(19)
(11)EP 3 514 795 B1

(12)EUROPEAN PATENT SPECIFICATION

(45)Mention of the grant of the patent:
13.05.2020 Bulletin 2020/20

(21)Application number: 18155279.5

(22)Date of filing:  06.02.2018
(51)Int. Cl.: 
G11C 11/54  (2006.01)
H01L 27/11524  (2017.01)
H03K 19/177  (2020.01)
G11C 16/04  (2006.01)
H01L 27/1157  (2017.01)

(54)

SUM-OF-PRODUCTS ACCELERATOR ARRAY

SUMMENPRODUKT-BESCHLEUNIGER-ARRAY

RÉSEAU D'ACCÉLÉRATEURS DE SOMME DE PRODUITS


(84)Designated Contracting States:
AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR

(30)Priority: 17.01.2018 US 201815873369

(43)Date of publication of application:
24.07.2019 Bulletin 2019/30

(73)Proprietor: MACRONIX INTERNATIONAL CO., LTD.
Hsinchu 300 (TW)

(72)Inventors:
  • LEE, Feng-Min
    300 Hsinchu (TW)
  • LIN, Yu-Yu
    300 Hsinchu (TW)

(74)Representative: Gunzelmann, Rainer 
Wuesthoff & Wuesthoff Patentanwälte PartG mbB Schweigerstraße 2
81541 München
81541 München (DE)


(56)References cited: : 
WO-A1-2012/015450
US-A1- 2016 336 064
US-A1- 2012 044 742
  
      
    Note: Within nine months from the publication of the mention of the grant of the European patent, any person may give notice to the European Patent Office of opposition to the European patent granted. Notice of opposition shall be filed in a written reasoned statement. It shall not be deemed to have been filed until the opposition fee has been paid. (Art. 99(1) European Patent Convention).


    Description

    BACKGROUND


    Field



    [0001] The present invention relates to circuitry that can be used to perform or support sum-of-products operations.

    Description of Related Art



    [0002] In neuromorphic computing systems, machine learning systems and circuitry used for some types of computation based on linear algebra, the sum-of-products function can be an important component. The function can be expressed as follows:



    [0003] In this expression, each product term is a product of a variable input Xi and a weight Wi. The weight Wi can be vary among the terms, corresponding for example to coefficients of the variable inputs Xi.

    [0004] The sum-of-products function can be realized as a circuit operation using cross-point array architectures in which the electrical characteristics of cells of the array effectuate the function.

    [0005] US 2016/336064 A1 discloses a neuromorphic computational circuitry including a network of variable resistive elements. Therein, an electronic element with a variable resistance, such as a resistive random access memory (RRAM), is used as the variable resistive element. The variable resistive element can thus provide different conductance states by changing the resistances of the resistive random access memory.

    [0006] For high-speed implementations, it is desirable to have a very large array so that many operations can be executed in parallel, or very large sum-of-products series can be performed. In the systems, there can be a very large number of inputs and outputs, so that the total current consumption can be very large.

    [0007] It is desirable to provide structures for sum-of-products functions suitable for implementation in large arrays, and that can be more energy-efficient.

    SUMMARY



    [0008] Some or all of the above problems are overcome by a device according to claim 1, and a method according to claim 12. Further aspects of the invention are defined in the dependent claims. A device is described that comprises an array of variable resistance cells, where a variable resistance cell in the array comprises a programmable threshold transistor and a resistor connected in parallel. The device can be operated so that an input voltage applied to the transistor and the programmable threshold of the transistor can represent variables of sum-of-products operations. The variable resistance of each variable resistance cell, in embodiments described herein, is a function of a voltage applied to the control gate of the programmable threshold transistor in the cell, a threshold of the programmable threshold transistor, and the resistor.

    [0009] The device in some embodiments includes a voltage sensing sense amplifier, configured to sense the voltage generated by the variable resistance cells, as a function of an applied current and the resistance of the variable resistance cells. In this manner, the magnitude of the current generated to produce the sum-of-products result can be limited or fixed, reducing the power consumption.

    [0010] The array can be implemented using cells consisting of one transistor and one resistor (1T-1R). Furthermore, embodiments described herein can implement the resistor as a buried implant resistor within the layout footprint of a single variable threshold transistor, in effect making an array of one transistor (IT) cells, for a very compact layout configured for sum-of-products operations with voltage sensing.

    [0011] Embodiments are described in which variable resistance cells in the array are configured in a plurality of strings of series-connected variable resistance cells. A plurality of word lines can be coupled to the strings of series-connected variable resistance cells. Word line drivers are connected to the plurality of word lines to apply variable gate voltages to the programmable threshold transistors in the variable resistance cells.

    [0012] Embodiments are described in which the programmable threshold transistors in the variable resistance cells comprise charge trapping memory transistors, such as floating gate transistors or dielectric charge trapping transistors.

    [0013] Embodiments are described in which the resistors in the variable resistance cells comprise a buried implant resistor connecting the current-carrying terminals (e.g. source and drain) of the programmable threshold transistor.

    [0014] A device for generating sum-of-products data is provided that includes an array of variable resistance cells, variable resistance cells in the array each comprising a programmable threshold transistor and a resistor connected in parallel, the array including n columns of cells including strings of series connected cells and m rows of cells, wherein n and m are non-zero integers. Control and bias circuitry are coupled to the array, including logic for programming the programmable threshold transistors in the array with thresholds corresponding to values of a weight factor Wmn for the corresponding cell. Input drivers are coupled to corresponding ones of the m rows of cells, the input drivers selectively applying inputs Xm to rows m. Column drivers are configured to apply currents In to corresponding ones of the n columns of cells. Voltage sensing circuits operatively coupled to the columns of cells.

    [0015] A system including a memory array and a sum-of-products accelerator array interconnected using a data path controller is described. The sum-of-products accelerator array comprises an array of programmable resistance cells. The memory array can be used in coordination with the sum-of-products accelerator array for configuration and operation of the sum-of-products functions.

    [0016] A method for operating an array of variable resistance cells to produce sum of products data comprises programming the programmable threshold transistors in the array with thresholds corresponding to values of a weight factor for the corresponding cell; selectively applying inputs to rows of cells in the array applying currents to corresponding ones of the columns of cells in the array; and sensing voltages on one or more of the columns of cells in the array.

    [0017] Other aspects and advantages of the present invention can be seen on review of the drawings, the detailed description and the claims, which follow.

    BRIEF DESCRIPTION OF THE DRAWINGS



    [0018] 

    Figure 1 is a functional diagram of a sum-of-products operation which can be a fundamental element of a neuromorphic computing system known in the prior art.

    Figure 2 illustrates a portion of an array of variable resistance cells configured for sum-of-products operations.

    Figure 3 is a schematic diagram of a variable resistance cell according to embodiments described herein.

    Figure 4 is a simplified cross-section of a variable resistance cell including a floating gate memory transistor and a buried implant resistor.

    Figure 5 is a simplified cross-section of a variable resistance cell including a dielectric charge trapping memory transistor and a buried implant resistor.

    Figures 6-9 illustrate stages in a manufacturing process for a variable resistance cell according to an embodiment described herein.

    Figures 10A and 10B are cross-section and layout views of variable resistance cells arranged in series in a NAND-like structure.

    Figures 11A and 11B are illustrated to demonstrate operation of a variable resistance cell.

    Figure 12 illustrates a string of variable resistance cells configured for implementation of a sum-of-products operation.

    Figure 13 is a simplified block diagram of sensing circuitry that can be utilized with an array of variable resistance cells for the purposes of sum-of-products operations.

    Figure 13A is a heuristic diagram referred to for the purposes of describing the sensing operations using the circuit of Figure 13.

