Analog conductivity state high-speed write-in method of resistive memory and application of analog conductivity state high-speed write-in method

Abstract

The invention provides an analog conductivity state high-speed writing method of a resistive memory, and belongs to the field of semiconductors, analog computation and integrated circuits. According to the method, based on a feedback loop in a continuous time domain formed by an electrical signal connection input module, a reference resistor module, a resistive memory module and a feedback module, the resistive memory can be autonomously programmed to reach any simulation conductivity state in continuous time; and the process of intelligently judging whether a target conductivity state is reached or not and terminating writing is realized through a feedback signal, multi-round reading verification is not needed, the programming speed of the device is improved, analog-to-digital conversion and an additional logic judgment circuit are not needed, the programming speed and efficiency of the memory array in the matrix operation accelerator can be greatly improved, and the programming cost is reduced. The potential of analog computation and in-memory computation is fully exerted, and the method has important significance and wide development prospects for accelerating neural network reasoning, training and other scenes needing a large number of device programming.

Description

Technical Field

[0001] The present invention belongs to the field of semiconductor, analog computing and integrated circuit, and relates to a high-speed writing method and application of analog conductivity state programming of a resistive memory. Background Art

[0002] With the development of artificial intelligence technology and the surge in data volume, traditional digital computers are limited by the architecture of memory-computation separation and serial digital algorithms, and are unable to cope with today's large and complex computing tasks. Analog computing circuits based on resistive memory arrays can perform in-situ matrix operations, such as matrix-vector multiplication, matrix inversion, matrix pseudo-inversion, and matrix eigenvectors. Benefiting from the elimination of data transfer in the in-memory computing architecture and the extremely high parallelism of analog computing, this method provides a strong guarantee for breaking through the von Neumann bottleneck and has been widely studied and applied as a neural network training and reasoning accelerator.

[0003] As the acceleration core of matrix operations, the memory array needs to be accurately and efficiently programmed to the target conductivity state to map the corresponding matrix elements. Due to the randomness of resistive memory, the currently widely used method is the write-verify method, that is, for each device, each write pulse must be read and verified once and the height and width of the next round of pulses must be determined. This usually takes dozens of rounds to program a device to the target state, which consumes a lot of time. In addition, the additional analog-to-digital conversion and digital logic operation modules will also generate a lot of power consumption and chip area overhead, ultimately reducing or even offsetting the advantages of analog in-memory computing. In addition, the existing high-speed self-terminating analog conductivity state writing scheme can only program a very limited number of conductivity states, usually 1-2 bits, which cannot meet the needs of analog in-memory computing. Therefore, a high-speed, analog-to-digital conversion-free, autonomously judged and terminated analog conductivity state writing scheme is urgently needed to be developed. Summary of the invention

[0004] In view of the problems existing in the above-mentioned prior art, the present invention proposes a high-speed writing method and application of the analog conductivity state of a resistive memory, which is based on a feedback loop in a continuous time domain composed of an electrical signal connecting an input module, a reference resistor module, a resistive memory module and a feedback module. The resistive memory can be autonomously programmed to reach any analog conductivity state in a continuous time, and the feedback signal can be used to intelligently determine whether the target conductivity state has been reached and terminate the writing process, without the need for an additional analog-to-digital converter and a digital logic operation module, thereby accelerating the programming process of the device.

[0005] The technical solution of the present invention is as follows:

[0006] A method for high-speed writing of analog conductivity state of resistive memory, the method is based on an analog circuit composed of an input module, a reference resistor module, a resistive memory module and a feedback module, each module is connected based on an electrical signal to form a feedback loop in a continuous time domain, the resistive memory is autonomously programmed to reach any analog conductivity state in a continuous time, and the feedback signal is used to realize intelligent judgment on whether the target conductivity state is reached and the writing process is terminated; the method comprises the following steps:

[0007] 1) Reset the conductivity state of the resistive memory

[0008] 2) Adjust the conductivity value of the reference resistance module and select the target conductivity state

[0009] 3) The input power supply of the input module provides a DC voltage to adjust the initial voltage value applied to the resistive memory.

