CN103106062B - The conforming method of calibrating optical Vector-Matrix Multiplier laser via - Google Patents

Abstract

The invention provides a kind of conforming method of calibrating optical Vector-Matrix Multiplier laser via.The parameter of other laser vias is demarcated and proofreaded to the method by a reference laser path, thus solve each laser via and output intensity inconsistency of corresponding to identical input vector of causing inconsistent due to characterisitic parameter.

Description

The conforming method of calibrating optical Vector-Matrix Multiplier laser via

Technical field

The present invention relates to optical information processing and optical computing technology field, particularly relate to a kind of conforming method of calibrating optical Vector-Matrix Multiplier laser via.

Background technology

Along with the development of science and technology, the raising of people's living standard, information to be processed gets more and more, and there has also been higher requirement and standard to real-time, accuracy.Since First computing machine is born, people constantly pursue the lifting of calculated performance, new technological level continues to bring out, new computer hardware structure is also constantly suggested, such as be specifically designed to the special DSP of digital signal processing, custom integrated circuit and programmable logic device (PLD) (FPGA), but this still can not meet the needs of current application fast development far away.

Due to the limitation of electric information processing in bandwidth, crosstalk, delay etc., sight is put into optical information processing and optical oomputing field by increasing researchist.At " OpticalComputing " (Chapter4.3, MITpress1988) in, propose the optical vector-matrix multiplier that one is referred to as " StanfordopticalVMM ", this structure can realize the parallel processing of signal, substantially increases the speed of signal transacting.The optical module of this structure is widely used in the middle of the research of various optical computer.At " High-speedandlow-powerelectro-opticalDSPcoprocessor " (JOSAA, Vol.26, Issue8, pp.A11-A20 (2009)) in, more detailed description has been done to the structure and application that adopt the light digital signal processor of optical vector-matrix multiplier, and the electricity DSP the fastest with present speed compares, result shows the operational performance of the operational performance of optical processor considerably beyond current electricity processor.At " Acoustoopticalmatrix-vectorproductprocessor:implementati onissues " (APPLIEDOPTICS, Vol.28, No.18) in, the detail that labor vector-matrix multiply musical instruments used in a Buddhist or Taoist mass realizes, to contrast detailed assessment and analysis on the parameter of the photoelectricity comprising modules related in system and the impact of these parameters on whole system performance.

But, realizing in process of the present invention, applicant finds existing optical vector-matrix multiplier, and there are the following problems: because the deficiency of device technology and the characteristic of device itself make the laser instrument required for optical vector-matrix multiplier work, detector and modulator matrices can introduce many performance inconsistencies when work, the power that different laser via exports same input vector is not identical, thus has influence on stability and the accuracy of whole optical vector-matrix multiplier operational performance.

Summary of the invention

(1) technical matters that will solve

For solving above-mentioned one or more problems, the invention provides a kind of method obtaining optical vector-matrix multiplier correction coefficient.

(2) technical scheme

According to an aspect of the present invention, a kind of conforming method of calibrating optical Vector-Matrix Multiplier laser via is provided.This optical vector-matrix multiplier comprises: the laser via on micro treatment module and the some roads by its control.The method is performed by micro treatment module, comprising: before actual operation, obtains the correction coefficient k of each laser via basis of reference laser via _[n]and b _[n]; And in actual operation process, by the input vector V correction coefficient k of present laser path _[n]and b _[n]be modified to V '=k _[n]v+b _[n].

(3) beneficial effect

As can be seen from technique scheme, the conforming method of calibrating optical Vector-Matrix Multiplier laser via of the present invention has following beneficial effect:

(1) each path laser instrument and problem of corresponding to the output intensity inconsistency of identical input vector of causing inconsistent due to characterisitic parameter is solved, realize the consistance of multi-path light source power, the stability of operational performance and accuracy improving Vector-Matrix Multiplier is had great significance;

(2) under the prerequisite not changing hardware, rely on control inputs electric current to realize the consistance of multi-path light source power, existing hardware module in multiplexing former optical vector-matrix multiplier, has saved cost and has reduced system bulk.

