DE4329317A1 - Method and system for transmitting data - Google Patents

Abstract

To reduce the detrimental effect of imperfect channel estimation, the coefficients of the estimated channel pulse responses are subjected to non-linear estimated value post-processing prior to the data detection and modified channel pulse responses are obtained in this way. The non-linear post-processing is, for example, a threshold value test in which those estimated coefficients whose magnitude falls below a specific threshold are set to zero during the data detection. Through the use of modified channel pulse responses, the effect of the imperfect channel estimation, in particular in data transmission systems with code-division multiplexing, can be reduced depending on the input-side interference and the current channel pulse response without the need to modify the channel estimator itself. <IMAGE>

Description

The invention relates to a method for transmission of messages according to the preambles of the claims 1, 4 and 5. Furthermore, the invention relates to a System for the transmission of messages.

There are already digital messaging systems with Burst transmission and code division multiplex known. As an application example For example, digital mobile radio comes into consideration. A large number of mobile systems intervene in mobile radio systems Participants to the transmission medium radio channel. The resulting multiple access problem can with the elementary multiple access method frequency division multiplex (Frequency Division Multiple Access, FDMA), time division multiplex (Time Division Multiple Access, TDMA) or code division multiplex (Code Division Multiple Access, CDMA) or with combinations this method can be solved, as z. B. in Meinke, H. H .; Gundlach, F. W .; Lange, K .; Löcherer, K. H .: Taschenbuch der High Frequency Technology, Vol. 3, Springer Verlag, Berlin 1986 is. Several subscribers transmit in CDMA mobile radio systems simultaneously in the same frequency band. The entire received signal is therefore due to the contributions of several Participants determined. By applying optimal estimation algorithms can separate the different subscriber signals and those sent by the individual participants Data are determined. Such an estimation algorithm is for example in Klein, A. and Baier, P. W .: Simultaneous cancellation of cross interference and ISI in CDMA mobile radio communications. 3rd IEEE International Symposium on  Personal, Indoor and Mobile Radio Communications (PIMRC 92), Boston, pp. 118-122, 1992.

To be able to use the estimation algorithms mentioned, the channel impulse responses of the radio channels of the individual participants be known. Obtaining this information will usually enables the recipient in that in the Transmit signals signal sections are faded in with the recipient has been agreed beforehand and therefore in the recipient are known and that the receiver from the response of the channel the channel impulse responses to these known signal sections determined. This is for example from the publications Steiner, B. and Baier, P. W .: Channel estimation in CDMA mobile radio systems using periodic test signals. COST 231 TD (92) 104, Helsinki, 1992; Crozier, S. N .: Short block data detection techniques employing channel estimation for fading time dispersive channels, dissertation Carleton University, Ottawa, 1990, and Baier, A .; Heinrich, G .; Wellens, U .: Bit synchronization and timing sensitivity in adaptive Viterbi equalizers for narrowband-TDMA digital mobile radio systems. Proc. IEEE Vehicular Technology Conference, Philadelphia, pp. 377-384, 1988. The Superimposed signal sections are called training signals or middle tambourine. The following is the designation Middle furniture used. It is assumed that the transfer done in blocks. A sent block is called Burst denotes and consists of a midamble for channel estimation and two data blocks.

Due to the unavoidable interference in radio transmission the exact cannot occur in the receiver Channel impulse responses, but only more or less exact Estimates of the channel impulse responses are obtained. These  imperfect channel estimation causes deterioration the transmission quality compared to the case perfectly known Channel impulse responses. In the upward stretch from CDMA mobile radio systems - that is the radio link over which several participants via different radio channels a base station address - affects the imperfect channel estimation more serious than in the downlink.

The harmful impact of an imperfect channel estimate can be reduced by reducing the transmission energy of the midamble increased by changing the amplitude and / or the The length of the middle tambour increased. This approach throws up the following new problems:

• - Increasing the midamble amplitude leads to transmission signals with a non-constant envelope. A constant envelope is, however, with a view to optimal design Transmitter amplifier desirable.
• - An extension of the midamble shortens that for data transmission available periods.
• - Increasing the midamble energy increases in undesirable Way the interference power at the receiver input of other participants of the system.