    Figure 14 is a simplified diagram of a reference voltage circuit that can be utilized with sensing circuitry like that of Figure 13.

    Figure 15 illustrates a configuration of an array of variable resistance cells, including a reference string.

    Figure 16 illustrates another configuration of an array of variable resistance cells, including a reference string and unused cells.

    Figure 17 illustrates another configuration of an array of variable resistance cells including two reference strings and unused cells.

    Figures 18-22 illustrates functional sets of variable resistance cells configured to implement terms of sum-of-products operations with multiple bit weights.

    Figure 23 is a simplified block diagram of a device including an array of variable resistance cells applied, for example, for neuromorphic memory.

    Figures 24-26 illustrates a system including a sum-of-products accelerator array, and various operations thereof.


    DETAILED DESCRIPTION



    [0019] A detailed description of embodiments of the present invention is provided with reference to the Figures 1-26.

    [0020] Figure 1 is a diagram of a sum-of-products operation, where the terms of the summation are the product of input Xi times a weight Wi, in this example, where i goes from 1 to 7. The weight Wi can differ over the terms of the summation. In operation, the weights can be assigned as a set of coefficients, and then the inputs applied to compute a summation that changes as the inputs change. Also, in algorithms executing a learning procedure, the weights can be changed from time to time as the learning procedures change coefficients to learn from of the summation that achieves a useable result.

    [0021] In the illustrated example, the output of the summation is applied to a sigmoid function to produce an output that ranges in a non-linear fashion between a minimum and a maximum such as between 0 and 1. This is a common model of a synapse for example used in neuromorphic computing. Other activation functions can be used as well, such as a logit function. The sum-of-products operation can be applied as well in configurations not neuromorphic or not otherwise considered to model neurological systems.

    [0022] Figure 2 is a schematic diagram of an array of variable resistance cells, where each of the cells in the array comprises a programmable threshold transistor (e.g. 12) and a resistor (e.g. 14) connected in parallel. In this illustration, the array includes four strings of variable resistance cells, where each string comprises four variable resistance cells in series between a summing node SUM1 to SUM4, and a reference line, such as ground (e.g. 26). Four word lines WL1 to WL4 are coupled to the control terminals of the variable resistance cells in each of the strings. As indicated in the figure, there can be any number of columns and summing nodes up to SUMn, and any number of word lines up to WLm. The variable resistance cells at column n and row m have weights Wnm set as a function of the programmable threshold Vt of the cell, the resistance Rnm of the resistor in the cell, and the current In in the column.

    [0023] A voltage applied to the word lines corresponds to the variable inputs X1 to X4, ...Xm In this manner, the variable resistance of each of the variable resistance cells in the strings is a function of a voltage applied on the word line to the control gate of the cell, a threshold of the programmable threshold transistor in the cell, the current in the cell, and the resistor.

    [0024] The summing nodes (SUM1 to SUM4, ... SUMn) are coupled to a voltage sensing sense amplifier to generate a signal representing the sum-of-products output of each of the strings. A current source 21-24 is coupled to each of the strings to apply a constant current, in a representative example, to each string during the sensing operation.

    [0025] Figure 3 is a schematic diagram of one variable resistance cell, such as that used in the array of Figure 2. The variable resistance cell includes a first current-carrying node 30, a second current-carrying node 31 and a control terminal 32. A programmable threshold transistor 35 and a resistor 36 are connected in parallel to the first and second current-carrying nodes. The programmable threshold transistor has a gate connected to the control terminal 32.

    [0026] A voltage VG on the control terminal 32 can be characterized as a gate voltage for the programmable threshold transistor 35. The control terminal 32 can correspond to a word line in the array shown in Figure 2. A voltage VS on the first current-carrying node 30 can be characterized as a source voltage for the cell. A voltage VD on the second current-carrying node 31 can be characterized as a drain voltage for the cell.

    [0027] In this example, a cell current IC is applied to the second current-carrying node 31, having a current amplitude that is set in design, or adjustable, to establish a voltage drop in the cells, depending on the voltage range of the voltage sense amplifier, and the resistance values in the cells for the resistor 36. The current amplitude can be tuned according to a particular embodiment of the array, so that a usable range of voltages can be generated on the string for supply to the summing node. Also, the magnitude of resistance of the resistor and the configuration of the programmable threshold transistor can be designed to operate with the selected current level and a specified sensing range.

    [0028] The programmable threshold transistor 35 can be implemented using a floating gate memory cell, a split gate floating gate memory cell, a dielectric charge trapping memory cell, such as a SONOS device or other types of dielectric charge trapping cells known as for example BE-SONOS and TANOS, and a split gate, dielectric charge trapping memory cell. Other programmable memory cell technologies, such as phase change memory, metal oxide memory, and so on, may be utilized as well.

    [0029] Also, in embodiments of the technology, the resistor 36 can be implemented as a buried implant resistor between source and drain terminals of the programmable threshold transistor 35.

    [0030] Figure 4 is a simplified cross-section of a floating gate device having a resistor connected in parallel with its channel, and the resistor being implemented using an ion implantation process which results in a buried implant resistor 110.

    [0031] In this example, the device is implemented on a substrate 100, which can be a p-type substrate. A source terminal 101 and a drain terminal 102 are implemented by n-type ion implantation in the substrate 100. The source terminal 101 and the drain terminal 102 have contacts 107, 108 formed thereon, coupled to a source node having voltage VS and a drain node having voltage VD. A p-type channel region 113 is disposed between the buried implant resistor 110 and a gate dielectric layer 105 (tunnel oxide) that overlies the substrate 100 between the source terminal 101 and the drain terminal 102. A floating gate polysilicon layer 103 is disposed over the gate dielectric layer 105. An interpoly dielectric 106 is disposed over the floating gate polysilicon layer 103, implemented in some embodiments using a multilayer structure comprising silicon oxide, silicon nitride, and silicon oxide layers (ONO). A control gate polysilicon layer 104 is disposed over the inter-poly dielectric 106. A contact layer 109 is formed over the control gate polysilicon layer 104. Sidewall structures (not numbered) are formed along the sidewalls of the gate stack.

    [0032] The structure shown in Figure 4 can be implemented using floating gate cell manufacturing technologies, modified by an additional doping step to form the buried implant resistor 110. The buried implant resistor 110 connects the source terminal 101 and the drain terminal 102 to act as a passive resistor. In this manner, the floating gate device and the buried implant resistor 110 provide a programmable threshold transistor and a resistor connected in parallel between a first current carrying terminal, the source terminal 101, and a second current carrying terminal, the drain terminal 102.

    [0033] In Figure 4, a current path 112 is illustrated that passes through the buried implant resistor 110 between the source terminal 101 and the drain terminal 102. Also, a current path 114 is illustrated, which is activated when the gate voltage and charge trapped in the floating gate, in combination with a source voltage Vs, result in current flow through the channel of the transistor.

    [0034] Thus, the device has a variable resistance (or variable conductance) that is a function of the resistance of the buried implant resistor 110, and the resistance of the channel of the floating gate device. The resistance of the channel of the floating gate device is a function of the gate voltage, and of the charge trapped in the floating gate.

    [0035] Figure 5 is a simplified cross-section of a dielectric charge trapping device having a resistor connected in parallel with its channel, and implemented using an ion implantation process which results in a buried implant resistor 210.