Accompanying drawing explanation

Fig. 1 for according to embodiment of the present invention calibrating optical Vector-Matrix Multiplier laser via coherence method based on the structural schematic block diagram of system;

Fig. 2 is the schematic diagram according to correcting laser via in embodiment of the present invention calibrating optical Vector-Matrix Multiplier laser via coherence method;

Fig. 3 is the process flow diagram according to obtaining correction coefficient step in embodiment of the present invention calibrating optical Vector-Matrix Multiplier laser via coherence method;

Fig. 4 is the sub-step process flow diagram according to obtaining sub-step D in correction coefficient step in embodiment of the present invention calibrating optical Vector-Matrix Multiplier laser via coherence method;

Fig. 5 is the process flow diagram step by step according to obtaining D1 step by step in correction coefficient step in embodiment of the present invention calibrating optical Vector-Matrix Multiplier laser via coherence method;

Fig. 6 is the structural schematic block diagram of system shown in Figure 1 under actual scene.

Embodiment

For making the object, technical solutions and advantages of the present invention clearly understand, below in conjunction with specific embodiment, and with reference to accompanying drawing, the present invention is described in more detail.It should be noted that, in accompanying drawing or instructions describe, similar or identical part all uses identical figure number.The implementation not illustrating in accompanying drawing or describe is form known to a person of ordinary skill in the art in art.

The present invention, in order to solve the problem of inconsistency of multipass laser output intensity, proposes a kind of conforming method of calibrating optical Vector-Matrix Multiplier laser via, to improve the computing consistance of each path in optical vector-matrix multiplier.

The present invention is based on an optical vector-matrix multiplier.This optical vector-matrix multiplier comprises the laser via on FPGA module and the some roads by its control.Wherein, each laser via comprises: D/A conversion unit, laser element, detector cells and AD conversion unit.Multiple D/A conversion unit can be discrete, also can be integrated D/A converter module on the same chip as shown in Figure 1.In like manner, multiple laser element can be discrete, also can be integrated laser array on the same chip as shown in Figure 1.Multiple detector cells can be discrete, also can be the integrated detector array on the same chip shown in Fig. 1.Multiple AD conversion unit can be discrete, also can be integrated analog-to-digital conversion module on the same chip as shown in Figure 1.

In this optical vector-matrix multiplier, digital vectors is sent into a laser via by FPGA module; In this laser via, digital to analog converter exports the electric signal drive laser unit corresponding to this digital vectors and launches the light intensity signal being loaded with digital vectors information, intensity signal is detected by detector cells, intensity signal is converted to digital quantity by AD conversion unit, and feed back input is to FPGA module again.

As shown in Figure 1, in optical vector-matrix multiplier except the above-mentioned modules mentioned, also comprise microprocessor module (MCU module), data storage cell, clock module, power module and temperature control module.Wherein, MCU module is for coordinating sequential relationship between modules and opening sequence.Data memory module is for being temporarily stored in the intermediate parameters corrected in consistance process.Clock module is used for controlling to keep synchronism between each module of whole system.Power module is used for providing operating voltage for unit and module.Temperature control module is used for adding monitoring temperature and control section for each road laser instrument, and Shi Mei road laser works at a constant temperature, ensures that the linear output character curve of single channel laser instrument is stablized, for interchannel linear Concordance is laid a good foundation.

It should be noted that, in the optical vector-matrix multiplier shown in Fig. 1, the FPGA module of employing controls.Those skilled in the art also can adopt other microprocessor to realize this function, such as: single-chip microcomputer, dsp chip etc.These modules all should comprise its more peripheral necessary module, such as memory modules etc.In the present invention, the above-mentioned module with control and data processing function can be referred to as micro treatment module.

In one exemplary embodiment of the present invention, provide a kind of conforming method of calibrating optical Vector-Matrix Multiplier laser via.The method comprises based on the optical vector-matrix multiplier shown in Fig. 1:

Step S100, before actual operation, obtains the correction coefficient k of each laser via basis of reference laser via _[n]and b _[n]; And

Step S200, in actual operation process, by the input vector V correction coefficient k of present laser path _[n]and b _[n]be modified to V '=k _[n]v+b _[n], as shown in Figure 2.

