Optical characteristic inspection device

Abstract

This optical characteristic inspection device comprises a detector (2) that receives emitted light from an inspection target object (900), a drive device that drives the inspection target object (900) or the detector (2) to position the measurement position thereof, and an optical axis adjusting unit (4) that, on the basis of position information (Ip) indicating the measurement position acquired from the drive device each time positioning is performed, and an electrical signal (Se) output from the detector (2), adjusts the optical axis of the inspection target object (900) by performing a search calculation for a position at which the maximum amount of light is obtained from the inspection target object (900), among the measurement positions, wherein the optical axis adjusting unit (4) executes the search calculation using a regression method employing nonlinear modeling, taking into consideration the amount of positional deviation caused by the drive device.

Description

Optical property inspection equipment

[0001] The present disclosure relates to an optical property inspection device.

[0002] When inspecting the optical characteristics of optical semiconductor elements such as semiconductor lasers and Mach-Zehnder modulators, it is necessary to precisely align the optical axis of the detector with respect to the output facet. However, the optical axis position varies with each inspection due to variations in the element characteristics and variations in the alignment accuracy of the element within the inspection equipment.

[0003] For example, when adjusting the optical axis between the output end face of an optical semiconductor element and a detector, the optical signal emitted from the optical semiconductor element is received by the detector, the received optical signal is converted into an electrical signal, etc., and the optical axis is adjusted while checking the signal strength. Therefore, when done manually, there are issues such as the inspection results being influenced by the skill of the operator and the optical axis adjustment taking time.

[0004] On the other hand, it is possible to operate the motor by program and perform automatic optical axis adjustment, but the general method of obtaining the point where the maximum light intensity is obtained by performing a full search within a specified range increases the time required for optical axis adjustment because signals are obtained in unnecessary ranges as well. Alternatively, an algorithm such as the hill-climbing method can be used to shorten the optical axis adjustment time. However, when the characteristics of the element before inspection, i.e., the element itself, are unknown, there is not necessarily a single peak, and methods such as the hill-climbing method may result in a locally optimal solution.

[0005] Therefore, it is conceivable to use a regression method such as Gaussian process regression (see, for example, Patent Document 1) to avoid convergence at a local optimum and to find a global optimum with fewer steps.

[0006] Japanese Patent Application Laid-Open No. 2004-85801 (paragraph 0060, Figures 12 and 14)

[0007] However, while typical regression methods take noise into account in the objective variable, they do not necessarily reach the global optimum when there is uncertainty in the explanatory variables, such as the motor's repeatability.

[0008] The present disclosure is intended to solve the above-mentioned problems, and aims to provide an optical property inspection device that can avoid convergence at a local optimum point and adjust the optical axis using a global optimum point obtained with fewer steps.

[0009] The optical property inspection device disclosed herein includes a detector that receives light emitted from an optical semiconductor element, a drive unit that drives and positions a measurement position of either the optical semiconductor element or the detector, and an optical axis adjustment unit that controls the drive of the drive unit and, each time positioning is performed, performs a search calculation for a position among the measurement positions at which the maximum light intensity is obtained from the optical semiconductor element based on position information indicating the measurement position obtained from the drive unit and light intensity information output from the detector, and adjusts the optical axis of the optical semiconductor element, wherein the optical axis adjustment unit performs the search calculation using a regression method with nonlinear modeling, taking into account the amount of positional deviation caused by the drive unit.

[0010] According to the optical property inspection device disclosed herein, convergence at a local optimum point can be avoided by performing processing that takes into account uncertainty in the explanatory variables, thereby making it possible to obtain an optical property inspection device that can adjust the optical axis by finding the global optimum point with fewer steps.

[0011] 4A and 4B are diagrams showing an initial stage of search and a state where the search has progressed further, respectively, for measured values ​​in optical axis adjustment using Gaussian process regression, a predicted average value based on the measured values, and a range twice the standard deviation.

[0034] FIG. 4B is a diagram showing a state where the search has been completed to a certain extent, for measured values ​​in optical axis adjustment using Gaussian process regression, a predicted average value based on the measured values, and a range twice the standard deviation.

[0035] FIG. 4B is a flowchart for explaining the operation of the optical property inspection apparatus according to the first embodiment.

[0036] FIG. 4C is a flowchart for explaining the operation of an optical property inspection apparatus according to the first embodiment in an optical axis adjustment step.

[0037] FIG. 4D is a block diagram showing the hardware configuration of a portion that executes arithmetic processing in the optical property inspection apparatus according to the first embodiment.

[0038] FIG. 4D is a block diagram for explaining the configuration of an optical property inspection apparatus according to a first modified example of the first embodiment.

[0039] FIG. 4E is a block diagram for explaining the configuration of an optical property inspection apparatus according to a second modified example of the first embodiment. FIG. 1 is a block diagram for explaining the configuration of an optical property inspection apparatus according to a second embodiment. FIG. 2 is a block diagram for explaining the configuration of an optical property inspection apparatus according to a modified example of the second embodiment. FIG. 3 is a block diagram for explaining the configuration of an optical property inspection apparatus according to a third embodiment at a front stage. FIG. 4 is a block diagram for explaining the configuration of an optical property inspection apparatus according to a second modified example of the third embodiment. FIG. 5 is a flowchart for explaining the operation of an optical property inspection apparatus according to the third embodiment. FIG. 6 is a block diagram for explaining the configuration of an optical property inspection apparatus according to a first modified example of the third embodiment. FIG. 7 is a block diagram for explaining the configuration of an optical property inspection apparatus according to a second modified example of the third embodiment. FIG. 8 is a block diagram for explaining the configuration of an optical property inspection apparatus according to a fourth embodiment. FIG. 9 is a block diagram for explaining the configuration of an optical property inspection apparatus according to a first modified example of the fourth embodiment. FIG. 10 is a block diagram for explaining the configuration of an optical property inspection apparatus according to a second modified example of the fourth embodiment. FIG. 11 is a block diagram for explaining the configuration of an optical property inspection apparatus according to a fifth embodiment.

