KR20190001433A - Integrate-and-fire Neuron Circuit and Operating Method Thereof - Google Patents

Abstract

The present invention includes an input unit, an ignition unit connected to the input unit and adapted to ignite a spike current according to an input voltage inputted within a unit time, and an output unit connected to the ignition unit and transmitting an output signal based on the spike current, And a threshold switching element connected in parallel to the capacitor and becoming a low resistance state when a voltage equal to or higher than a threshold value is applied, and a method of operating the same. The neuron circuit of the present invention has a simple structure with reduced processing time and cost, and enables high-density integration. Also, it can be applied to a portable artificial intelligent terminal because it can be operated with low power.

Description

FIELD OF THE INVENTION [0001] The present invention relates to a neuron circuit,

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a spark-ignition type neuron circuit and an operation method thereof, and more particularly to a spark-ignition type neuron circuit including a threshold switching element and an operation method thereof.

The von-Neumann structure, which is currently applied to computing technology, is composed of a CPU for operation, a memory for storing information, and an input / output device. The computer operates through the CPU and stores it in the memory. The data required for the operation is read from the memory back to the CPU. Therefore, in the process of retrieving and storing the data from the memory every time the operation is performed, a great amount of time and power are consumed. Such a von Neumann structure is suitable for an operation requiring accuracy such as numerical calculation, but is ineffective in an operation for simultaneously processing a large amount of data such as image recognition or deep learning (Deep Learning). For example, in 2012, Google needed 16,000 processors of von Neumann architecture to run software that automatically recognizes cat images.

On the other hand, the human brain is composed of more than 100 billion neurons, and the synapse between each neuron and neurons is over 100 trillion. Neurons constituting the human brain and neurons are connected in parallel, so that a large amount of data can be processed simultaneously by only 10 W of power consumption. Synapses allow a single neuron to be connected in parallel to send and receive signals to multiple neurons, as well as to increase conductivity in the presence of frequent stimuli and to decrease the conductivity in the case of rare stimuli, . Neuronal neurons have a signal processing function in which a pulse signal is output when the sum of input electrical signals exceeds a cell's action potential value. The signal processing function of neurons seems to play a pivotal role in complex brain operations such as pattern recognition and feature extraction. In recent years, research on next generation semiconductors that simulate the human brain has been accelerated.

Neuromorphic devices have a structure in which a large number of artificial neurons, such as the human nervous system, are connected in parallel by artificial synapses. In order to realize the synaptic function of the human brain, when a CMOS transistor structure is used, there is a limitation in integration due to occupying a large area of the CMOS transistor. Therefore, a novel memory device using a new memory element such as a magnetic random access memory, a phase-change random access memory, and a resistive switching random access memory has been actively studied. Due to the simple device structure of the metal / oxide / metal, the resistance change memory is expected to enable high density integration when applied to a neomorphic device.

SUMMARY OF THE INVENTION The present invention provides a spark-ignition type neuron circuit and a method of operating the same.

According to an aspect of the present invention, there is provided an input device including an input unit to which an input voltage is applied, a spark portion connected to the input unit and configured to spike a spike current according to an input voltage input within a unit time, And a threshold switching element which is connected in parallel to the capacitor and is in a low resistance state when a voltage equal to or higher than a threshold value is applied to the capacitor, and the output section converts the spike current into a spike voltage and outputs the spike current. to provide.

The threshold switching device includes a substrate, a lower electrode formed on the substrate, a transition metal oxide layer formed on the lower electrode, and an upper electrode formed on the transition metal oxide layer.

And the transition metal oxide layer has a threshold switching characteristic in a low current region of 0.01 占 내지 to 10 占..

The transition metal oxide layer may comprise any one of a hafnium oxide (HfO _2), titanium oxide (TiO _2), tungsten oxide (WO _3), nickel oxide (NiO), and aluminum oxide (Al ₂ O ₃₎ .

