WO2002067524A2

WO2002067524A2 - Pre-distortion in multiple-carrier transmitters

Info

Publication number: WO2002067524A2
Application number: PCT/DE2002/000449
Authority: WO
Inventors: Marc Schrader; Nabil Hentati
Original assignee: Robert Bosch Gmbh
Priority date: 2001-02-20
Filing date: 2002-02-07
Publication date: 2002-08-29
Also published as: DE10108101A1; WO2002067524A3

Abstract

The invention relates to a transmitter for transmitting signals over radio channels using orthogonal frequency multiplexing (OFDM) and to a method for transmitting signals over radio channels using orthogonal frequency multiplexing (OFDM). The inventive transmitter and method serve to iteratively modify filter coefficients of a pre-distorter (4). The differences of a measured characteristic curve having a straight line with regard to amount and phase, the interpolation points at which the differences are calculated, the respective interation step, and the filter coefficients enter the iterative calculation in order to calculate the new filter coefficients of the pre-distorter (4). The differences with regard to amount and phase are compared with threshold values in order to determine whether another iteration is necessary. By duplicating an OFDM symbol, the amplifier (8) is operated under saturation in order to examine the characteristic curve up to the inflection point. In one embodiment, filter coefficients are optimized by means of an offline simulation.

Description

Transmitter for sending signals over radio channels in orthogonal frequency division multiplexing (OFDM) and method for sending signals over radio channels in orthogonal frequency division multiplexing (OFDM)

State of the art

The invention is based on a transmitter for sending signals via radio channels in orthogonal frequency division multiplexing (OFDM) or on a method for sending signals via radio channels in OFDM according to the genre of the independent patent claims.

It is already known from M. Schrader and N. Hentati "Reduction of out-of-band radiation from transmission stages in the DAB-COFDM system", OFDM technical discussions, September 1998, Braunschweig, printed in the conference volume, that OFDM (Orthogonal Frequency Division Multiplex, English Orthogonal Frequency Division Multiplex) signals are predistorted according to the transmission properties of the amplifier in the transmitter, which is necessary because the OFDM signals due to the large difference between the small and the large amplitudes that occur in the OFDM signals, ie the dynamics or Amplitude variance, place high demands on the linearity of the amplifier in the transmitter, because all amplitudes of the OFDM signal are to be amplified linearly. In the above-mentioned article, a feedback system for predistortion was proposed, in which a part of the amplified OFDM signal is fed back and compared with a buffered OFDM signal in order to determine the transmission properties of the amplifier in the transmitter. The buffered OFDM signal is the OFDM signal, which is then amplified and fed back. Because the properties of the OFDM signal are very similar to those of a noise signal, sophisticated synchronization is necessary for the buffered OFDM signal and the amplified OFDM signal.

Advantages of the invention

The transmitter according to the invention for sending signals over radio channels in the OFDM or the method for sending signals over radio channels in the OFDM with the features of the independent claims has the advantage that a slow change in the predistortion is made by the iterative change of the filter coefficients of the predistorter , whereby the amplifier remains in the working point. In addition, there is an adaptive reaction to changes in the amplifier characteristics.

The measures and further developments listed in the dependent claims enable advantageous improvements to the transmitter or method for sending signals via radio channels in the OFDM specified in the independent patent claims.

Through an offline simulation of the calculated filter coefficients with respect to the predistortion, an optimization of the filter coefficients is achieved without having to optimize the signal path via which the sending OFDM signals to intervene. In particular, errors in the filter coefficient calculation are thus eliminated before the filter coefficients are brought into the signal path. This increases the quality of the transmitted data.

It is particularly advantageous that the measuring module carries out the comparison at support points with a straight line, whereby on the one hand only a calculation at the support points is necessary and on the other hand the comparison with the straight line detects whether there is still a non-linear behavior in the predistorter-amplifier system. This procedure is simple and requires little effort. The evaluation with the iteration step weights errors that occur more frequently, in particular at the beginning of an iteration, so that the convergence of the method is retained.

