A method for automatic testing of dynamic range of APD detector

Abstract

The application discloses a kind of dynamic range automatic test methods for APD detector, it includes: step one: reading configuration parameter, driving instrument instrument works and judging state;Step two, fitting power level straight line, obtains fitting parameter;Step three: carry out test point position registration;Step four: drive source table, light source works;Step five: collect minimum received light power Pmin;Step six: drive oscilloscope to read output signal and carry out pre-processing;Step seven: data is handled by convolutional neural network model analyzer and updates light attenuation parameter;Step eight: calculate the dynamic range Gd of current test point;Step nine: repeat step three to step eight until the dynamic range test of each test point of APD detector is completed.The application uses host computer program to combine deep convolutional neural network to assist dynamic range test, effectively improves test efficiency and precision.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of photodiode testing, in particular to a dynamic range automatic testing method for an APD detector. BACKGROUND

[0002] A higher reverse bias is applied to the APD (avalanche photodiode) to excite more carriers in the heterojunction built-in electric field, which greatly improves the light signal detection capability by several orders of magnitude. The high responsivity and high signal-to-noise ratio characteristics of the APD detector make up for the disadvantages of traditional photodiode detection of weak signals, enabling the deployment of modern detection equipment with high precision, high integration and high anti-interference. APD detectors are widely used in radar detection, three-dimensional sensing, data communication and instrument and metering.

[0003] Due to the manufacturing technology of APD detector design, preparation, packaging, etc., it is impossible to avoid the large discreteness of the finished photoelectric parameters. Accurate testing of the photoelectric parameters of the APD detector not only provides reliable feedback for the preparation conditions, but also provides basic performance evaluation for high-precision systems with APD detectors as the core. The test system for APD detectors involves instruments and equipment in key technical fields of optoelectronics, electromechanics and computer science. The system is complex, the debugging operation is tedious, and the manual testing precision is low, which has become a major factor restricting the efficient and accurate testing of APD detectors. The core of ATE (automatic test equipment) integrates the specific method of photoelectric parameter testing. The host computer program drives the instruments and meters through USB, RS232, GPIB and other communication interfaces to realize batch testing of the detector, reducing the complexity of debugging operation and improving testing efficiency and accuracy.

[0004] The dynamic range of the APD detector reflects the ability of the detector to detect the strongest and weakest light signals. The test system needs to provide the working voltage and current excitation of the detector, the input light signal after attenuation, and the amplitude of the APD detector output signal detected by the oscilloscope. How to drive the system modules to realize automatic testing of the dynamic range through the automatic testing control logic in the host computer program is a key technical problem for the ATE of the APD test system. In addition, the dynamic range test needs to accurately detect the saturation optical power of the APD detector. The current testing method is as follows: the light attenuation value corresponding to the saturation optical power of the standard APD detector is taken as a reference, a certain range of light attenuation values is selected, the attenuation value is changed in a fixed step, and the amplitude of the APD detector output signal is read by the oscilloscope at the same time, until the output signal amplitude no longer changes with the change of the light power, that is, the saturation optical power is measured. The problems of this method are: 1) the saturation state of the APD detector output signal needs to be judged manually; 2) when the light attenuation range increases, the number of tests increases and the time consumption is long. The limitations of the current testing method reduce the testing efficiency and accuracy, so there is an urgent need for a high-efficiency and high-precision dynamic range automatic testing method. SUMMARY

[0005] The present application aims to overcome the deficiencies of the prior art, and provides a dynamic range automatic testing method for an APD detector, which is driven by a host computer program to reduce the complexity of debugging operation, has high testing efficiency and precision, and is conducive to the deployment of an APD detector automatic testing system.

