Note: Descriptions are shown in the official language in which they were submitted.
CA 02394510 2002-06-14
WO 01/45288 PCT/RU99/00494
1
METHODS AND APPARATUS FOR SIGNAL SEARCHING USING CORRELATION
This invention relates to methods and apparatus for
searching for signals using correlation of a received signal
with a reference signal, in particular for detecting the
presence and code phase of wideband signals using CDMA (code
division multiple access) in CDMA wireless cellular
communications systems.
Background
Communications systems using CDMA signals have
advantages in terms of their capacity, frequency planning,
communications quality, security from unauthorized access, and
immunity to interference. However, a significant challenge in
CDMA system design arises from a need to achieve precise
synchronization between a desired signal, contained in a
received signal, and a locally generated reference signal. A
first step in this synchronization is a signal searching
process in which one or more parameters, such as the code phase
and frequency of a pseudo-noise (PN) signal which constitutes
the reference signal, are varied and hypotheses on the presence
of a desired signal are progressively evaluated. This uses
significant time and hardware resources of a CDMA system
receiver.
More particularly, for each possible set of values of
the parameters of the reference signal, referred to as an
offset position, or state of the reference signal, a
correlation is performed with a received signal and resulting
correlation values are evaluated to determine the likely
presence or absence of a desired signal with the respective
offset or position.
CA 02394510 2002-06-14
WO 01/45288 PCT/RU99/00494
2
It is known that communicated signals in CDMA
wireless cellular communications systems are subject to fading,
in which reduced RF signal amplitude and variations in phase
cause substantial degradation of the signal searching process.
To combat fast signal phase variance, or fast fading,
it is known for example from "DS-SS Code Acquisition in a Rapid
Fading Environment" by Manabu Mukai and Mutsumu Serizawa, IEEE
0-7803-2742-X/95, pages 281-285, 1995, to divide an
accumulation interval TA into a plurality of m consecutive
short intervals TCOH, so that TA = mTCOH. The duration of each
interval TCOH is selected to be sufficiently small that the
signal phase does not change appreciably during this interval,
which accordingly is referred to as a signal coherence
interval. During the accumulation interval TA correlation
results for a respective offset or position of the reference
signal are non-coherently accumulated. However, slow fading
can cause the accumulated correlation result, which is used for
determining the presence or absence of a desired signal with
the respective offset, to differ appreciably from its long-term
value, resulting in increased missed or false signal detection
probability. Increasing the length of the accumulation
interval TA to reduce this disadvantage would undesirably
increase the signal searching time.
Effects of slow fading can be reduced by using
diversity methods, for example spatial diversity techniques in
which signals having relatively independent fading
characteristics are received by two or more spaced antennas and
the correlation results for these signals are combined.
However, this undesirably increases complexity of the receiver,
and the use of plural spaced antennas may not be practical,
especially for portable receivers of small size.
CA 02394510 2002-06-14
WO 01/45288 PCT/RU99/00494
3
United States Patent No. 5,550,811 issued August 27,
1996 and entitled "Sync Acquisition and Tracking Circuit for
DS/CDMA Receiver" discloses a time diversity arrangement for
compensating for slow fading in which a selector switch
supplies correlation results for each position or parameter set
periodically to a respective combiner for performing a non-
coherent accumulation. The switching period is determined in
accordance with a fading period, so that the accumulated
correlation results are averaged relative to signal fading.
This arrangement has the disadvantage of being complex to
implement, requiring the selector switch, its control
arrangement, and as many combiners as there are parameter sets
or positions. For example, for a mobile station searching for
a pilot signal from a base station in an IS-95 CDMA wireless
cellular communications system there are N = 32768 possible PN
code phases or positions.
"CDMA. Principles of Spread Spectrum Communication"
by Andrew J. Viterbi, Addison-Wesley Communication Series,
1995, Section 3.4.1, "Single-Pass Serial Search" discloses a
signal searching method in which m correlation estimates in
consecutive adjacent signal coherence intervals TCOH are
accumulated during an accumulation interval TA = mTCOH. In an
IS-95 CDMA system with a signal bandwidth of 1.25 MHz at a
frequency of 800 MHz, the Rayleigh fading period is about 20 to
50 ms whereas the access channel signal accumulation interval
length may be from 1.2 to 2.4 ms. Accordingly, substantial
signal amplitude variations due to fading can occur within the
accumulation interval TA, resulting in increased missed and
false signal detection probability.