[0012] The optical property inspection device disclosed herein can inspect not only light-emitting elements such as semiconductor lasers that emit light themselves and function as light sources, but also modulators such as Mach-Zehnder modulators. Although the following description will be given using an example in which a semiconductor laser is used as the object to be inspected, when a modulator is the object to be inspected, a separate light source is provided to input an optical signal to the modulator.

[0013] 1 to 7 are diagrams for explaining the configuration and operation of an optical property inspection device according to a first embodiment, and Fig. 1 is a block diagram illustrating the spatial arrangement of an object to be inspected and a part that detects light, in order to explain the configuration of the optical property inspection device.

[0014] Figure 2 is a diagram simulating the two-dimensional distribution of the amount of light emitted from an optical semiconductor element, which is the object to be inspected (tested object), on a plane perpendicular to the optical axis, with the light appearing whiter as the light intensity increases, Figure 3A is a diagram showing the progression of search points during optical axis adjustment using a full search, and Figure 3B is a diagram showing the progression of search points during optical axis adjustment using Gaussian process regression.

[0015] 4A, 4B, and 5 show the measured values ​​(x), the predicted average value (solid line) based on the measured values, and the range twice the standard deviation (shaded) in optical axis adjustment using Gaussian process regression, with Fig. 4A showing the initial stage of search, Fig. 4B showing the state after further search, and Fig. 5 showing the state after the search has been completed to a certain extent, at which a re-search is started in processing that takes uncertainty into account. Fig. 6 is a flowchart for explaining the overall operation of the optical property inspection apparatus, and Fig. 7 is a flowchart for explaining the operation in the optical axis adjustment process.

[0016] As shown in FIG. 1, the optical property inspection device 1 in embodiment 1 is a device that aligns (optical axis adjustment) a detector 2 that detects light with the position of the optical axis Ao (optical axis position) relative to the output end face of an optical semiconductor element, which is the object 900, and inspects the optical properties of the object 900.

[0017] For the optical semiconductor element under test 900, the device is equipped with a device under test fixture 6 that positions and fixes the device under test 900, and a device under test operating power supply 7 that operates (emits light or modulates light) the device under test 900. The device under test fixture 6 may have a counterbore to accommodate the maximum tolerance of the optical semiconductor element. Such a design makes it possible to limit the optical axis adjustment range, which is expected to result in faster convergence of the regression calculation. Furthermore, if necessary for testing the optical semiconductor element, a temperature control element, such as a Peltier element, may also be provided. Furthermore, if it is difficult to fix the element in place due to the operating environment, a vacuum suction mechanism may also be provided.

[0018] The optical axis adjusting unit 4 sets a search point based on the amount of light (electrical signal Se) detected by the detector 2 and position information Ipd of the detector 2 output from the detector driving unit 3, and controls the driving of the detector driving unit 3 (outputs a driving signal Scd).The optical axis adjusting unit 4 also includes an inspection control unit 8 that controls the overall operation of the optical property inspection.

[0019] The detector 2 may be, for example, an optical fiber, a photodiode, a lensed fiber, a bulb-tipped fiber, etc. The material of the detector 2 is preferably selected appropriately in accordance with the wavelength of the light emitted from the object 900 to be inspected.

[0020] The detector driving device 3 drives the detector 2 in a horizontal direction (perpendicular to the optical axis Ao) and in the emission direction relative to the emission surface of the inspected object 900. A stepping motor, for example, is used to drive the detector driving device 3. Generally, the optical axis position of a laser diode or the like is about 10 μm, so it is preferable to use a device with a finer resolution than that. Furthermore, for directions that require more precise adjustment, a combination of stepping motors with different resolutions for the same direction may be used.

[0021] The optical axis adjustment unit 4 includes an optimal position search unit 41 that uses Gaussian process regression by nonlinear modeling using a kernel function to set a search point based on the electrical signal Se and position information Ipd, and a variable processing unit 42 that processes explanatory variables in consideration of uncertainty. It also includes a convergence determination unit 43 that determines the state of convergence in two stages, quasi-convergence and complete convergence, and a re-search position setting unit 44 that sets a re-search position when quasi-convergence is determined.

[0022] By operating the detector drive device 3 in cooperation with the optical axis adjustment unit 4 described above, the detector 2 can be moved to coordinates where a high output of light emitted from the object under test 900 can be obtained. When searching for the optimal point, the optimal position is searched for and set using, for example, position coordinates (position information Ipd: for example, XYZ) as explanatory variables and the light intensity (electrical signal Se) as a target variable. Furthermore, a function for adjusting the angle may be added to the detector drive device 3. In that case, angle information is also added to the explanatory variables.

[0023] Gaussian process regression is generally linear, but when there are multiple peaks, such as in optical semiconductor devices, linear modeling, including Gaussian process regression, is unsuitable, and Gaussian process regression, which performs nonlinear modeling by using a kernel function, is preferable.

[0024] Before explaining the operation of the optical property inspection device 1, we will explain the difference between a general search (full search) and Gaussian process regression, as well as the challenges involved in setting search points using general Gaussian process regression and the setting of search points by performing variable processing that takes uncertainty into account.

[0025] Consider the case where the light intensity distribution emitted from the optical semiconductor element serving as the object under test 900 has two peaks due to a peak split, as shown in Fig. 2. For example, when optical axis adjustment is performed by full search, search points (white circles) are set sequentially so as to cover the entire region, as shown in Fig. 3A. In this case, too, it is considered common practice to carry out a more detailed search for points closer to the peak.

