 |
Description  |
|
|
CROSS-REFERENCE TO A RELATED PATENT APPLICATION
This patent application is related to commonly assigned U.S. patent
application Ser. No. 08/606,285, filed Feb. 23, 1996, entitled "A
MULTI-USER ACQUISITION PROCEDURE FOR MULTIPOINT-TO-POINT SYNCHRONOUS CDMA
SYSTEMS", by S. Kingston et al. (Attorney Docket No. DUT 513).
FIELD OF THE INVENTION
This invention pertains generally to code division, multiple access (CDMA)
communication systems and, in particular, to direct-sequence (DS)
point-to-multipoint synchronous CDMA communications systems.
BACKGROUND OF THE INVENTION
In a CDMA communications system a plurality of user communication signals
can be transmitted within, i.e., share, a same portion of the frequency
spectrum. This is accomplished by providing a plurality of different
pseudonoise (pn) binary code sequences (e.g., one for each user) that
modulate a carrier, thereby "spreading" the spectrum of the resulting
waveform. In a given receiver all of the user signals are received, and
one is selected by applying an assigned one of the pn binary code
sequences to a correlator to extract only the signal energy intended for
the receiver, thereby "despreading" the received CDMA transmission. All
other (uncorrelated) user transmissions appear as noise.
One type of CDMA communication system is specified by a document referred
to as EIA/TIA/IS-95. The system as specified uses a plurality of base
stations that establish and maintain bidirectional direct-sequence (DS)
CDMA links with a plurality of mobile stations (e.g., cellular
telephones). One feature of the IS-95 system is the presence of a pilot
channel that is transmitted by each base station.
The pilot channel is an unmodulated, direct-sequence spread spectrum signal
that is transmitted continuously by each CDMA base station. The pilot
channel allows a mobile station to acquire the timing of the Forward CDMA
channel (i.e., from the base station to the mobile station), provides a
phase reference for coherent demodulation, and provides a reference for
signal strength comparisons between base stations for determining when to
handoff. The pilot pn sequence is defined as a pair of modified maximal
length PN sequences with period 2.sup.15 that are used to spread the
Forward CDMA channel and the Reverse CDMA channel. Different base stations
are identified by different pilot PN sequence offsets. A pilot pn sequence
offset index is defined to be in units of 64 pn chips, relative to a zero
offset pilot pn sequence. A pn chip is defined as one bit in the pn
sequence. The pilot strength is defined as the ratio of received pilot
energy to overall received energy.
Walsh functions are a class of 2.sup.N time orthogonal binary functions
that are used to establish orthogonality between the different pn binary
code sequences used by the pilot and user channels.
The use of the pilot channels, while providing certain advantages in a CDMA
system intended for use with mobile stations, may present disadvantages as
well, particularly in systems where the user transceivers are fixed as
opposed to mobile. For example, the pilot channels consume some amount of
the available pn code sequences and signal energy, all of which could be
otherwise allocated to the users of the system.
Also, in many detection approaches synchronization to the pn code timing
must be achieved before carrier phase-lock can occur. In this case a
non-coherent detection algorithm must be employed. Generally, non-coherent
detectors rely on energy detection within a fixed bandwidth, as a range of
code timing cells are searched. Upon locating the correct code timing, the
detector energy level rises above a predetermined threshold level. A
bit-sync loop then takes over to obtain the finer-resolution bit timing.
However, standard acquisition approaches are known to fail when the number
of users becomes large. This is due to the fact that the noise power
becomes comparable to the signal power when the user of a synchronous type
of CDMA system is not synchronized. As a result, it becomes very difficult
for the user's receiver to distinguish the correct pn timing phase from
the incorrect phases resulting from the increased noise.
As can be appreciated, the acquisition technique is an important aspect of
the receiver, in that its operation impacts the overall speed at which
synchronization to the forward link occurs. If the time required to
synchronize the user's receiver becomes excessive, the delay may be
considered as objectional by the user.
OBJECTS OF THE INVENTION
It is a first object of this invention to provide methods and apparatus to
enable a receiver to synchronize to a synchronous CDMA communications
system.
