Precision tuning of pages or words of non-volatile memory cells used in analog neural memory arrays in artificial neural networks and associated high-voltage circuitry.
By using non-volatile memory arrays and high-voltage circuits for precision tuning, the problems of low hardware energy efficiency and high computational complexity in artificial neural networks are solved, achieving efficient simulation calculations and accurate synaptic weight adjustment.
Patent Information
- Application Number
- CN202080098758.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Priority Date
- 2020-09-17
- Filing Date
- 2020-09-22
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2040-09-22
AI Technical Summary
Existing artificial neural network hardware technologies suffer from low energy efficiency and high computational complexity. In particular, when implementing a large number of synaptic connections, CMOS-implemented synapses are too large to meet the needs of high-performance information processing.
Using a non-volatile memory array as a synapse, precision tuning is achieved by configuring the memory cell array. High-voltage circuits are used to realize the precision tuning and weight adjustment of the analog neural network, reducing the need for separate multiplication and addition logic circuits.
It achieves efficient simulation computation, reduces energy consumption, and enables precise tuning of synaptic weights in neural networks, improving computational parallelism and energy efficiency.
Smart Images

Figure CN115298738B_ABST
Abstract
Description
[0001] Priority Statement
[0002] This patent application claims priority to U.S. Provisional Patent Application No. 62 / 993,008, filed March 22, 2020, entitled "Precision Tuning of Pages or Words and Associated High-Voltage Circuitry for Non-Volatile Memory Cells in Analog Neural Memory Arrays in Artificial Neural Networks"; and U.S. Patent Application No. 17 / 024,410, filed September 17, 2020, entitled "Precision Tuning of Pages or Words and Associated High-Voltage Circuitry for Non-Volatile Memory Cells in Analog Neural Memory Arrays in Artificial Neural Networks". Technical Field
[0003] This patent application discloses several embodiments for precision tuning of pages or words of non-volatile memory cells in analog neural memories. This patent application also discloses high-voltage circuitry and programming sequences used in the precision tuning process. Background Technology
[0004] Artificial neural networks mimic biological neural networks (the central nervous system of animals, especially the brain) and are used to estimate or approximate functions that can depend on a large number of inputs and are often unknown. Artificial neural networks typically consist of interconnected layers of "neurons" that exchange messages with each other.
[0005] Figure 1 An artificial neural network is illustrated, where circles represent the inputs or layers of neurons. Connections (called synapses) are indicated by arrows and have numerical weights that can be adjusted empirically. This allows the artificial neural network to adapt to its inputs and learn. Typically, an artificial neural network consists of layers with multiple inputs. There are usually one or more intermediate layers of neurons, and an output layer of neurons that provide the output of the neural network. Neurons at each level make decisions individually or collectively based on the data received from the synapses.
[0006] One of the major challenges in developing artificial neural networks for high-performance information processing is the lack of sufficient hardware technology. In reality, practical artificial neural networks rely on a large number of synapses to achieve high connectivity between neurons, i.e., very high computational parallelism. In principle, such complexity can be achieved using digital supercomputers or dedicated clusters of graphics processing units. However, compared to biological networks, these methods are generally energy inefficient, in addition to being costly, as biological networks consume far less energy primarily due to their ability to perform low-precision analog computations. CMOS analog circuits have been used in artificial neural networks, but given the large number of neurons and synapses, most CMOS-implemented synapses are excessively large.
[0007] The applicant previously disclosed an artificial (simulated) neural network utilizing one or more non-volatile memory arrays as synapses in U.S. Patent Application No. 15 / 594,439 (published as U.S. Patent Publication 2017 / 0337466), which is incorporated herein by reference. The non-volatile memory array operates as a simulated neuromorphic memory. As used herein, the term "neuromorphic" refers to a circuit that implements a model of a nervous system. The simulated neuromorphic memory includes a first plurality of synapses configured to receive a first plurality of inputs and generate a first plurality of outputs therefrom, and a first plurality of neurons configured to receive the first plurality of outputs. The first plurality of synapses includes a plurality of memory cells, wherein each memory cell includes: spaced-apart source and drain regions formed in a semiconductor substrate, wherein a channel region extends between the source and drain regions; a floating gate disposed over and insulated from a first portion of the channel region; and a non-floating gate disposed over and insulated from a second portion of the channel region. Each memory cell is configured to store weight values corresponding to a plurality of electrons on the floating gate. Multiple memory cells are configured to multiply a first plurality of inputs by stored weight values to generate a first plurality of outputs. An array of memory cells arranged in this manner may be called a vector matrix multiplication (VMM) array.
[0008] Examples of different non-volatile memory cells that can be used in a VMM will now be discussed.
[0009] Non-volatile memory cells
[0010] Various types of known nonvolatile memory cells can be used in VMM arrays. For example, U.S. Patent 5,029,130 (“130 Patent”), which is incorporated herein by reference, discloses an array of split-gate nonvolatile memory cells, which is a type of flash memory cell. Such memory cells 210 in... Figure 2 As shown in the figure. Each memory cell 210 includes a source region 14 and a drain region 16 formed in a semiconductor substrate 12, with a channel region 18 therebetween. A floating gate 20 is formed over and insulated from (and controls the conductivity of) a first portion of the channel region 18, and is formed over a portion of the source region 14. A word line terminal 22 (which is typically coupled to a word line) has a first portion disposed over and insulated from (and controlling the conductivity of) a second portion of the channel region 18, and a second portion extending upward and located over the floating gate 20. The floating gate 20 and the word line terminal 22 are insulated from the substrate 12 by a gate oxide. A bit line terminal 24 is coupled to the drain region 16.
[0011] The memory cell 210 is erased by applying a high positive voltage to the word line terminal 22 (where electrons are removed from the floating gate), which causes electrons on the floating gate 20 to tunnel from the floating gate 20 to the word line terminal 22 through the intermediate insulator via the Fowler-Nordheim tunnel.
[0012] Memory cell 210 is programmed by applying a positive voltage to word line terminal 22 and a positive voltage to source region 14 (where electrons are placed on the floating gate). Electron current flows from drain region 16 to source region 14 (source line terminal). When electrons reach the gap between word line terminal 22 and floating gate 20, they are accelerated and become excited (heated). Due to electrostatic attraction from floating gate 20, some heated electrons are injected into floating gate 20 through gate oxide.
[0013] Memory cell 210 is read by applying a positive read voltage to the drain region 16 and word line terminal 22 (which connects the portion of channel region 18 below the word line terminal). If the floating gate 20 is positively charged (i.e., electrons are erased), the portion of channel region 18 below the floating gate 20 is also turned on, and current flows through channel region 18, which is sensed as an erased state or a "1" state. If the floating gate 20 is negatively charged (i.e., programmed electronically), the portion of channel region below the floating gate 20 is mostly or completely turned off, and current does not flow (or very little current) through channel region 18, which is sensed as a programmed state or a "0" state.
[0014] Table 1 shows the typical voltage range that can be applied to the terminals of memory cell 110 to perform read, erase, and program operations:
[0015] Table 1: Figure 2 Operation of flash memory cell 210
[0016] WL BL SL Read 1 0.5-3V 0.1-2V 0V Read 2 0.5-3V 0-2V 2-0.1V erase Approximately 11-13V 0V 0V programming 1-2V 1-3μA 9-10V
[0017] "Read 1" is the read mode where the cell current is output on the bit line. "Read 2" is the read mode where the cell current is output on the source line terminal.
[0018] Figure 3 The memory cell 310 is shown, which is related to Figure 2 The memory cell 210 is similar, but with the addition of a control gate (CG) terminal 28. The control gate terminal 28 is biased at a high voltage (e.g., 10V) during programming, at a low or negative voltage (e.g., 0V / -8V) during erasure, and at a low or medium voltage (e.g., 0V / 2.5V) during reading. Other terminals are similar. Figure 2 That kind of bias.
[0019] Figure 4A quad-gate memory cell 410 is shown, comprising a source region 14, a drain region 16, a floating gate 20 over a first portion of a channel region 18, a select gate 22 (typically coupled to a word line WL) over a second portion of the channel region 18, a control gate 28 over the floating gate 20, and an erase gate 30 over the source region 14. This configuration is described in U.S. Patent 6,747,310, which is incorporated herein by reference for all purposes. Here, all gates except the floating gate 20 are non-floating gates, meaning they are electrically connected to or capable of being electrically connected to a voltage source. Programming is performed by heated electrons from the channel region 18 that inject themselves into the floating gate 20. Erasing is performed by electrons tunneling from the floating gate 20 to the erase gate 30.
[0020] Table 2 shows the typical voltage range that can be applied to the terminals of memory cell 410 for performing read, erase, and program operations:
[0021] Table 2: Figure 4 Operation of flash memory cell 410
[0022] WL / SG BL CG EG SL Read 1 0.5-2V 0.1-2V 0-2.6V 0-2.6V 0V Read 2 0.5-2V 0-2V 0-2.6V 0-2.6V 2-0.1V erase -0.5V / 0V 0V 0V / -8V 8-12V 0V programming 1V 1μA 8-11V 4.5-9V 4.5-5V
[0023] "Read 1" is the read mode where the cell current is output on the bit line. "Read 2" is the read mode where the cell current is output on the source line terminal.
[0024] Figure 5 Memory cell 510 is shown, except that it does not have the erase gate EG terminal. Memory cell 510 is similar to... Figure 4 The memory cell 410 is similar. Erasure is performed by biasing the substrate 18 to a high voltage and the control gate CG terminal 28 to a low voltage or a negative voltage. Alternatively, erasure is performed by biasing the word line terminal 22 to a positive voltage and the control gate terminal 28 to a negative voltage. Programming and reading are similar. Figure 4 As it is.
