WO2021096076A1

WO2021096076A1 - Apparatus and method for testing semiconductor device by using frequency output properties of semiconductor device and machine learning

Info

Publication number: WO2021096076A1
Application number: PCT/KR2020/014011
Authority: WO
Inventors: 김규태; 이국진; 남상진
Original assignee: 고려대학교 산학협력단
Priority date: 2019-11-14
Filing date: 2020-10-14
Publication date: 2021-05-20
Also published as: KR102225817B1

Abstract

The present invention relates to a technology for testing a semiconductor device by means of machine learning of the frequency output properties of a semiconductor device, and more specifically relates to technology that increases the test accuracy of a determination of the condition of a semiconductor device by measuring the waves of a current or voltage output via a semiconductor device, applying an MFCC-based filter to the measured waves, and thereby constructing a database using two dimensional array data of vector data, and carrying out machine learning with respect to the constructed database.

Description

Apparatus and method for testing semiconductor devices using frequency output characteristics and machine learning of semiconductor devices

The present invention relates to a technical idea of testing a semiconductor device by machine learning the frequency output characteristics of a semiconductor device, and more particularly, measuring a wave of current or voltage output through a semiconductor device, and measuring the wave of the measured wave by MFCC (Mel Frequency Cepstral Coefficient) A technology for determining and testing conditions of semiconductor devices by applying a similar filter.

Recently, as the size of the semiconductor device decreases and the difficulty of process and design increases, the problem of reliability has emerged, and discussions on how to determine the degree of deterioration of the semiconductor device are continuously being discussed.

Recently, numerous electronic products from smartphones to smart cars have been found in our lives.

Accordingly, various semiconductor devices in a product have their respective life expectancy, and a product supplier must be able to guarantee the normal operation of the product and the semiconductor device within the expected life span.

If normal operation within the expected life span is not guaranteed, it not only incurs additional costs, but also negatively affects the image of the product of the customer and the image of the company, thus affecting the reliability of the product and the degree of deterioration of the semiconductor device. The degree to which it follows is essential.

In particular, considering the effect of aging of the product on the performance of the semiconductor device due to the passage of time, electrical stress, or temperature change, etc., it is very important to ensure the normal operation of the semiconductor device within the expected life of the semiconductor device. It is essential.

In addition, various problems that occur as the size of a semiconductor device decreases have a great influence on the degradation of the device, and it is essential to accurately grasp this and classify and specify the degree of deterioration of the device.

Depending on the conditions of the semiconductor device (e.g., chemical material change, device size change, temperature, process equipment, process conditions, stress, metal material), the degree of stress and deterioration applied to the semiconductor device varies. It is difficult to accurately grasp the electrical characteristics of the device.

In other words, it is difficult to check the device information on the condition with only a simple measurement method. In particular, as semiconductor devices are repeatedly manufactured and scaled, it is difficult to accurately grasp them.

The present invention converts the fluctuation of a current or voltage signal of a semiconductor device into frequency domain data using a Mel Frequency Cepstral Coefficient (MFCC) similar filter, converts the converted data into a database, and converts the converted data into a database. The purpose of this study is to determine the conditions of semiconductor devices by performing machine learning.

The present invention performs machine learning to determine temperature, dimension, process conditions, channel material, gate material, degree of degradation, electrical properties, reliability, and interface traps. ), bulk trap, light emission degree, light emission wavelength, or lifespan.

The present invention divides frequency domain data into a plurality of frames, converts it into 2D array data to build a database, and improves test accuracy for semiconductor devices by performing machine learning on the built database. It is aimed at.

An object of the present invention is to improve test accuracy for a semiconductor device regardless of the material or size of the semiconductor device by using frequency domain data on the output of the semiconductor device.

According to an embodiment of the present invention, the semiconductor device test apparatus measures a current or voltage signal output from the test device based on the source application unit for applying a current or voltage to the device under test, and the applied current or voltage. , An amplification conversion unit for amplifying and converting the measured current or voltage signal into a machine learning signal, converting the amplified-converted machine learning signal into vector data having a vector form with respect to time, and converting the converted vector data into a plurality of frames A data conversion unit that divides into (frame) and converts it into 2D array data, and a machine learning unit that determines a condition of the device to be tested by performing machine learning on the 2D array data, wherein the measured current or voltage The signal may have a frequency form according to the condition of the device under test.