The neuron circuit can accelerate by decreasing the capacitance of the capacitor by reducing the time for the spark portion to spike the spark current by the same input signal or by lowering the threshold voltage of the threshold switching element.

The input of the neuron circuit may be connected to the outputs of one or more pre-neuron circuits and the output may be connected to the inputs of one or more post-neuron circuits.

According to another aspect of the present invention, there is provided a semiconductor memory device including an input unit, a speech unit, and an output unit, the memory unit including a capacitor and a capacitor connected in parallel to the capacitor, A neuron circuit comprising a threshold switching device, the neuron circuit comprising: a charging step of charging the capacitor with an input voltage applied to the input part; a charge step of charging the capacitor with no input voltage applied to the input part, Wherein the charging step is repeated before the capacitor is completely discharged and a spike current is ignited from the ignition part when a voltage equal to or higher than a threshold value is applied to the capacitor and the threshold switching element, To convert the spike current into a spike voltage And outputting the output signal of the neuron circuit.

The charging step, the discharging step, and the firing step may be performed in a low current region of 0.01 ㎂ to 10..

The operating method of the neuron circuit may accelerate the ignition step by reducing the capacitance of the capacitor or by reducing the threshold voltage of the threshold switching element.

The input voltage applied to the input may be a spike voltage that is emitted from the pre-neuron circuits, and the converted spike voltage at the output may be applied to the input of the post-neuron circuits to operate the novel Lomic system.

The neuron circuit of the present invention implements a signal processing function of a neuron constituting a human brain neural network using a nano-sized transition metal oxide-based threshold switching device and a simple structure of a capacitor. Because of the simple structure of the circuit and the high integration of the components, it can be used as an artificial system for a highly dense integrated neurocompic system. In addition, the threshold switching device of the present invention has the same structure as the resistance change memory, and it is possible to manufacture artificial neurons for signal processing and artificial synapses for storage and learning through the same process in the device integration process for a novel Lomonote system, It is thought that the unit price and the process time are greatly reduced.

It is believed that the method of operation of the neuron circuit of the present invention can significantly reduce power consumption as it operates at low currents. Also, since a plurality of pre-neuron circuits and a plurality of post-neuron circuits are connected in parallel, the present invention can be applied to various application devices for driving the novel Lomopic system with low power.

The technical effects of the present invention are not limited to those mentioned above, and other technical effects not mentioned can be clearly understood by those skilled in the art from the following description.

1 is a circuit diagram showing a unit neuron circuit according to an embodiment of the present invention.
2 shows a structure of a threshold switching device according to an embodiment of the present invention.
FIG. 3 is a graph showing energy dispersive spectroscopy results of (a) a threshold switching device according to an embodiment of the present invention, (b) a graph showing IV characteristics in a high current region, and ) Transmission electron microscope (TEM) photograph.
FIG. 4 is a graph showing a multi-level conductivity state according to application of a gate voltage and a source-drain voltage in a high-current region in a threshold switching device according to an embodiment of the present invention.
5 is a graph showing IV characteristics of a threshold switching device in a low current region and (b) IV characteristics in a low current region due to a voltage sweep and a current sweep Graph.
6 is a graph showing simulated values according to the Poole-Frenkel equation and IV characteristics in the low current region and the high current region.
7 is a diagram for explaining a principle of a threshold switching device according to an embodiment of the present invention showing a threshold switching characteristic in a low current region and a nonvolatile memory characteristic in a high current region .
FIG. 8 is a graph showing a spike current of a neuron circuit according to an embodiment of the present invention (a) a change in voltage (b) according to an applied voltage input to an input unit and (c) a spike current.
FIG. 9 is an explanatory diagram illustrating a novel Lomographic system according to an embodiment of the present invention.

While the invention is susceptible to various modifications and alternative forms, specific embodiments thereof are shown by way of example in the drawings and will herein be described in detail. Rather, the intention is not to limit the invention to the particular forms disclosed, but rather, the invention includes all modifications, equivalents and substitutions that are consistent with the spirit of the invention as defined by the claims.

