Quantization of channel state information in multiple antenna systems

• US 8,116,391 B2
• Filed: 05/29/2007
• Issued: 02/14/2012
• Est. Priority Date: 05/26/2006
• Status: Expired due to Fees

• Current Assignee: Quarterhill Inc. (f/k/a Wi-LAN Inc.)
• Original Assignee: Wi-LAN Inc.
• Inventors: Mielczarek, Bartosz, Krzymien, Witold A.
• Primary Examiner(s): Payne, David C.
• Assistant Examiner(s): NGUYEN, LEON VIET Q
• Application Number:
  US11/754,965
• Publication Number:
  US 20100266054A1
• Time in Patent Office: 1,722 Days
• Field of Search: 370/281, 370/329, 375/260, 375/265, 375/267, 375/298, 455/561, 455/562.1
• US Class Current: 375/260
• CPC Class Codes: H04B 7/0434 using multiple eigenmodes
• CPC Class Codes: H04B 7/0452 Multi-user MIMO systems
• CPC Class Codes: H04B 7/0456 Selection of precoding matrices or codebooks, e.g. using matrices antenna weighting
• CPC Class Codes: H04B 7/0478 Special codebook structures directed to feedback optimisation
• CPC Class Codes: H04B 7/0617 for beam forming
• CPC Class Codes: H04B 7/0632 Channel quality parameters, e.g. channel quality indicator [CQI]
• CPC Class Codes: H04B 7/0658 Feedback reduction
• CPC Class Codes: H04B 7/066 Combined feedback for a number of channels, e.g. over several subcarriers like in orthogonal frequency division multiplexing [OFDM]
• CPC Class Codes: H04L 1/0029 Reduction of the amount of signalling, e.g. retention of useful signalling or differential signalling power control H04W52/04
• CPC Class Codes: H04L 1/003 Adaptive formatting arrangements particular to signalling, e.g. variable amount of bits
• CPC Class Codes: H04L 1/0687 Full feedback
• CPC Class Codes: H04L 1/0693 Partial feedback, e.g. partial channel state information [CSI]
• CPC Class Codes: H04L 25/0204 of multiple channels

Abstract

A new vector quantization algorithm for reducing bit rate used for channel state information feedback in a variety of multiple antenna systems on flat and frequency selective channels. The receiver chooses an instantaneous capacity maximizing entry from a small finite set of predefined covariance matrices. The codebook of covariance matrices is constructed based on separate optimization of the channel eigenmodes and power division between them. If, for the given channel realization, one of the predefined covariance matrices provides higher capacity than equal power distribution the indexes of the two codebooks are fed to the transmitter, otherwise, the transmitter uses the open-loop approach. A vector quantization algorithm in which receiver chooses an instantaneous throughput maximizing modulation matrix from a finite set of predefined matrices (code-words). The codebook of modulation matrices is constructed based on joint optimization of the dominant channel eigenmodes of users and separate quantization of power levels.

Claims

1. A method of transmission between a transmitter and a plurality of receivers in a wireless multiple channel system where the transmitter transmits on at least n>

1 antennas, the method comprising the steps of;

each of the plurality of receivers identifying orthogonal modes of the respective channels between the transmitter and the respective receiver by estimating channel coefficients, storing the channel coefficients in a channel matrix and performing a transformation of the channel matrix;

each receiver quantizing the orthogonal modes of the channel by selecting a codebook index of a set of codebook indices of orthogonal mode vectors;

each receiver transmitting the codebook indices to the transmitter; and

upon receiving the codebook indices from the receivers, the transmitter selecting a group of k receivers, where k is selected from 1 up to and including n, and transmitting to the group of k receivers using a stored channel modulation matrix that is selected based upon the codebook indices received by the transmitter, in which

if the transmitter receives a codebook index corresponding to a full orthogonal mode matrix, the transmitter selects a stored channel modulation matrix based upon the received codebook index, selects a corresponding eigenmode distortion matrix and calculates estimated throughput for the receiver, and

if the transmitter receives codebook indexes corresponding to orthogonal mode vectors, the transmitter maps the indices to a composite modulation index, uses the composite modulation index to select a corresponding stored channel modulation matrix, selects a corresponding distortion matrix and calculates estimated throughput for the k receivers.

