Note: Descriptions are shown in the official language in which they were submitted.
1
~1'O 91 l0 ~ 831 PCT/S E90/0069'
;~u4~~~~
A METHOD OF DETERMINING SAMPLING TIRE POINTS
TECHNICAL FIELD
The present invention relates to a method of determining sampling
time point when transmitting symbol sequences with recurrent
synchronization sequences, wherein the symbol sequences are
transmitted as analog signals over a channel and are liable to be
subjected to disturbances during said transmission, said method
comprising the steps of
sampling the received, analog signals at recurrent signal
sampling time points which are selected in relation to a
synchronization time point common to a transmitter and a
receiver, wherein a time interval for transmission of a
symbol, a symbol time, includes a whole number of signal
sampling time points: and
- effecting channel correlation for calculating impulse
response for the channel with the aid of the known synchroni-
zation sequences and the sampled, received signals.
PRIOR ART
In the radio transmission of digital information, a number of
problems occur which must be solved in order to enable the
receiver to discern the information originally transmitted. One
example of these problems resides in transmitter and receiver
synchronization. This problem has found many solutions for
different applications and is well known to the skilled person.
Another problem is that the transmitted signals are liable to be
affected by various kinds of disturbances, for instance noise,
fading and multi-path propagation. The difficulties associated
herewith have been tackled in several ways. Thus, it is well known
to transmit a known synchronizing word and to calculate an impulse
response for the transmission channel between transmitter and re-
ceiver with the aid of the known synchronizing word. The trans-
mitted, unknown information can be interpreted by the receiver
~l~t) ~i.~tl"Si1 f CTiSE90i00b9'
2U4~~99
with the aid of the impulse response, and can be converted to an
acoustic signal for instance, through a plurality of signal
processing stages. A further example of the difficulties ex-
perienced with signal transition is one of controlling the re-
ceiver frequency in time with the transmitter frequency. This
difficulty has been recognized and a number of well-known methods
are found for controlling the frequency of the receiver. One
problem, on the other hand, which would not appear to have
awakened any particular interest is that of optimally utilizing
the signal strength of the transmitted signal during the aforesaid
transmission of digital information. It should be observed in this
respect that in the case of multipath propagation a transmitted
signal can be refound at several mutually different receiver time
points. Despite research, both in the patent literature and in
other sources, no publication has been found which deals with this
problem.
DISCLOSURE OF THE INVENTION
The present invention is based on the concept of optimally
utilizing the signal strength of a transmitted signal for the
purpose of simplifying the signal processing necessary in a
receiver. This optimization is achieved by selecting a time point
for sampling the tranmitted symbols. This choice is based on a
comparison of the energy content of different parts of the channel
impulse response.
The invention has the characterizing features set forth in the
accompanying Claims.
BRIEF DESCRIPTION OF THE DRAWINGS
An exemplifying embodiment of the invention will now be described
with reference to the accompany drawings, in which
Figure 1 is a block schematic illustrating a part of a mobile
telephony system;
3
~Z'O 91!0'831 PCT/SE90/OU69''
2U4~599
Figure ~< illustrates time slots for time-shared transmission of
information;
Figure 3 illustrates symbol sequences transmitted in a time-
shared time slot;
Figure 4 illustrates a complex number plan with symbol values;
Figure 5 is a block schematic illustrating a channel estimation
filter: and
Figure 6 is a diagram with an impulse response for the trans-
mission channel.
BEST MODE OF CARRYING OUT THE INVENTION
Figure 1 schematically illustrates a radio transmission system.
Signal processing, far instance channel coding, of the information
to be transmitted is effected in a unit 1 and the information is
sent to a digital/analog converter D/A in the form of digital
signals. The converter sends analog signals to a transmitting
radio unit RA1, which transmits the signals over a channel to a
receiving radio unit RA2. This unit sends the received signals to
an analog/digital converter A/D, in which sampling of the signal
takes place at a relatively high rate. Sampling takes place at
regular intervals at signal sampling time points, the number of
which is generally referenced n_, so as to obtain a sampled signal
S(n). That part of the radio transmission system described
hitherto is well known to the person skilled in this art. Synchro-
nization, channel correlation and sampling of the signal S(n)
takes place in a correlation-and-synchronization circuit KS, as
described in more detail hereinafter. The actual method in which
a sampling time point is selected during a sampling operation is
the object of the present invention. The sampling signals are sent
from the circuit KS for further signal processing, in the case of
the illustrated embodiment to an equalizer V, which produces
estimated symbols U. The inventive method of selecting sampling
Y
\1'O 91/0"R31 fCT/SE90/0069'
2044599
time points provides improved signal processing in the equalizer
V.