In addition, it is advantageous that, at the beginning of transmitter operation, the predistorter is loaded with initialization values as filter coefficients that have a linear characteristic, so that the actual amplifier characteristic in a first iteration step is determined by the measuring module without distortion by the predistorter.

It is also advantageous that an OFDM symbol is increased in amplitude by a multiplier or that a measurement signal is keyed into the OFDM signal, so that the amplifier is driven into the saturation range in order to increase the amplifier characteristic up to the break point, i.e. where the linear and nonlinear range of the amplifier characteristics meet.

drawing Exemplary embodiments of the invention are shown in the drawing and are explained in more detail in the description below. 1 shows a block diagram of the transmitter according to the invention, FIG. 2 shows a structure of a DAB frame, FIG. 3 shows an idealized characteristic curve of an amplifier, FIG. 4 shows a first flow diagram of the method according to the invention, FIG. 5 shows a second flow diagram of the method according to the invention and FIG. 6 shows a third flow diagram of the inventive method.

description

Orthogonal frequency division multiplex (OFDM) is a known and successful method for mobile radio applications. In OFDM, the signals to be sent are distributed over many subcarriers, these subcarriers being at a certain frequency spacing from one another, so that the signals distributed over the subcarriers do not interfere with one another. This behavior is described as orthogonal.

OFDM is therefore used for digital radio transmission methods, in particular for mobile reception, for example using car radios. These include DAB (Digital Audio Broadcasting), DVB (Digital Video Broadcasting) and DRM (Digital Radio Mondial). These broadcast transmission methods benefit from the property of OFDM that when frequency selective attenuation occurs, only a small part of the broadcast signal transmitted is disturbed because the broadcast signal has been distributed over a plurality of frequencies and only a portion of the signal which is transmitted on one frequency is disturbed , where a strong damping occurs. The disturbed signal component is corrected by error-detecting and corrective measures. About these error-detecting and correcting measures include error-detecting and correcting codes such as block codes or convolutional codes.

In OFDM, after the distribution of the signals to be transmitted to the subcarriers, a summation occurs in the time domain of the distributed signals, the amplitudes being able to add up in such a way that the amplitudes of the superimposed signal take on a very large value at certain times and on the other hand a very large one assume small value. This depends on the phase relationship of the additive signal components, whether the signals add constructively or destructively. An amplifier in the transmitter has the task of amplifying all amplitudes equally, so that no non-linear distortions occur.

In order to take into account the transmission properties of the amplifier in the transmitter, predistortion is provided. To determine a transmission characteristic of the amplifier, a signal that has been amplified by the amplifier must be compared with the original signal. The OFDM signal is a challenge due to the uncorrelated sequence of amplitudes that occur due to the addition of the individual signal components, since it is difficult to synchronize the original OFDM signal with the amplified OFDM signal.

For the amplification of the OFDM signals, the amplifier should only be operated in the linear range. If a signal that is transmitted at a certain frequency is given a non-linear characteristic, for example that of the amplifier, frequency components arise at multiples of this certain frequency. If these multiples are outside the transmission frequency spectrum, one speaks of out-of-band radiation, since signal energy is then transmitted outside the available spectrum and is therefore lost for signal transmission because a receiver receives the signal Filtered out band radiation. In addition, the out-of-band radiation interferes with other transmission systems that are used at frequencies at which the out-of-band radiation occurs.

If new frequency components are available within the available transmission frequency spectrum, unwanted signal components are demodulated in the receiver. So there is a crosstalk. As a result, the signal quality and thus the bit error rate of the received signal is significantly deteriorated. The bit error rate indicates how many bits are incorrectly detected per received bit. The error-detecting codes are used to determine the bit error rate. After the distribution of the signals to be transmitted to the subcarriers, the OFDM signal is thus present as a noise signal, with individual amplitude peaks being able to drive the amplifier of the transmitter into the non-linear range. A predistortion of the OFDM signal is therefore necessary so that the characteristic of the amplifier has no influence on the spectrum of the OFDM signal.