    [0036] In this example, the device is implemented on a substrate 200, which can be a p-type substrate. A source terminal 201 and a drain terminal 202 are implemented by n-type ion implantation in the substrate 200. The source terminal 201 and the drain terminal 202 have contacts 207, 208 formed thereon, coupled to a source node having voltage VS and a drain node having voltage VD. A p-type channel region 213 is disposed between the buried implant resistor 210 and a tunneling dielectric layer 205 that overlies the substrate 200 between the source terminal 201 and the drain terminal 202. A dielectric charge trapping layer 203 is disposed over the tunneling dielectric layer 205. A blocking dielectric 206 is disposed over the dielectric charge trapping layer 203. A control gate polysilicon layer 204 is disposed over the blocking dielectric 206. A contact layer 209 is formed over the control gate polysilicon layer 204. Sidewall structures (not numbered) are formed along the sidewalls of the gate stack.

    [0037] The structure shown in Figure 5 can be implemented using dielectric charge trapping memory cell manufacturing technologies, modified by an additional doping step to form the buried implant resistor 210. The buried implant resistor 210 connects the source terminal 201 and the drain terminal 202 to act as a passive resistor. In this manner, the dielectric charge trapping device and the buried implant resistor 210 provide a programmable threshold transistor and a resistor connected in parallel between the source terminal 201 and the drain terminal 202.

    [0038] In Figure 5, a current path 212 is illustrated that passes through the buried implant resistor 210 between the source terminal 201 and the drain terminal 202. Also, a current path 214 is illustrated, which is activated when the gate voltage and charge trapped in the dielectric charge trapping layer, in combination, result in current flow through the channel of the device.

    [0039] Thus, the device has a variable resistance (or conductance) that is a function of the resistance of the buried implant resistor 210, and the resistance of the channel of the dielectric charge trapping device. The resistance of the channel of the dielectric charge trapping device is a function of the gate voltage, and of the charge trapped in the dielectric charge trapping gate.

    [0040] In both of the embodiments of Figure 4 and Figure 5, a cell is illustrated that consists of one transistor and one resistor (1T-1R). Furthermore, the embodiments of Figure 4 and Figure 5 can implement the resistor as a buried implant resistor within the layout footprint of a single variable threshold transistor, in effect making an array of one transistor (1T) cells, for a very compact layout configured for sum-of-products operations with voltage sensing.

    [0041] In operation, the cells illustrated in Figures 4 and 5 can be characterized as follows.

    [0042] While the gate-to-source voltage VGS is less than the threshold voltage Vt, current can flow in the buried implant resistor but no transistor channel ("surface channel") is formed, allowing only current IB in the buried resistor. Thus, the current in the cell is equal to IB and the resistance of the cell is equal to the drain-to-source voltage VDS divided by the current IB.

    [0043] While the gate-to-source voltage VGS is greater than the threshold voltage Vt, both the surface channel current IS and the buried resistor current IB are induced. The channel resistance can be much less than the resistance of the buried resistor, and so IS can dominate when the transistor is on. Thus, the current In in the column is divided in the cell so that it is equal to the sum IS+IB, and the cell resistance is equal to the drain-to-source voltage VDS divided by the current In.

    [0044] Since the threshold of the floating gate or dielectric charge trapping cell is programmable, this cell resistance can emulate a product of a parameter X(i) represented by the gate voltage, and a parameter W(i) represented by the charge trapped in the cell, the resistance of the resistor in the cell and the cell current. The parameter W(i) can be a binary value, where the cell operates in one of two states (IB only higher resistance state and IB+IS lower resistance state). If the cell is operated in the linear region for FET behavior, then the parameter W(i) can be analog, and vary in range according the charge trapped in the cell.

    [0045] Figures 6-9 illustrate stages in a manufacturing process that can be used to implement a cell like that of Figure 4. In Figure 6, a substrate 300 is shown after formation of shallow trench isolation structures 301 and 302 that provide dielectric boundaries for a cell. Also, a well implant has been applied to form a p-type well represented by the boundary 303 providing a region in the substrate 300 in which the cells are formed. Different blocks of cells in an array can be implemented in separate blocks, allowing independent biasing of the wells for the separate blocks.

    [0046] Figure 7 illustrates a stage after a buried channel implant of n-type dopants such as phosphorus and arsenic are applied to form a buried implant resistor 304 between the shallow trench isolation structures 301 and 302.

    [0047] Figure 8 illustrates a stage after formation of the gate stack structures (floating gate 315, control gate 316, tunneling dielectric and interpoly dielectric along with sidewalls 320) and formation of source and drain regions 310 and 311 using an implant of n-type dopants.

    [0048] Figure 9 illustrates a stage in the manufacturing after formation of an interlayer dielectric 322 and interlayer contacts 325 and 326. The structures are formed in the illustrated example, using a process that forms silicide contacts on the source and drain regions, followed by a thin dielectric and an etch stop layer 321 over the gate stack and source and drain regions 310 and 311. The interlayer dielectric 322 is deposited, and vias are etched to form openings in which the source and drain contacts 325 and 326 are formed by tungsten deposition or other techniques.

    [0049] As can be seen, a variable resistance cell such as that shown in Figure 4 is manufactured according to these procedures. These same procedures, can be modified for the purposes of manufacturing a cell such as that shown in Figure 5, by modifying the gate stack including the gate dielectric, the charge trapping layer, the blocking layer and the control gate.

    [0050] Variable resistance cells having structures like that shown in Figures 4 and 5 can be arranged in series using patterned conductor layers connected to contacts 325, 326.

    [0051] Figures 10A and 10B are cross-section and layout views of variable resistance cells arranged in series in a NAND-like structure.

    [0052] Figure 10A shows a simplified cross-section of a substrate in which a series-connected string 400 of variable resistance cells is formed. The gate stacks 410-415, including charge trapping layers (floating gate or dielectric) and word lines, overlie the substrate and extend as elements of word lines in a direction perpendicular to the drawing page. In representative implementations, there may be for example 32 or 64 active word lines. In some embodiments, the series-connected strings may include a smaller number of active word lines or a larger number as suits a particular implementation. In some cases, there may be one or more dummy word lines which can be on opposing ends of the strings as is typical for example in high density NAND flash memory. Dummy word lines can be implemented for manufacturing quality or biasing purposes but not used in the sum-of-products operations of the string.

    [0053] In this example, the substrate is a p-type substrate, and current carrying terminals (i.e. source/drain terminals) of the variable resistance cells are implemented by n-type implants 420-427. In some high density embodiments, implants are not used in current carrying terminals between the cells, so the current carrying terminals rely on inversion for charge carriers as in the channel regions. No contacts are made directly to the current carrying terminals between all the cells in a NAND-like embodiment as illustrated.

    [0054] String select word lines 401 and 402 are disposed on opposing ends of the series-connected strings. Active regions 504 and 505 comprise n-type implants in the substrate used for bit line and common source line connections for the series-connected strings. The active regions 504 and 505 can be deeper implants or higher conductivity implants then the current carrying terminals of the variable resistance cells. A bit line contact 502 connects the active region 504 to a bit line in an overlying patterned conductor layer. A source line contact 503 connects the active region 505 to a source line in an overlying patterned conductor layer.

    [0055] An n-type buried implant resistor 451 is implemented, extending in this example from the edge of a channel of a select gate controlled by the bit line side string select word line 401, to the edge of the channel of a select gate controlled by source line side string select line 402. In this manner, the select gates operate to connect and disconnect the buried implant resistor 451 to the active regions 504, 505.

    [0056] In this example, a p-type guard layer 450, having a higher p-type impurity concentration than the channel regions of the variable resistance cells, is disposed between the channels and the buried implant resistor 451. The p-type guard layer 450 helps to shield the buried implant resistor 451from the gate voltage, and maintain the stability of the parallel resistance value.

