Apparatus for compensating amplitude and phase distortions in a common-wave radio system.

The invention relates to an arrangement according to the preamble of claim 1. B. are lines that are used as modulation feeders in radio networks. Analog signals that pass through the transmission link suffer amplitude and phase distortions. According to the prior art, these are equalized with fixed equalizer amplifiers.

The object of the invention is to provide an arrangement of the type mentioned at the outset which allows automatic, fast and accurate adaptive equalization and is of comparatively simple construction. The invention is described in claim 1. The further claims contain advantageous developments or developments of the invention.

In a single-wave radio network, the modulation signal is transmitted from a central station via modulation feeder lines to several transmitting stations, from which the signal is emitted with the same HF frequency and the same LF amplitude, LF phase and the same time position. In order to guarantee undisturbed and interference-free reception, the distortions and different runtimes of the modulation feeders must be compensated for. According to the prior art, this is done with the aid of suitable networks for equalizing the frequency response and for equalizing the signal transit time, which are individually adapted to the characteristics of the particular modulation feeder by manual comparison. Any change in the properties of the modulation feeder requires a new manual adjustment of the networks. It is obvious that this method is only suitable for dedicated lines or fixed radio connections and requires a lot of personnel to calibrate and maintain the compensation networks.

The advantages of the solution according to the invention are first of all in the saving of the engineering work required to measure the properties of the modulation feeder and the setting of the networks for equalization and for time equalization, and in addition also in the considerable saving of time in the adjustment itself. Furthermore, the self-adapting equalization does not permit use Fixed dial-up connections as modulation feeder. Furthermore, the accuracy and quality of the adaptive equalization structure is superior to manually set networks.

The invention will now be explained in more detail with reference to the figures. Show it:<ul id="ul0001" list-style="none"><li>FIG. 1: An example of a single-frequency radio network with three transmitting stations and conventional networks for equalization and for delay compensation.</li><li>FIG. 2: Basic block diagram of the adaptive equalizer network according to the invention with suitable means for controlling and storing the digital test signal on the receiver side of analog transmission links, e.g. B. in the modulation receiver of single-frequency transmitters.</li><li>FIG. 3: Block diagram for the control and storage of the digital test signal on the transmitter side of an analog transmission path, e.g. B. in the central station of a single wave radio network.</li><li>FIG. 4: Execution of the equalizer for bandpass equalization according to the invention.</li><li>FIG. 5: Example of the formation of the power density spectrum of the test signal by means of partial response coding.</li><li>FIG. 6: Time diagrams to explain the simultaneous emission of single-frequency transmitters.</li></ul>

FIG. 1 shows an example of a radio network according to the prior art. A central station Z supplies three synchronous transmission stations S1, S2 and S3 via the modulation feeders MZB1, MZB2 and MZB3. The modulation feeders are dedicated dedicated lines or radio connections. In order to achieve undistorted and simultaneous radiation from the transmitting stations, the signal propagation times are matched by the networks LAG1, LAG2 and LAG3, and the signal distortions due to the modulation feeders are eliminated using the equalizer filters E1, E2 and E3. As explained, the time equalization and equalizer currently have to be set manually. According to the invention, it is now proposed to carry out time equalization and equalization, which can be combined to form a common network ENT, using a self-adapting filter.

FIG. 2 shows an embodiment of the invention for the receiving side. It essentially consists of an adaptive transversal filter TVF and additional modules SR, TEST, VER and STE, which are required to adapt the transversal filter. The transversal filter is constructed in a known manner and consists of m coefficients and (m-1) delay elements with the delay τ. The delay r is calculated directly from the bandwidth f to be equalized_B the transmission path. The following applies:<maths id="math0001" num=""><img file="EP0072479A2_D0001.tif" /></maths>or<maths id="math0002" num=""><img file="EP0072479A2_D0002.tif" /></maths>

The difference between the two versions of equalization is explained below. Baseband equalization is initially assumed. The required number m of filter coefficients depends on the distortion of the transmission link. For typical applications, such as wire or TF lines, the number should be between 10 and 20. The invention is based on the idea of ​​equalizing the transmission links, which are designed for analog signals, by means of transversal filters, which are adapted with digital signals. This provides the advantages of fast, accurate and comparatively easy to implement equalization, which has long been used in digital transmission technology, also for analog transmission methods.

The transversal filter coefficients are adapted with the aid of an initially transmitted digital test signal which is known to the receiver side in undistorted form and is stored there in a module TEST. Since the test signal must have the character of a pseudo-noise sequence, the memory TEST can e.g. B. simply realize as a feedback shift register.