FIG. 1 shows a block diagram of an OFDM transmitter according to the invention. A data source 1 is connected to a data input of a source code 2. The source coding 2 transmits its output signals to an OFDM modulator 3. The first and second data outputs of the OFDM modulator 3 lead to first and second data inputs of a multiplier 5. The first data outputs of the multiplier 5 lead to a first data input of a predistorter 4 and on the other hand to a first data input of a measuring module 12. The second data output of the multiplier 5 leads on the one hand to a second data input of the predistorter 4 and on the other hand to a second data input of the measuring module 12. The first Data output of the predistorter 4 leads to a first data input of a digital-to-analog converter 30, while the second data output of the predistorter 4 leads to the second data input of the digital-to-analog converter 30. Output signals from the first data output of a processor 13 lead to a third data input of the predistorter 4 and the output signals from the second data output of the processor 13 lead to a fourth data input of the predistorter 4. The first and second data outputs of the digital-to-analog converter 30 lead to the first and second data input of the quadrature modulator 6. The output signals of the quadrature modulator 6 lead to the data input of an up-mixer 7, which in turn transfers its output signals to an amplifier 8. A first part of the output signals of the amplifier 8 leads to an antenna 9, while a second part of the output signals of the amplifier 8 leads to a down mixer 10. The output signals of the down mixer 10 lead to a quadrature demodulator 11. The first and second data outputs of the quadrature demodulator 11 each lead to first and second data inputs of an analog-digital converter 31. The first and second data outputs of the analog-digital converter 31 lead to third and fourth data inputs of the measuring module 12. The first and second data outputs of the measuring module 12 each lead to the first and second data inputs of the processor 13.

The data source 1 is used here to generate the data. The data source 1 is here a microphone with connected electronics for amplifying and digitizing the speech signals converted by the microphone. The microphone 1 converts sound waves into analog electrical signals, which are amplified and digitized by the electronics connected to the microphone. The digital data stream resulting from these voice signals leads into the Source coding 2. This source coding 2 is carried out on a processor.

The source coding 2 reduces the number of bits that have arisen from the voice signals by the source coding 2 taking redundancy from the digital data stream. Using psychoacoustic models, data that are not necessary for the reproduction of the speech signals are eliminated from the speech signals. The data stream reduced by the source coding 2 is then fed to the OFDM modulator 3. In addition to voice signals, other data such as text, image and video data can also be transmitted. A source coding specific to the type of data is then carried out here.

The OFDM modulator 3 first carries out a differential phase modulation of the signals to be transmitted. Differential quadrature phase shift keying (DQPSK) is used for this. The DQPSK is a digital modulation in which the phase change of the signal is modulated. The phase change is used as a modulation signal at a certain time interval, i.e. per bit. A phase change of +/- 90 ° is used here. Differential modulation methods have the advantage that no absolute value has to be determined in the receiver in order to demodulate the signals, since the transmitted information is contained in the phase change of the transmitted signals. A bit sequence of 110 therefore leads to a phase change of + 90 ° for the two ones and -90 ° for the zero.

In addition to the DQPSK, other differential or non-differential phase modulation methods can also be used. However, it is also possible, too Use amplitude modulation methods or frequency modulation methods here.

The DQPSK is a complex modulation method, since the bits of the bit stream which are fed into the OFDM modulator 3 are mapped to phase changes. If a phase of a signal is changed, a complex level is used for the graphic representation of the signals as a pointer, with a real part on the abscissa and an imaginary part on the ordinate. A signal with a phase of> 0 is rotated around this phase in the complex plane counterclockwise from the abscissa. If you change the phase by 90 ° four times, you are back at the output signal. Four modulation states that can be distinguished from one another are therefore possible with DQPSK.

