Background technology
In base station system, about 80% power consumption is produced by radio-frequency (RF) power amplification, therefore along with the development of digital mobile communication technology, high efficiency power amplifier become Ge great system equipment manufacturer must be satisfied one of requirement.Simultaneously, along with frequency spectrum resource growing tension, Modern Communication System more trends towards adopting has the more linear modulation mode of the non-constant envelope of spectral efficient compared with constant envelope modulation mode, and this modulation system is higher to the linearity of high power amplifier, so adopt this modulation system need to adopt linearization technique to improve the linearity of power amplifier.
At present, in power amplifier linearization technology, the technology of main flow is exactly pre-distortion technology, it is applied to the radio frequency part of analog communication system at first, along with DSP (Digital Signal Process, Digital Signal Processing) development of technology, also can in numeric field, realize, form DPD (Digital Pre-Distortion, digital pre-distortion) technology.DPD technology both can be applied in the baseband portion of digital communication system, also can be applied in radio frequency part, and DPD technology can also utilize adaptive principle to follow the tracks of compensating power amplifier because the environmental factor such as temperature, humidity changes the error causing.In a word, DPD technology not only can promote the efficiency of transmitter, reduces costs and reduced volume, also can effectively increase the linearity of transmitter, and with elevator system usefulness and communication quality, DPD technology is a kind of linearization technique that adapts to modern digital Communication Development.
For the ease of realizing, what predistortion structure adopted is LUT (Look-Up Table, look-up table) method.The principle of LUT is that pre-distortion characteristics is left in a data look-up table, read numerical value in table according to input signal, then process accordingly, draw the pre-distorted signals that needs output, numerical value in look-up table is constantly updated by adaptive Fast Convergent Algorithm, thereby can embody well the pre-distortion characteristics contrary with the nonlinear characteristic of power amplifier.Fig. 1 is a kind of structural representation of basic extraction look-up table.Wherein, X is input signal, and Y is output (feedback) signal.First determine according to demand power output, while operation for the first time, system is straight-through, form parameter is complete " 1 ", X '=X, signal, by after power amplifier, obtains first group of output data, feedback is carried out predistortion, according to min|X '-Y ' |, obtain form parameter, then continuous correction card parameter in iterative process, until output signal Y obtains satisfied effect, extract now form parameter as the look-up table under this power output for system.
Fig. 2 is the structural representation of pre-distortion system.As shown in Figure 2, pre-distortion system structure comprises two passages: data training passage and predistortion passage.Data training passage is a loop structure, its core is pre-distortion algorithms module, this pre-distortion algorithms module is processed the feedback signal after power amplifier and original input signal, obtains the pre-distortion characteristics of power amplifier, then obtains the form parameter of power amplifier predistortion characteristic.Predistortion passage is mainly realized forward signal index look-up table, draws pre-distorted signals X '.Taking complex gain pre-distortion algorithms as example, obtain predistortion value according to the performance number index look-up table of forward signal, forward signal X and predistortion value multiply each other and obtain pre-distorted signals X '.In the ideal case, after X ' mistake power amplifier, output Y=X.
Above-mentioned pre-distortion system is the performance number according to forward signal, utilize forward signal index look-up table to obtain pre-distorted signals X ', wherein the index value of look-up table is integer, but forward signal performance number differs and is decided to be integer, if removing to search index value by non-integral forward signal performance number is that the look-up table of integer can cause between the predistortion value that indexes and real predistortion value and has error, this error is brought into after X ', is likely further amplified by power amplifier, causes DPD performance to worsen.Especially in the time that forward direction signal power value is larger, the interval curvature of look-up table likely indexing is larger, also be that between the adjacent gear of look-up table, predistortion value is widely different, index error now may be very large, after this error and forward signal multiply each other, can further be amplified.This phenomenon is verified in reality test, when index error is larger, not only can cause DPD index variation, even can bring signal to cross to make an uproar after table the problem that the end lifts, and the stability of system is damaged.
Summary of the invention
Technical problem to be solved by this invention is the defect for prior art, proposes a kind of digital pre-distortion processing method and device, reduces index error, promotes the performance of digital pre-distortion system.
