Sunteți pe pagina 1din 1

is not bandlimited.

So when the pulse is sent through a baseband channel which acts as a


lowpass filter, it will get smeared beyond its original duration. When a sequence of amplitude
modulated pulses are transmitted through this channel, each pulse will experience intersymbol
interference (ISI) from its neighboring pulses. The smaller the pulse duration T the more severe
1
the ISI. However, if T > 2W where W is the cutoff frequency of the baseband channel ISI can
be avoided by choosing the transmitted pulses carefully so that the received pulses satisfy the
Nyquist condition for zero ISI. Thus the channel bandwidth limits the maximum pulse rate T1
to less than 2W pulses per second which is called the Nyquist signaling rate. If the channel
response is not known a priori, making the received pulses satisfy the Nyquist condition is not
possible. In this case, the ISI can be cancelled by filtering the received signal. This process is
called equalization. Equalization involves estimating the channel response and then using this
estimate to remove the ISI. To enable channel estimation, the information to be transmitted is
preceded by a preamble which is predetermined sequence of bits agreed upon by the transmitter
and receiver. This sequence is also called the training sequence. The training sequence also
aids in the initial estimation of the bit boundaries in the received signal. The receiver typically
detects the arrival of a packet by matching the training sequence. After the initial estimation
of the timing, the signal level transitions induced by the line coding schemes help update the
estimate in case of clock drifts. This updating is performed using a delay-locked loop (DLL)
circuit.

If we send one bit per pulse when the pulse rate is equal to the Nyquist rate, the maximum data
rate is limited to 2W bits per second. However, we can map multiple bits to a single pulse by
allowing more than two levels. For example, we could map 00 to −Ap(t), 01 to − A3 p(t), 10 to
A
3 p(t) and 11 to Ap(t). In that case, the maximum data rate becomes 4W bits per second. In
general, if the number of amplitude levels are 2m where m is a positive integer the maximum data
rate is 2W m bits per second. This seems to suggest that we can increase the achievable data
rate by increasing the number of levels. But this strategy does not work due to the presence
of noise. As the number of levels increase while keeping the maximum amplitude constant,
the levels will get closer to each other and can be confused for each other easily resulting in
demodulation errors. These ideas are due to Claude Shannon who proved that the maximum
data rate in an additive white gaussian noise (AWGN) channel with bandwidth W and SNR S
can support with arbitrarily small BER is less than the following value,

C = W log2 (1 + S) bits/second (3.2)

which is called the capacity of the channel. For our purposes, we assume that the a finite-
bandwidth baseband channel allows us to transmit at a finite data rate with a small but non-zero
BER where the data rates are directly proportional to the bandwidth.

In summary, the main physical layer issues in baseband channels are timing recovery and ISI
mitigation. Both of them influence the type of modulation scheme which is employed. For
example, line coding schemes which have frequent signal level transitions are chosen because
they aid timing recovery at the receiver. Transmitted pulse shapes are chosen to minimize ISI
at the receiver. So in fact it is the ease of demodulation in the presence of these issues which
is the factor determining the choice of modulation scheme. When pulse shaping to avoid ISI is
not feasible, equalization is used to remove it. Both equalization and initial timing recovery are
accomplished by the presence of a preamble which is sent immediately preceding the information-
bearing signal.

21

S-ar putea să vă placă și