 |
Description  |
|
|
BACKGROUND OF THE INVENTION
1. Field of the Invention
The field of the invention is that of the broadcasting of digital data
designed to be received notably by moving receivers in an urban
environment, namely in the presence of interferences or jamming, under
conditions of multiple propagation (RAYLEIGH process) generating a
phenomenon of fading.
The invention can be applied more particularly, but not exclusively, to a
system of digital sound broadcasting as described in the French patent
applications Nos. 86 09622 of Jul. 2, 1986 and 86 13271 of Sep. 23, 1986,
on behalf of the same applicants. This system of digital broadcasting,
presented in these prior patent applications, is based on the combined use
of a channel coding device and a method known as the COFDM system (coding
orthogonal frequency division multiplex system).
2. Description of the Prior Art
The modulation method proper of this prior art system consists in providing
for the distribution of the constituent digital elements of the data
signal in the frequency-time f-t space and in simultaneously emitting sets
of digital elements on M parallel broadcasting channels by means of a
multiplex of orthogonal carrier frequencies. This type of modulation makes
it possible to prevent two successive elements of the data train from
being emitted at the same frequency. This enables the absorption of the
fluctuating selectivity in frequency of the channel, by frequentially
dispersing the initially adjacent digital elements during the
broadcasting.
The prior art encoding method seeks, for its part, to enable the processing
of the samples coming from the demodulator to absorb the effect of
variation in amplitude of the signal received, due to the RAYLEIGH
process. This encoding is advantageously a convolutive encoding, possibly
concatenated with a REED-SOLOMON type encoding.
In a known way, the encoded digital elements are furthermore interlaced, in
time as well as frequency, in order to maximize the statistical
independence of the samples with respect to the Rayleigh process and the
selective character of the channel.
SUMMARY OF THE INVENTION
An aim of the present invention is to provide an optimized embodiment of
the frame structure of the broadcast signal so as to derive the maximum
benefit from the self-synchronization properties of the COFDM method, and
to maximize the resistance of the system to the pulsed interferences and
jamming.
This aim as well as others which shall appear subsequently are achieved by
means of a method for the broadcasting of digital data, notably for sound
broadcasting at a high throughput rate towards mobile receivers, of the
type providing for the distribution of said data in the form of digital
elements in the frequency f/time t space and the emission of frames of
symbols each formed by a multiplex of orthogonal carrier frequencies
modulated by a set of digital elements and broadcast simultaneously on M
parallel channels,
wherein the recovery of synchronization of each frame received is achieved
by analog synchronization means without prior extraction of a clock
signal.
According to an advantageous characteristic of the invention, the header of
each of said frames of symbols comprises an interval of silence, with the
duration of a digital symbol of the frame. This symbol of silence is
capable of being used as a means of synchronization of the demodulation
and/or as a means of analysis of the pulsed noise and of the jamming which
are characteristic of the channel.
According to another characteristic advantage of the invention, said frame
header comprises an unmodulated multiplex of said M orthogonal carrier
frequencies, with the duration of a digital symbol of the frame. This
unmodulated symbol may be used as a synchronization means and/or as a
phase reference for the J phase-modulated carriers of the digital train.
Preferably, said symbols are formed by means of a frequential interlacing
operation using a reversible deterministic function, said function
consisting in a method for shuffling the indices of said frequencies with
a maximization of the dispersal of the frequencies associated with
adjacent digital elements of the source data signal. Said shuffling of
indices advantageously consists in applying a function of the bit
inversion type to said binary encoded indices.
Preferably, the phase modulation done on the carriers is of the type with
four phase states, each carrier being modulated by a pair of digital
elements, and said pairs are formed by a source sequence of digital
elements in forming packets of 2J consecutive elements in said sequence,
and in associating the elements two by two in each packet according to a
criterion of maximization of dispersal of the adjacent digital elements of
the source sequence.
In an advantageous mode of the invention, the pairs are formed by splitting
each of the said data packets into two half-packets and by pairing the
same-order digital elements in each half-packet.