Fig. 3 is the process flow diagram according to obtaining correction coefficient step in embodiment of the present invention calibrating optical Vector-Matrix Multiplier laser via coherence method.As shown in Figure 3, above-mentioned steps S100 comprises:

Steps A, opens reference laser path Ch ₁, close except this reference laser path Ch ₁other outer laser vias;

Step B, FPGA module is respectively to the D/A conversion unit input digital vectors D in this reference laser path ₁₁and D ₁₂, receive and store the digital quantity P fed back by the AD conversion unit in this reference laser path ₁₁and P ₁₂;

Step C, opens laser channeling Ch to be corrected _n(n=2,3 ..., 16), close except this laser channeling Ch to be corrected _nother outer laser vias;

Step D, FPGA module obtains respectively when the digital quantity of AD conversion unit feedback in this laser via to be corrected is closest to P ₁₁and P ₁₂time, the digital vectors D of digital to analog converter input in this laser via to be corrected _n1and D _n2;

This step realizes in two steps, and in step D1, FPGA module obtains when the digital quantity of laser via feedback to be corrected is the most close to P ₁₁time, to the digital vectors D of D/A conversion unit input _n1; In step d 2, FPGA module obtains when the digital quantity of path feedback to be corrected is the most close to P ₁₂time, to the digital vectors D of D/A conversion unit input _n2.The executive mode of above-mentioned two steps is identical.Hereafter for step D1, concrete implementation is described.As shown in Figure 4, this step D1 is specifically divided into following sub-step:

Sub-step D1a, initialize digital vector D _n1=D ₁₁, initialization Flag1=Flag2=0;

In this sub-step, Flag1 and Flag2 be in order to do not understand under the error condition caused by AD conversion unit and D/A conversion unit precision introduce.Under known above-mentioned error condition, under can passing through rationally to arrange α situation, this Flag1 and Flag2 is not necessary yet.

Sub-step D1b, exports digital vectors D _n1, receive analog-to-digital conversion module and be converted to digital quantity P _n1;

Sub-step D1c, when | P _n1-P ₁₁| during > α, adopt the method for Step wise approximation to change digital vectors D _n1, re-execute step D1b; Otherwise, jump out circulation, obtain digital vectors D _n1and D _n2, wherein α is one larger between default tolerable error maximal value and the max value of error that caused by AD conversion unit/D/A conversion unit resolution limit.

As shown in Figure 5, this sub-step D1c can be divided into again as follows step by step:

D1c1 step by step, judges P _n1-P ₁₁whether be greater than ε, if so, enter step D1c2, otherwise enter step D1c4, wherein ε is default tolerable error maximal value, generally, and 1% of the output minimum power of ε set by laser via duty;

D1c2 step by step, judgement symbol variable Flag1 & Flag2=0, if so, performs step D1c3;

Otherwise, perform step D1c7;

If P _n1-P ₁₁> ε, then illustrate input vector D _n1should reduce, but now should first to indexed variable Flag1, Flag2 judges.If there is a variable to be 0, illustrate, program has only carried out increasing or decreasing operation, does not still make P _n1and P ₁₁difference minimum, enter step D1e.Flag1, Flag2 are non-zero else if, illustrate to input vector D _n1carry out successively decreasing and increment operation, but due to the lack of resolution of D/A converter, so can not P be made _n1and P ₁₁completely equal, but be under the D/A converter resolution adopted at optical vector-matrix multiplier, the most accurate result that can reach, therefore then enters step D1c7 for avoiding entering endless loop. this moment

D1c3 step by step, to vectorial D _n1once to successively decrease adjustment, and indexed variable Flag1 is carried out from add operation, re-execute sub-step D1b;

Indexed variable Flag1 carries out from add operation, shows to input vector D ₁₁carried out a decrement operations, thus be convenient to judge in subsequent process.Herein, indexed variable is not limited to carry out from add operation, if can make Flag1 be not now 0 any operation all can reach identical object.