The invention has for its object the harmful effect reduce the imperfect channel estimate without to increase the transmission energy of the Middle Ambe.

According to the invention the task in the method of the beginning mentioned type by the in the characterizing part of the claims 1, 4 and 5 specified features solved. A  advantageous system for message transmission is in the claim 19 specified.

The essence of the invention is particularly in the non-linear Postprocess an estimated channel impulse response seen.

Particularly advantageous options for choosing the weight functions g (x) are given in claims 5 to 7.

A special choice of the weight function g (x) is in claim 8 specified.

The definition of the deviation of the received signal from its Replication calculated in the receiver is in claims 10 and 11 indicated.

The use of a linear estimation algorithm used for data estimation is used is in claims 12 to 14 specified.

The realization of the data estimator as an optimal consequence estimator (MLSE) as described in Proakis, J.G .: Digital communications, McGraw-Hill, New York, 1989, is described in Claim 15 specified.

Finally, claims 16 to 18 relate to implementation the nonlinear estimate postprocessing a microprocessor or a digital signal processor.

The method and the function of a device according to the In the following, the invention will be explained in more detail with reference to drawings explained.  

Show it:

Fig. 1 a valve provided for the inventive process signal format,

Fig. 2 is a system model of the transmission path,

Fig. 3 shows the structure of a receiver and

Fig. 4 shows the structure of a receiver with non-linear post-processing.

In the signal format shown in Fig. 1 for the method for message transmission according to the invention it is assumed that the signals to be transmitted from a transmitter to a receiver are transmitted in blocks and that the transmission system is a mobile radio system, for. B. is a CDMA mobile radio system. Training signals M, which are also referred to as midambles, are shown in the data blocks to be transmitted. Each data block M divides a data block into two sub-blocks T1 and T2.

In the method according to the invention one of a time-discrete System model assumed. All calculations are done in equivalent low-pass range performed. Complex sizes are underlined. Vectors are represented by bold lowercase letters and matrices identified by bold capital letters.

In the system model of the transmission link shown in FIG. 2, K subscribers transmit, via K, generally different radio channels F with the channel impulse responses h ^(k) = ( h ₁ ^(k) , h ₂ ^(k) .. H _W ^(k) ), k = 1. . .K, data blocks d ^(k) = ( d ₁ ^(k) , d ₂ ^(k) .. D _N ^(k) ), k = 1. . .K. Before transmission, each data symbol with a modulator MO is a code c ^(k) ≡ ( c ₁ ^(k) , c ₂ ^(k) ... C _Q ^(k) ), k = 1. . .K, modulated linearly. With the additive disturbance n = ( n ₁, n ₂ _... N _{QN + W-1} ) and the NQ × N matrices

which indicate the linear relationship between the data symbols of a data block of a subscriber and the time-discrete transmission signal s _i ^{(K) of} this subscriber, and the (NQ + W-1) × NQ matrices

which describe the linear relationship between the time-discrete transmit signal s _i ^{(K) of} a subscriber and the e _i ^{(K) of} the receive signal caused only by this subscriber is the entire receive signal

The received signal arrives at the input of the receiver.  

The receiver shown in FIG. 3 contains a channel estimator KS, for example the one in Steiner, B. and Baier, PW: Channel estimation in CAMA mobile radio systems using periodic test signals. COST 231 TD (92) 104, Helsinki, 1992 corresponds to a linear data estimator DS and K quantizer Q. With the matrix

A = ( H ⁽¹⁾ C ⁽¹⁾ , H ⁽²⁾ C ⁽²⁾ ... H ^(K) C ^(K) ), (4)

into which the channel impulse responses h ^(k) and the codes c ^{(k) of} all K participants are included, and the vector

d = ( d ^{(1) T} , d ^{(2) T...} d ^{(K) T} ) ^T. (5)

that by lining up the elements of the vectors d ^(k) , k = 1. . . K, arises, is the received signal

e = Ad + n . (6)