The data converter may calculate a power spectral density by applying a Fourier transform to the divided frames.

The data converter may convert the vector data into the 2D array data by applying a discrete cosine transform using at least one filter to the calculated power spectral density.

The data conversion unit analyzes the frequency shape using at least one filter, and converts the converted vector data into the two-dimensional array data each including conditions of the device under test based on the analyzed frequency shape. can do.

The number of the at least one filter may be proportional to the determination accuracy of the condition of the device under test.

The machine learning signal has a text file format divided into a first column and a second column, the first column represents data according to time, and the second column is fluctuation data according to the change of time. Can represent.

The machine learning unit performs the machine learning to capture temperature, dimensions, process conditions, channel materials, gate materials, degradation degrees, electrical properties, reliability, and interfacial capture. Conditions of the device to be tested including at least one of (interface trap), bulk trap, degree of emission, emission wavelength, and lifetime may be determined.

The machine learning unit may divide and learn the 2D array data into a plurality of folders, and then perform cross validation on the plurality of folders.

According to an exemplary embodiment of the present invention, a method for testing a semiconductor device includes measuring a current or voltage signal output from the device under test based on a current or voltage applied to the device under test in an amplifying converter, and measuring the measured current or Amplifying and converting a voltage signal into a machine learning signal, in a data conversion unit, converting the amplified converted machine learning signal into vector data having a vector form with respect to time, and converting the converted vector data into a plurality of frames Dividing into two-dimensional array data, and determining a condition of the device under test by performing machine learning on the two-dimensional array data in a machine learning unit, wherein the measured current or voltage signal May have a frequency shape according to the condition of the device under test.

The step of dividing the transformed vector data into a plurality of frames and converting it into 2D array data includes applying a Fourier transform to the divided plurality of frames to achieve power spectrum density. Calculating spectral density) and applying a discrete cosine transform using at least one filter to the calculated power spectral density to transform the transformed vector data into the 2D array data It may include the step of.

The present invention converts the fluctuation of a current or voltage signal of a semiconductor device into frequency domain data using a Mel Frequency Cepstral Coefficient (MFCC) similar filter, converts the converted data into a database, and converts the converted data into a database. By performing machine learning on the semiconductor device, the conditions of the semiconductor device can be determined.

The present invention performs machine learning to determine temperature, dimension, process conditions, channel material, gate material, degree of degradation, electrical properties, reliability, and interface traps. ), a bulk trap, a degree of light emission, a light emission wavelength, or a condition of a semiconductor device including at least one of a lifetime, thereby improving test accuracy for the semiconductor device.

The present invention divides frequency domain data into a plurality of frames and converts it into 2D array data to build a database, and improves test accuracy for semiconductor devices by performing machine learning on the built database. I can.

According to the present invention, by using frequency domain data on the output of the semiconductor device, it is possible to improve test accuracy for the semiconductor device regardless of the material or size of the semiconductor device.

1 is a diagram illustrating an apparatus for testing a semiconductor device according to an embodiment of the present invention.

2 is a diagram illustrating an operation procedure of a semiconductor device testing apparatus according to an embodiment of the present invention.

3 is a diagram illustrating a process of converting a machine learning signal in a semiconductor device testing apparatus according to an embodiment of the present invention.

4 is a diagram illustrating a 2D array data conversion procedure of a semiconductor device testing apparatus according to an embodiment of the present invention.

5 to 7 are diagrams illustrating a machine learning procedure of a semiconductor device testing apparatus according to an embodiment of the present invention.

8 is a diagram illustrating the efficiency of a filter applied to a semiconductor device testing apparatus according to an embodiment of the present invention.

9 is a diagram illustrating a method of testing a semiconductor device according to an embodiment of the present invention.