[0006] To solve the above technical problems, the technical scheme provided by the present application is as follows: a dynamic range automatic testing method for an APD detector, comprising the following steps:

[0007] Step one: reading APD detector test basic configuration parameters, the host computer program transmits the initialization running parameters to the instruments and meters through a control protocol, the host computer program receives the instrument and meter state feedback, judges the working state of each instrument and meter, and if all the instrument and meter states are normal, step two is entered, otherwise the current test state is terminated;

[0008] Step two: driving the pulse laser, the optical attenuator and the optical power meter, respectively collecting the average optical power Pin of the optical power meter close to the pulse laser and the average optical power Pout of the APD detector image surface under different attenuation values multiple times, and respectively converting them into power levels Pin-dBm and Pout-dBm, and after taking Pin-dBm as the independent variable x and Pout-dBm as the dependent variable y, performing linear fitting to obtain the slope k and the intercept b, and saving them to the basic configuration parameters;

[0009] Step three: the host computer program obtains the APD detector image surface image center point and the light source light spot image position information from the infrared CCD camera image data, and drives the XY displacement table to perform test point position registration;

[0010] Step four: the host computer program reads the APD detector bias, the working voltage driving voltage current source table to provide the voltage, reads the light source repetition frequency f, the pulse width w and the attenuation value, and drives the pulse laser, the optical attenuator and the optical power meter to work;

[0011] Step five: the host computer program drives the voltage current source table to provide the APD detector allowable maximum bias voltage; turns off the light source output of the pulse laser, drives the oscilloscope to collect the detector output voltage root mean square Vrms, and obtains the responsivity Rv of the detector, and takes the Vrms / Rv calculation result as the minimum received optical power Pmin;

[0012] Step six: under the premise that the host computer program drives the voltage and current source table to provide the APD detector with the maximum allowed bias voltage and turn on the light source output, the host computer program reads the pre-constrained starting value As and ending value Ae of the optical attenuator, and transmits the median value Ac=(As+Ae) / 2 of the starting value and the ending value to the optical attenuator as the attenuation value to be executed; the host computer program drives the oscilloscope to read the APD detector output signal and pre-processes it;

[0013] Step seven: the pre-processed data is input into the host computer program deep neural network model analyzer; the output result of the host computer program deep neural network model analyzer is matched with the result matrix to update the starting value As and the ending value Ae of the optical attenuator attenuation, and a stop state is returned; according to the return state, the following judgment is made:

[0014] If the stop state is continue, return to step six;

[0015] If the stop state is stop, go to step eight;

[0016] Step eight: the host computer program collects the average optical power value of the optical power meter close to the pulsed laser under the current attenuation value state, and the result after linear correction processing is taken as the saturated optical power Pmax, and the dynamic range is obtained by the formula Gd=10lg(Pmax / Pmin);

[0017] Step nine: the host computer program reads the next test point, repeats steps three to eight to test the next test point, and the host computer program ends the test process after the dynamic range test of each test point of the APD detector is completed.

[0018] Preferably, the read APD detector test basic configuration parameters include the configuration XML file of the following parameter information: the electrical parameter configuration information including the APD detector bias, work, extended attenuation voltage and enable state; the optical parameter configuration information including the light source repetition frequency, pulse width, attenuation value; the detector image plane size, test point coordinates; the pre-calibrated object image pixel equivalent Cp, displacement pixel equivalent Cm; the fixed maximum attenuation value of the attenuator, the starting value and the ending value of the optical attenuator attenuation.

[0019] Preferably, when the APD detector is a single-pixel APD detector, the center point of the pixel is selected as the test point; when the APD detector is a multi-pixel APD detector, the position of each pixel offset from the center point of the image plane by a fixed distance is selected as the test point.

[0020] Preferably, the specific process of driving the XY displacement table to perform test point position registration is as follows:

[0021] The host computer program takes the two-dimensional coordinate point obtained at the center of the profile as the image center point of the APD detector photosensitive surface, and pre-calibrates the two-dimensional coordinates of the circular region containing the center and radius of the infrared CCD camera two-dimensional image as the light source light spot image position information;

[0022] The host computer program calculates the vector between the image center point of the APD detector photosensitive surface and the light source light spot image position information as the pixel distance vector Vp;

[0023] The host computer program reads the photosensitive surface size and test point coordinates and calculates the actual physical distance vector Vl of the test point coordinates offset from the photosensitive surface center;

[0024] The host computer program calculates the displacement vector R=Cm(Cp×Vl+Vp), wherein Cp is the pixel equivalent of the object image, and Cm is the displacement pixel equivalent;

[0025] The host computer program drives the XY displacement table to displace according to the vector R to complete the registration.