It is a further object of this invention to provide methods and apparatus
for implementing a synchronous CDMA system wherein a first forward channel
transmits a null (inactive) pn code sequence that is orthogonal to all
active codes, a second forward channel transmits an always-active pn code
sequence, and wherein a receiver uses either the first forward channel,
the second forward channel, or both of the forward channels to synchronize
to the forward CDMA link.
SUMMARY OF THE INVENTION
The foregoing and other problems are overcome and the objects of the
invention are realized by methods and apparatus in accordance with
embodiments of this invention, wherein a user terminal is provided with
circuitry and methods enabling the acquisition of a desired CDMA channel
in the presence of a plurality of synchronous interfering channels.
In one aspect, this invention teaches a method for synchronizing to a
forward channel in a CDMA system. The method includes steps of (a)
despreading a received CDMA signal with a first pn code that is known not
to be present in the received CDMA signal, (b) obtaining a measure of
received signal level, and (c) using the obtained measure of the received
signal level when setting a phase of a second pn code that corresponds to
a desired forward channel that is to be received. In this method the step
of despreading despreads a null channel, and the desired forward channel
is a continuously transmitted side-channel that provides system-level
information to all subscriber units or terminals.
In another aspect this invention teaches a method for acquiring a forward
channel in a point-to-multipoint CDMA system. This method includes the
steps of (a) despreading a received CDMA signal with a first pn code that
is known to be present in the received CDMA signal and obtaining a first
measure of received signal level; (b) despreading the received CDMA signal
with a second pn code that is known not to be present in the received CDMA
signal and obtaining a second measure of received signal level; and (c)
synchronizing to a desired channel using a difference between the first
and second signal levels.
In this method the first step of despreading despreads a continuously
transmitted side-channel, and the step of synchronizing synchronizes to
the side-channel. Also, the step of obtaining a first measure of received
signal level obtains a correlation peak, and the step of obtaining a
second measure of received signal level obtains a correlation null.
Preferably, the steps of despreading and obtaining are accomplished
iteratively over a range of n first pn code phase states and over a range
of n second pn code phase states. For a pn code phase state i of the
plurality n of first and second pn code states, the method further
determines a difference value between the first measure of received signal
level and the second measure of received signal level; compares the
difference value to a threshold value; and, if the difference value is
greater than the threshold value, sets a pn code generator to output a pn
code corresponding to the desired channel, wherein the outputted pn code
is set at the pn phase state i. Else, if the difference value is not
greater than the threshold value, the method instead increments the pn
phase state i and re-executes the steps of despreading and obtaining.
In this method the step of determining the difference value includes a step
of storing the determined difference value, and if none of the n
difference values are greater than the threshold value, the method further
includes a step of examining the stored difference values to select a
stored difference value that has a largest value; and a step of setting
the pn code generator to output the pn code corresponding to the desired
channel, wherein the outputted pn code is set to a pn phase state that
corresponds to the selected stored difference value.
In a further aspect, this invention teaches a synchronous CDMA
communication system that operates in accordance with the foregoing
methods.
BRIEF DESCRIPTION OF THE DRAWINGS
The above set forth and other features of the invention are made more
apparent in the ensuing Detailed Description of the Invention when read in
conjunction with the attached Drawings, wherein:
FIG. 1 is a simplified block diagram of a synchronous, DS-CDMA
communications system that is constructed and operated in accordance with
this invention, the system having a radio base unit (RBU) and a plurality
of subscriber units (SUs).
FIG. 2 is a block diagram of a first embodiment of the SU receiver detector
of FIG. 1, in particular a non-coherent square law detector.
FIG. 3 is a block diagram of a second embodiment of the SU receiver
detector of FIG. 1, in particular a non-coherent absolute value detector.
FIG. 4 is a graph that illustrates the relative mean acquisition time
performance of the single-user, multi-user, and difference tests of this
invention, for the case E.sub.s /N.sub.o =6 dB, P.sub.d =0.995, and
.alpha.-0.01, and for a range of numbers of active users.
