Note: Descriptions are shown in the official language in which they were submitted.
i38~
This invention relates to a method for transposing sub-
bands of equal width in a signal frequency band by means of modu-
lators followed by equal band-pass filters and additional modula-
tors.
Such methods are known, for example, from German Patent
24 26 451 which describes a circuit arrangement for obtaining fre-
quency scanning, independent of any transportation, for the harmon-
i.cs of basic speech-frequencies in sub~band transposition devices,
wherein the speech-signal-band is divided, at the transmitting end,
into sub-bands of equal width and a transmission~band is formed
by transposing the sub-bands, the sub-bands contained therein being
adapted to be inverted and the said transposition being reversed at
the receiving end; the said circuit arrangement being characterized
in that the ratio between the frequency and the lower limit of the
speech-band and half the width of the sub-band being as two whole
numbers to each other. This circuit arrangement comprises five
modulators which, by means of suitably selected upper carrier-frequ-
encies at the outputs from the subsequent unit-filter, break the
speech~band down into five sub-bands; with the aid of five more sub~
sequent modulators, the sub-bands thus obtained are converted back
into the original frequency-position in the speech band, and trans-
position of the sub-bands is achieved in that the modulators are
supplied with a selectable choice of five out of ten carrier-frequ-
encies.
DE-OS 26 52 607 describes a procedure for the concealed
transmission of information bearing signals, in which the known
processes of time scrambling and frequency band switching are inte-
,, ,','!, ' 1 ~
,
3~
grated. Here, too, -the costs involved for ring-type modulators,
band filters, memory matrices, frequency generators and control
are considerable in the analogue version~ In the digital version,
which operates according to the Weaver phase method it is true
that there is only one input and one output multiplier, but these
operate at high multiplication rates and they obtain their single
carriers from a sine/cosine table.
U.S. patent 2,408,692 describes a transmission system
that involves concealment by sub-band transposition in the frequ-
ency range of one of the sub-bands and subsequent transposition
into length-modulated pulses that are sent in time multiple.
The known band-pass method involves relatively high
circuit-costs~
It was therefore the purpose of the present invention to
provide a method of sub-band transpos.ition of the kind mentioned
above wh.ich can be achieved at less cost and permits discontinuous
ox dig.ital signal-processing.
The mekhod according to the invention .;s characterized
in that specific carrier-frequencies are permanently assiyned to
the modulators and in that arranged between the band-pass filters
and subsequent modulators is a switching matrix for scrambling the
sub-bands.
The method according to the present invention shows how
the frequency values for the carrier oscillations are to be selec-
ted, depending on the division into sub~bands, as well as the
carrier frequency, so that a time-discrete or digital signal
processing can be carried on in an inexpensive manner. In this
2 -
~G3~
connection, both s~uare as well as sinusoidal carriers can be used
at relatively low scanning frequencies and with simple modulator
circuitry. In a further form of the invention, values that are
favourable for the encryption of a CCITT speech channel are cited,
this being divided into five sub-bands.
The method according to ~he invention has the advantage
that there is no need to swi~ch over between the carrier-frequencies
fed to the modulators and this results in a saving in circuit-costs.
In a further development of the invention, the pass-band of the
band-pass filter is applied to one o~ the sub-bands, thus dispensing
with two modulators and the production of a carrier-frequencyO
~ccording to further configurations of the invention, information
is given for selection of frequency-values for carrier-waves as a
function of division into sub-bands and of scanning frequency, thus
making it possi~le to carry out discontinuous signal-processing
in an advantageous manner. In this connection it is possible to
use both square and sinusoidal carriers at relatively low scanning
frequencies and simple modulator circuitry. According to still
~urther configurations of the invention, favourable values are
yiven f~r encoding a CCITT speech-channel (CCITT= Comite Consultatif
International Téléphonique et Télégraphique) which is divided into
five sub-bands.
The invention is described hereinafore in conjunction with
the drawing attached hereto, wherein:
Figure 1 shows a block-diagram of a system according to the
invention for encoding speech by means of five-band transposition;
Figure 2 shows spectra according to fre~uency-inversion
2a
3~
with four sinusoidal carrier-waves and discontinuous si.gnal proces-
sing;
Figure 3 shows an example of a modulator arrangement;
Figure 4 shows spectra according to frequency-inversion
with a square-wave carrier and con~inuous signal processing.
