High-precision measurement system and calibration method

Abstract

To provide a high-precision measurement system and a calibration method.SOLUTION: A high-precision measurement system includes: a data collection circuit for acquiring a plurality of pieces of first output data corresponding to a plurality of pieces of first setting data; a machine learning circuit which is used for creating a machine learning model on the basis of the plurality of pieces of first setting data, the plurality of pieces of first output data, and a plurality of first calibration parameters between the plurality of pieces of first setting data and the plurality of pieces of first output data, and which is used for generating a second calibration parameter corresponding to second setting data on the basis of the machine learning model; and an output circuit for calibrating second output data corresponding to second input data to generate calibrated output data on the basis of the second calibration parameter.SELECTED DRAWING: Figure 1

Description

[[Technical Field]]

[0001] The present invention relates to a high-precision measurement system and a calibration method, and particularly relates to a high-precision measurement system and a calibration method that perform calibration via a machine learning model. [[Background Art]]

[0002] Before shipment, electronic devices or electronic components need to be electrically measured using an electronic measurement system. That is, the electronic measurement system provides the electrical energy required by the electronic device or electronic component, and verifies whether the function of the electronic device or electronic component is normal based on a feedback signal from the electronic device or electronic component.

[0003] However, due to factors such as hardware circuits, ambient temperature and ambient humidity, there is an error between the actual output electrical energy and the set output electrical energy of the electronic measurement system, so it is necessary to calibrate the actual output electrical energy of the electronic measurement system by measuring the feedback signal to make the actual output electrical energy of the electronic measurement system match the set output electrical energy.

[0004] As a conventional calibration method, when there is a linear relationship between the actual output electrical energy and the set output electrical energy, the point-slope method can be used to calibrate the actual output electrical energy. For example, Patent Document 1 discloses a calibration device for a DC source used for IC testing. This device calibrates the DC source for IC testing by adjusting the conversion gain and offset of an amplifier based on the potential difference generated by a current measuring resistor. However, in practice, there is no linear relationship between the actual output electrical energy and the set output electrical energy, so optimal calibration cannot be achieved if calibration is performed using the point-slope method. (Patent Document 1 is Japanese Unexamined Patent Publication No. 63-082382.) [[Summary of the Invention]] [[Means for Solving the Problem]]

[0005] The purpose of this document is to provide a simplified summary of the disclosure so that readers can understand the basic content of this disclosure. This document is not a complete overview of the disclosure and is not intended to identify key elements of the embodiments of the invention or to limit the scope of the invention.

[0006] One technical aspect of the present invention discloses a high-precision measurement system comprising: a data acquisition circuit for acquiring a plurality of first output data corresponding to a plurality of first setting data; a machine learning circuit coupled to the data acquisition circuit and used to create a machine learning model based on a plurality of first setting data, a plurality of first output data, and a plurality of first calibration parameters between the plurality of first setting data and the plurality of first output data, and used to generate second calibration parameters corresponding to second setting data based on the machine learning model; and an output circuit coupled to the machine learning circuit and used to calibrate second output data corresponding to second setting data based on the second calibration parameters and generate calibrated output data.

[0007] Another technical aspect of the present invention relates to a calibration method applicable to a high-precision measurement system, comprising the steps of: acquiring a plurality of first setting data, a plurality of first output data, and a plurality of first calibration parameters between the plurality of first setting data and the plurality of first output data; creating a machine learning model based on the plurality of first setting data, a plurality of first output data, and a plurality of first calibration parameters; generating second calibration parameters corresponding to second setting data based on the machine learning model; and calibrating second output data corresponding to second setting data based on the second calibration parameters to generate post-calibration output data. [Brief explanation of the drawing]

[0008] To make the above and other objectives, features, advantages, and embodiments of the present invention clearer and easier to understand, the accompanying drawings are described below. [Figure 1] This is a schematic diagram of a high-precision measurement system according to some embodiments of the present invention. [Figure 2] This is a flowchart of a calibration method according to several embodiments of the present invention. [Figure 3] This is a schematic diagram of a calibration method according to some embodiments of the present invention. [Figure 4] These are schematic diagrams of machine learning circuits according to several embodiments of the present invention. According to general working methods, the various features and elements in the drawings are not drawn to a fixed scale; rather, the drawing format is intended to best represent the specific features and elements related to the present invention. Furthermore, in different drawings, the reference numbers of the same or similar elements refer to similar elements / components. [Modes for carrying out the invention]