10 and 11 are diagrams illustrating an operating procedure of a semiconductor device test apparatus according to various embodiments of the present disclosure.

12A to 13 are diagrams for explaining classification accuracy of a semiconductor device testing apparatus according to an embodiment of the present invention.

14A to 14C are views for explaining measurement results of a current or voltage wave output through a semiconductor device of a semiconductor device testing apparatus according to an embodiment of the present invention.

Specific structural or functional descriptions of embodiments according to the concept of the present invention disclosed in the present specification are exemplified only for the purpose of describing embodiments according to the concept of the present invention, and embodiments according to the concept of the present invention They may be implemented in various forms and are not limited to the embodiments described herein.

Since the embodiments according to the concept of the present invention can apply various changes and have various forms, the embodiments will be illustrated in the drawings and described in detail in the present specification. However, this is not intended to limit the embodiments according to the concept of the present invention to specific disclosed forms, and includes changes, equivalents, or substitutes included in the spirit and scope of the present invention.

Terms such as first or second may be used to describe various elements, but the elements should not be limited by the terms. The terms are only for the purpose of distinguishing one component from other components, for example, without departing from the scope of the rights according to the concept of the present invention, the first component may be referred to as the second component, Similarly, the second component may also be referred to as a first component.

When a component is referred to as being "connected" or "connected" to another component, it is understood that it may be directly connected or connected to the other component, but other components may exist in the middle. It should be. On the other hand, when a component is referred to as being "directly connected" or "directly connected" to another component, it should be understood that there is no other component in the middle. Expressions describing the relationship between components, for example, "between" and "just between" or "directly adjacent to" should be interpreted as well.

The terms used in the present specification are only used to describe specific embodiments, and are not intended to limit the present invention. Singular expressions include plural expressions unless the context clearly indicates otherwise. In the present specification, terms such as "comprise" or "have" are intended to designate that the specified features, numbers, steps, actions, components, parts, or combinations thereof exist, but one or more other features or numbers, It is to be understood that the presence or addition of steps, actions, components, parts, or combinations thereof does not preclude the possibility of preliminary exclusion.

Unless otherwise defined, all terms used herein including technical or scientific terms have the same meaning as commonly understood by one of ordinary skill in the art to which the present invention belongs. Terms as defined in a commonly used dictionary should be construed as having a meaning consistent with the meaning of the related technology, and should not be interpreted as an ideal or excessively formal meaning unless explicitly defined in the present specification. Does not.

Hereinafter, embodiments will be described in detail with reference to the accompanying drawings. However, the scope of the patent application is not limited or limited by these embodiments. The same reference numerals shown in each drawing indicate the same members.

1 illustrates the components of a semiconductor device testing apparatus according to an embodiment of the present invention.

Referring to FIG. 1, a semiconductor device testing apparatus according to an embodiment of the present invention includes a source applying unit 110, an amplifying and converting unit 120, a data converting unit 130, and a machine learning unit 140.

The source application unit 110 according to an embodiment of the present invention may apply current or voltage to the device under test.

That is, the source application unit 110 includes a battery box in a shielded environment where external signals are blocked, and the battery box is in a shielded environment where external signals are blocked in order to accurately analyze weak signals. Apply voltage or current to the device under test.

For example, a current or voltage signal applied to the device under test has a specific fluctuation, and the specific vibration can be identified in a frequency domain.

The current or voltage applied by the source applying unit 110 according to an embodiment of the present invention will be described supplementarily with reference to FIG. 3.

The amplification converter 120 according to an embodiment of the present invention measures a current or voltage signal output from the device under test based on the current or voltage applied from the source application unit 110.

In addition, the amplification converter 120 may amplify and convert the measured current or voltage signal into a machine learning signal. Here, the measured current or voltage signal may have a frequency shape according to the condition of the device under test.

As an example, the amplification converter 120 includes a low-noise current/voltage preamplifier, and the low-noise preamplifier reads the micro signal applied from the battery box by python, one of the machine learning programs. It can be amplified and converted into a machine learning signal by amplifying into a possible frequency band.

