Description Method for Channel Estimation in a TDMA Mobile Radio System In TDMA-based CDMA mobile radio systems (TDMA = Time Division Multiple Access, CDMA = Code Division Multiple Access) a channel estimate is required to carry out the data estimation. The channel estimation in the TDMA-based systems is based on a training sequence, which is generally referred to as the midamble, the midamble being arranged between two data blocks. The unit consisting of the midamble and the two data blocks is called a burst.
From Bernd Steiner it follows: "A contribution to the mobile radio channel estimation with special consideration of synchronous CDMA mobile radio systems with joint detection", progress reports VDI, row 10, no. 337, Düsseldorf: VDI-Verlag 1995 that the long lm of the midamble in TD CDMA mobile radio systems, due to the algorithm used for channel estimation, directly from the number of active subscribers K and the maximum expected length W of the channel impulse response from the approximation
Lm = W (K +1) is determined. The length Lm of the midamble is therefore a limiting factor for the number of active participants that can be present in the uplink per burst.
If techniques such as "voice activity" or adaptive data rates are to be used, it may even be necessary to support more participants than are actually active per burst, in other words, to include them in the channel estimation.
If intelligent antennas are also used in the base station, i.e. antennas that act selectively in the direction of the mobile radio subscriber, for example, subscriber-specific midambles are also required in the downlink, so that the midamble length also becomes a limiting factor here.
Furthermore, it may be desirable in TD-CDMA mobile radio systems with a frequency reuse factor of r = 1 to include intercell interference signals from neighboring cells in the data selection. To do this, it is necessary to estimate channels to the neighboring stations or the mobile stations of the neighboring cells.
However, this is only possible if sufficiently long middle ambles are available. As a result, there is the unsatisfactory situation that the more channels that have to be estimated, the smaller the remaining portion of the burst available for the data transmission. Therefore, on the one hand, many channels should be estimated, but on the other hand, only as small a portion of the burst as possible should be sacrificed for the midamble. In principle, this problem does not only exist with the above-mentioned channel estimation method of the publication mentioned above, but also with all estimation methods.
In order to be able to achieve a sufficiently precise estimation result, a minimum number of measured values from the received midamble is required.
So far, this conflict has not been resolved satisfactorily. Rather, a compromise has so far been found between the midamble length and the burst portion available for data transmission. The mid-bamboo length is usually dimensioned such that at least as many channels can be estimated as CDMA codes can be detected simultaneously in one cell. The number of CDMA codes that can be detected simultaneously is essentially determined by the spreading factor used and the intercell interference.
Therefore, a useful compromise between mid-range length and the number of channels that can be estimated is only possible if the spreading factor is not too large, in other words if relatively few participants share the data transmission capacity of a burst. This means that the minimum data rate is too high for many applications and offers little scope for applying the above-mentioned capacity-increasing measures such as voice activity, adaptive data rates, antenna diversity and elimination of intercell interference.
The object of the invention is therefore to create a method and a device with which an estimation of an increased number of channels is possible without increasing the proportion of the burst for the midamble.
The object is achieved by the features of independent claims 1 and 10. Preferred embodiments of the invention are the subject of the dependent claims.
The maximum time interval between two channel impulse responses within which the channel impulse responses can be regarded as correlated is defined as the coherence time TK.
Furthermore, the coherence time TK is inversely proportional to the speed of a participant.
If the channel impulse responses of successive bursts are now almost the same, i.e. the time period Tauf in which the successive data bursts are transmitted is significantly smaller than the coherence time TK defined above, the reception values of several bursts that can be used for the channel estimation are used to estimate a single channel impulse response. Participants with low speed in particular meet the condition Tauf TK.
Short midambles (also called training sequences) can preferably be transmitted within the burst, the channel estimate then determining a current channel impulse response from the received values e (x) of several successive bursts x originating from the midambles. The received values of so many bursts must be taken into account in the channel estimation that the quality required for the channel estimation is achieved. This procedure is possible only if the channel impulse responses within the period in which the required receive values e (x), x = 1 ...
X are received, are largely identical, in other words the condition Tauf TK is fulfilled. Furthermore, the lower the speed of a mobile station, the shorter the midamble, since this increases the coherence time TK and the higher the data rate can be achieved.
The principle of piece-by-piece transmission of the midamble can consequently increase the number of participants. However, if you leave the length of the midamble within the bursts and interpret them as sections of a long midamble, it is possible to accommodate significantly more participants.