D1c4 step by step, judges P _n1-P ₁₁whether be less than ε, if so, perform step D1c5, otherwise, perform step D1c7;

D1c5 step by step, judgement symbol variable Flag1 & Flag2=0, if so, performs step D1c6; Otherwise, perform step D1c7;

In like manner, if P _n1-P ₁₁< ε, then illustrate input vector D _n1should increase, but now should first to the indexed variable Flag1 that step C is arranged, Flag2 judges, if there is a variable to be 0, illustrate, program has only carried out increasing or decreasing operation, does not still make P _n1and P ₁₁the absolute value of difference minimum, enter step D09.Flag1, Flag2 are non-zero else if, illustrate to input vector D _n1carry out successively decreasing and increment operation, but due to the lack of resolution of D/A converter, so can not P be made _n1and P ₁₁completely equal, but be under the D/A converter resolution adopted at optical vector-matrix multiplier, the most accurate result that can reach, therefore then enters step D1c7 for avoiding entering endless loop. this moment

D1c6 step by step, to vectorial D _n1carry out an incremental, and indexed variable Flag2 is carried out from add operation, re-execute sub-step D1b;

D1c7 step by step, corrects the first time of correction channel and terminates, store current D _n1;

It should be noted that, in this step, vector to be increased progressively and the step-length of decrement operations is determined according to the resolution of current laser power and selected analog-digital chip.Generally, this increases progressively identical with the step-length of successively decreasing, and in order to improve the precision of system compensation, often all selects the least significant bit (LSB) of analog-digital chip as the step-length increased progressively and successively decrease.For example: when selected digital to analog converter is 10bit resolution time, so digital quantity step-length is just 1/1023, and corresponding output current variable quantity can be arranged according to the characteristic of laser instrument and current margin.

The concrete implementation of step D2 and each sub-steps of step D1 similar, no longer repeated description herein.

Step e, according to following equation, FPGA module calculates and stores correction coefficient k _[n]and b _[n].

$\left\{\begin{array}{c}{k}_{\left[n\right]}{D}_{11}+b_{\left[n\right]}=D_{n1}\\ k_{\left[n\right]}{D}_{12}+b_{\left[n\right]}=D_{n2}\end{array}\right.$ (formula 1)

Step F, judges whether that all laser instrument paths all correct complete, if so, then terminates, otherwise, re-execute step C.

So far, the present embodiment is introduced complete.

In order to those skilled in the art can better understand technical scheme of the present invention and produce the reason of beneficial effect, special below following elaboration is carried out to theoretical foundation of the present invention:

Laser instrument is current drive-type device, after threshold current is exceeded to its input current, input current and output intensity linear, and the input vector of D/A converter and output current also linear, temperature is often embodied in the impact of laser instrument and changes the P-I slope of laser instrument, therefore can think that whole data load system is a linear system.The output intensity of laser instrument and the relation of input current as follows:

L _[n]=η _[n]* (I _[n]-I _{0 [n]}) (formula 2)

Wherein L _[n]represent the output intensity of the n-th road laser instrument, η _[n]represent the current-power conversion efficiency of the n-th road laser instrument, I _[n]represent the input current value of the n-th road laser instrument, I _{0 [n]}represent the threshold current of the n-th road laser instrument.When temperature changes, the output power of laser instrument can be expressed as:

L _[n]=a _[n]* η _[n]* (I _[n]-(I _{0 [n]}+ I _{t [n]})) (formula 3)

Wherein, a _[n]represent the influence coefficient that temperature is brought current-power conversion ratio, η _[n]represent the current-power conversion efficiency of expression n-th road laser instrument, I _[n]represent the input current value of the n-th road laser instrument, I _{0 [n]}represent the threshold current of the n-th road laser instrument, I _{t [n]}represent the skew of the laser threshold current that temperature causes.Formula (2) is launched, according to the linear transformation relation between the output current of D/A converter and the digital signal of input, following equivalent formulation can be obtained:

L _[n]=k _[n]* V+b _[n](formula 4)

Wherein: k _[n]=a _[n]* η _[n]* m (m represents the conversion coefficient of D/A converter) b _[n]=-(a _[n]* η _[n]* m*I _{0 [n]}+ a _[n]* η _[n]* I _{t [n]}), V represents that being input to DA digital to analog converter is used for the digital vectors value of drive laser.We can draw thus, when arranging laser works when linear work is interval, the factor (threshold current, current-power conversion ratio) affecting laser works curve can be unified two parameters for linear function.Each road laser instrument within specified temperatures, k _[n]and b _[n]be all fixing, we can show that the output intensity of laser instrument and input vector have stable linear relationship, and are independently mutually between path.Based on above theoretical analysis, we can learn to reach interchannel linear consistance, guarantee that identical input vector can export identical light intensity, take proposed a kind of optical vector-matrix multiplier to correct conforming method and calculate every path laser relative in the correction coefficient of reference channel, the vectorial V of input is revised (Fig. 3 adds the light-source system operating diagram after correction module), guaranteeing to have between path has unified vector-light intensity mapping relations, can reach the beneficial effect of expection.