[0009] To provide a more detailed and complete description of the present disclosure, illustrative descriptions of embodiments and specific examples of the present invention are provided below, but this is not the only form of carrying out or using specific examples of the present invention. The embodiments include features of several specific examples, and method steps and sequences for constructing and operating these specific examples. However, the same or equivalent functions and process sequences can also be achieved by utilizing other specific examples.

[0010] Unless otherwise defined herein, scientific and technical terms used herein have the same meanings as those understood and commonly used by those skilled in the art. Furthermore, where the context does not contradict the use of singular nouns, singular nouns are used herein with their plural forms, and plural nouns are used with their singular forms.

[0011] Furthermore, as used herein, "coupling" may refer to two or more elements being in direct physical or electrical contact with each other, or indirectly in physical or electrical contact with each other, or to two or more elements being in operation or acting with each other.

[0012] Figure 1 is a schematic diagram of a high-precision measurement system 10 according to some embodiments of the present invention. As shown in the figure, the high-precision measurement system 10 includes a data acquisition circuit 110, a machine learning circuit 130, and an output circuit 150. In terms of connections, the machine learning circuit 130 is coupled to the data acquisition circuit 110, and the output circuit 150 is coupled to the machine learning circuit 130. The operation method of the high-precision measurement system 10 will be described below with reference to Figure 2.

[0013] It should be noted that the embodiments of the present invention are not limited to the structure and operation shown in Figure 1, which is merely illustrative to illustrate one of the embodiments of the present invention in order to facilitate understanding of the technology of the present invention, and the scope of the present invention shall be based on the scope of the patent application. Any modifications and alterations made by a person skilled in the art to the embodiments of the present invention without departing from the spirit of the present invention shall still be included within the scope of the patent application.

[0014] Figure 2 is a flowchart of calibration method 200 according to several embodiments of the present invention. To facilitate understanding of calibration method 200, please also refer to Figure 1, and the detailed process of calibration method 200 will be described later.

[0015] In step S210, multiple first setting data, multiple first output data, and multiple first calibration parameters between the multiple first setting data and the multiple first output data are acquired. In some embodiments, step S210 is performed by the machine learning circuit 130 shown in Figure 1.

[0016] Refer to Figure 1 as well. In one embodiment, the output circuit 150 of the high-precision measurement system 10 generates an output voltage VT based on first setting data and outputs it to the device under test 900. For example, in one embodiment, the user operates the operation interface (not shown) of the high-precision measurement system 10 and sets the first setting data to 10 volts, expecting the output circuit 150 of the high-precision measurement system 10 to output an output voltage VT of 10 volts to the device under test 900. However, the actual output voltage VT output by the output circuit 150 may differ from the setting data of 10 volts due to factors such as the internal circuitry, ambient temperature, and humidity, meaning that the actual output voltage is not 10 volts. The above output voltage VT is output data that is generated and output in accordance with the setting data.

[0017] For example, in one embodiment, the output voltage VT generated and output by the output circuit 150 based on the setting data of 10 volts is 9.7 volts. The output voltage VT generated and output by the output circuit 150 based on the setting data of 20 volts is 20.3 volts. The data acquisition circuit 110 collects the above setting data and the corresponding output data (output voltage VT).

[0018] In some embodiments, the high-precision measurement system 10 generates and outputs multiple output voltages VT with different voltage values ​​based on multiple different first setting data. The voltage values ​​of the output voltages VT acquired by the data acquisition circuit 110 are then acquired by an external circuit (not shown) and fed back to the data acquisition circuit 110 via the external circuit.

[0019] Next, the machine learning circuit 130 acquires the above-mentioned multiple different first setting data and the corresponding multiple first output data, and then calculates multiple first calibration parameters between the setting data of different voltage values ​​and the corresponding multiple first output data using a point-oblique calibration method. In some embodiments, the multiple first calibration parameters include multiple gain values ​​and multiple offset values.