FIG. 5 is a graph that illustrates an average multi-user interference power
and average side-channel correlation power versus code timing offset, with
3 pole Butterworth transmitter and receiver filters, and 30 active users.
FIG. 6 is a graph that illustrates energy versus offset for a desired
subscriber unit, for multi-user interference, and for background noise.
FIGS. 7A-7D are graphs that illustrate energy versus offset curves for the
outputs of matched filters that are matched to the side-channel PN code
(dashed line) and a null-code (solid line) for the cases of high and low
SNR, as well as heavy and light loading. In these Figures the symbol
X=matched filter output for the side-channel code, and the symbol
.quadrature.=matched filter output for the null-code.
FIG. 8 is a logic flow diagram of a channel acquisition method in
accordance with this invention.
DETAILED DESCRIPTION OF THE INVENTION
Referring to FIG. 1, a synchronous CDMA communications system 10, which in
presently preferred embodiments of this invention is embodied as a fixed
wireless system (FWL), is considered herein to be a CDMA system wherein
forward link (FL) transmissions from a radio base unit (RBU) 12 for a
plurality of user or subscriber units (SUs) 14 are bit and chip aligned in
time, and wherein the SUs 14 operate in accordance with the teaching of
this invention for receiving the FL transmissions and for synchronizing to
one of the transmissions. The FWL is suitable for use in implementing a
telecommunications system that conveys voice and/or data between the RBU
12 and the SUs 14.
The RBU 12 includes circuitry for generating a plurality of user signals
(USER.sub.-- 1 to USER.sub.-- n), a side channel (SIDE.sub.-- CHAN) signal
that is continuously transmitted, and a NULL signal. Each of these signals
is assigned a respective pn spreading code and is modulated therewith
before being applied to a transmitter 12a having an antenna 12b. When
transmitted on the FL the transmissions are modulated in phase quadrature,
and the SUs 14 are assumed to include suitable phase demodulators for
deriving in-phase (I) and quadrature (Q) components therefrom. The
illustrated arrangement is for one frequency (carrier) channel, it being
realized that the RBU 12 is capable of transmitting a plurality of such
frequency channels. By example, each frequency channel includes up to 31
code channels, and has a center frequency in the range of 2 GHz to 3 GHz.
Each SU 14 includes an antenna 14a, a mixer 14b for down-converting the
received signal, a correlator 14c wherein the user's transmission is
obtained by despreading the received signal with a local pn code, and a
detector and correlator 14d. Suitable embodiments for the detector are a
non-coherent square law detector shown in FIG. 2 and also a non-coherent
absolute value-detector shown in FIG. 3. The SU 14 also includes a local
processor 14e that is responsible for managing the operation of the SU 14.
These management functions include generating a variable local oscillator
(LO) signal, such as is obtained from a voltage controlled oscillator
(VCO) 14f, and providing the pn binary code sequence that is assigned to
the SU 14 for despreading the user's signal. The processor 14e is also
responsible for executing one or more of the acquisition methods in
accordance with this invention. Although the SU 14 is capable of also
transmitting a DS-CDMA signal to the RBU 12 on a return link, these
functions are not germane to the teaching of this invention and are thus
not illustrated.
For the presently preferred embodiments of this invention the antennas 12b
and 14a have a line-of-sight relationship, the SUs 14 are fixed in
location with respect to the RBU 12, and the antennas 12b and 14a are
boresighted during installation of the SU 14. However, and as will be
discussed below, the teachings of this invention are not limited to only
this particular presently preferred arrangement.
The ensuing description assumes the use of a DS signal, a(t), with code
symbol duration T.sub.s, multiplied by a spreading sequence c(t), with
chip duration T.sub.c and a null-to-null bandwidth W.sub.c =2/T.sub.c. The
symbol rate for each SU 14 is fixed at 1/T.sub.s, and the chipping rate at
1/T.sub.c =P/T.sub.s. All of the pn codes are mutually orthogonal when
aligned, and are assumed to be accurately aligned during normal operation.