The frequential interlacing thus defined is advantageously combined with a
temporal interlacing achieved by the application of delays, the value of
which is assigned to each digital element by the application of a
reversible function of the index of the digital element, the delay
function (F) being such that the deinterlacing in the initial order and
the recovery of each element of the source sequence are achieved by the
application, to each digital element with a same index in the sequence
received, of a complementary delay value with respect to the depth of the
maximum interlacing of the delay function.
According to another characteristic of the invention, the method includes a
jamming extraction process comprising the following steps:
the received signal is analyzed during said symbol of silence, on the
spectrum covered by the J orthogonal carriers;
the frequencies affected by complex Gaussian noise are identified;
a correction and/or cancellation processing is done of the useful signal
received by said detected jammed signals.
Advantageously, said spectral analysis is complemented by a two-dimensional
filtering step in the time/frequency space, providing for a smoothing of
the results of said analysis on the useful extent of the analyzed
spectrum.
According to a complementary characteristic of the invention, the method is
of the type implementing a convolutive encoding of the data at the
transmitter, and a soft decision decoding through maximization of
likelihood at the receiver,
wherein said processing of correction and/or cancellation of the useful
signal received for said detected, jammed signals consists in informing
said soft decision making by means of the noise power detected at each of
said frequencies.
BRIEF DESCRIPTION OF THE DRAWINGS
Other features and advantages of the invention will appear from the
following description of an embodiment given by way of a non-restrictive
example, and from the appended drawings, of which:
FIG. 1 is a block diagram of a transmission-reception chain implementing
the method of the invention;
FIG. 2 gives a schematic view of the structure of a frame as broadcast by
the system of the invention;
FIGS. 3a, 3b respectively give a schematic view of, firstly, a standard
acquisition chain of synchronization by clock signal extraction and,
secondly, the principle of analog recovery of synchronization with two
stages according to the invention;
FIG. 4 gives a schematic view of a time-frequency interlacing/deinterlacing
chain, optimal in cooperation with the synchronization principle of the
invention;
FIG. 5 shows an advantageous mode of a convolutive temporal interlacing
that can be implanted in the chain of FIG. 4;
FIG. 6 illustrates an advantageous mode of a shuffling of frequency indices
compatible with the frequential interlacing of the chain of FIG. 4;
FIG. 7 gives a schematic view of a reception chain with extraction of
jamming, compatible with the high throughput rate digital broadcasting
method of the invention.
DESCRIPTION OF A PREFERRED EMBODIMENT
The different aspects of the embodiment which shall be described
hereinafter more particularly concern digital sound broadcasting towards
mobile receivers, as defined notably in the EUREKA Digital Audio
Broadcasting (DAB) program.
However, it is clear that the high throughput digital broadcasting
principle of the invention can be applied to any type of communications,
notably in channels subjected to the Rayleigh process such as, for
example, aircraft-satellite or other types of communications.
In the digital sound broadcasting application of the DAB, one aim may be,
for example, the transmission of sixteen stereophonic programs in an 8 MHz
wide frequency band with a digital throughput rate of the order of 100
kbits (after compression).
A transmission chain of the type described in the patent applications
mentioned in the introduction is shown in FIG. 1.
Each of the N(16) channels C.sub.0 to C.sub.n-1 undergoes an encoding 10 in
parallel, then a time-frequency interlacing 11 on a separate channel,
before being subjected jointly to a process 12 of temporal multiplexing
and OFDM modulation.
The encoding 10 is advantageously of the convolutive type. The
time-frequency interlacing 11 is aimed at shuffling the digital elements
of each channel in order to give them maximum independence with respect to
interferences and to the jamming of the broadcasting channel 13.
The OFDM modulation consists in the modulation of symbols each formed by a
multiplex of orthogonal frequencies broadcast simultaneously on J
channels. This operation can be achieved by a Fourier transform on the
encoded and interlaced digital sequence of each channel C.sub.i.
By way of example, in an 8 MHz frequency band, it is possible to define 512
separate 15 625 Hz carrier frequencies. Of these, 448 are usable, after
elimination of the central frequency of the spectrum and of the lateral
carriers (1/8th of the spectrum) to take the filtering constraints into
account.