The step duration of the digital test signal is exactly τ according to equation (1) or equation (2). The length of the test signal depends on the number m of coefficients and the chosen adaptation method. Typically, you should provide at least 10 - 30 times the coefficient number m as the step number of the test signal. If a time-variable transmission link is used, the test signal must be repeated periodically at a suitable time interval.

The start of the adaptation process is initiated by a control unit STE, which causes the control signal S1 to read out the test signal V from the memory TEST. Furthermore, the control unit STE switches a switch SR by means of a control pulse S2 in such a way that the received signal d reaches a summing point VER and an error signal E is obtained by subtracting the undistorted test signal W. With the aid of the error signal, the filter coefficients can be set according to a suitable known adaptation algorithm. An overview of common algorithms can be found e.g. B. in the publication "Advances in equalization for intersymbol interference" by JG Proakis from Advances in communication systems, Vol. 4, Academic Press, New York, 1975. With regard to the implementation effort, one should preferably choose the simple stochastic gradient algorithm, which is described in detail in the cited literature reference and therefore does not need to be explained here. After the digital test sequence has ended, the control unit STE ensures that the switch SR is reset to normal operation, the adaptively obtained setting of the filter coefficients is retained, and the memory TEST is switched off or returns to the initial state for a new adaptation.

It is clear that the time at which the adaptation begins determines the absolute transit time of the signal through the entire system comprising the transmission link and equalizer. When the invention is applied to the modulation feeders of a single-wave radio network, the timing of the control pulses at the start of the adaptation must therefore be selected such that a simultaneous radiation takes place from all single-wave transmission stations. With a known runtime of the modulation feeder, this is easy to achieve. If this requirement is not met, the invention can be extended to automatic absolute runtime compensation. This extension will be explained later.

FIG. 3 shows the means that the transmission of the digital test signal on the transmitter side requires. A control unit STS switches on the digital test signal, which can be called up from a memory TEST, in a manner similar to that on the receiver side by means of a switch SR. After completion of the test signal, the control STS ensures that the switch SR returns to the normal position and the test signal is switched off.

The previously assumed assumption that baseband equalization is carried out will not apply to many transmission links that are designed as bandpass systems. In this case, the equalizer must be implemented as a bandpass equalizer, as shown in principle in FIG. 4 is shown. Real and imaginary parts of the bandpass impulse response must be separated by transversal filters TVF and TVF_i be equalized, whereby for the imaginary part equalizer TVF_i a 90 ° phase shifter must be used. Furthermore, the comparison signal must be available as a bandpass signal, for which a modulator MOD is required. Finally, the same modulator must also be used in the control section on the transmitter side. Preferably one will choose a simple modulation method, e.g. B. binary phase keying. Analogously, the in FIG. 4th Design the bandpass equalization shown after prior downmixing as a baseband equalizer for the real and imaginary parts. Since this is the state of the art, this will not be discussed further.

With the arrangements described so far, it can happen with some transmission links that strong attenuation drops within the transmission bandwidth are not adequately equalized and consequently excessively high residual distortions remain. An example is shown in FIG. 5, in which the attenuation rises sharply towards the edges of the transmission range. The power spectrum of the error signal ε contains comparatively little information about these areas because of the damping leg breaks, so that the equalizer works imprecisely and incorrectly, especially in the frequency ranges in which it would have to be adapted particularly precisely. This undesirable effect can be counteracted by suitably shaping the power spectrum of the digital test signal, since one usually has rough information about the position and size of the attenuation drops in the transmission link. The partial response codes are particularly suitable for shaping the range of services. B. in the publication "Generalization of a Technique for Binary Data Communication" by ER Kretzmer, IEEE Trans. Commun. COM-14 (1966), pages 67-68. You have to select the code so that the power spectrum of the coded test signal has maxima precisely at the points where strong attenuation drops in the transmission link are to be expected. For the example from FIG. 4 With attenuation dips at the edges of the transmission area, the first order bipolar code (also known as the AMI code) is particularly suitable.

The partial response coding also offers the advantage of being able to use equalization in baseband technology even when the transmission link cannot transmit a constant component. You then have to select a code for the test signal that generates a signal power spectrum in the baseband without a DC component. In addition to the first-order bipolar code already mentioned, the second-order bipolar code is also particularly suitable here.

An alternative embodiment of the invention consists in using a recursive filter with transit times according to equation (1) or (2) instead of the transversal filter equalizer. The coefficients of the recursive filter can be set according to the same adaptation algorithms that apply to the transversal filter, preferably again using the stochastic gradient algorithm.