For the purpose of data estimation, the received signal should be usede with that Recipient known codesc ^(k) and the estimated channel impulse responses ^(k) an estimate of the vectord determined become. With the matrices, ^(k), k = 1. . . K, which from the estimated channel impulse responses^(k) analogous to Eq. (2a) and Eq. (2b) can be formed, the matrix

 = ( ⁽¹⁾ C. ⁽¹⁾, ⁽²⁾ C. ⁽²⁾ . . . ^(K) C. ^(K)), (7)

be formed. An estimation matrix should be made from Â

M _d = M _d ( Â ) (8)

are calculated and an estimate with the estimation matrix M _d

 =M _d e      (9)

can be determined from d .

For example, an estimate of minimal variance can be used to estimate the data. Are the elements n _{i of} the disturbance n uncorrelated, that is, applies

this is the case with a perfect channel estimate

 = (Â ^{· T} Â)^-1 Â ^{· T} e      (11)

an accurate estimate of the minimal variance of the vector d. Is the variance σ² of the input disturbancen known, can be used to get an estimate  an optimal linear Estimators are used. With the assumption

E {| d _n ^(k) | ²} = 1, k = 1. . . K, n = 1. . . N (12)

is

 = (Â ^{· T} Â+ σ²I)^-1 Â ^{· T} e      (13)

in the case of a perfect channel estimate, an optimal linear one estimate  of the vectordlike in Whalen, A. D .: Detection of signals in noise; Academic Press, San Dike, 1971 is described.

Because of the imperfect channel estimation, as already mentioned above, only estimates

_w ^(k) =H _w ^(k)+n _hw ^(k), w = 1. . . W, k = 1. . . K (14)

the channel coefficient _w ^(k) With

E {| n _hw ^(k) | ²} = (σ _hw ^(k) ) ² (15)

known, as in Steiner, B. and Baier, PW: Channel estimation in CDMA mobile radio systems using periodic test signals. COST231 TD (92) 104, Helsinki, 1992, and Crozier, SN, Falconer, DD; Mahmound, SA: Least sum of squared errors (LSSE) channel estimation. IEE Proceedings, Pt. F, vol. 138, pp. 371-378, 1991, the variances (σ _hw ^(k) ) ² decrease with increasing energy of the transmitted midambles. Due to the imperfect channel estimation, an incorrect estimation matrix M _{d is} calculated. A faulty estimation matrix acts like an additional disturbance on the input side.

The harmful effects of an imperfect channel estimate can by increasing the mean square error of the received signal compared to the error in case perfect known channel impulse responses are described. The middle one quadratic error is a measure of the deviation of the Received signal from its replica calculated in the receiver. For this, a dissertation Crozier, S. N .: Short-block  data detection techniques employing channel estimation for fading time dispersive channels. Dissertation., Carleton University, Ottawa, 1990.

With the discrete-time transmission signal S _i ^(k) , k = 1. . . K, the k-th participant is the received signal

The discrete-time transmission signals s _i ^(k) are generally not known in the receiver. For analysis purposes, it is assumed that both the time-discrete transmit signals s _i ^(k) and the coefficients of the channel impulse responses h _w ^{(k) are} known in the receiver. With the transmission quality customary in mobile radio, however, most of the data symbols are correctly detected, so that the assumption that the signals s _i ^{(k) are known} in the receiver is approximately fulfilled. With the knowledge of s _i ^(k) and h _w ^(k) the reception signal becomes in the receiver

replicated. The expected value

E {| e _i - ê _i | ²} = E {| n _i | ²} = σ² (18)

is referred to as the mean square error of the received signal and essentially determines the reception quality. In the case of channel impulse responses h ^(k) which are not perfectly known, the simulation of the received signal becomes

The following is from

E {| s _i ^(k) | ²} = 1, (20a)

