Communication Principles
- Frequency Spectrum
- Modulation and Demodulation
- Phase-locked loop
- Nyquist's Sampling Theorem
Notes:
Focus on the principles, not the formulas!
Frequency Transform
- given a complex function x(t), -oo < t < +oo$.
- j^2 = -1$
- e^{j2\pi ft} = \cos(2 \pi f t) + j\sin(2 \pi f t)$
- Fourier (frequency) transform:
x(t) = \int[-oo]^{+oo}X(f)e^{j2\pi ft}\,df$
- for some X(f), the frequency spectrum of x(t)$
- in every case we will look at, X(f)$ is symmetric about zero
Notes:
Give examples on the board, e.g. x(t) = cos 2 \pi f t, $X(f) =
delta(f)/2 + delta(-f) / 2$
Frequency Transform for
Periodic Functions
- if x(t) is periodic with period T (f = 1/T$)
- transform of x(t)$ is a sum instead of an integral:
x(t) = \sum[n=-oo]^{+oo}c[n] e^{j(2\pi n/T)t}$
cn = (1 / T) \int[-T/2]^{+T/2} x(t) e^{-j(2\pi n / T)t}\,dt$
Example of Periodic Functions:
Cosine Wave
- x(t) = \cos(2\pi f1 t)$
- since e^{j \theta} = \cos(\theta) + j \sin(\theta)$
- and \cos(-x) = \cos(x) whereas \sin(-x) = -\sin(x)$
- therefore \cos(\theta) = (1 / 2) (e^{j\theta} + e^{-j\theta})$
- therefore \cos(2\pi f1 t) = (1 / 2) (e^{j2\pi f1 t} + e^{-j2\pi f1 t})$
- which is a form of:
x(t) = \sum[n=-oo]^{+oo}c[n] e^{j(2\pi n/T)t}$
with c[(-1)] = c1 = (1 / 2) , and cn = 0 for all other n$.
Square Wave
- x(t) = 1 for (4n-1)T / 4 <= t < (4n + 1) T / 4$
- x(t) = 0$ otherwise
- decomposing into sine waves:
$x(t) = (1 / 2) + (2 / \pi) \cos ( (2 \pi / T) t) -
(2 / {3 \pi) } \cos ( (6 \pi / T) t) +
(2 / {5 \pi) } \cos ( (10 \pi / T) t) - ...$
which again is a form of
x(t) = \sum[n=-oo]^{+oo}c[n] e^{j(2\pi n/T)t}$
Notes:
Draw square wave and spectrum on board.
Dirac Impulse
delta[\epsilon](t) = 1/\epsilon for $-\epsilon/2 < t <
\epsilon/2, 0 for all other values of t$
x(s) = \int[-oo]^{+oo} x(t) delta(t - s) \, dt$
$\sum[n=-oo]^{oo} delta(t - nT) =
(1 / T) \sum[n=-oo]^{oo} e^{j(2\pi n/T)t}$
- dirac impulse samples the signal
- integral of dirac impulse is always one
Notes:
draw the dirac impulse on the board.
Draw the frequency spectrum of the dirac impulse on the board.
Amplitude Modulation
- x(t) = Ae^{j2\pi f1 t}$ is the signal to be modulated
- assume f0 > f1$ is the modulation frequency
- multiply x(t) by a sine wave \cos(2\pi f0 t)$ to give:
- y(t) = A/2 (e^{j2\pi(f1 + f0)t} + e^{j2\pi(f1 - f0)t}$
- Radio transmission: use more of the spectrum, use shorter antennas
Notes:
- Draw x(t) (sine wave).
- Draw frequency spectrum of x(t), single arrow
of scale A, f1$ to the right of the zero.
- Draw frequency spectrum
of y(t), two arrow of scale A/2, displaced f1$ to the right of the
f0 and -f0$.
- Draw modulated signal.
Amplitude Demodulation
- y(t) = x(t) \cos(2\pi f0 t)$ is the input signal.
- multiply y(t) by a sine wave \cos(2\pi f0 t) to give z(t)$
- result is z(t) = x(t)/2 + x(t)\cos(4\pi f0 t)$
- only retain frequencies in the range [-f1, f1]$
- if f0 > f1, result is x(t)/2$
Phase-Locked Loop
- voltage-Controlled oscillator: f(V) = f'0 + alpha V$
- transmission at frequency f0 =~ f'0$ (frequency of oscillator)
- multiply output of oscillator with incoming signal, average over time
- if phase is the same, average product will be positive, f'0$
will increase
- if phase is opposite, average product will be negative, f'0$
will decrease
- since f0 =~ f'0$, over the long term
f'0 will exactly track f0$
Notes:
Skip this slide if short of time (later than 1:05).
Nyquist's sampling theorem
- frequency-limited signal x(t)$ with all energy
between -f[max] < f < f[max] (e.g. f[max] = 20KHz$)
- Take period T < (1/2)f[max]$
(e.g. T = 20 u s$, corresponding to 50KHz)
- Sample x(t) at time nT for all -oo < n < oo$
- Given only the samples, I can reconstruct x(t)$ exactly.
Sampling in the Frequency Domain
Show picture C5
Notes:
- Explain that the reason we need to limit the frequencies of x$
is to avoid overlap between adjacent shifts in the frequency domain.
- Explain how limiting the frequency guarantees that we cannot "miss"
a significant event.
Homework
- Due Monday, September 14
- Problem 2-1.
- Problem 2-9. Assume:
- packet sizes of 1000 bits
- 10Mb/s$ bandwidth of IP links
- 100Mb/s$ bandwidth of ATM links
- 10 ms = 0.01s$ delay for ATM connection setup
- 10u s = 0.00001s$ delay through each ATM switch
- 100u s = 0.0001s$ delay through each IP router
- 50u s = 0.00005s$ delay through each link (ATM or IP)
Homework -- continued
Compute (and show your work) the break-even point (in number of
packets) for sending over IP versus sending over the ATM network.
Hints:
- Compute the delay for a single packet going through each of the
networks (you need to use all the above data except the connection
setup time).
- compute the difference in time between sending a single packet
over ATM and over IP
- compute the number of packets required to add up to the ATM
connection setup time.