In order to solve the problems of the technologies described above, the invention provides a kind of digital pre-distortion processing method, comprising:
According to predistortion model, ask for the predistortion multinomial coefficient of described predistortion model;
According to predistortion multinomial, obtain integer searching table and N decimal look-up table, described N decimal look-up table comprises that 1 rank Taylor's decimal look-up table is to N rank Taylor's decimal look-up table; Wherein, N is more than or equal to 1;
Index out respectively the predistortion value of integer searching table and N decimal look-up table according to forward signal performance number, calculate predistortion value corresponding to forward signal performance number according to Taylor's formula;
According to forward signal and predistortion value corresponding to forward signal performance number, obtain pre-distorted signals.
Further, described according to predistortion model, the predistortion multinomial coefficient of asking for described predistortion model comprises: the criterion of the error minimum between signal and forward signal based on feedback signal after predistortion model, ask for described predistortion multinomial coefficient.
Further, described method also comprises: calculate the n power value of decimal, the n power value of described decimal is stored in tables of data; When calculating predistortion value corresponding to forward signal performance number, the parameter that obtains using in computational process by data query table according to Taylor's formula.
Further, in the described process of calculating predistortion value corresponding to forward signal performance number according to Taylor's formula, adopt cut position method to realize division arithmetic.
Further, described according to forward signal and predistortion value corresponding to forward signal performance number, obtain pre-distorted signals and comprise: by after forward signal time delay certain hour, the predistortion value corresponding with described forward signal performance number multiplies each other and obtain described pre-distorted signals.
The present invention also provides a kind of digital predistortion process apparatus, comprising:
Multinomial acquisition module, for according to predistortion model, asks for the predistortion multinomial coefficient of described predistortion model;
Look-up table acquisition module, for according to predistortion multinomial, obtains integer searching table and N decimal look-up table, and described N decimal look-up table comprises that 1 rank Taylor's decimal look-up table is to N rank Taylor's decimal look-up table; Wherein, N is more than or equal to 1;
Computing module, for index out respectively the predistortion value of integer searching table and N decimal look-up table according to forward signal performance number, calculates predistortion value corresponding to forward signal performance number according to Taylor's formula;
Processing module, for according to forward signal and predistortion value corresponding to forward signal performance number, obtains pre-distorted signals.
Further, described multinomial acquisition module, specifically for the criterion of the error minimum between signal and the forward signal after predistortion model based on feedback signal, is asked for described predistortion multinomial coefficient.
Further, described device also comprises: memory module, for calculating the n power value of decimal, stores the n power value of described decimal in tables of data into; Described computing module calculates predistortion value corresponding to forward signal performance number, the parameter that obtains using in computational process by data query table specifically for working as according to Taylor's formula.
Further, described computing module, specifically for calculating according to Taylor's formula in the process of predistortion value corresponding to forward signal performance number, adopts cut position method to realize division arithmetic.
Further, described processing module is specifically for by after forward signal time delay certain hour, and the predistortion value corresponding with described forward signal performance number multiplies each other and obtain described pre-distorted signals.
In the present invention, forward signal performance number is divided into integer and decimal look-up table corresponding to index respectively, carry out matching according to Taylor's formula and approach the true predistortion value that forward signal is corresponding, can obviously improve index precision, in the time that Taylor series meet certain condition, in the span of any forward signal performance number, error is all less than the index error of prior art, has promoted the performance of digital pre-distortion system.
Embodiment
The invention provides a kind of digital pre-distortion processing method and device, its core concept is that forward signal performance number is divided into integer and decimal look-up table corresponding to index respectively, carries out matching approach the true predistortion value that forward signal is corresponding according to Taylor's formula.Concrete scheme of the present invention is: first according to predistortion model, ask for the predistortion multinomial coefficient of predistortion model; According to predistortion multinomial, obtain integer searching table and N decimal look-up table, wherein N decimal look-up table comprises that 1 rank Taylor's decimal look-up table is to N rank Taylor's decimal look-up table; Wherein, N is more than or equal to 1; Index out respectively the predistortion value of integer searching table and N decimal look-up table according to forward signal performance number, calculate predistortion value corresponding to forward signal performance number according to Taylor's formula; According to forward signal and predistortion value corresponding to forward signal performance number, obtain pre-distorted signals.In the present invention, forward signal performance number is divided into integer and decimal look-up table corresponding to index respectively, carry out matching according to Taylor's formula and approach the true predistortion value that forward signal is corresponding, can obviously improve index precision, in the time that Taylor series meet certain condition, in the span of any forward signal performance number, error is all less than the index error of prior art, has promoted the performance of digital pre-distortion system.