[0025] Figure 6 A tri-gate memory cell 610 is shown, which is another type of flash memory cell. Memory cell 610 and... Figure 4 The memory cell 410 is identical to the memory cell 610, except that the memory cell 610 does not have a separate control gate terminal. Except that no control gate bias is applied, the erase operation (erasing via the erase gate terminal) and read operation are similar. Figure 4 The programming operation is performed without a control gate bias, and as a result, a higher voltage must be applied to the source line terminals during the programming operation to compensate for the lack of a control gate bias.
[0026] Table 3 shows the typical voltage ranges that can be applied to the terminals of memory cell 610 for performing read, erase, and program operations:
[0027] Table 3: Figure 6 Operation of flash memory cell 610
[0028] WL / SG BL EG SL Read 1 0.5-2.2V 0.1-2V 0-2.6V 0V Read 2 0.5-2.2V 0-2V 0-2.6V 2-0.1V erase -0.5V / 0V 0V 11.5V 0V programming 1V 2-3μA 4.5V 7-9V
[0029] "Read 1" is the read mode where the cell current is output on the bit line. "Read 2" is the read mode where the cell current is output on the source line terminal.
[0030] Figure 7 The stacked gate memory cell 710 is shown, which is another type of flash memory cell. Memory cell 710 and... Figure 2 The memory cell 210 is similar, except that the floating gate 20 extends over the entire channel region 18, and the control gate terminal 22 (which will be coupled to the word line here) extends over the floating gate 20, separated by an insulating layer (not shown). Programming is performed using hot electron injection from the channel 18 to the channel region near the drain region 16, and erasure is performed using Fowler-Nordheim electron tunneling from the floating gate 20 to the substrate 12. Read operations operate in a similar manner to those previously described for memory cell 210.
[0031] Table 4 shows the typical voltage ranges that can be applied to the terminals of memory cell 710 and substrate 12 to perform read, erase, and program operations:
[0032] Table 4: Figure 7 Operation of flash memory cell 710
[0033] CG BL SL substrate Read 1 0-5V 0.1–2V 0-2V 0V Read 2 0.5-2V 0-2V 2-0.1V 0V erase -8 to -10V / 0V FLT FLT 8-10V / 15-20V programming 8-12V 3-5V / 0V 0V / 3-5V 0V
[0034] "Read 1" is a read mode in which the cell current is output on the bit line. "Read 2" is a read mode in which the cell current is output at the source line terminal. Optionally, in an array comprising rows and columns of memory cells 210, 310, 410, 510, 610, or 710, the source line can be coupled to a row of memory cells or two adjacent rows of memory cells. That is, the source line terminal can be shared by memory cells in adjacent rows.
[0035] Figure 8A dual-split-gate memory cell 810 is shown. The memory cell 810 includes: a floating gate (FG) 20 disposed on and insulated from a substrate 12; a control gate 28 (CG) disposed on and insulated from the floating gate 20; an erase gate 30 (EG) disposed adjacent to and insulated from the floating gate 20 and the control gate 28, and disposed on and insulated from the substrate 12, wherein the erase gate 30 is T-shaped, such that the apex corner of the control gate CG faces the inner corner of the T-shaped erase gate to improve erasure efficiency; and a drain region 16 (DR) adjacent to the floating gate 20 in the substrate (having bit line contacts 24 (BL) connected to the drain diffusion region 16 (DR)). The memory cells are formed as a pair of memory cells (A on the left and B on the right) sharing a common erase gate 30. This cell design is consistent with the above reference. Figures 2-7 The memory cell under discussion differs at least in that it lacks a source region beneath the erase gate EG, a select gate (also known as a word line), and a channel region for each memory cell. Instead, a single, continuous channel region 18 extends beneath two memory cells (i.e., from the drain region 16A of one memory cell to the drain region 16B of another). To read or program a memory cell, the control gate 28 of the other memory cell is raised to a sufficient voltage to turn on the underlying channel region portion via a voltage-coupled floating gate 20 (e.g., to read or program cell A, the voltage on FGB 20B is raised via a voltage coupling from CGB 28B to turn on the channel region portion beneath FGB 20B). Erasure is performed using Fowler-Nordheim electrons tunneling from the floating gate 20 to the erase gate 30. Programming is performed using hot electron injection from channel 18 to the floating gate 20, denoted as Programming 1 in Table 5. Alternatively, programming is performed using Fowler-Nordheim electrons tunneling from the erase gate 30 to the floating gate 20, which is represented as Programming 2 in Table 5. Alternatively, programming is performed using Fowler-Nordheim electrons tunneling from the channel 18 to the floating gate 20, in which case the conditions are similar to Programming 2, except that the substrate 12 is biased at a low or negative voltage while the erase gate 30 is biased at a low positive voltage.
[0036] Table 5 shows the typical voltage ranges that can be applied to the terminals of memory cell 810 for performing read, erase, and program operations. In this table, it is assumed that cell A (with terminals EG, CGA, and BLA) is used for the read, program, or erase operation in each row.
[0037] Table 5: Figure 8 Operation of flash memory cell 810
[0038]
[0039] To utilize memory arrays comprising one of the aforementioned types of non-volatile memory cells in artificial neural networks, two modifications were made. First, the circuitry was configured such that each memory cell could be individually programmed, erased, and read without adversely affecting the memory state of other memory cells in the array, as explained further below. Second, continuous (simulated) programming of the memory cells was provided.
[0040] Specifically, the memory state (i.e., the charge on the floating gate) of each memory cell in the array can be continuously changed from a fully erased state to a fully programmed state independently with minimal interference to other memory cells. In another embodiment, the memory state (i.e., the charge on the floating gate) of each memory cell in the array can be continuously changed from a fully programmed state to a fully erased state and vice versa, independently with minimal interference to other memory cells. This means that the cell storage device is analog, or at least can store one discrete value from many discrete values (such as 16 or 64 different values), which allows for very precise and individual tuning of all cells in the memory array, and makes the memory array ideal for storing and fine-tuning the synaptic weights of neural networks.
[0041] The methods and apparatus described herein can be applied to other non-volatile memory technologies, such as, but not limited to, FINFET split-gate flash or stacked-gate flash memory, NAND flash memory, 3D flash memory, SONOS (silicon-oxide-nitride-oxide-silicon, with charge trapped in nitride), MONOS (metal-oxide-nitride-oxide-silicon, with metal charge trapped in nitride), ReRAM (resistive RAM), PCM (phase-change memory), MRAM (magnetic RAM), FeRAM (ferroelectric RAM), OTP (double-layer or multi-layer programmable at one time), and CeRAM (associative electron RAM). The methods and apparatus described herein can also be applied to volatile memory technologies for neural networks, such as, but not limited to, SRAM, DRAM, and / or volatile synaptic cells.
[0042] Neural networks using non-volatile memory cell arrays
[0043] Figure 9 This conceptually illustrates a non-limiting example of a neural network using a non-volatile memory array in this embodiment. This example uses a non-volatile memory array neural network for a facial recognition application, but any other suitable application can also be implemented using a neural network based on a non-volatile memory array.
[0044] In this example, S0 is the input layer, which is a 32x32 pixel RGB image with 5-bit precision (i.e., three 32x32 pixel arrays, one for each color R, G, and B, with 5-bit precision per pixel). The synapse CB1 from the input layer S0 to layer C1 applies different sets of weights in some cases and shared weights in others, and scans the input image with a 3x3 pixel overlapping filter (kernel), shifting the filter by one pixel (or more than one pixel depending on the model). Specifically, the values of nine pixels in a 3x3 portion of the image (i.e., called the filter or kernel) are provided to synapse CB1, where these nine input values are multiplied by appropriate weights, and after summing the output of this multiplication, a single output value is determined by the first synapse of CB1 to generate the pixels of one of the feature maps in layer C1. The 3x3 filter is then shifted one pixel to the right within the input layer S0 (i.e., adding a column of three pixels to the right and releasing a column of three pixels to the left), thereby providing the nine pixel values from this newly positioned filter to synapse CB1, where they are multiplied by the same weights and the second single output value is determined by the associated synapse. This process continues until the 3x3 filter scans all three colors and all bits (precision values) across the entire 32x32 pixel image of the input layer S0. This process is then repeated using different sets of weights to generate different feature maps for C1 until all feature maps for layer C1 are computed.
[0045] At layer C1, in this example, there are 16 feature maps, each with 30x30 pixels. Each pixel is a new feature pixel extracted from the product of the input and the kernel, so each feature map is a two-dimensional array. Therefore, in this example, layer C1 consists of a 16-layer two-dimensional array (remember that the layers and arrays referred to in this article are logical relationships, not necessarily physical relationships; that is, the array does not have to be oriented as a physical two-dimensional array). Each of the 16 feature maps in layer C1 is generated by one set of sixteen different groups of synaptic weights applied to the filter scan. The C1 feature maps can all relate to different aspects of the same image features, such as boundary recognition. For example, the first map (generated using the first weight recombination, shared for all scans used to generate the first map) can recognize circular edges, the second map (generated using the second weight recombination, different from the first weight recombination) can recognize rectangular edges, or the aspect ratio of certain features, and so on.
