CN110137348B - A Multiplexing Multi-valued Resistive Switching Structure and Its Neural Network Formed - Google Patents
A Multiplexing Multi-valued Resistive Switching Structure and Its Neural Network Formed Download PDFInfo
- Publication number
- CN110137348B CN110137348B CN201910290334.XA CN201910290334A CN110137348B CN 110137348 B CN110137348 B CN 110137348B CN 201910290334 A CN201910290334 A CN 201910290334A CN 110137348 B CN110137348 B CN 110137348B
- Authority
- CN
- China
- Prior art keywords
- multiplexing
- mos
- resistance change
- resistive
- random access
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000013528 artificial neural network Methods 0.000 title claims description 13
- 108010001267 Protein Subunits Proteins 0.000 claims abstract 2
- 229910044991 metal oxide Inorganic materials 0.000 claims abstract 2
- 150000004706 metal oxides Chemical class 0.000 claims abstract 2
- 239000004065 semiconductor Substances 0.000 claims abstract 2
- 230000002093 peripheral effect Effects 0.000 claims description 17
- 210000002569 neuron Anatomy 0.000 claims description 16
- 210000004027 cell Anatomy 0.000 claims description 2
- 238000000034 method Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 239000010410 layer Substances 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 239000002356 single layer Substances 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
- 210000000225 synapse Anatomy 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G11—INFORMATION STORAGE
- G11C—STATIC STORES
- G11C13/00—Digital stores characterised by the use of storage elements not covered by groups G11C11/00, G11C23/00, or G11C25/00
- G11C13/0002—Digital stores characterised by the use of storage elements not covered by groups G11C11/00, G11C23/00, or G11C25/00 using resistive RAM [RRAM] elements
- G11C13/0021—Auxiliary circuits
-
- H—ELECTRICITY
- H10—SEMICONDUCTOR DEVICES; ELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
- H10N—ELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
- H10N70/00—Solid-state devices having no potential barriers, and specially adapted for rectifying, amplifying, oscillating or switching
- H10N70/011—Manufacture or treatment of multistable switching devices
-
- H—ELECTRICITY
- H10—SEMICONDUCTOR DEVICES; ELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
- H10N—ELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
- H10N70/00—Solid-state devices having no potential barriers, and specially adapted for rectifying, amplifying, oscillating or switching
- H10N70/20—Multistable switching devices, e.g. memristors
Landscapes
- Engineering & Computer Science (AREA)
- Manufacturing & Machinery (AREA)
- Semiconductor Memories (AREA)
Abstract
Description
技术领域technical field
本发明涉及阻变存储器,具体涉及一种多路复用多值阻变结构及其形成的神经网络。The invention relates to a resistive variable memory, in particular to a multiplexing multi-value resistive variable structure and a neural network formed therefrom.
背景技术Background technique
近年来,阻变存储器(RRAM)等新型存储器的发展成为存储器领域的一个新的推动力。阻变存储器因为功耗低,速度快,面积小的特点,在未来具备非常广阔的应用前景。除了传统的存储领域外,在人工智能领域的神经网络系统架构中,也可以起到重要的作用。阻变存储器现在已经存在研究,将单元往多值存储的方向发展,即一个单元通过不同的电阻大小,完成多比特数据的存储。In recent years, the development of new types of memory such as resistive RAM (RRAM) has become a new impetus in the field of memory. Due to the characteristics of low power consumption, high speed and small area, resistive memory has very broad application prospects in the future. In addition to the traditional storage field, it can also play an important role in the neural network system architecture in the field of artificial intelligence. Resistive variable memory has already been researched, and the unit is developed in the direction of multi-value storage, that is, one unit completes the storage of multi-bit data through different resistance sizes.
多值存储器的这种阻变的模拟特性,由于可以输出模拟量变化的电流值,在类脑神经网络系统中,是一种可以作为类脑神经突触的理想器件。目前,多值存储器还没有成熟的制备结果,所以现阶段可以采用多个单值的RRAM单元来模拟这种多值特性,如图1所示。The resistance-change analog characteristic of the multi-valued memory is an ideal device that can be used as a brain-like synapse in the brain-like neural network system because it can output the current value of the analog quantity change. At present, there is no mature preparation result of multi-valued memory, so at this stage, multiple single-valued RRAM cells can be used to simulate this multi-valued characteristic, as shown in Figure 1.