The reception chain comprises the steps of channel selection 14,
demodulation 15, frequency de-interlacing 16 and decoding 17 of the
de-interlaced channel.
The channel selection operation 14 is performed advantageously by Fast
Fourier Transform (FFT) so as to decimate the set of suitably interlaced
carriers to apply the OFDM demodulation operation only to the carriers of
the selected channel (see addition certificate No. 86 13721 already
referred to). After the time-frequency de-interlacing 16, a "soft"
decision Viterbi decoding 17 is advantageously applied.
The data frame, as broadcast through the channel 13 presents, according to
the invention, the structure of FIG. 2.
The frame is formed by a header 21 and N elementary channels 22 marked
C.sub.0 to C.sub.n-1 each formed by K symbols 23, marked S.sub.0 to
S.sub.k-1. Each symbol 23 is formed by a multiplex of J orthogonal
carriers. Each channel C.sub.i represents a particular data flow
independent of the information transmitted on the other channels.
The header 21 of the frame includes an "empty" or "blank" interval 24 which
is advantageously used to perform both an analog synchronization of the
frame and an extraction of the jamming of the broadcasting channel.
The possibility of achieving an analog synchronization recovery on an
"empty" symbol is a fundamental characteristic of the invention.
For, in existing systems working at a high throughput rate and as shown in
FIG. 3a, the recovery of synchronization is usually achieved in
synchronization at the binary level, on the received train, by means of a
clock 51 working with synchronization means 52 with locking. The recovered
synchronization drives a time base system 53 which opens windows 54 in the
wave train received to extract the useful frames therefrom. This type of
chain with locking of synchronization is made necessary by the need to
work with very high precision, typically of the order of .+-.5 ns for
throughput rates of 10 Mbits per second.
For equal throughput rates, the broadcasting method of the invention makes
it possible to work with considerably lower precision during the recovery
of synchronization. In effect, since each symbol is formed by a multiple
of J orthogonal carriers, the synchronization is achieved on symbols with
a width that is J times greater. Thus, in the case of the use of 448
carriers in parallel, the precision required at the recovery of
synchronization is about 4.5 .mu.s.
The assembly of FIG. 3b corresponds to the implementation of a two-stage
synchronization by recovery of two successive symbols of synchronization.
The first synchronization symbol recovered is the "blank" symbol 24 of the
frame header 21. The detection of an envelope 55 of the blank symbol sets
off the time base 56 which may be generated by a simple quartz-oscillator
57 at 12.5 kHz for symbols with a duration of 80 .mu.s. The time base 56
opens windows 58 in the digital train received so as to recover the second
recovery symbol 25.
This symbol is formed by an unmodulated multiplex of the J carrier
frequencies. It advantageously takes the form of a wobbling on the entire
spectrum covered by the carriers but may be formed by any multiplex with a
substantially constant envelope.
The aim of the second stage of synchronization is to make a more precise
resetting of the synchronization acquired at the first stage, by analysis
of the pulse response of the channel. The self-correlation of the wobbled
symbol 25 thus enables increased precision in synchronization to be
obtained. The detection of an envelope 61 of the second synchronization
signal, after analysis 62, resets the time base 56, and hence the
sequencing of the window 58 openings in the wave train received. The
recovery of this symbol, with a duration of 80 .mu.s, accomodates a
precision of .+-.2 .mu.s, and is therefore compatible with an analog
recovery chain.
The analysis of the pulsed response of the channel makes it possible to
take into account echo phenomena for the synchronization. Furthermore, a
safety interval is advantageously provided between each symbol of the
frame, with a view to absorbing this echo effects and limiting the
intersymbol interference phenomenon. The safety interval typically has a
value of 16 .mu.s, reducing the useful symbol period to 64 .mu.s.
The synchronization symbols 24, 25 of the header 21 of the frame may
further each have a distinct second function.
The blank symbol 24 may, in effect, serve to analyze the interferences and
jamming that affect the transmission channel in order to take them into
account, at the receiver, in the soft decision module as shall be seen
further below.