[0010] 4) Connect the resistive memory to form a complete feedback loop in the continuous time domain

[0011] 5) Program the resistive memory conductance state to near the target conductance state and realize the function of autonomous judgment and termination

[0012] In the feedback loop in the continuous time domain, the feedback module detects the conductivity difference between the memory and the reference resistor, and feeds back the conductivity difference to the input module through the output terminal in the form of a voltage signal, completing the feedback process of the electrical signal. In this process, the resistive memory is programmed, and its conductivity value continues to increase, approaching the target conductivity value, and the gap between the two becomes smaller and smaller. The input module will autonomously and continuously adjust the voltage value applied to the resistive memory according to the voltage signal obtained by feedback until the conductivity value of the resistive memory reaches near the target conductivity value. At this time, the voltage applied to the resistive memory will drop to 0, and the writing process is terminated. The entire judgment process and the termination of the writing process are autonomously realized by the feedback loop through the feedback of the electrical signal, so the resistive memory can be written to any analog conductivity state.

[0013] Furthermore, the resistive memory module of step 1) is composed of a single resistive memory, which is connected to a common node connected to the feedback module and the reference resistance module through a switch S1; a reset Reset voltage pulse is applied to the resistive memory, that is, the top electrode TE end of the resistive memory is grounded and a positive voltage pulse is applied to the bottom electrode BE end. After the Reset operation is completed, the resistive memory is adjusted to a low conductivity state.

[0014] Furthermore, the reference resistor module of step 2) includes a series of analog switches and fixed resistors, and the conductance value of the reference resistor module is adjusted by closing different switches or different numbers of switches, and the conductance value is used as the target conductance state to be written into the resistive memory.

[0015] Furthermore, the input module of step 3) includes an input power supply, a fixed resistor and an analog inverter constructed by an operational amplifier OA, the input power supply is directly connected to the inverting input terminal resistance of the analog inverter through a fixed resistor, and the output terminal of the analog inverter is connected to the resistive memory; the input power supply provides a DC voltage of a certain amplitude to the circuit, which is used to adjust the output terminal voltage of the analog inverter, that is, the initial voltage value applied to the resistive memory.

[0016] Furthermore, the feedback module in the feedback loop of step 4) is composed of another operational amplifier OA and a fixed resistor on the output feedback loop, the OA works as a positive feedback amplifier, the input end of the feedback module is directly connected to the output end of the reference resistance module, and is connected to the resistive memory module through the switch S1, the fixed resistor on the output feedback loop is directly connected to the inverting input resistance of the analog inverter in the input module, the feedback module is used to sense the difference between the conductance values ​​of the memory and the reference resistor, and is used for direct feedback of the voltage signal to form a feedback loop; closing the switch S1, the input module, the reference resistor module, the resistive memory module and the feedback module form a complete feedback loop in the continuous time domain.

[0017] Furthermore, the resistive memory is a resistive memory, a phase change memory, a magnetic memory or a ferroelectric memory.

[0018] The present invention also applies the above method to a resistive memory array, uses the resistive memory as a single device selected in the resistive memory array, and applies the feedback loop in the continuous time domain to an analog conductivity state write circuit of the resistive memory array; the analog conductivity state write circuit of the resistive memory array includes an N×N 1T1R array, a control circuit, the feedback loop in the continuous time domain, a driving circuit and a multiplexer for word lines WL / bit lines BL / source lines SL, the multiplexer near the bit lines BL and the source lines SL is equivalent to an external interface of the resistive memory module, the word line WL corresponding to the resistive memory determines whether the resistive memory module is connected to the feedback loop, and the control circuit is responsible for sending control signals to each module, controlling the connectivity and working mode between modules; applying the above method to the resistive memory array includes the following steps:

[0019] S1) Execute the reset operation on the selected resistive memory

[0020] First, a certain resistive memory in the 1T1R array is selected, that is, a control signal is input through the control circuit to open the WL corresponding to the resistive memory and the corresponding SL and BL are connected to the multiplexer, and a reset voltage pulse is applied to reset the selected resistive memory to a low conductance state;

[0021] S2) Adjust the reference resistance module and select the target conductivity state

[0022] The control circuit transmits a control signal to the reference resistor module, and adjusts the conductance value of the reference resistor module by closing different switches or different numbers of switches, and the conductance value is used as the target conductance state to be written into the resistive memory;

[0023] S3) controlled by the control circuit to adjust the input voltage of the input module in the feedback loop to provide an initial voltage drop of the memory;

[0024] S4) connecting the SL and BL corresponding to the memory to the feedback loop through a multiplexer, turning on the corresponding WL, connecting the selected resistive memory to the feedback loop, and forming a complete feedback loop in the continuous time domain;

[0025] S5) The selected resistive memory will be written near the target conductivity state in a continuous time, and realize the function of autonomous judgment and termination;

[0026] By repeating the above steps S1) to S5), all memories in the array can be written into the set simulated conductance state.

