CN1281602A - Method and receiving device for estimating channels in communications systems - Google Patents

Abstract

According to the invention, a received signal composed of data symbols is received by a receiving station. The received signal is decomposed into individual scan values at the receiver, and these scan values are compared with known data symbols to determine channel coefficients, wherein the individual known data symbols of the received signal are stored at the receiving station. Thus, for fast-moving mobile stations, a reduced number of channel coefficients are determined, thereby improving the evaluation accuracy.

Description

Method and receiving device for channel estimation in communication system

The invention relates to a method and a receiving device for channel evaluation in a communication system with transmission channels between radio stations, at least one of which is mobile.

In radio communication systems messages (for example speech, image information or other data) are transmitted via transmission channels, in radio communication systems this is done by means of electromagnetic waves via a radio interface. At this time, the emission of electromagnetic waves is carried out with a carrier frequency, which is in a frequency band arranged for each system. The frequency domain of the carrier frequency in GSM (Global System for Mobile Communications) is 900 MHz. For future radio communication systems, for example UMTS (Universal Mobile Telecommunications System) or other 3rd generation systems, frequencies are allocated in a frequency band of approximately 2000 MHz.

The emitted electromagnetic waves are damped due to reflection, diffusion, and loss of emission from the curvature of the Earth and similar causes. This reduces the receive power which is available at the receiving radio station. This damping is position-dependent and, in mobile radio stations, time-dependent. The faster the mobile station moves, the channel conditions of the transmission channel also vary greatly in short time intervals. In multipath propagation, various signal components experience different delays at the receiving radio station. The stated effects describe a single transmission channel for a communication connection.

From DE 195 49 148 known a kind of radio-communication system, this radio-communication system has utilized CDMA-subscriber classification method (CDMA code division multiple access), wherein radio interface is additional time-division multiple access-subscriber classification method (TDMA- time division multiple access). A JD detection method (joint detection) is used on the receiver side in order to improve the detection of the transmitted data when the CDMA codes of several users are known. It is known therein that at least two data channels can be assigned to a communication connection via a radio interface, each data channel being distinguishable by means of an individual spreading code.

It is known from GSM mobile radio communication networks that the transmitted data are transmitted as radio blocks (bursts) in time slots, the middle part of the radio block being transmitted with symbols known to the receiving radio station. This intermediate part can be used as a training sequence to coordinate the radio stations on the receiver side. With the aid of the intermediate part, the receiving radio station performs a channel evaluation, that is to say an evaluation of the channel impulse response for the different transmission channels. Optionally, the training sequences and useful data can be organized in parallel via channels named after different CDMA codes.

When the channel conditions change rapidly, the channel impulse response of the transmission channel also changes in the transmitted radio blocks, so that only a part of the radio blocks is used for channel estimation. A fixed number of equidistant channel coefficients is determined independently of the channel conditions. However, this channel estimation is inaccurate.

The object of the present invention is to specify a method for channel estimation and a receiving device which can reliably determine the channel coefficients despite the rapid movement of the mobile station. This object is achieved by a method with the features of claim 1 and a receiving device with the features of claim 11 . Advantageous refinements of the invention can be taken from the subclaims.

A method according to the invention for channel estimation between radio stations of a communication system having a transmission channel, wherein at least one radio station is mobile, receives a received signal consisting of data symbols from a receiving station. where the data symbols are altered by a transmission method such as spreading, modulation and channel distortion.

The received signal is decomposed into individual scan values at the receiving station and compared with known data symbols for determining the channel coefficients, the individual known data symbols of the received signal being stored at the receiving station. Thus, only a reduced number of channel coefficients is determined for mobile stations moving very quickly, for which the accuracy of the evaluation is improved. The erroneous ratio of short sub-blocks of the received signal to a long expected channel impulse response is eliminated and the accuracy of the estimated channel impulse response is improved. A larger number of channel coefficients is determined when the mobile station moves slower.

This type of channel evaluation also eliminates restrictions on the speed of the radio station and on the permissible channel conditions without simultaneously extending the training sequence or reducing the data rate. The present invention can be used not only for transmission between a base station and a mobile station, but also for transmission between mobile stations.

