Pulse or digital communications – Receivers – Particular pulse demodulator or detector
Reexamination Certificate
2001-12-20
2004-03-30
Ghayour, Mohammad H. (Department: 2631)
Pulse or digital communications
Receivers
Particular pulse demodulator or detector
Reexamination Certificate
active
06714607
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to joint demodulation of desired and interfering signals in Time Division Multiple Access (TDMA) wireless networks for both mobile and fixed applications.
2. Description of the Related Art
In mobile and fixed TDMA networks, frequencies may be reused throughout a given area. Cells are sufficiently separated to insure that co-channel interference is small relative to the desired signal. However, there are limits to how often a frequency can be reused since eventually the carrier-to-interference (C/I) ratio becomes too small for the receiver to properly process a received signal. Since reusing frequencies more often implies more channels per cell or sector, and a corresponding increase in capacity, there is great motivation to develop receivers which can operate at low values of C/I (e.g. 4 dB and below).
Many interference cancellation techniques other than joint demodulation require at least two antennas on the receive path. Although cellular and personal communication system (PCS) base station equipment typically employ multiple antennas on the receive path, mobile terminals and handset typically employ only a single antenna in light of aesthetics and space constraints.
In conventional receivers employing a single antenna, a number of signal processing techniques may be employed to develop an estimate of the desired signal. A common receiver implementation for wireless networks with multipath fading uses a Maximum Likelihood Sequence Estimator (MLSE). A detailed description of MLSE can be found in Forney, “Maximum Likelihood sequence estimation of digital sequences in the presence of intersymbol interference”, IEEE Transactions on Information Theory, Vol. IT-18, pp. 363-378, May 1972.
The MLSE attempts to generate all possible received signals based on all possible transmitted sequences. The resulting locally-generated estimates are compared to the signal that is actually received. The locally-generated signal that most closely matches the received signal indicates the most likely transmitted sequence. The baseband signal flow associated with such a receiver and its corresponding transmitter is shown in FIG.
1
.
A bit stream b(t) is input to a modulator
10
. Based on its input bit(s)-to-modulator symbol mapping, the modulator
10
produces an output m(t). The output m(t) is an input to a wireless channel
12
. For purposes of this discussion, the wireless channel
12
is assumed to be characterized as slow, frequency-selective Rayleigh fading. The “slow” description implies that the wireless channel
12
does not change significantly over a symbol time of the modulator
10
. The “frequency-selective” descriptor implies the existence of multipath, the delay of which from the main signal is a significant portion of the modulator symbol time (e.g. greater than 25%). The wireless channel
12
introduces intersymbol interference (ISI), which typically must be compensated for to achieve satisfactory receiver performance. The desired signal after traversing the wireless channel
12
is denoted as C(t).
A separate, co-channel interfering bit stream is denoted by b
i
(t). A modulator
14
produces an output m
i
(t) based on the interfering bit stream b
i
(t). The output m
i
(t) is an input to a wireless channel
16
. A resultant signal of the separate interfering signal path is denoted as I(t).
Although only one interfering signal is shown, there may be more than one interferer depending on frequency reuse of the underlying cellular or PCS network. On a TDMA burst-by-burst basis, there may be one or more interferers whose respective signal powers contribute significantly to the total interference power. For simplicity in discussion, only one interferer is shown.
A received signal r(t) is equal to a sum of C(t) and I(t). The received signal r(t) is input to a MLSE
20
. The MLSE
20
may use a Viterbi algorithm to implement the MLSE functionality. In this case, the MLSE
20
may also be referred to as a Viterbi equalizer. A key parameter of the Viterbi equalizer is the number of processing states. The number of processing states is typically calculated by M
L
, where M represents a modulator alphabet size, and L represents a memory order of the path between the modulator input and the input to the equalizer. For example, a link employing a binary modulation format, M=2, and a memory order of L=4 would require a 2
4
=16-state Viterbi equalizer. A particular state in this example is defined by the four previous modulator symbols, m
n−1
to m
n−4
, wherein the time dependence is accounted for in the subscripts. Assuming the modulator alphabet consists of +1 and −1, the states are defined by the 16 possible combinations of +1 and −1. A possible previous modulator output-to-state mapping is shown in FIG.
2
. This particular configuration implies that a current received signal r
n
can be estimated as a function of a current modulator output symbol m
n
and the four previous modulator output symbols m
n−1
to m
n−4
convolved with a channel impulse response (CIR) of the desired signal path.
In general, the signal estimate for state s can be expressed mathematically as:
r
^
n
=
∑
l
=
1
L
+
1
h
^
n
(
l
)
m
n
-
l
+
1
(
s
)
(
1
)
where ĥ
n
(l) is the l
th
coefficient of the estimate of the CIR for the desired signal path at time n, and m
(s)
is the previous modulator symbol sequence associated with state s plus the current symbol.
The computation performed for each processing state at each trellis stage within the Viterbi equalizer determines the distance squared between the actual received signal and the locally-generated estimates developed using the above equation. The equation for distance squared is:
d
2
=|real
(
r
n
)−
real
({circumflex over (r)}
n
)|
2
+imag
(
r
n
)−
imag
({circumflex over (r)}
n
)|
2
(2)
where real( ) and imag( ) are the real and imaginary parts of the respective quantities.
This calculation is often called a branch metric calculation. The branch metric calculation is added to the cumulative metric associated with each state to produce a new set of cumulative metrics as in a standard Viterbi algorithm. For binary modulation, there are two new cumulative metrics generated for each state: one for each of the two possible modulator symbols. Thus, for a 16-state equalizer, there are 32 new paths generated at each trellis stage. At each new state, there are two incident paths. The path having the minimum cumulative metric is selected as the surviving path. This implies that the selected path is a better match to the received signal sequence than the non-selected path. The detailed processing associated with a Viterbi equalizer is described in many references, such as Forney, “The Viterbi Algorithm”, Proceedings of the IEEE, Vol. 61, No. 3, March 1973, pp. 268-278. The end result is that the Viterbi equalizer produces an estimate of the transmitted bit stream b(t) that was input to the modulator
10
. Soft decisions would be employed if there was a following stage of channel decoding as is the case for Global System for Mobile Communications (GSM) and ANSI-136 TDMA networks. Soft decisions are described in Hagenauer et al., “A Viterbi Algorithm with Soft-Decision Outputs and its Applications”, CH2682-3/89/0000-1680, 1989, IEEE.
From the above discussion, it is apparent that the complexity of the equalizer is driven by the number of states, since the amount of processing and memory required increases significantly with this key parameter. Further, since the number of states is seen to grow exponentially with the memory order L, it is important to set L to just what is needed to properly process the received signal. Too high a value of L results in a prohibitive amount of processing, while too low a value of L results in poorer signal estimates and thus poorer performance.
Conventional receivers estimate the maximum amount of memory
Kobylinski Richard A.
Wolter David Randall
Brinks Hofer Gilson & Lione
Ghayour Mohammad H.
SBC Technology Resources Inc.
Williams Demetria
LandOfFree
Joint demodulation using a viterbi equalizer having an... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Joint demodulation using a viterbi equalizer having an..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Joint demodulation using a viterbi equalizer having an... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3217112