It will be appreciated that when an element such as a layer, region or substrate is referred to as being present on another element "on," it may be directly on the other element or there may be an intermediate element in between .

Although the terms first, second, etc. may be used to describe various elements, components, regions, layers and / or regions, such elements, components, regions, layers and / And should not be limited by these terms.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

Example 1: Neuron circuit

1 is a circuit diagram showing a unit neuron circuit according to an embodiment of the present invention.

Referring to FIG. 1, the neuron circuit includes an input unit 11, an ignition unit 13 electrically connected to the input unit 11, and an output unit 17 electrically connected to the ignition unit 13. The input unit 11 receives an electrical signal input from a signal generating source or a pre-neuron circuit for transmitting a signal. Alternatively, the input unit 11 can generate an input voltage that can be input to the ignition unit based on the input signal. The ignition unit 13 is electrically connected to the input unit 11 to receive the input voltage generated by the input unit 11 and to generate a spike current according to the accumulation and ignition model of the neuron. In other words, the ignition unit 13 may not transmit or transmit an electrical signal to other neural circuits or synapse circuits connected to the output unit based on the input voltage value received within a predetermined time. The output unit 17 may be electrically connected to the ignition unit 13 and may transmit the spiked current to other neuron circuits or synapse circuits to which the ignition unit 13 is connected. The output 17 may convert the spiked current into a form suitable for the circuit being received. For example, if the output 17 is connected to a post-neuron circuit, the spike current can be converted to the form of a spike voltage.

The ignition part 13 includes a capacitor 14 and a threshold switching element 15 connected in parallel to the capacitor 14. When the input voltage is applied to the ignition part 13, the capacitor 14 is charged with electric charge, and the voltage applied to both ends of the capacitor 14 and the threshold switching element 15 rises. On the other hand, when the input voltage is removed, the charge stored by the resistance of the threshold switching element 15 is consumed and the voltage applied to both ends of the capacitor 14 and the threshold switching element 15 decreases. When the input voltage is continuously applied to the ignition part 13, the charge stored in the capacitor 14 is completely discharged and charged again before returning to the initial state. The voltage applied to both ends of the capacitor 14 and the threshold switching element 15 can be equal to or higher than the threshold voltage value of the threshold switching element 15 by the application of the continuous input voltage by the transient response phenomenon. When a voltage equal to or higher than the threshold voltage value is applied to the threshold switching element 15, the threshold switching element 15 generates a low resistance state (LRS) and a spike type current. When the current in the form of a spike consumes all of the charge stored in the capacitor 14, the threshold switching element 15 is again in a high resistance state (HRS).

The output section 17 is electrically connected to the ignition section 13 and receives the spike current from the ignition section 13 and transmits the spike current to the post-neuron circuit or the signal output device. Alternatively, the output section 17 may convert the received spike current into a form suitable for the post-neuron circuit or signal output device. For example, the output 17 may include an operation amplifier 18 for converting the signal to be transmitted to the post-neuron circuit into a form of a spike voltage.

2 shows a structure of a threshold switching device according to an embodiment of the present invention.

2, the threshold switching element includes a lower electrode 21 formed on a substrate 10, a transition metal oxide layer 23 formed on a lower electrode 21, a transition metal oxide layer 23 formed on the transition metal oxide layer 23, And an upper electrode 25 formed thereon.

The substrate 20 may be a silicon substrate or a silicon on insulator (SOI) substrate. An interlayer insulating layer (not shown) may be formed between the substrate 20 and the lower electrode 21.

The lower electrode 21 may include tungsten, platinum, rubidium, iridium, aluminum, molybdenum, or titanium nitride. .