Here, the amplified machine learning signal is actually composed of classes for machine learning, and time and amplified current are included in each class folder, so that machine learning can be performed in the form of a text file.

According to an embodiment of the present invention, the data conversion unit 130 may convert the amplified-converted machine learning signal into vector data having a vector form with respect to time.

In addition, the data conversion unit 130 may divide the converted vector data into a plurality of frames and convert it into 2D array data.

For example, the data converter 130 may calculate a power spectral density by applying a Fourier transform to a plurality of divided frames.

That is, the present invention divides frequency domain data into a plurality of frames and converts it into 2D array data to build a database, and by performing machine learning on the built database, test accuracy for semiconductor devices is improved. Can be improved.

According to an embodiment of the present invention, the data conversion unit 130 applies a discrete cosine transform using at least one filter to the calculated power spectral density to convert the vector data into a two-dimensional array. Can be converted to data.

For example, at least one filter may include a Mel Frequency Cepstral Coefficient (MFCC) similar transformation filter.

That is, the data conversion unit 130 may analyze the frequency shape using at least one filter and convert the vector data into 2D array data each including conditions of the device under test based on the analyzed frequency shape. have. Here, a procedure for converting to 2D array data will be supplemented with reference to FIG. 4.

According to an embodiment of the present invention, the machine learning unit 140 may determine a condition of a device under test by performing machine learning on 2D array data.

For example, the conditions of the device under test include temperature, dimension, process conditions, channel material, gate material, degree of degradation, electrical properties, reliability, and interfacial trapping ( interface trap), bulk trap, degree of emission, emission wavelength, or lifetime.

That is, the present invention converts the fluctuation of a current or voltage signal of a semiconductor device into frequency domain data using a Mel Frequency Cepstral Coefficient (MFCC) similar filter, converts the converted data into a database, and By performing machine learning on the base, the conditions of the semiconductor device can be determined.

For example, the machine learning unit 140 may divide and learn 2D array data into a plurality of folders, and then cross-validate the plurality of folders.

Cross-validation is supplemented in the description of FIG. 7.

FIG. 2 illustrates a procedure of converting a frequency signal output from a device under test into 2D array data in order to test a device under test by a semiconductor device testing apparatus according to an embodiment of the present invention.

Referring to FIG. 2, in step S201, the semiconductor device testing apparatus applies a voltage or current to the device to be tested.

In step S202, the semiconductor device testing apparatus amplifies the current or voltage signal 210 output from the device under test using a low-noise current/voltage preamplifier.

For example, the current or voltage signal 210 output from the device under test has a specific fluctuation.

Here, the semiconductor device test apparatus amplifies the current or voltage signal and converts it into a machine learning signal so that the current or voltage signal can be recognized in Python.

In step S203, the semiconductor device testing apparatus reads and records the amplified machine learning signal.

In steps S204 and S205, the semiconductor device testing apparatus repeatedly performs normalizing and pre-emphasis on the recorded machine learning signal.

In step S206, the semiconductor device test apparatus divides the frame of the normalized and pre-emphasis frequency signal 220.

In step S207, the semiconductor device testing apparatus applies a window to the divided frequency signal 220, and applies a discrete cosine transform to the frequency signal to which the window is applied, and the divided frequency signal 230 ) To form.

In step S208, the semiconductor device testing apparatus applies a mel and log scale to the divided frequency signal 230 using at least one filter.

In step S209, the semiconductor device testing apparatus generates two-dimensional array data by applying a discrete cosine transform of the mel energy of the frequency signal to which the mel and log scale is applied.

3 illustrates the configuration and frequency characteristics of a current or voltage signal applied from a voltage applying unit of a semiconductor device testing apparatus and output from a device under test according to an embodiment of the present invention.

Referring to FIG. 3, a text configuration 300 and a waveform 310 of an applied signal are shown.

As an example, the text structure 300 is composed of a first column 302 and a second column 304, has a text file format, the first column 302 represents data according to the change of time, and the second Column 304 may represent fluctuation data over time.

The waveform 310 may represent a fluctuation according to a change in time as a change in frequency.