In addition to the differential QPSK, the OFDM modulator 3 distributes the signals to be demodulated to the subcarriers, so that an OFDM signal is produced. Since the DQPSK, which the OFDM modulator 3 performs, produces a complex signal, a first and a second data output from the OFDM modulator 3 are connected to a first and second data input of the multiplier 5 in order to generate two components of the signal, imaginary. and real part to be processed separately. The multiplier 5 multiplies a measurement signal by the OFDM signal. The multiplier 5 accordingly multiplies the measurement signal by the OFDM signal at certain times, so that the measurement signal changes the OFDM signal at these times. These times are predefined, for example every hour or once a day. Before the actual operation of the transmitter according to the invention, this measurement is carried out and then continued later at the specified times during the operation of the transmitter. The multiplier 5 has a memory which contains the measurement signal. The measurement signal is a factor here a two to double an OFDM symbol. It is thus advantageously achieved that the amplifier 8 is driven into saturation in order to determine the filter coefficients up to the break point of the amplifier characteristic curve. The amplifier characteristic curve relates the amplitude of the output signal and the input signal. The relationship between the input and output voltage is still linear down to the break point. With input voltages that are larger, it becomes non-linear. In addition to the measuring times at which the OFDM signal is doubled, the multiplier 5 multiplies the signal by one, so that the signal passes the multiplier unchanged. Alternatively, it is possible that the OFDM signal is evaluated with other factors as a measurement signal, as long as the amplifier 8 is driven into saturation.

With DAB, a zero symbol is provided for synchronization at the beginning of a DAB frame with which the DAB signals are transmitted. Figure 2 shows a DAB frame. A synchronization channel 40 at the beginning of the DAB frame has the zero symbol. In a so-called Fast Information Channel 41, information about the multiplex and other service information is transmitted. A so-called main service channel 42 has the data to be transmitted, such as audio programs and / or multimedia data.

The measurement signal is multiplied by the fast information channel 41, so that no other data which are transmitted in the DAB frame are overwritten. It is acceptable that the fast information channel 41 is changed with the measurement signal. Since the information contained in the Fast Information Channel is not absolutely necessary for the recipient and is often repeated anyway. The measurement signal only has to meet the requirement that it drives the OFDM signal into the saturation range of the amplifier. Furthermore, the measurement signal will have the length in time of at least one symbol or a multiple of a symbol. In addition, the measurement signal is synchronized with an OFDM symbol in the OFDM signals, so that two symbols are not impaired by the measurement signal. The measurement signal is therefore a DC voltage signal which acts as a constant factor by which the OFDM signal is multiplied.

The predistorted signals go to the predistorter 4 after the multiplier 5. The OFDM signals are predistorted using filter coefficients. When the transmitter is operated for the first time, initialization values are loaded into the predistorter 4, that is to say the filter coefficients are chosen to be linear, so that the amplifier characteristic can be measured in an unadulterated manner. The predistorter 4 does not pre-distort here.

The OFDM signal, which has the OFDM symbol by which the measurement signal has been multiplied, goes via the first and second data output as a complex signal from the predistorter 4 to a digital-to-analog converter 30, which converts the components of the complex signal into analog Converts signals, which then get into the quadrature modulator 6.

The complex OFDM signal with the OFDM symbol, which was multiplied by the measurement signal, is converted into a real signal with the quadrature modulator 6. The complex signal y (t), which is mathematically described as y (t) = a (t) + jb (t), is converted into a real signal x (t) by the following rule: (t) = a (t ) cos (ωt) - b (t) sin (ωt) Here ω is a frequency by which the OFDM signal is converted into an intermediate frequency by an upmix.

After the quadrature modulator 6, the up-mixing 7 follows, with the real OFDM signal now in the

Intermediate frequency range is implemented. The upmixer 7 therefore has an oscillator in order to generate the frequency by which the OFDM signal is to be shifted. Alternatively, direct mixing is also possible without the use of an intermediate frequency.

The OFDM signal converted into the intermediate frequency is fed into the amplifier 8 after the up-mixing 7 or is amplified in accordance with the transmission characteristic of the amplifier 8. After the amplifier 8, the OFDM signals arrive on the one hand at the antenna 9 in order to be sent therewith, and on the other hand to the downmixing 10, which reduces the amplified signal back to a baseband. This portion of the OFDM signal is therefore fed back. The portion is of course very small compared to the portion sent, for example less than one percent, since most of the signal energy is used to radiate the OFDM signals. The OFDM signal is decoupled using a directional coupler. The directional coupler has two lines which are placed in such a way that an electromagnetic decoupling of signal energy from one line to the other line is made possible.

