METHOD AND SYSTEM FOR SUBBAND INDICATOR SIGNALLING Field of the invention
The present invention relates to a method and system for subband indicator signalling. Background of the invention
In 3GPP Evolved UTRA (E-UTRA) frequency domain scheduling is an important feature. To enable such scheduling for downlink in case of frequency domain duplex, frequency dependent channel quality information needs to be transmitted from a mobile terminal to a base station (Node B) in some form. In E-UTRA a channel quality indicator (CQI) is transmitted to report the channel quality. The whole frequency band can be divided into subbands wherein the channel quality (CQI) then is estimated separately for each subband. Furthermore, in E-UTRA using MIMO precoding, a preferred precoding vector index (PVI) is transmitted from the receiver to the Node B to report the preferred MIMO precoding vector. Due to the frequency selectivity of the channel, the whole bandwidth need also in this case be divided into subbands, and the PVI is estimated separately for each subband.
The most straightforward way to transmit channel quality or precoding vector information is to transmit the PVIs and CQIs for all subbands . However , such a scheme may cause a considerable amount of signalling. Therefore, several alternatives have been proposed to reduce the amount of signalling. Such alternatives, of which only the first will be elaborated further here, include, among others: to only feedback information about the Q subbands which have the best CQI, to apply differential feedback information in time or frequency, to use bitmap techniques indicating which subband or subbands reflect a reported CQI/PVI value, to use a
hierarchical tree structure, and to use a set of orthogonal functions to approximate a frequency selective fading profile, see, for example, 3GPP TR 25.814 v.1.0.1 "Physical Layer Aspects for Evolved UTRA," December 2005. Regarding the first alternative, this can be further be elaborated by the following method: send one CQI value representative for the best Q subbands , send the indices of these subbands, and send a CQI representative for the whole frequency band. Such a method has been shown to give only a small degradation in throughput compared to a method wherein CQI values are transmitted for each and every subband, see 3GPP TSG RAN, Rl-060228, Huawei, "Sensitivity of DL/UL Performance to CQI-Compression with Text Proposal," January 2006.
An important issue in this method, however, is how to transmit the indices of the Q subbands for CQI reporting (and
PVI reporting in case of precoded MIMO) with a minimum amount of signalling. If there are N subbands, indexed 1, 2, ..., N, thus constituting a set of N subbands, a sequence of N bits can be transmitted, wherein the corresponding bit positions of the indices of the particular Q subbands may, e.g., be set as ones, while the remaining bit positions are set as zero. This method, however, requires rather extensive signalling.
One example of a prior art compression technique for bitmaps in general is run-length encoding, see A. Bookstein and S. T. Klein, "Construction of Optimal Graphs for Bit-Vector Compression," Proc. 13th annual international ACM SIGIR conference on Research and development in information retrieval, pp 327-342, 1990, Brussels, Belgium, wherein a set of consecutive ones/zeros is represented by a flag indicating one/zero and the number of ones/zeros. However, such a compression technique cannot be guaranteed to always compress the input, in particularly not when, e.g., 5 subbands of 24 are
indicated, in which case selected subbands may be arbitrarily distributed among the 24 possibilities.
A more efficient method relies on the fact that there are
/N\ (n\ ways to select Q subbands out of N subbands , where is [Q) [kj the binomial coefficient :
Thus , to convey the indices of Q out of N subbands , at
least bits are needed compared to N bits if a simple
bitmap representation is used. pel denotes the smallest integer greater than or equal to x. For example, if -V=24, and Q=S, 16 bits are needed instead of 24 bits for the bitmap. In this way, it is possible to obtain a significant data signalling reduction when signalling the indices of the subbands .
A straightforward method to map the possible subsets of Q subbands selected from the N subbands onto the minimum number of bits required is to tabulate them. The possible subsets of Q subbands raises fast with increasing Q. For example, in the case of iV=24, and Q=5 , there exist more than 40000 different subsets. However, arranging a table of possible subsets in a way that enables sufficiently fast mapping and de-mapping presents a problem.
Consequently, there exists a need for an improved method for signalling subband identities .