On mathematics, infinite Taylor (Taylor) progression that can micro-real variable function f (x) being defined on open interval has following power series:
Wherein, n! Represent the factorial of n; f
(n)(a) representative function f (x) is at the n order derivative at an a place.
The present invention mainly utilizes two important feature of Taylor series:
(1) Taylor series can be used for the value of approximate calculation function, the theoretical foundation of this predistortion value that is approaching to reality of the present invention; (2) differentiate of power series and integration can carry out item by item, and therefore summing function is relatively easy, the theoretical foundation that this carries out hierarchical index summation for the present invention.
For the pre-distortion system that adopts LUT method, in its look-up table, record the corresponding relation as forward signal performance number and the predistortion value of index value.In the present invention, establishing index value is x, and its corresponding predistortion value is f (x), and obtaining thus the infinite of look-up table can micro-real variable function f (x).For function f (x), provide 1 rank Taylors approximation formula, shown in (2):
f(x)≈f(x
k-1)+f
(1)(x
k-1)(x-x
k-1) (2)
In order further to improve approximation quality, provide 2 rank Taylors approximation formula, shown in (3):
In like manner, also can provide N rank Taylors approximation formula.
Based on above-mentioned theory basis, by reference to the accompanying drawings and preferred implementation technical solution of the present invention is elaborated.
Fig. 3 is the schematic flow sheet of digital pre-distortion processing method embodiment mono-provided by the invention.As shown in Figure 3, the present embodiment comprises the following steps:
Step 101, according to predistortion model, ask for the predistortion multinomial coefficient of predistortion model;
If input (forward direction) signal is X=(x
1x
2x
n), output (feedback) signal is Y=(y
1y
2y
n), the present embodiment adopts an example of following predistortion model formation:
X
DPD(n)=f
1(P
n)*X(n)+f
2(P
n-1)*X(n-1)+f
3(P
n-2)*X(n-2) (4)
Wherein P
nfor the mould value of input signal; f
m() (m=1,2,3) is the look-up table function of fitting of a polynomial.
Calculate predistortion multinomial coefficient (also claiming DPD parameter) criterion can be feedback signal Y after predistortion model, make RMS (Root Mean Square, root mean square) the error minimum between Y ' and X,
ε(n)=f
1(P
n)*Y(n)+f
2(P
n-1)*Y(n-1)+f
3(P
n-1)*Y(n-2)-X(n) (5)
Value minimum.Being write as matrix form has:
ε=(P.*Y)*W-X (6)
ω
m=[α
m0, α
m1..., α
mk] be f
mthe multinomial coefficient of (), its multinomial top step number is k.
Make the solution of the coefficient vector W of ε minimum be:
W=(P.*Y)
-1*X (7)
Can try to achieve thus predistortion multinomial coefficient.
In this step, the power that can first ask for separately P calculates, and to reduce operand, also makes follow-up Taylor series carry out can adopting corresponding fast algorithm in integer decimal index approach method, does not need to recalculate the power operation result of P.
Step 102, according to predistortion multinomial, obtain integer searching table and N decimal look-up table;
The coefficient of trying to achieve by step 101, by substitution predistortion multinomial and each derived function thereof, obtains integer searching table according to predistortion multinomial, obtains N decimal look-up table according to each derived function.