The baseband is the frequency range in which the data was generated. After the downmixing 10, a complex signal is again generated from the real signal in a quadrature demodulator 11, so that the quadrature modulator 11 has two data outputs, to each of which a data input of the analog-digital converter 31 is connected which digitizes the components of the complex signal. The digitized signals then arrive in the measuring module 12.

The measuring module 12 receives the OFDM signal with the symbol, which has been multiplied by the measuring signals, via its first and second data input. This OFDM signal was also amplified by the amplifier 8. Via its third and fourth data input, the measuring module 12 receives the OFDM signal with the OFDM symbol from the first and second data output of the multiplier 5, which was multiplied by the measuring signal. This symbol was temporarily saved for this. The thus modified OFDM signal, which is led from the multiplier 5 to the measuring module 12, is temporarily stored in the measuring module 12 until the same OFDM signal with the symbol that has been multiplied by the measuring signal is sent from the quadrature demodulator 11 to the measuring module 12 , This enables a comparison of the OFDM symbol, which was multiplied by the measurement signal, before and after that of the predistorter 4 and the amplifier 8.

By comparing the amount and phase of the OFDM signals upstream of the predistorter-amplifier system and after the predistorter-amplifier system, the characteristic curve of this system is determined by the measuring module 12. By comparing this characteristic with a straight line, the deviation of the predistorter-amplifier system from the desired linear behavior is determined, namely by amount and phase. In a linear system, the phase change ΔΦ should be 0 here. The connection to the multiplier 5 synchronizes the measuring module 12 so that the measuring module 12 is informed of which signals are to be compared. The measuring module 12 transfers the difference values according to amount and phase to a processor 13 via its first and second data output. The processor 13 compares that Difference values by amount and phase with threshold values in order to determine whether the predistorter 4 with its filter coefficients predistorted the OFDM signals to the extent that the overall system predistorter-amplifier had a linear characteristic. If the difference values are above the threshold values, there are still nonlinear components in the transmission characteristic, and a further correction of the filter coefficients is necessary. The processor 13 then recalculates the filter coefficients F according to the following equations:

F = | E | • exp ^Jφ

ak

\ F _{l + l} \ = \ F \ + δ -

The first equation generally describes the filter coefficients as complex quantities, the second equation giving the calculation for the amount of the new filter coefficient, while the third equation describes the calculation of the phase of the new filter coefficient. F denotes the filter coefficient in iteration step i and F _{1 + 1} the filter coefficient in iteration step i + 1. δ is an evaluation factor for the iteration step, the evaluation factor here being a fixed value less than one. As a result, strong deviations, which may be attributable to measurement errors and which can occur in particular at the beginning of the iteration, are strongly dampened and are therefore not very strongly included in the iteration. This evaluation factor δ thus leads to better convergence of the iteration method. Δk denotes the deviation with respect to the amount of the characteristic curve determined at the support points x in comparison to the straight line which has an incline 1 here, x thus denotes the support point and the factor 2 comes from a derivative. Φ. ^ Denotes the phase of the filter coefficient at the iteration point i and Φi ₊₁ denotes the phase at the iteration step i + 1. ΔΦ is the deviation of the measured characteristic of the predistorter-amplifier system with respect to the zero phase. The goal of the iteration is that Δk and ΔΦ are less than a tolerance threshold (ideally zero), which means that the predistorter-amplifier system is linear. The new filter coefficients F _{i + 1} are then transmitted via the first and second data outputs of the processor 13 to a third and fourth data inputs of the predistorter 4, so that the following OFDM signals are predistorted with the changed filter coefficients.