Summary of the invention It is an object of the present invention to provide a system and a method for signalling subband identities from a first transceiver to a second transceiver in a wireless data
communication system, which provides a more efficient way of obtaining a representation of a particular subset of subbands .
According to the present invention, groups of subsets are established, wherein each subset in a group contains the identities of the same number of subbands, and wherein subsets within a group are ordered, and identities of a subset of a set of subbands are signalled from the first transceiver to the second transceiver by signalling a representation of the identity of the group of the subset and a representation of the position of the subset in the group. For example, the subset of the subbands may be represented by a number r, wherein r is determined by the identity of the group of the subset and the position of the subset in the group.
This has the advantage that the first transceiver, e.g., a mobile terminal, in a simple and efficient manner can obtain a representation of the subset for transmission to the second transceiver, e.g. a base station. This further has the advantage that the representation may be obtained faster and using fewer resources in the mobile station. The signalling of a subset of subbands may be used, e.g., in signalling of CQI and/or PVI and/or any other subband specific parameters.
The identity of the group may be determined by a group specific offset . This has the advantage that the identity can be obtained in a simple manner. The method may further comprise the steps, wherein subbands are indexed ( 1 , ... , N) , of : a) ordering subsets Tn1,..., mQ, in a group such that the subset wherein the lowest index mi has the highest possible value n is arranged first in the group, followed by all subsets wherein the lowest index is n-1 and so on until the value of the lowest index is 1,
b) setting 1 = 1 , c) ordering subsets having an equal lowest index irij, or, if i>l, equal lowest indices πii,..., Jn1, such that the first subset is the one in which the lowest of the remaining indices m±+i has the highest possible value, then all subsets such that the lowest of the remaining indices has the next to highest possible value, and so on until the lowest of the remaining indices is /Hj+1, d) setting 1 = 1+1 and repeating step c) until i = Q-I. This has the advantage that the representation of a specific subset may be obtained even faster. Further, since these steps may be performed using mathematical calculations instead of a table look-up, considerable memory savings in a mobile terminal may be made. The present invention also relates to a system and a communication system.
The invention will be explained more fully below with reference to the appended drawings .
Brief description of the drawings Fig. 1 shows an exemplary wireless data communication system in which the present invention advantageously may be utilised.
Fig. 2 shows an example of possible communication resources in a communication system according to fig. 1. Detailed description of preferred embodiments
In fig. 1 is shown a communication system 1 wherein the present invention advantageously may be utilised. In the figure is shown a base station antenna 2, having capabilities to communicate with one or more mobile terminals 3. The communication resources consist of at least one frequency band.
which is divided into subbands . This is disclosed more in detail in fig. 2, in which a communication resource scheme suitable for use with the present invention is shown. As is shown in the figure, the frequency spectrum of the communication system is divided into 24 subbands [nlr ..., nN) , for example constituting equal portions of the frequency spectrum, as equal frequency subbands are preferred to facilitate resource management (for example, it is easier to allocate the available resources). However, division into non- equal frequency subbands is, of course, also possible.
The communication resources may be divided into time-slots (not shown) in the time domain, typically having a certain length, e.g. a number of OFDM symbols, and wherein a user may be allocated all or part of one or more time-slots in one or more subbands. Alternatively, a user may be allotted all or part of a subband, wherein one or more users continuously may be communicating at the same subband.
Only one base station is shown in the communication system in fig.l. As is apparent to a person skilled in the art, however, the communication system 1 may comprise a plurality of base stations , each providing coverage in part of the communication system. Further, the coverage area of one base station may be divided into sectors.
