 |
Description  |
|
|
This invention relates to CELLULAR communications systems and to
communications receiver devices for use in such systems.
BACKGROUND OF THE INVENTION
Cellular communications systems providing a telephony service are
attracting increasing commercial interest. In such systems a number of
mobile or portable telephone instruments are in radio communication with a
local base station. In a typical system, traffic is carried on the radio
link between the telephone instrument and the base station in a digitally
encoded format. This provides security of transmission and some resistance
to interference. Because of the high (UHF) radio frequencies used for
transmission between a mobile telephone and the base station, this
transmission is effectively restricted to line of sight.
In metropolitan or dense urban areas the direct line of sight between a
mobile telephone and the base station may be blocked by buildings and the
mobile telephone thus receives a distorted version of the transmitted
signal comprising a number of indirect multipaths or reflected signals.
These signals may have relative time delays of more than 5 microseconds
thus causing significant overlap of the various indirect signals. This can
cause effective loss or `break up` of the received signal.
A further difficulty, particularly in dense urban areas, is that of
spectrum allocation. Indeed, the availability of spectrum can restrict the
number of users that can be accommodated.
A recent approach to these problems has been the introduction of the code
division multiplex (CDM) transmission technique. This technique is
described, for example, by U Grob in IEEE Journal Selected Areas Comms.,
8, June 1990 pp 772-780 and by W C Y Lee in IEEE Transactions on Vehicular
Technology, 40 No 2, 2nd May 1991 pp 291-302.
In a CDM transmitter a number of data sources produce information bit
streams in parallel. Each bit stream is associated with a corresponding
binary spreading sequence which spreading sequence repeats after a
predetermined period. The spread bit streams from the source are modulated
onto a carrier and then added linearly to give a multiband composite RF
signal which is broadcast to the mobile telephones within the service area
of the transmitter. At each receiver, the composite signal is decomposed
to individual messages by correlation with a particular spreading sequence
associated with a particular message.
The technique allows a number of users to occupy the same channel thus
significantly increasing the effective spectrum availability. However,
with the rapidly increasing popularity of mobile communications services,
there is a need to accommodate further users within each channel.
It is an object of the invention to minimise or to overcome this
disadvantage.
A further object of the invention is to provide a receiver configuration
that permits an increased number of users to occupy the same channel.
SUMMARY OF THE INVENTION
According to the invention there is provided a receiver apparatus for a
code division multiplex (CDM) communications system in which a plurality
of broadcast data signals are provided each with a respective spreading
sequence, the receiver apparatus including a channel estimator adapted to
provide a channel impulse response for a said spreading sequence which
spreading sequence has been allocated to the receiver, means for
convolving the channel impulse response with the sequence allocated to the
receiver whereby to provide a matched sequence, and means for correlating
the matched sequence with the received signal whereby to recover a said
broadcast data signal.
According to the invention there is further provided a receiver apparatus
for a code division multiplex (CDM) communications system in which a
plurality of broadcast data signals on a common channel are provided each
with a respective spreading sequence, the receiver apparatus including
means for generating a copy of a said spreading sequence, that sequence
having been allocated to the receiver, a channel estimator incorporating a
matched filter adapted to provide a channel impulse response from the
allocated spreading sequence, means for convolving the channel impulse
response with said copy of the spreading sequence whereby to provide a
matched sequence, and an adaptive filter to which in use a received signal
is fed and whose coefficients correspond to the matched sequence whereby
to despread the received signal and recover said data.
According to a further aspect of the invention there is provided a code
division multiplex (CDM) communication system, including a base station
and a plurality of receiver stations, wherein the base station has means
for transmitting on a common channel a plurality of data signals provided
each with a respective spreading sequence, each spreading sequence being
allocated to a said receiver station, and wherein each receiver includes a
channel estimator adapted to provide a channel impulse response for the
spreading sequence allocated to that receiver, means for convolving the
channel impulse response with the sequence allocated to the receiver
whereby to provide a matched sequence, and means for correlating the
matched sequence with the received signal whereby to recover the broadcast
data signal.
