isl3873 Intersil Corporation, isl3873 Datasheet - Page 26

• Manufacturer Part Number: isl3873
• Description: Wireless Lan Integrated Medium Access Controller With Baseband Processor
• Manufacturer: Intersil Corporation

Channel Matched Filter (CMF) Description
The receive section shown in Figure 19 operates on the
RAKE receiver principle which maximizes the SNR of the
signal by combining the energy of multipath signal
components. The RAKE receiver is implemented with a
Channel Matched Filter (CMF) using a FIR filter structure with
16 taps. The CMF is programmed by calculating the Channel
Impulse Response (CIR) of the channel and mathematically
manipulating that to form the tap coefficients of the CMF.
Thus, the CMF is set to compensate the channel
characteristics that distort the signal. Since the calculation of
the CIR is inaccurate at low SNR or in the presence of strong
CW interference, the chip has thresholds (CR 36 to 39) that
are set to substitute a default CMF shape under those
conditions. This default CMF shape is designed to
compensate only the known transmit and receive non linearity.
PN Correlators Description
There are two types of correlators in the ISL3873 baseband
processor. The first is a parallel matched filter correlator that
correlates for the Barker sequence used in preamble,
header, and PSK data modes. This Barker code correlator is
designed to handle BPSK spreading with carrier offsets up
to 50ppm and 11 chips per symbol. Since the spreading is
BPSK, the correlator is implemented with two real
correlators, one for the I and one for the Q Channel. The
same Barker sequence is always used for both I and Q
correlators.
These correlators are time invariant matched filters
otherwise known as parallel correlators. They use one
sample per chip for correlation although two samples per
chip are processed. The correlator despreads the samples
from the chip rate back to the original symbol rate giving
10.4dB processing gain for 11 chips per symbol. While
despreading the desired signal, the correlator spreads the
energy of any non correlating interfering signal.
The second form of correlator is the parallel correlator bank
used for detection of the CCK modulation. For the CCK
modes, the 64 wide bank of parallel correlators is
implemented with a Fast Walsh Transform to correlate the 4
or 64 code possibilities. This greatly simplifies the circuitry
of the correlation function. It is followed by a biggest picker
which finds the biggest of 4 or 64 correlator outputs
depending on the rate. This is translated into 2 or 6 data
bits. The detected output is then processed through the
differential phase decoder to demodulate the last two bits
of the symbol.
Data Demodulation and Tracking
Description (DBPSK and DQPSK Modes)
The signal is demodulated from the correlation peaks
tracked by the symbol timing loop (bit sync) as shown in
Figure 18. The frequency and phase of the signal is
corrected using the NCO that is driven by the phase locked
26
ISL3873
loop. Averaging the phase errors over 10 symbols gives the
necessary frequency information for seeding the NCO
operation.
Data Decoder and Descrambler
Description
The data decoder that implements the desired DQPSK
coding/decoding as shown in Table 11. The data is formed
into pairs of bits called dibits. The left bit of the pair is the first
in time. This coding scheme results from differential coding
of the dibits. Vector rotation is counterclockwise for a positive
phase shift, but can be reversed with bit 7 or 6 of CR 1.
For DBPSK, the decoding is simple differential decoding.
The data scrambler and de-scrambler are self synchronizing
circuits. They consist of a 7-bit shift register with feedback of
some of the taps of the register. The scrambler is designed
to ensure smearing of the discrete spectrum lines produced
by the PN code.
One thing to keep in mind is that both the differential
decoding and the descrambling cause error extension or
burst errors. This is due to two properties of the processing.
First, the differential decoding process causes errors to
occur on pairs of symbols. When a symbol’s phase is in
error, the next symbol will also be decoded wrong since the
data is encoded in the change in phase from one symbol to
the next. Thus, two errors are made on two successive
symbols. Therefore up to 4 bits may be wrong although on
the average only 2 are. In QPSK mode, these may occur
next to one another or separated by up to 2 bits. In the CCK
mode, when a symbol decision error is made, up to 6 bits
may be in error although on average only 3 bits will be in
error. Secondly, when the bits are processed by the
descrambler, these errors are further extended. The
descrambler is a 7-bit shift register with two taps exclusive
or’ed with the bit stream. Thus, each error is extended by a
factor of three. Multiple errors can be spaced the same as
the tap spacing, so they can be canceled in the descrambler.
In this case, two wrongs do make a right. Given all that, if a
single error is made the whole packet is discarded anyway,
so the error extension property has no effect on the packet
error rate. It should be taken into account if a forward error
correction scheme is contemplated.
Descrambling is self synchronizing and is done by a
polynomial division using a prescribed polynomial. A shift
register holds the last quotient and the output is the exclusive-
or of the data and the sum of taps in the shift register.
PHASE SHIFT
+180
TABLE 11. DQPSK DATA DECODER
+90
-90
0
DIBIT PATTERN (D0, D1)
D0 IS FIRST IN TIME
00
01
11
10