[0020] In some embodiments, after calibration of the calibration parameter, the generated calibrated output data is the same as the setting data. That is, when the setting data is 10 volts, the calibrated output data is also 10 volts.

[0021] In some embodiments, the calculation formula for the point-slope method calibration is the following formula (1).

[0022] Formula (1): Y=AX+B.

[0023] In the above formula (1), Y is calibrated output data, X is first output data, A is a gain value, and B is an offset value. Through point-slope method calibration, output data can be calibrated to calibrated output data.

[0024] The above formula (1) is merely for illustrative explanation. In some other embodiments, the first calibration parameter may include a plurality of calibration formulas, a plurality of gain values and a plurality of offset values.

[0025] In step S230, a machine learning model is created based on a plurality of first setting data, a plurality of first output data and a plurality of first calibration parameters. In some embodiments, step S230 is executed by the machine learning circuit 130 in FIG. 1.

[0026] In some embodiments, the machine learning model includes a lookup table stored in a memory (not shown).

[0027] In one embodiment, the first calibration parameters (including gain values and offset values) and calibrated output data calculated by the machine learning circuit 130 based on a plurality of different first setting data and corresponding first output data are shown in the following lookup table 1.

[0028] Lookup Table 1 TIFF0007917588000001.tif51157

[0029] In the lookup table 1 above, the first output data consists of multiple first setting data acquired by the data acquisition circuit 110 of the high-precision measurement system 10 and the first output data corresponding to the first setting data. As shown in the lookup table 1 above, after the gain value and offset value are calibrated based on the above formula (1), the first output data is calibrated to the calibrated output data, so that the calibrated output data becomes the same as the first setting data.

[0030] In some embodiments, the error between the first setting data and the first output data is nonlinear, and therefore the first calibration parameters corresponding to different first setting data and first output data are also different.

[0031] In another embodiment, the lookup table 1 is a first setting data, a first output data, and corresponding first calibration parameters obtained when the output current of the high-precision measurement system 10 is a first current value (e.g., 5 amperes). In another embodiment, the machine learning circuit 130 can create another lookup table based on the first setting data, a first output data, and corresponding first calibration parameters obtained when the output current of the high-precision measurement system 10 is a second current value (e.g., 10 amperes). In other words, the setting data and corresponding output data may be voltage values / current values / power values, etc. The lookup table 1 described above is explained using only voltage values ​​as an example.

[0032] In some embodiments, when creating a machine learning model, the machine learning circuit 130 is used to create a machine learning model by generating a lookup table based on multiple environmental data acquired by the data acquisition circuit 110, and first calibration parameters (including gain values ​​and offset values) calculated based on multiple different first setting data and corresponding first output data, as well as the environmental data.

[0033] For example, let's assume that the above lookup table 1 is created at an ambient temperature of 27°C and an ambient humidity of 70%. At a different ambient temperature and humidity, the machine learning circuit 130 calculates calibration parameters based on the first setting data and the corresponding first output data obtained at yet another ambient temperature and humidity, and creates another lookup table.

[0034] In this way, by creating multiple lookup tables, the machine learning circuit 130 creates a machine learning model. Furthermore, the lookup table 1 described above is a lookup table for the output voltage VT. In some other embodiments, the lookup table created by the machine learning circuit 130 may be a lookup table created based on the output current setting data and output data of the high-precision measurement system 10, or it may be based simultaneously on the output current and output voltage setting data and output data of the high-precision measurement system 10, or it may be a lookup table created based on the input current setting data and output data of the high-precision measurement system 10, or it may be a lookup table created based on the input voltage setting data and output data of the high-precision measurement system 10. The embodiments of the present invention are not limited to those described above.

[0035] In step S250, a second calibration parameter corresponding to the second setting data is generated based on a machine learning model. In some embodiments, step S250 is performed by the machine learning circuit 130 shown in Figure 1.

[0036] In some embodiments, in step S250, after creating a machine learning model, the machine learning circuit 130 inputs second setting data into the machine learning model to obtain second calibration parameters corresponding to the second setting data. In some embodiments, the second setting data is data set by the user via an operation interface (not shown) and input into the high-precision measurement system 10.