[0046] Before transitioning from layer C1 to layer S1, activation function P1 (pooling) is applied, which pools the values from consecutive non-overlapping 2x2 regions in each feature map. The purpose of the pooling function is to average the neighboring locations (or, alternatively, use a max function) to, for example, reduce the dependence on edge locations and reduce the data size before moving to the next stage. At layer S1, there are 16 15x15 feature maps (i.e., sixteen different arrays, each 15x15 pixels). The synapse CB2 from layer S1 to layer C2 scans the map in S1 using a 4x4 filter, where the filter is shifted by 1 pixel. At layer C2, there are 22 12x12 feature maps. Before transitioning from layer C2 to layer S2, activation function P2 (pooling) is applied, which pools the values from consecutive non-overlapping 2x2 regions in each feature map. At layer S2, there are 22 6x6 feature maps. An activation function (pooling) is applied to the synapse CB3 from layer S2 to layer C3, where each neuron in layer C3 is connected to each mapping in layer S2 via a corresponding synapse in CB3. There are 64 neurons in layer C3. The synapse CB4 from layer C3 to the output layer S3 completely connects C3 to S3, meaning each neuron in layer C3 is connected to every neuron in layer S3. The output at S3 comprises 10 neurons, with the highest-output neuron determining the class. For example, this output could indicate the recognition or classification of the content of the original image.
[0047] Synapses for each layer are implemented using an array or a portion of an array of non-volatile memory cells.
[0048] Figure 10 This is a block diagram of a system that could be used for this purpose. The VMM system 32 includes non-volatile memory cells and serves as synapses between layers (such as...). Figure 6 (CB1, CB2, CB3, and CB4 in the original text). Specifically, the VMM system 32 includes a VMM array 33 (comprising non-volatile memory cells arranged in rows and columns), an erase gate and word line gate decoder 34, a control gate decoder 35, a bit line decoder 36, and a source line decoder 37, which decode the corresponding inputs to the non-volatile memory cell array 33. The inputs to the VMM array 33 may come from the erase gate and word line gate decoder 34 or from the control gate decoder 35. In this example, the source line decoder 37 also decodes the outputs of the VMM array 33. Alternatively, the bit line decoder 36 may decode the outputs of the VMM array 33.
[0049] The VMM array 33 serves two purposes. First, it stores weights that will be used by the VMM system 32. Second, the VMM array 33 efficiently multiplies the inputs with the weights stored in the VMM array 33 and adds them together at each output line (source line or bit line) to produce an output that will serve as the input to the next layer or the final layer. By performing multiplication and addition functions, the VMM array 33 eliminates the need for separate multiplication and addition logic circuits and is also highly efficient due to its in-situ memory computation. The weights of the filter (core) are mapped across one or more bit lines (columns) or across one or more rows of the single or multiple VMM arrays 33.
[0050] The output of VMM array 33 is provided to a differential summer (such as a summing operational amplifier or a summing current mirror) 38, which sums the output of VMM array 33 to create a single value for the convolution. Differential summer 38 is arranged to perform the summation of both the positive and negative weight inputs to output a single value.
[0051] The output values of the difference summer 38 are then summed and provided to the activation function circuit 39, which corrects the output. The activation function circuit 39 can provide a sigmoid, tanh, ReLU function, or any other nonlinear function. The corrected output value of the activation function circuit 39 becomes the next layer's (e.g., Figure 9 The elements of the feature map of layer C1 are then applied to the next synapse to produce the next feature map layer or the final layer. Thus, in this example, the VMM array 33 constitutes multiple synapses (which receive their input from existing neuron layers or from input layers such as an image database), and the summer 38 and activation function circuit 39 constitute multiple neurons.
[0052] Figure 10 The inputs to the VMM system 32 (WLx, EGx, CGx, and optional BLx and SLx) can be analog levels (e.g., current, voltage, or charge), binary levels, digital timing pulses (in which case a pulse-to-analog converter PAC may be required to convert the pulses to the appropriate input analog level), or digital bits (in which case a DAC is provided to convert the digital bits to the appropriate input analog level); the outputs can be analog levels, binary levels, digital pulses, or digital bits (in which case an output ADC is provided to convert the output analog level to digital bits).
[0053] Figure 11 A block diagram illustrating the use of a multi-layer VMM system 32 (here labeled VMM systems 32a, 32b, 32c, 32d, and 32e) is provided. Figure 11As shown, the input (denoted as Inputx) is converted from digital to analog by a digital-to-analog converter 31 and provided to the input VMM system 32a. The converted analog input can be voltage or current. The first-level input D / A conversion can be accomplished by using a function or LUT (lookup table) of appropriate analog levels of the matrix multiplier that maps Inputx to the input VMM system 32a. Input conversion can also be accomplished by an analog-to-analog (A / A) converter to convert an external analog input to a mapped analog input to the input VMM system 32a. Input conversion can also be accomplished by a digital-to-digital pulse (D / P) converter to convert an external digital input to one or more digital pulses mapped to the input VMM system 32a.
[0054] The output generated by the input VMM system 32a is provided as input to the next VMM system (hidden level 1) 32b, which in turn generates the output provided as input to the next VMM system (hidden level 2) 32c, and so on. The layers of the VMM system 32 serve as different layers of synapses and neurons in a convolutional neural network (CNN). Each VMM system 32a, 32b, 32c, 32d, and 32e can be an independent physical system comprising a corresponding non-volatile memory array, or multiple VMM systems can utilize different portions of the same physical non-volatile memory array, or multiple VMM systems can utilize overlapping portions of the same physical non-volatile memory array. Each VMM system 32a, 32b, 32c, 32d, and 32e can also be time-division multiplexed for different portions of its array or neurons. Figure 11 The example shown contains five layers (32a, 32b, 32c, 32d, 32e): one input layer (32a), two hidden layers (32b, 32c), and two fully connected layers (32d, 32e). Those skilled in the art will recognize that this is merely exemplary, and conversely, a system may include more than two hidden layers and more than two fully connected layers.
[0055] VMM array
[0056] Figure 12 A neuronal VMM array 1200 is shown, which is particularly suitable for Figure 3 The memory cell 310 shown serves as a synapse and component for neurons between the input layer and the next layer. The VMM array 1200 includes a memory array 1201 of non-volatile memory cells and a reference array 1202 of non-volatile reference memory cells (at the top of the array). Alternatively, another reference array may be placed at the bottom.
[0057] In VMM array 1200, control gate lines (such as control gate line 1203) extend vertically (therefore, reference array 1202 is orthogonal to control gate line 1203 in the row direction), and erase gate lines (such as erase gate line 1204) extend horizontally. Here, the inputs of VMM array 1200 are set on control gate lines (CG0, CG1, CG2, CG3), and the outputs of VMM array 1200 appear on source lines (SL0, SL1). In one embodiment, only even rows are used, and in another embodiment, only odd rows are used. The currents placed on each source line (SL0, SL1, respectively) perform a summation function of all currents from the memory cells connected to that particular source line.
[0058] As described herein with respect to neural networks, the non-volatile memory cells of the VMM array 1200 are preferably configured to operate in the subthreshold region.
[0059] In weak inversion, the non-volatile reference memory cell and non-volatile memory cell described in this paper are biased:
[0060]
[0061] Where Ids is the drain-to-source current; Vg is the gate voltage on the memory cell; Vth is the (effective or equivalent) threshold voltage of the memory cell; Vt is the thermal voltage = k*T / q, where k is the Boltzmann constant, T is the temperature in Kelvin, and q is the electron charge; n is the slope factor = 1 + (Cdep / Cox), where Cdep = the capacitance of the depletion layer, and Cox is the capacitance of the gate oxide layer; Io is the memory cell current at the gate voltage equal to the threshold voltage, and Io is related to (Wt / L)*u*Cox*(n-1)*Vt 2 Proportional, where u is the carrier mobility, and Wt and L are the width and length of the memory cell, respectively.
[0062] For I-to-V logarithmic converters that use memory cells (such as reference memory cells or peripheral memory cells) or transistors to convert input current Ids to input voltage Vg:
[0063] Vg = n * Vt * log[Ids / wp * Io]
[0064] Here, wp refers to the w in the reference memory cell or the peripheral memory cell.
[0065] For a memory array used as a VMM array for vector matrix multipliers, the output current is:
[0066] Iout = wa * Io * e (Vg) / nVt ,Right now
[0067] Iout = (wa / wp) * Iin = W * Iin
[0068] W = wa / wp = e (Vthp-Vtha) / nVt
[0069] Iin = wp * Io * e (Vg) / nVt
[0070] Here, wa = W for each memory cell in the memory array. Vthp is the (equivalent or effective) threshold voltage of the reference memory cell (or peripheral memory cell or transistor), and Vtha is the (equivalent or effective) threshold voltage of the array memory cell.
[0071] Word lines or control gates can be used as inputs to memory cells that accept input voltages.
[0072] Alternatively, the non-volatile memory cells of the VMM array described herein can be configured to operate in a linear region:
[0073] Ids=β*(Vgs-Vth)*Vds; β=u*Cox*Wt / L,
[0074] Wα(Vgs-Vth),
[0075] This means that the weight W in the linear region is proportional to (Vgs-Vth).
[0076] Word lines, control gates, bit lines, or source lines can be used as inputs to memory cells operating in a linear region. Bit lines or source lines can be used as outputs to memory cells.
[0077] For an I-to-V linear converter, a memory cell (such as a reference memory cell or a peripheral memory cell) or a transistor or resistor operating in the linear region can be used to linearly convert the input / output current into the input / output voltage.
[0078] Alternatively, the memory cells of the VMM array described herein can be configured to operate in a saturation region:
[0079] Ids = 1 / 2 * β * (Vgs - Vth) 2 β=u*Cox*Wt / L
[0080] Wα(Vgs-Vth) 2 This means that the weight W is related to (Vgs-Vth). 2 proportional
[0081] Word lines, control gates, or erase gates can be used as inputs to memory cells operating in saturation regions. Bit lines or source lines can be used as outputs of output neurons.
[0082] Alternatively, the memory cells of the VMM array described herein can be used in all regions or combinations thereof (subthreshold, linear, or saturated regions) of each or more layers of a neural network.
