Digital Signal Processing Reference
In-Depth Information
Indicative figures for an 802
.
11a/g receiver
Without proof, suppose that 10 bits are used as word length for the ad-
converter in the 802
.
11a/g receiver. The loading factor of the converter
[Tag07] is chosen to be
12 dB which means that, on average, only
8 out of the 10 bits are used to quantize the input signal. The main
reason for limiting the rms signal swing to about 25% of the full-scale
range of the converter is the necessity to cope with the peak-to-average
power ratio of the ofdm signal. Remember that, for the 52-subcarrier
802
.
11g standard, the peak papr of the time-domain signal is 17 dB.
However, the possibility that all subcarriers constructively interfere is
very low so a margin of only 8 dB - slightly more than 1 bit - should
be sufficient.
The second reason for limiting the signal swing is to cope with the
possibility of an incorrect gain setting in the variable gain amplifier
(vga) in front of the ad converter. In general, the automatic gain con-
trol (agc) of the receiver attenuates the gain of the vga in only a
discrete number of steps. Rather than taking risk of clipping, which
has a dramatic impact on the performance, the ad-converter is driven
at some back-off value from the maximum signal level. Summarizing,
8 bits of the converter are used to obtain a sufficiently high snr at the
output and the 2 remaining most significant bits are only used to cope
with peak signal values and the non-ideal behaviour of the agc.
−
Now, suppose that the power of the narrowband interferer is of the same order
of magnitude as the power of the ofdm transmission, say 20 dBm. If the dis-
tance between the (un)intended interferer and the ofdm receiver is larger than
or equal to the link distance of the ofdm transmission itself, the input power as
seen by the receiver will only rise by 3 dB or less. The headroom provided by
the 2 most significant bits in the ad-converter is more than sufficient to handle
this situation. The unwanted spur is removed from the spectrum in the digi-
tal back-end without further consequences on performance and life goes on.
Then, the distance between the interferer and the receiver is halved. Using the
free-space path loss model,
14
it follows that the interferer power available at
the antenna of the ofdm receiver becomes four times (6
.
02 adB) larger. Most
of the signal energy residing in the band-of-interest is useless, but clipping in
the ad-converter is guaranteed if no further action is undertaken (see inset).
One of the possible measures to cope with the increased spurious power is to
knit an extra bit on the word length of the converter. In best case scenario, doing
so will double the power consumption of the ad-converter [Wal99], which is a
14
PL
=
PL
d0
+
10
γ
·
log
10
(
d
/
d0
)
, with
γ
=
2
(
free space
)
···
4andPL
d0
≈
47dB [Fri46].
