DE4212300C2 - Method and device for the simultaneous estimation of channels in digital message transmission in multiple access systems - Google Patents

Description

The invention is based on a method for simultaneously estimating the discrete-time impulse responses of digital news channels after the Preamble of claim 1.

With digital message transmission via time variant and frequency selective channels are required in the receivers for signal equalization Information about the current channel status. This information can can be obtained in that the transmitter in addition to the signals for data transmission send additional test signals that are not modulated with data after transmission over the channels for channel estimation on the receiver side can be used, see e.g. B. / 1,2,3,4 /. Regarding the recovery of the information obtained from the channel estimation in the data estimation reference is made to / 5 /.

In the uplink of mobile radio systems, those different from K should mobile transmitters and each via a different from K Channels arriving at the receiver of the base station in the Base station can be evaluated. The multiple access that occurs here problem is traditionally caused by the multiple access methods FDMA (frequency division multiple access) and TDMA (time division multiple access) solved / 6 /. The signals from the individual mobile transmitters are in the process FDMA in the frequency domain disjoint, with the TDMA method, however, in Time range. A newer multiple access method, which due to its inherent frequency diversity benefits when transmitting over frequency selective channels and that u. a. for the mobile radio systems of third generation / 7 / could be considered is the CDMA method (code division multiple access), see e.g. B. / 8 /. In this process, the Signals from the individual mobile transmitters neither in the frequency nor in the time domain disjoint. For channel estimation with test signals, this means that in The receiver's responses to the test signals are not separated Form, but as a sum. You have the problem of Multiple channel estimation. The known methods of channel estimation / 1,2,3,4,9,10 /, in which a frequency or temporal separation of the  Channel response to the test signals is an essential requirement can therefore not be used with CDMA. This known method are designed so that only a single channel can be estimated, not however at the same time several channels that are neither in time nor in Frequency range are disjoint. With a known CDMA system / 11 / bypasses the described problem of multi-channel estimation in that in the receiver all CDMA signals received at the same time except one only sees as noise. With this procedure you do without however, on the a priori knowledge of the CDMA codes of the other participants and can therefore only accommodate a relatively small number of participants / 12 /.

The invention has for its object the problem of multiple channels estimate in CDMA systems can be solved at low cost. The invention The solution is outlined below. Here, time is discrete Signals that can be real or complex. Vectors and Matrices are identified by underlining. i is the discrete time.

Fig. 1 shows the K channels ( 1 ), their impulse responses

{h ^(k) (i)} = {h ^(k) (1), h ^(k) (2). . . h ^(k) (W _k ′)}, k = 1. . . K (1)

should be estimated with the sizes

W _k ′ ∈ IN, k = 1. . . K (2)

are the number of components of impulse responses. The number W _k 'of the components of a discrete-time impulse response is proportional to the duration of the associated continuous-time channel impulse response. With the sizes

W _k ∈ IN, W _k W _k ′, k = 1. . . K (3)

the impulse responses {h ^(k) (i)} according to Eq. (1) as vectors

write. The vectors h ^(k) according to Eq. (4) can become a vector

h = [ h ^{(1) T} , h ^{(2) T.} . . h ^{(K) T} ] ^T (5)

be united.

A test signal is placed in each of the channels (1) to be estimated, see FIG. 1

{a ^(k) (i)} = {. . . a ^(k) (1), a ^(k) (2), a ^(k) (3),. . .}, k = 1. . .K (6)

fed. The channel output signals caused by these test signals are summed up in a summer ( 2 ) to form the sum signal {s (i)}. An interference signal {n (i)} is added to the sum signal {s (i)} in the adder ( 3 ). {n (i)} represents the received noise, the noise of the receiver input stage, etc. At the output of the summer ( 3 ) you have the signal

{e (i)} = {s (i)} + {n (i)}. (7)

From this signal, the impulse responses {h ^(k) (i)}, k = 1. . . K, see Eq. (1) to be estimated.

There are endless possibilities to test signals (a ^(k) (i)) according to Eq. (6) to choose. The choice according to the invention, which is not obvious, consists in all K test signals from one and the same periodic basic signal

{p (i)} = {. . . p (P), p (1), p (2). . . p (P), p (1). . .} (8th)

the period  

according to the regulation

to generate. The main claim relates to this special choice of test signals {a ^(k) (i)}, k = 1. . . K.