[0026] On the other hand, when the optical axis is adjusted using Gaussian process regression, the area near the boundary (periphery) is searched first, and then points with high expected values ​​are set as search points, as shown in Figure 3B. This behavior is expected to result in a search point with high light intensity being reached with fewer search points than with a full search.

[0027] Although Gaussian process regression is a common technique, it is designed to take noise into account in the objective variable. However, in cases like this one involving motor drive, it is necessary to consider a certain amount of noise in the explanatory variables as well, such as uncertainty in the repeatability of a stepping motor.

[0028] Taking noise into account in such input variables is considered a difficult problem in statistics, and simply applying Gaussian process regression may not be enough to achieve stable, accurate, and high-speed adjustment. In other words, although optical axis adjustment can be completed at a fairly high speed, the calculation may not converge at the point where maximum light intensity is obtained because the input contains noise.

[0029] Therefore, in the optical property inspection device 1 disclosed herein, the drawbacks of Gaussian process regression are overcome by adding processing (variable processing by the variable processing unit 42) that takes into account that the position information (explanatory variable) of the stepping motor is probabilistically distributed depending on the repeatability and reproducibility of the stepping motor.

[0030] The future positional deviation of the stepping motor is considered to be strongly dependent on the current motor position and therefore to have a Markov property. Therefore, difficulties can be avoided by adding a process that takes the Markov property into account to the explanatory variables. For example, once the optical axis search has been completed to a certain extent (semi-convergence), an algorithm may be provided that performs a process (Process A) of probabilistically visiting a position considered to be the optimal point several times. Alternatively, an algorithm may be provided that performs a process (Process B) of varying the probability of visiting several times based on the expected value at which the maximum light intensity is obtained.

[0031] "Semi-convergence" refers to the timing at which it is determined that there is no point within the range where the search and variance calculation have converged at this point where a higher expected value can be obtained. In other words, it is the timing at which the maximum light intensity can be expected to be obtained by re-searching. The timing of semi-convergence will be explained using Figures 4A to 5. Note that the calculations (average predicted values) in the figures all take noise into account in the objective variables, so the function (solid line) estimated from the measurement data does not necessarily pass through the measurement points (x).

[0032] At the stage shown in Figure 4A, the number of searches is still insufficient, and it is not clear whether the x (explanatory variable) that yields the maximum y (objective variable) is in the range of -3 to -2, or in the range of 1 to 2. Furthermore, the range of values ​​that requires searching has not been narrowed sufficiently. At this point, there is a high possibility that re-searching points that have already been searched will be a waste of time, so it is not determined that quasi-convergence has occurred and no re-search is performed.

[0033] If we continue searching from here and reach the state shown in Figure 4B, we can see that the maximum point of y is likely to be in the range of x = (1.5 to 2.5). However, the data for x = 2 here contains an error due to the repeatability of the stepping motor. At this point, adding processing such as a re-search can minimize the effect of the repeatability of the stepping motor.

[0034] In addition, the operation of Process A: "probabilistically stopping at a position that is considered to be the optimal point a certain number of times" or Process B: "variable probability of stopping at a certain number of times depending on the expected value at which the maximum light intensity is obtained" that is performed when quasi-convergence is achieved will be explained using the case shown in Figure 5.

[0035] In this case, it can be seen that the average value + 2σ is likely to show the maximum value at two locations, near x = -3.0 and near x = 2.0. However, the reason that the maximum value is likely to be shown at x = -3.0 is simply because the variance is large. Therefore, in this case, the operation of Process B is to weight the value according to the variance and design it so that the value at x = 2.0, which has a smaller variance, is more likely to be searched again.

[0036] Process A corresponds to the operation of probabilistically re-searching for X = 2.0 or X = -3.0 (whichever occurs is determined by a random number or other method) at the timing of quasi-convergence. It is believed that this probability is set optimally by conducting experiments using actual objects. For example, the optimal probability is considered to be the probability that can be changed using various measurement targets, automatic optical axis adjustment can be performed, peak adjustment can be achieved, and the takt time can be minimized. Furthermore, rather than fixing the probability, it is also possible to devise a method in which the re-search point or the magnitude of variance in its vicinity can be used to affect the re-search probability in addition to the aforementioned probability.

[0037] Next, an operation that takes Markov property into consideration will be described. For example, in the above example, suppose that the point x = 2.0 is searched for by moving in the direction of increasing x. In this case, it is considered that the position accuracy will differ physically due to inertia when moving from x = 0.0 in the direction of increasing x and when moving from x = 4.0 in the direction of decreasing x. When moving from x = 0.0, inertia makes it easier to stop at a point slightly larger than x = 2.0, and when moving from x = 4.0, conversely, it makes it easier to stop at a point slightly smaller than x = 2.0. It is considered that the effect of noise on the objective variable can be minimized by an operation (processing) that takes into account the likelihood of position deviation due to inertia.

[0038] As described above, the operation of the optical characteristic inspection device equipped with the optical axis adjustment unit 4 configured to perform additional processing such as re-searching on the Gaussian process regression that has performed nonlinear modeling by using a kernel function will be described with reference to the flowcharts in FIGS. 6 and 7.

[0039] 6, the device under test 900 is positioned and fixed to the device under test fixture 6 or the like (step S100). This operation may be performed manually using tweezers or the like, or automatically using a suction collet or the like based on a command from the inspection control unit 8. Next, the device under test 900 is connected to the device under test operating power supply 7 for operating the device under test 900 (step S110). This connection is performed by bringing probes or the like into contact with the electrodes of the device under test 900.

[0040] After the device under test 900 is connected to the device under test operating power supply 7, the device under test operating power supply 7 is turned on to operate the device under test 900 (step S120). This operation may also be performed manually or automatically based on a command from the inspection control unit 8. In this case, the operating conditions are not important, but it is preferable to operate the device under conditions in which light is continuously emitted (or modulated) when adjusting the optical axis.