[0026] Preferably, the specific method of linear correction processing is:

[0027] The average optical power close to the pulsed laser is taken as the independent variable x to calculate the optical power y after linear correction processing by the formula y=(k×10lg(1000×x)+b) / (f×w).

[0028] Preferably, the specific method of pre-constraint processing is:

[0029] During the process of gradually reducing the attenuation value of the attenuator from the attenuation starting value, the APD detector output signal maintains the same saturated state;

[0030] During the process of gradually increasing the attenuation value of the attenuator from the attenuation ending value, the APD detector output signal maintains the same non-saturated state.

[0031] Preferably, the specific method of pre-processing the APD detector output signal is:

[0032] Sample Fm / Fl data points contained in a laser pulse duration from the oscilloscope waveform data, wherein Fm is the sampling rate of the oscilloscope, and Fl is the repetition frequency of the pulsed laser;

[0033] Deform the resampled waveform data into a two-dimensional matrix; and perform maximum and minimum normalization processing on the two-dimensional matrix.

[0034] Preferably, the host computer program deep neural network model analyzer adopts a convolutional neural network model, including two convolutional layers, two pooling layers with half width and height, a Flatten layer, three fully connected layers, the model input is a waveform data tensor in MHWC format, and the last layer of the model is a two-classification fully connected layer including non-saturated and saturated state values.

[0035] Preferably, the specific method for matching the output results of the host computer program deep neural network model analyzer with the result matrix to update the attenuation start value As and the attenuation end value Ae of the optical attenuator is as follows:

[0036] The host computer program collects the output results of the deep neural network model analyzer, and if the non-saturated state value is greater than the saturated state value, the attenuation end value Ae of the optical attenuator is equal to the current attenuation median value Ac, and the attenuation start value As remains unchanged; if the non-saturated state value is less than the saturated state value, the attenuation start value As of the optical attenuator is equal to the current attenuation median value Ac, and the attenuation end value Ae remains unchanged, so that the attenuation start value As and the attenuation end value Ae of the optical attenuator satisfy the pre-constraint processing in step seven.

[0037] The host computer program judges whether |As-Ae|<2x attenuation resolution of the optical attenuator, and if yes, the returned stop state is stop, otherwise the returned stop state is continue.

[0038] Compared with the prior art, the automatic test method of the present application has the following advantages: the automatic test method of the present application can realize automatic test of the dynamic range of a single or multiple test points of an APD detector, and the test process adopts a mode of driving instruments and meters by a host computer program, executing automatic test logic, and searching for saturated optical power by a deep convolutional neural network, so that the error rate and complexity introduced by manual operation can be excluded by only setting the basic configuration parameters of the APD detector in advance, thereby effectively improving the measurement efficiency and accuracy of the dynamic range parameter test of the APD detector; and the dynamic range of a single-pixel and multi-pixel APD detector can be tested, thereby facilitating subsequent compatibility transplantation on different types of APD detector test systems. BRIEF DESCRIPTION OF DRAWINGS

[0039] Figure 1 The host computer program drives instruments and meters for the present application.

[0040] Figure 2 The dynamic range automatic test method flowchart for the present application.

[0041] Figure 3 The deep neural network model data processing flowchart for the present application. DETAILED DESCRIPTION

[0042] In order to further explain the technical solutions of the present application, the following will be described in detail in combination with the drawings in the embodiments of the present application, but the protection scope of the present application is not limited by this.