According to an advantageous configuration of the invention, the known data symbols with scan values are compared with the data symbols in the transmitted training sequence, or the known data symbols are stored after data detection and subsequently compared with the original received The scan values are compared. It is therefore alternatively possible to return to known data symbols without distortion which allow a more accurate channel estimation at the expense of little data rate. In the second variant, accurate data evaluation is a prerequisite for the continuous derivation of the channel coefficients. Channel estimation can thus be improved during processing a part of radio blocks that does not include a training sequence.

A further feature of the invention provides that, when determining the channel coefficients, a reduction in the number of values of the known data symbols, that is to say the training sequence or the evaluated data symbols, is taken into account. Accurate and realistic evaluations of the channel coefficients are thus also obtained within a shortened time interval for the scan values to be processed.

It is also advantageous if the channel coefficients to be determined are selected such that the power of these channel coefficients is stronger on average than values that are not selected. This ensures that, despite the reduced number of channel coefficients, the transmission channel can still be described with sufficient accuracy and, using the channel coefficients, forms a good basis for subsequent data evaluation. For example, the selected value can be taken from a previously determined number of comparatively large channel coefficients. According to the invention, therefore, only the significant channel coefficients are reprocessed from the previous channel evaluation. The selection of the channel coefficients to be determined is repeated at relatively large distances in order to take into account the development of the meaning of the individual channel coefficients which are not calculated consecutively.

According to an advantageous refinement of the invention, the values of the selected channel coefficients are not equidistant from each other. A single channel impulse response corresponding to a different transmission channel can therefore also represent a large time spread of the channel impulse response (determined by multipath propagation), if the channel coefficients at the beginning and end of the channel impulse response are selected as significant of. It is not necessary to consider equidistance at this time.

An overlocalized equation e=Gh+n is constructed by reducing the number of values of certain channel coefficients, where G is the block order matrix and n is the noise component. The accuracy of channel estimation is improved due to over-localization. The accuracy determined therein can be traded off against realistic ones. It is also advantageous that the number of scan values considered in the channel evaluation can be reduced depending on the channel conditions of the transmission channel. Overlocalization is thus reduced, however the channel estimation becomes more realistic.

It is also advantageous that the consideration of a relatively long signal delay time results in a longer distance between the channel coefficients based on the signal delay time, without the need to increase the number of channel coefficients on the same scale. The channel impulse response can therefore also take into account special channel conditions (for example in channels) and is sufficient for the channel evaluation even with few scan values.

Exemplary embodiments of the invention are described in detail with reference to the drawings.

Which means

The block diagram of accompanying drawing 1 mobile wireless communication network,

A schematic diagram of the radio interface frame structure of accompanying drawing 2,

Figure 3 is a simplified diagram of the radio block,

The block diagram of the radio station receiver of accompanying drawing 4,

The block diagram of accompanying drawing 5 digital signal processing devices,

Figure 6 shows a block diagram of the channel assessment problem, and

Figure 7 is a process diagram of channel assessment.

The radio-communication system represented on accompanying drawing 1 corresponds in its structure to the known GSM-mobile radio communication network, and this is made up of a lot of mobile switching centers MSC that form a network with each other, and set up the entrance with the fixed network PSTN . Furthermore, the mobile switching centers MSC are connected to at least one base station controller BSC in each case. Each base station controller BSC in turn has the possibility of establishing a communication link with at least one base station BS. Such a base station BS is a radio station, which can establish a radio connection with a mobile station MS via a radio interface.

Figure 1 illustrates three radio connections V1...V3 for the transmission of useful information ni and signaling information si between three mobile stations MS and a base station BS, in which case a mobile station MS is assigned two data The channels DK1 and DK2 and the other mobile stations are each assigned a data channel DK3 and DK4. The operation and maintenance center OMC performs monitoring and maintenance functions for the mobile radio communication network and its components. According to the invention, the functionality of this structure is utilized by the radio communication system; this functionality can also be transferred to other radio communication systems, on which the invention can also be used.