The new choice of test signals {a ^(k) (i)}, k = 1. . . K according to Eq. (10) suggests a special way of performing the estimation as covered by claim 2.

As shown below, according to Eq. (10) generated test signals {a ^(k) (i)} for particularly simple implementations of the estimator. In FIG. 2, for the case K = 3, W ₁ = 3, W ₂ = 2, W ₃ = 5, the generation of the test signals {a ^(k) (i)}, k = 1,2,3, according to Eq . (10) illustrates.

From each of the test signals {a ^(k) (i)} according to Eq. (10) one can have a matrix with P rows and W _k columns

A ^(k) = (A _ÿ ^(k) ), k = 1. . . K, (11) (a)

A _ÿ ^(k) = a ^(k) (W _k + ij) (11) (b)

form. The matrices A ^(k) , k = 1. . . K, according to (11a) can in turn lead to a square matrix of dimension P

A = ( A ⁽¹⁾ , A ⁽²⁾ ... A ^(K) ) (12)

be summarized.

Due to the periodicity of the test signals {a ^(k) (i)} according to Eq. (10) is the sum {s (i)} at the output of the summer ( 2 ), see FIG. 1, also periodically with the period P according to Eq. (9). If one considers the output signal {e (i)} of the adder ( 3 ), see FIG. 1, for the duration P of a period of the sum signal {s (i)}, this section of the output signal {e (i)} can be compared with the Vector h according to Eq. (5), from which the interference signal {n (i)}, see Eq. (7), interference vector formed

n = [n (1), n (2). . . n (P)] ^T (13)

and with the matrix A according to (12) by the vector

e = A * h + n (14)

to be discribed.

Due to the construction rule given by the equations (10) to (12) of the matrix A , this matrix is circulating to the right. Hence, see e.g. B. / 13 /, is obtained with the right-hand circular matrix

C = A ^-1 (15)

an estimate  of the searched vectorH, see Eq. (4) and Eq. (5), according to the relationship

 =C. ·e. (16)

The right circulatory system of the matrixC. according to Eq. (15) allows the investigation from  according to Eq. (16) with a basis signal {p (i)} according to Eq. (8) inverse Filter / 2,4,10 / in a particularly simple way. Such one inverse filter can, as inFig. 3, advantageously by a correlator can be realized. The multipliers (4th) of Weight factors to be fed into the correlator  

g = [g (1), g (2). . . g (P)] (17)

be according to the elements

c = (c₁₁, c₁₂ ... c _1P ) (18)

the first line of the matrix C according to Eq. (17) chosen, ie

g = c . (19)

In the shift register (5) of the correlator is first clocked the reception signale according to Eq. (14) a. The switch (6) according to the institution Fig. 3 in position I. Then the shift register (5) of the correlator by flipping the switch (6) fed back into position II. One now clocks the correlator P times, for P see Eq. (9), so you get at the exit the correlator summer (7) the estimation  of the searched vectorH, out which the estimates of the impulse responses sought {h^(k)(i)}, k = 1. . . K, according to Eq. (5) and Eq. (4) emerge.

From Fig. 3, and the explanations given to show that it for estimating the channel impulse responses {h ^(k) (i)}, k = 1. . . K, is sufficient to have a section of the length P of the signal {e (i)} available. As a result, it is not necessary to send the test signals {a ^(k) (i)} periodically. Rather, it is sufficient to have finite sections of length

to send. This fact is the subject of claim 3.

Further refinements of the method are in claims 4 to 7 detected. Claims 8 and 9 relate to the implementation of the method.  