CA 02394510 2002-06-14
WO 01/45288 PCT/RU99/00494
4
An object of this invention is to provide a method
and apparatus which can facilitate signal searching in
communications systems such as CDMA systems.
Summary of the Invention
In accordance with one aspect of the invention, a
signal searcher for a CDMA communications system generates at
least one reference signal, and correlates the reference signal
with a received signal in correlation intervals each
sufficiently short that a phase of the received signal does not
change appreciably during the correlation interval. The
correlation results for the same code phase for a plurality of
said correlation intervals are accumulated, and the presence
and code phase of one or more desired signals in the received
signal are determined in dependence upon the accumulated
correlation results. The code phase of the reference signal is
changed through all of N possible code phases in N successive
correlation intervals constituting one signal scan cycle, the
correlation results being stored for accumulation over a
plurality of signal scan cycles. This aspect of the invention
also provides means, for example a digital signal processor,
for carrying out these functions.
Another aspect of the invention provides a method of
signal searching in which a received signal is correlated with
a reference signal and at least one parameter of the reference
signal is changed to produce respective correlation results for
different ones of N possible offsets between the received
signal and the reference signal, comprising the steps of: in a
scan cycle, producing a respective correlation result for each
of the N possible offsets from correlations between the
received signal and the reference signal with a respective
offset each during a correlation interval during which a phase
CA 02394510 2002-06-14
WO 01/45288 PCT/RU99/00494
of the received signal does not change appreciably, said at
least one parameter being changed between consecutive ones of
said correlations; and accumulating the correlation results for
a plurality of said correlations having corresponding offsets
5 in successive scan cycles to enable presence and offset of a
desired signal to be determined from the accumulated
correlation results.
In one embodiment of this method the correlation
results are accumulated for all N possible offsets between the
received signal and the reference signal. In another
embodiment this method comprises the step of determining in a
scan cycle largest correlation results for L of the N possible
offsets, where L is an integer less than N, the correlation
results being accumulated for only said L offsets. A further
embodiment comprises the step of determining in a scan cycle a
largest correlation result for each of a number L of groups
each of a number J of the N possible offsets, where L = N/J,
the correlation results being accumulated for only the largest
correlation result for each of said L groups. In each of these
last two embodiments, an identity of each respective offset,
for example a position count, can be stored in association with
the respective accumulated correlation result.
A further aspect f the invention provides a method of
detecting the presence and PN code phase of a desired signal in
a received signal of a CDMA communications system, comprising
the steps of: producing a reference signal with different ones
of N possible PN code phases in successive ones of N
correlation intervals in a scan cycle; correlating the received
signal with the reference signal during said correlation
intervals to produce respective correlation results;
accumulating at least some of the correlation results over
CA 02394510 2002-06-14
WO 01/45288 PCT/RU99/00494
6
successive scan cycles; and determining presence and code phase
of a desired signal from the accumulated correlation results.
The correlation results can be accumulated for all N
correlation intervals in each scan cycle. Alternatively, the
correlation results can be accumulated for only L of the
correlation intervals having greatest correlations in a scan
cycle, where L is an integer less than N, the method including
the steps of determining said greatest correlations and storing
an identity of each of said L correlation intervals in
association with the respective accumulated correlation
results. As a further alternative, the correlation intervals
can comprise L groups each of J correlation intervals in each
scan cycle, where L and J are integers and L = N/J, and
correlation results can be accumulated for only one of the
correlation intervals in each group of J correlation intervals
providing a greatest correlation in a scan cycle, the method
further comprising the steps of determining said greatest
correlations and storing an identity, of each correlation
interval providing said greatest correlation out of the
respective group of J correlation intervals, in association
with the respective accumulated correlation results.
The invention also provides a signal searcher for a
CDMA (code division multiple access) communications system,
comprising: a control unit; a reference signal generator
controlled by the control unit for generating a reference
signal with different ones of N code phases in respective ones
of N successive correlation intervals in a scan cycle; a
correlator for correlating a received signal with the reference
signal in the successive correlation intervals to produce
respective correlation results, each correlation interval being
sufficiently short that a phase of the received signal does not
CA 02394510 2002-06-14
WO 01/45288 PCT/RU99/00494
7
change appreciably during the correlation interval; and an
accumulator responsive to the control unit for accumulating the
correlation results from the correlator for each of a plurality
of corresponding correlation intervals in a plurality of scan
cycles to produce respective accumulated correlation results
from which the presence and code phase of a desired signal in
the received signal can be determined.