BRIEF DESCRIPTION OF THE DRAWINGS
Embodiments of the invention will now be described with reference to the
accompanied drawings in which:
FIG. 1 is a schematic diagram of a code division multiplexer (CDM)
transmitter;
FIG. 2 is a schematic diagram of a receiver according to one embodiment of
the invention;
FIG. 3 illustrates the matched filter response of the receiver of FIG. 2;
FIG. 4 is a timing diagram illustrating the operation of the receiver of
FIG. 2;
FIG. 5 is a further timing diagram; and
FIG. 6 shows an alternative receiver construction.
FIG. 7 shows a further receiver construction.
DESCRIPTION OF PREFERRED EMBODIMENT
Referring to FIG. 1, a CDM transmitter includes a plurality of data sources
11a to 11d each of which provides a respective sequence of binary
information bits, each bit being of duration T seconds. In the interests
of clarity, only four data sources have been shown in FIG. 1, but it will
be appreciated that in practice a large number may be provided. Each data
source 11 is coupled to one input of respective exclusive OR (XOR) gate
12a to 12d, the other input of that gate being fed with a respective
binary spreading sequence uniquely assigned to the respective data source.
The chipping rate C of these sequences is an integral multiple K of the
data source bit rate I/T and there are thus K spreading sequence chips for
each data sequence bit. All of the spreading sequences repeat after K
chips. In a typical arrangement, K has a value of 2.sup.N-1 where N is the
number of data sources. Advantageously, the spreading sequences comprise a
set of Gold codes which have the property of matched orthogonality. Within
a channel, each Gold code is allocated to a respective user. When each
data source stream is XOR'ed by its respective spreading sequence, it
becomes a sequence of repetitions of the spreading sequence with sign
reversals according to whether each data bit is a ONE or a ZERO. The
repetitive sequences thus generated are fed each to a corresponding
modulator 13a to 13d driven with a carrier frequency from a local
oscillator 14. The modulated signals are added linearly by adder 15, the
composite signal being fed to an antenna 16 for broadcasting to the
service area. A further data source 110 provides a pilot or reference
signal and is XOR'ed with a broadcast synchronisation sequence in gate
120. The synchronisation sequence uses e.g. a PN (pseudo-noise) code which
is received by the receiving station whereby to provide control timing
information. The PN sequence may be designed to have good auto-correlation
properties such that, when correlated with a replica of itself in a
matched filter in a receiver, the output function traces the channel
impulse response (CIR). The output of the gate 120 is fed to a respective
modulator 130 whose output is in turn fed to the adder 15 whereby that
output is combined with the outputs from the modulators 13a to 13d.
Referring now to FIG. 2, this shows a receiver circuit according to one
embodiment of the invention. The receiver comprises three components or
units, a channel estimator 21, a digital filter 22 and an adaptive matched
filter 23. Broadcast signals received by the antenna 24 are fed to a
further matched filter 210 incorporated in the channel estimator 23. This
filter 210 provides an estimate of the channel impulse response (CIR).
In the receiver of FIG. 2, the channel impulse response is estimated by
passing the broadcast synchronisation signal into the matched filter 210
incorporated in the channel estimator 21. The synchronisation signal may
be isolated from the traffic part of the data by providing that signal in
a special time slot which may easily be separated by gating. The
synchronisation signal is designed such that, in the absence of the
channel, the signal produces a matched filter response having a single
large peak and very low side lobes. In the presence of a non-ideal
channel, the matched filter response is convolved with the channel impulse
response (CIR) and is thus a replica of the CIR.
The data part of the traffic is separated from the broadcast sequence and
is passed to the adaptive matched filter 23. The purpose of this filter is
to select one out of many users by matching only to the particular user's
Gold code. For an ideal channel, when the CIR is an impulse, the Gold
codes assigned to different users will be orthogonal for zero time shift
and substantially complete isolation between users can be achieved.
Under non-ideal conditions tone effects become apparent in the matched
filter 23. These are:
(i) a loss of orthogonality between user codes due to time shifting
effects.
(ii) dispersion of the code among different multipaths with disparate time
lags causes a loss of output power and a loss of performance against
noise.