4 illustrates a procedure for converting a current or voltage signal output from a device under test into 2D array data by a semiconductor device testing apparatus according to an embodiment of the present invention.

Referring to FIG. 4, in step S401, the semiconductor device testing apparatus acquires a current or voltage signal output from the device under test.

In step S402, the semiconductor device testing apparatus applies normalizing and pre-emphasis to the acquired current or voltage signal to convert it into generalized vector data, divides the converted vector data into a preset frame size, and divides each frame. For this, a Fourier transform is performed and a power spectral density is calculated.

In step S403, the semiconductor device testing apparatus generates 2D array data by applying a discrete cosine transform using at least one filter to the calculated power spectral density.

For example, the 2D array data is comma-separated values (CSV) data composed of a plurality of rows and columns, and each of the 2D arrays may have a condition of a device under test as a value.

That is, the 2D array data may have a vector sequence data structure in which the condition of the device under test is an array structure.

5 illustrates a procedure for machine learning by using 2D array data by a semiconductor device testing apparatus according to an embodiment of the present invention.

Referring to FIG. 5, in step S501, the semiconductor device testing apparatus sequentially machine learns 2D array data.

In step S502, the semiconductor device testing apparatus performs learning of comparing and comparing the machine-learned data sequentially in step S501 with 2D array data.

In step S503, the semiconductor device testing apparatus identifies conditions of the device to be tested.

That is, the present invention performs machine learning to achieve temperature, dimension, process conditions, channel material, gate material, degree of degradation, electrical properties, reliability, and interfacial capture ( By determining conditions of a semiconductor device including at least one of interface trap), bulk trap, degree of light emission, light emission wavelength, and lifetime, test accuracy for a semiconductor device may be improved.

According to an embodiment of the present invention, the semiconductor device testing apparatus may use a hidden markov model, and it is possible to learn the conditions of the device under test by learning the conditions and classes of the device under test. have.

For example, the semiconductor device testing apparatus according to an embodiment of the present invention has an accuracy of up to 99.3% after learning with Si, doping, gate material, oxidation degree, temperature, channel material, QLED, OLED, luminescence degree, etc. Can be distinguished.

In addition, the semiconductor device testing apparatus according to an embodiment of the present invention may utilize 50 or more classes and 4000 or more data for machine learning.

6 illustrates an output screen when a semiconductor device testing apparatus according to an embodiment of the present invention learns 2D array data using Python.

The accuracy according to the result of the experiment using Python can be at least 85%, and conditions of semiconductor devices can be extracted with higher accuracy through additional parameter control.

In particular, even in various conditions, such as various conditions, substances, and degree of deterioration, not only the accuracy for each condition but also the overall accuracy exceeds 90%, it is possible to very accurately distinguish the state, material, and degree of deterioration of the device.

7 illustrates a procedure of cross validation for a plurality of folders by a semiconductor device testing apparatus according to an embodiment of the present invention.

Referring to FIG. 7, in step S701, the semiconductor device testing apparatus learns data by collecting training data from a folder Fold1 among a plurality of folders. On the other hand, among a plurality of folders, tests and learning are performed in parallel on the folder Fold2 and the folder Fold3.

In step S702, the semiconductor device testing apparatus performs a test on data in a folder Fold1 among a plurality of folders. Meanwhile, training data is collected for folders Fold 2 and Fold 3 among a plurality of folders to learn data.

In step 703, the semiconductor device testing apparatus collects test results for a plurality of folders (Fold1, Fold2, Fold3), and calculates an average accuracy of the collected test results.

That is, the semiconductor device testing apparatus performs a procedure of cross-verifying a plurality of folders.

8 illustrates accuracy efficiency according to the number of filters applied to the semiconductor device testing apparatus according to an embodiment of the present invention.

Referring to FIG. 8, a graph 800 shows a change in the number of filters on a horizontal axis, a change in accuracy on a vertical axis, and shows a high frequency region and a low frequency region as a legend.