The upper electrode 25 may include titanium nitride (TiN), tungsten (W), platinum (Pt), and tantalum nitride (TaN). And a metal layer (not shown) having high chemical reactivity between the upper electrode 25 and the transition metal oxide layer 23. For example, the metal layer may be any one of titanium (Ti), tantalum (Ta), and hafnium (Hf), but is not limited thereto. The metal layer absorbs oxygen present at the interface of the transition metal oxide layer 23 to modify the interface of the transition metal oxide layer 23 non-stoichiometrically.

The transition metal oxide layer 23 may be a material having a threshold switching characteristic in a low current region of 0.01 μA to 10 μA. For example, the transition metal oxide layer 23 of hafnium oxide (HfO _2), titanium oxide (TiO _2), tungsten oxide (WO _3), nickel oxide (NiO), and aluminum oxide (Al ₂ O ₃₎ thin film Lt; / RTI >

The threshold switching device has a simple structure of a metal / oxide film / metal and has a merit that a process is simple and high density integration is possible. The threshold switching element is a resistance change memory in a high current region, and a conductive filament is formed in the transition metal oxide layer 23 to operate as a resistance change memory. It can therefore be used as a component of a synapse circuit for storing weights. On the other hand, in the low current region of 10 ㎂ or less, threshold switching characteristics can be used as a constituent element of a spark - As the synapse circuitry and neuron circuitry use the same device, the benefits of process time and cost can be expected when building a neural system.

Experimental Example 1: Characteristic Measurement of Threshold Switching Element

In order to measure the characteristics of the threshold switching device, a 1-transistor and 1-threshold switching device configuration (1T-1R) was used.

First, a threshold switching device connected to a transistor was fabricated. A 400 nm thick titanium nitride lower electrode was formed on the silicon substrate. The lower electrode is electrically connected to the drain electrode of the 0.35 mu m transistor. A hafnium oxide transition metal oxide layer having a thickness of 6 nm was formed on the lower electrode using atomic beam deposition. An upper electrode including a 15 nm thick titanium layer and a titanium nitride layer was formed on the hafnium oxide transition metal oxide layer. The transistor prevents a problem such as overshoot occurring in the low current region when measuring the characteristics of the threshold switching element and electrically connects the lower electrode and the source of the threshold switching element to precisely control the operating current .

FIG. 3 is a graph showing energy dispersive spectroscopy results of (a) a threshold switching device according to an embodiment of the present invention, (b) a graph showing IV characteristics in a high current region, and ) Transmission electron microscope (TEM) photograph.

Referring to FIG. 3 (a), the energy dispersive spectroscopic analysis shows that the peak indicating the presence of the hafnium element is highest in the middle of the deposited tie element. In other words, it can be seen that the threshold switching element has a titanium nitride bottom electrode / hafnium oxide transition metal oxide layer / titanium-titanium nitride top electrode structure. Similarly, the structure of the threshold switching device can be confirmed through a transmission electron microscope photograph of FIG. 3 (c).

Referring to FIG. 3 (b), it can be seen that the threshold switching device exhibits the characteristics of the resistance change memory in the high current region where a current of 100 흐 flows. When a voltage having a positive value is applied to the upper electrode of the threshold switching element and subjected to a forming process, the threshold switching element becomes a low resistance state (LRS). It can be confirmed that when voltage sweeping with a negative voltage is applied to the threshold switching element in the low resistance state, the state is changed to the high resistance state (HRS). After that, if voltage-sweep is performed to have a positive value, a sudden increase in current can be confirmed at a set voltage (V _SET ) of about 1V. This is considered to be because the high resistance state was changed to the low resistance state again. This bipolar resistance change is due to the migration of oxygen vacancies in the transition metal oxide layer. In the forming step and the forming step, oxygen vacancies are produced in the transition metal oxide layer and are moved by the applied field to form conductive filaments. The conductive filament electrically connects the upper electrode and the lower electrode, resulting in a sudden increase in current. Conversely, when a negative voltage is applied in the reset step, the oxygen vacancies are pushed out of the conductive filament, and a gap is created between the conductive filament and the electrode. Therefore, the device returns to the high resistance state again.