2. The method of claim 1 in which the stored channel modulation matrix comprises a full matrix of orthogonal modes or a matrix of non-orthogonal modes.

3. The method of claim 2 in which, if a receiver of the one or more receivers receives a signal modulated using the full orthogonal mode matrix, the received signal is demodulated using a complementary full orthogonal mode matrix, and, if a receiver of the one or more receivers receives a signal modulated using the matrix of non-orthogonal modes, the received signal is demodulated using corresponding complementary orthogonal mode vectors.

4. The method of claim 1 in which at least one of condition A and condition B occurs where:

A is the condition that each receiver stores a set of orthogonal mode matrices that have been created by creating a list of possible orthogonal mode matrices, and by selecting a capacity enhancing set of orthogonal mode matrices, for a given quantization resolution, from the possible orthogonal mode matrices; and

B is the condition that each receiver stores a set of orthogonal mode vectors that have been created by creating a list of possible orthogonal mode matrices, separating the orthogonal mode matrices into component orthogonal mode vectors, and by selecting a distortion reducing set of orthogonal mode vectors, for a given quantization resolution, from the component orthogonal mode vectors.

5. The method of claim 1 in which, if the transmitter receives codebook indexes corresponding to orthogonal mode vectors, the choice of composite modulation index is independent of the order of the list of codebook indices.

7. The method of claim 4 in which each receiver has codebook indexes of orthogonal mode matrices and the stored channel modulation matrices at the transmitter are the same as the orthogonal mode matrices.

8. The method of claim 4, in which each receiver has codebook indexes of orthogonal mode vectors, and the stored channel modulation matrices at the transmitter are constructed by:

create a large list of possible orthogonal mode matrices;

separating the orthogonal mode matrices into component orthogonal mode vectors;

for each orthogonal mode vector, storing the codebook index of a matching pre-quantized mode vector from the codebook of orthogonal mode vectors at the receiver;

forming all possible subsets of the stored codebook indices;

within each subset mapping the stored codebook indices as a composite modulation index;

for each subset, calculating the optimum channel modulation matrix based on the corresponding orthogonal mode vectors;

grouping all the calculated optimum channel modulation matrices according to the corresponding composite modulation indices; and

calculating the spherical average of all orthogonal mode vectors within channel modulation matrices in each group and store the spherical averages in a channel modulation matrix that is mapped to the selected composite modulation index.

9. The method of claim 8 in which the choice of composite modulation index is independent of the order of the stored codebook indices.

10. The method of claim 9 in which this independence is obtained by ordering the indices and mapping the ordered set of indices to a composite modulation index.

11. The method of claim 1 in which the transmitter adjusts power applied to the antennas at the transmitter based on received mode gain information from one or more receivers.

12. The method of claim 11 further comprising each of the one or more receivers:

selecting a transformed component of the channel matrix that contains mode gains of the channels, the transformed component being either treated as a matrix or an ensemble of scalar gain values;

quantizing the transformed component by selecting from a stored codebook of gain matrices or gain values; and

forwarding the quantized gain matrix or gain values to the transmitter as mode gain information.

13. The method of claim 12 in which the transmitter adjusts power applied to the antennas at the transmitter based upon improving system throughput performance.

14. The method of claim 13 in which system throughput performance is calculated by the transmitter:

forming subsets of receivers;

within each subset of receivers, mapping the mode gain values as a composite power allocation index;

using the composite power allocation index to select a corresponding power allocation matrix from a set of stored power allocation matrices;

based on the selected power allocation matrix, choose a corresponding average gain matrix; and

based on the average gain matrices and knowledge of transmit power level at the receivers, calculate estimated throughput for the subset of receivers.

15. The method of claim 14 in which the choice of composite power allocation index is independent of the order of the mode gain values.