[0041] Next, as described above, the detector 2 receives light emitted from the object under test 900 while sequentially setting search points, and an optical axis adjustment process is performed (step S200). In the optical axis adjustment process, as shown in Fig. 7, variable processing is first performed taking noise into account for the objective variable (step S210). Then, using Gaussian process regression that performs nonlinear modeling using a kernel function, variance calculation is performed based on the setting of search points and the light intensity measurement results obtained by moving to the set search points (step S220).

[0042] At this time, the detector 2 is moved to a position (drive signal Scd) specified by the optimum position search unit 41 using the detector drive device 3, whereby the detector 2 is moved to a position where a statistically high output can be expected to be obtained. Note that in step S220, the motor drive may be controlled assuming positional deviation due to inertia, taking into consideration the above-mentioned Markov property.

[0043] The convergence state is determined (step S230), and if it is determined that convergence has not occurred, the process proceeds to step S220, where a re-search is performed. On the other hand, if it is determined that the optical axis search has been completed to a certain extent, that is, if it is determined that the search has been quasi-converged, a re-search position is set (step S250) using the above-mentioned processes A and B, and the process proceeds to step S220, where a re-search is performed. If it is determined that complete convergence has occurred, the optical axis adjustment process ends.

[0044] Once the optical axis adjustment process is completed and the optical axis is adjusted, an optical property inspection is performed (step S300: FIG. 6). In the optical property inspection, the output of the power supply 7 for operating the device under test is changed depending on the inspection item, or the detector drive device 3 is driven in a pattern different from the optical axis adjustment depending on the inspection item, to inspect the optical properties of the device under test 900. When the inspection is complete, the inspected product is removed (step S400), and the inspection ends.

[0045] This avoids convergence at local optima and enables optical axis adjustment using a global optimum obtained with fewer steps. In addition, in the case of nonlinear regression methods such as neural networks, processing that takes uncertainty into account, similar to Gaussian process regression made nonlinear using a kernel function, makes it possible to avoid convergence at local optima and enables optical axis adjustment using a global optimum obtained with fewer steps.

[0046] On the other hand, algorithms that search for an optimal point by moving multiple points, such as swarm particle optimization, are also unsuitable for optical element inspection. Furthermore, because it is unclear whether the object being measured has an optical peak close to a simple normal distribution, stochastic gradient descent is also unsuitable. Furthermore, because the relationship between optical peaks and position is generally nonlinear, it is considered difficult to apply linear regression alone. Linear modeling and probabilistic estimation methods are also unsuitable when multiple peaks exist. In other words, by using a nonlinear modeling method and performing processing that takes uncertainty into account, convergence at a local optimum can be avoided, enabling optical axis adjustment using a global optimum obtained with fewer steps.

[0047] In the optical property inspection apparatus 1 of the present disclosure, a portion that performs arithmetic processing, such as the optical axis adjustment unit 4 or the inspection control unit 8, may be configured as a single piece of hardware 400 including a processor 401 and a storage device 402, as shown in FIG. 8 . Although not shown, the storage device 402 includes a volatile storage device such as a random access memory and a non-volatile auxiliary storage device such as a flash memory. Alternatively, a hard disk auxiliary storage device may be included instead of the flash memory. The processor 401 executes a program input from the storage device 402. In this case, the program is input to the processor 401 from the auxiliary storage device via the volatile storage device. Furthermore, the processor 401 may output data such as calculation results to the volatile storage device of the storage device 402, or may store the data in the auxiliary storage device via the volatile storage device.

[0048] First Modification: In this first modification, an example will be described in which an object under test is driven instead of a detector. FIG. 9 is a block diagram simulating the spatial arrangement of the object under test and the part that detects light, in order to explain the configuration of an optical property inspection device according to the first modification. For the sake of brevity, the internal configuration of the optical axis adjustment unit will not be depicted in the following description, as compared to FIG. 1.

[0049] 9, the optical property inspection device 1 according to the first modification is provided with an inspected object driving mechanism 6M that movably supports the inspected object 900, instead of a fixing jig, for the inspected object 900. A signal Scs that controls the driving of the inspected object driving mechanism 6M is output from the optical axis adjustment unit 4 to the inspected object driving mechanism 6M, and position information Ips of the inspected object 900 is output from the inspected object driving mechanism 6M to the optical axis adjustment unit 4.

[0050] In this case, the movement direction (and position) of the light source, which is the object under test 900, is not limited to the XYZ directions of the Cartesian coordinate system, but may be the rθφ directions of the spherical coordinate system. Furthermore, in this first modified example, in order to correct the angle of incidence of light from the object under test 900 to the detector 2, it is preferable to use a detector having a function of correcting the angle of incidence, such as a bulb-tipped fiber, as the detector 2. In this first modified example, by moving the light source side, particularly in the angular direction, a global optimum solution can be obtained with a smaller amount of movement.

[0051] Furthermore, the object under test 900 and the object under test drive mechanism 6M preferably have a function for fixing the object under test 900, such as a vacuum suction mechanism, so that the light emission surface of the light source, which is the object under test 900, accurately follows the movement. Furthermore, since the light source side is moved, it is necessary to provide a mechanism for connecting the object under test 900 and the object under test operating power supply 7, such as a probe, with a mechanism for moving in the same way as the object under test 900.

[0052] In addition, although a mechanism for moving the detector 2 is not shown in the first modified example, a mechanism for moving the detector 2 may be provided.

[0053] Second Modification: In this second modification, an example will be described in which a lens is arranged in front of a detector. Fig. 10 is a block diagram illustrating the spatial arrangement of an object to be inspected and a part that detects light, in order to explain the configuration of an optical property inspection device according to the second modification.

