Communication Systems/Binary Modulation Schemes

This page discusses the binary modulation schemes and "keying".

What is "Keying?"

Square waves, sinc waves, and raised-cosine rolloff waves are all well and good, but all of them have drawbacks. If we use an optimal, matched filter, we can eliminate the effect of jitter, so frankly, why would we consider square waves at all? Without jitter as a concern, it makes no sense to correct for jitter, or even take it into consideration. However, since the matched filter needs to look at individual symbols, the transmitted signal can't suffer from any intersymbol interference either. Therefore, we aren't using the sinc pulse.

Since the raised-cosine roll-off wave suffers from both these problems (in smaller amounts, however), we don't want to use that pulse either.

So the question is, what other types of pulses can we send?

It turns out that if we use some of the techniques we have developed using analog signal modulation, and implement a sinusoidal carrier wave, we can create a signal with no inter-symbol interference, very low bandwidth, and no worries about jitter. Just like analog modulation, there are 3 aspects of the carrier wave that we can change: the amplitude, the frequency, and the phase angle. Instead of "modulation", we call these techniques keying techniques, because they are operating on a binary-number basis.

There is one important point to note before continuing with this discussion: Binary signals are not periodic signals. Therefore, we cannot expect that a binary signal is going to have a discrete spectra like a periodic squarewave will have. For this reason, the spectral components of binary data are continuous spectra.

Amplitude Shift Keying

In an ASK system, we are changing the amplitude of the sine wave to transmit digitial data. We have the following cases:

• Binary 1: $A_{1}\sin(f_{c}t)$
• Binary 0: $A_{0}\sin(f_{c}t)$

The simplest modulation scheme sets A0 = 0V (turning the transmitter off), and setting A1 = +5V (any random non-zero number turns the transmitter on). This special case of ASK is called OOK (On-Off keying). Morse code uses OOK.

Another common special case of ASK sets A1 to some positive number, and A0 to the corresponding negative number A0 = -A1. We will mention this case again later.

In ASK, we have the following equation:

$a(t)\sin(\omega t)$

by the principal of duality, multiplication in the time domain becomes convolution in the frequency domain, and vice-versa. Therefore, our frequency spectrum will have the following equation:

$A(j\omega )*\delta (t-\omega )$

where the impulse function is the fourier-transform of the sinusoid, centered at the frequency of the wave. the value for A is going to be a sinc wave, with a width dependent on the bitrate. We remember from the Signals and Systems book that convolution of a signal with an impulse is that signal centered where the impulse was centered. Therefore, we know now that the frequency domain shape of this curve is a sinc wave centered at the carrier frequency.

Frequency Shift Keying

In Frequency Shift Keying (FSK), we can logically assume that the parameter that we will be changing is the frequency of the sine wave. FSK is unique among the different keying methods in that data is never transmitted at the carrier frequency, but is instead transmitted at a certain offset from the carrier frequency. If we have a carrier frequency of $f_{c}$, and a frequency offset of $\Delta f$, we can transmit binary values as such:

• Binary 1: $A\sin((f_{c}+\Delta f)t)$
• Binary 0: $A\sin((f_{c}-\Delta f)t)$

Similar to ASK, we have FSK, which uses 2 different frequencies to transmit data. For now we will call them $\omega 1,\omega 2$. Using the same logic that we used above, the fourier representations of these waves will be (respectively):

$A_{1}(j\omega )*\delta (t-\omega 1)$

$A_{0}(j\omega )*\delta (t-\omega 2)$

With one sinc wave centered at the first frequency, and one sinc wave centered at the second frequency. Notice that A1 and A0 are the half-square waves associated with the 1s and the 0s, respectively. These will be described later.

Error Rate

The BER of coherent QPSK in the presence of gaussian and Rayleigh noise is as follows:

Gaussian Noise	Rayleigh Fading
${\frac {1}{2}}\operatorname {erfc} \left({\sqrt {\frac {E_{b}}{N_{0}}}}\right)$	${\frac {1}{2}}\left(1-{\sqrt {\frac {\gamma _{0}}{2+\gamma _{0}}}}\right)$

Phase Shift Keying

PSK systems are slightly different then ASK and FSK systems, and because of this difference, we can exploit an interesting little trick of trigonometry. PSK is when we vary the phase angle of the wave to transmit different bits. For instance:

• Binary 1: $A\sin(f_{c}t+\phi _{1})$
• Binary 0: $A\sin(f_{c}t+\phi _{0})$

If we evenly space them out around the unit-circle, we can give ourselves the following nice values:

• Binary 1: $A\sin(f_{c}t+0)$
• Binary 0: $A\sin(f_{c}t+\pi )$

Now, according to trigonometry, we have the following identity:

$\sin(f_{c}t+\pi )=-\sin(f_{c}t)$

So in general, our equations for each signal (s) is given by:

• $s_{1}(t)=A\sin(f_{c}t)$
• $s_{0}(t)=-A\sin(f_{c}t)$

Which looks awfully like an ASK signal. Therefore, we can show that the spectrum of a PSK signal is the same as the spectrum of an ASK signal.

There are two commonally used forms of Phase Shift keying Modulation:

Binary Phase Shift Keying (BPSK)

Quadrature Phase Shift Keying (QPSK)

Binary Phase Shift keying is set out above.

QPSK

Quadrature Phase Shift Keying utilises the fact that a cosine wave is in quadrature to a sine wave, allowing 2 bits to be simultaneously represented.

• Binary 11: $A\sin(f_{c}t+0)+\cos(f_{c}+\pi /2)$
• Binary 10: $A\sin(f_{c}t+0)+\cos(f_{c}-\pi /2)$
• Binary 01: $A\sin(f_{c}t+\pi )++\cos(f_{c}+\pi /2)$
• Binary 00: $A\sin(f_{c}t+\pi )++\cos(f_{c}-\pi /2)$

QPSK has the advantage over BPSK of requiring half the transmission band width for the same data rate, and error probability.

Error Rate

The BER of coherent BPSK in the presence of gaussian and Rayleigh noise is as follows:

Gaussian Noise	Rayleigh Fading
${\frac {1}{2}}\operatorname {erfc} \left({\sqrt {\frac {E_{b}}{N_{0}}}}\right)$	${\frac {1}{2}}\left(1-{\sqrt {\frac {\gamma _{0}}{1+\gamma _{0}}}}\right)$