CN111756478A - Method and device for realizing QR decomposition of matrix with low complexity - Google Patents
Method and device for realizing QR decomposition of matrix with low complexity Download PDFInfo
- Publication number
- CN111756478A CN111756478A CN202010590626.8A CN202010590626A CN111756478A CN 111756478 A CN111756478 A CN 111756478A CN 202010590626 A CN202010590626 A CN 202010590626A CN 111756478 A CN111756478 A CN 111756478A
- Authority
- CN
- China
- Prior art keywords
- data
- row
- max
- matrix
- decomposition
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Withdrawn
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L1/00—Arrangements for detecting or preventing errors in the information received
- H04L1/004—Arrangements for detecting or preventing errors in the information received by using forward error control
- H04L1/0045—Arrangements at the receiver end
- H04L1/0047—Decoding adapted to other signal detection operation
- H04L1/005—Iterative decoding, including iteration between signal detection and decoding operation
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L1/00—Arrangements for detecting or preventing errors in the information received
- H04L1/004—Arrangements for detecting or preventing errors in the information received by using forward error control
- H04L1/0045—Arrangements at the receiver end
- H04L1/0047—Decoding adapted to other signal detection operation
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Radio Transmission System (AREA)
- Variable-Direction Aerials And Aerial Arrays (AREA)
Abstract
The invention provides a method and a device for realizing matrix QR decomposition with low complexity, which comprises the following steps: s1, setting the maximum receiving antenna of the system as m _ max, the maximum layer number as n _ max, and setting m _ max as n _ max; s2, obtaining a complex channel to obtain a matrix H and a complex received signal vector y, wherein H includes m rows and n columns of complex data, m > is n, and y includes m rows and 1 columns of complex signals. The invention adopts a Givens rotating QR decomposition method, utilizes CORDIC to realize, has lower resource consumption and is convenient for hardware realization, uses the intensifying matrix to carry out QR decomposition, and avoids the calculation of an output Q matrix and (Q ^ T) y at the cost of increasing the operation amount of 1 column of data, realizes that a set of QR decomposition scheme is suitable for QR decomposition of any m (n +1) matrix by expanding any m (n +1) (m > ═ n) matrix into m _ max (m _ max +1) matrix, adopts the Givens rotating QR decomposition method, utilizes CORDIC to realize, has lower resource consumption and is convenient for hardware realization, uses the intensifying matrix to carry out QR decomposition, and increases the operation amount of 1 column of data at the cost of increasing the operation amount of 1 column of data, and avoids the calculation of the output Q matrix and (Q ^ T) y.
Description
Technical Field
The invention relates to the technical field of wireless communication, in particular to a method and a device for realizing matrix QR decomposition with low complexity.
Background
In a multi-signal detection system, matrix QR decomposition is a common technique. The QR decomposition can be realized in various ways, and common QR decomposition methods include Householder QR decomposition, Gram-Schmidt QR decomposition and QR decomposition based on Givens rotation. The prior art is based on three methods. But has the following disadvantages:
1. the Householder QR decomposition method and the Gram-Schmidt QR decomposition method have requirements on the number of multiplication dividers and dividers, have large bit width and consume resources comparatively.
2. The QR decomposition of Givens rotation can be realized by CORDIC (coordinate rotation digital computing method), and can be realized by only addition, shift and multiplication, thereby facilitating the hardware realization. However, in order to ensure performance, the number of iterations is large, a fixed number of iterations is used for all data, and power consumption is large.
3. Due to the fact that multiple combinations exist in the receiving antennas and the number m x n of layers, multiple sets of QR decomposition are needed to be achieved, and resources are wasted.
4. For the equation y ═ Hx, the QR decomposition of H is firstly carried out, then the equation is solved, the same operation needs to be carried out on a unit array I while the QR decomposition is carried out on H so as to output a Q matrix, and (Q ^ T) y needs to be calculated, which causes the waste of power consumption and resources.