According to the graph 800, when the number of filters in the low-frequency region is large, it is considered that the accuracy is large, so that the carrier contributing to the current or voltage is in the low-frequency region. Behavioral features can be extracted.

Accordingly, as the number of filters increases, the test accuracy of the semiconductor device test apparatus may increase.

That is, according to the present invention, by using frequency domain data on the output of the semiconductor device, it is possible to improve the test accuracy of the semiconductor device regardless of the material or size of the semiconductor device.

9 is a method for testing a semiconductor device according to an embodiment of the present invention measures a current or voltage signal having a specific frequency form according to a condition of a device under test, converts it into 2D array data, and converts the converted 2D array data. Illustrates the procedure of learning to accurately grasp the conditions of the device under test.

Referring to FIG. 9, in step 901, the method for testing a semiconductor device amplifies and converts an output signal from a device under test into a machine learning signal.

That is, the semiconductor device test method may measure a current or voltage signal output from the device under test based on the current or voltage applied to the device under test, and amplify and convert the measured current or voltage signal into a machine learning signal.

In step 902, the semiconductor device test method converts the machine learning signal into 2D array data.

That is, in the semiconductor device test method, the amplified-converted machine learning signal may be converted into vector data having a vector form over time, and the converted vector data may be divided into a plurality of frames and converted into 2D array data. .

In step 903, the method of testing a semiconductor device determines a condition of the device to be tested based on machine learning on the 2D array data.

That is, in the semiconductor device test method, sequential machine learning is performed on 2D array data, but after learning the 2D array data by dividing it into a plurality of folders, cross-validation is performed on the plurality of folders. Determine the conditions.

10 and 11 are diagrams illustrating an operating procedure of a semiconductor device testing apparatus according to various embodiments of the present disclosure.

Referring to FIG. 10, the semiconductor device testing apparatus applies a voltage to the semiconductor device in step 1001, amplifies the current measured from the semiconductor device in step 1002, and then the wave of the current measured in step 1003. Is measured, and a two-dimensional array data is generated from vector data by applying an MFCC-like filter to the measured wave.

In step 1004, the semiconductor device test apparatus machine learns the 2D array data using a hidden markov model.

For example, the step of machine learning 2D array data using the hidden Markov model includes the machine learning process described in FIG. 5.

In step 1005, the semiconductor device testing apparatus additionally machine-learns the machine-learned 2D array data using a neural network.

In step 1006, conditions of a semiconductor device related to channel material, gate material, chemical doping, and E-beam irradiation may be determined as a result of machine learning of the semiconductor device testing apparatus.

For example, the semiconductor device test apparatus may improve the accuracy of determining the conditions of the semiconductor device by using the hidden Markov model and the neuron network together to determine the conditions of the semiconductor device.

For example, in the case of using a hidden Markov model and a neuron network in the semiconductor device test apparatus, the crystal accuracy related to the conditions of the semiconductor device can be improved to about 96% or more.

FIG. 11 is a diagram illustrating an operation procedure of a semiconductor device test apparatus in which machine learning such as CNN is applied to the operation procedure of the semiconductor device test apparatus described in FIG. 10.

Referring to FIG. 11, the semiconductor device testing apparatus extracts feature data of a semiconductor device in step 1101, and features data using a machine learning neural network such as a convolution neural network (CNN) in step 1102. After machine learning is performed, feature data is learned using a hidden markov model in step 1103, and feature data is added using a neural network in step 1104. Learn.

The semiconductor device testing apparatus may determine conditions of a semiconductor device related to channel material, gate material, chemical doping, and E-beam irradiation based on the data learned in step 1102 and the feature data learned in step 1104.

According to an embodiment of the present invention, even when a model such as a CNN or a Gaussian Mixture Model (GMM) is used, the semiconductor device testing apparatus may exhibit an accuracy of about 90% or more in relation to the determination accuracy related to the conditions of the semiconductor device.

For example, when a specific voltage is applied to a semiconductor device or a nano-semiconductor device, the semiconductor device test apparatus may machine learn the condition of the semiconductor device by using a characteristic in which current/voltage fluctuation occurs from the device.

