DE19528068A1 - Data transmission method using multi-frequency QAM - dividing data bits into bit groups, obtaining amplitude values from sampled bits, and using amplitudes to modulate orthogonal carrier oscillations - Google Patents

Abstract

The method involves transmitting data signals by using a multi-frequency QAM principle. At the transmitter end, a number of consecutive data bits are divided into a number N of bit groups. The bit groups are sampled at a speed fT corresp. to the time frame T=1/fT of the data signals. Amplitude values are obtained from the sample values, and the amplitude values are used to modulate the orthogonal carrier oscillations of the same frequency. The frequency distance of adjacent carrier oscillations is equidistant or smaller than the sampling frequency. The resulting signals are added to the carrier oscillation mixture to be transmitted. At the receiver end, the transmitted signal is selectively filtered for the individual N bit groups. The data bits corresp. to the bit groups are obtained after sampling in a time frame T = 1/fT.

Description

Previously known modulation methods either work with Carrier vibrations of only one frequency (e.g. QAM) or with a plurality of carrier vibrations having frequencies gen (multifrequency method) In the latter is generally mean a variety of carrier frequencies (e.g. 256 or 512) is selected in order to achieve the so-called FFT algorithm for the realization of the dis to be able to apply the Fourier transform. Herewith there are advantages in terms of the number of multiples cations compared to monofrequency modulation ren. The carrier vibrations must ver evenly divides and its number is a power of two correspond. At a fixed time interval (according to the walking speed) the tears vibrations in phase and / or amplitude with the data converted in parallel. The pace This must exactly match the frequency spacing of the individual carrier vibrations match. With increasing The number of carriers therefore decreases for a given bandwidth Walking speed. Better use of the existing which bandwidth is not associated with it.

It is an object of the present invention to find a way show how a data transmission method with a more number of carrier frequencies can be formed to a to make better use of the available bandwidth.

This problem is solved by the in claim 1 given procedural features. The invention brings the advantage that the frequency spacing of the carriers vibrations less than the walking speed of the Is data signals and thus the bandwidth for a given type  number of different carrier frequencies better out can be used. Since the FFT algorithm cannot be used here the number of carrier frequencies is also necessary not to correspond to a power of two. So that the number the carrier frequencies are chosen so that by the fixed this number, the noise behavior is optimized.

Advantageous embodiments of the method according to the lying invention result from the patent claims 2 to 7.

In the following, the present invention is based on a theoretical derivation and Darge in drawings presented embodiments described in more detail.

Show:

Fig. 1 is a block diagram of a multi-frequency QAM transmitter,

Fig. 2 shows a first embodiment of a frequency-Mehrfre QAM receiver,

Fig. 3 shows a second embodiment for a multi-frequency QAM receiver and Fig. 4 shows a third exemplary embodiment for a multi-frequency QAM receiver.

The following is the transmission method theoretically derived according to the present invention.

A transmission signal is composed of individual carriers as follows vibrations together:

Consider a sample within the time interval valls kT and kT + T,

then you get:

where ω _i = 2πf₂ was set. The values f _i represent the individual carrier frequencies. The amplitude values a _i and b _i are obtained from the data to be transmitted with the aid of a suitable coding. In a four-valued transmission method with a _i , b _i ε {+ 1, -1} z. B. assigned two data bits each pair of numbers a _i , b _i .

Another transformation of (2) leads to T = 1 / f _T

Here, the expressions

and

than in the complex number plane around the angle

rotated pairs of numbers a _i (k), b _i (k). It is therefore considered to be a sample of the transmission signal