An idealized characteristic curve of an amplifier is shown in FIG. The input voltages are plotted on the abscissa, while the output voltages of the amplifier are plotted on the ordinate. Between points 80 and 90, the transmission characteristic of the amplifier is linear. Points 80 and 90 mark the breakpoints where the transmission characteristics of the amplifier pass into the respective non-linear range. In order to control the characteristic curve of the amplifier up to the break points, it is necessary, for example, to double an OFDM symbol, which then drives the amplifier into saturation, that is to say into the non-linear range. This is necessary because OFDM signals have a large amplitude variance extremely rarely reach the non-linear range or in the vicinity of the break point. The amplitudes of an OFDM signal are Rayleigh-distributed, ie here in particular that there are only a few amplitudes for large amplitude values. In order to be able to use an OFDM signal to test the amplifier characteristic down to the breakpoint, an OFDM symbol is simply multiplied by a factor of 2. Alternatively, other signals can be keyed into the OFDM signals which drive the amplifier into saturation, e.g. B. a sine signal. This can also be done iteratively, ie the measurement signal is gradually brought to the saturation area.

FIG. 4 shows the method according to the invention for sending signals over radio channels in orthogonal frequency division multiplex as a flow chart.

The data is generated in method step 14. This is done by means of a microphone as described above. However, other data sources are also possible, for example a computer with a keyboard. In method step 15, source coding is carried out, redundancy being taken from the speech signals, which are not necessary for a reconstruction of the speech data in the receiver. In method step 16, a modulation of the data stream is carried out after the source coding 15, a differential phase modulation being carried out here, as described above.

In method step 17, the data stream is distributed to various subcarriers by means of OFDM modulation. In method step 81, an existing OFDM symbol is optionally doubled by means of the multiplier 5 in order to drive the amplifier up to the saturation range. As shown above, other methods can also be used here. In method step 18, predistortion is carried out in accordance with the existing filter coefficients. A digital-analog conversion of the OFDM signal is carried out in method step 43. In method step 21, quadrature modulation is carried out in order to produce a real signal from the complex OFDM signal. In process step 22, the real signal is converted into the intermediate frequency. In method step 23, the converted OFDM signal is amplified by means of the amplifier 8. The amplified signal is sent in method step 24, while a further part of the amplified signal is mixed downwards again in method step 25 and converted into a complex signal again using a quadrature modulator in method step 26. In method step 44, an analog-digital conversion of the complex signal is carried out in order to carry out a comparison of the OFDM signal before and after the predistorter amplifier system in method step 27 in order to carry out the

Determine the transmission characteristic of this system and compare it with a straight line. The difference thus created is compared in step 82 with threshold values relating to the amount and the phase. If the existing differences are below the threshold values, then the method is run through again in method step 14. If, however, at least one threshold value is still exceeded by the differences, new filter coefficients are iteratively calculated in method step 84 in accordance with the rule described above. The method is then run through again with these new filter coefficients.

5 shows a second flow chart of the method according to the invention. It is limited to the essential procedural steps. In

Method step 100, the characteristic of the predistorter 4 and the amplifier 8 is measured by the measuring module 12. The characteristic curve of the amplifier 6 is calculated therefrom in method step 101. In method step 102, deviations of the measured signal from the characteristic at the support points are calculated, ie Δk, that is to say the deviation of the predistorted from the ideally linearized output power. In step 103, a Calculation of the filter coefficients using the equations above. In method step 104, the filter coefficients are then loaded into the predistorter 4. This method is then carried out iteratively, so that an optimization is achieved.

A development of the method according to the invention is shown in a third flowchart in FIG. 6, in that the filter coefficients are optimized separately from the actual signal path. A simulation is used for this so that, for example, extrapolation errors are not brought into the signal path via which the signals to be sent are carried. It is therefore an offline simulation.