[0054] 10, the optical property inspection device 1 according to the second modification has the same configuration as that described in the first modification, but with a lens 2op inserted between the exit surface of the inspected object 900 and the detector 2. With this configuration, the detector 2 does not necessarily need to have a function for correcting the incident angle of light.

[0055] The lens 2op may be, for example, a convex lens, a spherical-convex lens, or an aspherical lens. In addition, in this second modified example, a mechanism for moving not only the detector 2 but also the object under test 900 and the lens 2op may be provided. It is preferable to provide an appropriate movement mechanism depending on the space available in the device and the emission characteristics of the light source.

[0056] Second Embodiment In the second embodiment, instead of the single detector used in the first embodiment, a detector bundle is provided in which multiple detectors are bundled so that their incident surfaces are arranged in an arc shape. Figure 11 is a block diagram simulating the spatial arrangement of the object to be inspected and the part that detects light, in order to explain the configuration of the optical property inspection device according to the second embodiment. Note that the operation of setting the search points is the same as in the first embodiment, and Figures 2 to 7 described in the first embodiment are used.

[0057] The optical property inspection apparatus 1 according to the second embodiment is provided with a detector bundle 2G in which a plurality of detectors 2 are bundled together so that their incident surfaces are arranged in an arc shape, rather than the single detector 2 described in the first embodiment. The optical axis adjustment unit 4 sets a search point based on the light intensity (electrical signal SeG) detected by each detector 2 in the detector bundle 2G and position information IpG of the detector bundle 2G output from the detector drive device 3. Then, based on the set search point, the optical axis adjustment unit 4 controls the driving of the detector drive device 3 (outputs a drive signal ScG).

[0058] The detector bundle 2G is arranged so that the incident surface is arcuate in a plane including the optical axis Ao, and its position in the direction of the emitted light (distance from the light source) is shifted. This eliminates the need to move it in the direction of the emitted light. Therefore, the detector bundle moving device 3G can be reduced by one axis compared to the detector bundle moving device 3G described in the first embodiment, and the number of objective variables can be reduced, which has the effect of allowing the regression calculation to converge faster.

[0059] In the figure, the incident surface of each detector 2 constituting the detector bundle 2G is configured to be convex when viewed from the light source, but it may also be configured to be concave or the like, and it is preferable to determine the shape appropriately depending on the characteristics of the optical semiconductor element to be measured.

[0060] Modified Example: The arrangement of the incident surfaces of the detectors is not limited to an arc shape, as long as the distances from the light source are different. In this modified example, an example is described in which the incident surfaces of multiple detectors are arranged in a straight line so that they are at different distances from the light source. Figure 12 is a block diagram simulating the spatial arrangement of the object to be inspected and the part that detects light, in order to explain the configuration of an optical property inspection device according to this modified example.

[0061] The optical property inspection device 1 according to the modified example may be shifted in a straight line, as shown in Fig. 12. The distance of the detector bundle 2G from the light source is not changed in an arc shape with the center of the arrangement as the apex or base, but is changed monotonically, for example, from one edge toward the opposite edge. By adopting such a configuration, if the area of ​​the detector bundle 2G is sufficiently large, optical axis adjustment can be completed by simply moving it in one axial direction.

[0062] Third Embodiment. In the first and second embodiments, examples of performing optical axis adjustment without preliminary measurement have been described. In the third embodiment, an example of performing optical axis adjustment after acquiring light intensity mapping information will be described. FIGS. 13 to 15 are used to explain the configuration and operation of an optical property inspection device according to the third embodiment. FIG. 13 is a block diagram simulating the spatial arrangement of the object to be inspected and the part acquiring the light intensity mapping when acquiring the light intensity mapping information. FIG. 14 is a block diagram simulating the spatial arrangement of the object to be inspected and the part detecting light in a state where the equipment has been swapped after acquiring the light intensity mapping information. FIG. 15 is a flowchart for explaining the operation of the optical property inspection device. Note that the search point setting operation is the same as in the first embodiment, and FIGS. 2 to 6 described in the first embodiment are used.

[0063] As shown in FIG. 13 , the optical property inspection apparatus 1 according to the third embodiment places a camera 2C in a preliminary step relative to the emission surface of an object under test 900 fixed to an object under test fixing jig 6. The sensor of the camera 2C may be a CCD (Charge-Coupled Device) image sensor, a CMOS (Complementary Metal-Oxide-Semiconductor) sensor, or the like. The camera 2C is supported by a camera moving device 3C that can move in at least one axial direction (the direction of the arrow in the figure) to adjust the position of the camera 2C in the emission light direction. As a result, as shown in FIG. 15 , the camera 2C can acquire light intensity mapping information of the light emitted from the light source (step S130).

[0064] Based on the mapping information obtained here, the equipment is rearranged to a configuration such as that shown in Fig. 14 (step S140). At this time, in the optical axis adjustment (step S200), the movement range of the detector 2 is searched only within the range of the resolution of the camera 2C and the resolution of the stepping motor from the coordinate point where the maximum light intensity is obtained on the light intensity mapping. In other words, by obtaining the light intensity mapping information in advance, accurate optical axis adjustment can be performed with fewer search points.

[0065] First Modification: The device used to acquire light intensity mapping information is not limited to a camera. In this first modification, an example will be described in which a phosphor is used instead of a camera. Figure 16 is a block diagram simulating the spatial arrangement of an object to be inspected, a part for acquiring light intensity mapping information, and a part for detecting light, of an optical property inspection device according to the first modification.

[0066] In the optical property inspection device 1 according to the first modification, a phosphor 2L may be used instead of a camera when acquiring light intensity mapping information, as shown in Fig. 16. A phosphor 2L corresponding to the emission wavelength of the light source is arranged at the end of the light emitted from the light source (inspection object 900) so as to extend perpendicularly to the optical axis Ao. In addition, a detector 2 is arranged so as to penetrate the center of the phosphor 2L.