[0083] Figure 13 A neuronal VMM array 1300 is shown, which is particularly suitable for Figure 2 The memory cell 210 shown serves as a synapse between the input layer and the next layer. The VMM array 1300 includes a memory array 1303 of non-volatile memory cells, a reference array 1301 of first non-volatile reference memory cells, and a reference array 1302 of second non-volatile reference memory cells. The reference arrays 1301 and 1302, arranged along the column direction of the array, are used to convert current inputs flowing into terminals BLR0, BLR1, BLR2, and BLR3 into voltage inputs WL0, WL1, WL2, and WL3. In practice, the first and second non-volatile reference memory cells are diode-connected via a multiplexer 1314 (partially shown) through which current inputs flow. The reference cells are tuned (e.g., programmed) to a target reference level. The target reference level is provided by a reference microarray matrix (not shown) or from a band-based reference circuit.
[0084] Memory array 1303 serves two purposes. First, it stores the weights used by VMM array 1300 on their respective memory cells. Second, memory array 1303 efficiently multiplies the inputs (i.e., the current inputs provided in terminals BLR0, BLR1, BLR2, and BLR3, which reference arrays 1301 and 1302 convert into input voltages to provide to word lines WL0, WL1, WL2, and WL3) by the weights stored in memory array 1303, and then sums all the results (memory cell currents) to produce an output on the corresponding bit lines (BL0-BLN), which will be the input to the next layer or the final layer. By performing multiplication and addition functions, memory array 1303 eliminates the need for separate multiplication and addition logic circuits and is also highly efficient. Here, voltage inputs are provided on word lines (WL0, WL1, WL2, and WL3), and the output appears on the corresponding bit lines (BL0-BLN) during read (inference) operations. The current placed on each bit line in the bit lines BL0-BLN performs a summation function of the currents from all non-volatile memory cells connected to that particular bit line.
[0085] Table 6 shows the operating voltages used for the VMM array 1300. The columns in the table indicate the voltage applied to the word line for the selected cell, the word line for the unselected cell, the bit line for the selected cell, the bit line for the unselected cell, the source line for the selected cell, and the source line for the unselected cell, where FLT indicates floating, i.e., no voltage applied. The rows indicate read, erase, and program operations.
[0086] Table 6: Figure 13 Operation of VMM array 1300
[0087] WL WL - Not selected BL BL - Not Selected SL SL - Not selected Read 0.5-3.5V -0.5V / 0V 0.1-2V (Ineuron) 0.6V-2V / FLT 0V 0V erase Approximately 5-13V 0V 0V 0V 0V 0V programming 1-2V -0.5V / 0V 0.1-3uA Vinh approximately 2.5V 4-10V 0-1V / FLT
[0088] Figure 14 A neuronal VMM array 1400 is shown, which is particularly suitable for Figure 2 The memory cell 210 shown serves as a synapse and component for neurons between the input layer and the next layer. The VMM array 1400 includes a memory array 1403 of non-volatile memory cells, a reference array 1401 of first non-volatile reference memory cells, and a reference array 1402 of second non-volatile reference memory cells. Reference arrays 1401 and 1402 extend in the row direction of the VMM array 1400. The VMM array is similar to VMM 1300, except that in VMM array 1400, word lines extend in the vertical direction. Here, inputs are set on word lines (WLA0, WLB0, WLA1, WLB2, WLA2, WLB2, WLA3, WLB3), and outputs appear on source lines (SL0, SL1) during read operations. The current placed on each source line performs a summation function of all currents from the memory cells connected to that particular source line.
[0089] Table 7 shows the operating voltages used for the VMM array 1400. The columns in the table indicate the voltage applied to the word line for the selected cell, the word line for the unselected cell, the bit line for the selected cell, the bit line for the unselected cell, the source line for the selected cell, and the source line for the unselected cell. The rows indicate read, erase, and program operations.
[0090] Table 7: Figure 14 Operation of VMM array 1400
[0091]
[0092] Figure 15 A neuronal VMM array 1500 is shown, which is particularly suitable for... Figure 3The memory cell 310 shown serves as a synapse and component for neurons between the input layer and the next layer. The VMM array 1500 includes a memory array 1503 of non-volatile memory cells, a reference array 1501 of first non-volatile reference memory cells, and a reference array 1502 of second non-volatile reference memory cells. Reference arrays 1501 and 1502 are used to convert current inputs flowing into terminals BLR0, BLR1, BLR2, and BLR3 into voltage inputs CG0, CG1, CG2, and CG3. In practice, the first and second non-volatile reference memory cells are diode-connected via a multiplexer 1512 (partially shown), through which current inputs flow via BLR0, BLR1, BLR2, and BLR3. Each multiplexer 1512 includes a corresponding multiplexer 1505 and a common-source cascode transistor 1504 to ensure a constant voltage on the bit lines (such as BLRO) of each of the first and second non-volatile reference memory cells during a read operation. The reference cells are tuned to a target reference level.
[0093] Memory array 1503 serves two purposes. First, it stores weights that will be used by VMM array 1500. Second, memory array 1503 efficiently multiplies the inputs (current inputs provided to terminals BLR0, BLR1, BLR2, and BLR3, which reference arrays 1501 and 1502 convert into input voltages to be provided to control gates CG0, CG1, CG2, and CG3) by the weights stored in the memory array, and then sums all the results (cell currents) to produce an output that appears on BL0-BLN and will be the input to the next layer or the final layer. By performing multiplication and addition functions, the memory array eliminates the need for separate multiplication and addition logic circuits and is also highly efficient. Here, the inputs are provided on control gate lines (CG0, CG1, CG2, and CG3), and the outputs appear on bit lines (BL0–BLN) during read operations. The currents placed on each bit line perform a summation function of all currents from the memory cells connected to that particular bit line.
[0094] The VMM array 1500 implements unidirectional tuning for the non-volatile memory cells in the memory array 1503. That is, each non-volatile memory cell is erased and then partially programmed until the desired charge is reached on the floating gate. This can be performed, for example, using the precise programming techniques described below. If too much charge is placed on the floating gate (causing an incorrect value to be stored in the cell), the cell must be erased, and the sequence of partial programming operations must restart. As shown, two rows sharing the same erase gate (such as EG0 or EG1) need to be erased together (this is called page erasure), and thereafter, each cell is partially programmed until the desired charge is reached on the floating gate.
[0095] Table 8 shows the operating voltages used for the VMM array 1500. The columns in the table indicate the voltages applied to the word lines for selected cells, word lines for unselected cells, bit lines for selected cells, bit lines for unselected cells, control gates for selected cells, control gates for unselected cells in the same sector as the selected cell, control gates for unselected cells in different sectors from the selected cell, erase gates for selected cells, erase gates for unselected cells, source lines for selected cells, and source lines for unselected cells. Rows indicate read, erase, and program operations.
[0096] Table 8: Figure 15 Operation of VMM array 1500
[0097]
[0098] Figure 16 This illustrates a neuronal VMM array 1600, which is particularly suitable for... Figure 3The memory cell 310 shown serves as a synapse and component for neurons between the input layer and the next layer. The VMM array 1600 includes a memory array 1603 of non-volatile memory cells, a reference array 1601 of first non-volatile reference memory cells, and a reference array 1602 of second non-volatile reference memory cells. EG lines EGR0, EG0, EG1, and EGR1 extend vertically, while CG lines CG0, CG1, CG2, and CG3 and SL lines WL0, WL1, WL2, and WL3 extend horizontally. Similar to VMM array 1600, but unlike VMM array 1600, VMM array 1600 implements bidirectional tuning, where each individual cell can be fully erased, partially programmed, and partially erased as needed to achieve the desired amount of charge on the floating gate due to the use of individual vertical EG lines. As shown in the figure, reference arrays 1601 and 1602 convert the input currents in terminals BLR0, BLR1, BLR2, and BLR3 into control gate voltages CG0, CG1, CG2, and CG3 to be applied to memory cells in the row direction (through the operation of reference cells connected via diodes of multiplexer 1614). Current outputs (neurons) are in bit lines BL0-BLN, where each bit line sums all currents from non-volatile memory cells connected to that particular bit line.
[0099] Table 9 shows the operating voltages used for the VMM array 1600. The columns in the table indicate the voltage applied to the word lines for the selected cell, the word lines for the unselected cell, the bit lines for the selected cell, the bit lines for the unselected cell, the control gate for the selected cell, the control gate for the unselected cell in the same sector as the selected cell, the control gate for the unselected cell in a different sector from the selected cell, the erase gate for the selected cell, the erase gate for the unselected cell, the source line for the selected cell, and the source line for the unselected cell. The rows indicate read, erase, and program operations.
[0100] Table 9: Figure 16 Operation of VMM array 1600
[0101]
[0102] The inputs of a VMM array can be analog levels, binary levels, timing pulses, or digital bits, and the outputs can be analog levels, binary levels, timing pulses, or digital bits (in which case an output ADC is needed to convert the output analog level current or voltage into digital bits).
[0103] For each memory cell in the VMM array, each weight W can be implemented by a single memory cell, a differential cell, or a hybrid memory cell (the average of two or more cells). In the case of differential cells, two memory cells are needed to implement the weight w as a differential weight (w = w+ – w-). In the case of two hybrid memory cells, two memory cells are needed to implement the weight w as the average of the two cells.
[0104] Non-volatile memory cells must be programmed and erased accurately and precisely to ensure that the correct amount of charge is placed on the floating gate to store the correct weight in the cell.
