DE19701011C1 - Channel estimating method for mobile communication channel - Google Patents

Abstract

The method involves using a transmitter and a receiver in which data symbols are detected, which are transferred through a time-variable, multi-path mobile communication channel. A linear channel estimate is performed in combination with a non-linear modulation method. The training signal of the linear channel estimate and the data signal are preferably transmitted in a time-multiplexed operation. The linear channel estimate is preferably performed through an optimised estimate method, whereby a training signal is multiplied in the transmitter with a frequency-dependent transfer function, and a channel pulse response is calculated from the resulting transfer function by a Fourier transformation.

Description

The invention relates to a method for channel estimation tion for the purpose of detection of data symbols by ei NEN time variants and multi-way transmission channel are transmitted with non-linear modulation methods, ge according to the preamble of claim 1. Furthermore relates the invention relates to an arrangement for performing the Procedure.

In radio transmission systems, this is generally within the transmission bandwidth frequency-selective transmission function or the equivalent low-pass impulse response of the over transmission channel time-variant due to multipath propagation, Shading effects and Doppler shifts due to moving Transmitter, receiver and spreader as well as changes in the atmosphere propagation conditions. This applies e.g. B. for dir radio, satellite radio and other radio services and especially for mobile communications.

The changes in the propagation conditions lead to ei radio channel with a large coherence bandwidth compared to Signal bandwidth of the transmission system to an almost free quench independent channel transmission function, its attenuation however is subject to considerable fluctuations. As a countermeasure to improve the reception conditions here z. B. An tennendiversity and / or frequency hopping are applied.

In contrast, a channel has a smaller coherence band wider than the signal bandwidth of the transmission system frequency selective transfer function. This frequency selec tive fading, due to the multipath valued differences in runtime of the individual paths  causes intersymbol interference in digital transmission limit of the symbols of the received signal. Especially in the mobile This frequency-selective transmission function is additional radio Lich time-varying, the rate of change from Doppler spectrum of the channel is dependent. The intersymbol-in Interference through multipath propagation can occur in the receiver Exploitation of this multipath diversity by means of suitable Ent strain measures can be processed in a useful way.

The reception signal is also used in cable transmission systems through the frequency-dependent transfer function of the channel linearly distorted, so that equalization measures are required are. In contrast to radio transmission systems, time is variance here however negligible.

When transmitting digital signals via time variant Channels, e.g. B. non-stationary radio channels occur through the multipath propagation, transit time differences of over the individual detours of the multipath channel received signal components on the signal detection of relatively ho bit rates (e.g. in the GSM system) to extremely strong and time-variant neighboring character disturbances of the transmitted symbols to lead. In doing so, the mutual temporal shift of signal components several characters or symbols of the useful infortnation. To the signal under such influences recognizing symbols are adaptive echo equalizers necessary, for example, the Viterbi algorithm for quickly find a sufficient equalization optimum apply.

In the recipient can look at the effort and the requested transmission quality aimed for different detection principles pien such as B. threshold detection, correlation receiver and maximum likelihood sequence estimation (MLSE) will. For all of these methods, however, the reception must signal to be adapted to the detection principle by a ge  opened eye diagram for single symbol decision or Threshold decision or to improve the effective Signal / noise ratio through approximately known par tial response signals for the decision of entire sequences.

Since the radio channel can be viewed as a linear system, the received signal is subject to linear distortion. To the Adaptation of the received signal to the required detection signal, an adaptive linear equalizer or nonlinear equalizer for channels with spectral zeros len are used, which are caused by the channel linear distortions as much as possible reduced or Optimization of the effective signal-to-noise ratio to the ge selected detection methods. The Teilsi Gnale the multi-way radio channel largely constructively overlaid device. In the MLSE method, the equalization is implicit in the Algorithm included.

Because of the rapid changes in the channel transmission radio tion, especially in mobile communications, the equalizer / detec Adapt the gate to the new channel sufficiently quickly without that iterative procedures are used. With such time-variant fading channels is therefore a quick adjustment equalizer, an adaptive channel estimator is required which approximates the effective channel impulse response telt. With this reference, the ones to be set directly Coefficients of the adaptive equalizer / detector determined.

