CN100392638C - Fixed-point composite signal modulus value normalization - Google Patents
Fixed-point composite signal modulus value normalization Download PDFInfo
- Publication number
- CN100392638C CN100392638C CNB2004100095277A CN200410009527A CN100392638C CN 100392638 C CN100392638 C CN 100392638C CN B2004100095277 A CNB2004100095277 A CN B2004100095277A CN 200410009527 A CN200410009527 A CN 200410009527A CN 100392638 C CN100392638 C CN 100392638C
- Authority
- CN
- China
- Prior art keywords
- normalization
- fixed
- calculate
- plural
- displacement
- 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.)
- Expired - Fee Related
Links
Landscapes
- Mobile Radio Communication Systems (AREA)
Abstract
The present invention relates to a fixed-point composite signal modulus value normalization method which is applied to a fixed-point digital signal processor and comprises the following steps: in the step 1, a normalized factor table K(n) is generated and is loaded in the fixed-point digital signal processor; in the step 2, according to the modulus square of a fixed point complex number Y required to be normalized, the fixed-point digital signal processor calculates and acquires revision displacement bit number p and looks up the offset address r of the normalized factor table K(n); in the step 3, the fixed-point digital signal processor makes the fixed point complex number Y multiplied by a normalized factor K(r) acquired by looking up the table and carries out correction displacement treatment to obtain a modulus value normalized complex number Z. Using the method of the present invention can carry out modulus value normalization with very high precision on complex signals of the modulus value in large dynamic range. Moreover, the method occupies few resources and needs small operand.
Description
Technical field
The present invention relates to digital processing field, particularly relate to the smart antenna downlink weights and obtain and method for transmitting based on the fixed point complex signal modulus value normalization processing of fixed-point dsp.
Background technology
Fixed-point dsp is simple in structure, cheap with it, is used widely in digital processing field.The operand of fixed point digital signal process chip has symbol or unsigned int numerical table to show with scale-of-two.The fixed point digital signal process chip given word length determined the dynamic range and the processing accuracy of the scale-of-two integer number that it can be represented.How to make full use of the given word length of fixed point digital signal process chip, reduce truncation error in the calculating process and cumulative errors and be key and difficult point problem during fixed point digital signal is handled.
The normalization of fixed point modulus of complex number value is often used in signal of communication is handled.For example in the smart antenna exploitation of systems such as PHS, WCDMA or TD-SCDMA, descending weights need be made the fixed point modulus of complex number value normalized of degree of precision.The signal of the receiving equipment receiving antenna array of common said system, handle up, the descending restore one's right value that obtains subscriber signal through intelligent algorithm, descending restore one's right value is being composed to needing before the transmitter through the fixed point normalized, control its dynamic range, this process can realize on digit chips such as fixed DSP, FPGA or CPLD.The smart antenna of said system if normalization as a result the precision deficiency can bring a series of problems, too big such as descending synthetic power swing, overall performance descends; Particularly the mould value is too big as a result in normalization, when causing descending weights overflow, will cause unforeseen result.Therefore in said system, need the fixed point modulus of complex number value method for normalizing that adopts precision higher, to satisfy system to descending performance demands.
And the plural modulus value normalization problem of fixing a point will be used the fixed point nonlinear operation.Nonlinear operation can realize by direct calculating or look-up table as evolution, index, logarithm and trigonometric function etc.In general look-up table is effective, more general comparatively fast.But for nonlinear operation, construct very difficulty of suitable form, often size is bigger to use the form of conventional method structure, and nonlinear operation as a result precision be difficult to control.Use conventional look-up table to be difficult to guarantee at present,, all can obtain high-precision fixed point modulus value normalization result for the fixed point complex signal of big dynamic mode value taking under the less situation of resource.
Summary of the invention
Technical matters to be solved by this invention is to provide the smart antenna downlink weights to handle and obtains and launching technique, fixed point complex signal modulus value normalization based on look-up table is handled, fixed-point dsp (digital signal processor at antenna system, DSP) carry out in composing after the fixed point normalized of the descending weights of base station intelligent antenna and launch after giving smart antenna, obtain the descending weights of high precision, overcome descending weights precision deficiency, the technical matters that smart antenna brings when emission.