但是这类结构应用在神经网络中,会存在以下一些缺点。首先,每一个单值RRAM单元都需要配置一个灵敏放大器(SA)作为数据读出,其次,需要有数字-模拟转换器(DAC)将单值数据数字信号数据转化成一个具有多值特性的模拟量。这些额外的电路结构会提高整个神经网络系统在面积和功耗上的开销,即便平均到单个RRAM上,开销也会比较大。因此,设计一种可以将RRAM的电阻特性通过一些简单的附加电路结构,直接转化成具有多值模拟性质的电学量,从而节省面积和功耗,在现有的技术条件下是一件具有现实意义的事情。However, when this type of structure is applied to a neural network, there are some disadvantages as follows. First of all, each single-valued RRAM unit needs to be equipped with a sense amplifier (SA) for data readout, and secondly, a digital-to-analog converter (DAC) is required to convert the single-valued data digital signal data into an analog with multi-valued characteristics. quantity. These additional circuit structures will increase the overhead of the entire neural network system in terms of area and power consumption. Even if it is averaged on a single RRAM, the overhead will be relatively large. Therefore, it is realistic to design a method that can directly convert the resistance characteristics of RRAM into electrical quantities with multi-valued analog properties through some simple additional circuit structures, thereby saving area and power consumption. meaningful things.
发明内容Contents of the invention
本发明的目的是提供一种多路复用多值阻变结构及其形成的神经网络,利用了简化的MOS选择单元来实现多值的模拟量输出,进而实现电路模块复用,减少了灵敏放大器和数字-模拟转换器等电路模块,节省了电路面积。The purpose of the present invention is to provide a multiplexing multi-valued resistive variable structure and the neural network formed thereof, which utilizes a simplified MOS selection unit to realize multi-valued analog output, thereby realizing multiplexing of circuit modules and reducing sensitivity Circuit modules such as amplifiers and digital-to-analog converters save circuit area.
为了实现上述目的,本发明采用如下技术方案:一种多路复用多值阻变结构,包括阻变存储器单元和MOS选择单元,所述阻变存储器单元包括M个阻变存储器子单元,所述MOS选择单元包括N个并联的复用MOS管;其中,M和N均为大于1的整数;In order to achieve the above object, the present invention adopts the following technical solution: a multiplexing multi-valued resistive switch structure, including a resistive switch memory unit and a MOS selection unit, the resistive switch memory unit includes M resistive switch memory subunits, the The MOS selection unit includes N parallel multiplexing MOS transistors; wherein, M and N are integers greater than 1;
所述阻变存储器子单元包括阻变存储器和控制MOS管,其中,所述阻变存储器的一端连接该阻变存储器子单元对应的控制信号,另一端连接所述控制MOS管的漏极,所述控制MOS管的栅极连接该阻变存储器子单元对应的选择信号,且M个控制MOS管的源极与所述N个复用MOS管的漏极共同连接至同一节点,所述N个复用MOS管的源极共同连接至输出端口,每个复用MOS管的栅极连接对应的导通信号。The resistive memory subunit includes a resistive memory and a control MOS transistor, wherein one end of the resistive memory is connected to the control signal corresponding to the resistive memory subunit, and the other end is connected to the drain of the control MOS transistor. The gate of the control MOS transistor is connected to the selection signal corresponding to the RRAM subunit, and the sources of the M control MOS transistors and the drains of the N multiplexing MOS transistors are connected to the same node, and the N The sources of the multiplexed MOS transistors are commonly connected to the output port, and the gates of each multiplexed MOS transistor are connected to corresponding conduction signals.
进一步地,所述N个复用MOS管在导通时,对应的电导值不同。Further, when the N multiplexing MOS transistors are turned on, the corresponding conductance values are different.
进一步地,所述N个复用MOS管在导通时,对应的电导值分别为GT,2GT,22GT至2N-1GT,通过控制所述导通信号,使得所述MOS选择单元可以产生2N-1个不同的电导值,其中,GT大于零。Further, when the N multiplexing MOS transistors are turned on, the corresponding conductance values are G T , 2G T , 2 2 G T to 2 N-1 G T , and by controlling the conduction signal, all The MOS selection unit can generate 2 N -1 different conductance values, wherein G T is greater than zero.