FIG. 4 is a graph showing a multi-level conductivity state according to application of a gate voltage and a source-drain voltage in a high-current region in a threshold switching device according to an embodiment of the present invention.

When a high current of 100 가 flows to the threshold switching element and the transistor, the threshold switching element has a multi-level resistance change characteristic by the above-described principle. Increasing the magnitude of the gate voltage results in the formation of larger conductive filaments, thus increasing the conductivity. On the other hand, when the negative reset voltage (-1.1 V) is applied as a pulse, the gap between the conductive filament and the electrode gradually increases, and the conductivity decreases. In other words, depending on the electrical signal input, the conductivity of the threshold switching element may have a multi-level value, which is similar to a synapse whose conductivity increases or decreases with frequency of use. Therefore, the threshold switching device can be used as a synapse device that stores the weight value according to the frequency of use.

5 is a graph showing IV characteristics of a threshold switching device in a low current region and (b) IV characteristics in a low current region due to a voltage sweep and a current sweep Graph.

Referring to FIG. 5 (a), it can be seen that threshold switching characteristics are observed instead of the memory change, which was confirmed in FIG. 4, when the threshold switching element operates in a low current region of 10 낮 by lowering the gate bias of the transistor. That is, although the threshold switching element is in the low resistance state at V _TH , when the voltage applied to the threshold switching element decreases again to become the hold voltage (V _HOLD ), the low resistance state returns to the high resistance state. That is, the memory characteristics of the threshold switching device disappear. The IV curve showing this threshold switching characteristic was continuously observed in the low current region of 10 ㎂ or less.

Referring to FIG. 5 (b), a snap back indicating a sudden state change with current-sweep was clearly observed. These threshold switching characteristics were also observed in transition metal oxides such as VO ₂ , NbO ₂ and TiO ₂ .

FIG. 6 is a graph showing other simulated values in the Poole-Frenkel equation and I-V characteristics in the low current region and the high current region.

A Poole-Frenkel-based analytical model was used to understand the high-resistance conduction mechanism in the low current region.

I = 2qAN _T x? Z / t ₀ x exp (- (E _C -E _F ) / kT) sinh (qV _A ? Z / 2dkT)

Wherein q is a charge amount, A is the area, N _T is the trap density, △ z is the average distance between the traps, t ₀ is the trapped electrons to escape time, k is Boltzmann's constant, T is temperature, d is a hafnium oxide-transition metal The thickness of the oxide layer and (E _C -E _F ) is the trap energy. The trap energy was assumed to be 0.3 eV in consideration of oxygen vacancies having a positive value in the hafnium oxide layer.

Referring to FIG. 6, it can be confirmed that the graph showing the calculated value when the trap density is as low as about 3 × 10 ¹⁷ cm ^-3 is very similar to the IV curve in the high-resistance state of the threshold switching element. Also, the graph showing the calculated values when the trap density is as high as about 1 × 10 ²¹ cm ^-3 confirmed that the threshold switching device is in good agreement with the IV curve when the resistance change memory characteristic is shown. The coincidence between the calculated value and the experimental value means that the resistance change memory characteristic and the threshold switching characteristic are related to the number of oxygen vacancies according to the allowable current value of the threshold switching device.

7 is a diagram for explaining a principle of a threshold switching device according to an embodiment of the present invention showing a threshold switching characteristic in a low current region and a nonvolatile memory characteristic in a high current region .

When the threshold switching element undergoes a foaming step in a low current region, only a small amount of oxygen vacancies are formed, and the formed oxygen vacancies are dispersed with a wide interval. Thus, a high voltage is required to form the conductive filament. Also, as shown in Fig. 7 (a), a weak conductive filament is formed by oxygen vacancies distributed at a low density. It is considered that the weak conductive filament spontaneously disappears as the voltage applied to the threshold switching device is removed, causing a threshold switching phenomenon.