5. For the application scenario of solving the equation, the output R matrix generally needs to be solved again, division is needed during solving, and resource overhead of division is large, so that the QR result is inconvenient to use.
Disclosure of Invention
The invention aims to provide a method and a device for realizing matrix QR decomposition with low complexity, which can effectively solve the problems in the background technology.
In order to achieve the purpose, the invention is realized by the following technical scheme: a method for realizing matrix QR decomposition with low complexity comprises the following steps:
and S1, setting the maximum receiving antenna of the system as m _ max, and setting the maximum layer number as n _ max, wherein m _ max > is equal to n _ max.
S2, obtaining the complex channel to obtain a matrix H and a complex received signal vector y, wherein H includes m rows and n columns of complex data, where m > is n, y includes m rows and 1 column of complex signals, and constitutes an amplification matrix H1 ═ H y.
S3, and a pair matrix H2, starting from the first row, passes through the first CORDIC iteration unit, each CORDIC iteration unit having the same structure.
S4, the data of the second row and the data of the second column of the second row are processed by the second CORDIC iteration unit, starting with the second row, for each row outputted in S3.
Further, according to the operation procedure in S2, the augmented matrix H1 includes m rows and n +1 columns, the complex matrix H1 is augmented into m _ max rows and m _ max +1 columns of the complex matrix H2, and the complex matrix H2 is augmented by filling 0
Further, in the initialization m _ max row, the m _ max +1 column matrix RR is all zero, and the number of CORDIC iterations is 16.
Further, according to the operation step in S3, the method further includes the steps of:
s301, one line of data a enters an iteration unit, and the whole line of data a is rotated by using a CORDIC, so that the first data a is converted into real data from complex data.
S302, carrying out CORDIC algorithm rotation on the row data a updated in the S301 and the corresponding row position data in the RR to enable the first row data a in the S301 to be 0, updating current data output, and updating the corresponding row position data in the RR.
Further, according to the operation step in S4:
s401, the m _ max step: and (3) each row output in the step (m _ max-1) starts from the second row, and data starting from the second column of each row passes through the (m _ max-1) CORDIC iteration list.
S402, the m _ max +1 step: and (4) each row output in the step (m _ max) passes the data of the second column of each row through the (m _ max) CORDIC iteration units from the second row.
S403, the m _ max +2 step: and after the data in the m _ max row completely go through the m _ max +1 step, judging whether the first data a output by the CORDIC iteration unit is real (a), and then judging whether the first data a is updated to 1/a according to div _ en, if the a is 0, the first data a is unchanged, and after updating, respectively outputting non-zero elements in the ith row of the upper triangular matrix RR from the ith CORDIC iteration unit.
S404, the m _ max +3 step: the first n +1 columns of elements of the first m rows are output from the RR matrix to form an upper triangular matrix R output.
A low complexity implementation matrix QR decomposition apparatus, comprising: the detection system comprises a decomposition unit, a first unit, a second unit, two registers and a reciprocal module.
Further, the decomposition unit is configured to perform QR decomposition on a channel matrix of the received signal; the first unit is used for making the same rotation on the whole row of data of the CORDIC, and the rotation enables the first data of the row to become real numbers; the second unit is used for the CORDIC2 to rotate the first column data of the K-th row and the first column data of the first row, so that the data of the K-th row and the first column become 0, and the other data of the K-th row and the first row are rotated in the same way; the two registers can be matched with parameters to monitor data in each iteration process of the CORDIC, and when the data meet the conditions, the iteration is ended in advance, so that excessive iteration times and power consumption waste are avoided; the reciprocal module: the diagonal elements of the flexibly configurable output are R11 or 1/R11.