进一步地,所述N个复用MOS管中的电导值满足: 其中,Ron表示所述阻变存储器处于低阻状态时的电阻值,Roff表示所述阻变存储器处于高阻状态时的电阻值。Further, the conductance value in the N multiplexing MOS transistors satisfies: Wherein, R on represents the resistance value when the RRAM is in a low-resistance state, and R off represents the resistance value when the RRAM is in a high-resistance state.
进一步地,通过控制所述控制信号,使得所述M个阻变存储器子单元中只有一个处于可读写状态,所述多路复用多值阻变结构可以产生2N+1-2个电阻值。Further, by controlling the control signal so that only one of the M resistive variable memory subunits is in a readable and writable state, the multiplexed multi-valued resistive variable structure can generate 2 N+1-2 resistors value.
进一步地,所述控制MOS管和复用MOS管的源极和漏极可以互换。Further, the source and drain of the control MOS transistor and the multiplexing MOS transistor can be interchanged.
进一步地,所述多路复用多值阻变结构还包括外围电路,用于控制M个阻变存储器子单元中只有一个处于可读写状态。Further, the multiplexed multi-valued resistive switch structure further includes a peripheral circuit for controlling only one of the M resistive switch memory subunits to be in a readable and writable state.
进一步地,所述外围电路还包括存储单元,用于存储N个复用MOS管的导通和断开状态。Further, the peripheral circuit further includes a storage unit for storing the ON and OFF states of the N multiplexed MOS transistors.
一种神经网络包括多值阻变阵列、A个前神经元电路、B个后神经元电路和外围控制电路,所述多值阻变阵列包括A行B列的如上所述的多路复用多值阻变结构,每一行的多路复用多值阻变结构连接一个前神经元电路,每一列的多路复用多值阻变结构连接一个后神经元电路,所述外围控制电路用于控制选择其中一个多路复用多值阻变结构,且该多路复用多值阻变结构的控制和输出由与之对应的前神经元电路和后神经元电路协同完成,其中,A和B均为大于1的整数。A neural network includes a multi-valued resistive variable array, A front neuron circuits, B rear neuron circuits and peripheral control circuits, the multi-valued resistive variable array includes multiplexing as described above in A row and B column Multi-valued resistive structure, the multiplexed multi-valued resistive structure of each row is connected to a pre-neuron circuit, the multiplexed multi-valued resistive structure of each column is connected to a post-neuron circuit, and the peripheral control circuit uses One of the multiplexed multi-valued resistive structures is selected for control, and the control and output of the multiplexed multi-valued resistive structure is completed by the corresponding pre-neuron circuit and post-neuron circuit, wherein, A and B are both integers greater than 1.
本发明的有益效果为:本发明利用了简化的MOS选择单元来实现多值的模拟量输出,进而实现电路模块复用,减少了灵敏放大器和数字-模拟转换器等电路模块,节省了电路面积;此外,由于采用多路复用结构,复用MOS管作为复用部分,面积和功耗开销将会分摊到每一个阻变存储器子单元上,那么所需要的平均的功耗花费也会得到显著降低。The beneficial effects of the present invention are: the present invention utilizes the simplified MOS selection unit to realize multi-valued analog output, thereby realizing multiplexing of circuit modules, reducing circuit modules such as sensitive amplifiers and digital-to-analog converters, and saving circuit area ; In addition, due to the multiplexing structure, the multiplexing MOS tube is used as the multiplexing part, and the area and power consumption overhead will be allocated to each resistive memory subunit, so the required average power consumption will also be obtained Significantly lower.
附图说明Description of drawings
附图1为现有技术中基于RRAM单值存储单元构成的多值阻变单元示意图。Figure 1 is a schematic diagram of a multi-valued resistive variable unit based on a RRAM single-valued storage unit in the prior art.
附图2为本发明一种多路复用多值阻变结构的示意图。Figure 2 is a schematic diagram of a multiplexed multi-valued resistive switching structure of the present invention.
附图3为本发明一种多路复用多值阻变结构的外围架构示意图。Figure 3 is a schematic diagram of the peripheral architecture of a multiplexed multi-valued resistive switching structure according to the present invention.
附图4为本发明一种神经网络的示意图。Accompanying drawing 4 is the schematic diagram of a kind of neural network of the present invention.