    [0057] Figure 10B is a plan view diagram of series connected strings of variable resistance cells implemented as shown in Figure 10A. Common reference numerals (401, 402, 410-415) are given to the gate stacks (including word lines) and select lines. Likewise, the common reference numbers 502, 503 are given to bit line and source line contacts.

    [0058] Figure 10B shows two series connected strings arranged in parallel with bit lines 500, 501, which are implemented in overlying patterned conductor layers above the gate stacks 410-415.

    [0059] Embodiments of sum-of-products arrays utilizing variable resistance cells can have very large arrays, including arrays having thousands or millions of variable resistance cells. Manufacturing techniques that are utilized for large-scale NAND devices can be applied, with the addition of steps for implementation of the buried implant resistors or other resistor structures, in the manufacturing of large sum-of-products arrays implemented in NAND-like structures as shown in Figures 10A and 10B. Operating techniques for writing (program and erase) weights to the programmable resistance cells can be like those utilized for large-scale NAND devices can be applied. As mentioned above, the programmable resistance cells can be operated in an analog mode. In the analog mode, peripheral circuitry for sensing circuitry and signal routing can be complex.

    [0060] The peripheral circuitry can be simplified by configuring the programmable resistance cells in the array of cells to operate in a "binary" mode. The programmable threshold transistors can store binary state. The current applied to the columns can be constant, or applied in a fixed number of binary levels. The resistors in the programmable resistance cells can be constant throughout the array, or be implemented in a fixed number of binary levels of resistance.

    [0061] Binary mode operation can allow simplification of the peripheral circuits, by reducing the complexity of the programming algorithms needed to program the threshold in the cells, the current sources used to apply current to the columns in the array and the sensing circuitry used to generate output values.

    [0062] Figure 11A shows a circuit schematic of a single programmable resistance cell. Figure 11B provides an IV curve (a current versus voltage plot) by which operation of the cell in a one bit per cell, binary mode can be understood. The cell as current carrying nodes 600 and 601. An input node 602 is connected to the gate of a programmable transistor as discussed above. The resistance of the parallel resistor in the cell is set at a value Rmn, where m corresponds to the row of the cell, and n corresponds to the column of the cell.

    [0063] The graph in Figure 11B shows two voltage versus current traces. The first voltage versus current trace corresponds to a "1" cell weight Wmn in which the cell has a low threshold Low Vt. The second trace corresponds to a "0" cell weight Wmn in which the cell has a high threshold High Vt. When the input value is low, so that the Low Vt is greater than the input voltage, the transistor in the cell is off and conducts very low current for either binary weight of the cell. When the input value is high, so that the Low Vt is less than the input voltage, and the input voltage is less than the High Vt, then the transistor in the cell is on if the weight is "1" corresponding to a Low Vt cell, and the transistor is off if the weight is "0" corresponding to a High Vt cell.

    [0064] When the transistor is off, a larger voltage drop VdLg dominated by the voltage drop caused by current through the resistor IRmn is induced. When the transistor is on, a smaller voltage drop VdSm which can be considered to be close to 0 V, dominated by the voltage drop caused by current through the transistor channel, is induced. This relationship is illustrated in the following TABLE 1.
    TABLE 1
    Vn (voltage drop)Weight Wmn
    Transistor state "0" High Vt (HVt) "1" Low Vt (LVt)
    "1" High : HVt > Input > LVt OFF, (IR, VdLg ) ON, (0, VdSm )
    "0" Low : LVt > Input OFF, (IR, VdLg) OFF, (IR, VdLg)


    [0065] The binary operation can be extended to a string of variable resistance cells as illustrated in Figure 12. In Figure 12, three cells are illustrated on a single string in a column n of the array. The column receives a constant current In and input values X1 to X3 on respective rows. The voltage drop in the column depends on the weights W1n, W2n and W3n of the individual cells in the column, and the input values. This example uses three individual cells to implement three terms XiWi for i going from 1 to 3 of a sum-of-products operation, to generate a voltage Vn representing the sum.

    [0066] With three input variables, shown in the first column, and three potential weights, shown across the second row of the table, and assuming a constant current and constant resistor value for each of the cells, the variations on the voltage drop Vn on the column can be seen (assuming VdSm is close to ("∼") zero) in the following TABLE 2.
    TABLE 2
    VnWeight [W1n W2n W3n]
    Input [x1 x2 x3][000][001][010][011][100][101][110][111]
    [000] 3 VdLg 3 VdLg 3 VdLg 3 VdLg 3 VdLg 3 VdLg 3 VdLg 3 VdLg
    [001] 3 VdLg ∼ 2 VdLg 3 VdLg ∼ 2 VdLg 3 VdLg ∼ 2 VdLg 3 VdLg ∼ 2VdLg
    [010] 3 VdLg 3 VdLg ∼ 2 VdLg ∼ 2 VdLg 3 VdLg 3 VdLg ∼ 2VdLg
    [011] 3 VdLg ∼ 2 VdLg 2 VdLg ∼ VdLg 3 VdLg 2 VdLg ∼ VdLg
    [100] 3 VdLg 3 VdLg 3 VdLg 3 VdLg ∼ 2 VdLg ∼ 2 VdLg ∼ 2 VdLg
    [101] 3 VdLg 2 VdLg 3 VdLg ∼ 2 VdLg ∼ 2 VdLg ∼ VdLg 2 VdLg ∼ VdLg
    [110] 3 VdLg 3 VdLg ∼ 2 VdLg ∼ 2 VdLg ∼ 2 VdLg ∼ 2 VdLg ∼ VdLg ∼ VdLg
    [111] 3 VdLg ∼ 2 VdLg ∼ 2 VdLg ∼ VdLg ∼ 2 VdLg ∼ VdLg ∼ VdLg ∼ 0


    [0067] By setting the sensing reference voltages according to these four levels of Vn, the voltage across the columns can be translated into numerical outputs between 0 and 3, as shown in the following TABLE 3.
    TABLE 3
    Voltage drop (3 - Vn / VdLg)Weight [W1n W2n W3n]
    Input [x1 x2 x3][000][001][010][011][100][101][110][111]
    [000] 0 0 0 0 0 0 0 0
    [001] 0 1 0 1 0 1 0 1
    [010] 0 0 1 1 0 0 1 1
    [011] 0 1 1 2 0 1 1 2
    [100] 0 0 0 0 1 1 1 1
    [101] 0 1 0 1 1 2 1 2
    [110] 0 0 1 1 1 1 2 2
    [111] 0 1 1 2 1 2 2 3


    [0068] As the number of rows providing unique inputs and number of columns of cells increases, the array can generate complex sum-of-productss while relying on "binary" operation (i.e., programming the transistors to a low threshold or a high threshold) of the individual programmable resistance cells.

    [0069] In some embodiments, multibit binary weights may be stored in some or all of the cells in the array, adding further resolution to programmable weights of the cells.

    [0070] Figure 13 is a block diagram of sensing circuitry that can be utilized with an array of variable resistance cells configured for sum-of-products operation, with voltage sensing as described above. Sensing circuitry in this example includes a sense amplifier 650 implemented for example using an operational amplifier or other type of comparator. The inputs to the sense amplifier 650 include the voltage Vn and a reference voltage Vref on line 652. The voltage Vn is developed on a selected column, and can be delivered through a buffer 651. The buffer 651 can be implemented for example using a unity gain configured operational amplifier or other voltage to voltage amplifier. The reference voltage Vref on line 652 is provided by a reference voltage circuit 655, which is configured to sequence through a set of reference voltages corresponding to each of the voltage levels to be distinguished by the sense amplifier 650 in response to sequencing signals on line 656. The reference voltage circuit 655 can receive input voltages Vmax and Vmin which can determine minimum and maximum voltages to be developed on line 652 as the reference voltage Vref.