The loss of orthogonality is remedied by the adaptive filtering and this
will be discussed below. The loss of output power is remedied by matching
the filter 23 to the observed distorted version of the users code.
Normally, the user's Gold code is fed to the filter taps of the filter 23
in an unmodified form, apart from time reversal and complex configuration,
and the filter correlates the code with the traffic data, Since the
contents of the filter taps are constant for an ideal channel, this need
be set up only once. When the channel is imperfect e.g. as a result of
multipath reception, the internal replica of the user's Gold code is
predistorted to match the code to the distorted Gold code arriving over
the channel. This is effected by convolution of the local Gold code
replica with the CIR estimate, this convolution having effected with the
digital filter 22. The estimated channel impulse response is fed to the
tap weights of the filter 22 which then convolve the ideal Gold code with
this estimated response. This process need be updated only at a rate
consistent with the rate of change of the channel impulse response. For a
non-moving use the time period between updates can be several minutes.
The response of the matched filter 210 is illustrated in FIG. 3. In the
presence of multipath signals, the matched filter produces several outputs
in cascade, each output being a replica of the ideal channel response with
different amplitude and phase weightings. The signal power is dispersed
between these multiple outputs and is at a low level in any one of them.
The channel estimator combines these outputs coherently to provide an
effective diversity gain.
As described above, the output of the channel estimator 21 (FIG. 2) is fed
to the digital filter 22 where the channel impulse response is convolved
with the spreading sequence allocated to that receiver, i.e. the user's
Gold code, to produce a modified Gold code sequence. This sequence is a
replica of the Gold code sequence as actually observed at the receiver and
so is a matched sequence for the user. The co-efficients of this modified
and extended sequence are then used as filter co-efficients for the
adaptive matched filter 23 whereby to despread the signal and recover the
user data.
FIG. 4 illustrates the mode of operation of the adaptive modification of
the filter of the receiver of FIG. 2 whereby the filter is provided with
variable coefficients. When this technique is employed with a multipath
channel where intersymbol interference and interference from other users'
codes are significant, the adaption of the filter significantly reduces
the effects of these sources of interference. The sequence of data bits
has been separated into even and odd bits for clarity. Each bit is
extended in time by the multipath from N chips to N+E and the
channel-matched filter does a weighted sum over the extended duration N+E
of each data bit.
The output of the channel matched filter depends both on its total length
and the coefficients. The assumption is made that the multipath duration
is less than one data bit length T (E<N spreading chips) and the target
bit is positive. There are three possible conditions to be considered.
(I) If the matched filter has full length, L=N+E, designed for maximum
correlator energy output on the desired transmission then, when bit 2 of
the figure is in the correct position and is being output, the filter also
partially overlaps bits 1 and 3 on either side as is clear in the figure.
Therefore its output has four possible values depending on the signs of
the preceding and succeeding data bits.
(II) If the receiving filter has a truncated length of N or less
L.ltoreq.N, then it can be located in time so as to only overlap one other
bit. This will occur when the filter is matched to the back end of the
spreading sequence so that its correlation of the desired bit sequence
starts after the tail end of the preceding bit sequence has disappeared
and before the subsequent bit has started.
(III) If L.ltoreq.N-E then the filter can be located so as not to overlap
any surrounding bits. This would solve the IS problem, however in this
case, if E is comparable to N there is a considerable loss in matching
performance due to direct loss of energy and there would be enhanced
susceptibility to thermal noise.
These three possibilities are illustrated in FIG. 5. When the receiver
filter has length N+E, and in intermediate conditions between
N-E.ltoreq.L.ltoreq.N+E, there is interference from two other bits. One
way to treat the situation is to consider that there is no inter-bit
interference introduced by the channel but each source uses one of four
different spreading codes dependent on the signs of the data surrounding
the target bit.