图中:1存储单元,2外围电路,3前神经元电路,4后神经元电路,5标准工艺电路层,6阻变存储器子单元,7多路复用多值阻变结构,8外围控制电路。In the figure: 1 storage unit, 2 peripheral circuit, 3 front neuron circuit, 4 post neuron circuit, 5 standard process circuit layer, 6 resistive variable memory subunit, 7 multiplexing multi-value resistive variable structure, 8 peripheral control circuit.
具体实施方式Detailed ways
为使本发明的目的、技术方案和优点更加清楚,下面结合附图对本发明的具体实施方式做进一步的详细说明。In order to make the purpose, technical solution and advantages of the present invention clearer, the specific implementation manners of the present invention will be further described in detail below in conjunction with the accompanying drawings.
如附图2所示,本发明提供的一种多路复用多值阻变结构,包括阻变存储器单元和MOS选择单元。阻变存储器单元包括M个阻变存储器子单元,每一个阻变存储器子单元由通识的1T1R结构组成,即包括阻变存储器和控制MOS管,其中,阻变存储器的一端连接该阻变存储器子单元对应的控制信号,另一端连接控制MOS管的漏极,控制MOS管的栅极连接该阻变存储器子单元对应的选择信号,且M个控制MOS管的源极与N个复用MOS管的漏极共同连接至同一节点。其中,控制MOS管作为阻变存储器子单元的控制管,通过外部加到控制MOS管栅极上的电平高低作为是否选择该阻变存储器子单元的判决条件。上述控制MOS管的源极和漏极可以互换,M为大于1的整数。As shown in FIG. 2 , a multiplexing multi-valued resistive switch structure provided by the present invention includes a resistive switch memory unit and a MOS selection unit. The resistive memory unit includes M resistive memory subunits, and each resistive memory subunit is composed of a common 1T1R structure, that is, includes a resistive memory and a control MOS tube, wherein one end of the resistive memory is connected to the resistive memory The control signal corresponding to the subunit, the other end is connected to the drain of the control MOS transistor, the gate of the control MOS transistor is connected to the selection signal corresponding to the resistive variable memory subunit, and the sources of the M control MOS transistors are connected to the N multiplexed MOS transistors. The drains of the tubes are commonly connected to the same node. Wherein, the control MOS transistor is used as the control transistor of the resistive variable memory subunit, and the level level externally applied to the gate of the control MOS transistor is used as a judgment condition for whether to select the resistive variable memory subunit. The source and drain of the above-mentioned control MOS transistors can be interchanged, and M is an integer greater than 1.
MOS选择单元包括N个并联的复用MOS管;N个复用MOS管的源极共同连接至输出端口,每个复用MOS管的栅极连接对应的导通信号。通过外部加在复用MOS管栅极上的电平高低作为是否选择该复用MOS管的判决条件,上述复用MOS管的源极和漏极可以互换,N为大于1的整数。The MOS selection unit includes N multiplexing MOS transistors connected in parallel; the sources of the N multiplexing MOS transistors are commonly connected to the output port, and the gate of each multiplexing MOS transistor is connected to a corresponding conduction signal. The level applied externally to the gate of the multiplexing MOS transistor is used as the judgment condition for whether to select the multiplexing MOS transistor. The source and drain of the above multiplexing MOS transistor can be interchanged, and N is an integer greater than 1.
