Digital Signal Processing Reference
In-Depth Information
Theoretically, the best way to accomplish this would be to use a matched fil-
ter receiver, where the input signal is correlated with a template pulse. Several
strong arguments can be formulated against this approach. One of the prereq-
uisites of using a matched filter is that the shape of the input waveform is
known in advance. However, assuming that the path from the input terminal of
the transmitting antenna all the way down to the receiver can be approximated
by a flat frequency- and a linear phase-characteristic is a huge mistake. The
waveform of the pulses depends on the matching between the transmitter (re-
ceiver) and its antenna, the geometrical radiation point of a certain frequency
on the antenna surface and whether pulses overlap or not (as a result of multi-
path interference). Under the assumption of a los channel, the initial radiation
characteristics of the antennas will also drastically change if an electromag-
netic responsive material is brought in the neighborhood of the transmitter or
the receiver.
Even if an acceptably accurate template pulse waveform could be constructed,
there is still a second and potentially far more serious problem related to the
demodulation of pulses using a matched filter. This time, the problem is only
brought to the surface by diving deep into the circuit level of a receiver design.
Remember that our original purpose was to bring (part of) the demodulation
process closer to the antenna of the receiver. For example, in a heterodyne re-
ceiver, the downconversion mixer is located just behind the lna. At this point,
the signal-of-interest is still extremely weak, in the order of a few (tens of)
microvolts. In order to obtain a high conversion gain, the amplitude swing of
the lo-inputs
12
of the mixer is kept as high as possible. Any practical im-
plementation will suffer from spurious frequencies at the mixer output due to
lo-feedthrough and self-mixing (Figure 5.6).
Fortunately, in the case of a single frequency in the lo-signal this results in a
dc-offset
13
and some spurious signals at the lo-frequency and at some higher
harmonics of f
lo
(Figure 5.6). However, doing the same in the correlation filter
of a pulse-based radio system is doomed to fail catastrophically, for the follow-
ing reason. The power spectral density of the template pulse waveform applied
to the mixer inputs of an analog correlator spans the entire pulse spectrum. As
a result of self-mixing and leakage, the output spectrum of the mixer will thus
contain unwanted components over the entire range from dc to twice the max-
imum pulse frequency. Since the power of the locally generated template sig-
nal is much stronger than the incoming rf-signal, these spurious components
could cause clipping further down the chain. Even if the pulses are spaced
12
lo: local oscillator.
13
This dc-offset signal is the biggest architectural problem in zero-if receivers.