In the presently preferred embodiments of this invention the pn codes are
selected from a set of randomized Walsh-Hadamard codes. The teaching of
this invention is not, however, limited to only signals having these
characteristics. By example, the set of pn spreading codes can be selected
from any set that exhibits low cross-correlation at zero relative shift.
In the absence of multi-user interference, the received signal can be
written as r(t)=a(t-r)c(t-r)+n(t), where n(t) represents white gaussian
channel noise. r(t) is multiplied by successive shifts of the spreading
sequence, c(t-m.delta.t), in an attempt to estimate the timing offset, r.
The detector filter 17 of FIGS. 2 and 3 is a finite (T.sub.s) integrator
with an equivalent noise bandwidth W.sub.s =1/T.sub.s. Within this
bandwidth, the effective noise power spectral density, 1/2N.sub.o, is
unchanged by the integrator. When c(t-m.delta.t) is synchronous with r(t),
the pn spreading code is collapsed, and the integrator produces mean
values of the DS signal, a(t-r). Assuming non-synchronous operation, most
of the signal energy falls outside the bandwidth of the detector filter 17
and, to a good approximation, the detector input can be taken as white
gaussian noise only. For worst case non-synchronous operation, the signal
energy is reflected back into the spread bandwidth, W.sub.c.
The sum of n samples of a squared gaussian process is central .chi..sup.2
distributed with n degrees of freedom, and has a bandwidth (near DC) twice
that of the detector filter 17. The central .chi..sup.2 distribution with
n degrees of freedom can be written
##EQU1##
where .sigma. is the variance of the gaussian process.
Given spread code synchronization (i.e., m.delta.t.apprxeq.r), the spectrum
of the received signal is collapsed to the data bandwidth, thus increasing
the total energy within the detector filter 17. Because the phase of the
received signal rotates freely with respect to the reference, this
waveform is not DC, even for a constant energy signal. It would thus
appear that the output of the detector should again be central .chi..sup.2
distributed, but with a larger mean value than for the noise only case.
However, the phase rotation rate is assumed to be slow enough that the
signal component can be treated as deterministic, and hence the ith
independent sample of the detector filter output is distributed normally
with mean s.sub.i and variance .sigma..sup.2 determined by the noise
component. The integrator output is then characterized by a non-central
.chi..sup.2 distribution with n degrees of freedom, and a non-centrality
parameter:
##EQU2##
where S represents the average signal power. The non-central .chi..sup.2
distribution with n degrees of freedom can be written as:
##EQU3##
Finally, a summation of n samples from block 21 is compared in block 23 to
a preset threshold, t.sub.h. The threshold and the number of samples are
set to achieve a probability of detection.gtoreq.P.sub.d simultaneously
with a probability of false alarm.ltoreq..alpha.. Give t.sub.h, P.sub.d
can be determined by integrating the non-central .chi..sup.2 distribution
from t.sub.h to .infin.. The probability of false alarm, .alpha., is
determined by integrating the central .chi..sup.2 distribution from
t.sub.h to .infin.. As the integration time, T, becomes larger, the spread
between the two distributions also becomes larger. At some point, the
required detection probability can be achieved while simultaneously
achieving the target false alarm probability. The minimum number of
samples required to achieve the target detection and false alarm
probability criteria determines the observation window of the detector,
T=nT.sub.s.
However, normally some time uncertainty will exist, and some number (Q) of
time cells must be searched. The mean acquisition time is then determined
as follows: Given a H.sub.o case (a noise only case), the time required to
search a single time cell is given by the expression
T.sub.o =T+.alpha.KT,
where K is referred to as the cost for a false alarm. If it is assumed
that, upon exceeding the threshold, a verification procedure is begun, and
that this verification procedure includes a second integration of length
KT, which is long enough to assure that the probability of a second false
alarm is negligible, then T.sub.o =(1+.alpha.K)T. On average, (1/2Q-1)
time cells must be searched before encountering the correct cell. Next,
the time required to search a time cell given by a H.sub.1 case (a signal
present case), is given by the expression
T.sub.1 =P.sub.d (1+K)T+(1-P.sub.d)(1/2QT.sub.o +T.sub.acq).