E {( s _i ^(k) ) ^· s _j ^(m) } = δ _ÿ δ _km (20b)

went out. Eq. (20a) means that the envelope of the time-discrete Transmission signal of each participant is constant. This applies when modulation types with constant Envelope can be used. Eq. (20b) means that the time-discrete transmission signals of the K participants among themselves are uncorrelated and the different elements of the time-discrete transmission signal of a subscriber uncorrelated are. The acceptance of uncorrelated transmission signals the participant is obvious. The assumption of uncorrelated different Elements of the discrete-time transmission signal Due to the code spread, the participant only approximates to. With the above assumptions

The relationship

specifies the factor by which the signal-to-interference ratio changes of the estimates _n ^(k) due to the effect of the imperfect Channel estimate reduced. In mobile radio systems with CDMA is to strive for a not too long duration of the participants' middle ambitions, thus the associated loss of data rate can be tolerated remains. Assuming that the duration of the Middle Amambes, those with the same amplitude as the data symbols sent just enough to all _w ^(k) appreciate, applies Crozier, S.N .: Short-block data detection techniques employing channel estimation for fading time dispersive channels. Dissertation., Carleton University, Ottawa, 1990 and Steiner, B .; Klein, A .: Channel and data estimation in synchronous CDMA mobile radio systems with interference elimination, Kleinheubacher reports, 1992,

so that v according to Eq. (22) to

v2 (24)

will, d. H. the SNR degradation based on the estimated data the imperfect channel estimate is greater or equal 3 dB. The equality in Eq. (24) occurs when the Codes of the midambles used for channel estimation ideal Have auto and cross correlation properties.

A not obvious way to reduce the harmful There is an effect of an imperfect channel estimate therein, the estimates determined by the channel estimator _w ^(k)  including a quality measure of a non-linear To undergo estimated postprocessing and the  resulting modified estimates _w ^(k) to determine the Estimation matrix M_d to use.

The resulting structure of the receiver shown in FIG. 4 comprises a linear channel estimator US, an arrangement for non-linear estimation postprocessing NN and a data estimator DS with K quantizers Q. Modifications are made with the quality measure z _w ^(k) and the monotonically increasing weight function g (x) Estimates

_w ^(k) = _w ^(k)g (| _w ^(k)| z_w ^(k)      (25)

educated. A non-linear post-processing of the estimates _w ^(k) then reduces the harmful effect of the imperfect Channel estimate if the mean square error

compared to Eq. (21) decreased. A suitable quality measure z _w ^(k) is the reciprocal

the standard deviation of an estimate _w ^(k). With z_w ^(k) to Eq. 27 grows E {| _w ^(k)| z_w ^(k)with increasing quality one Estimate _w ^(k). In general, the recipient does not Standard deviation σ_hw ^(k)but an estimate of the standard deviation σ_hw ^(k) before, so that even an estimate of the Quality measure is available.  

With the Heavyside step function h (x) and the constant ζ becomes a special weight function g (x)

g (x) = h (ζx-1) (28)

chosen, d. H. all estimates _w ^(k)whose amount is smaller as ζz_w ^(k), are set to zero and those that are not set to zero Estimates _w ^(k) are unchanged as _w ^(k) accepted. The constant ζ from Eq. (28) is a normalized threshold, with of the estimates _w ^(k) be compared. Through non-linear Estimation postprocessing can then be a profit in comparison achieved using unprocessed estimates, if only a few, a priori unknown coefficients of a estimated channel impulse response Achieve total energy of the channel impulse response. Typically must be the number W of the channel coefficients to be estimatedH _w ^(k)  of a subscriber be so large that channel impulse responses long duration, as z. B. occur in hilly terrain, can be estimated in line with expectations. For this purpose, the GSM Recommendation 05.05: Transmission and Reception pointed out. The spreading conditions are in rural areas much cheaper and the duration of a channel impulse response for this reason much less than in hilly terrain. By applying a non-linear Postprocessing of estimated channel impulse responses can be a simple, non-parametric channel estimator in a CDMA Mobile radio system can be used by the choice of Weight function and taking into account the disorder flexible to the effective duration of a channel impulse response can be adjusted.