To achieve these goals, the invention provides a kind of smart antenna downlink weights processing and obtain and launching technique, its characteristics are, comprise the steps:
Step 1 produces a normalization coefficient form K (n),
Wherein
Expression rounds, and M is the mould value of the fixed point plural number Z after the system requirements normalization, n=1, and 2 ..., 2
2* m-1, m are normalized factor, and α is an appropriate integer, make the maximal value of form K (n) be in the process range of signal handling equipment; And be loaded in the fixed-point dsp;
Step 2, according to the mould of the plural Y of the normalized fixed point of need square | Y|
2, this fixed-point dsp calculating binary representation | Y|
2Number of significant digit q, calculate to obtain according to number of significant digit q and parameter f and revise displacement figure place p, wherein parameter f=log
2(2
2* m); Look into the offset address r of this normalization coefficient form K (n), wherein
Expression rounds;
Step 3, this fixed-point dsp plural Y that will fix a point multiply by the normalization coefficient K (r) that tables look-up and obtain, and multiply by by revising the correction shifted divisor σ that displacement figure place p calculates again, and obtains the plural Z behind the modulus value normalization;
Step 4, the plural Z after the normalization composes to smart antenna, and the plural Z of smart antenna after according to normalization carries out the synthetic and emission of descending power.
The present invention is by fixed-point dsp (the digital signal processor at antenna system, DSP) carry out the fixed point normalized of the descending weights of base station intelligent antenna in, obtain the high descending weights of precision, descending weights are being composed to smart antenna and emission, overcome existence in the known technology because descending weights precision deficiency causes degradation technical matters under the antenna system overall performance.
Describe the present invention below in conjunction with the drawings and specific embodiments, but not as a limitation of the invention.
Description of drawings
Fig. 1 is the process flow diagram of descending weights normalized of the present invention.
Embodiment
In the present invention, with the PHS antenna system is the method for normalizing that example illustrates fixed point modulus of complex number value of the present invention, it is mainly used in the fixed point normalized of the descending weights of PHS base station intelligent antenna, the PHS base station adopts fixed DSP to read the received signal of four aerial arrays, and it is up to go out four-way by the smart antenna algorithm computation, descending weights, fixed DSP writes descending weights before the descending modulation FPGA then, at first by fixed point modulus of complex number value method for normalizing of the present invention with the modulus value normalization of descending weights scope to system requirements, and the descending weights after the normalization are composed to behind the smart antenna, the descending weights of smart antenna after according to normalization carry out the synthetic and emission of descending power.
Utilize modulus value normalization method provided by the invention, if given normalized factor m and normalized mode value M, then the absolute error of the mould value after the normalization will for:
Relative error is
In conjunction with the characteristics of fixed-point dsp, general m and M all get 2 integral number power, and the multiplication and division fixed-point arithmetic of m or M can be finished by displacement like this, are easy to realize on the fixed point chip.
See also Fig. 1, show the implementation method of modulus value normalization provided by the invention, comprise the steps:
Step 101: produce normalization coefficient form K (n) by method provided by the invention;
Step 102: according to the mould of the plural Y of the normalized fixed point of need square | Y|
2, calculate the offset address r that revises displacement figure place p and look into normalization coefficient form K (n);
Step 103: the plural Y that fixes a point multiply by normalization coefficient K (r), and do to revise displacement, obtains the plural Z behind the modulus value normalization.
Described step 101 is used for producing normalization coefficient table K (n) required for the present invention.If the mould value of the fixed point after system requirements normalization plural number Z is M, and the relative error of the M that allows of system is e, then basis
Can calculate
Determined the value of m again by the size of λ, wherein m equals 2 positive integer time power, its value greater than and approach λ most.At last, make n=1,2 ..., 2
2* m-1, and calculate
Obtain required normalization coefficient form K (n), wherein
Expression rounds, and is generally to prevent normalization overflow as a result, and normal here employing rounds downwards; α is an appropriate integer, makes the maximal value of form K (n) be in the process range of signal handling equipment.Normalization coefficient form K (n) can be loaded in digital signal processor or other peripherals, so that use in the step of back.
Described step 102 is calculated the offset address r that looks into normalization coefficient form K (n) and is revised displacement figure place p according to the fixed point plural number Y of input.