literature

/ 1 / Plagge, W .; Poppen, D .: New method for measuring the channel impulse response and carrier synchronization in digital mobile functions. Frequency 44 (1990) 7-8, pp. 217-221
/ 2 / Koo, D .: Method and Apparatus for Communication Channel Identification and Signal Restoration. U.S. Patent No. 5047,859, 1991
/ 3 / Koch, W .: Receiver for digital transmission system. German Patent No. 40 01 592 A1, 1991
/ 4 / Ruprecht, J .: Maximum-Likelihood Estimation of Multipath Channels. Hartung-Gorre, Constance, 1989
/ 5 / Proakis, JG: Digital Communications. McGraw-Hill, New York, 1983
/ 6 / Ochsner, H .: The future pan-European digital mobile radio system, Part 1: GSM recommendations and services. Bulletin SEV VSE 79 (1988) 11, pp. 603-608
/ 7 / Steele, R .: Deploying Personal Communication Networks. IEEE Communications Magazine 28 (1990), 9, pp. 12-15
/ 8 / Baier, PW; Jung, P .; Klein, A .: CDMA - an inexpensive multiple access method for frequency-selective and time-variant mobile radio channels. Telecommunications, Electronics 41 (1991) 6, pp. 223-227 and 234
/ 9 / Lüke, HD: Sequences with perfect periodic auto and cross correlation functions. Frequency 40 (1986) 8, pp. 215-220
/ 10 / Rohling, H .; Borchert, W .: On the mismatched filter design for periodic binary phase coded signals. ntz Archive 10 (1988) 5, pp. 111-117
/ 11 / Salmasi, A .; Gilhousen, KS: On the System Design Aspects of Code Division Multiple Access (CDMA) applied to digital cellular and personal Communication Networks. Proc. IEEE Vehicular Technology Conference (1991), pp. 57-62
/ 12 / Crozier, SN: Short-Block Data Detection Techniques employing Channel Estimation for fading Time dispersive Channels. Dissertation, Carleton University, 1990, p. 24
/ 13 / Marple, SL: Digital Spectral Analysis. Prentice-Hall, New Jersey, 1987.

Claims (10)

1. Method for simultaneous estimation of the time-discrete real or complex channel impulse responses {h ^(k) (i)} = {h ^(k) (1), h ^(k) (2). . . h ^(k) (W ′ _k )}, k = 1. . . K, the lengths W ′ _k ∈ IN of K channels ( 1 ) of the digital message transmission from the discrete-time signal {e (i)} = {s (i)} + {n (i)}, where

• - i is the direct time,
• - W _k ′ ∈ IN is the number of components of the discrete-time channel impulse response {h ^(k) (i)},
• the components h ^(k) (i) can be samples of a continuous-time channel impulse response,
• - {n (i)} is an interference signal and
• - {s (i)} is the sum of the channel output signals which arise when a time-discrete real or complex periodic test signal {a ^(k) (i)} = {. . . a ^(k) (1), a ^(k) (2), a ^(k) (3). . .}, k = 1. . . K, in each of the channels ( 1 ),

characterized in that
with the sizes W _k ∈ IN, k = 1. . . K, the K test signals {a ^(k) (i)} from a periodic discrete-time base signal {p (i)} = {. . . p (P), p (1), p (2). . . p (P), p (1). . .} of the period according to the regulation are formed, the number of components W _k 'W _k and the sizes of the BedingungenW _k W _k', k =. 1 . . Ker filling. 2. The method according to claim 1, characterized in that one with the impulse response of a faulty replacement channel Estimates components for which the estimate assumes that it responds to the base signal {p (i)} as an input signal with the signal {e (i)}, and whose impulse response is achieved by stringing the impulse responses {h ^{(k )} (i)}, k = 1. . . K, the K channels ( 1 ) is created. 3. The method according to one or both of the above claims, characterized in that only finite sections {a ^(k) (1), a ^(k) (2). . . a ^(k) (N)}, k = 1. . . K, with Components of the periodic test signals {a ^(k) (i)} are fed into the K channels ( 1 ), it being possible to send signals before and / or after these sections which are not used for channel estimation but for other purposes, e.g. B. the data transmission. 4. The method according to one or more of the above claims, characterized characterized in that the signal {e (i)} is stored before further processing  is and can be fed to further processing, which is not done in real time. 5. The method according to one or more of the above claims, characterized characterized in that the signal {e (i)} is one for the base signal {p (i)} undergoes inverse filtering, d. H. a filter whose response to the base signal {p (i)} apart from a single non-vanishing Value per period that occurs periodically at the interval of period P is zero. 6. The method according to one or more of the above claims, characterized in that the generation of the test signals {a ^(k) (i)}, k = 1. . . K and / or the processing of the signal {e (i)} in the time domain. 7. The method according to one or more of claims 1 to 5, characterized in that the generation of the test signals {a ^(k) (i)}, k = 1. . . K and / or the processing of the signal {e (i)} in the frequency domain. 8. The method and device according to one or more of the above claims, characterized in that the processing of the signal {e (i)} with a Correlator is done. 9. The method and device according to one or more of the above claims, characterized in that the generation of the test signals {a ^(k) (i)}, k = 1. . . K and / or the processing of the signal {e (i)} with signal processors and / or microprocessors.