According to the state of the art, the prerequisite for the Ent distortion of neighboring sign interference an estimate of the pulse answer or the complex filter coefficients of the Funkka nals, which takes place with the help of a channel estimator, which the Radio channel by evaluating one in each radio block (e.g. TDMA or FDMA block) included training sequence describes. These complex filter coefficients, who the obtained by correlation and evaluate the by the  Multipath propagation caused temporal distribution of the emp capture components, which are also used as the complex impulse response of the Radio channel is called, the distance of the sequence of the transmitted symbols. Each of these coefficients, called the tap coefficient the complex integral the level value of all caused by multipath propagation Signal components that are in the evaluation period of a Symbol duration fall, and are adjusted using standard adjustment Algorithms such as LMS (Least Mean Square) or RLS (Recursive Least Square).

The tap coefficients are fed to the signal detection and provide a cal for the duration of a radio block basis for a sufficiently reliable detection of the symbols transmitted in this block.

A receiving system for digital transmission with linear mo dulation systems has a transmitter and receiver egg equalizer / detector and a channel estimator.

The channel estimator generally uses the input signal and the output signal of the equalizer / detector to a Ka nal estimate for the time variant impulse response for each time to observe.

The output signal can only be used as the sole reference for the channel estimator can be used sensibly if it is pre has given statistical properties to the fault radio tion to be able to evaluate. Because of the feedback is done in this configuration the equalizer setting is iterative and therefore requires a longer adaptation time. Therefore divorce this procedure because of the rapid change in mobile radio ka for an equalization algorithm for mobile radio applications gen out.  

Procedures are therefore used for rapidly changing channels set, for which the channel estimator on the receiver side only uses the input signal and then the equalizer rer / detector acts. As a reference, only the Input signal used. So in the broadcast signal and there with a reference known at the receiving location also in the received signal is contained in the transmission signal a frame structure that consists of a specified sequence of Training sequences for the channel estimator and useful signals sequences for evaluation in the equalizer / detector.

Since these are usually non-stationary radio channels delt, the tap coefficients have the changes of the Radio channel can be adjusted per radio block. Wrong decisions the stability of the signal detection is endangered when adapting tion and ultimately lead to detection errors. For the change to be able to follow changes in the radio channel per radio block, be each radio block sits a training sequence.

The receiver is synchronized with the frame of the transmitter, to be able to evaluate the training sequences correctly. The Training sequence, which ver in the receiver as a priori knowledge is available, is evaluated in the channel estimator. In known Systems (e.g. GSM) the channel estimate is used to determine an approximation for the channel impulse response using correla tion carried out. This is the impulse response of the channel folded with the part of the receiver signal that contains the training sequence according to the framework. The channel estimator only provides a good approximation for that Channel impulse response when the autocorrelation function of the Transmitted signal is pulse-shaped. This procedure is therefore only usable for linear modulation methods.

The training sequence for channel estimation must be set up in this way be that their symbol sequence after RF modulation requires the  properties of the AKF (auto correlation function) owns.

The echo equalizer according to the described prior art are only used for linear RF modulation types bar. The Gaussian Minimum Shift Keying (GMSK) modulation, how it is used for the GSM system can serve as a limit apply to linear modulation because it is only low Exhibits non-linearities and with the echo equalizers delivers sufficiently good values in the state of the art.

In the result of the channel estimator, however, occur due to the limited bandwidth of the signal and the continuous time, possible delays due to reusable correlation benzipfel, which is the approximation of the channel impulse response fundamentally falsify. This will equalize / detec gate not optimally adjusted. This is especially true when Si low process gain signals are used, whose Peak / sidelobe ratio after the correlation is small.

However, the suitability of the conventional equalizer method works with increasing non-linearity of the modulation methods, such as For example, the 4-CPM (Continuous Phase Modulation) is lost ren, because with increasing non-linearity of the modulation ver driving the quasi-pulsed autocorrelation properties ten of the training sequence are lost after demodulation or depend on the phase shift of the modulation method.

Another type of channel estimation is the so-called optimal estimate tongue, in which the impulse response of the radio channel according to DE 42 33 222 C2 is determined and the no requirements on the AKF properties of the training sequence provides. The Training sequence selected so that its spectrum none Has zeros, an envelope that is as constant as possible and a cyclic phase curve for its inner part  sits, and that at the beginning and end of the training sequence the same phase condition exists.

DE 43 29 317 A1 describes a method for transmission of news where the coefficients of ge estimated channel impulse responses before data detection be subjected to non-linear estimation postprocessing.