Signal demodulation in the presence of intersymbol interference (ISI) but
in the absence of other interference can best be done using a sequential
decision algorithm such as a Viterbi decoder though, for the small amount
of ISI expected, straightforward thresholding of the matched filter output
would be nearly as good. However the data bits in question can also be
other users data. Should it be required to make a modified receiving
filter which is near-orthogonal to other users' sequences then, based on
the concept of the four spreading codes introduced above, there are four
different cross correlation values to be considered rather than just the
one.
If the data bits are spread with codes having N chips then the sequence for
a given data bit becomes a complex vector V with N elements. In the
absence of multipath the matched filter output, at the instant of output
sampling, can be represented as an inner product of the matched filter
weights with the total input sequence and, by linearity, is a summation of
the inner products of the weights with each of the individual sequences
present in the input data.
The theory of linear algebra indicates that a given N-vector can be
orthogonal to at most N-1 other linearly independent vectors. Thus a
filter with N coefficients can be made orthogonal to N-1 spreading
sequences yet retain a response to its desired sequence. Using Gold codes
of duration N=2k-1 bits, N-1 users can share a channel which has no ISI
and achieve orthogonality.
In case (I) above and intermediate cases with less than full length
filters, each sequence can be distorted by multipath in four ways at the
instant of sampling the filter output and given m actual users, a total of
4 m code variants (4(m-1) of which represent interference) is observed by
the filter. In order that a mismatched receiver filter can be made
orthogonal to all these variants of the data from the other m-1 users, we
require that
(N-1)>4(m-1) or
m.ltoreq.(N+3+/4 for exact orthogonality.
In case (ii) above with a shortened receiver filter it is only necessary to
be orthogonal to 2(m-1) other codes so the corresponding condition is
(N-1).gtoreq.2(m-1) or
m.ltoreq.(N+1)/2
The way in which the receiver matched filter can be automatically modified
to achieve the desired orthogonality condition is now described.
Up-link receiver at the base station.
Here the mobile or remote users will all have different multipaths and, if
the base station has to compute a filter directly from the ensemble of
channel impulse responses, then it must be in a position to continuously
update these CIR's. This is a formidable task, for several reasons:
The users will require a considerable data overhead in the provision of
training sequences in their transmissions. The measurement of uplink CIR's
is problematic in any case because of the very interference which is
addressed in this paper and extended duration training sequences, i.e.
much longer than the standard spreading sequences, may be necessary to
give sufficient carrier to interference ratios.
The base station has the computation burden to track the multipath
behaviour of a large number of mobiles.
To reduce the computational task, we prefer to employ a feedback loop to
adaptive update the weights without explicit knowledge of the channel
impulse responses. There are a large number of algorithm suitable for this
purpose and typically a least squares algorithm will minimise the means
square errors between the matched filter output and the ideal output. A
decision feedback equaliser can be used in this situation and FIG. 6 shows
such a system in outline.
Downlink to mobiles.
Here the situation is less onerous since although all users on the downlink
will have different CIR's they can share a common high-energy channel
impulse response measurement signal, i.e. the broadcast sequence of FIG.
1, and it would be feasible to measure CIR and compute a matched filter
directly. Alternatively an adaptive filter similar to that in the base
station could be used.
As well as training the weights of the receiver filter, the length of the
filter can be varied so as generate the conditions described above with
reference to FIG. 4. Optimism of a performance criterion such as mean
square error will achieve a compromise condition between high cancellation
of strong or highly correlated interference and little extra cancellation
of weak interference.
An alternative receiver construction is shown in FIG. 7 In this
arrangement, the adaptive matched filter of the receiver of FIG. 2 is
replaced by first and second integrate and dump filters 41, 42
respectively. In this receiver, the estimated channel impulse response is
convolved with the receiver Gold code, as described above with reference
to FIG. 2, whereby to generate a replica of the code as distorted by the
channel. As the convolved sequence has an extended duration, the two
integrators are used alternatively with some overlap of their integration
periods. The mode of operation of this receiver is analogous to the
receiver of FIG. 2 described above.
By providing increased discrimination between users, the technique
described above allows the number of users sharing a channel to be
increased. The receovers may comprise mobile receivers or they may
comprise e.g. office data terminals.
* * * * *
|
|
 |
|
 |
Description |
|