    [0071] Figure 13A is a chart that illustrates heuristically the sensing operation. Given voltages Vmax and Vmin, circuitry of Figure 14 can generate reference voltages at a plurality of levels within a sensing range as indicated by the chart. The voltage Vn developed on a selected column in the array can fall at a level within the sensing range that has a voltage difference above the voltage Vmin. The sensing circuitry determines a level for the voltage Vn, which in this case is higher than each of the reference voltages V1 to V5 and lower than the reference voltage V6. It can therefore assign a digital value to the voltage Vn corresponding to the reference voltage V6.

    [0072] The output of the sense amplifier 650 comprises a sequence of signals that correspond with the input reference voltage level. These signals can be stored in registers 660, that are provided to an arithmetic logic unit 661 or other type of processing circuitry like a digital signal processor, general purpose processor, or the like, where further arithmetic operations can be executed to further the sum-of-products operations. For example, the outputs generated on a plurality of columns of the array can be combined for the purposes of generating a single term of the sum-of-products operation, depending on how the array of programmable resistance cells configured as discussed below.

    [0073] Figure 14 is a block diagram of a reference voltage circuit which can be utilized with a sense amplifier arrangement like that of Figure 13. In the diagram of Figure 14, a reference column or reference columns 665 in the array of programmable resistance cells, or using cell structures like those used in the array, can be arranged to provide one or both of the voltages Vmax and Vmin. Voltages Vmax and Vmin this example are applied to a resistive voltage divider 666, which produces a plurality of reference voltage levels at nodes between the resistors in the divider 666. The nodes responding to the reference voltage levels are coupled to a selector 667. The selector 667 is responsive to the sequencing signals on line 656 to provide a sequence of reference voltages Vref on line 652, coupled to the sense amplifier 650 in the configuration of Figure 13.

    [0074] Figure 15 illustrates one configuration of a reference column configured for the purposes of generating voltage Vmin that can be used as described with reference to Figures 13 and 14 for the purposes of generating the reference voltages used in the sensing circuitry. In the example, a three cell operating string 680 on column n is configured for a sum-of-products operation, where the cells have inputs X1 to X3 and weights W1n, W2n and W3n. The weights are programmed into the operating string 680 according to the terms of the sum-of-products operation to be executed. The voltage generated by a current In through the string is designated Vn.

    [0075] Reference column 681 is implemented in the array, using three cells that can have electrical characteristics matching those of the three cells used in the operating string 680. In order to generate the voltage Vmin, the weights of the cells in the reference string 681, designated W1ref, W2ref and W3ref, are all set to the value (in this case "1") of corresponding to the low threshold state. The resistors in the cells of the reference string 681 can have a constant value R that matches the value R in the cells in the operating string 680. The inputs to the cells in the reference string 681 are tied together and coupled to a voltage VON during operation so that all the cells in the reference string 681 are turned on, and generate the small voltage drop VdSm. Thus the voltage Vmin in this example will be equal to about 3*VdSm, or three times the small voltage drop of the unit cell used in the operating string 680. TABLE 4 below illustrates an operating example for a given input and weight configuration for the operating string (computation column) and the reference column.
    TABLE 4
    Computation columnReference columnVoltage difference for each bit
    InputWeightVoltage dropInput / WeightVoltage drop
    x1 1 W1n 1 VdSm 1/1 VdSm VdSm - VdSm = 0
    x2 0 W2n 1 VdLg 1/1 VdSm VdLg - VdSm = Vd
    x3 1 W3n 0 VdLg 1/1 VdSm VdLg-VdSm = Vd


    [0076] In an embodiment in which a reference string is used to generate only Vmin, the value Vmax used by the reference voltage circuitry can be set at a high enough value to provide a good operating margin for the device. The example shown in Figure 15 is based on a string including three variable resistance cells.

    [0077] In embodiments of the present technology, the variable resistance cells can be implemented in large-scale arrays using NAND -like technologies. Thus, any given column of cells that is coupled in a string can have, for example, 16, 32, 64 or more cells. In the configuration of any given sum-of-products operation, less than all of the cells in a given column may be utilized.

    [0078] Figure 16 illustrates an example configuration including operating string 690 in operating column n and a reference string 691 in a reference column with a plurality of unused cells in region 692 on the operating and reference columns. The reference column in this example is configured for the purposes of generating voltage Vmin that can be used as described with reference to Figures 13 and 14 for the purposes of generating the reference voltages used in the sensing circuitry.

    [0079] In the example shown, a three cell operating string 690 on the operating column n is configured for a sum-of-products operation, where the cells on the operating string 690 have inputs X1 to X3 and weights W1n, W2n and W3n. The weights are programmed into the operating string 690 according to the terms of the sum-of-products operation to be executed. Unused cells on the operating column n are given inputs Y1 and Y2 and weights W4n and W5n. The voltage generated by a current In through the string is designated Vn. The inputs Y1 and Y2 and weights W4n and W5n are configured so that the unused cells in the operating column n are turned on during the sum-of-products operation.

    [0080] Reference string 691 is implemented in the array or in a reference array, in a reference column using three cells that can have electrical characteristics matching those of the three cells used in the operating string 690. Unused cells on the reference column including the reference string 691 have weights W4ref and W5ref. In order to generate the voltage Vmin, the weights of the cells in the reference string 691, designated W1ref, W2ref and W3ref, and of the cells having weights W4ref and W5ref in the unused part of the column, are all set to the value (in this case "1") of corresponding to the low threshold state. The resistors in the cells of the reference string 691 in the unused part of the column in region 692 can have a constant value R that matches the value R in the cells in the operating string 690 and the value R of the cells in the region 692 in the same column is the operating string 690. The inputs to the cells in the reference string 691 including the unused cells are tied together and coupled to a voltage VON during operation so that all the cells in the column including the reference string 691 are turned on, and generate the small voltage drop VdSm. Thus the voltage Vmin in this example with five cells in the string, will be equal to about 5VdSm, or five times the small voltage drop of the unit cell used in the operating string 680. With more cells in the string, the value of Vmin will be shifted accordingly.

    [0081] TABLE 5 below illustrates an operating example for a given input and weight configuration for the operating string (computation column) and the reference column, for the configuration of Figure 16.
    TABLE 5
    Computation columnReference columnVoltage difference for each bit
    InputWeightVoltage dropInput / WeightVoltage drop
    X1 1 W1n 1 VdSm 1/1 VdSm VdSm - VdSm = 0
    X2 0 W2n 1 VdLg 1/1 VdSm VdLg - VdSm = Vd
    X3 1 W3n 0 VdLg 1/1 VdSm VdLg - VdSm = Vd
    Y1 1 Wy1n 1 VdSm 1/1 VdSm VdSm - VdSm = 0
    Y2 1 Wy2n 1 Vdsm 1/1 VdSm VdSm - VdSm = 0


    [0082] Figure 17 illustrates an example configuration in which both the voltages Vmin and Vmax are generated. In this configuration, an operating string 700 in an operating column n includes three cells as in the examples of Figure 15 and Figure 16. Thus, the operating column n is configured for a sum-of-products operation, where the cells in the operating string 700 have inputs X1 to X3 and weights W1n, W2n and W3n. The weights are programmed into the operating string 700 according to the terms of the sum-of-products operation to be executed. Unused cells on the operating column n are given inputs Y1 and Y2 and weights W4n and W5n. The voltage generated by a current In through the string is designated Vn. The inputs Y1 and Y2 and weights W4n and W5n are configured so that the unused cells in the operating column n are turned on during the sum-of-products operation.