The wobbled synchronization symbol may, for its part, serve as a phase
reference for the decoding of the useful signals 23 of the frame. In
effect, advantageously, the reference phase of each of the J carriers of
the multiplex is locked in a distinct and specific way, so as to make it
possible to restore each component of the multiplex to the receiver in
differential demodulation. Advantageously, the locking of the reference
phases is expressed by the formula:
.phi.k=.pi.k.sup.2 /N
with k=0 to N: index of each frequency.
N: total number of frequencies of the multiplex (N=512 in the present
example).
Any other mode of computation of the locking phases is suitable, provides
that it makes it possible to discriminate the information conveyed by each
of the carriers of the multiplex.
If necessary, the header 21 of the frame has a third symbol 26 which is a
carrier of information such as the list of the local frequencies of
emissions for the channel considered. A mobile receiver is then capable of
getting automatically and permanently locked into the most powerful local
transmitter, by means of a specific device for the analysis of this
information.
In all, the principle of analog and implicit synchronization of the frames,
in the invention, makes it possible to avoid the drawbacks of the existing
systems using synchronization words that are recognized at the binary
level (consumption of throughput, risks of poor recognition of the word,
total loss of the frame in the event of synchronization error). This
determining advantage is added on to the optional possibility of a
bi-functional use of the symbols of synchronization, as has just been
presented.
The frame structure thus achieved results, according to the invention, from
a dual operation of temporal interlacing and frequential interlacing of
the source sequence (FIG. 4).
The number of binary elements per channel of one and the same frame coming
to the input of the interlacing system depends on the number J of carriers
per symbol, the number K of symbols per channel and the number of states
of the modulation applied to each carrier. In the case of a modulation
with four phase states, the size of the blocks P.sub.l of data (l
designating the index of the frame) presented at each frame at the input
of this system is 2.multidot.J.multidot.K bits.
By way of example, if J=448 and K=9 (k=0 to 8 designating the order number
of the symbol in the channel), we obtain blocks P.sub.l of data equal to 8
064 bits.
Let P.sub.l,i be the index i bit of the block P.sub.l (i=0 to
2.multidot.J.multidot.K-1)
The temporal interlacing consists in forming a block Q.sub.l, of the same
size as P.sub.l, the index i of which, marked Q.sub.l,i is defined as
follows:
Q.sub.l,i =P.sub.l-f(i),i.
On the choice of the function f(.) depends the depth and efficiency of the
interlacing. In general, the image of f is the set F={0,1 . . . , m-1},
where m designates the temporal depth of the interlacing.
An example of an open-ended temporal interlacing is shown in FIG. 5. The
example shown is of the type applying to a process with inversion of bits
defined by the following interlacing function:
n: reciprocal number associated with n:
if n is a number varying from 0 to 2.sub.p-1, written in the form:
##EQU1##
the associated number n is equal to:
##EQU2##
The block P.sub.l is interlaced according to the diagram of FIG. 5 so as to
form a block Q.sub.l. The double change-over switch 81, 82 symbolizes the
application of the interlacing function by successive switching over of
each of the delay blocks 83. Let q.sub.l-1 be the i.sup.th element of the
block Q.sub.l. We have the relationship:
q.sub.l-1 =P.sub.l-R(i/16),i.
The depth of interlacing is therefore 16 frames.
Clearly, this example is given purely as a non-restrictive illustration.
The elements of the block Q.sub.l are assigned to the K symbols of the
channel considered in the frame 1 as follows:
The block Q.sub.l is split up into K packets of 2J bits in ascending order
of the index i and these packets are assigned to the K symbols of the
channel considered, according to the following principle:
##STR1##
The frequential interlacing consists in assigning the 2J bits of each
packet to the J carriers forming the symbol associated with the packet
considered. These 2J bits are assembled in J pairs which are bijectively
associated with the carriers of the symbol, according to the particular
relationship that defines the interlacing.