Described step 103 is finally finished the modulus value normalization process of the plural Y of fixed point.
Wherein, two kinds of optional implementation methods are arranged, are described as follows successively for step 2 and step 3:
Method one:
In step 2, establish the fixed point plural number Y=a+b * j that will carry out modulus value normalization, wherein
Make f=log
2(2
2* m).We calculate the mould square of Y | Y|
2=a
2+ b
2, and calculate binary representation | Y|
2Number of significant digit q.If q<f then gets and revises displacement figure place p=0; If q 〉=f and q-f are odd number, then get and revise displacement figure place p=q-f+1; If q 〉=f and q-f are even number, then get and revise displacement figure place p=q-f.After having determined to revise displacement figure place p, the offset address r that looks into normalization coefficient form K (n) can pass through
Calculate, here
Expression rounds downwards.
In the step 3, we can obtain normalization coefficient κ=K (r) to be offset location r by tabling look-up, and can calculate correction shifted divisor σ=2 by revising displacement figure place p simultaneously
-(α+p/2), the then final normalization that requires is Z=Y * κ * σ as a result.Usually calculate Y * κ when calculating Z earlier, with abundant assurance computational accuracy, the α+p/2 bit that then product moved to right can be realized Y * κ and the multiplication mutually of revising shifted divisor σ.
Method two:
In step 2, establish the fixed point plural number Y=a+b * j that will carry out modulus value normalization, wherein
Make f=log
2(2
2* m).We calculate the mould square of Y | Y|
2=a
2+ b
2, and calculate binary representation | Y|
2Number of significant digit q.If q<f then gets and revises displacement figure place p=0; If q 〉=f then gets and revises displacement figure place p=q-f.After having determined to revise displacement figure place p, the offset address r that looks into normalization coefficient form K (n) can pass through
Calculate, here
Expression rounds downwards.
In the step 3, we can obtain normalization coefficient κ=K (r) to be offset location r by tabling look-up, and can calculate correction shifted divisor σ=2 by revising displacement figure place p simultaneously
-(α+p/2)Calculate Z '=Y * κ, then when q-f was even number, the final normalization that requires is Z=Z ' * σ=Y * κ * σ as a result; When q-f is odd number, the final normalization result who requires
Wherein, with the step execution in the end usually of the multiplication mutually of revising shifted divisor σ, can realize by α that product is moved to right+p/2 bit; When q-f is odd number, multiply by
Component, can obtain by approximation method, for example multiply by 724,9 bits again move to right product.
In addition, in the process of tabling look-up of the present invention, can make full use of the resource that digital signal processor provides,, obtain as specific instruction etc. | Y|
2Number of significant digit q, and obtain to revise displacement figure place p by method provided by the invention.Will | Y|
2The p position that moves to right makes it fall into interval [m, 2
2* m) in, thereby obtain looking into the offset address r of normalization coefficient table K (n), tabling look-up to obtain normalization coefficient κ.
Even the dynamic range of input fixed point plural number is bigger, utilize method for normalizing provided by the invention can guarantee that still normalization result's relative error satisfies requirement of system design.For the big more number of mould value, the number of significant digit of its mould square is big more, and therefore the figure place p that is cut is also many more, but normalization result's relative error remains unchanged.
Certainly; the present invention also can have other various embodiments; under the situation that does not deviate from spirit of the present invention and essence thereof; those of ordinary skill in the art work as can make various corresponding changes and distortion according to the present invention, but these corresponding changes and distortion all should belong to the protection domain of the appended claim of the present invention.
Claims (10)
1. smart antenna downlink weights are handled and are obtained and launching technique, it is characterized in that, comprise the steps:
Step 1 produces a normalization coefficient form K (n),
Wherein
Expression rounds, and M is the mould value of the fixed point plural number Z after the system requirements normalization, n=1, and 2 ..., 2
2* m-1, m are normalized factor, and α is an appropriate integer, make the maximal value of form K (n) be in the process range of signal handling equipment; And be loaded in the fixed-point dsp;
Step 2, according to the mould of the plural Y of the normalized fixed point of need square | Y|
2, this fixed-point dsp calculating binary representation | Y|
2Number of significant digit q, calculate to obtain according to number of significant digit q and parameter f and revise displacement figure place p, wherein parameter f=log
2(2
2* m); Look into the offset address r of this normalization coefficient form K (n), wherein
Expression rounds;
Step 3, this fixed-point dsp plural Y that will fix a point multiply by the normalization coefficient K (r) that tables look-up and obtain, and multiply by by revising the correction shifted divisor σ that displacement figure place p calculates again, and obtains the plural Z behind the modulus value normalization;
Step 4, the plural Z after the normalization composes to smart antenna, and the plural Z of smart antenna after according to normalization carries out the synthetic and emission of descending power.