The aforedescribed radio transmission system may, for instance,
form a part of a time-shared mobile telephony system. Subscribers
in this system are regularly assigned recurring time slots 1,--
--, P as illustrated in Figure 2, in which T signifies time. One
of the subscribers has been assigned the time slot numbered H and
the symbol sequences designated SS1, SS2, SS3,---, are transmitted
in this time slot. Each symbol sequence includes a synchronization
sequence SY and a data sequence D and together take up the length
of a time slot designated TO in Figure 3. The transmitted signals
may be modulated in accordance with QPSK-modulation, as illusrated
in Figure 4, for instance. In a complex number plan, with the axes
designated I and Q, the four possible values of the symbols are
marked one in each square with the binary digits 00, O1, 10 and 11.
In the case of the aforesaid QPSK-modulation, the time taken to
transmit a symbol, a symbol time Ts, is equal to the time for two
binary digits.
Various kinds of disturbances are liable to occur during trans-
mission of the symbols over the channel, for instance such
disturbances as multipath propagation, as indicated with double
signal paths in Figure 1. These disturbances change from one
signal sequence to the immediately following sequence. In order to
enable interpretation of the transmitted information contained in
the data sequence D, the impulse response of the channel is
determined in a known manner for each signal sequence. This is
achieved by correlating the known synchronization sequence SY in
the receiver with the received, sampled values S(n) in the
synchronization sequence. Correlation is carried out in a filter,
as illustrated in Figure 5. The filter has delay units 2, filter
coefficients 3 and summators 4. The filter coefficients have the
values SYO-°-SYK-1 corresponding to the known synchronization
word, the length of which is a K symbol sampling intervals. The
received, sampled synchronization word S(n) is delayed in the
delay unit 2, so as to subsequently obtain signals
5
V'O 9110"831 fCT/SE90/0069'
204459
S{n-N)---S{n-{K-1)N) which are stepwise delayed by one symbol
sampling interval. The delayed signals are multiplied with their
respective coefficients and summated in the sum.~nator 4. Subsequent
to dividing with the value K in a circuit 5, the values
C2 (n-(K-1)N) are obtained in the sampled impulse response for the
channel between the radio units RA1 and RA2.
Figure 6 illustrates the sampled impulse response obtained in the
aforedescribed manner for the synchronization sequence in the
symbol sequence SS1. As in Figure 2, T designates the time and C2
generally designates the energy for the discrete correlation
values of the impulse response,~these values being marked with
columns at the signal sampling time point n. The impulse response
has a length of L+M x N sample, which have been numbered from 0 to
L-1+M x N in the Figure. In the illustrated case, N designates the
number of signal sampling time points n for each symbol and
according to the illustrative example N is equal to 2. The length
of a channel estimate for the equilizer V in symbol times Ts is
designated M, and in the case of the illustrative example, M is
equal to 3 . The length M x Ts of the channel estimate is determined
by the magnitude of the time dispersion possessed by the channel,
so that the equilizer V will be able to equalize dispersions which
range up to M x Ts. The letter L designates the number of signal
sampling time points over which the correlation must be carried
out in order to ensure that the impulse response will cover a large
and rapid change in the transmission properties of the channel.
Normally; an interval which covers L example is called a corre-
lation window. According to the sample illustrated in Figure 6,
L=11 and the signal sampling time points n of the impulse response
have been numberd from 0 to 16.
As mentioned in the introduction, sampling of the signal S(n)
takes place in the correlation-and-synchronization circuit KS.
This sampling takes place in step with the symbol timing at symbol
sampling time points having an interval of one symbol time Ts
between two mutually adjacent samples. The impulse response is
also sampled in step with the symbol timing to a channel es-
timate, the length.o~ which is selected to M symbol times Ts in
6
~i'O 91/0'.831 PCT/SE90/0069'
2044599
accordance with the aforegoing. It is possible, in accordance with
the invention, to select a plurality of different channel
estimates from the impulse response in Figure 6, this selection
being effected in the following manner: A first sampling of the
impulse response commences at the signal sampling time point n =
0. Sampling continues in the symbol sampling time points at each
alternate sampling time point n = 2, n = 4 up to n = 6, where
according to the illustrated embodiment N = 2 and M = 3. This
channel estimate is marked in Figure 6 with heavily drawn columns.