Claims (21)

1. A method for the time-discrete transmission of signals from a transmitting side to a receiving side, the signals being able to be transmitted simultaneously via generally different channels, the signals from the transmission being linearly modulated onto the signals, and estimates of the channel impulse responses being available in the respective receiver and for each component of these estimates there is a quality measure, characterized in that the components of the estimates of the channel impulse responses are subjected to an estimated postprocessing before the signals are detected, so that modified channel impulse responses are generated. 2. The method according to claim 1, characterized in that the estimated postprocessing is non-linear. 3. The method according to claim 2, characterized in that the non-linear post-processing is designed as threshold processing, at of those estimated components, the amount of which is a certain one Below the threshold, be set to zero. 4. Method for time-discrete message transmission, at that from a transmitting device to a receiving device at the same time K data blocksd ^(k) = (d₁^(k),d₂^(k) . . .d _N ^(k)), k = 1. . . K, about K, generally different channels with the impulse responses H ^(k) = (H₁^(k),H₂^(k) . . .H _W ^(k)), k = 1. . .K, are transferred, whereby

• - Before the transmission, each of the data blocks d ^(k) is assigned a code c ^(k) ≡ ( c ₁ ^(k) , c ₂ ^(k) ... c _Q ^(k) ), k = 1. . .K, is modulated linearly so that the received signal e = ( e ₁, e ₂... E _{NQ + W-1} ), taking into account an additive interference signal n = ( n ₁, n ₂ _... N _{NQ + W- 1} ) with the NQ × N matrices C ^(k) = ( C _ÿ ^(k) ), k = 1. . . K, and with the (NQ + W-1) × NQ matrices H ^(k) = ( H _ÿ ^(k) ), k = 1. . .K, form assumes
• - Estimates in the recipient ^(k) = ( ₁^(k), ₂^(k) . . . _W ^(k)), k = 1. . . K, the Channel impulse responsesH ^(k) are present
• - in the receiver for each component _w ^(k), w = 1. . . W, a measure of quality e.g._w ^(k), k = 1. . . K, which with increasing quality of the Estimate _w ^(k) increases monotonously, or an estimate of Quality measure z_w ^(k) is present
  characterized in that
• - Modified in the receiver with a weight function g (x) Channel impulse responses ^(k), k = 1. . . K, with the components _w ^(k) = _w ^(k)g (| _w ^(k)| z_w ^(k)) are formed  
• - and as estimates ^(k) = ( ₁^(k), ₂^(k) . . . _N ^(k)), k = 1. . . K, the data blocks ^(k) those blocks ^(k) serve with the (NQ + W-1) × NQ Matrices ^(k) = ( _ÿ ^(k)), k = 1. . . K, the deviation of the signal of received signale minimize.

5. Method for time-discrete message transmission, at that from a transmitting device to a receiving device at the same time K data blocksd ^(k) = (d₁^(k),d₂^(k) . . .d _N ^(k)), k = 1. . . K, about K, generally different channels with the impulse responses H ^(k) = (H₁^(k),H₂^(k) . . .H _W ^(k)), k = 1. . . K are transmitted, whereby