A transmitter structure shown in FIG. 1 can be derived on the basis of this equation (5). In this figure, a QAM multi-frequency transmission device is shown as an example, which receives data signals to be transmitted serially and receives them first, for example with the aid of a shift register having parallel outputs, into parallel data signal groups, ie into parallel bit groups. N separate QAM signal branches are connected to these parallel outputs, each of which is assigned to one of N different carrier frequencies. Each of these QAM signal branches has on the input side an encoder connected to outputs of the shift register. The number of connected outputs and thus the number of data bits to be converted into an above-mentioned number pair a _i , b _i depends on the set of values of the transmission method used. In a previously mentioned four-value transmission method, each of the encoders is connected to two outputs of the shift register, as is indicated as an example in FIG. 1. In this example, it is also assumed that in the respective encoder (i), in addition to the determination of the number pair a _i , b _i, its complex rotation according to Eq. (4a) and (4b) is carried out so that the coded transmission data words α _i , β _i are provided by the respective encoder at separate outputs. It is advisable to make certain restrictions with regard to the choice of carrier frequencies and walking speed. Is z. B. 90 ° symmetry for the values α _i (k), β _i (k) in the complex plane is required, then the condition

with a suitable value for λ, κ = 0, 1, 2
be fulfilled. The condition is obtained from this

f _i = f _T ^λ / ₄ with λ = {1, 2,. . . }.

The values α _i (k), β _i (k) provided in this way by the encoder of the respective QAM signal branch (i) are sampled by a schematically indicated scanning device. The resulting sample α _i (k) or β _i (k) is then in accordance with equation (5) with the instantaneous value of the carrier oscillation cos (2πf _i / f _{T *} j / w) or sin (2πf _i / f _T ^* / w) multiplied so that the respective QAM signal branch (i) provides two orthogonal carrier oscillations modulated with the sample values α _i (k), β _i (k).

The provided by the individual QAM signal branches modulated carrier vibrations become a summing device device Σ fed, which additively these carrier vibrations overlaid. The resulting carrier vibration Ge is then mixed using a digital / analog converter D / A fed to a transmission medium.

To derive a suitable recipient structure all samples within a step interval

kT t _i <(k + 1) T

combined into a column vector. Received in matrix form one with Eq. (5) the following expression:

In short form:

y = CD, (7)

wherein the column vector y contains the samples of the transmission signal, the matrix C the trigonometric function values and the column vector d the rotated numerical values α and β according to equation (6). This system of equations can be solved by the method of the smallest quadratic error. One first obtains by multiplying by the transposed matrix C ^T

C ^T · C · d = C ^T y (8)

From this one obtains for the data vector

d = [C ^T · C] ^-1 C ^T · y (9)

This relationship results, for example, in a receiver structure according to FIG. 2. Thereafter, the analog received signal is first converted digitally with the aid of an analog / digital converter A / D and then fed in parallel to a number N of receiving branches corresponding to the number N of the carrier frequencies . Each of the receiving branches has a pair of filters connected in parallel on the input side, a filter of the receiving branch (i) denoted by h _i having the above-mentioned sample value α _i (k), the remaining filter q _i on the other hand having the sample value β _i (k) is arranged.

The filtered signals occurring at the output of the filter pair hi , qi of the receiving branch (i) are sampled by a schematically illustrated sampling device in a time pattern corresponding to the step duration T of the data signals in order to obtain the samples α _i (k), β _i (k). The non-recursive filtering corresponds to a vector multiplication, as is to be carried out according to equation (9) for the determination of each data value. The filter coefficients of the filters hi , qi are defined by the two vectors h _i and q _i resulting from equation (9).

The two sample values α _i (k), β _i (k) provided by the filter pair hi , qi are rotated by the phase angle 2πf _i / f _{T to} recover the data values α _i and b _i , which is a complex multiplication by exp (2πf _i / f _T ) speaks accordingly. For this purpose, a corresponding multiplier is indicated in each of the reception branches (i) in FIG. 2. With the aid of a decision logic connected downstream of the respective multiplier, the received data values a _i and b _i are finally obtained and, after a decoding (not shown), data bits which match the transmitted data bits in the case of interference-free transmission.

The complex multiplication with exp (2πf _i / f _T ) can be simplified if the order of phase shift and decision is reversed, since the decided values can be represented with a shorter word length. FIG. 3 shows a receiver structure corresponding to the receiver structure shown in FIG. 2, in which this interchange is carried out.