[0105] What is required is a VMM system that allows for precise tuning on a word or page basis, where a word typically comprises 8-64 memory cells, each storing multiple logic bits (i.e., multilevel cells, each storing, for example, 3-8 bits); and a page comprises 64 memory cells, each storing multiple logic bits. Further required is high-voltage circuitry to generate the required voltage. Further required is an improved programming sequence to minimize the occurrence of undesirable effects such as erase gate interference and control gate interference. Summary of the Invention
[0106] This patent application discloses several embodiments for precision tuning of pages or words of non-volatile memory cells in analog neural memory. This patent application also discloses a high-voltage circuit for forming a high voltage applied to the terminals of the non-volatile memory cells during the precision tuning process. This patent application further discloses a programming sequence for applying voltage to the terminals to minimize the occurrence rate of interference during precision tuning.
[0107] In one embodiment, the analog neural memory system includes an array of non-volatile memory cells arranged in rows and columns, each non-volatile memory cell including word line terminals, bit line terminals, and erase gate terminals; a plurality of word lines, each word line coupled to a word line terminal of a row of the non-volatile memory cell; a plurality of bit lines, each bit line coupled to a bit line terminal of a column of the non-volatile memory cell; and a plurality of erase gate enable transistors, each erase gate enable transistor coupled to an erase gate terminal of a word of the non-volatile memory cell.
[0108] In another embodiment, the analog neural memory system includes an array of non-volatile memory cells arranged in rows and columns, each non-volatile memory cell including word line terminals, bit line terminals, and erase gate terminals; multiple word lines, each word line coupled to the word line terminals of a row of non-volatile memory cells; multiple bit lines, each bit line coupled to the bit line terminals of a column of non-volatile memory cells; and multiple erase gate enable transistors, each erase gate enable transistor coupled to the erase gate terminal of a page of non-volatile memory cells, the page comprising two words of non-volatile memory cells in adjacent rows.
[0109] In another embodiment, the analog neural memory system includes an array of non-volatile memory cells arranged in rows and columns, each non-volatile memory cell including word line terminals, bit line terminals, and erase gate terminals; multiple word lines, each word line coupled to a word line terminal of a row of non-volatile memory cells; multiple bit lines, each bit line coupled to a bit line terminal of a column of non-volatile memory cells; and multiple erase gate lines, each erase gate line coupled to an erase gate terminal of a page of non-volatile memory cells, the page comprising two words of non-volatile memory cells in adjacent rows.
[0110] In another embodiment, a method for tuning a word or page of a non-volatile memory cell includes erasing the word or page of the non-volatile memory cell; performing soft programming of the word or page of the non-volatile memory cell; performing coarse programming of the word or page of the non-volatile memory cell; performing fine programming of the word or page of the non-volatile memory cell; and identifying any fast bits in the word or page of the non-volatile memory cell such that those bits can be programmed using a sequence different from the normal (neither fast nor slow) bits.
[0111] In another embodiment, a method for programming words of non-volatile memory cells in an array of non-volatile memory cells arranged in rows and columns, each non-volatile memory cell including a control gate terminal, a source line terminal, and an erase gate terminal, the method comprising: raising a voltage ramp on the control gate terminal of the word of the non-volatile memory cell during a first time period; raising a voltage ramp on the source line terminal of the word of the non-volatile memory cell during a second time period following the first time period; and raising a voltage ramp on the erase gate terminal of the word of the non-volatile memory cell during a third time period following the second time period.
[0112] In another embodiment, a method for programming words of non-volatile memory cells in an array of non-volatile memory cells arranged in rows and columns, each non-volatile memory cell including a control gate terminal, a source line terminal, and an erase gate terminal, the method comprising: during a first time period, raising a voltage ramp on the control gate terminal of the word of the non-volatile memory cell to an intermediate voltage; during a second time period following the first time period, raising a voltage ramp on the source line terminal of the word of the non-volatile memory cell; during a third time period following the second time period, raising a voltage ramp on the erase gate terminal of the word of the non-volatile memory cell; and during a fourth time period following the third time period, further raising a voltage ramp on the control gate terminal of the word of the non-volatile memory cell.
[0113] In another embodiment, a method for programming words of non-volatile memory cells in an array of non-volatile memory cells arranged in rows and columns, each non-volatile memory cell including a control gate terminal, a source line terminal, and an erase gate terminal, the method comprising: during a first time period, raising a voltage ramp on the control gate terminal of the word of the non-volatile memory cell to a first intermediate voltage; during a second time period following the first time period, raising a voltage ramp on the source line terminal of the word of the non-volatile memory cell to a second intermediate voltage; during a third time period following the second time period, raising a voltage ramp on the erase gate terminal of the word of the non-volatile memory cell, and further raising a voltage ramp on the source line terminal of the word of the non-volatile memory cell, and further raising a voltage ramp on the control gate terminal of the word of the non-volatile memory cell to a third intermediate voltage; and during a fourth time period following the third time period, further raising a voltage ramp on the control gate terminal of the word of the non-volatile memory cell.
[0114] In another embodiment, a method for programming words of non-volatile memory cells in an array of non-volatile memory cells arranged in rows and columns, each non-volatile memory cell including a control gate terminal, a source line terminal, and an erase gate terminal, the method comprising: during a first time period, raising a voltage ramp on the source line terminal of the word of the non-volatile memory cell; and during a second time period after the first time period, raising a voltage ramp on the erase gate terminal of the word of the non-volatile memory cell and raising a voltage ramp on the control gate terminal of the word of the non-volatile memory cell.
[0115] In another embodiment, the analog neural memory system includes an array of non-volatile memory cells arranged in rows and columns, each non-volatile memory cell including word line terminals, bit line terminals, and erase gate terminals; multiple word lines, each word line coupled to the word line terminals of a row of non-volatile memory cells; multiple bit lines, each bit line coupled to the bit line terminals of a column of non-volatile memory cells; and a circuit decoder block including an erase gate decoder for providing erase gate decoding signals, a control gate decoder for providing two control gate signals, and a source line decoder for providing source line decoding signals.
[0116] In another embodiment, the analog neural memory system includes an array of non-volatile memory cells arranged in rows and columns, each non-volatile memory cell including word line terminals, bit line terminals, and erase gate terminals; multiple word lines, each word line coupled to the word line terminals of a row of non-volatile memory cells; multiple bit lines, each bit line coupled to the bit line terminals of a column of non-volatile memory cells; and a circuit decoder block including an erase gate decoder for providing erase gate decoding signals, a control gate decoder for providing four control gate signals, and a source line decoder for providing source line decoding signals.
[0117] In another embodiment, the analog neural memory system includes an array of non-volatile memory cells arranged in rows and columns, each non-volatile memory cell including word line terminals, bit line terminals, and erase gate terminals; multiple word lines, each word line coupled to the word line terminals of a row of non-volatile memory cells; multiple bit lines, each bit line coupled to the bit line terminals of a column of non-volatile memory cells; and a circuit decoder block including an erase gate decoder for providing two erase gate decoding signals, a control gate decoder for providing eight control gate signals, and a source line decoder for providing source line decoding signals. Attached Figure Description
[0118] Figure 1 This illustrates existing artificial neural networks.
[0119] Figure 2 This illustrates a split-gate flash memory cell from the prior art.
[0120] Figure 3 This illustrates another prior art split-gate flash memory cell.
[0121] Figure 4 This illustrates another prior art split-gate flash memory cell.
[0122] Figure 5 This illustrates another prior art split-gate flash memory cell.
[0123] Figure 6 This illustrates another prior art split-gate flash memory cell.
[0124] Figure 7 This illustrates a stacked gate flash memory cell from the prior art.
[0125] Figure 8 This illustrates a prior art dual-split-gate memory cell.
[0126] Figure 9 Different layers of an exemplary artificial neural network using one or more VMM arrays are shown.
[0127] Figure 10 A VMM system including a VMM array and other circuitry is shown.
[0128] Figure 11 An exemplary artificial neural network using one or more VMM systems is shown.
[0129] Figure 12 An implementation scheme for a VMM array is shown.
[0130] Figure 13 Another implementation of the VMM array is shown.
[0131] Figure 14 Another implementation of the VMM array is shown.
[0132] Figure 15 Another implementation of the VMM array is shown.
[0133] Figure 16 Another implementation of the VMM array is shown.
[0134] Figure 17 An implementation scheme of a VMM array with erasure capability on a page-based basis is shown.
[0135] Figure 18 An implementation scheme of a VMM array with erasing capability on a word-based basis is shown.
[0136] Figure 19 Another implementation of a VMM array with erasing capability on a word-based basis is shown.
[0137] Figures 20A to 20B This demonstrates a bidirectional tuning algorithm based on pages or words.
[0138] Figure 21A This demonstrates the fast bit algorithm.
[0139] Figure 21B The slow bit algorithm is shown.
[0140] Figure 22 The VMM system is shown.
[0141] Figure 23 The high-voltage generation block is shown.
[0142] Figure 24 The circuit with the highest voltage is shown.
[0143] Figure 25 Another high-voltage generation block is shown.
[0144] Figure 26 Another high-voltage generation block is shown.
[0145] Figures 27A to 27B An implementation scheme for a sloping descent circuit is shown.
[0146] Figure 28 The circuit shown is a ramp circuit.
[0147] Figures 29A to 29C An implementation scheme for a high-voltage decoding circuit is shown.
[0148] Figures 30A to 30D An implementation scheme for a high-voltage decoding circuit is shown.
[0149] Figure 31 The programming sequence is shown.
[0150] Figure 32 This shows another programming sequence.
[0151] Figure 33 This shows another programming sequence.
[0152] Figure 34 This shows another programming sequence.
[0153] Figure 35 The erasure sequence is shown.
[0154] Figures 36A to 36C An implementation scheme for a high-voltage decoding circuit is shown.
[0155] Figure 37 A high-voltage latching circuit is shown. Detailed Implementation
[0156] The artificial neural network of this invention utilizes a combination of CMOS technology and non-volatile memory arrays.