The base station BS is connected to an antenna arrangement consisting of three individual transmitters. Each individual transmitter transmits to an area of the radio cell served by the base station BS. Alternatively, however, a relatively large number of individual transmitters can be used (according to matched antennas), so that spatial user separation methods according to the SDMA method (Space Division Multiple Access) can also be used.

The communication link of useful information ni and signaling information si between base station BS and mobile station MS is subject to multipath propagation, which is caused, for example, by reflections from buildings superimposed on the direct transmission path. Signal components of the same subscriber signal belonging to different propagation paths thus meet at the receiving station at different points in time (delayed propagation).

Starting from the movement of the mobile station MS, multipath propagation together with other disturbances leads to a time-dependent overlapping of signal components of a subscriber signal on different propagation paths at the receiving mobile station MS. Furthermore, starting from here, the subscriber signals of different base stations BS are superimposed on a channel at the receiving location to form a received signal rx. The task of the receiving mobile station MS is to perform channel estimation on the transmission channel of the user signal. The channel estimation is also accurate and realistic enough when the mobile station MS moves quickly, and the useful information ni transmitted in the user signal. Let the data symbols of the information si and the data of the management information be detected.

The frame structure of the radio interface can be seen from Figure 2. According to TDMA, the composition is a distribution in the wideband frequency domain, for example a bandwidth of B=1.6 MHz is arranged in a plurality of time slots ts, for example 8 time slots ts1 to ts8. Each time slot ts in the frequency domain B constitutes a channel. In the channels arranged for the transmission of useful data, the information of a plurality of communication links is transmitted in radio blocks. A number of frequency domains B are assigned to the radio communication system according to the FDMA (Frequency Division Multiple Access) composition.

According to FIG. 3, the radio block for the transmission of useful data consists of a data part dt with a data symbol d, on which an intermediate part m known to the recipient is inserted. The data d are spread with a separate fine-grained structure of the communication link, a spreading code (CDMA coding), so that the recipient, for example K data channels DK1, DK2, DK3, . . . DKK, can be separated by this CDMA component. The sender assigns each symbol to each data channel DK1 , DK2 , DK3 , . . . DKK a certain energy E.

A single symbol of data d is spread with Q blocks, sub-segments Q of period Tc are transmitted within symbol period Ts. The Q blocks are formed here as individual CDMA codes. The middle part m with training sequence tseq is composed of L blocks with the same period Tc. In addition, a guard time guard with period Tg is arranged in the time slot ts to compensate for different signal runtimes of the communication connections of consecutive time slots ts.

In the wideband frequency domain B, consecutive time slots ts are grouped according to the frame structure. In this way, the eight time slots ts are combined into a frame, in which case certain time slots of the frame are formed as channels for the transmission of useful data and returned to be used again by a group of communication links. Other channels, such as frequency synchronization or time synchronization for the mobile station MS, are not included in each frame, but are inserted at predetermined points in time within a multiframe.

For example, the parameters of the radio interface are as follows:

Period of radio block 577μs

Number of blocks per middle part m 243

Protection time Tg 32μs

Data symbol 33 for each data part N

Symbol period Ts 6.46μs

Number of blocks per symbol Q 14

Block cycle Tc 6／13μs

The parameters can be adjusted differently in the uplink direction (MS->BS) and downlink direction (BS->MS).

The receiver according to FIG. 4 is a radio station, which can be both a base station BS and a mobile station MS. On the receiving side, the receiving device according to the invention is used for channel evaluation. A corresponding launcher that can be realized is known, for example, from German patent document DE 197 34 936.

The receiving path of the device is described in detail in FIG. 4 . In the submodule E1 the receive signal rx is converted from the transmit frequency band into the low-pass range and split into a real and an imaginary component. An analog low-pass filter is performed on the submodule E2, and the receive signal with 13/3 MHz and a word width of 12 bits is then overscanned by a factor of two at the submodule E3.

Digital filtering takes place at the submodule E4 with a channel-splitting filter with a bandwidth of 13/6 MHz and the highest possible contour slope. Finally, the twice overscanned signal is decelerated 2:1 on the submodule E4.