The basic block circuit-diagram of a speech-encoding sys-
tem based upon five-band transposition is shown in Figure 1. The
input spectrum of speech-signal Xin~f) to be encoded may be burden-
ed with unwanted amounts of
- 2b -
~.~8~38~
higher-frequency spectra which must be suppressed, possibly by a low-pass
fi]ter TP 1 at the input-end so that, for example, with discontinuous signal
processing, with predetermined scanning frequency fS, the scanning theorem
may be maintained. The resulting band-defined spectrum F(f) is divided, in the
next step, with the aid of similar band-pass filters, into five sub-bands. This
requires prior modulation with appropriate carrier-frequencies ~ 2 etc..
'I'he actual scrambling takes place in the next step by means of a swîtching
(scrambling) matrix M controlled by a random-number generator ZG, sub-bands
Xv (f) v = 1, 2 ... 5 being transposed at random in short time-intervals accor-
ding to (i, j, k, 1, m)C~l, 2, 3, 4, 5). Sub-bands ~Xv ~f) for v = i, j, k, 1,
m) are then transposed again with the same frequencies as the input-modulators
and thus appear in scrambled sequence in the original frequency-range of speech-
spectrum X(f). Possible harmonics of the output-spectrum, for example un-
wanted products of modulation or parts of the spectrum recurring periodically
during discontinuous signal processing, are eliminated by low-pass filter TP2.
Y(f) is the spectrum of the encoded continuous output signal. The system for
decoding encoded speech signals also has the block-structure shown in Figure 1
but, or correct decoding, random-number generator ZG at the receiving end must
have the same random sequence as the transmitter and must be synchronized
therewi~h.
The m~thod according to the invention also makes it possible to
dispense with two modulators if the pass-band of unit band-pass filter BPl to
BP5 is applied to one of the five sub-bands and if one of the carrier-frequencies
at the input and output-end receives the value 0.
Figure 2a shows the speech-signal spectrum (telephone quality), the
normal position of the continuous signal being fully represented and the addi-
tional parts of the spectrum appearing symmetrically with whole-number multiples
-- 3 --
3~C3
of scanning frequency fS (here the inverted position is shown below
fS) during discontinuous processing.
The CCITT speech-channel is established within the limits
of 0.3 to 3~4 kHz and, as a result of the division into five sub-
bands, sub-band width B = 620 ~Iz. The arithmetical mean frequencies
of the sub-bands are then:
fM~' = (2v - 1) . 2 ~ 300 Hz (1)
for v = 1 to 5. In the case of simple carrier-frequency production,
and for discontinuous signal-processing, it is desirable to modify
the mean frequencies slightly with:
fM = v . b (2)
The speech-spectrum to be scrambled is thus extended from B2 to
112-(0.31 to 3.41 kHz).
In order to transpose the sub-bands of the spectrum acc-
ording to Figure 2a, pertaining to the continuous signal, to the
same fre~uency-band by means o~ square wave carriers, basic frequ-
encies fv for v = 1 to 5 must be arranged e~uidistantly at inter-
vals B. ~ccordiny to the formula:
~v (nv ~ 1) . B
for nv = ~ 1, 2 etc. v = 1 to 5, 1 = ~ IN and i = 0, 1, ... , 1 - 1
as freely selectable parameters. In the simplest case one obtains
fv = nv . B for 1 = 1 and i = 0. In this case, the minimal value
of the square wave carrier frequencies must not exceed 4~, other-
wise spectral overlapping occurs as a result of the Hermiticity of
the spectra and this interferes with the signals. This condition
also appears as a result of the line-spectrum of the square wave
carrier according to Fourier. The above-mentioned restriction does
- 4 -
not apply for nv = corresponding to fv = ~ since in this case
there is no frequency-inversion. The overlappings are clear to
see in Figures 2b to 2d, in which a frequency inversion is indica-
ted by f2 = 2B, f3 = 3B, f4 = 4B, and they are indicated as foldover
products.
A foldover product can also be seen in ~igure 2e; this
is not, however, disruptive since it is above the carrier frequency
f9 = 9B.
Figure 3 shows a simple modulator with 2 multipliers,
which manages simply with the two values V2 and 1, each with two
signs and 0, and in which the scanning value x(kT) that is to be
evaluated is passed by a selector switch, which is controlled by
the drive logic of the analogue switch, to the corresponding one
of the two inputs of a multiplier amplifier or to zero potential~
Figure 4 shows in detail the spectral relationships for
continuous
4a -
3~3~
signal processing.
The same spec~ral relationships may be achieved, in the frequency-
range reproduced in Figure 4, even with a discontinuous system, i-f scanning
frequency fs is suitably selectedO In this connection, fs mus~ be such that all
of the harmonics of all square wave carriers fv for v = 1 to 5 must end up sym-
metrical with half the scanning frequency. This ensures that no further unwan-
ted lines are added, as a result of the periodicity of discontinuous signals, to
the lines in the square wave carrier-spec~rum which are in any case present.