The first term corresponds to a successful detection, in which case a
verification procedure is also required. The second term corresponds to a
detection failure (in which case the acquisition procedure must be started
from the beginning, but increased by 1/2QT.sub.o (since all time cells
must now be searched). The acquisition time is therefore given by the
expression
T.sub.acq =(1/2Q-1)T.sub.o +T.sub.1.
Substituting and rearranging, one arrives at the expression
##EQU4##
Hence, given the signal-to-noise ratio (SNR), the required detection
probability, P.sub.d, the false alarm probability, .alpha., and a cost
factor, one can compute the necessary integration time, T. Then, given T
and the uncertainty, Q, the mean acquisition time, T.sub.acq, can be
determined.
For a particular value of E.sub.s /N.sub.o at the input filter, the input
signal-to-noise ratio is given by
##EQU5##
Assuming that the signal is AGC controlled at the output of the SU receiver
filter 15, then S+N.sub.c =1, from which it follows that
##EQU6##
where S is the signal power, and N.sub.c the noise power at the output of
W.sub.c. In addition, N.sub.o =2N.sub.c /W.sub.c. The noise power at the
output of the detector filter 17, W.sub.s, is then N.sub.s =N.sub.o
W.sub.s /2, from which the signal-to-noise ratio in the energy detector
19a or 19b can be calculated as SNR.sub.s =S/N.sub.s. The probability
distribution functions for each hypothesis are then completely determined
by the parameters:
##EQU7##
The foregoing analysis is now generalized to a multi-user channel of most
interest to this invention. By example, there may be M.ltoreq.30 active
users (SUs 14) in the CDMA communication system 10, each of which receives
coded information symbols from the RBU 12, with an assigned length P=32
code. All of the pn codes are mutually orthogonal when aligned, and are
assumed to be accurately aligned during normal operation. The symbol rate
for each SU 14 is fixed at 1/T.sub.s, and the chipping rate at 1/T.sub.c
=P/T.sub.s. The RBU 12 transmits all active channels together
synchronously and uniformly, and hence the channel power levels and timing
offsets received at any one SU 14 are substantially equal.
In addition to the user channels, there are two additional channels which
all SUs 14 make use of. One, which is never active, is referred to as the
above-described null-channel. Although the null-channel is assigned a
unique pn code (referred to herein as the null-code), which is orthogonal
to every active code, the null-code is not actually transmitted. That is,
the null-code can be considered as a "missing" code. This is schematically
shown in FIG. 1 by the open switch (SW) placed in the NULL signal path.
The switch (SW) could also be placed in the pn.sub.-- null path, such that
the pn code assigned to the null channel (pn.sub.-- null) does not reach
the associated multiplier (spreader) 12c.
The second channel, referred to above as the side-channel, is always
active. The side channel provides side-information to the SUs 14 (e.g.,
system access information, etc.). The side-channel is used as well for
code-synchronization, as described below.
In a single-user acquisition technique or test (SUT), only a side-channel
statistic is employed, while in a multi-user acquisition technique or test
(MUT), the null-channel statistic is employed. It is also within the scope
of this invention to employ two different statistics; one corresponding to
the null-channel and the other to the side-channel, as explained below.
Single-user technique: For this measurement, the side-channel pn code is
used as the SU 14 correlation code, c(t-m.differential.t), as shown in the
non-coherent square law detector block diagram of FIG. 2 and also the
non-coherent absolute value detector block diagram of FIG. 3. The
single-user technique makes use of this statistic alone in an attempt to
locate a correlation peak between the copy of the side-channel pn code
stored in the SU 14 and the active side-channel itself. The maximum point
on this peak occurs when the codes are aligned.
The multi-user technique: For this measurement, the null-code is used as
the correlation code. The multi-user test makes use of this statistic
alone in an attempt to find a correlation null (as opposed to peak)
between the copy of the null-code stored in the SU 14 and the active
channels. The correlation null results from the mutual orthogonality of
every possible active code with the null-code, and coincides with code
alignment.