    [0083] A Vmin reference column includes a reference string 701 and unused cells in the region 703 of the array. The reference string 701 includes three cells that can have electrical characteristics matching those of the three cells used in the operating string 700. Unused cells on the Vmin reference column include the reference string 691 weights W4Lref and W5Lref· In order to generate the voltage Vmin, the weights of the cells in the reference string 701, designated W1Lref, W2Lref and W3Lref, and of the unused cells in the Vmin column having weights W4Lref and W5Lref in the unused part of the column, are all set to the value (in this case "1") corresponding to the low threshold state. The resistors in the cells of the reference string 701 and in the unused part of the column can have a constant value R that matches the value R in the cells in the operating string 700 and the value R of the cells in the region 703 in the operating column n. The inputs to the cells in the Vmin reference column including reference string 701 and the unused cells are tied together and coupled to a voltage VON during operation so that the transistors in all the cells in the Vmin reference column including the reference string 701 are turned on, and assuming Iref is equal to In, generate the small voltage drop VdSm. Thus, the voltage Vmin in this example with five cells in the string, will be equal to about 5VdSm, or five times the small voltage drop of the unit cell used in the operating string 700. With more cells in the string, the value of Vmin will be shifted accordingly.

    [0084] A Vmax reference column includes a reference string 702 and unused cells in the region 703 of the array. The reference string 702 includes three cells that can have electrical characteristics matching those of the three cells used in the operating string 700. Unused cells on the Vmax reference column including the reference string 702 have weights W4Href and W5Href. In order to generate the voltage Vmax, the weights of the cells in the reference string 702, designated W1Href, W2Href and W3Href, and of the unused cells in the Vmin column having weights W4Href and W5Href in the unused part of the Vmax reference column, are all set to the value (in this case "0") corresponding to the high threshold state. The resistors in the cells of the reference string 701 and in the unused part of the column can have a constant value R that matches the value R in the cells in the operating string 700 and the value R of the cells in the region 703 in the operating column n. The inputs to the cells in the Vmax reference column including string 702 are tied together and coupled to a voltage VOFF during operation, and the unused cells are coupled to the voltage VON, so that the transistors in three cells in the column including the Vmax reference string 701 are turned off, and assuming Iref is equal to In, generate the large voltage drop VdLg. Thus, the voltage Vmax in this example with five cells in the string, will be equal to about 3*VdLg, or three times the large voltage drop of the unit cell used in the operating string 700. With more cells in the string, the value of Vmax will be shifted accordingly.

    [0085] TABLE 6 below illustrates an operating example for a given input and weight configuration for the operating string (computation column) and the reference column, for the configuration of Figure 17.
    TABLE 6
    Computation column (Vn)Reference column, Low boundary (Vmin)Reference column, High boundary (Vmax )Voltage difference for each bit
    InputWeightVoltage dropInput / WeightVoltage dropInput / WeightVoltage drop
    X1 1 W1n 1 VdSm 1/1 VdSm 1/0 VdLg VdSm - VdSm = 0
    X2 0 W2n 1 VdLg 1/1 VdSm 1/0 VdLg VdLg - VdSm = Vd
    X3 1 W3n 0 VdLg 1/1 VdSm 1/0 VdLg VdLg - VdSm = Vd
    Y1 1 W4n 1 VdSm 1/1 VdSm 1/1 VdSm VdSm - VdSm = 0
    Y2 1 W5n 1 VdSm 1/1 VdSm 1/1 VdSm VdSm - VdSm = 0


    [0086] In the embodiments described with reference to Figures 12 and 15-17, the array of programmable resistance cells is configured into functional sets that have one input Xi, and include one member cell, each cell in the operating column implementing a term XiWi having one bit binary weight Wi, that is determined by the threshold of the programmable threshold transistor in the cell. The resistance R of the resistors in the cell, and the current In in the string are constants.

    [0087] In some embodiments, the array of programmable resistance cells can be configured into functional sets having one input and multiple members to implement a term XiWi of the sum-of-products operation in which the weight Wi can be values other than a one bit binary "0" or "1", such as a multiple bit binary value, using a single bit value programmed in the programmable transistors in the cells.

    [0088] Figures 18-22 illustrate some example configurations of functional sets that implement multiple bit binary values.

    [0089] Figure 18 illustrates a functional set including three member cells on a single string in a column n of the array. The column receives a constant current In. Input value Xm is connected to the gates of the transistors in the cells in all three rows. In this example, the resistances of the resistors R1,mn, R2,mn and R3,mn in the three cells in the set are different. Thus the resistor R3 has a resistance R, the resistor R2 has a resistance 2*R, and the resistor R1 has a resistance 4*R. Thus, the weight of the functional set has a three bit binary value, ranging from 0 to 7, based on combinations making the effective resistance vary from 0R (all transistors in the functional set on) to 7R (all transistors in the functional set off). The term of the sum-of-products operation implemented using the functional set of Figure 18 can be characterized as Xm(W14R + W22R + W3R). In other embodiments, a function set of cells in the array like that of Figure 18 can have more than three members connected in a column with a common input Xm.

    [0090] An array of cells as discussed above can be configured using logic circuits to implement terms of a sum-of-products operation using many functional sets configured to implement many forms of the terms of the operation.

    [0091] Figure 19 illustrates a functional set including three member cells in one row of the array on three different columns n1, n2 and n3 of the array. Each of the three columns receives a constant current In. Input value Xm is connected to the gates of the transistors in the cells in the row. In this example, the resistances of the resistors R1 mn, R2,mn and R3,mn in the three cells in the set are different. Thus the resistor R3 has a resistance R, the resistor R2 has a resistance 2*R, and the resistor R1 has a resistance 4R. The voltage generated in each of the column Vn1, Vn2 and Vn3 are summed in peripheral circuitry to provide the sum output of the term.

    [0092] The term of the sum-of-products operation implemented using the functional set of Figure 19 can be characterized as Xm(W1I4R + W2I2R + W3IR), with each column producing a voltage representing one part of the term. Thus, the weight of the functional set has a three bit binary value, ranging from 0 to 7, based on combinations making the voltage vary from 0IR (all transistors in the functional set on) to 7IR (all transistors in the functional set off).

    [0093] Peripheral circuitry configured to perform the sum can include analog summing amplifiers or digital logic. In one example, the voltage on each column can be sensed in sequence, and the results of each sensing step added in arithmetic logic as illustrated in Figure 13 for example.

    [0094] In other embodiments, a function set of cells in the array like that of Figure 19 can have more than three members connected in a column with a common input Xm.

    [0095] Figure 20 illustrates a functional set including three member cells in one row of the array on three different columns n1, n2 and n3 of the array. Input value Xm is connected to the gates of the transistors in the cells in the row. In this example, the resistances of the resistors R1,mn, R2,mn and R3,mn in the three cells in the set are the same. Each of the three columns receives a different constant current In. Thus a current source provides I3 to column 3 with a current I, a current source provides I2 to column 2 with a current 2I, and a current source provides I1 to column 1 with a current 4I. The voltages Vn1, Vn2 and Vn3 generated in the columns of the functional set are summed in peripheral circuitry to provide the sum output of the term. Thus, the weight of the functional set has a three bit binary value, ranging from 0 to 7, based on combinations making the output values vary from 0IR (all transistors in the functional set on) to 7IR (all transistors in the functional set off).

    [0096] The term of the sum-of-products operation implemented using the functional set of Figure 19 can be characterized as Xm(W14IR + W22IR + W3IR), with each column producing a voltage representing one part of the term.

    [0097] Peripheral circuitry configured to perform the sum can include analog summing amplifiers or digital logic. In one example, the voltage on each column can be sensed in sequence, and the results of each sensing step added in arithmetic logic as illustrated in Figure 13 for example.