[0037] In some embodiments, the machine learning circuit 130 obtains at least two of the first calibration parameters from a lookup table of the machine learning model based on second setting data, and obtains a second calibration parameter using an interpolation method.

[0038] In one embodiment, the machine learning circuit 130 obtains at least two first setting data points that are closest to the second setting data points from the lookup table of the machine learning model based on the second setting data points, and obtains a second calibration parameter in the form of an interpolation method based on the first calibration parameter corresponding to at least two first setting data points.

[0039] For example, refer to lookup table 1. In process S250, if the second setting data is 15 volts, the machine learning circuit 130 obtains that the first setting data closest to the second setting data are 10 volts and 20 volts. Subsequently, the machine learning circuit 130 calculates the gain value corresponding to the second setting data of 15 volts in the form of interpolation. The filename is TIFF0007917588000002.tif1011, and the offset value corresponding to the second setting data for 15 volts is It can be determined that the file is TIFF0007917588000003.tif911.

[0040] Furthermore, refer to Figure 3, for example. Figure 3 is a schematic diagram of a calibration method according to some embodiments of the present invention. In Figure 3, G(I1, V2) is the gain value when the current value in the first setting data is I1 and the voltage value is V2, G(I2, V2) is the gain value when the current value in the first setting data is I2 and the voltage value is V2, G(I1, V1) is the gain value when the current value in the first setting data is I1 and the voltage value is V1, and G(I2, V1) is the gain value when the current value in the first setting data is I2 and the voltage value is V1.

[0041] The above gain values ​​G(I1, V2), G(I2, V2), G(I1, V1), and G(I2, V1) are gain values ​​obtained by the machine learning circuit 130 based on the lookup table in the machine learning model, and the voltage and current values ​​for the above gain values ​​G(I1, V2), G(I2, V2), G(I1, V1), and G(I2, V1) and the voltage value V* and current value I* for the calculated gain value G(I*, V*) are at least two of the closest values ​​in the lookup table. That is, the current values ​​I1 and I2 are the current values ​​closest to the current value I* in the lookup table, and the voltage values ​​V1 and V2 are the voltage values ​​closest to the voltage value V* in the lookup table.

[0042] Next, the machine learning circuit 130 determines the gain value of the second setting data (i.e., current value I* and voltage value V*) based on the above gain values ​​G(I1, V2), G(I2, V2), G(I1, V1), and G(I2, V1) and the following equations (2) to (5). Retrieve TIFF0007917588000004.tif716.

[0043] Formula (2): TIFF0007917588000005.tif840

[0044] Formula (3): TIFF0007917588000006.tif1083

[0045] Formula (4): TIFF0007917588000007.tif982

[0046] Formula (5): TIFF0007917588000008.tif966

[0047] In step S270, the second output data corresponding to the second setting data is calibrated based on the second calibration parameter to generate calibrated output data. In some embodiments, step S270 is performed by the output circuit 150 in Figure 1.

[0048] In some embodiments, the machine learning circuit 130 acquires a second calibration parameter corresponding to the second setting data, and then the output circuit 150 calibrates the second output data based on the second calibration parameter to generate calibrated output data.

[0049] In some embodiments, the machine learning circuit 130 calibrates the second output data using the point-diagonal method formula (for example, formula (1) above).

[0050] For example, the gain value corresponding to the second setting data of 15 volts acquired by the machine learning circuit 130 in process S250 is The filename is TIFF0007917588000009.tif911, and the offset value corresponding to the second setting data for 15 volts is If the filename is TIFF0007917588000010.tif911, the machine learning circuit 130 calculates the calibrated output data based on equation (1) as shown in the following equation. TIFF0007917588000011.tif1147

[0051] Y2 is the output data after calibration, and X2 is the second output data generated by the output circuit 150 in the case of non-calibration, corresponding to the second setting data of 15 volts. TIFF0007917588000012.tif911 is a gain value, TIFF0007917588000013.tif910 is an offset value. Using the point-oblique calibration method, the second output data can be calibrated so that the calibrated output data matches the set data.