• - Before the transmission, each of the data blocks d ^(k) is assigned a code c ^(k) = ( c ₁ ^(k) , c ₂ ^(k) ... c _Q ^(k) ), k = 1. . . K, is modulated linearly, so that the received signal e = ( e ₁, e ₂ _... C _{NQ + W-1} ), taking into account an additive interference signal n = ( n ₁, n ₂ _... N _{NQ + W-1} ) with the NQ × N matrices C ^(k) = ( C _ÿ ^(k) , k = 1 _... K, and with the (NQ + W-1) × NQ matrices H ^(k) = ( H _ÿ ^(k) , k = 1 _... K, form assumes
• - Estimates in the recipient ^(k) = ( ₁^(k), ₂^(k) . . . _W ^(k)), k = 1. . . K, the Channel impulse responsesH ^(k) are present
• - in the receiver for each component _w ^(k), w = 1. . . W, a measure of quality e.g._w ^(k), k = 1. . . K, the higher the quality of the estimate _w ^(k) increases monotonously, or an estimate of the quality measure e.g._w ^(k) is present
  characterized in that
• - Modified in the receiver with a weight function g (x) Channel impulse responses , k = 1. . . K, with the components _w ^(k) = _w ^(k)g (| _w ^(k)|_w ^(k)) are formed
• - and as estimates ^(k) = ( ₁^(k), ₂^(k) . . . _N ^(k)), k = 1. . . K, the data blocks d ^(k) those blocksd ^(k) serve with the (NQ + W-1) × NQ Matrices ^(k) = ( _ÿ ^(k)), k = 1. . . K, the deviation of the signal of received signale minimize.

  6. The method according to claim 4 or 5, characterized in that the weight function g (x) growing monotonically with x. 7. The method according to any one of claims 4 to 6, characterized in that the weight function g (x) is steady. 8. The method according to claim 4 or 5, characterized in that the weight function g (x) is discontinuous. 9. The method according to any one of claims 4 to 8, characterized in that the quality measure z_w ^(k) of the Reciprocal of the standard deviation of the estimate _w ^(k) from the real valueH _w ^(k) is or the quality measure_w ^(k) the estimate of the Reciprocal of the standard deviation of the estimate _w ^(k) from the real valueH _w ^(k) is. 10. The method according to any one of claims 4 to 9, characterized in that with the Heavyside'schen Jump function h (x) and one of the reception quality dependent constant ξ the weight function g (x) = h (ξx-1). 11. The method according to any one of claims 4 to 10, characterized in that the size as a reducing deviation of the received signal e from the signal ê is used. 12. The method according to any one of claims 4 to 11, characterized in that the size to be minimized deviation of the received signal e from the signal ê is used. 13. The method according to any one of claims 4 to 12, characterized in that to determine continuous value Estimates _n ^(k), n = 1. . . N, the elements d _n ^(k), n = 1. . . N, the data blocksd ^(k) an expectant Minimal variance estimator is used. 14. The method according to any one of claims 4 to 13, characterized in that to determine continuous value Estimates _n ^(k), n = 1. . . N, the elements d _n ^(k), n = 1. . . N, the data blocksd ^(k) an optimal linear Estimator is used. 15. The method according to any one of claims 4 to 13, characterized in that to determine continuous value Estimates _n ^(k), n = 1. . . N, the elements d _n ^(k), n = 1. . . N, the data blocksd ^(k) matched filters be used. 16. The method according to any one of claims 4 to 15, characterized in that for determining Estimates _n ^(k) of the elementsd _n ^(k) of the data blocksd ^(k) an optimal one Impact estimator is used.   17. The method according to any one of claims 4 to 16, characterized in that the components _w ^(k) of the modified channel impulse responses ^(k) with a digital Signal processor can be calculated. 18. The method according to any one of claims 4 to 17, characterized in that the components _w ^(k) of the modified channel impulse responses ^(k) with a microprocessor be calculated. 19. System for time-discrete transmission from one transmission side to a receiving side, the news about in general different channels can be transmitted simultaneously, each of the messages from the transmission being associated with it Code is modulated linearly, and in each case Receiver estimates of the channel impulse responses are available and a quality measure for each component of these estimates is present, characterized in that on the receiving end a digital signal processor is provided which Components of the estimates of the channel impulse responses before Detection of messages from an estimated postprocessing undergoes so that modified channel impulse responses are generated become. 20. System according to claim 19, characterized in that the digital signal processor a non-linear Performs estimated postprocessing. 21. System according to claim 19 or claim 20, characterized in that the signal processor a linear channel estimator (KS), an arrangement for Estimated postprocessing (NN) and a linear data estimator (DS) includes.