The invention provides a method and a device for realizing matrix QR decomposition with low complexity. The method has the following beneficial effects:
(1) the invention comprises the following steps: the QR decomposition method adopting Givens rotation is realized by utilizing CORDIC, the resource consumption is low, the hardware realization is convenient, the intensifying matrix is used for QR decomposition, the cost of 1 column of data operation amount is increased, and the calculation of an output Q matrix and (Q ^ T) y is avoided.
(2) The invention comprises the following steps: a set of QR decomposition schemes is applied to QR decomposition of any m (n +1) matrix by expanding any m (n +1) (m > ═ n) matrix into m _ max (m _ max +1) matrix.
(3) The invention comprises the following steps: the data in each iteration process of the CORDIC is monitored through the two registers which can be matched with parameters, and when the conditions are met, the iteration is ended in advance, so that excessive iteration times and power consumption waste are avoided, and the iteration times can be effectively reduced for a large number of specific numbers. Due to the fact that the register can be configured, flexible selection of performance and power consumption is facilitated.
(4) The invention comprises the following steps: the system is provided with a configurable high-precision low-complexity reciprocal solving module, so that the diagonal elements capable of being flexibly configured and output are R11 or 1/R11, when an equation is solved after inversion, only one multiplication is needed, if the inversion is not solved, division is needed, and an output mode is configured and determined by a user.
(5) The invention comprises the following steps: q' Q is obtained by QR decomposition of the matrix A, B is QR, the error between B and A is about one thousandth, and the error between D and the unit matrix is about one thousandth, so that the performance requirement is met.
Description of the drawings:
FIG. 1 is a general flow chart of the present invention;
FIG. 2 is a QR decomposition overall flow diagram of the present invention;
FIG. 3 is a flow chart of a CORDIC unit according to the present invention;
fig. 4 is a block diagram of the system of the present invention.
Detailed Description
The invention is illustrated below with reference to specific examples. It will be understood by those skilled in the art that these examples are for illustrative purposes only and are not intended to limit the scope of the present invention in any way.
Example 1: referring to FIGS. 1-4: a method for realizing matrix QR decomposition with low complexity comprises the following steps:
the maximum receiving antenna of the system is set as m _ max, the maximum layer number is set as n _ max, and m _ max > is set as n _ max, and the implementation steps are as follows:
the method comprises the following steps: a complex channel estimation matrix H and a complex received signal vector y are obtained, H comprising m rows and n columns of complex data, where m > is n and y comprises m rows and 1 column of complex signals, forming an amplification matrix H1 ═ H y, H1 comprising m rows and n +1 columns. The complex matrix H1 is expanded into m _ max row m _ max +1 column complex matrix H2, which is expanded by filling 0. Initializing m _ max row m _ max +1 column matrix RR as all zeros, CORDIC iteration number is 16, die _ para, diebabspeara, div _ en.
Step two: for matrix H2, the first CORDIC iteration block is traversed starting from the first row, each CORDIC iteration block having the same structure.
A row of data a enters an iteration unit, and the whole row of data is rotated by using a CORDIC, so that the first data a is converted into real data from complex data. In the iteration process, when the iteration number is greater than die _ para and the updated abs (imag (a (1))) < dieebspara is iterated, the current iteration is ended.
The updated row of data a, the position data corresponding to the corresponding row in RR are rotated (CORDIC algorithm) so that the first data a of the current data becomes 0. And updating the current data output and updating the corresponding position data of the corresponding row in the RR. In the iteration process, when the iteration number is greater than die _ para and the updated abs (real (a) < diebscara) is iterated, the current iteration is ended.
And step three, each row output in the step two starts from the second row, and data starting from the second column of each row passes through a second CORDIC iteration unit.
…
The m _ max step: and (3) each row output in the step (m _ max-1) starts from the second row, and data starting from the second column of each row passes through the (m _ max-1) th CORDIC iteration unit.
The m _ max +1 step: and (4) each row output in the step (m _ max) passes the data of the second column of each row through the (m _ max) CORDIC iteration units from the second row.