Suppose the coefficient vector W=[α asking for
0, α
1, α
2, α
3...], integer searching table is normalized to the integer searching table index value of regulation according to putting forward table data peaks, and predistortion value corresponding to each integer searching table index value can be according to predistortion polynomial computation, as shown in the formula:
f(x)=α
0+α
1x+α
2x
2+α
3x
3+α
4x
4+α
5x
5+α
6x
6+... (8)
What 1 rank Taylor's decimal look-up table was corresponding is the polynomial single order derived function of predistortion, and in this form, predistortion value corresponding to each decimal look-up table index value can be calculated according to the following formula:
f
(1)(x)=α
1+2α
2x+3α
3x
2+4α
4x
3+5α
5x
4+6α
6x
5+... (9)
What 2 rank Taylor's decimal look-up tables were corresponding is the polynomial second order derived function of predistortion, and in this form, predistortion value corresponding to each decimal look-up table index value can be calculated according to the following formula:
f
(2)(x)=2α
2+6α
3x+12α
4x
2+20α
5x
3+30α
6x
4+... (10)
The rest may be inferred, and what N rank Taylor's decimal look-up table was corresponding is the polynomial N of predistortion rank derived functions, can calculate predistortion value corresponding to each decimal look-up table index value according to this N rank derived function.
Step 103, index out respectively the predistortion value of integer searching table and N decimal look-up table according to forward signal performance number, calculate predistortion value corresponding to forward signal performance number according to Taylor's formula;
After obtaining integer searching table and N decimal look-up table, search in each look-up table corresponding with it predistortion value using forward signal performance number as index value, these predistortion value substitution Taylor's formulas are carried out to matching and approach predistortion value corresponding to forward signal performance number.
Step 104, according to forward signal and predistortion value corresponding to forward signal performance number, obtain pre-distorted signals.
Fig. 4 is the block diagram of 1 rank Taylor's look-up table index scheme in digital pre-distortion processing method embodiment bis-provided by the invention.In the present embodiment, describe as an example of 1 rank Taylor's look-up table index example.
First, the criterion of the error minimum between signal and forward signal based on feedback signal after predistortion model, asks for the predistortion multinomial coefficient of predistortion model in native system; According to this predistortion multinomial, obtain integer searching table and 1 rank Taylor's decimal look-up table.As shown in Figure 4, LUT represents integer searching table, LUT
(1)represent 1 rank Taylor's decimal look-up table, for example: LUT (35) represents the predistortion value of index value 35 correspondences in integer searching table, LUT
(1)(35) the predistortion value of index value 35 correspondences in expression 1 rank Taylor's decimal look-up table.
Then, calculate forward signal performance number, adopt 1 rank Taylor's formula (being formula (2)) to calculate predistortion value corresponding to forward signal performance number:
If forward signal performance number is 36.6, there is x
k-1=36, substitution formula (2) has:
f(36.6)≈LUT(36)+LUT
(1)(36)(36.6-36) (11)
Finally, in order to ensure that forward signal is corresponding one by one with the predistortion value indexing, by after forward signal time delay certain hour, multiply each other and obtain pre-distorted signals with the predistortion value of trying to achieve.Wherein, delay value is determined according to the actual conditions of specific implementation link.
Fig. 5 is the block diagram of 2 rank Taylor's look-up table index schemes in digital pre-distortion processing method embodiment tri-provided by the invention.In the present embodiment, describe as an example of 2 rank Taylor's look-up table index example.
First, the criterion of the error minimum between signal and forward signal based on feedback signal after predistortion model, asks for the predistortion multinomial coefficient of predistortion model in native system; According to this predistortion multinomial, obtain integer searching table, 1 rank Taylor's decimal look-up table and 2 rank Taylor's decimal look-up tables.As shown in Figure 5, LUT represents integer searching table, LUT
(1)represent 1 rank Taylor's decimal look-up table, LUT
(2)represent 2 rank Taylor's decimal look-up tables, for example: LUT (35) represents the predistortion value of index value 35 correspondences in integer searching table, LUT
(1)(35) the predistortion value of index value 35 correspondences in expression 1 rank Taylor's decimal look-up table, LUT
(2)(35) the predistortion value of index value 35 correspondences in expression 2 rank Taylor's decimal look-up tables.
Then, calculate forward signal performance number, adopt 2 rank Taylor's formulas (being formula (3)) to calculate predistortion value corresponding to forward signal performance number:
If forward signal performance number is 36.6, there is x
k-1=36, substitution formula (3) has:
Finally, in order to ensure that forward signal is corresponding one by one with the predistortion value indexing, by after forward signal time delay certain hour, multiply each other and obtain pre-distorted signals with the predistortion value of trying to achieve.Wherein, delay value is determined according to the actual conditions of specific implementation link.