[0157] Improved VMM system with page- or word-based tuning
[0158] Figure 17A VMM array 1700 is shown. The VMM array 1700 enables bidirectional tuning of pages of non-volatile memory cells. Here, an exemplary page 1701 comprises two words, each in a different row. A word comprises multiple (e.g., 8-64) memory cells. A particular word can comprise only one or a few cells. Pairs of adjacent rows share a source line, such as SL0 or SL1. All cells in page 1701 share a common erase gate line controlled by an erase gate enable transistor 1702, which controls the supply of voltage to the erase gate terminal EGW of all cells in the exemplary page set 1701. Here, all cells in page 1701 can be erased simultaneously. The cells in page 1701 can then be bidirectionally tuned via programming and erase operations, and some cells in page 1701 can be unidirectionally tuned via programming operations. Programming operations can include the precise programming techniques described below with respect to Figures 20 and 21. If too much electronic charge is placed on the floating gate (which will cause incorrect current values to be stored in the cell, i.e., the current value is lower than the expected current value), the cell must be erased and part of the sequence of programming operations must be restarted.
[0159] Figure 18 A VMM array 1800 is shown. The VMM array 1800 enables bidirectional tuning of words of non-volatile memory cells. Here, an exemplary word 1801 comprises multiple cells in a row. All cells in word 1801 share a common erase gate line controlled by an erase gate enable transistor 1802, which controls the supply of voltage to the erase gate terminals of all cells in word 1801. Here, all cells in word 1801 can be erased simultaneously. Subsequently, bidirectional tuning of the cells in word 1801 is possible through programming and erasing operations. Programming operations can include precise programming techniques described below. When too much electronic charge is placed on the floating gate (causing an incorrect current value to be stored in the cell, i.e., a current value lower than the expected current value), the cell must be erased, and part of the sequence of programming operations must be restarted.
[0160] Figure 19A VMM array 1900 is shown. The VMM array 1900 enables bidirectional tuning of words of non-volatile memory cells. Here, an exemplary word 1901 comprises two half-words of cells. Each half-word belongs to a row sharing an erase gate. All cells in word 1901 share a common erase gate line connected to the erase gate terminal EGW. Unlike VMM arrays 1800 and 1700, there is no erase gate enable transistor. Here, all cells in word 1901 can be erased simultaneously. Subsequently, bidirectional tuning of the cells in word 1901 is possible through programming and erasing operations. Programming operations can include the precise programming techniques described below. When too much electron charge is placed on the floating gate (causing an incorrect current value to be stored in the cell, i.e., a current value lower than the expected current value), the cell must be erased, and part of the programming sequence must be restarted.
[0161] although Figure 17 , Figure 18 and Figure 19 Not shown, but as described by reference in U.S. Provisional Patent Application No. 62 / 981,757, filed February 26, 2020, entitled “Ultra-precision tuning of analog neural memory cells in deep learning artificial neural networks,” which is incorporated herein by reference, pull-down bit lines, tuning bit lines (for ultra-fine programming), virtual bit lines, and redundant bit lines can be used.
[0162] Figure 20A This demonstrates the tuning page / word algorithm 2000, which can be applied to... Figures 17 to 19 The VMM arrays are 1700, 1800, or 1900.
[0163] First, erase the word or page (step 2001). Second, perform deep programming on unused cells (step 2002). Deep programming is used to program the cells to a state (or target level) that has insignificant cell current during a read operation, for example, in the <pA range. Third, perform coarse programming on the cells within the word or page (step 2003). Coarse programming is used to program the cells to a coarse target level, for example, within 50%-500% of the target with a large (coarse) increment voltage and / or program current and / or program timing. Fourth, perform fine programming on the cells within the word or page (step 2004). Fine programming is used to program the cells to a fine target level, for example, within + / -5%-30% of the target with a small (fine) increment voltage and / or program current and / or program timing. Fifth, optionally perform ultra-fine programming on the cells within the word or page (step 2005). Ultra-fine programming is used to program the cells to the final target level with precise, very small increment voltage and / or program current and / or program timing. The percentage achieved within the final target level in coarse / fine / ultra-fine programming is traded off against the magnitude of the increment level and / or program timing to minimize noise, such as noise from program quantization noise (increment magnitude), disturb noise, various coupling noises, FG-FG coupling noise, etc.
[0164] Figure 20B shows a tuned page / word algorithm 2050 that can be applied to Figures 17 to 19 the VMM array 1700, VMM array 1800, or VMM array 1900 in. The tuned page / word algorithm 2050 is similar to Figure 20A the tuned / page word algorithm 2000 in, with the difference that the tuned page / word algorithm 2050 further includes steps for handling fast bits or slow bits. Steps 2001 to 2005 appear as shown in Figure 20A . In step 2006, or alternatively in step 2003, it is determined whether the word or page contains any fast bits or slow bits (step 2006). A fast bit is a bit that requires a shorter programming period than a normal bit to reach a desired level, and a slow bit is a bit that requires a longer programming period than a normal bit to reach the desired level. Fast bits can be detected by monitoring the programming rate (programming speed) of the cells, such as by measuring ΔIr / Δtprog (the change in current over time, for example, the change in current over K consecutive pulses) > a predetermined R factor, for example, K=2 pulses. Slow bits can be detected by monitoring the programming rate ΔIr / Δtprog < a predetermined R factor. If no, the algorithm stops. If yes, the fast bit or slow bit cells are identified and marked. Subsequently, any programming of the word or page will utilize the fast bit or slow bit algorithm, as discussed below with reference to Figure 21A and Figure 21B .
[0165] The applicant previously disclosed various techniques for performing coarse programming, fine programming, and ultrafine programming in U.S. Provisional Patent Application No. 62 / 981,757, filed on February 26, 2020, entitled “Ultra-precise tuning of analog neural memory cells in deep learning artificial neural networks,” which is incorporated herein by reference.
[0166] Figure 21A The fast bit algorithm 2100 is illustrated. This is performed when it has been determined that a word or page contains one or more fast bit units, also known as fast bits. First, the word or page is erased or partially erased (step 2101). Second, the fast bits are programmed and verified (step 2102). This is performed, for example, under conditions less than the default (or constant) voltage increment and / or less than the default program current and / or less than the default timing. Next, the normal bits (i.e., all bits not marked as fast bits or slow bits) are programmed and verified (step 2103). This is performed, for example, under the default settings of coarse / fine / ultra-fine voltage increments and / or program current and / or timing.
[0167] Figure 21B The slow bit algorithm 2150 is shown. This is performed when it has been determined that a word or page contains one or more slow bit cells. First, the word or page is erased or partially erased (step 2151). Second, the slow bit cells are programmed and verified (step 2152). This is performed, for example, under conditions greater than the default voltage increment and / or greater than the default program current and / or greater than the default timing. Next, the normal bits (i.e., all bits not marked as slow or fast bits) are programmed and verified (step 2153). This is performed, for example, under default settings of coarse / fine / ultra-fine voltage increments and / or program current and / or timing.
[0168] In one implementation, the slow bit cell is first tuned to avoid interference with other cells in the same page / word, followed by the fast bit cell, and then the normal bit cell.
[0169] In one implementation, the slow bit cell is tuned first (to avoid interfering with other cells in the same page / word), followed by the normal bit cell, and then the fast bit cell.
[0170] Figure 22A block diagram of a VMM system 2200 is shown. The VMM system 2200 includes a VMM array 2201, a row decoder 2202, a high-voltage decoder 2203, a column decoder 2204, a bitline driver 2205, input circuitry 2206, output circuitry 2207, control logic unit 2208, and a bias generator 2209. The VMM system 2200 further includes a high-voltage generation block 2210, which includes a charge pump 2211, a charge pump regulator 2212, and a high-voltage level generator 2213. The VMM system 2200 further includes a (programming / erasing, or weighted tuning) algorithm controller 2214, analog circuitry 2215, control logic unit 2216, and test control logic unit 2217. The systems and methods described below can be implemented in the VMM system 2200.
[0171] Input circuitry 2206 may include circuitry such as a DAC (digital-to-analog converter), DPC (digital-to-pulse converter), AAC (analog-to-analog converter, such as a current-to-voltage converter), PAC (pulse-to-analog level converter), or any other type of converter. Input circuitry 2206 can implement normalization, linear, or nonlinear up / down scaling functions or arithmetic functions. Input circuitry 2206 can implement a temperature compensation function for the input. Input circuitry 2206 can implement activation functions such as ReLU or sigmoid functions. Output circuitry 2207 may include circuitry such as an ADC (analog-to-digital converter, used to convert the analog output of a neuron to digital bits), AAC (analog-to-analog converter, such as a current-to-voltage converter), APC (analog-to-pulse converter), or any other type of converter. Output circuitry 2207 can implement activation functions such as ReLU or sigmoid functions. The output circuit 2207 can implement statistical normalization, regularization, up / down scaling functions, statistical rounding, or arithmetic functions (e.g., addition, subtraction, division, multiplication, shifting, logarithms) on the neuron output. The output circuit 2207 can also implement temperature compensation functions on the neuron output or array output (such as bitline output) to keep the array's power consumption approximately constant or to improve the accuracy of the array (neuron) output, for example, by keeping the IV slope approximately the same.
[0172] Figure 23 The high voltage generation block 2300 is shown, which is relative to Figure 22An embodiment of the high-voltage generation block 2210 is described. The high-voltage generation block 2300 includes a charge pump 2211 and a charge pump regulator 2212, which generate various high voltages and are controlled by an enable signal labeled EN_CP. The charge pump 2211 and charge pump regulator 2212 provide the necessary high voltages to the control gate high-voltage generator 2301, the erase gate high-voltage generator 2302, and the source line high-voltage generator 2303, which are controlled by enable signals labeled EN_CGGEN, EN_EGGEN, and EN_SLGEN, respectively, and provide high-voltage signals to the control gate line, erase gate line, and source line as needed during programming, erasing, or reading operations within the VMM array.