请继续参阅附图2,阻变存储器自身存在低阻状态和高阻状态下的两种阻值,分别为Ron和Roff,控制MOS管在导通时也存在电阻RT,该电阻远小于阻变存储器低阻状态和高阻状态下的两种阻值。通过外部加在阻变存储器子单元对应的控制信号上的电平高低作为是否选择该阻变存储器子单元的判决条件,复用MOS管的电导值可以通过调整其宽长比来调整,为了使得本发明中多路复用多值阻变结构可变的阻值尽量多,设置N个复用MOS管在导通时,对应的电导值不同。具体可以设置N个复用MOS管在导通时,对应的电导值呈2的指数阶梯状分布,通过控制上述的导通信号,即选择电路中复用MOS管的个数,其中,通过导通信号的变化,可以使得0-N个复用MOS管连接在本发明多值阻变结构中,从而使得MOS选择单元可以产生2N-1个不同的电导值;且2N-1个不同的电导值可以作为RRAM自身电阻变化的一种有效补充,当只有一个阻变存储器子单元处于读写状态时,由于其对应的阻变存储器具有低阻和高阻两种状态,本发明结构一共可以产生(2N+1-2)个不同阻值,同时可以把RRAM阻值变化看作阻值变化的最高位。Please continue to refer to Figure 2. The RRAM itself has two resistance values in a low-resistance state and a high-resistance state, which are R on and R off respectively. When the control MOS transistor is turned on, there is also a resistance R T , which is far from It is smaller than the two resistance values of the RRAM in the low-resistance state and the high-resistance state. By externally adding the level of the control signal corresponding to the resistive memory subunit as the judgment condition of whether to select the resistive memory subunit, the conductance value of the multiplexed MOS transistor can be adjusted by adjusting its width-to-length ratio, in order to make In the present invention, the variable resistance values of the multiplexed multi-value resistive structure are as many as possible, and when N multiplexed MOS transistors are turned on, the corresponding conductance values are different. Specifically, it can be set that when the N multiplexed MOS transistors are turned on, the corresponding conductance values are distributed in an exponential ladder shape of 2. By controlling the above-mentioned turn-on signal, the number of multiplexed MOS transistors in the circuit is selected. The change of the pass signal can make 0-N multiplexing MOS transistors connected in the multi-value resistance switching structure of the present invention, so that the MOS selection unit can generate 2 N -1 different conductance values; and 2 N -1 different conductance values The conductance value of RRAM can be used as an effective supplement for RRAM's own resistance change. When only one RRAM subunit is in the read-write state, since its corresponding RRAM has two states of low resistance and high resistance, the structure of the present invention has a total of (2 N+1 -2) different resistance values can be generated, and at the same time, the change of RRAM resistance value can be regarded as the highest bit of resistance value change.
本发明中上述提及的控制信号用于控制与之连接的阻变存储器处于低阻状态或者高阻状态;选择信号用于控制在M个阻变存储器子单元中选择某一个阻变存储器子单元处于可读写状态;导通信号用于控制对应的复用MOS管是否需要导通。The control signal mentioned above in the present invention is used to control the resistive memory connected to it to be in a low-impedance state or a high-impedance state; the selection signal is used to control the selection of a certain resistive memory subunit among the M resistive memory subunits It is in a readable and writable state; the conduction signal is used to control whether the corresponding multiplexed MOS transistor needs to be conducted.
优选地,将N个复用MOS管的导通电导值设为GT,2GT至2N-1GT,通过控制施加在复用MOS管上对应的导通信号,使得MOS选择单元可以产生2N-1个不同的电导值,其中,GT大于零。与RRAM串联后,可以得到的阻值为一共(2N+1-2)个。因为Ron<Roff,通过设计使得 当只有一个阻变存储器子单元处于读写状态时,由于其对应的阻变存储器具有低阻和高阻两种状态,本发明的多路复用多值阻变结构一定可以得到共(2N+1-2)种不同的电阻值。Preferably, the on-conductance values of the N multiplexed MOS transistors are set to G T , 2G T to 2 N-1 G T , and by controlling the corresponding conduction signal applied to the multiplexed MOS transistors, the MOS selection unit can 2 N -1 different conductance values are produced, where G T is greater than zero. After being connected in series with RRAM, the resistance that can be obtained is A total of (2 N+1 -2) pieces. Since R on < R off , by design such that When only one resistive variable memory subunit is in the read-write state, since its corresponding resistive variable memory has two states of low resistance and high resistance, the multiplexing multi-valued resistive variable structure of the present invention must be able to obtain a total of (2 N +1 -2) different resistor values.
本发明利用了简化的MOS选择单元来实现多值的模拟量输出,进而实现电路模块复用,减小了灵敏放大器和数字-模拟转换器等电路模块,节省了电路面积;此外,由于采用多路复用结构,复用MOS管作为复用部分,面积和功耗开销将会分摊到每一个阻变存储器子单元上,那么所需要的平均功耗花费也会得到显著降低。The present invention utilizes the simplified MOS selection unit to realize multi-valued analog output, thereby realizing multiplexing of circuit modules, reducing circuit modules such as sensitive amplifiers and digital-to-analog converters, and saving circuit area; in addition, due to the use of multiple Multiplexing structure, multiplexing MOS transistors as the multiplexing part, the area and power consumption overhead will be allocated to each resistive memory subunit, and the required average power consumption will also be significantly reduced.