[0043] As shown in Figure 1 , Figure 2 , the present embodiment proposes a dynamic range automatic testing method for APD detector, which is based on APD photoelectric parameter testing system, the system includes infrared CCD camera, pulsed laser, optical attenuator, optical power meter, voltage and current source meter, XY displacement table, oscilloscope, the upper computer program automatically tests the control logic to drive instruments and meters, reduces the complexity of manual operation, improves the test speed and accuracy. Including the following steps:

[0044] Step one: read the APD detector 7 test basic configuration parameters, the upper computer program 5 transmits the initialization running parameters to the instruments and meters through the control protocol, the upper computer program receives the instrument and meter state feedback, judges the working state of each instrument and meter, all instrument and meter state is normal, then enters step two, otherwise the current test state is terminated;

[0045] Step two: drive pulsed laser 1, optical attenuator 2, optical power meter 3, respectively multiple times collect the average optical power Pin of optical power meter 3 close to pulsed laser 1 and the average optical power Pout of APD detector 7 image surface under different attenuation values, and respectively convert into power level Pin-dBm, Pout-dBm, after taking Pin-dBm as independent variable x and Pout-dBm as dependent variable y, linear fitting is carried out to obtain slope k and intercept b, and saved to basic configuration parameters;

[0046] Step three: the upper computer program 5 obtains the APD detector 7 image surface image center point and light source light spot image position information from the infrared CCD camera 4 image data, drives the XY displacement table 6 to carry out test point position registration;

[0047] Step four: the upper computer program reads the APD detector 7 bias, working voltage, and drives the voltage and current source meter 8 to provide voltage, reads the light source repetition frequency f, pulse width w, attenuation value and drives the pulsed laser 1, optical attenuator 2, optical power meter 3 to work, and then enters step five;

[0048] Step five: the upper computer program 5 first drives the voltage and current source meter 8 to provide the maximum bias voltage allowed by the APD detector 7; then turn off the light source output of the pulsed laser 1, drive the oscilloscope 9 to collect the root mean square voltage Vrms of the APD detector 7 output voltage, obtain the responsivity Rv of the APD detector 7, and take Vrms / Rv calculation result as the minimum received optical power Pmin;

[0049] Step six: the host computer program 5 first drives the voltage and current source table 8 to provide the APD detector 7 with the maximum allowed bias voltage, and then starts the output of the pulsed laser 1 light source; then the host computer program 5 reads the pre-constrained attenuation starting value As and the attenuation ending value Ae of the optical attenuator 2, at this time, the output signal under the attenuation starting value As is in a saturated state, and the output signal under the attenuation ending value Ae is in a non-saturated state; the host computer program 5 transmits the midpoint Ac=(As+Ae) / 2 of the starting value and the ending value to the optical attenuator 2 as the attenuation value to be executed; then the host computer program 5 drives the oscilloscope 9 to read the output signal of the APD detector 7 and performs preprocessing; finally, step seven is entered;

[0050] Step seven: the preprocessed data is first input to the host computer program 5 deep neural network model analyzer; then the result matrix after solving is collected for matching to update the attenuation starting value As and the attenuation ending value Ae of the optical attenuator 2, and a stop state is returned; then the stop state is judged according to the return state; if the stop state is continue, step six is entered again; if the stop state is stop, step eight is entered;

[0051] Step eight: first, the host computer program 5 collects the average optical power value of the optical power meter 3 close to the pulsed laser 1 under the current attenuation value state, and the result after linear correction processing is taken as the saturated optical power Pmax. Then the dynamic range is obtained by the formula Gd=10lg(Pmax / Pmin). Finally, step nine is entered.

[0052] Step nine: the host computer program 5 reads the next test point, and repeats steps three to eight to test the next test point, until the dynamic range test of each test point of the APD detector 7 is completed, and then the host computer program 5 ends the test process.

