ac m2 ( t )]
2
2
The term m ( t ) is not a big problem if the modulation index is small. This term
disappears and for audio broadcasting this term makes no discernible difference.
One by-product of this method is that if no carrier is included, we can still recover
the carrier. This technique can also be used to recover the carrier. Take a signal
squaring it gives
1 2
[m ( t ) m2 ( t ) cos( 2 c t )]
2
The second term is the carrier at twice its frequency which we recover by filtering
at this frequency.
m(t) cos(wct)
Squaring
Circuit Filter at 2wc Divide by 2 cos(wct)
Envelope Detector
The envelope of a signal is its maximum value over a set sampling period. A diode circuit
used most often to detect the envelope of AM signals is the simplest and the universal
method of demodulating AM signals. The prerequisite for the use of this demodulation
method is the presence of a strong carrier and high SNR. Excessive amount of noise
causes severe envelope fluctuations and makes this method less effective. We all know of
the AM radio’s vulnerability to noise and other atmospheric perturbations.
Vi R C Vo
The envelope detector is basically a Diode-RC circuit as shown above. The signal
is applied to the terminals of the circuit. The Diode conducts as the voltage(amplitude)
increases and the capacitor charges up. Now as the voltage begins to go down, resistor
discharges and the capacitor lets go of its charge. The cycle continues and each charge of
the capacitor indicates the maximum value over that period. In fact the capacitor
discharges slightly between cycles as shown in the figure below but this can be
compensated for easily.
Summary
Baseband Signal - The baseband signal is usually the message signal. It has a bandwidth
of B. See Figure 20.
Passband Signal - The passband signal is one that has been multiplied by a carrier. It is
centered at the carrier frequency and has a bandwidth of 2B.
Double Sideband - When both sidebands and the carrier is transmitted, this is called the
AM or DSB modulation. DSB signals which are passband signals have a bandwidth of
2B.
Single Sideband - When either by filtering or phasing only one band is transmitted the
signal is called SSB. It has a bandwidth of B.
Vestigial Sideband - VSB is used for video broadcasting. VSB is a compromise between
SSB and DSB and has a bandwidth of .666B.
Figure 20 - follows
__________________________________
Charan Langton, Nov 4, 1998
Previous Tutorials are kept at the Advanced Systems Web site under CAP.
Thanks much to Eric Arakaki and Dave Watson for their invaluable comments and edits.
Figure 20 - AM Waveforms
2
2
1
fm = -
m 1( t ) 1
1
0.6 2
0 t 3
2. 2.
3
3
1.5
c1( t )
fc = -8
m 1( t )
1.5
3 3
0 t 3
3. 3.
DSB Waveform dsb(t ) (1 km1(t ))c1(t ) k = 60%
-9, -8, -7
2.5
2.5
1.13
dsb ( t )
0.25
m e( t )
1.63
3 3
0 t 3
ak2 1
4. 4. DSB waveform Overmodulated k = 150%
Note the envelope of the signal is no longer same as the baseband signal trace, hence
there is no way to demodulate it from the envelope of this signal.
4
4
2.25
-9, -8, -7
dsbo ( t )
m e3( t ) 0.5
m 1a( t )
1.25
3 3
0 t 3
3
3
1.5
ssbu2 ( t ) -9
m e( t )
1.5
3 3
0 t 3
-7
3
3
1.5
ssbl2 ( t )
m e( t )
1.5
3 3
0 t 3
7. 7. DSB-Suppressed Carrier dsbsc(t ) 2c1(t )m1(t ) ssbl (t ) ssu(t )
3
3
1.5
dsbsc2 ( t ) -9, -8, -7
m e( t )
1.5
3 3
0 t 3