CN105847194A - QRD structure based on MGS - Google Patents

Abstract

A QRD structure based on MGS comprises a first module CORE-GEQRT and a second module CORE-TTQRT. The first module CORE-GEQRT is used for performing QR decomposition on an input 2*2 matrix [a1,a2) for outputting q1, r11, r22, sign(f(A)), r12 and q2; wherein [q1,q2] is a Q matrix; r11, r12 and r22 form an upper triangular matrix R, sign(f(A))=(a21a12-a11a22)/r11. The two obtained R matrixes are used as the input of the second module CORE-TTQRT and [1,0;01] and [0,0;0,0], thereby forming a matrix [a11,a12,1,0;0,a22,0,1;a31,a32,0,0;0,a42,0,0], and finally obtaining a 4*2 matrix. The QRD structure based on the MGS has advantages of simple structure, integral performance improvement, etc.

Description

A kind of QRD structure based on MGS

Technical field

Present invention relates generally to extensive multi-antenna technology field, refer in particular to a kind of based on the Schmitt orthogonalization improved Novel 2 × 2 tile QR of (Modified Gram Schmidt, MGS) algorithm decompose (QR Decomposition, QRD) knot Structure.

Background technology

Extensive multi-antenna technology (multiple input multiple output, MIMO) is the pass of next generation communication technology (5G) One of key technology.In extensive mimo system, base station end, equipped with up to a hundred antennas, can service tens users (letter simultaneously For the sake of list, each user only has 1 antenna).This with traditional mimo system in 4 base station antennas, service 4 The situation of single-antenna subscriber (being called for short 4 × 4MIMO system) is compared, and more antenna provides more spatial multiplexing gain and divides Diversity gain.And, in extensive mimo system, simple linear signal processing process just can reach near-optimization performance.

In extensive mimo system, by accommodating more with the bigger degree of freedom of offer at the substantial amounts of antenna of base station equipment Information.Therefore, extensive MIMO compares traditional mimo system and can preferably improve the availability of frequency spectrum, channel capacity and company Connect reliability.Compared with traditional mimo system, several times of the matrix dimension of the base band signal process algorithm of extensive MIMO is very To tens times.Especially relate to matrix inversion operation and QR decomposition etc. when algorithm.In order to overcome extensive matrix complex degree bottleneck, Need to design a kind of new structure and accelerate the matrix algorithm unit of key.

QRD obtains in mimo systems and uses widely, and works known to some, and QRD becomes at transmitting terminal One requisite assembly.In general, QRD is used to decompose channel response matrix H and becomes unitary matrice Q and Individual upper triangular matrix R.In extensive MIMO, the number of user and the antenna number of base station change in a scope the biggest. The dimension of H changes in a scope the biggest, and it needs QRD hardware configuration to decompose the matrix of this multidimensional.Just at present and Speech, the most already present QRD hardware configuration is concentrated mainly on the matrix of one or several fixing dimension.Therefore, QRD hardware configuration has the most great meaning to wireless communication system in future flexibly.

As depicted in figs. 1 and 2, traditional MGS hardware algorithm structure is broadly divided into two modules of DP, TP.Wherein DP mould Block is mainly used to produce q_jAnd r_jj, TP module is mainly used to produce the matrix updatedAnd r_ji, DMP is mainly used to produce q_j, r_jj,r_ji。

To the rectangular array vector a inputted in DP module_j, first, each of which element is separately input to squaring module M1, puts down In side module M2, squaring module M3, squaring module M4, the result obtained is stored in Buffer1.Then, the result that will obtain Being separately input in adder M1, adder M2, the result obtained is separately input in adder M3, and adder M3 obtains To result be stored in Buffer2.The result obtained in Buffer2, is input to open in more number module M1 and obtains result r_jj。Buffer3 In the result that obtains as divisor, a_jIt is separately input to divider M1, divider M2, division as dividend each of which element Device M3, divider M4 obtain result q_jIt is stored in Buffer4.