Another embodiment for a receiver structure is obtained when the system of equations (8) is transformed. With the help of an appropriate transformation, the Vek goal equation on the form

Dd = By (10)

are brought, the matrix D a lower triangular dimension trix with the values 1 as diagonal elements. With The received data values such as determine as follows:

With y _j the samples of the received signal and b _ÿ and d _{ÿ denote} the elements of the matrices B and D be. If the decided values d _j are used for the execution of equation (11), this results in a decision feedback feedback structure indicated in FIG. 4. With this receiver structure, the analog received signal is digitized by an analog / digital converter A / D and a number of filters corresponding to the number of samples to be determined is supplied. Since, in the assumed example, N carrier vibrations are provided and two samples α, β (α₁,..., Α _N , β₁,..., Β _N ) are to be determined for each of these carrier vibrations, there are 2N with g _i to g _2N designated Fil ter whose filter coefficients are defined by equation (11) and which each form the input of a separate receiving branch. In each of these reception branches, the associated filter is followed by a scanning device which is illustrated schematically in FIG. 4. This scans the signal value that occurs at the output of the associated filter and changes over time in a time pattern corresponding to the step duration T of a data signal. This scanning takes place synchronously in all receiving branches. In addition, decision logic is provided in each of the receive branches.

As can be seen from equation (11) and the formation of the matrix D as a triangular matrix, in the first of the reception branches the received signal filtered and filtered by the filter g1 is fed directly to the associated decision logic and is used to determine a sample value. This sample may be, for example, the sample value α 1 (k). In the other, each for the determination of one of the other sample values (α₂ (k),..., Α _N (k), β₁ (k),..., Β _N (k)) is provided between the filter and Decision logic provided a summing device. This summing means are accordingly their allocation to the reception branches 2 to 2 N with Σ2 referred to Σ2N. The respective summing device is, on the one hand, the received signal filtered and sampled by the associated filter and, on the other hand, all the sample values determined in the preceding receiving branches are each multiplied by a value of the matrix D assigned to the respective sample value through the filter g2 filtered, sampled receive signal the value α₁ (k) _* d₂₁ supplied. In a corresponding manner, the summing device Σ2N of the receiving branch 2 N with the sampled received signal filtered by the filter g2N and the values α₁ (k) _* d _{2N, 1} ; α₂ (k) _* d _{2N, 2} ; . . . ; α _2N-1 (k) _* d _{2N, 2B-1} applied.

The determined in this way in the individual reception branches samples α₁ (k),. . . , β _N (k) are then fed to a connection logic VL, which samples the carrier frequencies f 1,. . , f _N assigned value pairs α₁ (k), β₁ (k); . . . ; α _N (k), β _N (k) forms. The individual pairs of values α _i (k), β _i (k) (i = 1,..., N) are then each separated by a phase angle 2πf _i / f _T (i = 1,..., N ) rotated in order to recover the originally transmitted data values a _i , b _i (i = 1,..., N) and the original data bits by decoding (not shown). The phase shift corresponds to a complex multiplication of the respective pair of values α _i (k), β _i (k) by exp (2πf _i / f _T ), as shown in FIG. 4 by the connection logic VL connected to the individual Multipliers assigned to pairs of values is specified.

Even with the use of a receiving device according to FIG. 4, it is expedient, with respect to step speed and carrier frequencies comply with certain conditions, in order to simplify the complex multiplications.

With 90 ° symmetry:

f _i = f _T ^λ / ₄ with λ = {1, 2,. . . }

If you choose carrier frequencies with equidistant distances, then follows

f _i = f _u + i · Δf with i = 1,. . . , N

To comply with the above condition, the following applies first

The following are possible for the frequency spacing keiten:

• a) Δf = f _T. With f _u = 0, this corresponds to the usual multi-frequency process.