In a first embodiment of the searcher described
below, the accumulator comprises a buffer for storing an
accumulated correlation result for each of said N code phases.
In a second embodiment, the signal searcher comprises a unit
for determining in a scan cycle largest correlation results for
L of the N code phases, and the accumulator comprises a buffer
for storing an accumulated correlation result for each of said
L code phases and an associated count identifying the
respective code phase, where L is an integer less than N.
In a third embodiment, the correlation intervals
comprise L groups each of J correlation intervals in each scan
cycle, where L and J are integers and L = N/J, the signal
searcher further comprising a detector for determining a
greatest correlation result for each group of J correlation
intervals in a scan cycle and for providing a count identifying
a corresponding correlation interval in the respective group,
wherein the accumulator comprises a buffer for storing the
correlation result and the count associated therewith for each
of the L groups, and a combiner for increasing the stored
correlation result for each of the L groups in at least one
subsequent scan cycle by the correlation result for the same
code phase identified by said count. In this case the combiner
can be arranged to increase the stored correlation result in
the respective subsequent scan cycle only if the detector
CA 02394510 2002-06-14
WO 01/45288 PCT/RU99/00494
8
determines that the correlation result for the same code phase
identified by said count is a greatest correlation result for
the respective group of J correlation intervals.
The signal searcher can include a decision unit for
determining a maximum one or more of the accumulated
correlation results to determine the presence and code phase of
one or more desired signals in the received signal.
Embodiments of the invention can provide significant
advantages compared with the prior art discussed above. In
particular, they are simple to implement and can provide
substantial improvements in missed and false signal detection
probability, and can facilitate the detection of multiple
desired signals such as pilot signals from a plurality of base
stations in a CDMA cellular communications system, due to their
relative immunity to fading environments. The second and third
embodiments described below also have reduced memory
requirements, which is of particular importance for cases where
the number N of code phases is very large.
Brief Description of the Drawings
The invention will be further understood from the
following description by way of example with reference to the
accompanying drawings, in which similar elements in different
figures are denoted by the same reference numerals, and in
which:
Fig.l is a block diagram of a known signal searcher;
Fig. 2 is a block diagram of a known correlator used
in the signal searcher of Fig. 1;
Fig. 3 is a time diagram illustrating signal fading;
CA 02394510 2002-06-14
WO 01/45288 PCT/RU99/00494
9
Fig. 4 is a time diagram illustrating operation of
the signal searcher of Fig. 1;
Fig. 5 is a block diagram of a signal searcher in
accordance with a first embodiment of the invention;
Fig. 6 schematically illustrates a serial buffer of
the signal searcher of Fig. 5;
Fig. 7 is a time diagram illustrating operation of
the signal searcher of Fig. 5;
Fig. 8 is a block diagram of a signal searcher in
accordance with a second embodiment of the invention;
Fig. 9 schematically illustrates a buffer memory of
the signal searcher of Fig. 8;
Fig. 10 is a flow chart representing a data updating
algorithm;
Figs. 11 and 12 are diagrams illustrating data
updating in operation of the signal searcher of Fig. 8;
Fig. 13 is a graph comparing the performance of the
signal searcher of Fig. 8 with that of Fig. 1;
Fig. 14 is a block diagram of a signal searcher in
accordance with a third embodiment of the invention;
Fig. 15 schematically illustrates a serial buffer of
the signal searcher of Fig. 14;
Fig. 16 schematically illustrates a maximum detector
of the signal searcher of Fig. 14;
CA 02394510 2002-06-14
WO 01/45288 PCT/RU99/00494
Fig. 17 schematically illustrates a comparison and
combining unit of the signal searcher of Fig. 14; and
Fig. 18 is a flow chart illustrating operations of
the signal searcher of Fig. 14.
5 Detailed Description
Referring to the drawings, Fig. 1 illustrates a known
wideband signal searcher for example for detecting signals in a
CDMA cellular communications system. The signal searcher
comprises a reference pseudo-noise (PN) signal generator 10, a
10 correlator 12, an accumulator 14, a decision unit 16, and a
timing control unit 18.
The signal searcher serves to detect a desired RF
signal supplied to an input 20 of the correlator 12, the RF
signal having a nominal carrier frequency f0 and comprising a
sum of two components I and Q in phase quadrature, these
components being modulated by spectrum spreading PN sequences
or PN codes PNI and PNQ respectively.