The method of optimal estimation is described below with the aid of a transmitter / receiver arrangement which comprises a transmitter with an output signal S _S (t), a channel h (t) and a receiver with an input signal S _E (t), a channel estimator with a function c (t) and an output signal S _c (t). The following applies to the received signal S _E (t) at the input of the receiver:

S _E (t) = S _S (t) .h (t) (1)

The channel estimator output is:

S _C (t) = S _S (t) .h (t) .c (t) (2)

Ideally, the following should apply if the transmission channel is not to be estimated in a band-limited manner:

S _c (t) = h (t) (3)

From this follows for the transmission signal and the channel estimator impulse response:

S _S (t) .c (t) = δ (t) (4)

Dirac function.

This is only possible if both S _S (t) and c (t) are not band-limited. The following applies in the frequency domain:

S (jω) C (jω) = 1

For real transmission systems, the transmission channel must but only within the bandwidth of the transmission signal Carrier frequency to be known.  

The following transfer functions must be observed:

H (jω) = transfer function of the non-band-limited transmission channel, and
H _sch (jω) = evaluation function for band limitation of the transfer function to be estimated (= rect _{BHF / 2} (jω)).

The following calculation is made for the example of a right-angled evaluation function. This is for simplification only. The method can be applied to any band-limited evaluation functions.

H (jω) H _sch (jω) (6)

estimated transfer function

Fourier transform results in:

h (t) .h _sch (t) = h (t) .S _S (t) .c (t) (7)

Fourier transform results in:

H (jω) .S (jω) .C (jω) (8)

The following must apply to band-limited channel estimation:

h _sch (t) = (sin (π B _HF t)) / (π B _HF t) = S _S (t) .c (t) (9)

Fourier transform results in:

This gives the required channel estimator transfer function:

C (jω) = 1 / S (jω) rect _{BHF / 2} (jω) (11)

Fourier transform results in c (t).

C (jω) follows directly from the transmission spectrum. At C (jω) who who have no special demands - for example on the correction lation properties - provided. The only requirement is that S (jω) within the transmission band or at periodi shear transmission at the discrete frequencies to be considered zen has no zeros. Thus the broadcast spectrum be arbitrary.

The invention is in view of the problem described matik the task, a procedure and an arrangement indicate that a non-linear modulation method reliable channel estimation and determination of impulse response the propagation distance or the tap coefficients allowed, in order to achieve a sufficiently secure signal detection.

According to the invention, this object is achieved by means of of claim 1 and 5 specified features.

In an advantageous embodiment of the invention The training signal and the data signal in the Time division multiplex transmission. In this embodiment of the method according to the invention, the channel estimation can be un certain conditions both by correlation estimation tion as well as by the optimal estimate.

According to a further advantageous embodiment of the inventions The method according to the invention is based on the linear channel estimation the method of optimal estimation. This ent there are some restrictions on the correlation egg properties of the training sequence that are related to the channel estimation are connected by the correlation, and the optimal  can be done with training signals that only are subject to the conditions of claim 5.

The arrangement according to claim 8 enables in advantageously a linear channel estimation in Kombina tion with a non-linear modulator.

The arrangements according to claims 9 or 10 are two variations anten for the implementation of the method according to claim 2, wherein the embodiment according to claim 10 still has the advantage that with an I / Q modulator Darge each modulation can be set so that for the modulation and demodula tion of the useful signals and the training sequences the same com components can be used.

Further advantageous embodiments of the invention result from the remaining subclaims.

Embodiments of the invention are based on the enclosed described drawings. Show it:

Figure 1 shows a first embodiment of the arrangement according to the invention with a linear channel estimator and a non-linear modulator.

FIG. 2 shows a discrete-time and discrete-time training sequence as used in the arrangement of FIG. 1;

Fig. 3 shows another embodiment, carried out at the time multiplexed operation of an arrangement according to the invention;

Fig. 4 shows a training and data sequence as it is processed in the order of Fig. 3;

Fig. 5 shows an alternative embodiment of an arrangement accelerator as the invention for time division multiplex operation.

In Fig. 1 it is assumed that the equivalent base band is considered, in which case mixers, which would otherwise be Darge, are omitted and the signals are presented complex. It is therefore a representation of the transmission link in the equivalent baseband.

The arrangement according to FIG. 1 has a transmitter which contains a non-linear modulator 2 , which is fed with a base band signal S _BB (t), as shown in FIG. 2. The baseband signal S _BB (t) comprises data signals D and a training signal sequence T. It is assumed for optimal estimation and correlation that the training signal spectrum is a complex test signal with N spectral lines, which is periodic, has an envelope that is as constant as possible and has no zeros .