[0053] The key methods in steps one to nine are further described below in combination with example parameters:

[0054] In this embodiment, the basic configuration parameters in step one are configuration XML files containing the following parameter information: an APD detector 7 of four quadrants is selected, the maximum allowed bias voltage is-200V, the minimum allowed bias voltage is-1V, the working voltage is+12V and-12V, and the extended attenuation voltage is-9V; the repetition frequency of the pulsed laser 1 is set to 10KHz, the pulse width is set to 50ns, the attenuation starting value As of the optical attenuator 2 is 10dB, at this time the light source power is large and the output signal of the APD detector 7 is in a saturated state, the attenuation ending value Ae is 50dB, at this time the light source power is small and the output signal of the APD detector 7 is in a non-saturated state; the image plane size of the APD detector 7 is 10mm, and the center points of the XY direction axes in each quadrant are selected as test points; the pre-calibrated object image pixel equivalent Cp and displacement pixel equivalent Cm are selected.

[0055] In this embodiment, the host computer program 5 drives the XY displacement table 6 to perform the test point position registration. First, the host computer program 5 visually identifies the APD detector 7 profile, and obtains the two-dimensional coordinate point at the center of the profile as the image center point of the APD detector 7 image plane, which is the physical center point of the APD detector 7 image plane observed in the infrared CCD camera 4; then the host computer program 5 calculates the pixel coordinate difference between the APD detector 7 image center point and the pre-marked light source light spot image position information as the pixel distance vector Vp; secondly, the host computer program 5 reads the image plane size and the current test point coordinate and calculates the actual physical distance vector Vl of the test point coordinate offset from the image plane center, if the selected test point is the center point of the XY direction axis in a certain quadrant, then Vl=(2.5mm, 2.5mm) or (2.5mm, -2.5mm) or (-2.5mm, 2.5mm) or (-2.5mm, -2.5mm); then the host computer program 5 calculates R=Cm(Cp×Vl+Vp), thereby converting the pixel distance in the image into the corresponding physical distance and movement direction of the XY displacement table 6; finally, the host computer program 5 drives the XY displacement table 6 to move according to the vector R, completing the registration.

[0056] In this embodiment, before testing, the average optical power of the optical power meter 3 near the pulsed laser 1 and the APD detector 7 image plane under different attenuation values needs to be collected multiple times respectively, and recorded as Pin and Pout respectively; then the power levels Pin-dBm and Pout-dBm corresponding to Pin and Pout are calculated, and linear fitting is performed on Pin-dBm and Pout-dBm to obtain the slope k and intercept b of the straight line, so that when formal automatic testing is performed, the average optical power at the APD detector 7 image plane does not need to be tested using the optical power meter 3, only the average optical power near the pulsed laser 1 needs to be obtained, and the average optical power can be calculated through the linear mapping relationship established between the two. The current pulse repetition frequency f of the pulsed laser 1 and the pulse width w are read, the obtained average optical power is taken as the independent variable x, and the formula y=(k×10lg(1000×x)+b) / (f×w) is used for calculation. This formula not only performs linear mapping calculation, but also converts the average optical power into peak optical power, and finally completes the linear correction processing.

[0057] In this embodiment, reference is made to Figure 3The waveform data 10 collected by the host computer program 5 from the oscilloscope 9 is data containing multiple cycles, and then the waveform data 10 is resampled to obtain single-cycle waveform data 11 by following the Fm / Fl value, wherein Fm is the sampling rate of the oscilloscope, and Fl is the repetition frequency of the pulsed laser; then the resampled single-cycle waveform data 11 is deformed into a two-dimensional matrix 12 with equal rows and columns, and the two-dimensional matrix 12 is normalized by maximum and minimum; the host computer program 5 inputs the processed data into a convolutional neural network model 13 containing CNN feature extraction and FC two-class full connection layer; finally, the output result of the network is a matrix represented by two values for non-saturated and saturated states. If the non-saturated state value is greater than the saturated state value, the attenuation end value Ae ’ of the optical attenuator 2 is equal to the current attenuation value Ac, and the attenuation start value As ’ remains unchanged; if the non-saturated state value is less than the saturated state value, the attenuation start value As ’ of the optical attenuator 2 is equal to the current attenuation value Ac, and the attenuation end value Ae ’ remains unchanged, so that the attenuation start value As ’ and the attenuation end value Ae ’ of the optical attenuator 2 meet the pre-constraint processing in step six; finally, if |As-Ae|<2×attenuation resolution of the optical attenuator 2, the optical attenuator 2 is attenuated to the Ac calculated again, and at this time, the Ac is not satisfied by the current optical attenuator 2. Therefore, if the judgment is true, the returned stop state is stop, otherwise the returned stop state is continue.