On the other hand, referring to Fig. 7 (b), in the high current region, the oxygen vacancies formed in the foaming step have a high density, and consequently the conductive filament is larger and more durable. Conductive filaments of high density oxygen vacancies can maintain a stable low resistance state without voltage application. This explanation explains why a set voltage (V _SET = 1V) lower than the threshold voltage (V _TH = 1.5V) when the threshold switching element exhibits a resistance change memory characteristic. In the high current region, the oxygen vacancies have a high density, so that the interval between the oxygen vacancies is narrow and the conductive filament is more easily formed.

EXPERIMENTAL EXAMPLE 2: Experimental experiment on neuron circuit

A unit neuron circuit was fabricated by connecting a threshold switching element having a top electrode structure of titanium nitride lower electrode / hafnium oxide transition metal oxide layer / titanium-titanium nitride described in Experimental Example 1 and a capacitor in parallel. In the input part of the neuron circuit, a voltage generating source that periodically generates a pulse voltage of a predetermined magnitude is connected, and the output part of the neuron circuit is grounded.

FIG. 8 is a graph showing a spike current of a neuron circuit according to an embodiment of the present invention (a) a change in voltage (b) according to an applied voltage input to an input unit and (c) a spike current.

8 (a) is a graph showing an input voltage transmitted to an input part of a neuron circuit. Since the electrical signal applied to the input section is in the form of a spike voltage, the input section of the neuron circuit can be constituted solely of resistors without a circuit configuration for signal conversion. A voltage of 3 V was applied at time intervals of 10 mu s.

Referring to FIG. 8 (b), a change in the voltage applied to the threshold switching element can be confirmed by charging and discharging the capacitor. When the pulse voltage is transmitted to the input of the neuron circuit, the capacitor is charged and then discharged quickly. When the pulse voltage is applied again before the capacitor is fully discharged, the charge is recharged to the capacitor and more charge is charged than when the first pulse voltage is applied. The magnitude of the voltage applied to the threshold switching element is changed by the charge stored in the capacitor. The pulse voltage is repeatedly applied to repeatedly charge and discharge the capacitor, and the peak value of the voltage applied to the threshold switching element gradually increases. The conductive filament is formed in the transition metal oxide layer of the threshold switching element and the threshold switching element is in a low resistance state when the applied voltage becomes equal to or higher than the threshold voltage. However, the conductive filament formed in the low current region is unstable and disappears when there is no applied voltage, so that the threshold switching element returns to the high resistance state. Therefore, when the threshold switching element is in the low resistance state, the current flows instantaneously through the neuron circuit to generate a spike type current.

Referring to FIG. 8 (c), when the threshold switching element is in the low resistance state as described above, currents flowing across both ends of the neuron circuit instantaneously increase and the spike current flows. The generation of such a spike current can be accelerated by lowering the threshold voltage value of the threshold switching element or decreasing the capacitance of the capacitor. By adjusting the threshold voltage value or the capacitance of the capacitor to adjust the spark sensitivity of the spike current of each neuron circuit, the weight value can be utilized as a pre-stored synaptic neuron circuit.

Example 2: A neuromorphic system using a neuron circuit

FIG. 9 is an explanatory diagram illustrating a novel Lomographic system according to an embodiment of the present invention.

The input of the neuron circuit 33 may be electrically connected to a signal source such as a plurality of pre-neuron circuits 31a, b ..., a synapse circuit or a sensor. The input may be a conductive wire for simply receiving an input signal originating from a signal source, such as a plurality of pre-neuron circuits, synaptic circuits, or sensors, in the form of a spike voltage. Alternatively, the input signal may be a circuit for forming a spike voltage based on the input signal if the input signal is not suitable for the ignition part.

A spike voltage received by the input section or an input voltage generated based on the input section input signal is applied to the ignition section. The ignition portion can ignite the spike current based on the input voltage value received within a certain time.