[0052] In some embodiments, the output circuit 150 calibrates the second output data in PWM (pulse width modulation) form based on a second calibration parameter to generate calibrated output data. Various forms of calibrating the second output data are all within the embodiments of the present invention, and the embodiments of the present invention are not limited to PWM.

[0053] It should be noted that the present invention is not limited to the process shown in Figure 2, which is merely illustrative to illustrate one embodiment of the present invention in order to facilitate understanding of the technology of the present invention, and the scope of the present invention shall be based on the scope of the patent application. Modifications and alterations made by persons skilled in the art to embodiments of the present invention without departing from the spirit of the present invention are still included within the scope of the patent application.

[0054] See also Figure 4. Figure 4 is a schematic diagram of a machine learning circuit 130 according to some embodiments of the present invention. As shown in Figure 4, the machine learning circuit 130 includes an analog-to-digital conversion circuit 131, a lookup table processing circuit 135, and a calibration circuit 137. The lookup table processing circuit 135 is coupled to the analog-to-digital conversion circuit 131, and the calibration circuit 137 is coupled to the lookup table processing circuit 135.

[0055] In some embodiments, the analog-to-digital conversion circuit 131 is used to convert the output data acquired by the data acquisition circuit 110 from analog data to digital data. The lookup table processing circuit 135 is used to input the output data converted to digital data into a machine learning model to obtain calibration parameters. In some embodiments, the lookup table processing circuit 135 obtains calibration parameters by performing interpolation calculations using the moving average method.

[0056] The calibration circuit 137 is used to generate a PWM control signal based on calibration parameters and to transfer the PWM control signal to the output circuit 150 shown in Figure 1, so that the output circuit 150 generates a calibrated output signal based on the PWM control signal.

[0057] In some embodiments, the high-precision measurement system 10 further includes memory (not shown) for storing machine learning models and lookup tables for access by the machine learning circuit 130.

[0058] As can be seen from the embodiments of the present invention described above, applying the present invention offers the following advantages. The high-precision measurement system and calibration method shown in the embodiments of the present invention calculate calibration parameters in response to changes in different environmental variables and circuit variables in the form of a machine learning model, and calibrate the output data to post-calibration output data, so that the data actually output by the high-precision measurement system is the same as the set data. Furthermore, the embodiments of the present invention can continuously learn, update, and modify calibration parameters in the form of a machine learning model. The above embodiments can quickly acquire calibration parameters and effectively reduce calibration errors.

[0059] In some embodiments, the machine learning circuit 130 described above may be integrated into an electronic device or electronic system. In some embodiments, the server may be implemented as a cloud server including a central processing unit. In some embodiments, the server may be a single processor or a collection of multiple microprocessors, but is not limited thereto, and includes memory and input / output circuits.

[0060] In some embodiments, the machine learning circuit 130 may be a central processor unit (CPU), a microprocessor (MCU), a digital signal processor (DSP), a field-programmable gate array (FPGA), a server, or other arithmetic circuits, processing circuits, or elements having data access, data computation, data storage, data transmission / reception, or similar functions.

[0061] In some embodiments, the machine learning circuit 130 includes a processor and an input / output circuit. The processor may be a circuit or element having data access, data computation, data storage, or similar functions. The input / output circuit may be a circuit or element having data transmission and reception, or similar functions.

[0062] In some embodiments, the data acquisition circuit 110 may be a current detection circuit, a voltage detection circuit, a temperature detection circuit, a humidity detection circuit, or other circuit or element having a similar function of data detection, data transmission / reception, or similar function. In some embodiments, the data acquisition circuit 110 acquires input data and environmental data via an external current detection circuit, a voltage detection circuit, a temperature detection circuit, a humidity detection circuit, or other circuit or element having a similar function.

[0063] In some embodiments, the high-precision measurement system 10 further includes a display circuit (not shown) for displaying setting data, output data, and post-calibration output data.

[0064] In some embodiments, the output circuit 150 may be a PWM drive circuit, or other circuit or element having current output / voltage output / power output, or other circuit or element having the same or similar function.