Typically, and as is assumed in the description
below, a PN code phase is assumed to be a timing parameter with
respect to which the signal search is performed. However, it
can be appreciated that additional parameters, such as the
input signal frequency f, may be involved. In any event, for
carrying out a signal search a number of N offsets, positions,
or states (referred to balance simply at positions) are
provided each corresponding to a respective set of parameters
within a search range or uncertainty area; in the case of a PN
code phase search a respective PN code phase value corresponds
to each position, and the distance between PN code phases of
adjacent ones of the N positions is not more than one
elementary symbol, or chip, of the communications system. The
CA 02394510 2002-06-14
WO 01/45288 PCT/RU99/00494
11
reference signal generator 10 produces and supplies to
correlator 12 the PN code sequences PNI and PNQ and a reference
frequency signal cos 2~fOt, where t represents time in
accordance with timing pulses supplied from the timing control
unit 18.
Referring to Fig. 2, the correlator 12 comprises
multipliers 22, 24, 26, 28, 30, and 32, a quadrature phase
shifter 34, low pass filters (LPFs) 36 and 38, an inverter 40,
accumulating combiners 42 and 44, squaring units 46 and 48, and
a combiner 50. The reference frequency signal cos 20fOt is
multiplied by the signal from the input 20 in the multiplier
22, and is phase shifted by the phase shifter 34 and the result
multiplied by the signal from the input 20 in the multiplier
24, and the outputs of these multipliers are low pass filtered
by the LPFs 36 and 38 respectively to produce I and Q
demodulation channel signals respectively. These signals are
supplied respectively to a first input of the multipliers 26
and 38 and to a first input of the multipliers 28 and 32; a
second input of the multiplier 26 is supplied with the PN code
sequence PNI, a second input of each of the multipliers 28 and
is supplied with the PN code sequence PNQ, and a second
input of the multiplier 32 is supplied with the PN code
sequence PNQ after inversion by the inverter 40.
The outputs of the multipliers 26 and 28 are combined
25 and accumulated by the accumulating combiner 42, and the
outputs of the multipliers 26 and 28 are combined and
accumulated by the accumulating combiner 44, in each case over
a signal coherence interval TCOH which is described below, to
produce correlation values YI and YQ respectively. These
30 correlation values are squared by the squaring units 46 and 48
respectively, and the squared values are added by the combiner
CA 02394510 2002-06-14
WO 01/45288 PCT/RU99/00494
12
50 to produce a correlation estimate Y (Y = YI2 + YQ2 ) for the
signal coherence interval TCOH.
The correlation estimates Y for a number m of
consecutive coherence intervals are accumulated in the (non-
coherent) accumulator 14 (Fig. 1), to produce an accumulated
value Z which is supplied to the decision unit 16 to produce
(for example by comparison with a threshold) at an output of
the signal searcher a decision result as to whether or not a
desired signal has been detected. If no desired signal has
been detected, then via the timing control unit 18 the
parameters of the reference signal generator 10 are changed to
a next one of the positions, and the search process described
above is repeated. This process continues for each of the N
positions in turn to produce successive values Z1 to ZN, and is
subsequently repeated cyclically, until a desired signal has
been detected.
Fig. 3 illustrates by a curve 52 the amplitude of a
desired signal in a fading environment as a function of time,
represented by successive signal coherence intervals TCOH which
are small in relation to the fading rate. To the same time
scale, Fig. 4 illustrates time intervals for determining the
values Zl, Z2, D ZN, in N successive accumulation periods each
of duration TA corresponding to m consecutive signal coherence
intervals TCOH, for a total scan cycle of duration NTA. Fig. 4
illustrates this for the case of m = 6.
A disadvantage of the known signal searcher and
method described above is that the value Z for each of the N
positions is determined over m consecutive signal coherence
intervals. In a fading environment such as for a CDMA cellular
communications system, signal fading can seriously adversely
affect determinations within such consecutive intervals, so
CA 02394510 2002-06-14
WO 01/45288 PCT/RU99/00494
13
that substantial random variations of the determined values Z
(relative to a long term average for each respective position)
are possible, with a resulting increased probability of missed
signal detection and false signal detection. To reduce this,
it becomes necessary to increase the value m, resulting in a
proportional increase in the signal search time and scan cycle
duration. This in turn results, in a CDMA cellular
communications system, in increased intra-system interference,
increased system access time and reduced system capacity.