There is obtained in this way a channel estimate of length M x Ts,
the total energy Eke(n) of which can be expressed generally by the
relationship '
M
Eke (n) = E C2 (n+N x i)
i=0
which constitutes a comparison value for the energy of the channel
estimate. The next sampling of the impulse response commences at
n - 1 and new values of the energy Eke(n) are subsequently
calculated up to n = L-1, in the case of the illustrated embodiment
n = 10. There is obtained in this way an L number of comparison
values Eke(n) of which one has a largest magnitude and is desig-
nated E ke(n). Those symbol sampling time points in the impulse
response which give a channel estimate with this maximum energy
has been marked with a cross in Figure 6. The channel estimate
having the comparison value E~ke(n) is selected and the first
sampling time point in the selected channel estimate is selected
as the sampling time point.. In the case of the illustrated
embodiment of Figure 6, the sampling 'rime point n = 8 is selected,
which according to the aforegoing applies for the symbol sequence
SS1.
According to the invention, the sampling time point can also be
calculated in the following alternative manner. That signal
sampling time point of the signal sampling time points n in which
the impulse response has maximum amplitude C2max(n) is sought and
constitutes the selected sampling time point. The comparison value
wo m.~o~sm
in this sampling time point can be expressed with the simple
relationship
E t (n) = a x CYmax (n)
where a is a constant. C2max (n) is marked with a ring in the Figure
6 example and the corresponding sampling time point is n = 9. This
alternative method of selecting the sampling time point is
beneficial when the impulse response has a single correlation
value C2 (n) which dominates over the remaining correlation values.
A combination of the two aforedescribed methods of selecting
sampling time points also lies within the purview. The comparison
value E~ke(n) and the comparison value E~t(n) are calculated in
accordance with the aforegoing. The largest of these values Emax
is selected and the corresponding signal sampling time point nmax
constitutes the selected sampling time point.
The aforedescribed inventive method of selecting a sampling time
point for one of the signal sequences according to the example SS1
has the advantage of simplifying the following signal processing
step in, for instance, the equalizer V. It is possible, however,
that the transmitted signal of the Figure 1 illustration has been
subjected to fading, i.e. the signal strength has fallen radically
over a short time interval due to signal interference. If the
fading occurs during the synchronization sequence SY, the selected
channel estimate and the selected sampling time point will not be
representative of the remainder of the symbol sequence. This
weakness is particularly noticeable in transmission systems which
have long symbol sequences extending over several milliseconds.
This weakness is counteracted in accordance with the present
invention by calculating an estimated value nest () for the
sampling time point iteratively. The maximum energy value, for
instance Emax' and the corresponding sampling time point nmax is
subsequently calculated for the sequences SS1, SS2, SS3-- . The
estimated sampling time point for the symbol sequence numbered i
is calculated in accordance with the relationship
nest ( j ) = nest ( -1 ) + B ( nmax nest ( -1 ) )
8
wo 9mo~sm P 204599
In this case, nest( 1) is the estimated sampling time point from
the preceding symbol sequence; nmax belongs to the symbol sequence
numbered j and B is a weighting function. This weighting function
may, for instance, assume the value B = 80 when Emax exceeds or is
equal to a threshhold value EO, while in other cases B is equal to
0. Other average value formations can also be made. In general,
the estimated sampling time pcint nest () will lie between two
signal sampling time points n and the signal sampling time point
which lies nearest nest() is selected as the sampling time point.
It should be noted that all time points of the receiver, for
instance the signal sampling time points, are calculated in
relation to a synchronization time point Tsync of a frame clock,
which is controlled in a known manner.
The invention has been described in the aforegoing with reference
to an exemplifying embodiment applied with time-shared mobile
telephony. It will be understood, however, that the invention can
also be applied with other signal transmission systems as soon as
recurrent synchronization sequences are transmitted. The inter-
vals between the synchronization sequences may have varying
lengths.