[0065] Specific embodiments of the present invention are disclosed in the above embodiments, but these are not intended to limit the invention. Those skilled in the art can make various modifications and alterations without departing from the principles and spirit of the invention. Therefore, the scope of protection of the present invention should be limited to that defined by the appended scope of the patent application. [Explanation of Symbols]

[0066] 10: High-precision measurement system 110: Data acquisition circuit 130: Machine Learning Circuits 150: Output circuit VT: Output voltage 900: Device under test 200: Calibration method S210, S230, S250, S270: Process G(I1, V2), G(I2, V2), G(I1, V1), G(I2, V1): Gain values G(I*, V*), G1, G2: Gain values I1, I2, I3, I4, I*: Current values V1, V2, V3, V4, V*: Voltage values 131: Analog-to-digital conversion circuit 135: Lookup Table Processing Circuit 137: Calibration circuit

Claims

1. A data acquisition circuit for acquiring a plurality of first setting data and a plurality of first output data actually generated and output by an output circuit based on the plurality of first setting data, wherein the plurality of first setting data and the plurality of first output data are voltage values ​​or current values, A machine learning circuit coupled to the data acquisition circuit calculates a plurality of first calibration parameters based on the plurality of first output data and the plurality of first setting data, wherein the plurality of first calibration parameters include a plurality of first gain values ​​and a plurality of first offset values, and the calibration output data obtained by calibrating the plurality of first output data by point oblique calibration using the first gain values ​​and the first offset values ​​is identical to the corresponding first setting data. The machine learning circuit creates a lookup table as a machine learning model, and the lookup table records the plurality of first setting data, as well as the first gain value and the first offset value. The machine learning circuit further receives second setting data, obtains corresponding second calibration parameters from the lookup table based on the second setting data, and the second calibration parameters include a second gain value and a second offset value. The output circuit coupled to the machine learning circuit, which actually generates and outputs second output data based on the second setting data, and calibrates the second output data by point-oblique calibration based on the second gain value and the second offset value to generate calibrated output data that is the same as the second setting data, A high-precision measurement system including [specific components / features].

2. The high-precision measurement system according to claim 1, wherein the machine learning circuit is further used to obtain at least two of the first setting data closest to the second setting data from the lookup table based on the second setting data, and to obtain the second gain value and the second offset value by interpolation based on the first gain value and the first offset value corresponding to the at least two first setting data.

3. The high-precision measurement system according to claim 1, wherein the data acquisition circuit is used to acquire a plurality of environmental data, including ambient temperature and ambient humidity, and the machine learning circuit is used to create the lookup tables corresponding to different ambient temperatures and ambient humidity.

4. A calibration method applicable to high-precision measurement systems, A step of acquiring a plurality of first setting data and a plurality of first output data actually generated and output by an output circuit based on the plurality of first setting data, wherein the plurality of first setting data and the plurality of first output data are voltage values ​​or current values. A step of calculating a plurality of first calibration parameters based on the plurality of first setting data, wherein the plurality of first calibration parameters include a plurality of first gain values ​​and a plurality of first offset values, and the calibration output data obtained by calibrating the plurality of first output data by point oblique calibration using the first gain values ​​and the first offset values ​​is made identical to the corresponding first setting data. A step of creating a lookup table as a machine learning model, wherein the lookup table records the plurality of first setting data, and the first gain value and the first offset value corresponding to each of the first setting data, A step of receiving second setting data and obtaining corresponding second calibration parameters from the lookup table based on the second setting data, wherein the second calibration parameters include a second gain value and a second offset value. The output circuit actually generates and outputs second output data based on the second setting data, and the second output data is calibrated by the point-oblique calibration method based on the second gain value and the second offset value to generate calibrated output data that is the same as the second setting data. A calibration method that includes this.

5. The calibration method according to claim 4, further comprising the steps of obtaining at least two first setting data from the lookup table that are closest to the second setting data based on the second setting data, and obtaining the second gain value and the second offset value by interpolation based on the first gain value and the first offset value corresponding to the at least two first setting data.

6. The calibration method according to claim 4, further comprising the step of acquiring a plurality of environmental data including ambient temperature and ambient humidity, and creating the lookup tables corresponding to the plurality of first setting data, different ambient temperatures, and ambient humidity.

Images