An example of frequential interlacing is illustrated in FIG. 6 wherein:
i represents the index of the elements of the sequence that have undergone
the temporal interlacing 41 and are introduced in the frequential
interlacing module 42;
.tau. represents the index of the elements after frequential interlacing
42;
the column .tau.=F(i) illustrates the implementation of the interlacing
function by inversion of bits on the indices of the 512 carrier
frequencies forming each multiplex.
It will be noted that each elementary symbol of modulation is formed by a
selection of 448 carriers forming a sub-set of the set:
{f.sub.j =f.sub.0 +jDf}(j=0 to 511)f.sub.o designates an arbitrary
frequency and Df the difference between each carrier.
This sub-set is the set of carriers fj, the index j of which meets the
condition (1):
3.ltoreq.j.ltoreq.480,j.noteq.256 (1)
This choice is warranted by constraints related to the feasibility of
certain analog functions of the receiver. The elimination of the central
carrier overcomes the problem of the continuous drift of the analog
digital circuits, and the elimination of the lateral carriers of the
spectrum (7/8 of the total spectrum) overcomes the edge effects of the
cut-off filters.
Let j be the i+1.sup.th number meeting the condition (1) in the list of the
indices 0 to 511 classified in their reciprocal ascending order. This
relationship defines the function j=F(i). The frequential interlacing is
characterized by by the relationships:
if E(i/448) is an even value, then u.sub.j,k =q.sub.l,i
if E(i/448) is an odd value, then v.sub.j,k =q.sub.l,i
with i=0 to 8 063, k=E(i/896) and j=F(R(i/448))
and with E(p/q): integer part of p/q.
R(p/q): remainder of the division of p by q.
In these relationships, (u.sub.j,k, v.sub.j,k) designates the couple of
binary elements determining the phase of the carrier fj of the order k
symbol. (Each carrier undergoes a four phase state modulation).
In other words, each of the pairs of binary elements is formed by the
digital elements source sequence in forming packets of 2M constituent
elements in said sequence, and in associating the elements two by two in
each packet according to a criterion of maximization of dispersal of the
adjacent digital elements of the source sequence. The pairs of digital
elements are formed by splitting each of the data packets into two
half-packets and by matching the same order digital elements in each
half-packet.
According to an essential characteristic of the invention, provided that
the temporal interlacing function F and the frequential interlacing
function G are appropriately chosen with respect to each other, the
de-interlacing operation is done implicitly by the application of the
combined function (GoF).sup.-1. This results, firstly, from the simplicity
of synchronization described further above, which enables the implicit
obtaining of the sequence received from the transmission channel 43 with
immediate knowledge of the index of the symbols in the sequence and,
secondly, the complementarity of the two functions of temporal interlacing
and frequential interlacing.
The reconstruction of the blocks Q.sub.l is then done very simply in the
de-interlacing module in using the bijective character of the frequential
interlacing. With the receiver using a differential demodulation, the data
are restored without any problem of phase ambiguity.
The principle of the temporal de-interlacing 45 consists in the
application, to the binary elements of each block Q.sub.l, of the
complementary delay with respect to the depth of the interlacing of the
delay undergone at transmission. The knowledge of this complementary delay
is expressed by m-f(i)-1, and is deduced directly from the index i of the
binary element and a priori knowledge of the function f(.). No
synchronization other than that of the multiplex itself is needed to do
the de-interlacing.
The diagram of FIG. 4 shows this mechanism for a single symbol of
modulation.
Should the channel decoder work in soft decision mode, the de-interlacing
is actually applied not to binary elements but to words (generally
four-bit words) representing the estimation, by the demodulator, of the
bits received.
As mentioned further above, the blank symbol 24 of the frame header 21 can
be used to identify and characterize the jamming of the transmission
channel, and to take it into account in the restoration of the signal
received notably within a soft decision decoding process.
FIG. 7 shows the reception branch of the transmission system of the
invention. The hatched modules therein illustrate the implementation of
this supplementary function of taking the jamming into account with
respect to the known reception chain of FIG. 1.
The explanation of FIG. 7 first of all requires reminder of the chief
characteristics of the signal transmitted in general.
The signal transmitted is formed by a sequence of modulation symbols
forming a multiplex of N orthogonal carriers.