[0067] When performing optical axis adjustment, the camera 2C captures an image of the phosphor 2L with the light source emitting light, thereby obtaining mapping data regarding the spread of light from the light source based on the light emission state of the phosphor 2L, and simultaneously performing optical axis adjustment. In other words, the light intensity distribution information acquisition (step S130) described in FIG. 15 is performed in parallel in step S200, and the equipment replacement (step S140) is skipped. In this case, the light emission state of the phosphor 2L is used as the objective variable, and a statistical method such as Gaussian process regression can be used to quickly complete optical axis adjustment.

[0068] In this case, the phosphor 2L has an area that takes into consideration the spread of light emitted from the light source, specifically, 1×1 cm 2 In this modification, the optical axis may be adjusted in a dark room in order to detect the light emission state of the phosphor 2L more precisely.

[0069] Second Modification: Alternatively, the connection between the photoelectric converter that converts optical signals into electrical signals and the optical axis adjustment unit may be via an AD conversion board. Figure 17 is a block diagram simulating the spatial arrangement of the inspected object and the light detection unit of the optical property inspection device according to the second modification.

[0070] In the optical property inspection device 1 according to the second modification, as shown in Fig. 17, the optical axis adjustment unit 4A is configured by a computer equipped with an AD conversion board. This electrically connects the computer and the photoelectric converter 5 via the AD conversion board. For example, by synchronizing the motor movement period with the timing of signal acquisition from the AD conversion board, it becomes possible to continuously acquire the output of the photoelectric converter 5, which converts the optical output So at a specific position into an electrical signal Se. This configuration makes it possible to acquire electrical signal data even while moving to a point where a statistically high output is obtained, and as a result, regression can be performed using a large amount of data, which can be expected to result in rapid convergence of the regression calculation.

[0071] Fourth Embodiment In this fourth embodiment, an example will be described in which equipment for measuring the position and angle of an object under test is provided for optical axis adjustment. Figure 18 is a block diagram illustrating the configuration and operation of an optical property inspection device according to the fourth embodiment, simulating the spatial arrangement of a part for checking the installation state of an object under test and a part for detecting light. Note that the optical axis adjustment operation is the same as in the first embodiment, and reference is made to Figures 2 to 5 and 7 described in the first embodiment, and Figure 15 described in the third embodiment.

[0072] 18, the optical property inspection apparatus 1 according to the fourth embodiment has a detector 2 supported by a detector drive device 3 on the emission surface side of an optical semiconductor element (light source) that is an object under inspection 900. In this case, a single detector 2 may be used, or detectors 2 may be bundled together to increase the effective light receiving area.

[0073] The light source, which is the object under inspection 900, is imaged using camera 6C before the optical axis is adjusted. Image processing technology is then used to obtain the image information of the light source, and position information such as the position information of the light output end face and the tilt angle of the light source are obtained. This corresponds to replacing the light intensity distribution information acquisition (step S130) described in FIG. 15 of the third embodiment with light source arrangement information acquisition. Furthermore, the equipment replacement (step S140) is skipped.

[0074] The range of movement of the detector 2 is limited based on the light source position information obtained by image processing. With this configuration, the optical axis adjustment can be completed more quickly. Furthermore, machine learning may be performed using the light source position information, tilt angle, and converged position information Ip of the detector 2 (detector drive device 3) obtained by image processing as learning data. By using machine learning, it is possible to set the initial position of the detector 2 to a point closer to the optimal point, and it is also possible to further limit the range of movement of the detector 2.

[0075] First Modification: To acquire the positional information of the light source, a laser displacement meter or the like may be used in combination with a camera. In this first modification, an example will be described in which the positional information of the light source is acquired by combining a camera and a one-dimensional laser displacement meter. Figure 19 is a block diagram simulating the spatial arrangement of the inspected object, the part for acquiring the positional information of the inspected object, and the part for detecting light in the optical property inspection device according to the first modification.

[0076] In the optical property inspection device 1 according to the first modification, a camera 6C and a one-dimensional laser displacement meter 6D are combined as a configuration for acquiring position information, as shown in Fig. 19. When, for example, a plate-shaped optical semiconductor element is used as the light source, which is the inspected object 900, the one-dimensional laser displacement meter 6D is positioned so that laser light is directed at the principal surface from a position separated along the normal to the principal surface. With this configuration, it is possible to acquire height information from the one-dimensional laser displacement meter 6D in addition to the position and tilt information of the light source acquired by the camera 6C.

[0077] Furthermore, by acquiring multiple points as light source height information using the one-dimensional laser displacement meter 6D, it is possible to acquire information on the vertical tilt of the light source. By inputting the light source height information and vertical tilt information into the prediction algorithm, it is possible to predict an initial position closer to the optimal point, and optical axis adjustment can be completed more quickly. In this case, machine learning or the like may also be used to optimize the initial position.

[0078] Second Modification: A two-dimensional laser displacement meter may be used as the laser displacement meter. Fig. 20 is a block diagram simulating the spatial arrangement of an object to be inspected in an optical property inspection device according to the second modification, a part for acquiring the object's positional information, and a part for detecting light. The optical property inspection device 1 according to the second modification is provided with a two-dimensional laser displacement meter 6D2 as shown in Fig. 20, instead of the combination of camera 6C and one-dimensional laser displacement meter 6D in the first modification.

[0079] This configuration makes it possible to acquire the surface morphology of the light source, which is the object under test 900, and simultaneously acquire the tilt of the light source's emission surface, tilt in the vertical direction, and position information in the horizontal direction, i.e., placement information. When determining the initial position of the detector 2 based on the information obtained by the two-dimensional laser displacement meter 6D2, it becomes possible to set the initial position near the point where the maximum light intensity is obtained. Furthermore, because tilt information is available, it is possible to reduce the search range. As a result, it becomes possible to reduce inspection time and calculation time.