17. The method of claim 14 in which the stored codebook of gain matrices or gain values is formed by:

if the full mode gain matrix is to be quantized, creating a large list of possible mode gain matrices and selecting a set of mode gain matrices; and

if gain values are to be quantized, creating a large list of possible mode gain matrices, separating the mode gain matrices into their component scalar values and performing probabilistic scalar quantization with a selected resolution to select a set of mode gain values.

18. The method of claim 17 in which stored power allocation matrices at the transmitter are created by:

creating a large list of possible scalar mode gains;

for each possible scalar mode gain, storing the codebook gain index of a matching pre-quantized mode gain value;

forming all possible subsets of the stored codebook gain indexes;

within each subset, mapping the stored codebook gain indexes as a composite power allocation index;

for each subset, calculating the power allocation matrix based on the corresponding mode gains;

grouping all the calculated power allocation matrices according to the corresponding composite power allocation indexes; and

calculating the average of all gain values within matrices in each group and storing the gain values in a power allocation matrix that is mapped to the selected composite power allocation index.

19. The method of claim 18 in which the choice of composite power allocation index is independent of the order of the gain indexes.

21. The method of claim 1 comprising the further step of using a more detailed codebook in addition to the ordinary codebook, with each entry in the original codebook corresponding to multiple entries in the more detailed codebook, such that if the channel is expected to stay sufficiently close to its previous state over the time period between sendings of feedback that the entry from the coarse codebook will not change, the receiver, instead of sending a new index from the coarse codebook, sends the index of an entry of the fine codebook in the subset of the codebook corresponding to the previously sent entry of the coarse codebook.

22. The method of claim 1 in which the system transmits using OFDM, and the method is applied to sets of subcarriers jointly.

23. The method of claim 3 in which, after processing of the received signal, each receiver re-aligns recovered values of the received signal by reversing the signs of the recovered values when the orthogonal modes of the channel have been quantized to opposite mode vectors.

24. The method of claim 4 further comprising the step of aligning the orthogonal mode vectors of the orthogonal mode matrices in a single hemisphere by reversing the sign of any vectors that are not in that hemisphere.

25. The method of claim 8 further comprising the step of aligning the component orthogonal mode vectors in a hemisphere by reversing the sign of any vectors that are not in that hemisphere.

26. The method of claim 17 in which the stored power allocation matrices are the same as the mode gain matrices at the receiver.

27. The method of claim 1 in which the codebook modulation matrices is created by the steps of:

creating plural random channel states with a channel matrix for each simulated user;

selecting one or more eigenmode from each channel matrix;

quantizing the modulation vectors associated with each eigenmode by selecting indices from the codebook of modulation vectors;

mapping each set, unique up to ordering, of indices thus obtained to a unique index of the set;

creating a modulation matrix for each of the channel states; and

for each index of a set, averaging the corresponding modulation matrices to obtain the entry in the codebook corresponding to that index.

28. The method of claim 1 in which the subset of receivers chosen is chosen on the basis of performance.

29. The method of claim 1 in which the set of codebook indices comprises at least a first set of coarse codebook indexes, and a second set of fine codebook indexes, and the method further comprising:

selecting at least a first codebook index from the coarse codebook indexes and transmitting the first codebook index to the transmitter;

the transmitter transmitting to the receiver using a modulation matrix based upon the first codebook index;

the receiver, upon receiving a transmission using a modulation matrix based upon the first codebook index, selecting a second codebook index from the fine codebook indexes and transmitting the second codebook index to the transmitter.

30. The method of claim 29 in which the receiver, upon receiving a transmission using a modulation matrix based upon a coarse codebook index, selecting whether to send a fine codebook index based upon an estimate of the rate of change of the multiple wireless channels.

6. The method of 5 in which the independence is ensured by ordering the indices and assigning a composite modulation index to each such ordered list of indices.

16. The method of 15 in which this independence is obtained by ordering the mode gain values and mapping the ordered set of values to a composite index.

20. The method of 19 in which this independence is obtained by ordering the gain indexes and mapping the ordered set to a composite index.