[0027] The beneficial effects of the present invention are as follows:

[0028] The present invention discloses a method and application for writing analog conductivity states of resistive memory at a high speed. Based on a feedback loop in a continuous time domain formed by connecting an input module, a reference resistor module, a resistive memory module, and a feedback module with an electrical signal, the resistive memory can be programmed to the vicinity of any analog conductivity value at an extremely high writing speed in a continuous time. Compared with the traditional write-verify method, the circuit does not require an iterative process, a read verification process, or multiple rounds of writing, thus saving the overhead of circuits such as analog-to-digital converters and logic judgment circuits, and greatly improving the speed and energy efficiency of device programming. Compared with other self-terminating one-step writing schemes, this method greatly increases the programmable conductivity states of the device, which helps to further bring into play the advantages of analog in-memory computing technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 It is a flow chart of the analog conductance state high-speed writing method of the resistive memory of the present invention;

[0030] Figure 2 It is a feedback loop block diagram used in the analog conductance state high-speed writing method of the resistive memory of the present invention;

[0031] Figure 3 It is a feedback loop circuit diagram and its principle adopted by the analog conductance state high-speed writing method of the resistive memory of the present invention;

[0032] Figure 4 It is a schematic diagram of applying the analog conductivity state high-speed writing method of the resistive memory of the present invention to a variable resistor array. DETAILED DESCRIPTION

[0033] In order to more clearly illustrate the purpose, technical solutions and advantages of the present invention, the following is a further detailed description in conjunction with the accompanying drawings. The description herein is only used to explain the present invention and is not intended to limit the present invention.

[0034] The present invention discloses a method for high-speed writing of analog conductivity state of a resistive memory. The method is based on an analog circuit composed of an input module, a reference resistor module, a resistive memory module and a feedback module. The modules are connected based on electrical signals to form a feedback loop in a continuous time domain. The resistive memory is autonomously programmed to reach any analog conductivity state in a continuous time, and the feedback signal is used to realize the process of intelligently judging whether the target conductivity state is reached and terminating the writing. The flow chart is as follows: Figure 1 As shown, the following steps are included:

[0035] 1) Reset the conductivity state of the resistive memory in the resistive memory module

[0036] The resistive memory module is composed of a single resistive memory, which is connected to the common node of the feedback module and the reference resistance module through a switch S1; a reset voltage pulse is applied to the resistive memory, that is, the top electrode TE end of the resistive memory is grounded and a positive voltage pulse is applied to the bottom electrode BE end. After the reset operation is completed, the resistive memory is adjusted to a low conductance state;

[0037] 2) Adjust the conductivity value of the reference resistance module and select the target conductivity state

[0038] The reference resistor module includes a series of analog switches and fixed resistors, and the conductance value of the reference resistor module is adjusted by closing different switches or different numbers of switches, and the conductance value is used as the target conductance state to be written into the resistive memory;

[0039] 3) The input power supply of the input module provides a DC voltage to adjust the initial voltage value applied to the resistive memory.

[0040] The input module includes an input power supply, a fixed resistor and an analog inverter constructed by an operational amplifier OA. The input power supply is directly connected to the inverting input terminal resistance of the analog inverter through a fixed resistor, and the output terminal of the analog inverter is connected to the resistive memory. The input power supply provides a DC voltage of a certain amplitude for the circuit to adjust the output terminal voltage of the analog inverter, that is, the initial voltage value applied to the resistive memory.

[0041] 4) Connect the resistive memory to form a complete feedback loop in the continuous time domain

[0042] The feedback module is composed of another operational amplifier OA and a fixed resistor on the feedback loop at the output end. The working mode of the OA is a positive feedback amplifier. The input end of the feedback module is directly connected to the output end of the reference resistor module and is connected to the resistive memory module through a switch S1. The fixed resistor on the feedback loop at the output end is directly connected to the inverting input resistance of the analog inverter in the input module. The feedback module is used to sense the difference in conductance between the memory and the reference resistor, and is used for direct feedback of the electrical signal (voltage signal) to form a feedback loop. The switch S1 is closed, so that the input module, the reference resistor module, the resistive memory module and the feedback module form a complete feedback loop in the continuous time domain.