In order to those skilled in the art can implement the present invention smoothly, below by the present invention be described in detail based on all parts of optical vector-matrix multiplier.

Array of source 1

Array of source 1 provides light source for whole system, can select LED light emitter, VSCEL laser instrument, select corresponding way according to the system requirements of vector matrix-multiplier.

As shown in Figure 6, in actual scene, have selected the 6001A series Distributed Feedback Laser of Xeston company, this laser works is at 1310nm wave band, and high frequency response is good, and carries integrated TEC and thermistance element, peak power output reaches 31mW, meets the demand of system.

The drive part of light source is the MAX3669 laser driving chip of a FPGA, Maxim company of XC5VSX50T by the Virtex serial model No. of Xilinx company, MAX5182DA converter and temperature-control circuit form jointly.Signal flow is such, vector matrix operand takes out by Microprocessor S3C44B0X FPGA from storer, by the D/A converter control circuit in FPGA operand is input in MAX5182 and produces corresponding current signal, current signal Direct driver laser instrument makes it export to represent the laser intensity of operand, then utilizes optical vector matrix multiplier to carry out multiply-add operation.Because the work of laser instrument, need to provide the threshold current that stable, the laser driving chip Max3669 that have employed Maxim company provides high-precision threshold current for laser instrument.Before system intersection, we measure the characteristic of laser instrument, obtain the current versus output power curve of two-way various lasers, as can be seen from curve we, after the drive current of laser instrument is greater than threshold current, laser instrument is a linear convergent rate device, but the current versus output power curve of various lasers is inconsistent, therefore must carry out calibration operation to each road laser instrument before calculating.

The Virtex5 Series FPGA of the Xilinx company adopted in system and the MCU based on ARM9 kernel of Samsung can also replace with the product of other semiconductor manufacturer's same kind.

Detector array 2

The operation result of detector array 2 pairs of Vector-Matrix Multipliers carries out opto-electronic conversion, because native system uses the extra system cycle to carry out demarcation and the self-correcting of system, thus correct the consistance cycle can the detector array of multiplex optical Vector-Matrix Multiplier.

As shown in Figure 6, detector array has selected the G7150 series 16 road detector arrays of Japanese Bin Song company, the light intensity exported after Vector-Matrix Multiplier computing is input to ADC modulus conversion chip ADS5474 after one across resistance amplification, be converted to corresponding digital signal by ADS5474, be stored into and output to TFT display in the SDRAM of microprocessor S3C2440 and by result and show.Because native system is the demarcation and the self-correcting that use the extra system cycle to carry out system, so can this photoelectronic detecting array multiplexing in the correction consistance cycle, supposing the system be operated in self-calibration and self-correcting cycle then microprocessor can calculate accordingly the result of ADS5474 conversion, calculate the modifying factor of corresponding way, and regulate the input digital quantity of digital to analog converter Max5812 to be fed back on the input current of laser instrument, to keep consistance and the accuracy of the luminous power of every road laser instrument by this modifying factor.

AD and D/A conversion module

D/A conversion module is used for being that the electric current input quantity drive laser of laser instrument sends the multiply-add operation that light intensity carries out vector matrix by arithmetic operation number conversion, and AD conversion module is used for the simulation trial results conversion that detector obtains to become digital quantity storages to show or outputs to microprocessor and carries out self-correcting and self-calibration operates.

As shown in Figure 6, in actual scene, in this module, D/A converter have employed the max5182 of Maxim company, the current-output type DAC of this device to be a chip resolution be 10bit, by the DA control interface realized in FPGA, revised computing vector is converted to current signal, drive laser.What AD converter adopted is the ADS5474 modulus conversion chip of TI company, and the sampling rate of this chip is 400MSPS.

Temperature control unit

Temperature control unit obtains the operational temperature information of current lasers by heat sensitive sensor and compares with the temperature value preset, then both differences are sent and processed by PID circuit, export via PID control circuit and control electric current and drive TEC module to heat laser instrument or freeze, ensure that the stability of laser. operating temperature.