The channel quality for a specific mobile station may vary substantially among the various subbands , even between closely located subbands , e.g., between ni and n2 or between Xi1 and n3 in fig. 2. Consequently, in order to utilise communication resources as efficiently as possible, it is necessary that the mobile station communicates channel parameters, such as channel quality (CQI) and a preferred precoding vector index (PVI) to the base station. As the parameters may vary substantially from subband to subband, this information needs to be transmitted to
the base station. However, as stated above, signalling channel parameters for all subbands require extensive signalling, and a way of reducing this signalling is to only transmit the identities of those subbands that exhibit the most favourable properties. Alternatively, if all but a few channels provide satisfactory properties, subband identities of those subbands providing poor channel properties could be signalled instead, as this would offer the base station enhanced flexibility when allocating channel resources . Still, as it is inherent in a wireless communication system to utilise available bandwidth as efficiently as possible, also the amount of data needed to transmit identities of a limited number of subbands should be signalled to the base station as efficient as possible, i.e., if each subset of Q subbands is considered as a bitmap of N bits, Q of them being ones, the problem to be solved is how to transmit indices of Q subbands out of a total of N subbands with a minimum number of bits and in a way that allows fast mapping from the subset of indices to the sequence of bits as well as fast de-mapping. The mapping can be viewed as a compression from a bitmap of N bits, where the original bitmap has the restriction that there are exactly Q ones . The problem to be solved is then to devise an efficient compression of the bitmap.
As has been disclosed above, to convey the indices of Q out of
N subbands (2)
bits are needed to signal the indices to the base station (node B) from the mobile station. Although this is a substantial improvement as compared to using a simple bitmap, as described above, it is still very important that the mobile station in a short period of time and with a minimum load on processing
resources is capable of determining the particular representation that represents Q subbands out of N subbands .
For example, if, as in fig. 1, there are 24 subbands, i.e. iV=24, and five particular ones of these (£?=5) are to be transmitted to the base station, there are more than 40 000 possible combinations. The number N may, however, be even larger, usually depending on the bandwidth. For a bandwidth of 20 MHz, N may be 48. Further, it is common to allow transmission of identities (indices) of an arbitrary number of subbands up to a given maximum. Using the same example, if any number of subbands up to 5 are allowed, there are more than 55000 possibilities. If each of these possibilities is represented by a number, it is difficult to obtain the particular number to transmit to the base station (e.g., in binary form) in a sufficiently quick and efficient manner.
The present invention provides a method for mapping a subset of subband indices selected from a total set of N subbands on a minimum number of bits in a way that enables fast mapping and de-mapping. In the general case, the allowed numbers Q of subbands in the subset is given by the set M ={mx,m2,...,nij} of unique non- negative integers where 0 ≤m. ≤N, j =1,2,...,J .
The number of bits required to map all possible subsets is then
The degree of reduction in signalling due to this mapping depends on N and the set M. For the above example, where there are in total N = 24 subbands and 0 to 5 subbands are in the subset of best subbands, i.e. ^={0,1,2,3,4,5}, 0, 1, 2, 3, 4 or 5 subbands can be indexed using only
instead of _V=24 bits.
According to the invention, each subset of indices is labelled by a number r. E.g., a subset may comprise the indices
1,7,12,15,21. The subset of indices is mapped on, e.g., a sequence of bits as a binary representation of r. However, it is clear to anyone skilled in the art that the number r can also be mapped on a sequence of symbols from another alphabet, for example, as a ternary representation or a decimal representation .
In the case when r is mapped on a binary sequence of N bits , the range of r must be within [0, 2^-1].
According to the present invention the range of r is divided into J intervals such that all values of r within interval j represent subsets of πij indices. The smallest r within interval j is called the jth offset Oj. Each interval must be large enough to contain all subsets with πij indices, i.e.
(5)
must be valid. The minimum offsets possible are given by
The invention will be further exemplified using an example wherein N=S, M=Im1=O, mz=l , m3=2} , and N=A. In this example, the offsets are O1=O, O2=I, O3=6. Consequently, r=0 for the empty subset, 1 = r = 5 for subsets containing one index, and 6 = r
= 15 for subsets of two indices, as is shown in Table 1.
Table 1
Accordingly, the number r labelling a subset of Q=πij indices can be obtained as the sum of the offset Oj plus the position of the subset in a group of subsets . The above definition of Oj is exemplary and may be -• defined in numerous other ways . For example, Oj may be calculated as the last position in a group of subsets, wherein the number r is obtained as Oj subtracted by a certain number. Using the above method of dividing the range of r into intervals, or groups, a considerably quicker "look-up" may be obtained, i.e., the particular r to be transmitted is identified by the mobile station more efficient as compared to the prior art. In this way, the base station will receive a more accurate measurement , in particular when the mobile station is moving fast or the propagation properties are subject to frequent and sudden changes. Further, the
present invention also has the advantage that the processor load in the mobile terminal is reduced.