To the rectangular array vector a inputted in TP module_iAnd q_j, wherein i is incremented by (i≤3) and each q successively_jCorresponding a₀、a₁、 a₂、a₃, a_iSuccessively as input, with q_jIt is separately input to multiplier M1, multiplier M2, multiplier M3, multiplier M4, the result obtained is stored in Buffer1.From Buffer1, read result be deposited into adder M4, adder M5 respectively The result obtained is stored in and is input to adder M6, and the result obtained is stored in Buffer2.The result obtained from Buffer2 and Column vector q of input_jIn each element be separately input to multiplier M5, multiplier M6, multiplier M7, multiplier M8, The result obtained is deposited in Buffer3.Using the result that obtains in Buffer3 as subtrahend a_iIt is separately input to subtract as minuend Musical instruments used in a Buddhist or Taoist mass M1, subtractor M2, subtractor M3, subtractor M4, obtain result

In above-mentioned traditional structure, extensive MIMO matrix dimension is relatively big, causes relating to significantly increasing during QRD algorithm The computational complexity of base station,

Summary of the invention

The technical problem to be solved in the present invention is that the technical problem existed for prior art, and the present invention provides a kind of structure Simply, the QRD structure based on MGS of overall performance can be improved.

For solve above-mentioned technical problem, the present invention by the following technical solutions:

A kind of QRD structure based on MGS, including the first module CORE_GEQRT and the second module CORE_TTQRT, described First module CORE_GEQRT is used for 2 × 2 matrix [a to input₁,a₂] carry out QR decomposition output by MGS algorithm q₁,r₁₁,r₂₂,sign(f(A)),r₁₂,q₂, wherein [q₁,q₂] it is Q matrix, r₁₁,r₁₂,r₂₂Composition upper triangular matrix R, Sign (f (A))=(a₂₁a₁₂-a₁₁a₂₂)r₁₁；Using two R matrixes obtaining as the second module CORE_TTQRT input with [1,0；01]、[0,0；0,0] matrix [a is collectively constituted₁₁,a₁₂,1,0；0,a₂₂,0,1；a₃₁,a₃₂,0,0；0,a₄₂, 0,0], finally give the matrix of 4 × 2.

As a further improvement on the present invention: described first module CORE_GEQRT includes:

First 2 × 2 matrixes of input are stored in buffer1, the matrix-vector a being stored into by selector₁Element a₁₁As taking advantage of Two inputs, a of musical instruments used in a Buddhist or Taoist mass 1₂₁Two inputs as multiplier 2.Two outputs that will obtainWithAs adder 1 Input, adder 1 obtains resultIt is stored in buffer2；Result adder 1 obtained reads input from buffer2 Carry out out radical sign module 1, the result that radical sign computing obtains will be openedIt is stored in buffer4.By vector a₁Make the dividend of divider, r₁₁It is separately input in divider 1 and divider 2 as divisor, result q obtained₁It is stored in buffer5；

After vector is input to buffer1, read vector a from buffer1₁While read vector a₂, first by a₁₁Become -a₁₁, then by data-a₁₁, a₁₂, a₂₁, a₂₂It is stored in buffer2；WhenWhen being input to out radical sign module 1, a₁₂, a₂₁ Input, a as multiplier 1₂₂,-a₁₁As the input of multiplier 2, export two obtained as adder 1 is defeated Entering, the output f (A) obtained the most at last is stored in buffer3；F (A) is read from buffer3 the computing that carries out taking absolute value, The value obtained is input in buffer4；Meanwhile, the symbol value as output Sign of f (A) is taken, by | f (A) | with r₁₁From buffer4 Middle reading inputs in dividing module 3, by operation result r₂₂It is stored in buffer5；

After f (A) has calculated, by element a₁₁, a₁₂It is input to multiplier 1, by element a₂₁, a₂₂Input multiplier 2, will The output obtained is as the input of adder 1, and the result that adder 1 obtains is stored in buffer4；The result that adder 1 is obtained Take out from buffer4 and r₁₁Jointly it is linked into the input of dividing module 4, obtains result r₁₂It is stored in buffer5；

When obtaining q₁After Sign (f (A)), being entered into and be input in selector 2, wherein Sign (f (A)) believes as condition Number.When Sign (f (A)) is timing q₂₁=q₁₂, q₂₂=-q₁₁, when Sign (f (A)) is q time negative₂₁=-q₁₂, q₂₂=q₁₁.? The q arrived₂As output.