In method steps 105 and 106, the characteristic curve of the amplifier 8 and the predistorter 4 is determined by sampling the signals in the signal path and subsequent calculation. The method curve calculated in this way is stored in method step 107. In method step 108, the filter coefficients are calculated as shown above. The calculated characteristic curve is then loaded in method step 109 in order to simulate a predistortion in method step 110 using the calculated filter coefficients and the characteristic curve. The deviations are then calculated therefrom in method step 111, so that the deviation Δk is then compared in method step 112 with a predetermined maximum deviation Δk _max . If Δk is above Δk _max , the process jumps back to step 108 in order to further optimize the filter coefficients. If Δk is below Δk _raax , the filter coefficients are optimized. Now in method step 113 the optimized filter coefficients are loaded into the predistorter 4 in order to transmit the signals to be sent with these filter coefficients pre-distort. This process is also carried out iteratively on an ongoing basis.

Claims

Expectations

1. Transmitter for sending signals over radio channels in orthogonal frequency division multiplexing (OFDM), the transmitter having an OFDM modulator (3) that generates OFDM signals, a predistorter (4) that predistorts the OFDM signals using filter coefficients, one Up-mixer (7), which converts the pre-distorted OFDM signals into an intermediate frequency, an amplifier (8), which amplifies the converted OFDM signals, an antenna (9), which transmits a first part of the amplified OFDM signals via the radio channels, and a downward mixer (10) which mixes a second part of the amplified OFDM signals down from the intermediate frequency into the baseband, characterized in that the transmitter has a measuring module (12) which, in iteration steps, mixes down the OFDM signals with the OFDM signals of the OFDM modulator (3) are compared and a characteristic curve is determined therefrom, and that a processor (13) is provided which for each iteration step from the respective K ennlinie and the filter coefficients new filter coefficients for the predistorter (4) determined, the processor (13) aborts the iteration depending on the comparison between the downmixed OFDM signals and the OFDM signals of the OFDM modulator (3).

2. Transmitter according to claim 1, characterized in that the processor simulates the filter coefficients for an optimization.

3. Transmitter according to claim 1 or 2, characterized in that the measuring module (12) determines the characteristic at support points, that the processor (13) compares the characteristic curve with a straight line at the support points in order to determine a difference in amount and phase in each case that the processor (13) determines the new filter coefficients from the differences, the support points, the respective iteration step and the filter coefficients if the difference is in each case greater than a predetermined threshold value.

4. Transmitter according to claim 3, characterized in that the predistorter (4) has initialization values as the filter coefficients at the start of operation of the transmitter.

5. Transmitter according to claim 4, characterized in that the initialization values of the predistorter (4) form a straight line.

6. Transmitter according to one of the preceding claims, characterized in that the transmitter has a multiplier (5) which increases an existing OFDM signal or probes a measurement signal, whereby the amplifier (8) is driven into saturation.

7. Method for sending signals over radio channels in orthogonal frequency division multiplexing (OFDM), whereby OFDM signals are generated, the OFDM signals being predistorted using filter coefficients, the predistorted OFDM signals being converted into an intermediate frequency are implemented, the converted OFDM signals being amplified, a first part of the amplified OFDM signals being transmitted via radio channels, a second part of the amplified OFDM signals being downmixed into the baseband, characterized in that the downmixed OFDM signals are compared with the generated OFDM signals in iteration steps and thus a characteristic curve is determined that new filter coefficients are calculated for each iteration step from the respective characteristic curve and the filter coefficients and that the iteration is dependent on the comparison between the down-mixed OFDM signals and the generated ones OFDM signals is canceled.

8. The method according to claim 7, characterized in that the filter coefficients for the predistortion are simulated in order to achieve an optimization.

9. The method according to claim 6, characterized in that the characteristic curve at the support points is determined, that the characteristic curve at the support points is compared with a straight line in order to determine a difference in amount and phase in each case, that from the differences, the support points, the each iteration step and the filter coefficients, the new filter coefficients are determined if the difference is greater than a predetermined threshold value.

10. The method according to claim 7, characterized in that initialization values for the filter coefficients are used in the first iteration step.

11. The method according to claim 8, characterized in that the initialization values form a straight line.

2. The method according to claim 9, characterized in that a signal is generated which drives the amplifier (8) into saturation.