Therefore, the problem of the number of subscribers normally limited by the channel estimation in TD-CDMA mobile radio systems can be regarded as solved.
Furthermore, given the validity of the Tauf TK estimate, it is not absolutely necessary to use midambles within a burst, but the use of preambles is generally possible. Because of their property that they are not disturbed by the interference of previous symbols, preambles can advantageously be chosen to be shorter than midambles, which, for example, increases the data rate. If the preamble and the midamble are of equal length, the use of preambles enables a higher number of participants.
Preferred embodiments of the invention are described below with reference to the drawings.
Fig. +1 shows the division of a midamble into several shorter midambles, and Fig. 2 shows the use of periodic midamble codes.
1 shows the division of a midamble m into a plurality of midamble m (x), where x runs from 1 to X. By sending the midamble m in sections in successive bursts, the midamble of a burst is shortened, so that the number of channels that can be estimated is increased due to the short midambles.
In the case described here by way of example, the midamble m consists of X = 5 blocks of length W. The received signal originating only from the not yet divided midamble m is calculated for any subscriber k
The total received signal originating from the middle tones of the participants results from the superimposition of the signals of the K participants
Furthermore, the reception sequence em is shorter by W elements than the midamble, since reception values with interference from the data blocks cannot be used for channel estimation.
So that the same algorithm can be used for channel estimation after splitting the midamble over several bursts as previously, which is based on the reception sequence Sm of the long KW, the signal em must also be generated by the transmission of the midamble in sections. In the example in FIG. 1, the signal 2 is composed of four sub-signals Sm = [em (l) ... em (4)] when the midamble m is distributed over X = 4 bursts.
The received values of em (1) result from the transmission sequence of the midamble part m (1) and, since the channel impulse response has the length W, from the W-1 previous samples from block m (0). The received values of em (2) result from the transmission sequence m (2) and, since the channel impulse response has the length W, from the W-1 previous samples from block m (1). The same applies to the received values em (3) and e = (4).
After the transmission of the four bursts, as shown in FIG. 1, the identical signal em = [em (1) ... em (4)] can be composed of the four reception sequences if the channel impulse response is constant. The channel estimation can be done inexpensively according to a method described in the article mentioned above.
The estimation of the channel impulse response can be carried out in the steady state after the reception of each burst and thus of each section em (x) if, for example, an older em (1) is replaced by a just received section of the same name. The result of the channel estimation is therefore adaptively tracked to the slowly changing channel.
Fig. 2 shows how blocks of the midamble emerge from a periodic midamble basic code described in the above-mentioned publication. If the supported number of subscribers K is a multiple of the number X of bursts to which the midamble is divided, the midamble distribution on the channel estimator manifests itself as a cyclical interchanging of the midambles used by the subscribers. Exactly the same channel estimator can be used for each burst, which uses the midamble reception signals of the last X bursts.
The estimated value h of the channel impulse responses is obtained with the constant, right-circulating matrix G-1, which depends only on the midamble codes used, for h = G 1 em (3) Since only a part of the reception sequence em is newly acquired per burst, it is not absolutely necessary to perform a full channel estimate on each burst. Since it is a linear estimation algorithm, one can instead determine the difference between the old and new reception sequence Sm directly from the difference between the old and new reception sequence and update the estimation result h.
In the example in FIG. 2, X = K = 4. The periodic basic code PGC is divided into four components, each beginning with ml, mW + lt m2W + lt and m3W + l. The periodic basic code PGC ends with the character mKW. The basic code PGC is always periodically shifted by one block between the participants T1, T2, T3 and T4, the participants ml sending the components ml and mW + l in the midamble M1, while the participants T2 send the components mW + and m2W + l
The subscriber T3 sends the components m2W + l and m3W + l and the subscriber T4 the components m2W + l and ml in this order. Correspondingly, the 8 components of the periodic basic code PGC which are arranged above it in FIG. 2 are sent in the middle messages M2 to M4. The midamble M1 is sent in the first burst B1, while the midambles M2 to M4 are sent in the corresponding subsequent second to fourth bursts. 2, the components of the first midamble M1 sent in the first burst B1 are shown in the partial image at the bottom right.
The invention can be applied to any mobile radio system with a TDMA component, and is therefore not restricted to use in TD-CDMA systems.