The m _ max +2 step: after the steps of 2 to m _ max +1 are completed for all the data in the m _ max row, the first data a (1) output by the CORDIC iteration unit is equal to real (a (1)), and whether the data is updated to 1/a (1) is determined according to div _ en, and if the a (1) is 0, the data is not changed. After updating, the ith row nonzero elements of the upper triangular matrix RR are respectively output from the ith CORDIC iteration unit.
The m _ max +3 step: the first n +1 columns of elements of the first m rows are output from the RR matrix to form an upper triangular matrix R output.
A low complexity implementation matrix QR decomposition apparatus, comprising: the detection system comprises a decomposition unit, a first unit, a second unit, two registers and a reciprocal module, wherein the decomposition unit is used for carrying out QR decomposition on a channel matrix of a received signal; the first unit is used for making the same rotation on the whole line of data of the CORDIC, and the rotation enables the first data of the line to become real numbers; the second unit is used for the CORDIC2 to rotate the first column data and the first row data of the K-th row, so that the data of the K-th row and the first column become 0, and the other data of the K-th row and the first row are rotated in the same way; the two registers can be matched with parameters to monitor data in each iteration process of the CORDIC, and when the data meet the conditions, the iteration is ended in advance, so that excessive iteration times and power consumption waste are avoided; a reciprocal module: the diagonal elements of the flexibly configurable output are R11 or 1/R11.
For a certain LTE receiver, the maximum receiving antenna is set to be 4, the maximum number of receiving layers is 4, when the actual receiving antenna is 4 and the number of receiving layers is 2, div _ en is configured to be 0, and the received signal is set to be
The corresponding channel estimate is
Then the value of y-H x + n,
n is a 2 x 1 matrix, and an ML solution of x is required to be solved.
Inputting H and y into QR decomposition in the scheme, and outputting to obtain an upper triangular matrix of 4 x 3
It can be known that the ML solution of y ═ H × x + n is equivalent to the equation
The solution of (1). Is reversely pushed to x1＝R23/R22,x2＝(R13-R12*x2)/R11。
Example 2: referring to FIGS. 1-4: a method for realizing matrix QR decomposition with low complexity comprises the following steps:
the maximum receiving antenna of the system is set as m _ max, the maximum layer number is set as n _ max, and m _ max > is set as n _ max, and the implementation steps are as follows:
the method comprises the following steps: a complex channel estimation matrix H and a complex received signal vector y are obtained, H comprising m rows and n columns of complex data, where m > is n and y comprises m rows and 1 column of complex signals, forming an amplification matrix H1 ═ H y, H1 comprising m rows and n +1 columns. The complex matrix H1 is expanded into m _ max row m _ max +1 column complex matrix H2, which is expanded by filling 0. Initializing m _ max row m _ max +1 column matrix RR as all zeros, CORDIC iteration number is 16, die _ para, diebabspeara, div _ en.
Step two: for matrix H2, the first CORDIC iteration block is traversed starting from the first row, each CORDIC iteration block having the same structure.
A row of data a enters an iteration unit, and the whole row of data is rotated by using a CORDIC, so that the first data a is converted into real data from complex data. In the iteration process, when the iteration number is greater than die _ para and the updated abs (imag (a (1))) < dieebspara is iterated, the current iteration is ended.
The updated row of data a, the position data corresponding to the corresponding row in RR are rotated (CORDIC algorithm) so that the first data a of the current data becomes 0. And updating the current data output and updating the corresponding position data of the corresponding row in the RR. In the iteration process, when the iteration number is greater than die _ para and the updated abs (real (a) < diebscara) is iterated, the current iteration is ended.
And step three, each row output in the step two starts from the second row, and data starting from the second column of each row passes through a second CORDIC iteration unit.
…
The m _ max step: and (3) each row output in the step (m _ max-1) starts from the second row, and data starting from the second column of each row passes through the (m _ max-1) th CORDIC iteration unit.