2. smart antenna downlink weights according to claim 1 are handled and are obtained and launching technique, it is characterized in that in step 1, the production process of the value of normalized factor m is as follows:
If the relative error of the M that system allows is e, then basis
Can calculate intermediate parameters
Determined the value of given normalized factor m again by the size of intermediate parameters λ, wherein m equals 2 positive integer time power, its value greater than and approach λ most.
3. smart antenna downlink weights according to claim 2 are handled and are obtained and launching technique, it is characterized in that, this normalization coefficient table K (n) adopts and rounds downwards.
4. smart antenna downlink weights according to claim 3 are handled and are obtained and launching technique, it is characterized in that, and in the described step 2, this fixed-point dsp calculating binary representation | Y|
2Number of significant digit q, calculate to obtain according to number of significant digit q and parameter f and revise displacement figure place p, specifically comprise:
If carry out the fixed point plural number Y of modulus value normalization;
Calculate to need the mould square of the plural Y of normalized fixed point | Y|
2, and calculate binary representation | Y|
2Number of significant digit q;
If q<f then gets and revises displacement figure place p=0; If q 〉=f and q-f are odd number, then get and revise displacement figure place p=q-f+1; If q 〉=f and q-f are even number, then get and revise displacement figure place p=q-f.
5. smart antenna downlink weights according to claim 4 are handled and are obtained and launching technique, it is characterized in that described step 3 comprises:
Obtain normalization coefficient κ=K (r) by the offset address r that tables look-up;
Calculate correction shifted divisor σ=2 by revising displacement figure place p simultaneously
-(α+p/2)
Calculate the plural Z=Y * κ * σ behind the modulus value normalization.
6. smart antenna downlink weights according to claim 5 are handled and are obtained and launching technique, it is characterized in that, usually calculate Y * κ when calculating the plural Z behind the modulus value normalization earlier, with abundant assurance computational accuracy, the α+p/2 bit that then product moved to right can be realized Y * κ and the multiplication mutually of revising shifted divisor σ.
7. smart antenna downlink weights according to claim 3 are handled and are obtained and launching technique, it is characterized in that, and in the described step 2, this fixed-point dsp calculating binary representation | Y|
2Number of significant digit q, calculate to obtain according to number of significant digit q and parameter f and revise displacement figure place p, specifically comprise:
If carry out the fixed point plural number Y of modulus value normalization;
Calculate to need the mould square of the plural Y of normalized fixed point | Y|
2, and calculate binary representation | Y|
2Number of significant digit q;
If q<f then gets and revises displacement figure place p=0; If q 〉=f then gets and revises displacement figure place p=q-f.
8. smart antenna downlink weights according to claim 7 are handled and are obtained and launching technique, it is characterized in that described step 3 comprises:
Skew location r obtains normalization coefficient κ=K (r) by tabling look-up;
Calculate correction shifted divisor σ=2 by revising displacement figure place p simultaneously
-(α+p/2)
Calculate Z '=Y * κ;
When q-f was even number, the final normalization that requires is Z=Z ' * σ=Y * κ * σ as a result; When q-f is odd number, the final normalization result who requires
9. smart antenna downlink weights according to claim 8 are handled and are obtained and launching technique, it is characterized in that, with the step execution in the end usually of the multiplication mutually of revising shifted divisor σ, can realize by α that product is moved to right+p/2 bit; When q-f is odd number, multiply by
Component, can obtain by approximation method.