The output may be in electrical communication with the spark portion to deliver spiked currents to other neuron circuits 33a, b ...., synaptic circuits or output devices connected. The output may be a conductive wire for delivering the spike current to the plurality of circuits. Or the output may be a circuit for converting the spiked current into a form suitable for the circuit being received.

A plurality of circuits are connected to the input unit and the output unit, so that a new Lomo pick system can be formed. The neuromotor system consists of sensors, neuron circuits, synapse circuits and output devices connected in parallel. One neuron circuit determines whether or not to ignite by combining signals inputted from a plurality of circuits or sensors, and the ignited electrical signals are output to a plurality of circuits or output devices to enable parallel processing of data. Since the neuron device according to the present invention can be driven with low power and high density integration, the neuromotion system can be applied to a next generation artificial intelligence computer and a portable terminal.

11: input unit 13:
15: Threshold switching element 14: Capacitor
17: output section 18: operational amplifier
20: substrate 21: lower electrode
23: transition metal oxide layer 25: upper electrode
31 a, b: pre-neuron circuit 33: neuron circuit
35 a, b: post-neuron circuit

Claims (14)

An input unit for transmitting an input signal to generate an input voltage;
A spark portion connected to the input portion and generating a spike current according to an input voltage input within a unit time; And
And an output unit connected to the ignition unit and transmitting an output signal based on the spike current,
Wherein the ignition portion comprises a capacitor and a threshold switching element connected in parallel to the capacitor and being in a low resistance state when a voltage equal to or greater than a threshold value is applied. The method according to claim 1,
The threshold switching device
Board;
A lower electrode formed on the substrate;
A transition metal oxide layer formed on the lower electrode; And
And a top electrode formed on the transition metal oxide layer. 3. The method of claim 2,
Wherein the transition metal oxide layer has a threshold switching characteristic in a low current region of 0.01 占 내지 to 10 占.. 3. The method of claim 2,
Neuron circuit of the transition metal oxide layer comprises one of hafnium oxide (HfO _2), titanium oxide (TiO _2), tungsten oxide (WO _3), nickel oxide (NiO), and aluminum oxide (Al ₂ O ₃₎ . The method according to claim 1,
And the time for which the spark portion is sparked by the same input signal is accelerated by lowering the capacitance of the capacitor. The method according to claim 1,
And the time for which the spark portion generates the spike current by the same input signal is accelerated by lowering the threshold voltage of the threshold switching element. The method according to claim 1,
Wherein the input is connected to one or more signal sources or pre-neuron circuits that carry signals. The method according to claim 1,
Wherein the output is coupled to one or more post-neuron circuits or signal output devices. And a threshold switching element connected in parallel to the capacitor and being in a low resistance state when a voltage equal to or higher than a threshold value is applied, the circuit comprising:
A charging step of charging the capacitor with an input voltage applied to the input unit;
A discharging step of discharging the charge charged in the capacitor with time when there is no input voltage applied to the input unit;
Wherein the charging step is repeated before the capacitor is completely discharged and a spike current is generated from the ignition unit when a voltage equal to or higher than a threshold value is applied to the capacitor and the threshold switching element; And
And an output step of converting a spike current generated in the ignition step into a spike voltage. 10. The method of claim 9,
Lt; RTI ID = 0.0 > 10 A, < / RTI > 10. The method of claim 9,
Thereby reducing the capacitances of the capacitors and accelerating the ignition step. 10. The method of claim 9,
Wherein the threshold voltage of the threshold switching device is reduced to accelerate the ignition step. 10. The method of claim 9,
Wherein the input voltage applied to the input is a spike voltage that is generated from a plurality of signal generators or pre-neuron circuits. 10. The method of claim 9,
Wherein the spike voltage converted at the output is applied to a plurality of post-neuron circuits or an input of the signal output devices.

Images