请参阅附图3,本发明中多路复用多值阻变结构还包括外围电路2,其中,外围电路2中又包括存储单元1,外围电路2用于确保在使用时,每次只能选择其中的一个阻变存储器子单元,使其处于可读写状态,其余的阻变存储器子单元均处于关闭状态;其中的存储单元1用于存储M个复用MOS管的导通和断开状态。具体可以采用非挥发存储单元来记录相应复用MOS管的开关状态,保证阻值在下电后不会丢失。Please refer to accompanying
请参阅附图4,本发明上述多路复用多值阻变结构形成的神经网络,包括多值阻变阵列、A个前神经元电路3、B个后神经元电路4和外围控制电路8,多值阻变阵列包括A行B列的多路复用多值阻变结构7;多路复用多值阻变结构在工艺制备上,阻变存储器子单元6单独为一层,在垂直方向的标准工艺电路层5上,制作相应的复用MOS管、控制MOS管以及其余需要使用MOS管搭建的电路结构。每一行的多路复用多值阻变结构7连接一个前神经元电路3,每一列的多路复用多值阻变结构7连接一个后神经元电路4,外围控制电路8用于控制选择其中一个多路复用多值阻变结构,且该多路复用多值阻变结构的控制和输出由与之对应的前神经元电路和后神经元电路协同完成,多路复用多值阻变结构的模拟输出可以由后神经元直接接收处理,与典型的RRAM存储器阵列相比,省略了相应的灵敏放大器电路结构。Please refer to accompanying drawing 4, the neural network formed by the multiplexing multi-valued resistive structure of the present invention includes a multi-valued resistive array, A
神经网络还可以集成一个适当规模的非挥发存储阵列,用来存储需要掉电记忆的阻值信息。因为多值阻变阵列理论上存在多个并联的多路复用多值阻变结构,所以相当于使用了一套外围电路实现了多路多路复用多值阻变结构的控制,因此是一个节省面积和功耗的神经网络系统。The neural network can also integrate a non-volatile storage array of appropriate size to store resistance information that needs power-down memory. Because the multi-value resistive variable array theoretically has multiple parallel multiplexed multi-valued resistive structures, it is equivalent to using a set of peripheral circuits to realize the control of multiple multiplexed multi-valued resistive structures, so it is A neural network system that saves area and power.
以上所述仅为本发明的优选实施例,所述实施例并非用于限制本发明的专利保护范围,因此凡是运用本发明的说明书及附图内容所作的等同结构变化,同理均应包含在本发明所附权利要求的保护范围内。The above are only preferred embodiments of the present invention, and the embodiments are not intended to limit the scope of patent protection of the present invention, so all equivalent structural changes made by using the description and drawings of the present invention should be included in the same reason Within the protection scope of the appended claims of the present invention.
Claims (9)
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910290334.XA CN110137348B (en) | 2019-04-11 | 2019-04-11 | A Multiplexing Multi-valued Resistive Switching Structure and Its Neural Network Formed |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910290334.XA CN110137348B (en) | 2019-04-11 | 2019-04-11 | A Multiplexing Multi-valued Resistive Switching Structure and Its Neural Network Formed |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110137348A CN110137348A (en) | 2019-08-16 |
CN110137348B true CN110137348B (en) | 2023-01-31 |
Family
ID=67569680
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910290334.XA Active CN110137348B (en) | 2019-04-11 | 2019-04-11 | A Multiplexing Multi-valued Resistive Switching Structure and Its Neural Network Formed |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110137348B (en) |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4378595A (en) * | 1980-03-25 | 1983-03-29 | The Regents Of The University Of California | Synchronous multivalued latch |
CN1848294A (en) * | 2005-04-12 | 2006-10-18 | 株式会社东芝 | Non-volatile semiconductor memory for storing multi-valued data |
JP2008227267A (en) * | 2007-03-14 | 2008-09-25 | Fujitsu Ltd | Resistance change memory forming method, resistance change memory, and resistance change memory manufacturing method |
CN101783170A (en) * | 2009-01-21 | 2010-07-21 | 中国科学院微电子研究所 | Circuit and method for driving resistance transition type memory to realize multi-value storage |
CN102543147A (en) * | 2012-01-18 | 2012-07-04 | 北京大学 | Reading circuit and reading method of multilevel storage circuit |
CN104681085A (en) * | 2015-03-03 | 2015-06-03 | 中国科学院微电子研究所 | Resistive random access memory based on flip coding circuit and corresponding data storage method |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP4728726B2 (en) * | 2005-07-25 | 2011-07-20 | 株式会社東芝 | Semiconductor memory device |
US20140159770A1 (en) * | 2012-12-12 | 2014-06-12 | Alexander Mikhailovich Shukh | Nonvolatile Logic Circuit |