Conditions for this are:
fv m for m IN = 1, 2 O.O ~
In relation to half the scanning frequency, spectral lines then appear
in the scanned square wave carriers at the following frequencies:
for m odd f25 + 2~ f and
,.
for mv even ~ 2~ o fv~
raspectively for ~ an elemen~ from the ser.ies of whole numbers Z = ~...-1, 0, +1
Ø~. The scanning ~requenc~ is therefore:
~s = 2nV . mv B = k . B.
Tabl~ 1
k fs/kHz nl n2 n3 n4 n5
S60 347,2 4 5 - 7 8
lZ60 781,2 5 6 7 - g
1008 624,96 4 7 8 9
1008 624,96 6 7 8 ~ -
1680 1041,6 4 S 6 7 8
Table 1, above, shows cost-effective values for k, relevant scanning
frequency fs, and for nv. Represented in the first four lines are solutions
3~
with no transposition of sub-bands 3, 4 and 5, which saves modulators.
In addition to the reduction in cost, the respective scanning
frequency is lower.
According to the invention, scanning frequencies may be lowered s~ill
further if sinusoidal carrier-waves, which can be assembled into corresponding
series from only a few scanning values differing in amount, are used for the
transposition. Here again, relationships ~4) and ~5) appl~. The following
appl~ to t~e scanning frequenc~:
fs = KGv ~1. nv ~ i} . 1 ~ v = 1, O.. 5
or
fs = ---r-~i k IN.
Because of the scanning theorem, however, k/l > 11 must be maintained~
;ln order to avoid spectrum-disturbing overlapping. Table ~ below shows pos-
sible carrier-frequencies, in rela~ion to scanning frequency fs, as a f~lction
of n, and for favourable values of i and 1, for the values 1/1 = 15, 20, 24,
39 and 36, in five corresponding columns.
Table ~
n fv/fs fv/fsfv~fs fv/fsfv~fs = 1 n - i
00 0 0 0 0
01
02 1/10 1/10 1/12
03 1/6 1/8 1/10 1/12
0~ 1/5 1/6
05 3/10 1/4 1/6
06 3/10 1/~ 1/5 1/6
07
08 1/2 2/5 1/3
~6
E;3~
n fv/fs . fv/fs fv/fs ~v/fs f /fs = 1 O nk ~ i
.. . . .. .
09 3/8 3/10 1/4
1/2 5/12 1/3
12 1/2 2/5 1/3
1/2 5/12
18 1/2
k/l 15 20 24 30 36
0 0 0 0
1 2
It may be gathered from ~able fY that if k/l = 15 with 1 = 2 and i = 1
only four of the permissible carrier-frequencies arise. Similarly, the last
result shown for k/l = 36 with i = 0 is useless, since a maximum cf only four of
the permissible carrier-re~uencies end up in an~ desired re~uency intervals of
width lOB, so that at the most four sub~bands can be separa~ed ~ith equal fil-
ters. Spectra after requençy-transposi~ion are glven in ~igure 2 for ~/1 = 24.
I:n the case of. values k/l = 24 and 307 ~avourable data for an arrangemen~ are
listed in Table Z ~belo~3 ~r the carrier~requencies, again relative to t~e
scanning frequenc~) for indiv.idual sub-bands, as is the mean f.requency of the
band-pass filterO
Table Z
Sub-band fv/fs fv/fs fv~fS v/
. .
lf~ l/6 113, 114 l/5 3/lo
2 1/8 3/~ , 3/lQ l/3, 1/5
3 l/3 l/6 1lr3 l/6
4 3~8 l/8 l/lo 2/5
0, 5/12 1/2, 1/12 2/5 1/l~
~7
386~
Sub-band fv/fS fv/fS f fv/fS
M 5 . B 7B 7B 8B
k/l 24 24 30 30
As regards discontinuous signal processing, the most cost-effective
case arises with the choice of the follo~ing band-mean~frequency of the band-
pass filter:
f~ = 5 B,
since in this case a sub~band signal can be obtained with no frequency transpo-
sition.
The modulators may be in the ~orm o a multiplying circui~ which
multiplies the signal for fv/fs~ or whole number multiples thereof~ in ~ime wi~h
the scanning frequency, according to Table 3 below, with the factors given
therein:
Table 3
fv/fs 1/2 1/3 1/~ 1/5 1/6 1/8 1/10 1/12
0 ~`1 0 0 -1
l1actors -1 0.5 0 1 ~0.5 ~ 0.268
-0~618 ~~ +0O61~ +0.732