[0058] In this embodiment, when one test point is completed, if multiple test points are specified, the host computer program 5 drives the instrument to automatically jump to the next test point without manual intervention, effectively improving the test efficiency; in addition, the method of the present application combines the deep learning classification network model 13 to accelerate the search for the saturated optical power of the APD detector 7, and improves the accuracy and efficiency of the dynamic range test.

[0059] The above is only one embodiment of the present application, and does not limit the present application in any form. Therefore, any equivalent structure using the contents of the present application specification and drawings, directly or indirectly applied to other related technical fields, is also within the scope of the patent protection of the present application.

Claims

1. An automatic testing method for the dynamic range of an APD detector, characterized in that, Includes the following steps: Step 1: Read the basic configuration parameters of the APD detector test. The host computer program transmits the initialization operation parameters to the instruments through the control protocol. The host computer program receives the status feedback of the instruments and judges the working status of each instrument. If all instruments are in normal status, proceed to Step 2; otherwise, the current test state is terminated. Step 2: Drive the pulsed laser, optical attenuator, and optical power meter. Collect the average optical power Pin near the pulsed laser and the average optical power Pout of the APD detector image sensor surface at different attenuation values ​​multiple times. Convert the power levels Pin-dBm and Pout-dBm respectively. After using Pin-dBm as the independent variable x and Pout-dBm as the dependent variable y, perform linear fitting to obtain the slope k and intercept b, and save them to the basic configuration parameters. Step 3: The host computer program obtains the position information of the center point of the APD detector image sensor surface and the light source spot image from the infrared CCD camera image data, and drives the XY displacement stage to perform test point position registration. Step 4: The host computer program reads the APD detector bias and operating voltage, the voltage provided by the driving voltage and current source meter, reads the light source repetition frequency f, pulse width w, and attenuation value, and drives the pulsed laser, optical attenuator, and optical power meter to work. Step 5: The host computer program drives the voltage and current source meter to provide the maximum allowable bias voltage of the APD detector; turns off the light source output of the pulse laser, drives the oscilloscope to collect the root mean square voltage Vrms of the detector output voltage, obtains the detector responsivity Rv, and uses the Vrms / Rv calculation result as the minimum received optical power Pmin. Step Six: With the host computer program driving the voltage and current source meter to provide the maximum allowable bias voltage of the APD detector and the light source output turned on, the host computer program reads the pre-constrained attenuation start value As and end value Ae of the optical attenuator, and sends the median value Ac = (As + Ae) / 2 of the start and end values ​​to the optical attenuator as the attenuation value to be executed; the host computer program drives the oscilloscope to read the APD detector output signal and perform preprocessing. Step 7: The preprocessed data is input into the deep neural network model analyzer of the host computer program; the output of the deep neural network model analyzer of the host computer program is matched with the result matrix to update the attenuation start value As and end value Ae of the optical attenuator, and the stop state is returned; the following judgment is made based on the return state: If the stop status is to continue, then return to step six; If the stop status is "stopped", proceed to step eight; Step 8: The host computer program collects the average optical power value of the optical power meter near the pulsed laser under the current attenuation value, and uses the result of the average optical power value after linear correction as the saturation optical power Pmax. The dynamic range is obtained by processing it with the formula Gd=10lg(Pmax / Pmin). Step 9: The host computer program reads the next test point and repeats steps 3 to 8 to test the next test point until the dynamic range test of each test point of the APD detector is completed and the host computer program ends the test process.