JP2015076556A (en) * | 2013-10-10 | 2015-04-20 | ソニー株式会社 | Memory unit, writing method and reading method |
-
2019
- 2019-04-11 CN CN201910290334.XA patent/CN110137348B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4378595A (en) * | 1980-03-25 | 1983-03-29 | The Regents Of The University Of California | Synchronous multivalued latch |
CN1848294A (en) * | 2005-04-12 | 2006-10-18 | 株式会社东芝 | Non-volatile semiconductor memory for storing multi-valued data |
JP2008227267A (en) * | 2007-03-14 | 2008-09-25 | Fujitsu Ltd | Resistance change memory forming method, resistance change memory, and resistance change memory manufacturing method |
CN101783170A (en) * | 2009-01-21 | 2010-07-21 | 中国科学院微电子研究所 | Circuit and method for driving resistance transition type memory to realize multi-value storage |
CN102543147A (en) * | 2012-01-18 | 2012-07-04 | 北京大学 | Reading circuit and reading method of multilevel storage circuit |
CN104681085A (en) * | 2015-03-03 | 2015-06-03 | 中国科学院微电子研究所 | Resistive random access memory based on flip coding circuit and corresponding data storage method |
Also Published As
Publication number | Publication date |
---|---|
CN110137348A (en) | 2019-08-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Sun et al. | XNOR-RRAM: A scalable and parallel resistive synaptic architecture for binary neural networks | |
CN110111827B (en) | Multi-value resistive random access memory based on multiple single-value resistive random access memories | |
CN108092658A (en) | A kind of logic circuit and its operating method | |
TW200426839A (en) | Nonvolatile memory cell and nonvolatile semiconductor memory device | |
CN111478703B (en) | Processing circuit and output current compensation method based on memristive cross array | |
CN110569962B (en) | A convolutional computing accelerator based on 1T1R memory array and its operation method | |
TW200304235A (en) | Multiple data state memory cell | |
CN102882513B (en) | Full adder circuit and chip | |
US11171650B2 (en) | Reversible logic circuit and operation method thereof | |
CN107194462A (en) | Three-valued neural networks cynapse array and utilize its neuromorphic calculating network | |
CN107424647B (en) | A Memristor-Based Speech Storage and Classification System | |
CN114400031A (en) | Complement mapping RRAM (resistive random access memory) storage and calculation integrated chip and electronic equipment | |
CN110827898B (en) | A memristor-based voltage-resistance reversible logic circuit and its operation method | |
CN112489709B (en) | Two-step writing operation method of resistive random access memory array | |
Yan et al. | Understanding the trade-offs of device, circuit and application in ReRAM-based neuromorphic computing systems | |
CN113949385A (en) | Analog-to-digital conversion circuit for RRAM storage and calculation integrated chip complement quantization | |
CN109814837B (en) | LFSR circuit based on resistive variable memory and its pseudo-random data sequence generation method | |
CN110137348B (en) | A Multiplexing Multi-valued Resistive Switching Structure and Its Neural Network Formed | |
CN109660250B (en) | Multi-state gate based on resistive random access memory | |
CN114254743A (en) | A circuit for parallel multiply-accumulate operation in binary neural network based on RRAM array | |
CN117219139A (en) | Implementation method of physical unclonable function based on self-activated resistive random access device | |
CN110572149A (en) | A kind of Toffoli gate circuit and its operation method | |
Alshaya et al. | Passive Selectorless Memristive Structure with One Capacitor-One Memristor | |
CN113222131B (en) | A Synaptic Array Circuit with Signed Weight Coefficient Based on 1T1R | |
CN102882509B (en) | Carry circuit and chip |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |