Amplitude Modulation
Amplitude Modulation
Amplitude Modulation
Amplitude Modulation
Contents
Slide 1 Amplitude Modulation
Slide 2 The Envelope and No Overmodulation
Slide 3 Example for Single Tone Modulation
Slide 4 Measuring the Modulation Index
Slide 5 Transmitted vs. Message Power in s(t)
Slide 6 Powers in Single Tone Case (cont.)
Slide 7 Spectrum of an AM Signal (cont.)
Slide 8 Demodulating by Envelope Detection
Slide 9 Square-Law Envelope Detector (cont.)
Slide 10 Sampling Rate for Square-Law Detector
Slide 11 Hilbert Transforms and Complex
Envelope
Slide 12 Hilbert Transforms (cont.)
Slide 13 Pre-Enveope or Analytic Signal
Slide 14 Spectrum of Pre-Envelope
Slide 15 Pre-Envelope of AM Signal
Slide 16 Envelope Detector Using the Hilbert
Transform
5-ii
Slide 31 Experiment 5.4 Envelope Detector Using
the Hilbert Transform (cont. 1)
Slide 32 Experiment 5.4 Envelope Detector Using
the Hilbert Transform (cont. 2)
5-iii
✬ ✩
where
• m(t) is the baseband message.
• c(t) = Ac cos ωc t is called the carrier wave.
• The carrier frequency, fc , should be larger
than the highest spectral component in m(t).
• The parameter ka is a positive constant called
the amplitude sensitivity of the modulator.
✫ ✪
5-1
✬ ✩
The Envelope and No
Overmodulation
• e(t) = Ac |1 + ka m(t)| is called the envelope of
the AM signal. When fc is large relative to
the bandwidth of m(t), the envelope is a
smooth signal that passes through the
positive peaks of s(t) and it can be viewed as
modulating (changing) the amplitude of the
carrier wave in a way related to m(t).
✫ ✪
carrier phase and frequency.
5-2
✬ ✩
0.5
−0.5
−1
−1.5
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Normalized Time t /T
m
✫ ✪
5-3
✬ ✩
emin = Ac (1 − µ)
✫ ✪
5-4
✬ ✩
✫ ✪
5-5
✬ ✩
Ps = 0.5A2c + 0.25A2c µ2
✫ ✪
5-6
✬ ✩
◗ ✑
◗ ✑
◗ ✑ ω
−W 0 W
(a) Fourier Transform of Baseband Signal
S(ω)
Ac Ac
ka M (ω +ωc ) ka M (ω −ωc )
✁ ✁
2 2
πAc πAc
✁
☛
✁ ✻ ✁
✁
☛ ✻
◗ ✑ ◗ ✑
◗ ✑ ◗ ✑
◗ ✑ ◗ ✑ ω
−ωc −W −ωc −ωc +W 0 ωc −W ωc ωc +W
✫ ✪
5-7
✬ ✩
Demodulating an AM Signal by
Envelope Detection
Method 1: Square-Law Demodulation
Lowpass
Filter
s(t) p y(t)
✲ (·)2 ✲ H(ω) ✲ (·) ✲
✫ ✪
5-8
✬ ✩
✫ ✪
5-9
✬ ✩
✫ ✪
5-10
✬ ✩
✫ ✪
5-11
✬ ✩
✫ ✪
5-12
✬ ✩
Example 1.
The pre-envelope of x(t) = cos ωc t is
Example 2.
Let m(t) be a lowpass signal with cutoff frequency
W which is less than the carrier frequency ωc .
Then the pre-envelope of x(t) = m(t) cos ωc t is
✫ ✪
5-13
✬ ✩
s+ (t) = Ac [1 + ka m(t)]ej ωc t
s̃(t) = Ac [1 + ka m(t)]
e(t) = |x̃(t)|
✫ ✪
5-15
✬ ✩
ŝ2 (t)
✲ −j sign ω ŝ(t)
✲ (·) 2
s(t) ❄
✐ ✲
p
✲
e(t)
+ (·)
✲ ✻
(·)2
s2 (t)
✫ ✪
5-16
✬ ✩
Chapter 5, Experiment 1
Making an AM Modulator
Write a C program for the TMS320C6713 to:
1. Initialize McBSP0, McBSP1, and the codec as
in Chapters 2 and 3, and use the left channel.
2. Read samples m(nT ) from the codec at a 16
kHz rate.
Convert the samples into 32-bit integers
by shifting them arithmetically right by 16
bits. The resulting integers lie in the range
±215 .
3. AM modulate the input samples to form the
sequence
✫ ✪
5-17
✬ ✩
✫ ✪
5-18
✬ ✩
✫ ✪
5-19
✬ ✩
✫ ✪
5-20
✬ ✩
Experiment 5.2 Making a Square-Law
Detector (cont.)
✫ ✪
to the Line Out.
5-21
✬ ✩
Experiment 5.2.1
Observing the Square-Law Detector
Output with No Input Noise
• Attach the signal generator to the ADC input
and the oscilloscope to the DAC output.
• Set the signal generator to create an AM
wave with a sinusoidal modulating signal with
a frequency between 100 and 400 Hz. Use a 3
kHz carrier frequency.
• Capture and plot the AM input signal and
the output of your demodulator.
Experiment 5.2.2
Observing the Square-Law Detector
Output with Input Noise
Experiment demodulating signals corrupted by
additive, zero mean, Gaussian noise.
✫ ✪
5-22
✬ ✩
✫ ✪
5-23
✬ ✩
✫ ✪
5-24
✬ ✩
Gaussian RV’s (cont. 1)
float v;
v = (float) rand()/(RAND_MAX + 1);
6
Gaussian RV’s (cont. 2)
2
− r2
Let v = 1 − e 2σ u(r). Then, for
✫ ✪
5-25
✬ ✩
✫ ✪
5-26
✬ ✩
✫ ✪
5-27
✬ ✩
1 − az −1
✫ ✪
5-28
✬ ✩
Experiment 5.3
Hilbert Transform Experiments
1. Design an N = 29 tap Hilbert transform filter
using remez.exe. Use a sampling rate of
16,000 Hz. Use just one band extending from
1000 Hz to 7000 Hz with a gain of 1. The tap
at delay (N − 1)/2 = 14 is called the center
tap. There are 14 taps for delays 0, . . . , 13
before the center tap and 14 taps for delays
15, . . . , 28 after the center tap.
2. Plot the amplitude response of your filter in
dB vs. frequency.
3. Write a C program to implement your filter
using the DSK. Take the input samples from
the signal generator from the left codec input
channel. Send the filter output samples to the
left codec output channel. Also send the
delayed input samples at the center tap of the
filter to the right codec output channel.
✫ ✪
5-29
✬ ✩
4. Set the signal generator for a sine wave with
a frequency in the 1000 to 7000 Hz passband.
Display the left and right line outputs on two
oscilloscope channels.
5. Use the oscilloscope phase difference
measuring function to measure the phase shift
between the left and right channel outputs.
Vary the input frequency over the passband
and check that the phase difference remains
90 degrees.
Experiment 5.4
Making an Envelope Detector Using the
Hilbert Transform
• Write a program for the DSK to implement
the envelope detector shown on Slide 5-16.
Again, assume a carrier frequency of 3 kHz
and a baseband message bandlimited to 400
Hz. The codec is ac coupled to the line out
connector, so the dc component after the
square root box will automatically be
✫ ✪
removed.
5-30
✬ ✩
Experiment 5.4
Hilbert Transform Detector (cont. 1)
• You can design the Hilbert transform filter
with the program REMEZ.EXE.
– Use an odd number, N , of filter taps.
– Good results can be achieved with
REMEZ.EXE by using just one band with a
lower cutoff frequency f1 and upper cutoff
frequency f2 chosen to pass the AM signal.
Choosing the band to be centered in the
Nyquist band also seems to improve the filter
amplitude response generated by
REMEZ.EXE. To center the band, choose the
upper cutoff frequency to be
f2 = 0.5fs − f1
✫ ✪
factor.
5-31
✬ ✩
Experiment 5.4
Hilbert Transform Detector (cont. 2)
• Select N so that the amplitude response of
the filter is quite flat over the signal passband
and ripples caused by incomplete cancellation
of the 2fc components are essentially invisible
in the demodulated output.
• You can also design the Hilbert transform
filter with the program WINDOW.EXE. Try
using the Hamming and Kaiser windows. You
can make a tradeoff between the transition
bandwidth and the out-of-band attenuation
with the Kaiser window.
• The FIR filter designed with either program
will have a delay equal to the delay from the
input to the center tap of the filter,
T (N − 1)/2. Therefore, s(nT ) must be
delayed by this amount to match the delay in
ŝ(nT ). This can be accomplished by taking
s(nT ) from the point in the delay-line of the
✫ ✪
Hilbert transform filter at its center tap.
5-32
✬ ✩
✛ L delays ✲ ✛ L delays ✲
❄
s[n] s[n − L]
✲ z −1 ✲ ··· ✲ z −1 ✲ z −1 ✲ ··· ✲ z −1
h0 h1 hL−1 hL hL+1 hN −1
❄ ❄ ❄ ❄ ❄ ❄ ❄
+
✲ y[n] = ŝ[n − L]
✫ ✪
5-33
✬ ✩
The Delay of the FIR Hilbert
Transform Filter (cont. 1)
The frequency response for the filter is
N
X −1 L−1
X
H ∗ (ω) = hn e−jωnT = hn e−jωnT
n=0 n=0
2L
X
+ hL e−jωLT + hn e−jωnT
n=L+1
✫ ✪
5-34
✬ ✩
The Delay of the FIR Hilbert
Transform Filter (cont. 2)
The taps of the Hilbert transform filter have odd
symmetry about the center tap. That is
✫ ✪
the sum.
5-35
✬ ✩
FV (v)
0 1
F (y)
v = F (y)
0 Y = F −1 (V ) y = F −1 (v)
Y = F −1 (V )
FY (y) = P (Y ≤ y) = P (F −1 (V ) ≤ y)
= P (V ≤ F (y)) = FV (F (y)) = F (y)
✫ ✪
5-37
✬ ✩
Another Solution to the Transformation
The probability density function (pdf) for V is
the rectangle
0 for v < 0
fV (v) = 1 for 0 ≤ v ≤ 1
0 for v > 1
fV (v)
0 1
✫ ✪
5-39
✬ ✩
By the two to two transformation theorem
fR,Θ (r, θ) 1 2
− x +y
2
fX,Y (x, y) = = e 2σ 2
∂(x,y)
∂(r,θ) 2πσ 2
1 2
− x2 1 2
− y2
= √ e 2σ √ e 2σ
σ 2π σ 2π
Performance of the AM Envelope
Detectors in the Presence of
Additive Noise
Suppose the baseband message, m(t), is a lowpass
signal with cutoff frequency W , that is, its
spectrum is zero except for ω ∈ [−W, W ]. The
transmitted AM signal is
✫ ✪
5-41
✬ ✩
The Square-Law Envelope Detector
The square-law envelope detector first forms
r2 (t) = [a(t) + nI (t)]2 cos2 (ωc t)]
− 2[a(t) + nI (t)]nq (t) sin(ωc t) cos(ωc t)
+ n2Q (t) sin2 (ωc t)
= [a(t) + nI (t)]2 0.5[1 + cos(2ωc t)]
− [a(t) + nI (t)]nq (t) sin(2ωc t)
+ n2Q (t)0.5[1 − cos(2ωc t)]
= 0.5{[a(t) + nI (t)]2 + n2Q (t)}
+ 0.5{[a(t) + nI (t)]2 − n2Q (t)} cos(2ωc t)
− [a(t) + nI (t)]nq (t) sin(2ωc t) (7)
The first line of (7) is a lowpass signal with cutoff
frequency 2W . The second and third lines are
bandpass signals centered around 2ωc with
spectra confined to the band
[2ωc − 2W, 2ωc + 2W ]. The square-law envelope
detector extracts the lowpass first line by passing
r2 (t) through a lowpass filter that cuts out the
✫ ✪
components centered around 2ωc .
5-42
✬ ✩
✫ ✪
5-43
✬ ✩
✫ ✪
5-44
✬ ✩
Additional Comments
The output of the envelope detectors before the
final square-root when expanded is
✫ ✪
5-45
✬ ✩
The noise you add to the signal has a power
spectral density of σ 2 extending over the entire
Nyquist band. Thus the difference in noise
bandwidths from no filtering in the Hilbert
transform detector and lowpass filtering in the
square-law detector account for your different
performances.
As an additional optional experiment you could
add a receive bandpass filter to your envelope
detectors and compare the performances.
✫ ✪
5-46