The m _ max +1 step: and (4) each row output in the step (m _ max) passes the data of the second column of each row through the (m _ max) CORDIC iteration units from the second row.
The m _ max +2 step: after the steps of 2 to m _ max +1 are completed for all the data in the m _ max row, the first data a (1) output by the CORDIC iteration unit is equal to real (a (1)), and whether the data is updated to 1/a (1) is determined according to div _ en, and if the a (1) is 0, the data is not changed. After updating, the ith row nonzero elements of the upper triangular matrix RR are respectively output from the ith CORDIC iteration unit.
The m _ max +3 step: the first n +1 columns of elements of the first m rows are output from the RR matrix to form an upper triangular matrix R output.
A low complexity implementation matrix QR decomposition apparatus, comprising: the detection system comprises a decomposition unit, a first unit, a second unit, two registers and a reciprocal module, wherein the decomposition unit is used for carrying out QR decomposition on a channel matrix of a received signal; the first unit is used for making the same rotation on the whole line of data of the CORDIC, and the rotation enables the first data of the line to become real numbers; the second unit is used for the CORDIC2 to rotate the first column data and the first row data of the K-th row, so that the data of the K-th row and the first column become 0, and the other data of the K-th row and the first row are rotated in the same way; the two registers can be matched with parameters to monitor data in each iteration process of the CORDIC, and when the data meet the conditions, the iteration is ended in advance, so that excessive iteration times and power consumption waste are avoided; a reciprocal module: the diagonal elements of the flexibly configurable output are R11 or 1/R11.
For a multi-signal detection system, the maximum receiving antenna is set as m _ max, and the maximum receiving layer number is set as n _ max (m _ max)>N _ max), when the actual receiving antenna is m, the number of receiving layers is n (m)>N), the configuration div _ en is 0, and the received signal is set to
The corresponding channel estimate is
Then the value of y-H x + n,
n is an n x 1 matrix, and an ML solution of x is required to be solved.
Inputting H and y into QR decomposition in the scheme, and outputting to obtain an upper triangular matrix of m x (n +1)
It can be known that the ML solution of y ═ H × x + n is equivalent to the equation
The solution of (1). Is reversely pushed to
xn＝Rn(n+1)/Rnn，
Example 3: referring to FIGS. 1-4: a method for realizing matrix QR decomposition with low complexity comprises the following steps:
the maximum receiving antenna of the system is set as m _ max, the maximum layer number is set as n _ max, and m _ max > is set as n _ max, and the implementation steps are as follows:
the method comprises the following steps: a complex channel estimation matrix H and a complex received signal vector y are obtained, H comprising m rows and n columns of complex data, where m > is n and y comprises m rows and 1 column of complex signals, forming an amplification matrix H1 ═ H y, H1 comprising m rows and n +1 columns. The complex matrix H1 is expanded into m _ max row m _ max +1 column complex matrix H2, which is expanded by filling 0. Initializing m _ max row m _ max +1 column matrix RR as all zeros, CORDIC iteration number is 16, die _ para, diebabspeara, div _ en.
Step two: for matrix H2, the first CORDIC iteration block is traversed starting from the first row, each CORDIC iteration block having the same structure.
A row of data a enters an iteration unit, and the whole row of data is rotated by using a CORDIC, so that the first data a is converted into real data from complex data. In the iteration process, when the iteration number is greater than die _ para and the updated abs (imag (a (1))) < dieebspara is iterated, the current iteration is ended.
The updated row of data a, the position data corresponding to the corresponding row in RR are rotated (CORDIC algorithm) so that the first data a of the current data becomes 0. And updating the current data output and updating the corresponding position data of the corresponding row in the RR. In the iteration process, when the iteration number is greater than die _ para and the updated abs (real (a) < diebscara) is iterated, the current iteration is ended.
And step three, each row output in the step two starts from the second row, and data starting from the second column of each row passes through a second CORDIC iteration unit.