The semiconductor device test apparatus divides the current/voltage fluctuation into a specific frame such as a time domain frame, and then converts the current/voltage fluctuation into a power spectrum density through a Fourier transform, respectively, A filter that is sensitive to a specific frequency can be used to amplify the converted power spectrum density.

In addition, the semiconductor device test apparatus may perform a featureization of a two-dimensional array by extracting a coefficient through a discrete cosine transform.

In addition, the semiconductor device test apparatus learns through a machine learning application having various algorithms to infer various conditions such as channels, oxide films, device size, temperature, process conditions, degree of deterioration, and metal type. And can be distinguished.

The feature extraction mechanism used by the semiconductor device testing apparatus according to an embodiment of the present invention may depend on the level, distribution, and density of traps present in the semiconductor device.

A semiconductor device testing apparatus according to an embodiment of the present invention extracts fluctuation noise generated from a semiconductor device as a feature.

In the semiconductor device testing apparatus according to an embodiment of the present invention, by machine learning fluctuation noise using machine learning and deep learning algorithms, channels related to semiconductor devices, oxide films, device sizes, temperatures, process conditions, and Various conditions such as the degree of deterioration and the type of metal can be inferred and distinguished.

12A and 12B illustrate classification accuracy according to types of machine learning applications that can be utilized in the semiconductor device testing apparatus according to an embodiment of the present invention.

Referring to the graph 1200 of FIG. 12A, classification accuracy using a hidden markov model (HMM) and a neural network (NN) is shown.

In particular, if the result of learning more than 20,000 pieces of data is checked, the result of learning more than 20,000 pieces of data has very excellent classification accuracy, and may be the same as the graph 1210 of FIG. 12B.

Referring to the graph 1210 of FIG. 12B, classification accuracy of a hidden markov model (HMM), a neural network (NN), and a convolution neural network (CNN) is shown.

HMM (hidden markov model) represents 90.52%, NN (neural network) represents 95.52%, CNN (convolution neural network) represents 94.77%, respectively.

13 shows score differences according to labels in a semiconductor device testing apparatus according to an embodiment of the present invention.

Referring to the graph 1300 of FIG. 13, it can be seen that three labels are shown, and a specific label has very different noise characteristics and is thus clearly distinguished from other labels.

14A illustrates 2D array data 1400 obtained by classifying ReS2 h-BN related to a first semiconductor device by a semiconductor device testing apparatus according to an embodiment of the present invention.

14B illustrates two-dimensional array data 1410 in which ReS2 SiO2 related to a second semiconductor device is classified by the semiconductor device testing apparatus according to an embodiment of the present invention.

14C illustrates 2D array data 1420 obtained by classifying graphene related to a third semiconductor device by a semiconductor device testing apparatus according to an embodiment of the present invention.

14A to 14C illustrate two-dimensional array data for a test object related to a semiconductor device, but are not limited thereto. -beam, h-BN MoS2 300 K, SiO2 MoS2 300 K, SiO2 mono MoS2 after E-beam, SiO2 multi MoS2 after E-beam, SiO2 multi MoS2, h-BN MoS2 25 K, SiO2 MoS2 25 K, etc. Array data can be created.

The apparatus described above may be implemented as a hardware component, a software component, and/or a combination of a hardware component and a software component. For example, the devices and components described in the embodiments include, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable array (FPA), It may be implemented using one or more general purpose or special purpose computers, such as a programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions. The processing device may execute an operating system (OS) and one or more software applications executed on the operating system. Further, the processing device may access, store, manipulate, process, and generate data in response to the execution of software. For the convenience of understanding, although it is sometimes described that one processing device is used, one of ordinary skill in the art, the processing device is a plurality of processing elements and/or a plurality of types of processing elements. It can be seen that it may include. For example, the processing device may include a plurality of processors or one processor and one controller. In addition, other processing configurations are possible, such as a parallel processor.