The received signal e thus obtained consists essentially of two parts, namely a part em for channel evaluation (training sequence tseq with known data symbols t) and parts e1 and e2 for data evaluation. In sub-module E5 , the channel coefficient h ^(k) of the channel impulse response is evaluated with the aid of the known intermediate basic codes m of all data channels transmitted in the time slot.

In submodule E6 the parameters b ^{(k) of} the adapted filter for each data channel are determined using the CDMA code c (k). Interference caused by the intermediate part m ^(k) is limited to a minimum in the received data block e1/2 used for data evaluation in the submodule E7. This is possible by knowing h ^(k) and m ^(k) .

In the submodules E8 to E12 the data symbols d are determined by means of the pseudo-transpose of the integrated channel matrix A. A further solution is to decompose the integrated channel matrix A into singular values (singular value decomposition). Other solutions involve additional optimal criteria, such as the minimum mean square error criterion (MMSE) instead of the zero force (ZF) criterion. In addition, anti-coupling structures are also possible. These solutions can also be combined with each other.

The calculation of the cross-correlation A ^*T A is performed in the sub-module E8. Because A ^{* T} A has a Toeplitz-structure, only a small part of the matrix needs to be operated here, and this can be extended to use the entire matrix. When the moving part is moving slowly, this matrix A ^*T A is large because one chooses large sub-blocks. In this matrix, only a small part of A ^{* T} A is calculated. When the matrix A ^*T A becomes smaller when moving faster, one sometimes chooses the entire operation in order to achieve a low-noise result. In the sub-module E9, A ^*T A is Cholesky-decomposed into H ^*T H, where H is an upper triangular matrix. Since A ^*T A is Toeplitz-structured, H is also approximately Toeplitz-structured and the entire operation is not required. The vector s represents the reciprocal of the H quadrature unit, which can be used advantageously in the formula calculator.

In the submodule E10, the received symbol sequence e1/2 and b ^(k) are matched filtered (matched filter). The sub-module 11 realizes the formula group resolver 1 of H ^*T *z1/2=e1/2. In the submodule E13 the evaluated data d1/2 are demodulated, raised to the third power and then convolutionally decoded with a Viterbi coder. The decoded data block e ^(k) E13 is selectively input to the first data sink D1 or via the source decoder E14 to the second data sink D2. In addition, the decoded data blocks e ^(k) E13 are fed back to the submodule E5 , which uses these data blocks to track the estimated channel coefficient h.

The receiver (see accompanying drawing 5) performs digital filtering on the received signal rx in a digital low-pass filter after analog processing, that is, amplification, filtering, and convergence to the baseband in the HF-part. A part of the digitized received signal e, which is represented by a vector em of length L=M*W and does not include the interference of the data part dt, is transmitted to digital processing means comprising a channel estimator KS. The data evaluation of all communication links is performed jointly in a joint detection data evaluation device, wherein a detailed description can be obtained from German patent document DE 197 34 936.

The channel evaluation in sub-module E5 is described in detail below. The submodule E5 includes a channel estimator KS, a storage device SP and a control device SE, between which information can be exchanged. The data symbols t of the training sequence tseq known from the intermediate part m and the already detected data symbols d of the trace evaluation are stored in the memory means SP. The control device SE can be switched individually by the user between various modes of channel evaluation, while the channel evaluation is carried out in the channel estimator KS. The channel coefficient h is determined from the scan values e1 . . . e16 of the received signal e by means of channel estimation.

In Fig. 6 the channel estimation problem is represented. The data symbols d to be transmitted are varied by the transmission method, are represented by the combined channel matrix A and appear as scanning values e at the receiving radio station. When the channel evaluation determines the evaluated combined channel matrix A, the receiving radio station attempts to determine the evaluation value d for the transmitted data symbol d. For example the optimization criterion is the minimum mean square deviation of the scan value e from the received signal e to be evaluated. d = t when channel estimation is performed from the training sequence tseq, and d has been evaluated for return coupling and d = d when tracking.