[0080] Fifth Embodiment In each of the above embodiments, an example has been described in which optical axis adjustment is performed one by one on optical semiconductor elements as the object to be inspected. In this fifth embodiment, an example will be described in which a plurality of light sources are inspected at once as the object to be inspected. Figure 21 is a block diagram illustrating the configuration and operation of an optical property inspection device according to the fifth embodiment, and is a model of the spatial arrangement of two light sources as the object to be inspected and the part that detects light. Note that the optical axis adjustment operation is the same as in the first embodiment, and Figures 2 to 7 described in the first embodiment are used.

[0081] 21, the optical property inspection apparatus 1 according to the fifth embodiment uses two optical semiconductor elements 900a and 900b as the object under test 900. The detector used is a detector bundle 2G that bundles together a plurality of detectors 2. For simplicity, two light sources are used, but three or more light sources may be used.

[0082] The power supply 7 for operating the device under test may, for example, alternately apply pulse signals with known periods to the optical semiconductor elements 900a and 900b, and use the pulse signals as a trigger to acquire the electrical signal SeG from the photoelectric converter 5. With this configuration, it becomes possible to distinguish and acquire the optical signals from the multiple optical semiconductor elements 900a and 900b without moving the detector bundle 2G.

[0083] Furthermore, the destination of detector bundle 2G is determined by using Gaussian process regression through nonlinear modeling using the kernel function described above, and processing is performed taking uncertainty into consideration. For example, detector bundle 2G is first moved so that the regression calculation for one optical semiconductor element of inspected device 900 converges, and the information obtained when detector bundle 2G was moved for the first time is reused for the regression calculation from the second optical semiconductor element.

[0084] By using such a configuration, it is expected that the average convergence time of the regression calculation per optical semiconductor element can be shortened. In this case, it is preferable that the performance of the computer used as the optical axis adjustment unit 4, such as the specifications of the memory and CPU (Central Processing Unit), be appropriately selected depending on the number of optical semiconductor elements. Furthermore, multiple computers may be used as necessary.

[0085] In this configuration, when optical fibers are used as detectors, it is not necessary to use only one, and a bundle of optical fibers may be used similarly to the detector bundle 2 G. Furthermore, if the pitch between the light sources supported by the fixture 6 for fixing the object under test falls within a range of, for example, the pitch distance of about ±1 mm, a detector or detector bundle corresponding to the number of light sources matching the pitch interval between the optical semiconductor elements may be provided.

[0086] Furthermore, the plurality of optical semiconductor elements do not necessarily need to be arranged in a plane, but may be arranged three-dimensionally. As in the fourth embodiment, the object under test 900 may be imaged and height information obtained using the camera 6C, one-dimensional laser displacement meter 6D, two-dimensional laser displacement meter 6D2, etc., and the information may be used to determine the movement destination of the detector bundle 2G or to limit the movement range.

[0087] Furthermore, the initial position of the detector bundle 2G may be determined by previously acquiring information on the light intensity distribution of the light emitted from the light source using the camera 2C, phosphor 2L, etc., as in the second embodiment or its modified example. The field of view of the camera 2C and the area of ​​the fluorescent surface of the phosphor 2L used in this case are appropriately determined according to the arrangement of the multiple light sources.

[0088] Although various exemplary embodiments and examples are described in this disclosure, the various features, aspects, and functions described in one or more embodiments are not limited to the application of a particular embodiment, but may be applied to the embodiments alone or in various combinations. Therefore, countless modifications not illustrated are contemplated within the scope of the technology disclosed in this specification. For example, this includes cases where at least one component is modified, added, or omitted, or where at least one component is extracted and combined with components of another embodiment.

[0089] As described above, the optical property inspection apparatus 1 of the present disclosure includes the detector 2 that receives light emitted from the optical semiconductor element (object under inspection 900), a drive device (detector drive device 6, detector bundle moving device 6G, camera moving device 3C, object under inspection drive mechanism 6M) that positions and drives the measurement position of either the optical semiconductor element (object under inspection 900) or the detector 2, and an optical axis adjustment unit 4 that controls the drive of the drive device and performs a search calculation for a position among the measurement positions at which the maximum light intensity is obtained from the optical semiconductor element (object under inspection 900) based on position information Ip indicating the measurement position obtained from the drive device and light intensity information (electrical signal Se) output from the detector 2 each time positioning is performed, and adjusts the optical axis of the optical semiconductor element (object under inspection 900), and the optical axis adjustment unit 4 performs the search calculation using a regression method based on nonlinear modeling (for example, Gamma process regression in which nonlinear modeling is performed using a kernel function, or a neural network) taking into account the amount of positional deviation caused by the drive device. This reduces the influence of noise, avoids convergence at a local optimum even when peak splits occur, and enables optical axis adjustment using a global optimum obtained with fewer steps.

[0090] In this case, the influence of noise can be further reduced by providing the optical axis adjustment unit 4 with a convergence determination unit 43 that determines the convergence state of the search calculation and a re-search position setting unit 44 that sets a re-search position when the convergence determination unit 43 determines that convergence has occurred to a certain extent.

[0091] Furthermore, if the re-search position setting unit 44 sets the re-search position from among multiple positions that may show the maximum light amount by weighting using variance or by allocating using random numbers, the influence of noise can be more reliably reduced.

[0092] If an angle correcting lens 2op is disposed between the optical semiconductor element (inspection object 900) and the detector 2, correction of the angle of incidence becomes unnecessary.

[0093] If the detector is a detector bundle 2G that bundles together detectors at different distances from the emission surface of the optical semiconductor element (test object 900), the objective variable can be reduced by omitting movement in the bundled direction (overlapping detectors 2), allowing for faster regression calculations.