An advantageous extension of the invention in the application to single-wave radio networks is to ensure an automatic absolute runtime compensation of the modulation feeders. As already mentioned, the runtime over the entire system of modulation feeder and equalizer is determined by the time at which the adaptation process begins, and thus by the timing of the control signal. Within certain limits, which are roughly defined by the temporally effective equalizer length (m-1) · τ, the signal delay can therefore be varied by changing the timing of the control signal. In particular, the required time-synchronous radiation from the single-wave transmission stations is achieved by the time-synchronous start of the adaptation in all transmission stations. Of course, you have to extend all runtimes to the longest permissible runtime in the network.

As a particularly simple implementation of the time-synchronous adaptation, the reference to an external and very precise time signal is proposed, such as is broadcast in the Federal Republic of Germany by the time signal transmitter DCF 77. In FIG. 6, the principle is explained on the basis of a time diagram. First you need in the control parts STS and STE of FIG. 2 and FIG. 3 additional devices for receiving the time signal and for forming synchronous pulse trains for time marking. These additional devices generate time-synchronous pulse sequences in the control center (FIG. 6A) and in the transmitting stations (FIG. 60) from the time signal. The time interval of the pulse sequence corresponds exactly to the signal duration τ to be specified_sig. This should be chosen slightly larger than the maximum runtime of the modulation feeder, in order to allow possible statistical fluctuations and measurement inaccuracies even at the maximum runtime, and on the other hand to take into account delays due to the execution of control operations. In addition, points of view for conveniently deriving the pulse sequence from the time signal can also be incorporated here.

It is now agreed that the beginning of the digital test signal in the control center always coincides with the occurrence of the short synchronizing pulses (or the positive or negative edges of the pulses etc.) (FIG. 6B).

The start of the adaptation at the single-wave transmission stations is also linked to the occurrence of the synchronizing pulses. Regardless of the time delay τ_MZB&lt;τ_sig the ^vripped digital test signal _d arrives at the comparison point VER (FIG. 6D), the control STE ensures the correct start of the adaptation by switching the undistorted comparison signal W to the comparison point only when the next time marker occurs (FIG. 6E). It is clear that in this way the same signal delay τ_sig is set for all broadcasting stations.

Depending on the transmission distortions within the range of time compensation that is possible per se, runtimes may exist for which the achievable equalization quality is poor. This effect can be avoided or at least significantly suppressed if the delay times in the equalizer filter are set to r '<maths id="math0003" num=""><img file="EP0072479A2_D0003.tif" /></maths>with t according to Eq. (1) or Eq. (2) shortened. In order not to reduce the effective equalizer length too much at the same time, the choice n-2 or n-3 is appropriate.

Finally, the case may arise that the required transit time equalization reaches or even exceeds the effective temporal equalizer length. In this case, the equalizer can only maintain the required signal delay due to a large loss in equalization quality or no longer at all. An additional runtime unit is then connected in a conventional manner upstream of the equalizer network. However, the runtime adjustment only needs to be carried out roughly, because the exact setting is made adaptively by the equalizer.

The rough runtime compensation can also be automated easily. As delay elements in the modulation receivers z. B. bucket chain circuits installed, the clock is adjustable. The higher the clock frequency, the faster an analog signal travels through the bucket chains, the less the delay. In addition to the receiver for the transmitter DCF 77 (or similar) and the device for generating the clock pulse sequences - cf. above - the modulation receivers still need a counter. This is restarted with every clock pulse. The digital test signal is sent out by the control center with a clock pulse and stops the counter as soon as it has reached the modulation receiver. Proportional to the meter reading reached, the cycle of the bucket chain circuits can now be set automatically. The digital test signal can then be used for adaptive equalization.

It is disadvantageous if the counter runs continuously; however, it must be in operation at the latest with the clock pulse at which the control center sends out the digital test signal. With selectable modulation feeders, this problem can be solved by only starting the counter when a connection has been switched through. After setting the clock of the bucket chain circuits, the counter can be stopped again.

As an alternative, you can only start the counter with the incoming digital test signal and stop it with the next clock pulse. The clock of the bucket chain circuits is then set inversely proportional to the counter reading. The counter only runs for a short time.

Instead of the bucket chains, passive runtime networks with switchable taps can also be used.

Another option for automating the rough runtime pre-adjustment is to send a time stamp from the control center via the modulation feeder before sending out the digital test signal. To do this, e.g. B. the center and the single-wave transmitter digital clocks that are synchronized by the transmitter DCF 77 (or a similar device). The clock gives the current time in binary form. In the control center, the current time is presented in a logic circuit for the longest possible runtime of the modulation feeders and then sent to the single-frequency transmitters. These have a time comparator in which the incoming time presented is compared with the current time. The determined time difference indicates by how much the modulation signal still has to be delayed so that it is emitted synchronously by all single-wave transmitters. The rough runtime pre-adjustment is set accordingly.