With regard to the noise behavior, it is advisable to choose f _i = 3/4 f _T when using the receiving devices according to FIGS. 2 and 3. When white noise is fed in, there is an increase in noise of <4 dB at N = 4 carrier frequencies compared to f _i = f _T.

The realization cost of the receiving device of FIG. 4, although higher than for the receiving devices of FIGS. 2 and 3, but here there are advantages be züglich of the noise behavior.

In a receiving device according to FIG. 4 3/4 f _T versus .DELTA.f = f _T rises for .DELTA.f = the noise figure only slightly with to participating carrier signal Of. For N = 5 this results in a value of approximately 2 dB. The bandwidth saving is about 30%. A further reduction in the noise figure can be achieved in the receiving device according to FIG. 4 by changing the sequence when determining the sample values α 1,. . , β _N can be achieved. If the first carrier frequency is only subjected to a one-dimensional data value (d 1 = 0), the noise figure can still be greatly reduced. It should be noted that the transmission capacity is correspondingly lower. This results in N = 4 carrier frequencies and Δf = 1/2 f _T, a noise increase of less than 4 dB. Taking the bandwidth savings into account, this is an acceptable value.

In the derivation of the receiving devices explained above, an ideal transmission medium was assumed, ie the received signal was assumed to be undistorted. The specified receiving devices can also be used for transmission media that are not free from distortion. The impulse responses of the
10 non-recursive receive filters can be extended beyond the step duration (number of coefficients <w).

In addition, it is appropriate to use the filter coefficients using adaptive iterative methods len in order to optimally adapt to the transmission medium to guarantee.

Finally, it should be pointed out that it is the previously only generally mentioned data signals can trade any digital information under which for example, text and video signals and voice formations in digital form.

Claims (7)

1. Method for transmitting data signals according to the multi-frequency QAM principle,
in which a fixed number of successive data bits is subdivided into a plurality N of bit groups, each with the same number of data bits,
each of the N bit groups are assigned two orthogonal carrier oscillations of the same frequency (f _i , with i = 1,..., N), the frequency spacing (Δf _i ) of neighboring carrier oscillations is in each case equidistant and less than the step speed (f _T ) of the data signals is
the data bits belonging to the N bit groups are scanned simultaneously in a time pattern (T = 1 / f _T ) corresponding to the step speed (f _T ) of the data signals,
from the samples obtained for the respective bit group by coding, two amplitude values (α _i , β _i ) forming an amplitude pair are derived, by means of which one of the two orthogonal carrier oscillations assigned to the respective bit group is modulated,
a carrier vibration mixture is formed and transmitted by additive superimposition of all modulated carrier vibrations,
and in which the transmitted carrier oscillation mixture is filtered selectively for the individual N bit groups at the receiving end,
the filtered signals obtained therefrom for the N bit groups are simultaneously scanned in a time grid (T = 1 / f _T ) corresponding to the step speed (f _T ) of the data signals
and then the N amplitude pairs (α _i , β _i ) from the sampled values and by decoding them the data bits belonging to the N bit groups are recovered. 2. The method according to claim 1, characterized, that the frequency spacing of the carrier vibrations is equidistant is set. 3. The method according to claim 1 or 2, characterized in that the number of bit groups and thus the number of carrier vibrations (f _i ) and the frequency spacing (Δf _i ) be adjacent carrier vibrations are determined in accordance with the required noise behavior. 4. The method according to any one of claims 1 to 3, characterized in that the frequency spacing (Δf _i ) adjacent carrier vibrations gene 3/4 corresponds to the walking speed (f _T ) of the data signals. 5. The method according to any one of claims 1 to 3, characterized in that the frequency spacing (Δf _i ) of adjacent carrier vibrations corresponds to 1/2 the step speed (f _T ) of the data signals. 6. The method according to any one of claims 1 to 5, characterized, that the number of bit groups and thus the number of the sen assigned carrier vibrations is fixed with N = 4. 7. The method according to any one of claims 1 to 5, characterized, that the number of bits groups and thus the number of bits its assigned carrier vibrations is set to N = 5.