2. The automatic dynamic range testing method for APD detectors as described in claim 1, characterized in that, The read APD detector test basic configuration parameters include the following configuration XML file information: electrical parameter configuration information containing APD detector bias, operation, extended attenuation voltage and enable status; optical parameter configuration information containing light source repetition frequency, pulse width, attenuation value; detector image sensor size and test point coordinates; pre-calibrated object-image pixel equivalent Cp and displacement pixel equivalent Cm; and fixed maximum allowable attenuation value of the attenuator, optical attenuator attenuation start value and end value.

3. The automatic dynamic range testing method for APD detectors as described in claim 1, characterized in that, When the APD detector is a single-pixel APD detector, the center point of the pixel is selected as the test point; when the APD detector is a multi-pixel APD detector, the position of each pixel offset from the center point of the image-sensitive surface by a fixed distance is selected as the test point.

4. The automatic dynamic range testing method for APD detectors as described in claim 1, characterized in that, The specific process of registering the test point positions using the XY displacement stage is as follows: After the host computer program visually recognizes the detector outline, it will use the two-dimensional coordinate point obtained at the center of the outline as the center point of the APD detector image sensor surface image, and pre-calibrate the two-dimensional coordinates of the circular area containing the center and radius in the two-dimensional image of the infrared CCD camera as the position information of the light source spot image. The host computer program calculates the pixel distance vector Vp between the center point of the APD detector's image sensor surface and the position information of the light source spot image. The host computer program reads the image sensor surface size and test point coordinates, and calculates the actual physical distance vector Vl between the test point coordinates and the center of the image sensor surface. The host computer program calculates the displacement vector R = Cm(Cp × Vl + Vp), where Cp is the image pixel equivalent and Cm is the displacement pixel equivalent. The host computer program drives the XY displacement stage to move according to vector R to complete the registration.

5. The automatic dynamic range testing method for an APD detector as described in claim 1, characterized in that, Specific methods for linear correction processing: The average optical power near the pulsed laser is used as the independent variable x and substituted into the formula y=(k×10lg(1000×x)+b) / (f×w) to calculate the optical power y after linear correction.

6. The automatic dynamic range testing method for APD detectors as described in claim 1, characterized in that, The specific method for pre-constraint processing is as follows: As the attenuator is gradually reduced from the initial attenuation value, the output signal of the APD detector remains in a saturated state. As the attenuator is gradually increased from its attenuation limit, the output signal of the APD detector remains in a non-saturated state.

7. The automatic dynamic range testing method for an APD detector as described in claim 1, characterized in that, The specific method for preprocessing the output signal of the APD detector is as follows: Sampling Fm / Fl data points from the oscilloscope waveform data constitutes all the data contained in the duration of a laser pulse, where Fm is the sampling rate of the oscilloscope and Fl is the repetition frequency of the pulsed laser; The resampled waveform data is transformed into a two-dimensional matrix; the two-dimensional matrix is ​​then subjected to max-min normalization.

8. The automatic dynamic range testing method for an APD detector as described in claim 1, characterized in that, The host computer program's deep neural network model analyzer uses a convolutional neural network model, which includes two convolutional layers, two pooling layers with halved height and width, one Flatten layer, and three fully connected layers. The model input is a waveform data tensor in MHWC format, and the model ends with a two-class fully connected layer containing unsaturated and saturated state values.

9. The automatic dynamic range testing method for an APD detector as described in claim 1, characterized in that, The specific method for matching the output results of the deep neural network model analyzer in the host computer program with the result matrix to update the attenuation start value As and end value Ae of the optical attenuator is as follows: The host computer program collects the output of the deep neural network model analyzer. If the value of the unsaturated state is greater than the value of the saturated state, the attenuation end value Ae of the optical attenuator is equal to the current attenuation median value Ac, while the attenuation start value As remains unchanged. If the value of the non-saturated state is less than the value of the saturated state, then the starting value As of the optical attenuator is equal to the current median value Ac, and the ending value Ae of the attenuation remains unchanged, so that the starting value As and the ending value Ae of the optical attenuator satisfy the pre-constraint processing in step seven; the host computer program judges |As-Ae|<2×the attenuation resolution of the optical attenuator. If it is true, the returned stop state is stop; otherwise, the returned stop state is continue.

Images