Generally, temperature control unit 4 is made up of jointly heat sensitive sensor, TEC (semiconductor refrigerating unit) and PID backfeed loop.This temperature control unit both can be realized by discrete device, and commercial modularized circuit chip also can be adopted to build.

As shown in Figure 6, in actual scene, in the present embodiment, this part has mainly used a laser instrument Special temp controlling chip MAX1978 of Maxim company, and the temperature of device, by inner integrated integral differential circuit, can control at 0.001 DEG C by this chip accurately.Thermistor in addition needed for temperature-controlled portion and semiconductor cooling device TEC have all been integrated in the inside of laser instrument.By arranging the input resistance of the corresponding pin of MAX1978 before system electrification, presetting the temperature of good berth, in the process of system works, just can be detected the working temperature of current lasers by the resistance or voltage of testing thermistor two ends.

The modularization chip Max1978 of the Maxim company that temperature control modules adopts can also complete PID control form by discrete device designed, designed

Microprocessor control module

Microcontrol processor control module 5: mainly comprise FPGA and MCU, wherein FPGA is responsible for controlling mould/number, D/A switch unit, operand is input to D/A from storer with drive laser output intensity, MCU is responsible in correction and scalar period, the result obtain every road detector to be calibrated and reference value carry out difference operation, and calculate the modified value of every road laser instrument link, be stored in predetermined address.FPGA with MCU realizes mutual communicating by the mode of shared drive.

As shown in Figure 6, in actual scene, microprocessor control module is made up of jointly the FPGA of a Virtex5 series based on ARM9 kernel S3C2440 and Xilinx company of Samsung, self-calibration algorithm in this paper mainly realizes in this microprocessor, and adopts the mode of shared drive to realize communication and exchanges data between S3C2440 and FPGA.

Power module

As shown in Figure 6, in actual scene, power module adopts PTH12030, PTH05000 and PTH08080 system of being respectively of TI company to provide the voltage of 12V, 5V, 3.3V.

Above in each module, FPGA is connected with MAX5182 by digital-to-analog conversion control unit interface, and this control unit interface comprises a clock line, and a heel piece selects signal wire, and a sample enable signal and 10 data lines realize input and the control of digital signal amount.FPGA is connected with microprocessor S3C2440 by data address bus, the operational data in Flash is outputted in MAX5182.Current signal after transformation is connected to the modulating current input end of laser instrument by MAX5182, controls the power of laser instrument output current.

It should be noted that, the definition of the above-mentioned implementation to each element is not limited in various concrete structures, shape or the implementation mentioned in embodiment, and those of ordinary skill in the art can replace it with knowing simply, such as:

(1) detector array can select multi-path detector integrated on sheet, and the mode be coupled by the discrete photoelectric detector of multichannel and multi-path fiber array also can be selected to build;

(2) for digital-to-analogue and analog-to-digital conversion module, both can select multi-path analog-digital chip, also can select discrete but path analog-digital chip, so that the expansion of system;

In sum, the present invention obtains the method for optical vector-matrix multiplier correction coefficient, the parameter of other laser vias is demarcated and proofreads by a reference passageway, thus solve each laser via and output intensity inconsistency of corresponding to identical input vector of causing inconsistent due to characterisitic parameter, the stability of operational performance and accuracy improving Vector-Matrix Multiplier is had great significance.Meanwhile, rely on control inputs electric current to realize the consistance of multi-path light source power, existing hardware module in multiplexing former Vector-Matrix Multiplier, has saved cost and has reduced system bulk.

Above-described specific embodiment; object of the present invention, technical scheme and beneficial effect are further described; be understood that; the foregoing is only specific embodiments of the invention; be not limited to the present invention; within the spirit and principles in the present invention all, any amendment made, equivalent replacement, improvement etc., all should be included within protection scope of the present invention.