As a further improvement on the present invention: described second module CORE_TTQRT includes: include three module QR and the 4th Module Column update；Described three module QR is fully pipelined architecture, first clock cycle first to its input matrix A_1=[a₁₁,1；a₃₁, 0] and carry out QR operation, second clock cycle to its input matrix A_2=[a₂₂,1；a₄₂, 0] and carry out QR Operation；First, 2 × 2 matrixes of input are stored in buffer1, the matrix-vector a being stored into by selector₁Element a₁₁Make Two inputs, a for multiplier 3₂₁As two inputs of multiplier 4, meanwhile, take a₂₁Symbol as output Sign；Will Two outputs obtainedWithAs the input of adder 2, adder 1 obtains resultIt is stored in buffer2；By addition The result input that device 2 obtains carries out out radical sign module 2, will open the result that radical sign computing obtainsMake For output；By vector a₁Make the dividend of divider 5, r₁₁It is input to divider 6, result q obtained as divisor₁As defeated Go out.Wherein r₁₂=q₁₁, r₂₂=q₁₂；

Obtain Sign and q₁After, by Sign and q₁As the input of selector, wherein Sign is as selecting signal.When Sign is just Time be output as q₂₁=q₁₂, q₂₂=-q₁₁, when Sign is for being output as q time negative₂₁=-q₁₂, q₂₂=q₁₁；The q obtained₂As output；

After obtaining the Q matrix of A_1, to a₁₂, a₃₂Carry out row by Column update and update operation；By a₁₁, q₁₂As The input a of multiplier 5₂₁, q₂₂Input multiplier 6, by a₁₁, q₁₁Input multiplier 7, by a₂₁, q₂₁Input multiplier 8； The result input summer 3 that multiplier 5 and multiplier 6 obtain obtains resultThe result that multiplier 7 and multiplier 8 obtain Input summer 2 obtains resultObtain matrixTake matrixMiddle submatrixCarry out and A_1 After same operation, to vectorIt is updated operation, obtains the R matrix of 2 × 2.

Compared with prior art, it is an advantage of the current invention that:

The QRD structure based on MGS of the present invention, simple in construction, can improve overall performance, solves extensive MIMO square Battle array dimension is relatively big, causes this problem of computational complexity relating to significantly increasing base station during QRD algorithm.Due to QRD watt Sheet algorithm is very suitable for future broadband wireless communication systems, and its bottleneck is the calculating of 2 × 2 tiles, therefore, proposed by the invention 2 × 2 tile structures extremely there is meaning.

Accompanying drawing explanation

Fig. 1 is the hardware architecture diagram of DP in tradition MGS algorithm.

Fig. 2 is the hardware architecture diagram of TP in tradition MGS algorithm.

Fig. 3 is present invention hardware architecture diagram of CORE_GEQRT in concrete application example.

Fig. 4 is present invention hardware architecture diagram of CORE_TTQRT in concrete application example.

Detailed description of the invention:

Below with reference to Figure of description and specific embodiment, the present invention is described in further details.

The MGS structure of the present invention includes: CORE_GEQRT and CORE_TTQRT, and wherein CORE_GEQRT is used for input 2 × 2 matrix [a₁,a₂] carry out QR decomposition output q by MGS algorithm₁,r₁₁,r₂₂,sign(f(A)),r₁₂,q₂, wherein [q₁,q₂] it is Q matrix, r₁₁,r₁₂,r₂₂Composition upper triangular matrix R, sign (f (A))=(a₂₁a₁₂-a₁₁a₂₂)r₁₁.Two the R matrixes obtained are made Input and [1,0 for CORE_TTQRT；01]、[0,0；0,0] matrix is collectively constituted [a₁₁,a₁₂,1,0；0,a₂₂,0,1；a₃₁,a₃₂,0,0；0,a₄₂, 0,0], finally give the matrix of 4 × 2.

In concrete application example, as it is shown on figure 3, the hardware of CORE_GEQRT consists of:

First 2 × 2 matrixes of input are stored in buffer1, the matrix-vector a being stored into by selector₁Element a₁₁As taking advantage of Two inputs, a of musical instruments used in a Buddhist or Taoist mass 1₂₁Two inputs as multiplier 2.Two outputs that will obtainWithAs adder 1 Input, adder 1 obtains resultIt is stored in buffer2.Result adder 1 obtained reads input from buffer2 Carry out out radical sign module 1, the result that radical sign computing obtains will be openedIt is stored in buffer4.By vector a₁Make the dividend of divider, r₁₁It is separately input in divider 1 and divider 2 as divisor, result q obtained₁It is stored in buffer5。

After vector is input to buffer1, read vector a from buffer1₁While read vector a₂, first by a₁₁Become -a₁₁, then by data-a₁₁, a₁₂, a₂₁, a₂₂It is stored in buffer2.WhenWhen being input to out radical sign module 1, a₁₂, a₂₁ Input, a as multiplier 1₂₂,-a₁₁As the input of multiplier 2, export two obtained as adder 1 is defeated Entering, the output f (A) obtained the most at last is stored in buffer3.F (A) is read from buffer3 the computing that carries out taking absolute value, The value obtained is input in buffer4；Meanwhile, the symbol value as output Sign of f (A) is taken, by | f (A) | with r₁₁From buffer4 Middle reading inputs in dividing module 3, by operation result r₂₂It is stored in buffer5.