…
The m _ max step: and (3) each row output in the step (m _ max-1) starts from the second row, and data starting from the second column of each row passes through the (m _ max-1) th CORDIC iteration unit.
The m _ max +1 step: and (4) each row output in the step (m _ max) passes the data of the second column of each row through the (m _ max) CORDIC iteration units from the second row.
The m _ max +2 step: after the steps of 2 to m _ max +1 are completed for all the data in the m _ max row, the first data a (1) output by the CORDIC iteration unit is equal to real (a (1)), and whether the data is updated to 1/a (1) is determined according to div _ en, and if the a (1) is 0, the data is not changed. After updating, the ith row nonzero elements of the upper triangular matrix RR are respectively output from the ith CORDIC iteration unit.
The m _ max +3 step: the first n +1 columns of elements of the first m rows are output from the RR matrix to form an upper triangular matrix R output.
A low complexity implementation matrix QR decomposition apparatus, comprising: the detection system comprises a decomposition unit, a first unit, a second unit, two registers and a reciprocal module, wherein the decomposition unit is used for carrying out QR decomposition on a channel matrix of a received signal; the first unit is used for making the same rotation on the whole line of data of the CORDIC, and the rotation enables the first data of the line to become real numbers; the second unit is used for the CORDIC2 to rotate the first column data and the first row data of the K-th row, so that the data of the K-th row and the first column become 0, and the other data of the K-th row and the first row are rotated in the same way; the two registers can be matched with parameters to monitor data in each iteration process of the CORDIC, and when the data meet the conditions, the iteration is ended in advance, so that excessive iteration times and power consumption waste are avoided; a reciprocal module: the diagonal elements of the flexibly configurable output are R11 or 1/R11.
Configuring diven as 1, setting the maximum receiving antenna as m _ max and the maximum receiving layer number as n _ max (m _ max) for a certain multi-signal detection system>N _ max), when the actual receiving antenna is m, the number of receiving layers is n (m)>N), the received signal is set to
The corresponding channel estimate is
Then the value of y-H x + n,
n is an n x 1 matrix, and an ML solution of x is required to be solved.
Inputting H and y into QR decomposition in the scheme, and outputting to obtain an upper triangular matrix of m x (n +1)
It can be known that the ML solution of y ═ H × x + n is equivalent to the equation
The solution of (1). Is reversely pushed to xn＝Rn(n+1)*Rnn，,
The foregoing is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, various changes and modifications can be made without departing from the inventive concept of the present invention, and these changes and modifications are all within the scope of the present invention.
Claims (7)
1. A method for realizing matrix QR decomposition with low complexity is characterized by comprising the following steps:
s1, setting the maximum receiving antenna of the system as m _ max, the maximum layer number as n _ max, and setting m _ max as n _ max;
s2, obtaining a complex channel to obtain a matrix H and a complex received signal vector y, where H includes m rows and n columns of complex data, where m > is n, y includes m rows and 1 column of complex signals, and an amplification matrix H1 is [ H y ];
s3, making the matrix H2 pass through a first CORDIC iteration unit from the first row, wherein each CORDIC iteration unit has the same structure;
s4, the data of the second row and the data of the second column of the second row are processed by the second CORDIC iteration unit, starting with the second row, for each row outputted in S3.
2. The method of claim 1, wherein the augmented matrix H1 comprises m rows and n +1 columns, and the augmented matrix H1 is augmented as m _ max rows and m _ max +1 columns of the complex matrix H2 according to the operation procedure in S2, and the augmented matrix H2 is augmented by filling 0.
3. The method of claim 2, wherein the matrix RR is all zeros at m _ max row and m _ max +1 column of initialization m _ max, and the number of CORDIC iterations is 16.