However, the above method can be enhanced even further by introducing an additional step, which will be described in the following.
In the alternative embodiment, the additional step consists of sorting the subsets in each group in a particularly efficient manner.
First, the indices of a subset are sorted. The sorted subset of Q indices out of the total number of indices N1 wherein the N indices ranges from 1 to N may be expressed as
r io i l≤sk ≤N kϊto ■ wherein . (7)
Next , the subsets in a group are ordered in the following way: The first subset in the group is the one with sO=N-Q+l, followed by all subsets with S0=N-Q and so on until so=l.
In a next step all subsets of indices wherein so=x, the subsets of indices are ordered such that one first finds all subsets such that sχ=N-Q+2 , then all subsets such that S1=N-Q+! , and so on until S1=X+!. The same procedure to order the subsets of indices is repeated for each index k in the subset. For Q=O, only a single set exists and no ordering is needed, r is hence given by the offset Oj plus sum of the number of all subsets with Q indices larger than s0 and the number of all subsets with Q-I indices larger than S1 and so on. The mapping for the previous example N=5 , M=Im1=O, m2=! , Λ73=2}, and offsets C1=O, O2=I, O3=6 is illustrated in Table 2.
Table 2 Mapping example
The above mapping (ordering) of subsets allows for fast mapping and de-mapping, even without need of a full table, as will be shown below. Further, it is, of course, equally possible to arrange the list such that one first finds ≤ro=l, followed by all subsets with so=l+l and so on until sQ=N-Q+l , whereupon the following equations will be changed correspondingly.
Continuing with the arrangement as disclosed in table 2, if sk=x, then it comes after all subsets with the same indices Sj, j<k, and with x < sk = N in the list. The number of such
1) in the range from x+1 to N. In case sk has its highest possible value, i.e. sk =N-Q+k+l, there are of course no
subsets with higher values of sk. In that case N-sk = Q-k -lbut
(Q~k-1\ is not mathematically defined. Therefore, the notation { Q-k ) may be simplified by defining the extended binomial
It is now easy to calculate r for a given subset of πij indices by evaluating and summing the numbers of all possible subsets with higher values of the indices and by adding the offset Oj. The mapper takes the subset PA/^Q 1 as input and creates a number _r in the range 0≤r≤2N-l using the following equation:
where O
1 is obtained by ( 6 ) . The number r can now be
transferred to the Node B using N bits.
Accordingly, the present invention has the substantial advantage that r can be calculated using a mathematical expression without a need for storing or producing a table each time a subset of subband indices is to be signalled to a base station.
The present invention further has the advantage that it is equally, or substantially equally easy to extract r from the received sequence of bits in the receiver (base station) , and this de-mapping algorithm will now be described. The task of the de-mapper is to extract r from the received sequence of
bits, detect J and then regenerate the subset of indices \sk/^0 "1 This is performed in two steps, according to the following:
Step 1 j is given as the largest integer such that Oj ≤r . Find j and calculated :
r'=r-Oj .
Step 2
Given j and the number r' from Step 1 set
and the indices P
A-J
AI
O W^1
now t*
e found by executing the following algorithm, given in a generic software code form: for £=0 to Q-I
Find the largest integer
s
k = x
/•' = r'-m end
An example of a more detailed algorithm is shown below:
V =1
Λmin x for A:=0 to Q-I
X =*„
/N-x\ m = ι \Q-kl while m > r' x = x +1
end s
k=x
r' = r'-m end
/n\ The mapper and de-mapper need the binomial coefficients
\k) for n = 0,...,N -1 and k =0,l,...,max{m;-} to perform the above operations. These coefficients can be pre-calculated and stored in a table.
As has been disclosed above, the present invention provides an efficient method for finding a representation of a particular subset of subband indices to be transmitted to a base station. The present invention further provides an efficient method for retrieving r in a receiver.