Claims (7)

1. the conforming method of calibrating optical Vector-Matrix Multiplier laser via, is characterized in that, this optical vector-matrix multiplier comprises: micro treatment module and the some laser vias controlled by it; The method is performed by micro treatment module, comprising:

Before actual operation, obtain the correction coefficient k of each laser via basis of reference laser via _[n]and b _[n]; Comprise:

Digital vectors D is inputted respectively in reference laser path ₁₁and D ₁₂, receive and store the digital quantity P fed back by this reference laser path ₁₁and P ₁₂;

The digital quantity fed back when laser via to be corrected is closest to P ₁₁and P ₁₂time, obtain the digital vectors D to this laser via input to be corrected respectively _n1and D _n2;

According to following equation, by digital vectors D ₁₁, D ₁₂, D _n1and D _n2calculate the correction coefficient k of this laser via to be corrected _[n]and b _[n], wherein, described correction coefficient k _[n]and b _[n]for in described optical vector-matrix multiplier calculating process, its input vector V is modified to V'=k _[n]v+b _[n], to keep the output-consistence of itself and reference laser path:

$\left\{\begin{array}{c}{k}_{\&lsqb;n\&rsqb;}{D}_{11}+b_{\&lsqb;n\&rsqb;}=D_{n1}\\ k_{\&lsqb;n\&rsqb;}{D}_{12}+b_{\&lsqb;n\&rsqb;}=D_{n2}\end{array}\right.$

And, in actual operation process, by the described correction coefficient k of input vector V of present laser path _[n]and b _[n]be modified to V'=k _[n]v+b _[n].

2. method according to claim 1, is characterized in that, the described digital quantity when laser via feedback to be corrected is closest to P ₁₁time, obtain the digital vectors D to this laser via input to be corrected _n1step comprise:

Step D1a, initialize digital vector D _n1=D ₁₁;

Step D1b, exports digital vectors D _n1, receive analog-to-digital conversion module and be converted to digital quantity P _n1;

Step D1c, as ︱ P _n1-P ₁₁during ︱ > α, the method for Step wise approximation is adopted to change digital vectors D _n1, re-execute step D1b; Otherwise, jump out circulation, obtain digital vectors D _n1and D _n2, wherein α is one larger between default tolerable error maximal value and the max value of error that caused by AD conversion unit, D/A conversion unit resolution limit.

3. method according to claim 2, is characterized in that, described step D1a also comprises: initialization flag amount Flag1=Flag2=0; Described step D1c comprises:

D1c1 step by step, judges P _n1-P ₁₁whether be greater than ε, if so, enter step D1c2, otherwise enter step D1c4, wherein ε is default tolerable error maximal value, is set to 1% of the output minimum power set by laser via duty;

D1c2 step by step, judgement symbol variable Flag1 & Flag2=0, if so, performs step D1c3; Otherwise, perform step D1c7;

D1c3 step by step, to vectorial D _n1once to successively decrease adjustment, make indexed variable Flag1 be not 0, re-execute sub-step D1b;

D1c4 step by step, judges P _n1-P ₁₁whether be less than ε, if so, perform step D1c5, otherwise, perform step D1c7;

D1c5 step by step, judgement symbol variable Flag1 & Flag2=0, if so, performs step D1c6; Otherwise, perform step D1c7;

D1c6 step by step, to vectorial D _n1carry out an incremental, make indexed variable Flag2 be not 0, re-execute step D1b;

D1c7 step by step, terminates the correction of correction channel, stores current D _n1.

4. method according to claim 3, is characterized in that, the step-length of described increasing or decreasing is identical, all gets the minimum resolution of corresponding AD conversion unit, D/A conversion unit.

5. method according to claim 3, is characterized in that, in described D1c3 step by step and step D1c6, makes indexed variable Flag1 or Flag2 not be 0 be operating as from add operation.

6. method according to any one of claim 1 to 5, is characterized in that, each laser via comprises: D/A conversion unit, laser element, detector cells and AD conversion unit;

In the process of described optical vector-matrix multiplier computing, digital vectors is sent into a laser via by micro treatment module; In this laser via, digital to analog converter exports the electric signal drive laser unit corresponding to this digital vectors and launches the light intensity signal being loaded with digital vectors information, intensity signal is detected by detector cells, intensity signal is converted to digital quantity by AD conversion unit, and feed back input is to micro treatment module again.

7. method according to claim 6, is characterized in that, described micro treatment module is FPGA module, single-chip microcomputer or DSP module;

In described optical vector-matrix multiplier: multiple laser element, multiple detector cells, multiple AD conversion unit and multiple D/A conversion unit are discrete unit or are integrated in array on same chip.