A signal searcher in accordance with a first
embodiment of the invention is illustrated in Fig. 5, and
comprises a reference signal generator 10 and a correlator 12
which can be as described above, a recirculating accumulator 60
comprising a combiner 62 and a serial buffer 64, a decision
unit 66, and a timing control unit 68. In contrast to the
timing control unit 18 of Fig. 1 which produces an output pulse
to change the parameters of the reference signal generator 10
once every m intervals TCOH, the timing control unit 68
produces an output pulse for each interval TCOH to change the
parameters of the reference signal generator after every such
signal coherence interval, so that a scan cycle through all of
the N positions or states is completed every N intervals TCOH.
The output pulses of the timing control unit 68 are also
supplied to a clock input C of the serial buffer 64.
The correlation estimate Y produced at the output of
the correlator 12 for each signal coherence interval TCOH is
supplied to one input, and an output ZOUT of the serial buffer
64 is supplied to a second input, of the combiner 62, which
adds these to produce at its output a result ZIN which it
supplies to an input of the serial buffer 64. Fig. 6
illustrates the serial buffer 64, which comprises N buffer
CA 02394510 2002-06-14
WO 01/45288 PCT/RU99/00494
14
stages 701 to 70N each for storing a respective correlation
sum, connected in series with one another in the manner of a
shift register and all clocked by pulses supplied to the clock
input C, at the periodicity of the intervals TCOH. An output
of the final buffer stage 70N constitutes the output ZOUT of
the serial buffer 64.
Fig. 7, which can be compared with Fig. 4 described
above, illustrates the resulting operation of the signal
searcher of Fig. 5. In each of m scan cycles, each of duration
NTCOH and comprising N signal coherence intervals, the
reference signal generator 10 is changed through all of its N
positions and the contents of the stages 70 of the serial
buffer 64 are shifted cyclically through the buffer, in each
case being supplemented by having the current correlation
estimate Y, for the respective reference signal generator
position, added by the combiner 62. The contents of the buffer
stages 70 of the serial buffer 64 can be initially set to zero,
and after m scan cycles are accumulated for m signal coherence
intervals. However, as shown in Fig. 7 this accumulation to
determine a corresponding value Zi (i = 1 to N) takes place
over non-adjacent signal coherence intervals, so that the
effects of signal fading as shown in Fig. 3 are substantially
reduced by averaging. These accumulated correlation values Zi
are conveniently supplied sequentially from the serial buffer
64 as its output ZOUT, a maximum one, or maximum ones, of these
values being determined in the decision unit 66 and compared
with a threshold to detect the presence and offset (code phase)
of one or more desired input signals.
It can be appreciated that instead of being supplied
serially from the buffer 64 to the decision unit 66, the values
Zi can (for suitable values of N) be supplied in parallel or in
CA 02394510 2002-06-14
WO 01/45288 PCT/RU99/00494
a combined serial and parallel manner. Furthermore, it can be
appreciated that instead of the search cycle being carried out
in discrete groups of m scan cycles each of N signal coherence
intervals, the values Zi can be updated iteratively in an
5 ongoing manner with each scan cycle of N signal coherence
intervals. In this case the signal ZOUT fed back to the
combiner 62 can be reduced by weighting in known manner. In
any event, it can be appreciated that the averaging of the
accumulated correlation values Zi provided by the searcher of
10 Fig. 5 reduces the disadvantages due to signal fading discussed
above.
A signal searcher in accordance with a second
embodiment of the invention is illustrated in Fig. 8, and
comprises the reference signal generator 10, correlator 12,
15 combiner 62, and timing control unit 68 which can be as
described above, a counter 72, a buffer memory 74, a data
updating unit 76, and a decision unit 78. As described above,
the timing control unit 68 produces an output pulse for each
interval TCOH to change the parameters of the reference signal
generator after every such signal coherence interval; so that a
scan cycle through all of the N positions is completed every N
intervals TCOH. The output pulses of the timing control unit
68 are also supplied to a clock input C of the data updating
unit 76, and to an input of the counter 72, which is a modulo-N
counter which counts these pulses to provide via its output,
connected to the data updating unit 76, a position count i
representing the current position in a scan cycle.
