Digital Signal Processing Reference
In-Depth Information
conversion of the baseband modulated signal to (and from) the rf carrier cen-
ter frequency. This is due to the fact that analog circuits can be tuned to the
frequency of interest, while maintaining a narrow operating bandwidth around
this particular frequency. Apart from extra losses due to the non-idealities of
the building blocks, tuning an analog circuit to a higher frequency has no di-
rect impact on its power efficiency. For a digital system however, processing a
baseband signal or the equivalent rf-passband signal
does
make a big differ-
ence. In order to handle the rf frequency components, the rf signal must be
processed at Nyquist sampling rate. In a digital discrete-time sampled system,
there is no equivalent to a tuned analog circuit. Of course the reader might ar-
gue that something like subsampling exists, but this technique basically comes
down to baseband signal processing glued to a somewhat obfuscated frequency
conversion step (frequency folding). To the knowledge of the author, there is
no such thing as passband logic hardware.
In the transmitter, the conversion from a mathematical data representation to
the analog signal that is applied to the antenna terminals is performed by
the digital-to-analog converter (dac) and some other building blocks such
as the rf power amplifier. In the receiver, the modulated rf-signal with low
fractional bandwidth is first shifted to a sufficiently low if or baseband fre-
quency and is then converted to the digital domain. Also noteworthy to men-
tion here is that in the front-end of a receiver, out-of-band blocker filtering
occurs before frequency conversion and amplification. By removing unwanted
frequency bands early in the signal chain, the linearity requirements and power
consumption of the remaining components are drastically reduced.
Due to the large dynamic range of the input signals that arrive at the receiver,
this pre-filtering step is commonly accomplished by surface acoustic wave
(saw) filters [Har69]. This type of passive filters provides a sometimes impres-
sive ratio between the passband insertion loss and the attenuation in stopband
(typically more than 40 dB [Fre99]). The disadvantage of saw filters is that
they cannot be tuned to the desired frequency during operation, so they are typ-
ically designed only to block frequencies which are out of the frequency band
of interest. In-band interferers are still present at the output of the preselection
filter. This is why in a superheterodyne receiver architecture (Figure 1.7), the
signal is first converted to a fixed intermediate frequency,
12
where it is filtered
using a ceramic channel select filter (455 kHz/4
.
5 MHz/10
.
7 MHz). Finally,
the if-signal can be further converted to baseband.
Of course, all those frequency conversion steps and off-chip saw filters do
not make the superheterodyne receiver the most favourite architecture among
12
The if-frequency should be chosen high enough so that mirror frequencies fall out of the passband of the
band select filter.