10. obtain and launching technique according to claim 5 or 8 described smart antenna downlink weights processing, it is characterized in that, by inciting somebody to action | Y|
2The p position that moves to right makes it fall into interval [m, 2
2* m) in, thereby obtain looking into the offset address r of normalization coefficient table K (n), tabling look-up to obtain normalization coefficient κ.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CNB2004100095277A CN100392638C (en) | 2004-09-08 | 2004-09-08 | Fixed-point composite signal modulus value normalization |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CNB2004100095277A CN100392638C (en) | 2004-09-08 | 2004-09-08 | Fixed-point composite signal modulus value normalization |
Publications (2)
Publication Number | Publication Date |
---|---|
CN1746877A CN1746877A (en) | 2006-03-15 |
CN100392638C true CN100392638C (en) | 2008-06-04 |
Family
ID=36166418
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CNB2004100095277A Expired - Fee Related CN100392638C (en) | 2004-09-08 | 2004-09-08 | Fixed-point composite signal modulus value normalization |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN100392638C (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP5101378B2 (en) * | 2008-04-16 | 2012-12-19 | ソニー株式会社 | Reader / Writer, Communication Processing Device, Communication Processing Method, Data Management System, and Communication System |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2002099999A1 (en) * | 2001-05-31 | 2002-12-12 | Magn0Lia Broadband, Inc. | Communication device with smart antenna using a quality-indication signal |
CN1503491A (en) * | 2002-11-25 | 2004-06-09 | 深圳市中兴通讯股份有限公司上海第二 | Intelligent receiving-transmitting diversity method and device for antenna |
-
2004
- 2004-09-08 CN CNB2004100095277A patent/CN100392638C/en not_active Expired - Fee Related
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2002099999A1 (en) * | 2001-05-31 | 2002-12-12 | Magn0Lia Broadband, Inc. | Communication device with smart antenna using a quality-indication signal |
CN1503491A (en) * | 2002-11-25 | 2004-06-09 | 深圳市中兴通讯股份有限公司上海第二 | Intelligent receiving-transmitting diversity method and device for antenna |
Also Published As
Publication number | Publication date |
---|---|
CN1746877A (en) | 2006-03-15 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US6237014B1 (en) | Correlator method and apparatus | |
CN101174200A (en) | 5-grade stream line structure of floating point multiplier adder integrated unit | |
CN105183424A (en) | Fixed-bit-width multiplier with high accuracy and low energy consumption properties | |
CN100555212C (en) | The carry calibration equipment of a kind of floating dual MAC and multiplication CSA compressed tree thereof | |
CN112434801B (en) | Convolution operation acceleration method for carrying out weight splitting according to bit precision | |
CN109144473B (en) | Decimal 3:2 compressor structure based on redundant ODDS number | |
CN110688086A (en) | Reconfigurable integer-floating point adder | |
CN116400883A (en) | Floating point multiply-add device capable of switching precision | |
CN116661733A (en) | Multiplier and microprocessor supporting multiple precision | |
CN110222305B (en) | Logarithmic function calculation system and method based on hyperbolic CORDIC | |
US20110153995A1 (en) | Arithmetic apparatus including multiplication and accumulation, and dsp structure and filtering method using the same | |
CN100392638C (en) | Fixed-point composite signal modulus value normalization | |
CN110825346A (en) | Low-logic-complexity unsigned approximate multiplier | |
CN207115387U (en) | XIU accumulator registers, XIU accumulator registers circuit and electronic equipment | |
CN103399727A (en) | Hardware integersaturation detector, method for detectingsaturation, and hardware device thereof | |
CN103036577A (en) | Low-complexity code circuit structure of low density parity check (LDPC) code | |
US20070294330A1 (en) | Systems, methods and computer program products for providing a combined moduli-9 and 3 residue generator | |
US7330867B2 (en) | Method and device for floating-point multiplication, and corresponding computer-program product | |
CN102545910B (en) | A kind of jpeg huffman decoding circuit and coding/decoding method thereof | |
US20140136582A1 (en) | Method and apparatus for digital automatic gain control | |
WO2023078364A1 (en) | Operation method and apparatus for matrix multiplication | |
US20220067120A1 (en) | Device for computing an inner product of vectors | |
US20040122878A1 (en) | Circuits and methods for implementing approximations to logarithms | |
CN115543911B (en) | Method for calculating computing capacity of heterogeneous computing equipment | |
CN116263875A (en) | Convolution processing method, convolution processing system and terminal equipment |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
C17 | Cessation of patent right | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20080604 Termination date: 20130908 |