Channel estimation in mobile radio systems must take into account the movement of the mobile station MS. Movement causes a Doppler-shift, which can be described by the amplitude- and phase-variation of the individual channel parameters h (a1 complex value), see channel model in DE 196 35 271. If the same distance is used for many channel coefficients h, the value of a single channel coefficient h varies, whereas the profile of the channel impulse response remains constant. The channel evaluation can thus also be carried out continuously for relatively fast-moving mobile stations MS by processing the detected data symbols d during a radio block with the tracking channel coefficient h.

The value of the channel coefficient h when moving slowly is based on the average of the channel estimates over a number of radio blocks based on the middle part m. If the frequency is busy only radio blocks of the same carrier frequency are considered. At average speeds the middle part is compressed and at comparatively high speeds the tracking of the channel coefficients also takes place outside the middle part m formed by the training sequence tseq. The channel evaluation mode is switched on the receiving device via the control device SE in dependence on the evaluation accuracy, the rate of change of the measured channel impulse response or the rate of change of the reserve time (time advance) corresponding to the signal runtime.

For tracking purposes the data part of the radio block is divided into sub-blocks so that it can be assumed that the channel conditions do not change significantly during the sub-blocks. Data detection uses the previously determined channel coefficient h and returns the known data symbols d (e13 in Fig. 4) for which sub-blocks are detected for channel estimation. In order to minimize the delay of the channel evaluation, the sub-blocks can be overlapped so that, for example, a first sub-block contains data symbols 1 to 10, a second sub-block contains data symbols 2 to 11, etc.

The problem occurs in the short sub-blocks with known data symbols d, and the ratio of the much longer channel impulse response to the limited number of known data symbols t in the middle part m based on the training sequence tseq. In order to formulate the channel estimate with sufficient accuracy, it is sought to use a localized formulae.

Channel estimation (step 1 in Figure 7) deals with the time-dependent received signal during the training sequence tseq: $e\left(t\right)=\underset{k}{\mathrm{\Σ}}{c}_k\left(t\right)\×h_k(t,\mathrm{\τ})\×d_k\left(t\right)+\mathrm{no}\left(t\right),$

where n(t) is the noise part and x is the sign of the convolution operation. In matrix notation this formula can be rewritten as, e=Ad+n, where the combined channel matrix A reflects the influence of spreading by the CDMA-code c and the modulation by the transmission channel with the channel coefficient h to be evaluated.

This matrix formula is solved using the least squares solution (step 2 in Figure 7): $\stackrel{\⩓}{d}=A^{+}e={\left(A^{\′}A\right)}^{-1}{A}^{\′}e,$

Where A ⁺ is matrix A, the comprehensive transposition of A' is a closed matrix A, and d is a transformed vector of data symbol d (＾ means evaluation value).

After the data evaluation, if it is necessary to track the channel coefficient h when the mobile station MS is moving very quickly (step 3 in FIG. 7 ), it is possible to recalculate the sub-data blocks of the part containing useful data in the received signal e: $e\left(t\right)=\underset{k}{\mathrm{\Σ}}{c}_k\left(t\right)\×h_k(t,\mathrm{\τ})\×d_k\left(t\right)+\mathrm{no}\left(t\right),$

Among them, c and h are synthesized into a combined block order g and its matrix writing method is: e=Gh+n, then the channel coefficient h is based on $\stackrel{\⩓}{h}=G^{+}e={\left(G^{\′}G\right)}^{-1}{G}^{\′}e$ Evaluation is performed (step 6 of Figure 7).

Many transmission channels can be described with sufficient accuracy by relatively power-weak channel coefficients h, which are sometimes distributed over large temporal distances determined by signal delays.

The delay (position) of these significant channel coefficients h is determined in step 4 (see FIG. 7). Then only the number of channel coefficients h that have been reduced is determined during subsequent sub-block processing.

For example, the formula group of scan values e1...e16 with three user signals A, B, C and 16 received signals e to be solved has the following form:

如果将加上虚线的单元删除，则得出缩短的公式组：

这个公式组与减少信道系数h=hs的数量非常一致，则按照公式 $\stackrel{\⩓}{h_s}=G_S^{+}e$ 使评估的精度(附图7步骤5)得到改善。This set of formulas can easily be overlocalized and lead to noisy evaluation results. The block sequence matrix G and the vector of the channel coefficient h can be drawn with a dotted line, and only the significant channel coefficient h is reserved.