After f (A) has calculated, by element a₁₁, a₁₂It is input to multiplier 1, by element a₂₁, a₂₂Input multiplier 2, will The output obtained is as the input of adder 1, and the result that adder 1 obtains is stored in buffer4.The result that adder 1 is obtained Take out from buffer4 and r₁₁Jointly it is linked into the input of dividing module 4, obtains result r₁₂It is stored in buffer5.

When obtaining q₁After Sign (f (A)), being entered into and be input in selector 2, wherein Sign (f (A)) believes as condition Number.When Sign (f (A)) is timing q₂₁=q₁₂, q₂₂=-q₁₁, when Sign (f (A)) is q time negative₂₁=-q₁₂, q₂₂=q₁₁.? The q arrived₂As output.

As shown in Figure 4, in concrete application example, CORE_TTQRT includes two modules of QR and Column update. QR module is fully pipelined architecture, first clock cycle first to its input matrix A_1=[a₁₁,1；a₃₁, 0] and carry out QR operation, Second clock cycle to its input matrix A_2=[a₂₂,1；a₄₂, 0] and carry out QR operation.First, by 2 × 2 matrixes of input It is stored in buffer1, the matrix-vector a being stored into by selector₁Element a₁₁Two inputs, a as multiplier 3₂₁As Two inputs of multiplier 4, meanwhile, take a₂₁Symbol as output Sign.Two outputs that will obtainWithAs addition The input of device 2, adder 1 obtains resultIt is stored in buffer2.Result input adder 2 obtained carries out out radical sign mould Block 2, will open the result that radical sign computing obtainsAs output.By vector a₁Make divider 5 Dividend, r₁₁It is input to divider 6, result q obtained as divisor₁As output.Wherein r₁₂=q₁₁, r₂₂=q₁₂。

Obtain Sign and q₁After, by Sign and q₁As the input of selector, wherein Sign is as selecting signal.When Sign is Timing is output as q₂₁=q₁₂, q₂₂=-q₁₁, when Sign is for being output as q time negative₂₁=-q₁₂, q₂₂=q₁₁.The q obtained₂As defeated Go out.

After obtaining the Q matrix of A_1, to a₁₂, a₃₂Carry out row by Column update and update operation.By a₁₁, q₁₂As The input a of multiplier 5₂₁, q₂₂Input multiplier 6, by a₁₁, q₁₁Input multiplier 7, by a₂₁, q₂₁Input multiplier 8. The result input summer 3 that multiplier 5 and multiplier 6 obtain obtains resultThe result that multiplier 7 and multiplier 8 obtain Input summer 2 obtains resultObtain matrixTake matrixMiddle submatrixCarry out and A_1 After same operation, to vectorIt is updated operation.Obtain the R matrix of 2 × 2.

Below being only the preferred embodiment of the present invention, protection scope of the present invention is not limited merely to above-described embodiment, all belongs to Technical scheme under thinking of the present invention belongs to protection scope of the present invention.It should be pointed out that, the ordinary skill for the art For personnel, some improvements and modifications without departing from the principles of the present invention, should be regarded as protection scope of the present invention.

Claims (3)

1. a QRD structure based on MGS, it is characterised in that include the first module CORE_GEQRT and the second module CORE_TTQRT, described first module CORE_GEQRT is used for 2 × 2 matrix [a to input₁,a₂] carried out by MGS algorithm QR decomposes output q₁,r₁₁,r₂₂,sign(f(A)),r₁₂,q₂, wherein [q₁,q₂] it is Q matrix, r₁₁,r₁₂,r₂₂Composition upper triangular matrix R, Sign (f (A))=(a₂₁a₁₂-a₁₁a₂₂)/r₁₁；Using two R matrixes obtaining as the second module CORE_TTQRT input with [1,0；01]、[0,0；0,0] matrix [a is collectively constituted₁₁,a₁₂,1,0；0,a₂₂,0,1；a₃₁,a₃₂,0,0；0,a₄₂, 0,0], finally give the matrix of 4 × 2.