It has been found that the single-user test may be reduced in effectiveness
when M approaches the maximum number of allowed active users. This is due
to the fact that the multi-user interference level drops just as the
side-channel correlation signal rises, in a nearly self-canceling fashion.
Similarly, when M is small, it has been found that the multi-user test may
be reduced in effectiveness, because the correlation null can be masked by
ambient noise.
In order to compensate for these effects, it is preferred to use a
difference between these two statistics when acquiring the FL. Hence, for
large M, the single-user statistic dominates, and for small M, the
multi-user statistic dominates. But, in either case, the difference
between the two statistics changes significantly (i.e., becomes larger) as
the pn code timing approaches the lock point. This advantage is mitigated
somewhat by the doubling of an observation window, T=2nT.sub.s. However,
and referring now to FIG. 4, it can be readily seen that the use of the
difference test provides an overall advantage in mean acquisition time
when compared over the full range of possible numbers of active users.
The detection in accordance with this invention is thus preferably
performed on two probability distribution functions (pdf) resulting from
the difference between the two statistics described above. As in the
single-user detection case, one pdf corresponds to a H.sub.o case, or
codes mis-aligned case, the other to a H.sub.1 case, or codes aligned
case. In either case, and because the two statistics are computed
sequentially, the noise components are essentially uncorrelated. The pn
codes are length P=32, and the timing is searched in the SU 14 in
half-chip increments (.delta.t=T.sub.c /2), thus giving Q=64 possible
positions. The time cell yielding the best signal level will therefore, at
worst, be off by 1/4T.sub.c.
Given the H.sub.o case, both the single-user and multi-user tests yield
ambient gaussian white noise and multi-user interference. Since the
multi-user codes are uncorrelated with the test code, both noise
components are assumed gaussian, and hence central a .chi..sup.2
distributed at the detector output. The channel noise variance is N.sub.s
=1/2N.sub.o W.sub.s, where 1/2N.sub.o =N.sub.c /W.sub.c is the two-sided
noise spectral power density. Hence N.sub.s =N.sub.c T.sub.c /2T.sub.s
=1/2N.sub.c /P. The worst case multi-user interference power is I.sub.s
=1/2I.sub.o W.sub.s, where 1/2I.sub.o is the two-sided interference
spectral density.
It should be noted that the interference density varies over the
null-to-null bandwidth of the receiver filter 15 of FIGS. 2 and 3. Of most
interest, however, is the density near DC. The equivalent interference
bandwidth at this power density is 1/2W.sub.c =1/T.sub.c. Hence,
1/2I.sub.o =I.sub.c T.sub.c, and I.sub.s =I.sub.c T.sub.c /T.sub.s
=I.sub.c /P. For the H.sub.o hypothesis, then
##EQU8##
where S/P represents the contribution of the side-channel.
Given the H.sub.1 case, the single-user test yields the ambient white
gaussian noise term and a deterministic signal component, and hence
contributes a non-central .chi..sup.2 distributed random variable at the
output of the detector 12d. The multi-user test yields ambient white
gaussian noise which contributes a central .chi..sup.2 distributed random
variable. This assumes, of course, that the multi-user interference can be
ignored due to the mutual orthogonality of the aligned pn spreading codes.
However, and because of filtering, the pn codes are typically not strictly
orthogonal. It can be shown that, with 3-pole butterworth filtering, the
interference null depth is approximately 25 dB upon pn code alignment.
Thus, for worst case fades (E.sub.s /N.sub.o .apprxeq.3 dB), the
interference null can be well below the ambient white gaussian noise
level. Another factor to consider results from the fact that the worst
case alignment offset of 1/4T.sub.c will have the effect of weakening the
interference null and the signal correlation peak. From the graphs shown
in FIG. 5, the multi-user correlation null can be seen to lose as much 14
dB and the single-user correlation peak about 2 dB. Thus, these effects
are taken into consideration. It can be assumed however, that for a worst
case fade, the noise variance is approximately .sigma..sub.1.sup.2
=N.sub.s for both the single-user | | |