    [0098] In other embodiments, a function set of cells in the array like that of Figure 20 can have more than three members connected in a column with a common input Xm.

    [0099] Figure 21 illustrates a functional set including three member cells in one row of the array on three different columns n1, n2 and n3 of the array. Input value Xm is connected to the gates of the transistors in the cells in the row. In this example, the resistances of the resistors R1,mn, R2,mn and R3,mn in the three cells in the set are the same. Each of the three columns receives a same constant current In. The voltages Vn1, Vn2 and Vn3 generated in the columns of the functional set divided by 4, 2 and 1, respectively, and then summed in peripheral circuitry to provide the sum output of the term. Thus, the weight of the functional set has a three bit binary value, ranging from 0 to 7, based on combinations making the output values vary from 0*IR (all transistors in the functional set on) to 7IR (all transistors in the functional set off).

    [0100] The term of the sum-of-products operation implemented using the functional set of Figure 21 can be characterized as Xm(W14IR + W22IR + W3IR), with each column producing a voltage that is divided in peripheral circuit to represent one part of the term.

    [0101] Peripheral circuitry configured to perform the sum can include analog summing amplifiers or digital logic. In one example, the voltage on each column can be sensed in sequence, and the results of each sensing step added in arithmetic logic as illustrated in Figure 13 for example.

    [0102] In other embodiments, a function set of cells in the array like that of Figure 21 can have more than three members connected in a column with a common input Xm.

    [0103] Figure 22 illustrates a functional set including four member cells, including two cells in one row of the array and two cells in a second row of the array, on two different columns n1 and n2 of the array. Input value Xm is connected to the gates of the transistors in all cells in both rows of the functional set. In this example, the resistances of the resistors R1,mn, R2,mn, R3,mn and R4,mn in the four cells in the set are different. Thus the resistors R3 and R4 have resistance R, the resistors R1 and R2 have resistance 4R. Each of the two columns receives a different constant current In. Thus, a current source provides I2 to column 2 with a current I, a current source provides I1 to column 1 with a current 2I. The voltages Vn1 and Vn2 generated in the two columns of the functional set are summed in peripheral circuitry to provide the sum output of the term.

    [0104] The term of the sum-of-products operation implemented using the functional set of Figure 22 can be characterized as Xm(W12I4R + W2I4R + W32IR + W3IR), with each column producing a voltage representing one part of the term. Thus, the weight of the functional set has a four bit binary value, ranging from 0 to 15, based on combinations making the output values vary from 0IR (all transistors in the functional set on) to 15IR (all transistors in the functional set off).

    [0105] Peripheral circuitry configured to perform the sum can include analog summing amplifiers or digital logic. In one example, the voltage on each column can be sensed in sequence, and the results of each sensing step added in arithmetic logic as illustrated in Figure 13 for example.

    [0106] In other embodiments, a function set of cells in the array like that of Figure 22 can have more than three members connected in a column with a common input Xm.

    [0107] Other functional set configurations can be utilized as well.

    [0108] A large array of programmable resistance cells can be configured between operations to perform complex sum-of-products operations with various functions for the terms of the summation, as needed for each computation executed. Also, the coefficients (i.e. weights) of the terms of the summation can set in nonvolatile form in the transistors in the cells, and changed by programming and erasing operations as needed for each computation executed.

    [0109] Figure 23 is a simplified chip block diagram of an integrated circuit 901 including a sum-of-products array with voltage sensing, and buried channel cells such as illustrated in Figures 5 and 6 and Figure 10A/10B, configured as a neuromorphic memory array 960.

    [0110] A word line driver 940 is coupled to a plurality of word lines 945. The driver comprises, for example, digital-to-analog converters in some embodiments that produce an input variable x(i) for each selected word line, or in the alternative a binary word line driver can be apply binary inputs. A column decoder 970 is coupled via lines 965 to one or more layers of strings of series-connected cells arranged along columns in the array 960 for selecting strings for reading sum-of-products data from and writing parameter data to the memory array 960. Addresses are supplied on bus 930 from control logic (controller) 910 to decoder 970 and driver 940. Voltage sensing sense amplifiers are coupled to the column decoder via lines 975, and are in turn coupled to buffer circuits 980. Current sources applying the load currents In are coupled with the sensing circuits. A program buffer can be included with the sense amplifiers in circuits 980 to store program data for two-level or multiple-level programming of the programmable threshold transistors in the cells. Also, the control logic 910 can include circuits for selectively applying program and inhibit voltages to the strings in the memory in response to the program data values in the program buffer.

    [0111] Sensed data from the sense amplifiers are supplied via second data lines 985 to data buffer 990, which is in turn coupled to input/output circuits 991 via a data path 993. The sense amplifiers can comprise operational amplifiers configured to apply unity gain or a desired gain level, and provide analog outputs to digital-to-analog converters or other signal processing or signal routing circuits. Additional arithmetic units and routing circuits can be included to provide for arrangement of multiple layers of strings of cells into neuromorphic circuits.

    [0112] Also, arithmetic units and routing circuits can be included to provide for arrangement of the layers of strings into matrix multiplication units.

    [0113] Input/output circuits 991 drive the data to destinations external to the integrated circuit 901. Input/output data and control signals are moved via data bus 905 between the input/output circuits 991, the control logic 910 and input/output ports on the integrated circuit 901 or other data sources internal or external to the integrated circuit 901, such as a general purpose processor or special purpose application circuitry, or a combination of modules providing system-on-a-chip functionality supported by the memory array 960.

    [0114] In the example shown in Figure 23, control logic 910, using a bias arrangement state machine, controls the application of supply voltages generated or provided through the voltage supply or supplies in block 920, for sum-of-products read operations, and parameter writing operations to set parameters, such as cell weights, represented by charge trapping levels including, for charge trapping cells and floating gate cells, erase, verify and program bias voltages. The control logic 910 is coupled to the buffer 990 and the memory array 960.

    [0115] The control logic 910 can be implemented using special-purpose logic circuitry as known in the art. In alternative embodiments, the control logic comprises a general-purpose processor, which can be implemented on the same integrated circuit, which executes a computer program to control the operations of the device. In yet other embodiments, a combination of special-purpose logic circuitry and a general-purpose processor can be utilized for implementation of the control logic.

    [0116] Figures 24-26 illustrates a configuration of system 1000 including a memory array 1002 and a sum-of-products accelerator array 1001 interconnected using a data path controller 1003. The sum-of-products accelerator array 1001 comprises an array of programmable resistance cells according to any of the embodiments described above. The memory array can comprise a NAND flash array, an SRAM array, a DRAM array, a NOR flash array, or other type of memory that can be used in coordination with the sum-of-products accelerator array 1001.

    [0117] The system can receive input/output data from outside the system as shown in Figure 24, and route the data to the memory array. The data can include configuration data used to configure the functional sets of cells that implement the terms of one or more sum-of-products operations, weights for the functional sets in the array used for the operations, and input values for the sum-of-products operations.

    [0118] As represented in Figure 25, data from the memory array 1002 can be transferred to the sum-of-products accelerator array 1001, using a direct data path that can be controlled using the data path controller 1003. Alternatively, a data path through the data path controller 1003 can be used to transfer data from the memory array 1002 to the sum-of-products accelerator array 1001, as suits a particular implementation..

    [0119] As represented in Figure 26, output data from the sum-of-products accelerator array can be applied through the data path controller 1003 to the input output data paths for the system 1000. The input output data paths for the system 1000 can be coupled to a processing unit configured to compute weights, provide inputs and utilize the outputs of the sum-of-products accelerator array.

    [0120] Also, the output data from the sum-of-products accelerator array 1001 can be routed through the data path controller 1003 back to the memory array 1002 for use in iterative sum-of-products operations.