If the cells with dotted lines are removed, the shortened set of formulas is obtained:

This formula group is very consistent with reducing the number of channel coefficients h=hs, then according to the formula $\stackrel{\⩓}{h_{\mathrm{the\; s}}}=G_S^{+}e$ The accuracy of the assessment (step 5 of Figure 7) is improved.

This procedure can also be used on signal delays that can be represented by extending the channel impulse response. The number of channel coefficients h remains the same, however its distance is enlarged.

Claims (15)

1. the method for channel estimating in the communication system between base station (BS) and mobile radio station (MS) with transmission channel, wherein not only base station (BS) but also mobile radio station (MS) can send and receive,

Therein,

-receiving station (MS BS) go up to receive the received signal (e) that includes the data symbol that is changed by transmission method,

-the recipient received signal (e) is decomposed into single scan values (e1 ... e16),

-receiving station (MS, BS) go up storage received signal (e) single known data symbol (d, t),

-in order to determine that channel coefficients (h) is with scan values (e1 ... e16) and known data symbol (d t) compares,

-wherein determine to reduce the channel coefficients (h) of quantity for moving fast mobile radio station (MS).

2. according to the method for claim 1, therein, with the scan values (e1 of known data symbol (t) with data symbol ... e16) on the training sequence that is sent out (tseq), compare.

3. according to the method for claim 1, therein, (d) stores after Data Detection with known data symbol, and with the scan values (e1 of original reception ... e16) compare.

4. the method that one of requires according to aforesaid right therein, will be considered the quantity of minimizing of the numerical value of piece sequential matrix (Gs) when determining channel coefficients.

5. according to the method for claim 4, therein, the digital average that selected channel coefficient (h) is reduced quantity is better than on power does not have chosen numerical value.

6. according to the method for claim 5, therein, chosen numerical value is to take out from the bigger quantity of the channel coefficients of determining before this (h).

7. according to the method for one of aforesaid right requirement, therein, the distance of the chosen numerical value of the piece sequential matrix (Gs) that is reduced is not equidistant.

8. the method that one of requires according to aforesaid right, therein, the quantity of the channel coefficients (h) that is determined by minimizing has proposed the formula group e=Gh+n that very consistent preparation is found the solution, and wherein n is that noise section and G are the piece sequential matrix.

9. the method that one of requires according to aforesaid right, therein, the scan values (e1 that when channel estimating, considers ... e16) quantity depends on the channel condition of transmission channel and reduces.

10. the method that one of requires according to aforesaid right, therein, the distance between channel coefficients (h) on the signal delay time basis becomes big, does not need the quantity of channel coefficients (h) is amplified with same yardstick.

11. the receiving system of communication system,

Have a storage device (SP) and be used to store the scan values (e1 of received signal (e) ... e16), received signal comprises the data symbol that is changed by transmission method, and be used to store received signal (e) known data symbol (d, t) and

Have a channel estimator (KS) and be used for channel estimating transmission channel,

-wherein in order to determine that channel coefficients (h) is with scan values (e1 ... e16) with known data symbol (d, t) compare and

-wherein determine the minimizing quantity of channel coefficients (h) for moving fast mobile radio station (MS).

12. according to the receiving system of claim 11, therein, the chosen digital average of the minimizing quantity of channel coefficients (h) is better than on power does not have chosen numerical value.

13. according to the receiving system of claim 11 or 12, therein,

Scan values (the e1 that when channel estimating, considers ... e16) quantity depends on the channel condition of transmission channel and reduces.

14. according to the receiving system of one of claim 11 to 13, therein,

With a plurality of subscriber signals during through different transmission channel, it is a received signal (e) that these transmission channels are gone up overlapping at receiving system (EE) subscriber signal on a channel.

15. according to the receiving system of one of claim 11 to 14, therein,

The learning sequence (tseq) that will comprise given data symbol (t) transmits on identical channel with subscriber signal.

Images