[0094] The system is provided with a camera 2C that captures the light intensity distribution of the emitted light, and the optical axis adjustment unit 4 narrows down the range of the search calculation based on the light intensity distribution information of the emitted light acquired from the camera 2C, thereby reducing the number of calculations and enabling faster regression calculations.

[0095] Alternatively, a phosphor 2L that receives the emitted light and a camera 2C that photographs the light-emitting state of the phosphor 2L may be provided, and the optical axis adjustment unit 4 may narrow down the range in the search calculation based on the light emission distribution information of the phosphor 2L obtained from the camera 2C, thereby reducing the number of calculations and enabling faster regression calculations.

[0096] The optical axis adjustment unit 4 is provided with an element camera (camera 6C) that photographs the optical semiconductor element (object under test 900), and the optical axis adjustment unit 4 sets the initial position in the search calculation based on information indicating the position of the optical semiconductor element (object under test 900) obtained from the element camera (camera 6C), thereby reducing the number of calculations and enabling high-speed regression calculations.

[0097] The optical axis adjustment unit 4 is provided with laser displacement meters (one-dimensional laser displacement meter 6D, two-dimensional laser displacement meter 6D2) for measuring the position of the optical semiconductor element (object under test 900), and even if the optical axis adjustment unit 4 sets the initial position in the search calculation based on information indicating the position of the optical semiconductor element (object under test 900) obtained from the laser displacement meter, the number of calculations can be reduced and regression calculations can be performed at high speed.

[0098] The object under test 900 is provided with an object under test operating power supply 7 that operates a plurality of optical semiconductor elements 900a, 900b individually, and the optical axis adjustment unit 4 controls the object under test operating power supply 7 so that the plurality of optical semiconductor elements 900a, 900b emit light at mutually different timings. If the result of a search calculation for one of the plurality of optical semiconductor elements 900a, 900b is used for a search calculation for the other element, the number of calculations can be reduced and regression calculations can be performed at high speed.

[0099] 1: Optical property inspection device, 2: Detector, 2C: Camera, 2G: Detector bundle, 2L: Phosphor, 2op: Lens, 3: Detector drive device, 3C: Camera moving device, 3G: Detector bundle moving device, 4: Optical axis adjustment unit, 41: Optimum position search unit, 42: Variable processing unit, 43: Convergence determination unit, 44: Re-search position setting unit, 5: Photoelectric converter, 6: Inspection object fixing jig, 6C: Camera (element camera), 6D: One-dimensional laser displacement meter (laser displacement meter), 6D2: Two-dimensional laser displacement meter (laser displacement meter), 6M: Inspection object drive mechanism, 7: Inspection object operating power supply, 8: Inspection control unit, 900: Inspection object (optical semiconductor element), Ip: Position information, Scd: Drive signal, Se: Electrical signal.

Claims

1. An optical property inspection device comprising: a detector that receives light emitted from an optical semiconductor element; a drive unit that drives and positions the measurement position of either the optical semiconductor element or the detector; and an optical axis adjustment unit that controls the drive of the drive unit and, each time positioning is performed, performs a search calculation for the position among the measurement positions at which the maximum amount of light is obtained from the optical semiconductor element based on position information indicating the measurement position obtained from the drive unit and light amount information output from the detector, and adjusts the optical axis of the optical semiconductor element, wherein the optical axis adjustment unit performs the search calculation using a regression method with nonlinear modeling, taking into account the amount of positional deviation caused by the drive unit.

2. The optical characteristic inspection device according to claim 1, characterized in that the optical axis adjustment unit is provided with a convergence determination unit that determines the convergence state of the search calculation, and a re-search position setting unit that sets a re-search position when the convergence determination unit determines that convergence has occurred to a certain extent.

3. The optical property inspection device described in claim 2, characterized in that the re-search position setting unit sets the re-search position from among multiple positions that may show the maximum light amount by weighting using variance or by allocating using random numbers.

4. An optical characteristic inspection device according to any one of claims 1 to 3, characterized in that an angle correction lens is disposed between the optical semiconductor element and the detector.

5. An optical characteristic inspection device according to any one of claims 1 to 4, characterized in that the detector comprises a detector bundle in which detectors at different distances from the light emitting surface of the optical semiconductor element are bundled together.

6. An optical property inspection device according to any one of claims 1 to 5, characterized in that it is provided with a camera that photographs the light intensity distribution of the emitted light, and the optical axis adjustment unit narrows down the range in the search calculation based on the light intensity distribution information of the emitted light obtained from the camera.

7. An optical property inspection device according to any one of claims 1 to 5, characterized in that it comprises a phosphor that receives the emitted light, and a camera that photographs the light emission state of the phosphor, and the optical axis adjustment unit narrows down the range in the search calculation based on light emission distribution information of the phosphor obtained from the camera.

8. An optical property inspection device according to any one of claims 1 to 7, characterized in that it is provided with an element camera that photographs the optical semiconductor element, and the optical axis adjustment unit sets an initial position in the search calculation based on information indicating the position of the optical semiconductor element obtained from the element camera.

9. An optical property inspection device according to any one of claims 1 to 8, characterized in that it is provided with a laser displacement meter that measures the position of the optical semiconductor element, and the optical axis adjustment unit sets an initial position in the search calculation based on information indicating the position of the optical semiconductor element obtained from the laser displacement meter.

10. An optical property inspection device according to any one of claims 1 to 9, characterized in that the device under test is provided with an operating power supply for operating the plurality of optical semiconductor elements individually, the optical axis adjustment unit controls the operating power supply for the device under test so that the plurality of optical semiconductor elements emit light at different timings, and the result of the search calculation for one element of the plurality of optical semiconductor elements is used in the search calculation for the other elements.

Images