4. The method for implementing QR decomposition of matrices with low complexity according to claim 1, further comprising the following steps according to the operation steps in S3:
s301, enabling a row of data a to enter an iteration unit, and rotating the whole row of data by using a CORDIC (coordinate rotation digital computer) to enable the first data a to be converted into real data from complex data;
s302, carrying out CORDIC algorithm rotation on the row data a updated in the S301 and the corresponding row position data in the RR to enable the first row data a in the S301 to be 0, updating current data output, and updating the corresponding row position data in the RR.
5. The method for implementing QR decomposition of matrices according to claim 1, wherein according to the operating steps in S4:
s401, the m _ max step: starting from the second row, each row output in the step m _ max-1 passes the data starting from the second column of each row through the m _ max-1 CORDIC iteration list;
s402, the m _ max +1 step: each row output in the m _ max step starts from the second row, and data starting from the second column of each row passes through the m _ max CORDIC iteration units;
s403, the m _ max +2 step: after the data in the m _ max row completely go through the m _ max +1 step, the first data a output by the CORDIC iteration unit is real (a), whether the data is updated to 1/a is judged according to div _ en, if the a is 0, the data is unchanged, and after the updating, the non-zero elements in the ith row of the upper triangular matrix RR are output from the ith CORDIC iteration unit respectively;
s404, the m _ max +3 step: the first n +1 columns of elements of the first m rows are output from the RR matrix to form an upper triangular matrix R output.
6. A low complexity realization matrix QR decomposition apparatus, comprising: the detection system comprises a decomposition unit, a first unit, a second unit, two registers and a reciprocal module.
7. The apparatus of claim 6, wherein the low complexity realization matrix QR decomposition device is characterized in that: the decomposition unit is used for carrying out QR decomposition on the channel matrix of the received signal; the first unit is used for making the same rotation on the whole row of data of the CORDIC, and the rotation enables the first data of the row to become real numbers; the second unit is used for the CORDIC2 to rotate the first column data of the K-th row and the first column data of the first row, so that the data of the K-th row and the first column become 0, and the other data of the K-th row and the first row are rotated in the same way; the two registers can be matched with parameters to monitor data in each iteration process of the CORDIC, and when the data meet the conditions, the iteration is ended in advance, so that excessive iteration times and power consumption waste are avoided; the reciprocal module: the diagonal elements of the flexibly configurable output are R11 or 1/R11.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010590626.8A CN111756478A (en) | 2020-06-24 | 2020-06-24 | Method and device for realizing QR decomposition of matrix with low complexity |
| CN202011000444.7A CN111901071B (en) | 2020-06-24 | 2020-09-22 | Method and device for realizing QR decomposition of matrix with low complexity |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010590626.8A CN111756478A (en) | 2020-06-24 | 2020-06-24 | Method and device for realizing QR decomposition of matrix with low complexity |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111756478A true CN111756478A (en) | 2020-10-09 |
Family
ID=72677122
Family Applications (2)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010590626.8A Withdrawn CN111756478A (en) | 2020-06-24 | 2020-06-24 | Method and device for realizing QR decomposition of matrix with low complexity |
| CN202011000444.7A Active CN111901071B (en) | 2020-06-24 | 2020-09-22 | Method and device for realizing QR decomposition of matrix with low complexity |
Family Applications After (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011000444.7A Active CN111901071B (en) | 2020-06-24 | 2020-09-22 | Method and device for realizing QR decomposition of matrix with low complexity |
Country Status (1)
| Country | Link |
|---|---|