The transmitter emits a transmission signal S _S (t), which is passed on via the transmission channel h (t) and received by the receiver as an input signal S _E (t). The input signal S _E (t) is output at a demodulator 6 and a linear channel estimator 8 . The output signal of the demodulator is detected and equalized in an equalizer / detector, which also receives the output signal of the linear channel estimator 8 and is thus controlled.

The following signals must therefore be considered in order to assess the advantages of this arrangement:

S _BB (t) = baseband signal containing a training signal.
S _S (t) = transmission signal
h (t) = impulse response of the transmission channel
S _E (t) = received signal

If S _BB (t) is a value-continuous signal, the most important properties of the transmission link are as follows. The training signal in the transmission signal S _S (t) has a pulse-shaped AKF. For this purpose, S _BB (t) must be selected accordingly before demodulation, which follows directly from S _S (t) and the demodulation method. The channel estimation with respect to the signal S _E (t) with the known training signal in S _S (t) is carried out by optimal estimation in parallel with the demodulation process. The tap coefficients and amplitudes of the channel, that is to say the function h (t), then result from the optimal estimate.

With a discrete-value transmission signal S _S (t), S _BB (t) for digital transmission at the sampling times is a discrete-value signal, as is S _S (t). The training signal in S _S (t) must be a periodically continued signal after the modulation, which follows, for example, from a binary data signal. The training sequence in S _BB (t) is also a periodically continued signal in this example a binary signal. The discrete spectrum of S _S (t), which corresponds to the training sequence, only has to satisfy the boundary condition that it must not have zeros according to equation 11.

The value and time discrete training sequence shown in FIG. 2 has a maximum multipath delay ΔTmax of 5 symbols. The forerunners and followers correspond to the maximum delay ΔTmax to be estimated. In this way, overlaps with the random data can be avoided in the estimation. In this baseband signal, S _S (t) is a periodically continued signal of the training signal with any spectrum that is periodically continued at the edges, but which is known in the receiver. The optimal estimate can be carried out in the channel estimator whenever there are no zeros in the spectrum. In this case, the received signal S _E (t) is sampled and the spectrum is calculated symbolically. According to equation 7, this spectrum is divided by z consecutive symbols by the spectrum of the training sequence. After an inverse Fourier transformation, an estimate for the impulse response follows.

The only boundary condition valid here is that the period of the training signal core az before the modulation in S _BB (t) must be the same as the period of the corresponding part in S _S (t). With angle modulation, the phase at the end must be the same as the phase at the beginning. For an effective channel estimation, the following must apply to the training signal:

S _S (t) = S _S (t) + z T _symbol )

Furthermore, the spectrum of the training signal in S _S (t) must not include any zeros

n × 1 / (z T _symbol )

have within the measuring range.

An arrangement with which the method according to the invention can be carried out in a very general manner, the restrictions relating to the period of the modulated signal, that this being the same as the period of the baseband signal, does not apply, is shown in FIG. 3. The periodic training signal is generated in a generator 12 , while the data signal S _BB (t) is fed in parallel to a non-linear modulator 14 . The output signals of the generator 12 and the modulator 14 are passed on via a transmission signal switch 16 and the transmission channel 18 to a reception signal switch 20 . In the receiver, the received signal S _E (t) reaches the channel estimator 22 and the non-linear demodulator 24 . The output signal of the demodulator passes to the equalizer / detector 26 , which receives the output signal of the channel estimator 22 as a control signal.

Fig. 4 shows the baseband signal and the switch positions of the changeover switch 16 , 20 , from which the time-division multiplex operation for the transmission of the training signal and the baseband signal can be seen. By the switch, which can also be replaced by a gating function, either the training signal or the data signal are emitted in synchronization, that is, the switch is synchronized with the receiving frame in the receiver. Because of the time-division multiplexing and the linear channel estimation parallel to the non-linear modulation method for the data signals, the training signal can be evaluated in this case either by correlation or by optimal estimation, since there are no restrictions with regard to the period of the modulated signal that these are equal to the period of the base band signal is in the choice of this signal as in the previous embodiments. The process that can be carried out with this arrangement is completely independent of the demodulation process.

Finally, FIG. 5 shows an arrangement with I / Q modulators for performing the method according to the invention with time division multiplexing.