[0043] 5) Programming the resistive memory conductance state to near the target conductance state and realizing the function of autonomous judgment and termination;

[0044] In the feedback loop, the feedback module detects the conductivity difference between the memory and the reference resistor, and feeds back the conductivity difference to the input module through the output terminal in the form of a voltage signal, completing the feedback process of the electrical signal. In this process, the resistive memory is programmed, and its conductivity value continues to increase, approaching the target conductivity value, and the gap between the two becomes smaller and smaller. The input module will autonomously and continuously adjust the voltage value applied to the resistive memory according to the voltage signal obtained by feedback until the conductivity value of the resistive memory reaches near the target conductivity value. At this time, the voltage applied to the resistive memory will drop to 0, which means that the writing process is terminated. The entire judgment process and the termination of the writing process are autonomously realized by the feedback loop through the feedback of the electrical signal, so the resistive memory can be written to any analog conductivity state.

[0045] The resistive memory may be a resistive memory, a phase change memory, a magnetic memory or a ferroelectric memory.

[0046] Figure 2 is a block diagram of the feedback loop used in the analog conductivity state high-speed writing method of the resistive memory; the loop is Figure 1The feedback loop mentioned above is composed of an input module (S201), a reference resistor module (S202), a resistive memory module (S203) and a feedback module (S204). S201 is used to define the initial conditions imposed by the method on the resistive memory, preventing the resistive memory from being programmed to an extremely high conductivity state within a time scale of hundreds of picoseconds under high voltage conditions, and ensuring that the time constant of the scheme can match the time constant of the resistive memory. The S202 module is used to select the conductivity of the reference resistor, which is the target conductivity state of the memory, meaning that the resistive memory of the S203 module will be programmed to the vicinity of the conductivity state. The S204 module is used for direct feedback of electrical signals (voltage signals) to form a feedback loop for judging the difference between the current conductivity value of the resistive memory and the target conductivity value. Once the target conductivity value is reached, the module will reduce or even remove the electrical stimulation applied to the S203 module to achieve the function of autonomous judgment and termination. It is worth noting that the programming process is implemented in one step in continuous time, thus greatly improving the programming speed.

[0047] Figure 3 This is the feedback loop circuit diagram and principle used in the analog conductivity state high-speed writing method of the resistive memory of the present invention. The reference resistance module is composed of a series of analog switches and fixed resistors. By closing different switches or different numbers of switches, the conductivity value of the reference resistor can be adjusted. The input module is composed of an analog inverter constructed by input voltage, fixed resistors and an OA, and the feedback module is composed of an OA and a fixed resistor. The OA works in the positive feedback amplifier (PFA). These modules are connected to form a feedback loop. In the present invention, the agreed target conductivity value is expressed as G target , the resistance memory conductivity is G mem , the conductivity of other fixed resistors is G0, V1, V2, V3 and V4 represent the input terminal of the inverter, the positive input terminal of PFA, the reverse input terminal and output terminal node of inverter OA respectively, V in is the input voltage, V out is the output node of PFA. Among them, the input module is connected to the feedback module and the reference resistor module through the V1 node, and is connected to the resistive memory module through the node V4. The resistive memory module is connected to the common node V2 of the feedback module and the reference resistor module through the switch S1. When the loop is working, if both OAs are in the linear amplification area, according to the "virtual short, virtual open" characteristics of the operational amplifier, V2 is 0, V1 = -V4, and the current flowing from the V2 node to the input end of the amplifier is 0, which means that at this time the voltage drop of the reference resistor and the memory is the same, and the current flowing is also the same. At this time, the two should have the same or similar conductance values. Before the circuit reaches stability, that is, before the conductance of the memory approaches the target conductance value, the PFA works in the saturation area, and the dynamic equation of the circuit should be analyzed, and the convention Gtarget / G mem =c,G target / G0=d, according to Kirchhoff's law and Ohm's law, we can get:

[0048] (V in -V1)G0=(V1-V out )G0+(V1-V3)G0+(V1-V2)G target (V2-V4)G mem =(V1-V2)G target

[0049] (V1-V3)G0=(V3-V4)G0

[0050] The physical meanings of the above three equations are: Input voltage V in The input current provided to the V1 node is equal to the sum of the currents flowing out of the node. The currents flowing through the reference resistor and the memory are the same, and the currents flowing through the two fixed resistors in the analog inverter are the same. Combined with the OA open-loop transfer function equation: V out =L(s)V2, V4 = -L(s)V3, where L0 is the DC gain multiple, and ω0 is the 3-dB bandwidth. Combining the above five equations and omitting the small terms, the dynamic equation of the circuit can be obtained:

[0051]

[0052] That is, equation (1) in the attached figure. When c = 1, the coefficients of the last two terms in equation (1) become 0. At this time, the contribution of the previous small terms to the second-order frequency domain equation cannot be ignored, so the dynamic equation of the circuit becomes:

[0053]

[0054] This is equation (2) in the attached figure. According to the dynamic equation of the circuit, when c≤1, that is, when the conductance of the memory is greater than the target conductance value, the poles of the circuit are distributed on the left side of the complex plane. At this time, the voltage values ​​of the V2 and V4 nodes will evolve to 0, and both OAs are in the linear amplification region, which means that the circuit can automatically reduce the voltage difference between its two ends to 0 when the variable resistor reaches the target conductance, and play the role of automatic judgment and termination. When c>1, that is, when the conductance of the variable resistor is less than the target conductance value, the poles of the circuit are distributed on the right side of the complex plane. At this time, the voltage of the V2 node is not 0, which means that the PFA is in the saturation region. At this time, V out =V DD , V DD is the supply voltage of the circuit. According to Kirchhoff's current law, the voltage drop across the variable resistor can be calculated as: At this time, the circuit performs a programming process on the variable resistor. Therefore, before the variable resistor is connected to the circuit, the c value is infinite, and the voltage difference between the V2 and V4 nodes is Therefore, by adjusting V in To adjust the initial voltage drop of the loop to adapt to different types of resistive memories.

[0055] Figure 4 It is a schematic diagram of the analog conductivity state high-speed writing method of the resistive memory of the present invention applied to a variable resistor array, which includes a currently most widely used N×N 1 transistor 1 resistor (1T1R) resistive memory array, an input circuit, a control circuit, a feedback loop proposed by the present invention, a WL / BL / SL drive circuit and a multiplexer and other modules. The control circuit is responsible for sending control signals to each module, controlling the connectivity and working mode between the modules. When the circuit is working, the control circuit transmits a digital signal corresponding to the target conductivity state to the feedback loop, adjusts the conductivity state of the reference resistor module to the target conductivity value, and then the control circuit controls and adjusts the input voltage of the input module in the feedback loop to provide the initial voltage drop of the memory, and then the control circuit transmits an address signal to the multiplexer of WL / BL / SL, so that the corresponding resistive memory can be connected to the feedback loop, and the resistive memory can be written to the target analog conductivity state. Repeating and performing the above operations in sequence can write all the memories in the array to the set analog conductivity state. The input circuit and ADC module are used for voltage signal input and output signal reading in the subsequent calculation mode and do not participate in the programming process of the 1T1R array.

[0056] The embodiments described above are not intended to limit the present invention. Any person skilled in the art may make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention is defined by the scope of the claims.

Claims

1. A method for high-speed writing of analog conductivity state of resistive memory, characterized in that: The method is based on an analog circuit consisting of an input module, a reference resistor module, a resistive memory module and a feedback module. The modules are connected based on electrical signals to form a feedback loop in a continuous time domain. The resistive memory is autonomously programmed to reach any analog conductivity state in a continuous time, and the feedback signal is used to realize an intelligent judgment on whether the target conductivity state is reached and the writing process is terminated. The method comprises the following steps: 1) Reset the conductivity state of the resistive memory 2) Adjust the conductivity value of the reference resistance module and select the target conductivity state 3) The input power supply of the input module provides a DC voltage to adjust the initial voltage value applied to the resistive memory. 4) Connect the resistive memory to form a complete feedback loop in the continuous time domain 5) Program the resistive memory conductance state to near the target conductance state and realize the function of autonomous judgment and termination In the feedback loop in the continuous time domain, the feedback module detects the conductivity difference between the memory and the reference resistor, and feeds back the conductivity difference to the input module through the output terminal in the form of a voltage signal, completing the feedback process of the electrical signal. In this process, the resistive memory is programmed, and its conductivity value continues to increase, approaching the target conductivity value, and the gap between the two becomes smaller and smaller. The input module will autonomously and continuously adjust the voltage value applied to the resistive memory according to the voltage signal obtained by feedback until the conductivity value of the resistive memory reaches near the target conductivity value. At this time, the voltage applied to the resistive memory will drop to 0, and the writing process is terminated. The entire judgment process and the termination of the writing process are autonomously realized by the feedback loop through the feedback of the electrical signal, so the resistive memory can be written to any analog conductivity state.

2. The analog conductance state high-speed writing method of a resistive memory as claimed in claim 1, characterized in that: The resistive memory module of step 1) is composed of a single resistive memory, which is connected to a common node connected to the feedback module and the reference resistance module through a switch S1; a reset voltage pulse is applied to the resistive memory, that is, the top electrode TE end of the resistive memory is grounded and a positive voltage pulse is applied to the bottom electrode BE end. After the Reset operation is completed, the resistive memory is adjusted to a low conductivity state.

3. The analog conductance state high-speed writing method of a resistive memory as claimed in claim 1, characterized in that: The reference resistor module in step 2) includes a series of analog switches and fixed resistors. The conductance value of the reference resistor module is adjusted by closing different switches or different numbers of switches. The conductance value is used as the target conductance state to be written into the resistive memory.

4. The analog conductance state high-speed writing method of a resistive memory as claimed in claim 1, characterized in that: The input module of step 3) includes an input power supply, a fixed resistor and an analog inverter constructed by an operational amplifier OA, the input power supply is directly connected to the inverting input terminal resistance of the analog inverter through a fixed resistor, and the output terminal of the analog inverter is connected to the resistive memory; the input power supply provides a DC voltage of a certain amplitude for the circuit to adjust the output terminal voltage of the analog inverter, that is, the initial voltage value applied to the resistive memory.

5. The analog conductance state high-speed writing method of a resistive memory as claimed in claim 1, characterized in that: The feedback module in the feedback loop of step 4) is composed of another operational amplifier OA and a fixed resistor on the output feedback loop. The OA works as a positive feedback amplifier. The input end of the feedback module is directly connected to the output end of the reference resistor module and is connected to the resistive memory module through a switch S1. The fixed resistor on the output feedback loop is directly connected to the inverting input resistance of the analog inverter in the input module. The feedback module is used to sense the difference in conductance between the memory and the reference resistor, and is used for direct feedback of the voltage signal to form a feedback loop. Closing the switch S1 allows the input module, the reference resistor module, the resistive memory module and the feedback module to form a complete feedback loop in the continuous time domain.

6. The analog conductance state high-speed writing method of a resistive memory as claimed in claim 1, characterized in that: The resistive memory is a resistive memory, a phase change memory, a magnetic memory or a ferroelectric memory.

7. A method for high-speed writing of analog conductance state of a resistive memory array, characterized in that: The resistive memory as claimed in claim 1 is used as a single device selected in a resistive memory array, and the feedback loop in the continuous time domain as claimed in claim 1 is applied to the analog conductivity state writing circuit of the resistive memory array; the analog conductivity state writing circuit of the resistive memory array includes an N×N 1T1R array, a control circuit, the feedback loop in the continuous time domain, a driving circuit and a multiplexer of a word line WL / bit line BL / source line SL, the multiplexer near the bit line BL and the source line SL is equivalent to the external interface of the resistive memory module, the word line WL corresponding to the resistive memory determines whether the resistive memory module is connected to the feedback loop, and the control circuit is responsible for sending control signals to each module, controlling the connectivity and working mode between modules; the analog conductivity state high-speed writing method of the resistive memory as claimed in claim 1 is applied to the resistive memory array, comprising the following steps: S1) Execute the reset operation on the selected resistive memory First, a certain resistive memory in the 1T1R array is selected, that is, a control signal is input through the control circuit to open the WL corresponding to the resistive memory and the corresponding SL and BL are connected to the multiplexer, and a reset voltage pulse is applied to reset the selected resistive memory to a low conductance state; S2) Adjust the reference resistance module and select the target conductivity state The control circuit transmits a control signal to the reference resistor module, and adjusts the conductance value of the reference resistor module by closing different switches or different numbers of switches, and the conductance value is used as the target conductance state to be written into the resistive memory; S3) controlled by the control circuit to adjust the input voltage of the input module in the feedback loop to provide an initial voltage drop of the memory; S4) connecting the SL and BL corresponding to the memory to the feedback loop through a multiplexer, turning on the corresponding WL, connecting the selected resistive memory to the feedback loop, and forming a complete feedback loop in the continuous time domain; S5) The selected resistive memory will be written near the target conductivity state in a continuous time, and realize the function of autonomous judgment and termination; By repeating the above steps S1) to S5), all memories in the array can be written into the set simulated conductance state.

Images