[0173] Figure 24 A high-voltage maximum circuit for supplying a high-voltage power supply is shown, which identifies and outputs the highest voltage between a first high voltage HV1 and a second high voltage HV2. Comparator 2401 receives HV1div and HV2div, which are level-shifted, down-divided (i.e., low-voltage) versions of HV1 and HV2, respectively. When HV1 is greater than HV2, comparator 2401 outputs a high signal on line COMPO, and when HV1 is less than HV2, it outputs a low signal on line COMPO. The output of comparator 2401 on line COMPO is provided to high-voltage level shifters 2402 and 2403. When the output of comparator 2401 is low, high-voltage level shifter 2402 outputs a low signal on line 2402B, and high-voltage level shifter 2403 outputs a low signal on line 2403A. When the output of comparator 2401 is high, high-voltage level shifter 2402 outputs a low signal on line 2402A, and high-voltage level shifter 2403 outputs a low signal on line 2403B. PMOS transistors 2404, 2405, 2406, and 2407 are configured as shown, with their gates coupled to a corresponding line among lines 2402A, 2402B, 2403A, and 2403B. When high-voltage level shifter 2402 outputs a low signal on line 2402B, and high-voltage level shifter 2403 outputs a low signal on line 2403A, the high-voltage output 2408 will be equivalent to HV2, less than any voltage drop across PMOS transistors 2406 and 2407. When the high-voltage level shifter 2402 outputs a low signal on line 2402A and the high-voltage level shifter 2403 outputs a low signal on line 2403B, the high-voltage output 2408 will be equivalent to HV1, which is less than any voltage drop across PMOS transistors 2404 and 2405.
[0174] Figure 25A high-voltage generation block 2500 is shown, which is another embodiment of the high-voltage generation block 2412. Here, the high-voltage generation block 2500 includes a charge pump and a regulator 2501, enabled by the signal EN_CP, a high-voltage increment reference generator 2503, and a high-voltage buffer operational amplifier 2502. By partially trimming the output voltage HVSUP fed from the output of the charge pump 2501 to the input of the high-voltage operational amplifier HVOPA, the voltage output of the charge pump regulator 2503 can be controlled based on the signal sent to the gate of the MOS transistor of the high-voltage increment reference generator 2503.
[0175] Figure 26 A high-voltage generator block 2600 is shown, which is another embodiment of the high-voltage generation block 2412. The high-voltage generation block 2600 receives an input VIN (low-voltage signal) and generates an output HV (high-voltage signal), and includes an operational amplifier 2603, a variable resistor 2601, and a variable resistor 2602, wherein the gain of the operational amplifier 2601 depends on the values of the variable resistors 2601 and / or 2602. Therefore, the high-voltage increment value is controlled by the values of the variable resistors 2601 and / or 2602.
[0176] Figure 27A A ramp-down control circuit 2700, configured as shown, includes a clamping PMOS transistor 2701, an enabling NMOS transistor 2702, and a current-biased NMOS transistor 2703. The ramp-down circuit 2700 receives a voltage HV to ramp down at the source of the PMOS transistor 2701 and generates an output signal VHV_DET at the drain of the PMOS transistor 2701. This output signal VHV_DET has a peak value of HV and ramps down towards ground in response to a signal ENRMP, which is fed from low to high to the gate of the enabling NMOS transistor 2702. When ENRMP is high, the current controlled by the current-biased NMOS transistor 2703 ramps down the output signal VHV_DET from the HV value to the value Vcas + Vt PH.
[0177] Figure 27B Another ramp-down control circuit 2750 is shown, which includes a cascode NMOS transistor 2751, a shunt NMOS transistor 2753 (providing ramp current), an enable NMOS transistor 2572, a current source 2754, and a capacitor 2755. The ramp-down rate at the HV node is controlled by I / C (= I or the capacitance of the reference current source 2754 / capacitor 2755).
[0178] Figure 28A ramp-up circuit 2800 configured as shown is illustrated, comprising an NMOS cascode transistor 2801, an enable NMOS transistor 2802, a current shunt NMOS transistor 2803, a current source 2805, and a capacitor 2804. The ramp-up circuit 2800 shunts current through the NMOS transistor 2803 and controls the ramp-up rate of the HV node by the I / C ratio (I = Ireference current source 2805 / capacitance of capacitor 2804). The ramp rate at the HV node ensures that the current injected through capacitor 2804 is equal to the current injected through current source 2805.
[0179] Figure 29A , Figure 29B and Figure 29C A VMM high-voltage decoding circuit is shown, which includes components adapted to... Figure 2 The word line decoder circuit 2901, the source line decoder circuit 2904, and the high-voltage level shifter 2908 are used together with the memory cells of the type shown.
[0180] Figure 29A The word line decoder circuit 2901 includes a PMOS select transistor 2902 (controlled by signal HVO_B) and an NMOS deselect transistor 2903 (controlled by signal HVO_B) configured as shown. HVSUP is a high voltage supply, such as from a charge pump and regulator. When HVO_B is enabled, WLSUP provides the voltage supply for the word line WL.
[0181] Figure 29B The source-to-pole decoder circuit 2904 includes an NMOS monitoring transistor 2905 (controlled by signal HVO), a drive transistor 2906 (controlled by signal HVO), and a deselect transistor 2907 (controlled by signal HVO_B), configured as shown. When signal HVO is high, the voltage appearing on line SLSUP is transferred to line SL and appears on the monitoring line SL_MON. When signal HVO_B is high, line SL is pulled down.
[0182] Figure 29C The high-voltage level shifter 2908 receives an enable signal EN and outputs a high-voltage signal HVO and its compensation HVO_B between HVSUP (e.g., 12V) and HVSUP_LOW (e.g., 0V when HVSUP is, for example, equal to an intermediate level of about 5V, or 2.5V when HVSUP is, for example, 12V). For example, HVO can be 5V and HVO_B can be 0V; or HVO can be 12V and HVO_B can be 2.5V.
[0183] Figures 30A to 30D A VMM high-voltage decoding circuit is shown, which includes components adapted to work with... Figure 3 The erase gate decoder circuit 3001, control gate decoder circuit 3004, source line decoder circuit 3007, and high voltage level shifter 3011 are used together with memory cells of the type shown.
[0184] Figure 30A and Figure 30B The erase gate decoder circuit 3001 and the control gate decoder circuit 3004 in the middle are used with Figure 29A The word line decoder circuit 2901 in the middle has the same design.
[0185] Figure 30C The source line decoder circuit 3007 in Figure 29 uses the same design as the source line decoder circuit 2904 in Figure 29.
[0186] Figure 30D The high-voltage level shifter 3011 in the figure uses the same design as the high-voltage level shifter 2908 in Figure 29.
[0187] Figures 31 to 34 The programming sequence illustrates the voltages applied to the control gate terminal, source line terminal, and erase gate terminal of one or more non-volatile memory cells during programming operations.
[0188] Figure 31 Programming sequence 3100 is shown, in which the control gate voltage CG ramps up during a first time period, the source line voltage SL ramps up during a second time period, and the erase gate voltage EG ramps up during a third time period. During a fourth time period, all three voltages plateau at their peaks, and then the ramp sequence reverses. During a fifth time period, the erase gate voltage EG ramps down, during a sixth time period, the source line voltage SL ramps down, and during a seventh time period, the control gate voltage CG ramps down. Programming sequence 3100 minimizes the occurrence rate of erase gate interference.
[0189] Figure 32 Programming sequence 3200 is shown, in which the control gate voltage CG ramp rises to its intermediate value during a first time period, the source line voltage SL ramp rises to its peak value during a second time period, and the erase gate voltage EG ramp rises to its peak value during a third time period, and the control gate voltage CG ramp rises to its peak value during a fourth time period. During a fifth time period, all three voltages plateau at their peak values, and the control gate voltage CG ramp falls to its intermediate value during a sixth time period, the erase gate voltage EG ramp falls during a seventh time period, the source line voltage SL ramp falls during an eighth time period, and the control gate voltage CG ramp falls to ground during a ninth time period.
[0190] Figure 33The programming sequence 3300 is shown, wherein the gate voltage CG is first controlled to rise to a first intermediate value during a first time period, and then the source line voltage SL is controlled to rise to a second intermediate value during a second time period. Subsequently, the gate voltage CG is controlled to rise to a third intermediate value during a third time period, while the source line voltage SL rises to its peak voltage, and the erase gate voltage EG rises to its peak voltage. Finally, the gate voltage CG is controlled to rise to its peak voltage during a fourth time period. Then, all three voltage plateaus occur at their peaks during a fifth time period. Subsequently, the gate voltage CG is controlled to fall to a third intermediate value during a sixth time period, then the source line voltage SL falls to a second intermediate value during a seventh time period, then the gate voltage CG falls to a first intermediate value during an eighth time period, then the erase gate voltage EG falls to ground during a ninth time period, then the source line voltage SL falls to ground during a tenth time period, and finally, the gate voltage CG falls to ground during an eleventh time period.
[0191] Figure 34 Programming sequence 3400 is shown, in which the source line voltage SL rises to its peak during a first time period, and then the control gate voltage CG rises to its peak during a second time period, while the erase gate voltage EG rises to its peak. Subsequently, all three voltage plateaus occur at their peaks during a third time period. Then, the control gate voltage CG decreases during a fourth time period, while the erase gate voltage EG decreases, followed by a decrease in the source line voltage SL during a fifth time period. Programming sequence 3400 minimizes the occurrence of control gate interference.