The buffer memory 74 is shown in greater detail in
Fig. 9. In contrast to the serial buffer 64 having N stages as
described above, the buffer memory 74 contains a typically much
smaller number L of stages 801 to 80L, each of which has two
CA 02394510 2002-06-14
WO 01/45288 PCT/RU99/00494
16
memory fields, for a respective one of L correlation values Z1
to ZL and an associated position count i1 to iL. The
respective memory fields of the stages 80 are connected in
series with one another between inputs for respective signals
ZIN and iIN and outputs for respective signals ZOUT and TOUT,
but are individually clocked by clock signals Cl to CL supplied
from the data updating unit 76, and supply the respective
signals Z1 to ZL to the data updating unit 76. The data
updating unit 76 is also supplied with the signals ZOUT and
TOUT from the buffer memory 74, and produces on a line 82 a
signal which is either zero or the correlation value ZOUT as
described below. The signal on the line 82 is supplied to the
second input of the combiner 62, the first input of which is
supplied with the current correlation estimate Y as described
above, and the output of which constitutes the signal ZIN to
the buffer memory 74. The signal iIN to the buffer memory 74
is constituted by the current position count i from the counter
72. The current correlation estimate Y is also supplied to the
data updating unit 76.
As described below, the number L may be of the order
of one-fifth or one-tenth the number of positions N, so that
the capacity of the buffer memory 74 can be much smaller than
that of the serial buffer 64. This is especially significant
for large values of N.
The operation of the signal searcher of Fig. 8 is
described below with additional reference to the flow chart of
Fig. 10, comprising steps 81 to 87, and the data updating
diagrams of Figs. 11 and 12, in each of which upper and lower
parts indicate contents of the buffer memory stages
respectively before and after the data updating.
CA 02394510 2002-06-14
WO 01/45288 PCT/RU99/00494
17
In a step 81, the data updating unit 76 determines a
minimum (but non-zero) one Zn of the correlation values
currently stored in the buffer memory 74, and determines the
position in of this. If there is more than one equal minimum
value, the value and position of any of the minima is used. It
can be appreciated that this determination can be carried out
during the interval TCOH for which the current correlation
estimate Y is being produced, because it only involves stored
data.
In a subsequent step 82, the unit 76 determines
whether the current position count i is equal to the position
TOUT (= iL) from .the output of the buffer memory 74. If it is
not, as will be the case during a first scan cycle through the
N positions, in a step 83 the unit 76 determines whether the
current correlation estimate Y is less than or equal to the
determined minimum value Zn stored in the buffer memory 74. If
so, there is no updating of data for the current position count
i and an exit from the flow chart of Fig. 10 is reached.
If in the step 83 it is determined that Y > Zn, then
steps 84 and 85 are executed to update data in the buffer
memory 74 in the manner represented in Fig. 11, for which it is
assumed for example that n = 4, i.e. that Z4 is the minimum
value determined in the step 81. In this case the unit 76
produces a zero signal on the line 82, so that as indicated by
step 84 the combiner 62 outputs the current correlation
estimate Y as the correlation value ZIN. As shown by step 85,
this value and its position count iIN = i are stored in the
first stage 801 of the buffer memory, and the contents of the
buffer memory stages 801 to 80n-1 are shifted into the stages
802 to 80n respectively by clock pulses C1 to Cn. The previous
contents of the buffer memory stage 80n = 804 are overwritten,
CA 02394510 2002-06-14
WO 01/45288 PCT/RU99/00494
18
and the contents of the buffer memory stages 80n+1 to 80L are
unchanged because clock pulses Cn+1 to CL are not produced. In
this manner, in the first scan cycle of N signal coherence
intervals TCOH the L largest correlation estimates Y and their
respective positions i are stored in the L stages of the buffer
memory 74.
If it is determined at the step 82 that the current
position count i is equal to the position TOUT (= iL) from the
output of the buffer memory 74, then steps 86 and 87 are
executed to update data in the buffer memory 74 in the manner
represented in Fig. 12. In this case the unit 76 produces the
correlation value ZOUT on the line 82, so that as indicated by
step 86 the combiner 62 outputs the sum of the current
correlation estimate Y and the correlation value ZOUT as the
correlation value ZIN. As shown by the step 87, all of the
stages of the buffer memory are then updated by shifting by the
production of clock pulses C1 to CL. In this case the
arrangement operates substantially as a recirculating
accumulator for the correlation values Z1 to ZL and their
positions i1 to iL stored in the buffer memory 74.
After the desired number m of scan cycles, during
each of which the parameters of the reference signal generator
10 are changed after every interval TCOH to provide the same
advantages with respect to fading as described above for the
first embodiment, the buffer memory 74 contains accumulated
substantially maximum correlation values for L positions, and
the locations i of these positions, out of the total N
positions. These accumulated correlation values Z1 to ZL and
their positions i1 to iL are then supplied serially from the
buffer memory 74 as shown in Fig. 8, or in parallel or series-
parallel in a similar manner to that discussed above, to the
CA 02394510 2002-06-14
WO 01/45288 PCT/RU99/00494
19
decision unit 78. The decision unit 78 determines a maximum
one, or maximum ones, of these values for example by comparison
with a threshold to detect the presence and offset (code phase)
of one or more desired input signals, and supplies the
resulting correlation values (if desired) and position
information i to its output.