Fig. 6 is the block diagram of 2 rank Taylor's look-up table index prioritization schemes in digital pre-distortion processing method embodiment tetra-provided by the invention.The present embodiment is prioritization scheme embodiment illustrated in fig. 5.In Fig. 6, illustrate by formula (8) and obtained integer searching table LUT, obtained 1 rank Taylor's decimal look-up table LUT by formula (9)
(1), obtain 2 rank Taylor's decimal look-up table LUT by formula (10)
(2).
The concrete optimization part of the present embodiment comprises:
(1) calculating according to Taylor's formula in the process of predistortion value corresponding to forward signal performance number, adopt cut position method to realize division arithmetic, for example: 1/2! This division arithmetic can be realized by cut position method, can improve like this processing speed, saves the processing time;
(2), for the required square operation of 2 rank Taylor's index, the multiplier consuming in order to save square operation, can be decided to be 10 by fractional part precision
-1, like this, can calculate in advance the square value of decimal, by the square value of decimal (from 0.1
2to 0.9
2totally 9 square values) store in tables of data.When calculate predistortion value corresponding to forward signal performance number according to Taylor's formula, obtain square operation result required in computational process by data query table.Multiplier can be saved like this, processing speed can also be improved.For the required n power value of high-order Taylor index more, also can adopt similar approach.
Fig. 7 is the structural representation of digital predistortion process apparatus one embodiment provided by the invention.As shown in Figure 7, the present embodiment comprises: multinomial acquisition module 11, look-up table acquisition module 12, computing module 13 and processing module 14, wherein:
Multinomial acquisition module 11, for according to predistortion model, is asked for the predistortion multinomial coefficient of predistortion model;
Look-up table acquisition module 12, for according to predistortion multinomial, obtains integer searching table and N decimal look-up table, and this N decimal look-up table comprises that 1 rank Taylor's decimal look-up table is to N rank Taylor's decimal look-up table; Wherein, N is more than or equal to 1;
Computing module 13, for index out respectively the predistortion value of integer searching table and N decimal look-up table according to forward signal performance number, calculates predistortion value corresponding to forward signal performance number according to Taylor's formula;
Processing module 14, for according to forward signal and predistortion value corresponding to forward signal performance number, obtains pre-distorted signals.
Further, multinomial acquisition module 11 can, specifically for the criterion of the error minimum between signal and the forward signal after predistortion model based on feedback signal, be asked for predistortion multinomial coefficient.
The present embodiment can also comprise: memory module 15, for calculating the n power value of decimal, stores the n power value of decimal in tables of data into; Computing module 13 can calculate predistortion value corresponding to forward signal performance number, the parameter that obtains using in computational process by data query table specifically for working as according to Taylor's formula.
Further, computing module 13 can, specifically for calculating according to Taylor's formula in the process of predistortion value corresponding to forward signal performance number, adopt cut position method to realize division arithmetic.
Processing module 14 can be specifically for by after forward signal time delay certain hour, and the predistortion value corresponding with forward signal performance number multiplies each other and obtain pre-distorted signals.
In digital pre-distortion processing method provided by the invention and device, forward signal performance number is divided into integer and decimal look-up table corresponding to index respectively, carry out matching according to Taylor's formula and approach the true predistortion value that forward signal is corresponding, can obviously improve index precision.First can improve look-up table index precision by linear interpolation, and then adopt Taylor's decimal look-up table index, further improve look-up table index precision.Show by emulation, in the time that Taylor series meet certain condition, in the span of any forward signal performance number, error is all less than the index error of prior art, has promoted the performance of digital pre-distortion system; In addition, the resource that this scheme is brought increases also little, and owing to adopting the prioritization schemes such as fast algorithm, can not reduce the ageing of this scheme processing.The present invention has realized and has improved digital pre-distortion system stability, improves the object of digital pre-distortion performance.
Finally; it should be noted that: what enumerate above is only specific embodiments of the invention; certainly those skilled in the art can change and modification the present invention; if these amendments and modification all should be thought protection scope of the present invention within belonging to the scope of the claims in the present invention and equivalent technologies thereof.