The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, and the like alone or in combination. The program instructions recorded on the medium may be specially designed and configured for the embodiment, or may be known and usable to those skilled in computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic media such as floptical disks. -A hardware device specially configured to store and execute program instructions such as magneto-optical media, and ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine language codes such as those produced by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like. The hardware device described above may be configured to operate as one or more software modules to perform the operation of the embodiment, and vice versa.

The software may include a computer program, code, instructions, or a combination of one or more of these, configuring the processing unit to operate as desired or processed independently or collectively. You can command the device. Software and/or data may be interpreted by a processing device or, to provide instructions or data to a processing device, of any type of machine, component, physical device, virtual equipment, computer storage medium or device. , Or may be permanently or temporarily embodyed in a transmitted signal wave. The software may be distributed over networked computer systems and stored or executed in a distributed manner. Software and data may be stored on one or more computer-readable recording media.

As described above, although the embodiments have been described by the limited drawings, various modifications and variations are possible from the above description to those of ordinary skill in the art. For example, the described techniques are performed in a different order from the described method, and/or components such as systems, structures, devices, circuits, etc. described are combined or combined in a form different from the described method, or other components Alternatively, even if substituted or substituted by an equivalent, an appropriate result can be achieved.

Therefore, other implementations, other embodiments, and those equivalent to the claims also fall within the scope of the claims to be described later.

Claims

A source applying unit for applying a current or voltage to the device under test;

An amplification converter measuring a current or voltage signal output from the device under test based on the applied current or voltage, and amplifying and converting the measured current or voltage signal into a machine learning signal;

A data conversion unit converting the amplified-converted machine learning signal into vector data having a vector form with respect to time, and converting the converted vector data into two-dimensional array data by dividing the converted vector data into a plurality of frames; And

A machine learning unit that determines a condition of the device to be tested by performing machine learning on the two-dimensional array data,

The measured current or voltage signal has a frequency form according to the condition of the device under test.

Semiconductor device testing device.
The method of claim 1,

The data conversion unit calculates a power spectral density by applying a Fourier transform to the divided frames.

Semiconductor device testing device.
The method of claim 2,

The data conversion unit converts the vector data into the 2D array data by applying a discrete cosine transform using at least one filter to the calculated power spectral density.

Semiconductor device testing device.
The method of claim 1,

The data conversion unit analyzes the frequency shape using at least one filter, and converts the converted vector data into the two-dimensional array data each including conditions of the device under test based on the analyzed frequency shape. doing

Semiconductor device testing device.
The method of claim 4,

The number of the at least one filter is proportional to the determination accuracy of the condition of the device under test

Semiconductor device testing device.
The method of claim 1,

The machine learning signal has a text file format divided into a first column and a second column, the first column represents data according to time, and the second column is fluctuation data according to the change of time. Indicating

Semiconductor device testing device.
The method of claim 1,

The machine learning unit performs the machine learning to capture temperature, dimensions, process conditions, channel materials, gate materials, degradation degrees, electrical properties, reliability, and interfacial capture. (interface trap), bulk trap (bulk trap), luminescence degree, luminescence wavelength, or determining the conditions of the device to be tested including at least one of a lifetime

Semiconductor device testing device.
The method of claim 1,

The machine learning unit divides and learns the 2D array data into a plurality of folders, and then cross-validates the plurality of folders.

Semiconductor device testing device.
Measuring a current or voltage signal output from the device under test based on the current or voltage applied to the device under test, and amplifying and converting the measured current or voltage signal into a machine learning signal;

Converting the amplified-converted machine learning signal into vector data having a vector form with respect to time, and converting the converted vector data into 2D array data by dividing the converted vector data into a plurality of frames; And

In the machine learning unit, comprising the step of determining a condition of the device to be tested by performing machine learning on the two-dimensional array data,

The measured current or voltage signal has a frequency form according to the condition of the device under test.

Semiconductor device test method.
The method of claim 9,

Dividing the converted vector data into a plurality of frames and converting the converted vector data into 2D array data,

Calculating a power spectral density by applying a Fourier transform to the divided frames; And

Converting the transformed vector data into the 2D array data by applying a discrete cosine transform using at least one filter to the calculated power spectral density.

Semiconductor device test method.