As indicated above, the fact that successive
correlation estimates which are accumulated and stored for
respective positions are separated in each case by N signal
coherence intervals TCOH reduces the effects of fading on the
operation of the signal searcher of Fig. 8. However, a missed
detection probability is increased because the first scan cycle
is used to determine the positions for which correlation
estimates are accumulated, this being dependent upon the ratio
between the number of positions N and the buffer memory
capacity L. Fig. 13 illustrates computer simulation results in
this respect, showing a missed detection probability as a
function of m for a Rayleigh fading channel with N = 675. An
upper curve 88 represents performance of the known signal
searcher of Fig. 1, and three lower curves 90 represent the
substantially improved performance for the signal searcher of
Fig. 8 for different ratios of L to N. As can be appreciated
from these curves, the signal searcher of Fig. 8 provides a
substantial improvement in missed signal detection probability,
which is not significantly degraded for values of L above about
N/10, so that there can also be a substantial reduction in
buffer memory capacity.
A signal searcher in accordance with a third
embodiment of the invention is illustrated in Fig. 14, and
comprises a reference signal generator 10, correlator 12,
timing control unit 68, position counter 72, and decision unit
CA 02394510 2002-06-14
WO 01/45288 PCT/RU99/00494
substantially as described above, except that the position
counter 72 in this case is a modulo-J counter where J is an
integer as described below, and the position determination of
the decision unit 78 is modified as described below. As
5 described above, the timing control unit 68 produces an output
pulse for each interval TCOH to change the parameters of the
reference signal generator 10 after every such signal coherence
interval. The position counter 72 counts these pulses on a
clock pulse line C and, after every J pulses, produces an
10 output pulse on a clock pulse line Cl. The pulses on this line
C1 could alternatively be provided directly from the timing
control unit 68.
The signal searcher of Fig. 14 also comprises a
maximum detector 92 and a recirculating accumulator 94
15 comprising a comparison and combining unit 94 and a serial
buffer 96. As shown in Fig. 15, the serial buffer 98 comprises
L stages 1001 to 100L, each of which has two memory fields, for
a respective one of L correlation values Z1 to ZL and an
associated position count j1 to jL. The respective memory
20 fields of the stages 100 are connected in series with one
another between inputs for respective signals ZIN and jIN and
outputs for respective signals ZOUT and jOUT, and are all
commonly clocked by clock signals on the clock pulse line C1.
The signal searcher of Fig. 14 is particularly
advantageous when N is very large, for example N = 32768, and
for searching in cases where a plurality of signals may need to
be detected, for example for detecting pilot signals of several
base stations. In the latter case the PN code phases of
different pilot signals are separated by J of the N positions;
for example J = 64 in the case of PN code chips for the forward
pilot channel of an IS-95 cellular communications system. The
CA 02394510 2002-06-14
WO 01/45288 PCT/RU99/00494
21
integers J, L, and N are related by the equation J = N / L, the
value of N being increased if necessary to ensure that J and L
are integers.
In the signal searcher of Fig. 14, the maximum
detector 92 is supplied with the correlation estimate Y from
the correlator 12 for each signal coherence interval TCOH, the
clock pulses C from the timing control unit 68, and a position
count j from the counter 72, and determines a maximum one YMAX
of the correlation estimates Y within each subset of J of the N
positions being searched, providing at its outputs this maximum
correlation estimate YMAX and a value jMAX of the position
count j for this maximum.
To this end, as shown in Fig. 16 the maximum detector
comprises a comparator 102, a memory cell 104 for storing a
correlation estimate Y, and a memory cell 106 for storing an
associated position count j. It should be noted that the
memory cell 106 is required to store numbers only up to J,
requiring relatively few bits compared with the number of bits
which would be required for storing position counts up to N
(e. g. 6 bits for J = 64 compared with 15 bits for N = 32768).
The same applies to the number fields j1 to jL of the L stages
100 of the serial buffer 98, resulting in a substantial
reduction in the memory capacity required for storing position
counts.