QRD structure based on MGS the most according to claim 1, it is characterised in that described first module CORE_GEQRT Including:

First 2 × 2 matrixes of input are stored in buffer1, the matrix-vector a being stored into by selector₁Element a₁₁As taking advantage of Two inputs, a of musical instruments used in a Buddhist or Taoist mass 1₂₁Two inputs as multiplier 2；Two outputs that will obtainWithAs adder 1 Input, adder 1 obtains resultIt is stored in buffer2；Result adder 1 obtained reads input from buffer2 Carry out out radical sign module 1, the result that radical sign computing obtains will be openedIt is stored in buffer4；By vector a₁Make the dividend of divider, r₁₁It is separately input in divider 1 and divider 2 as divisor, result q obtained₁It is stored in buffer5；

After vector is input to buffer1, read vector a from buffer1₁While read vector a₂, first by a₁₁Become -a₁₁, then by data-a₁₁, a₁₂, a₂₁, a₂₂It is stored in buffer2；WhenWhen being input to out radical sign module 1, a₁₂, a₂₁ Input, a as multiplier 1₂₂,-a₁₁As the input of multiplier 2, export two obtained as adder 1 is defeated Entering, the output f (A) obtained the most at last is stored in buffer3；F (A) is read from buffer3 the computing that carries out taking absolute value, The value obtained is input in buffer4；Meanwhile, the symbol value as output Sign of f (A) is taken, by | f (A) | with r₁₁From buffer4 Middle reading inputs in dividing module 3, by operation result r₂₂It is stored in buffer5；

After f (A) has calculated, by element a₁₁, a₁₂It is input to multiplier 1, by element a₂₁, a₂₂Input multiplier 2, will The output obtained is as the input of adder 1, and the result that adder 1 obtains is stored in buffer4；The result that adder 1 is obtained Take out from buffer4 and r₁₁Jointly it is linked into the input of dividing module 4, obtains result r₁₂It is stored in buffer5；

When obtaining q₁After Sign (f (A)), being entered into and be input in selector 2, wherein Sign (f (A)) believes as condition Number；When Sign (f (A)) is timing q₂₁=q₁₂, q₂₂=-q₁₁, when Sign (f (A)) is q time negative₂₁=-q₁₂, q₂₂=q₁₁, The q arrived₂As output.

QRD structure based on MGS the most according to claim 1 and 2, it is characterised in that described second module CORE_TTQRT includes: include three module QR and the 4th module Column update；Described three module QR is full flowing water Structure, first clock cycle first to its input matrix A_1=[a₁₁,1；a₃₁, 0] and carry out QR operation, second clock cycle To its input matrix A_2=[a₂₂,1；a₄₂, 0] and carry out QR operation；First, 2 × 2 matrixes of input are stored in buffer1, pass through The matrix-vector a that selector is stored into₁Element a₁₁Two inputs, a as multiplier 3₂₁Two as multiplier 4 defeated Enter, meanwhile, take a₂₁Symbol as output Sign；Two outputs that will obtainWithAs the input of adder 2, add Musical instruments used in a Buddhist or Taoist mass 1 obtains resultIt is stored in buffer2；Result input adder 2 obtained carries out out radical sign module 2, will open radical sign fortune The result obtainedAs output；By vector a₁Make the dividend of divider 5, r₁₁As removing Number is input to divider 6, result q obtained₁As output；Wherein r₁₂=q₁₁, r₂₂=q₁₂；

Obtain Sign and q₁After, by Sign and q₁As the input of selector, wherein Sign is as selecting signal；When Sign is just Time be output as q₂₁=q₁₂, q₂₂=-q₁₁, when Sign is for being output as q time negative₂₁=-q₁₂, q₂₂=q₁₁；The q obtained₂As output；

After obtaining the Q matrix of A_1, to a₁₂, a₃₂Carry out row by Column update and update operation；By a₁₁, q₁₂As The input a of multiplier 5₂₁, q₂₂Input multiplier 6, by a₁₁, q₁₁Input multiplier 7, by a₂₁, q₂₁Input multiplier 8； The result input summer 3 that multiplier 5 and multiplier 6 obtain obtains resultThe result that multiplier 7 and multiplier 8 obtain Input summer 2 obtains resultObtain matrixTake matrixMiddle submatrixCarry out and A_1 After same operation, to vectorIt is updated operation, obtains the R matrix of 2 × 2.