Let f.sub.k be the set of carrier frequencies considered with:
f.sub.k =f.sub.o +k/T.sub.s,k=0 to N-1
where T.sub.s represents the duration allocated to a modulation symbol.
We then define an orthogonal base of elementary signals
.psi..sub.j,k (t) with k=0 to N-1, j=-.infin. to +.infin.
.psi..sub.j,k (t)=g.sub.k (t-jT.sub.s)
with 0.ltoreq.t.ltoreq.T.sub.s :g.sub.k (t)=e.sup.2i.pi.fkt
elsewhere: g.sub.k (t)=0.
Let us then take a set of complex numbers C.sub.j,k taking its values in a
finite alphabet, and representing the signal of transmitted data.
The associated OFDM signal is written:
##EQU3##
In the case concerning this application, and cohesively with the preceding
descriptions, the transmitted signals C.sub.j,k have a constant modulus.
This means, in other words, that each of the carriers of the multiplex
undergoes a phase modulation.
The transmission channel can be modelized according to the relationship:
Y.sub.j,k =H.sub.j,k C.sub.j,k +N.sub.j+k
where H.sub.j,k is the complex response of the channel at the point (j,k)
of the time-space frequency, and N.sub.j,k is a complex Gaussian noise
with:
N.sub.j,k =N.sub.Ij,k +iN.sub.Qj,k
and
E(N.sub.Ij,k).sup.2 =E(N.sub.Qj,k).sup.2 =.sigma..sup.2.sub.j,k
where E() represents the mathematical expectation.
It can then be shown that the implementation of a decoding according to a
maximum likelihood criterion a posteriori consists in the maximization, on
C.sub.j,k, under the constraint of the code of linkage of the symbols
C.sub.j,k of the expression:
Re(Y.sub.j,k H*.sub.j,k C*.sub.j,k /.sigma..sup.2.sub.j,k)
where Re(.) represents the real part of a complex number.
The essential element of this analysis relates to the fact that the noise
power .sigma..sup.2.sub.j,k at every point (j,k) of the time-frequency
space coming into play in the decoding process. When the code linking the
elements C.sub.j,k is a convolutive code and when the decoder used is a
soft decision Viterbi decoder, the knowledge of the noise power generated
by the channel and by the reception device therefore forms a major
weighting parameter with respect to the optimization of the decoding. This
parameter does not come into play in the particular case of a white noise,
such that:
.sigma..sup.2.sub.j,k =.sigma..sup.2, irrespectively of j and k.
However, if a jammer affects the signal, the weighting has the effect of
"erasing" the corresponding carriers in varying degrees in the same way as
a fading on these same carriers. This property is a specific feature of
the COFDM system which makes it extremely attractive in channels highly
disturbed by industrial stray signals or noise, the nature of which may be
pulsed or recurrent in frequency (localized field of the time-frequency
space).
The implementation of this process of measurement and identification of the
jamming, then of weighting of the resemblance coefficients consists in
performing a spectral analysis of the noise on the empty symbol 21.
This analysis is achieved through a discrete Fourier transform 71 using the
digitized signal 72 obtained at the output of an ADC 73.
If {f.sub.k }.sub.k=0, . . . , N-1 designates the set of carriers used in
the COFDM signal, it would appear to be necessary to analyze the noise on
a comb of spectral lines, of which {f.sub.k } forms a sub-set. Since,
furthermore, a Fourier transform is used for the demodulation of the
signal proper, the use of the same transform for the spectral analysis 71
of the noise is a technical solution that is perfectly suited to the
problem and does not call for the implementation of any additional
functions.
This operation should be complemented by a bi-directional filtering in the
time-frequency space, the role of which is to provide for a smoothing of
this measurement in order to obtain an estimation of the mean value of the
weighting parameter.
The computation of the parameter 1/.sigma..sup.2.sub.j,k (75) gives a
complementary piece of information of weighting of the metrics 76
associated with the demodulated symbol and designed to be used in the soft
decision decoding step 17 (Viterbi decoding).
* * * * *
|
|
 |
|
 |
Description |
|