The comparator 102 has inputs coupled to the input
and output of the memory cell 104, so that it compares each
input correlation estimate Y with the current contents YMAX of
the memory cell 104. In the event that Y > YMAX, the
comparator 102 supplies an output to enable inputs E of the
memory cells 104 and 106, to write inputs W of which the clock
pulses C are supplied so that the larger correlation estimate Y
CA 02394510 2002-06-14
WO 01/45288 PCT/RU99/00494
22
is stored in the memory cell 104 and its position count j is
stored in the memory cell 106. After J signal coherence
intervals TCOH defining the respective position subset, the
stored maximum YMAX and its position jMAX are supplied to the
unit 96 of the recirculating accumulator 94, if necessary via
clocked buffers which are not shown.
The signals ZOUT and jOUT from the output of the
serial buffer 98 and the signals YMAX and jMAX from the maximum
detector 92 are supplied to the unit 96, one form of which is
illustrated in Fig. 17. As shown in Fig. 17, the unit 96
comprises comparators 108 and 110, a switch 112, a combiner
114, and a selector 116 having outputs for the signals ZIN and
jIN to the inputs of the serial buffer 98. Fig. 18 shows a flow
chart of the operation of the unit 96, comprising steps 121 to
125 which are referred to below.
The comparator 108 is supplied with and compares the
position counts jMAX and jOUT and, if they are equal as
determined at step 121, closes the switch 112 to supply the
correlation estimate YMAX to one input of the combiner 114, to
another input of which the correlation value ZOUT is supplied.
The combiner 114 sums its inputs to produce an output Z1. The
selector 116 supplies Z1 or YMAX as the signal ZIN, and jOUT or
jMAX as the signal jIN, under the control of the comparator 110
which compares YMAX and ZIN. The selector 116 has the switch
positions shown in Fig. 17 except when the comparator 116
determines that YMAX > Zl. Thus in this case, as shown by
block 122 in Fig. 18, the selector 116 has the switch positions
shown to provide outputs ZIN = Z1 = ZOUT + YMAX and jIN = jOUT
(which is also equal to jMAX). Consequently, when the maximum
correlation estimate YMAX occurs with the same position count
jMAX in the same position subset in repeated scan cycles, these
CA 02394510 2002-06-14
WO 01/45288 PCT/RU99/00494
23
correlation estimates are accumulated in a similar manner to
that described above.
If in the step 121 it is determined that jMAX is not
equal to jOUT, then the switch 112 remains open and the
combiner 114 produces the signal Zl = ZOUT. In this case the
comparator 110 determines at the step 123 whether YMAX > Z1
and, if so, controls the selector 116 to adopt its alternative
switch position in which it passes the signals YMAX and jMAX to
its outputs ZIN and jIN respectively, as shown by the step 124.
Otherwise, the comparator 112 controls the selector 116 to
adopt the switch position shown in which it passes the signals
Z1 = ZOUT and jOUT to its outputs ZIN and jIN respectively, as
shown by the step 125 in Fig. 18.
After m scan cycles each of N signal coherence
intervals TCOH as described above, the contents of the serial
buffer 98 are supplied serially to the decision unit 78, and
one or more maximum values ZI, where I is an integer from 1 to
L denoting a respective one of the position subsets and
corresponding buffer stages 100, are determined and compared
with a threshold to determine the presence or absence of one or
more signals to be detected. The code phase or position of
each such signal is J(I - 1) + jI, where jI is the position
count or value jOUT associated with the determined maximum.
Although as described above each correlation result
stored in the serial buffer 98 is only increased when the
respective position count jMAX is the same as the stored
position count jOUT, this need not necessarily be the case.
Instead, in the first signal scan cycle the maximum correlation
result YMAX and its position jMAX could be determined by the
maximum detector 96 as described above, and in subsequent ones
of the m signal scan cycles the correlation result Y for the
CA 02394510 2002-06-14
WO 01/45288 PCT/RU99/00494
24
respective position count jMAX in the respective group could be
accumulated regardless of whether or not it is the maximum
correlation result for that group in the respective scan cycle.
However, it is expected that significant performance
degradation will occur to the detection probability since the
decision for the maximum position in each subset is based on
the first scan cycle only and no accumulation is utilized.
Although embodiments of the invention have been
described above in terms of specific physical devices such as
counters and comparators, it can be appreciated that these may
be replaced, and the embodiments of the invention more easily
implemented, by one or more digital signals processors or
application-specific integrated circuits.
It will also be appreciated that numerous other
changes, variations, and modifications may be made to the
specific embodiments of the invention described by way of
example above without departing from the scope of the claims.