| CN (2) | CN111756478A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113141233A (en) * | 2021-03-11 | 2021-07-20 | 西安电子科技大学 | Channel matrix sequencing QR decomposition processing method and system |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8223872B1 (en) * | 2007-04-04 | 2012-07-17 | Marvell International Ltd. | Reuse of a matrix equalizer for the purpose of transmit beamforming in a wireless MIMO communication system |
| US9318813B2 (en) * | 2009-05-22 | 2016-04-19 | Maxlinear, Inc. | Signal processing block for a receiver in wireless communication |
| CN102111350A (en) * | 2009-12-25 | 2011-06-29 | 中国电子科技集团公司第五十研究所 | FPGA device for matrix QR decomposition |
| JP5787527B2 (en) * | 2011-01-18 | 2015-09-30 | キヤノン株式会社 | Signal processing circuit and ultrasonic diagnostic apparatus |
| CN102624653B (en) * | 2012-01-13 | 2014-08-20 | 清华大学 | Extensible QR decomposition method based on pipeline working mode |
| CN103293519B (en) * | 2013-05-10 | 2015-07-15 | 东南大学 | Method and system for error correction of channels I/Q based on pipeline working mode |
| CN107203491A (en) * | 2017-05-19 | 2017-09-26 | 电子科技大学 | A kind of triangle systolic array architecture QR decomposers for FPGA |
-
2020
- 2020-06-24 CN CN202010590626.8A patent/CN111756478A/en not_active Withdrawn
- 2020-09-22 CN CN202011000444.7A patent/CN111901071B/en active Active
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113141233A (en) * | 2021-03-11 | 2021-07-20 | 西安电子科技大学 | Channel matrix sequencing QR decomposition processing method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111901071B (en) | 2021-04-06 |
| CN111901071A (en) | 2020-11-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| RU2404513C2 (en) | Efficient filter weight computation for mimo system | |
| CN110855585B (en) | Channel estimation method of millimeter wave large-scale MIMO system | |
| WO2010135745A1 (en) | Signal processing block for a receiver in wireless communication | |
| US20080155002A1 (en) | Combined fast fourier transforms and matrix operations | |
| CN103516643A (en) | MIMO detecting preprocessing device and method | |
| US11018729B2 (en) | Structured-pipelined CORDIC for matrix equalization | |
| CN111901071B (en) | Method and device for realizing QR decomposition of matrix with low complexity | |
| Patel et al. | A low-complexity high-speed QR decomposition implementation for MIMO receivers | |
| Gao et al. | Low-complexity block coordinate descend based multiuser detection for uplink grant-free NOMA | |
| CN105703813B (en) | A kind of method for precoding of mimo system | |
| Kurniawan et al. | Multidimensional Householder based high-speed QR decomposition architecture for MIMO receivers | |
| Berthe et al. | Low-complexity soft-output signal detector based on adaptive pre-conditioned gradient descent method for uplink multiuser massive MIMO systems | |
| CN106664267B (en) | Method and apparatus for decoding | |
| US20220374500A1 (en) | Finite resolution decomposition of a matrix and matrix-vector multiplication | |
| US7974370B2 (en) | Parallel processing for single antenna interference cancellation | |
| Castaneda et al. | FPGA design of approximate semidefinite relaxation for data detection in large MIMO wireless systems | |
| Lin et al. | Accelerating Next-G Wireless Communications with FPGA-Based AI Accelerators | |
| CN107229596B (en) | Non-pipeline fast Fourier transform processor and operation control method thereof | |
| US20190171613A1 (en) | Method and Apparatus for A Vector Memory Subsystem for Use with A Programmable Mixed-Radix DFT/IDFT Processor | |
| US11973562B2 (en) | Internal data transfer in a multiple-antenna communication system | |
| Fan et al. | Approximate message passing detector based upon probability sorting for large-scale GSM systems | |
| Gizzini et al. | Deep Neural Network Augmented Wireless Channel Estimation for Preamble-based OFDM PHY on Zynq System on Chip | |
| Bian et al. | Joint activity-delay detection and channel estimation for asynchronous massive random access | |
| CN112953607B (en) | Method, medium and equipment for eliminating quantization noise of MIMO-OFDM system | |
| Yang et al. | Design of low-complexity FFT processor for MIMO-OFDM based SDR systems |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20201009 |
|
| WW01 | Invention patent application withdrawn after publication |