[0192] Figure 35 An erase sequence 3500 is shown, in which the suppressor control gate or suppressor source line (CG-inh or SL-inh, applied to unselected cells during operation) slopes up during a first time period, followed by an erase gate voltage EG slope up during a second time period. Then, during a third time period, all voltages plateau at their peaks. Subsequently, during a fourth time period, the erase gate voltage EG slopes down, followed by a fifth time period where the suppressor control gate or suppressor source line (CG-inh or SL-inh) slopes down. This sequence is applicable, for example, to arrays suitable for bidirectional tuning, such as... Figure 12 , Figure 14 , Figure 16 , Figure 19 .
[0193] Figure 36A , Figure 36B and Figure 36CA high-voltage decoder 3600 utilizing decoding sub-circuit blocks 3001 (EG dec), 3004 (CGdec), and 3007 (SL dec) in Figure 30 is shown. Different arrangements of the sub-circuit blocks are shown to optimize different configurations and optimizations.
[0194] Figure 36A A circuit decoder block 3601 including a circuit decoder block 3602 is shown. The circuit decoder block 3602 includes an EG dec that provides one EG decoding signal, a CG dec that provides two CG decoding signals, and an SL dec that provides one SL decoding signal.
[0195] Figure 36B The circuit decoder block 3611, including the circuit decoder block 3612, is shown. The circuit decoder block 3612 includes an EG dec that provides one EG decoding signal, a CG dec that provides four CG decoding signals, and an SL dec that provides one SL decoding signal.
[0196] Figure 36C The circuit decoder block 3621, which includes the circuit decoder block 3622, is shown. The circuit decoder block 3622t includes an EG dec that provides two EG decoding signals, a CG dec that provides eight CG decoding signals, and an SL dec that provides one SL decoding signal.
[0197] Figure 37 A high-voltage latch-up circuit 3700 is shown. The high-voltage latch-up circuit 3700 includes a cross-coupled high-voltage transistor inverter formed by PMOS transistors 3711 and 3712, NMOS transistors 3713 and 3714, and enable transistors 3715 and 3716. When VHVSUP 3701 is at an intermediate voltage (e.g., approximately 1.8V-5V), enable signals ENB 3706 and EN 3705 are logic signals (e.g., 0V / Vdd); and when VHVSUP 3701 is at a voltage greater than the intermediate HV voltage (e.g., >5V), and when VHVSUP_LOW 3702 is at an intermediate level (such as 1.8V), enable signals ENB 3706 and EN 3705 are at another intermediate voltage (e.g., both signals EN 3705 = ENB 3706 = Vdd = 1.8V). Generate output HVOUT 3704 and its compensation signal HVOUTB3703.
[0198] It should be noted that, as used herein, the terms “above” and “on” both encompass “directly on” (without intermediate material, elements, or space between) and “indirectly on” (with intermediate material, elements, or space between). Similarly, the term “adjacent” includes “directly adjacent” (without intermediate material, elements, or space between) and “indirectly adjacent” (with intermediate material, elements, or space between), “mounted to” includes “directly mounted to” (without intermediate material, elements, or space between) and “indirectly mounted to” (with intermediate material, elements, or space between), and “electrically coupled to” includes “directly electrically coupled to” (without intermediate material or elements electrically connecting the elements together) and “indirectly electrically coupled to” (with intermediate material or elements electrically connecting the elements together). For example, forming an element “above the substrate” can include forming an element directly on the substrate without intermediate material / elements between them, and forming an element indirectly on the substrate with one or more intermediate materials / elements between them.
Claims
1. A method for programming words of non-volatile memory cells in an array of non-volatile memory cells arranged in rows and columns, each non-volatile memory cell including a control gate terminal, a source line terminal, and an erase gate terminal, the method comprising: During the first time period, the voltage ramp on the control gate terminal of the word in the non-volatile memory cell is increased. During a second time period following the first time period, the voltage ramp on the source line terminal of the word of the non-volatile memory cell is increased. as well as During a third time period following the second time period, the voltage ramp on the erase gate terminal of the word in the non-volatile memory cell is increased.
2. The method according to claim 1, further comprising: During a fourth time period following the third time period, the voltage ramp on the erase gate terminal of the word of the non-volatile memory cell is reduced; During a fifth time period following the fourth time period, the voltage ramp on the source line terminal of the word of the non-volatile memory cell is reduced; as well as During a sixth time period following the fifth time period, the voltage ramp on the control gate terminal of the word of the non-volatile memory cell is reduced.
3. The method of claim 1, wherein the array of non-volatile memory cells forms an analog neural memory.
4. A method for programming words of non-volatile memory cells in an array of non-volatile memory cells arranged in rows and columns, each non-volatile memory cell including a control gate terminal, a source line terminal, and an erase gate terminal, the method comprising: During the first time period, the control gate voltage ramp on the control gate terminal of the word of the non-volatile memory cell is raised to an intermediate voltage; During a second time period following the first time period, the source line voltage on the source line terminal of the word of the non-volatile memory cell is ramped up. During a third time period following the second time period, the erase gate voltage on the erase gate terminal of the word of the non-volatile memory cell is ramped up. as well as During a fourth time period following the third time period, the control gate voltage ramp on the control gate terminal of the word of the non-volatile memory cell is increased.
5. The method of claim 4, further comprising: During a fourth time period following the third time period, the control gate voltage ramp on the control gate terminal of the word of the non-volatile memory cell is reduced to an intermediate voltage. During a fifth time period following the fourth time period, the erase gate voltage on the erase gate terminal of the word of the non-volatile memory cell is ramped down; During a sixth time period following the fifth time period, the source line voltage on the source line terminal of the word of the non-volatile memory cell is ramped down. as well as During a seventh time period following the sixth time period, the control gate voltage ramp on the control gate terminal of the word of the non-volatile memory cell is reduced.
6. The method of claim 4, wherein the array of non-volatile memory cells forms an analog neural memory.
7. A method for programming words of non-volatile memory cells in an array of non-volatile memory cells arranged in rows and columns, each non-volatile memory cell including a control gate terminal, a source line terminal, and an erase gate terminal, the method comprising: During the first time period, the voltage ramp on the control gate terminal of the word of the non-volatile memory cell is raised to a first intermediate voltage; During a second time period following the first time period, the voltage ramp on the source line terminal of the word of the non-volatile memory cell is raised to a second intermediate voltage. During a third time period following the second time period, the voltage ramp on the erase gate terminal of the word of the non-volatile memory cell is increased, and the voltage ramp on the source line terminal of the word of the non-volatile memory cell is further increased, and the voltage ramp on the control gate terminal of the word of the non-volatile memory cell is further increased to a third intermediate voltage. as well as During a fourth time period following the third time period, the voltage ramp on the control gate terminal of the word of the non-volatile memory cell is increased.
8. The method of claim 7, further comprising: During a fifth time period following the fourth time period, the voltage ramp on the control gate terminal of the word of the non-volatile memory cell is reduced to the third intermediate voltage; During a sixth time period following the fifth time period, the voltage ramp on the source line terminal of the word of the non-volatile memory cell is reduced to the second intermediate voltage; During a seventh time period following the sixth time period, the voltage ramp on the control gate terminal of the word of the non-volatile memory cell is reduced to the first intermediate voltage; During the eighth time period following the seventh time period, the voltage ramp on the erase gate terminal of the word of the non-volatile memory cell is reduced; During the eighth time period following the seventh time period, the voltage ramp on the source line terminal of the word of the non-volatile memory cell is reduced; as well as During a ninth time period following the eighth time period, the voltage ramp on the control gate terminal of the word of the non-volatile memory cell is reduced.
9. The method of claim 7, wherein the array of non-volatile memory cells forms an analog neural memory.
10. A method for programming words of non-volatile memory cells in an array of non-volatile memory cells arranged in rows and columns, each non-volatile memory cell including a control gate terminal, a source line terminal, and an erase gate terminal, the method comprising: During the first time period, the voltage ramp on the source line terminal of the word of the non-volatile memory cell is increased. as well as During a second time period following the first time period, the voltage ramp on the erase gate terminal of the word of the non-volatile memory cell is increased, and the voltage ramp on the control gate terminal of the word of the non-volatile memory cell is increased.
11. The method of claim 10, further comprising: During a third time period following the second time period, the voltage ramp on the erase gate terminal of the word of the non-volatile memory cell is lowered and the voltage ramp on the control gate terminal of the word of the non-volatile memory cell is lowered. as well as During a fourth time period following the third time period, the voltage ramp on the source line terminal of the word of the non-volatile memory cell is reduced.
12. The method of claim 10, wherein the array of non-volatile memory cells forms an analog neural memory.
13. A method for programming words of non-volatile memory cells in an array of non-volatile memory cells arranged in rows and columns, each non-volatile memory cell including a control gate terminal, a source line terminal, and an erase gate terminal, the method comprising: The voltage ramp on the control gate terminal of the word of the non-volatile memory cell is increased during a time period with multiple ramp steps. as well as The voltage ramp on the source line terminal of the word of the non-volatile memory cell is increased.
14. The method of claim 13, further comprising: The voltage ramp on the erase gate terminal of the word in the non-volatile memory cell is increased.
15. The method of claim 13, wherein the array of non-volatile memory cells forms an analog neural memory.
Citation Information
Patent Citations
Deep learning neural network classifier using non-volatile memory array
US11308383B2
Deep Learning Neural Network Classifier Using Non-volatile Memory Array
US20170337466A1
Single transistor non-valatile electrically alterable semiconductor memory device
US5029130A
Flash memory cells with separated self-aligned select and erase gates, and process of fabrication
US6747310B2
Neural network matrix multiplication in memory cells
US20200020393A1