The arrangement has a changeover switch 30 with four inputs and two outputs in the transmitter. The inputs of the switch 30 are in this case the I-channel signal of the training signal, the I-channel signal of the baseband signal, the Q-channel signal of the training signal and the Q-channel signal of the baseband signal. Depending on the position A or B of the switch, these signal components are selectively applied to the low-pass filter 32 or 34 and from there to the I-channel modulator component 36 or the Q-channel modulator component 38 . The two modulator components 36 , 38 are connected to a local oscillator 40 , a phase shifter 42 which shifts the phase by 90 ° lies between the local oscillator 40 and the Q-channel modulator component 38 . The output signals of the modulator components 36 , 38 are added in a mixer 44 and then sent. The receiver is essentially symmetrical to the transmitter, and the input signal is given to an I-channel demodulator component 46 and a Q-channel demodulator component 48, respectively, which are connected to a local oscillator 50 , between the local oscillator and the Q-channel demodulator part 48, a phase shifter 52 is provided, which effects a phase shift of 90 ° be. The output signals of the demodulator components 46 and 48 pass through low-pass filters 54 , 56 to a reception signal switch 58 and from there to the channel estimator or the equalizer / detector (not shown).

With this arrangement, too, the method is completely independent from the modulation process, and the training signal is linearly modulated in the I / Q modulator. Because every modulati can be represented with an I / Q modulator, can use the same components for modulation and demodulation are used, only the control of the components differently.

Claims (11)

1. A method for channel estimation of mobile radio channels using a transmitter and a receiver in which data symbols are detected which are transmitted by a time-variant and multipath-afflicted mobile radio channel, characterized in that a linear channel estimation is carried out in combination with a non-linear modulation method. 2. The method according to claim 1, characterized, that a linear channel estimation training signal and a Data signal are transmitted in time-division multiplex operation. 3. The method according to claim 1, characterized, that the linear channel estimates according to the method of Op is estimated. 4. The method according to claim 3, characterized, that a training signal in the optimal estimation method with a frequency-dependent transfer function in the transmitter is multiplied that the reception spectrum by a trai ningssignalspektrum is divided, and that from that transfer function obtained using a Fourier Transformation the channel impulse response is calculated. 5. The method according to claim 4, characterized, that the training signal spectrum with a complex test signal N is spectral lines, which is periodic, a con as possible constant envelope and has no zeros. 6. The method according to claim 4 or 5,  characterized, that a training sequence has a cyclical course in its inner part and the same phase state at the beginning and at End has. 7. The method according to claim 1, characterized, that a 4-CPM method was chosen as the modulation method becomes.   8. Arrangement for performing the method according to one of claims 3 to 7, characterized in that in the transmitter a non-linear modulator ( 2 ) and in the receiver ger a non-linear demodulator ( 6 ) a linear channel estimator ( 8 ) with the transmitted signal fed who, and a demodulator ( 6 ) and the channel estimator ( 8 ) downstream equalizer / detector ( 10 ) are provided, and that the channel estimator ( 8 ) is laid out for the optimal estimate. 9. Arrangement for performing the method according to claim 2, characterized in that in the transmitter a generator ( 12 ) for a linear training signal and a non-linear modulator ( 14 ) and a transmit signal switch ( 16 ) and in the receiver a receive signal switch ( 20 ) , A linear channel estimator ( 22 ), a non-linear demodulator ( 24 ) and an equalizer / detector ( 26 ) are seen before, and that the changeover switches ( 16 , 20 ) are switched synchronously, the data signal and the training signal being time-division-multiplexed be transmitted. 10. Arrangement for performing the method according to claim 2, characterized in that in the transmitter a data signal generator, a training signal generator, a transmission signal switch ( 30 ) and an I / Q modulator ( 36 , 38 ) and in the receiver an I / Q- Demodulator ( 46 , 48 ), a linear channel estimator, an equalizer / detector and a receive signal switch ( 58 ) are provided, and that the switch ( 30 , 58 ) are switched synchronously, the data signal and the training signal being transmitted in time-division multiplexing mode. 11. The arrangement according to claim 10, characterized in that the changeover switches ( 30 , 58 ) are designed so that they the one modulator component ( 36 or 46 ) of the I / Q modulator, the I-channel signal and the other modulator component ( 38 or 48 ) of the I / Q modulator supply or remove the Q-channel signal.