    [0121] The system 1000 including the memory, the sum-of-products accelerator array and the data path logic can be implemented on a single integrated circuit in some embodiments. Also, the system 1000 can include on the same or different integrated circuits, arithmetic logic units, digital signal processors, general-purpose CPUs, state machines and the like that are configured to take advantage of the sum-of-products accelerator array 1001 during execution of computer processes.

    [0122] A method for using an array of programmable resistance cells according to any of the embodiments described herein can be executed using a system like that of Figures 24-26, using logic implemented on the same integrated circuit, coupled to the integrated circuit, or a combination of both which performs a configuration step in which the functional sets of cells in the array are programmed with respective weights, and an operating step, and which the array is used to produce the sum of products data.

    [0123] A method for operating an array of variable resistance cells to produce sum of products data comprises programming the programmable threshold transistors in the array with thresholds corresponding to values of a weight factor for the corresponding cell; selectively applying inputs to rows of cells in the array applying currents to corresponding ones of the columns of cells in the array; and sensing voltages on one or more of the columns of cells in the array.

    [0124] Such a method can include configuring cells in the array into functional sets of cells including one or more members; where the functional set implementing respective terms of a sum of products function. Each functional set can receive a corresponding input term and can be programmed with a weight that is a function of the programmable thresholds of the one or more members of the functional set. The functional set can be configured in a variety of ways, such as described above with reference to Figures 18-22. In this way, programmable resistance cells in the array which are operated with weights configured using a one bit binary mode as individual cells, can be configured into functional sets that have multibit weights. The multibit weights can be configured using functional sets of cells that have respective resistors with different resistances, using different current levels during sensing on different columns in the functional set, using arithmetic logic to combine voltages sensed on respective columns in the functional set with different weights, and otherwise as described herein.

    [0125] Also, in some embodiments, the system can be operated to use a reference column of cells to generate a column reference voltage, or a low column reference voltage and a high column reference voltage as suits a particular implementation. The method can include generating sensing reference voltages as a function of the one or more column reference voltages. The sensing operation can include comparing voltages on selected columns of cells with the sensing reference voltages to generate outputs indicating voltage levels on the selected columns.

    [0126] While the present invention is disclosed by reference to the preferred embodiments and examples detailed above, it is to be understood that these examples are intended in an illustrative rather than in a limiting sense. It is contemplated that modifications and combinations will readily occur to those skilled in the art, which modifications and combinations will be within the scope of the following claims.


    Claims

    1. A device for generating sum-of-products (901), comprising:

    an array of variable resistance cells (960), the array (960) including n columns of cells including strings of series connected cells and m rows of cells, wherein n and m are non-zero integers;

    m input drivers (991) coupled to corresponding ones of the m rows of cells, the input drivers configured to selectively apply inputs Xm to rows m;

    n column drivers (970) configured to apply currents In to corresponding ones of the n columns of cells; and

    voltage sensing circuits (980) operatively coupled to the n columns of cells to sense the voltage representing the sum-of-products of the inputs Xm by weight factors Wmn,

    characterized in that

    each variable resistance cell comprises a programmable threshold transistor (12; 35) and a resistor (14; 36) connected in parallel, wherein

    the device for generating sum-of-products (901) further comprises control and bias circuitry (910) coupled to the array (960), including logic for programming the programmable threshold transistors in the array with thresholds corresponding to values of the weight factor Wmn for the corresponding cell.


     
    2. The device (901) of claim 1, wherein the programmable threshold transistors (12; 35) are charge trapping memory transistors, and the control and bias circuitry (910) is configured to apply program bias voltages to the programmable resistance cells in the array (960) to selectively set the threshold in the programmable threshold transistors (12; 35) in the cells.
     
    3. The device (901) of any of the preceding claims, wherein the resistor (14; 36) in the variable resistance cell comprises a buried implant resistor in the charge trapping memory transistor.
     
    4. The device (901) of any of claims 1 and 2, wherein the programmable threshold transistors (12; 35) in the variable resistance cells are floating gate charge trapping transistors or dielectric charge trapping transistors, and the resistor (14; 36) in the variable resistance cell comprises a buried implant resistor in the floating gate charge trapping transistor or the dielectric charge trapping transistor, respectively.
     
    5. The device (901) of any of the preceding claims, wherein cells in the array (960) are configured into functional sets including one or more members, implementing respective terms X¡W¡ of a sum-of-products function, each functional set receiving a corresponding term input Xi and term weight Wi, the term weight being a function of the programmable thresholds of the one or more members of the functions set.
     
    6. The device (901) of any of claims 1 to 4, wherein cells in the array (960) are configured into a functional set including a plurality of members connected in series on column n, and the resistors (14; 36) in the cells in the functional set have different values of resistance (fixed ratio of R), the functional set implementing a term X¡W¡ of a sum-of-products function, receiving a corresponding term input Xi and having term weight Wi, the term weight being a function of the programmable thresholds of the plurality of members of the functions set.
     
    7. The device (901) of any of claims 1 to 4, wherein cells in the array (960) are configured into a functional set including a plurality of members arranged on different columns in the array (960), and the resistors (14; 36) in the cells in the functional set have different values of resistance, the functional set implementing a term XiWi of a sum-of-products function, receiving a corresponding term input Xi and having term weight Wi, the term weight being a function of the programmable thresholds of the plurality of members of the functions set; and
    the voltage sensing circuits (980) include voltage summing circuits operably connected to the sum voltages generated on columns of cells in the functional set, wherein optionally,
    the voltage summing circuits are operably connected to the sum voltages generated on columns of cells in the functional set, and apply fixed ratios of weight factors in the summing to the voltages on the different columns, the term weight being a function of the programmable thresholds of the plurality of members of the function set and of the fixed ratios.
     
    8. The device (901) of any of claims 1 to 4, wherein cells in the array (960) are configured into a functional set including plural strings of multiple cells, the plural strings being arranged on different columns in the array (960), and the n column drivers (970) are configured to apply currents In to corresponding ones of the n columns of cells and apply different magnitudes of current to the different columns, the functional set implementing a term XiWi of a sum-of-products function, receiving a corresponding term input Xi and having term weight Wi, the term weight being a function of the programmable thresholds of the plurality of members of the functions set and the different magnitudes of current; and
    the voltage sensing circuits (980) include voltage summing circuits operably connected to the sum voltages generated on columns of cells in the functional set.
     
    9. The device (901) of any of claims 1 to 4, wherein cells in the array (960) are configured into a functional set including cells arranged on different columns in the array (960), the resistors (14; 36) in the cells in the functional set have different values of resistance, and the n column drivers (970) are configured to apply currents In to corresponding ones of the n columns of cells apply different magnitudes of current to the different columns, the functional set implementing a term XiWi of a sum-of-products function, receiving a corresponding term input Xi and having term weight Wi, the term weight being a function of the programmable thresholds of the plurality of members of the function set, the different values of resistance, and the different magnitudes of current; and
    the voltage sensing circuits (980) include voltage summing circuits operably connected to the sum voltages generated on columns of cells in the functional set.
     
    10. The device (901) of any of the preceding claims, wherein, the array (960) includes a reference column of cells configured to generate a column reference voltage; the device (901) including:

    a circuit to generate sensing reference voltages as a function of the column reference voltage; and

    wherein the voltage sensing circuits (980) include comparators configured to compare voltages on selected columns of cells with the sensing reference voltages to generate outputs indicating voltage levels on the selected columns.


     
    11. The device (901) of any of claims 1 to 9, wherein the array (960) includes a first reference column of cells configured to generate a low column reference voltage and a second reference column of cells configured to generate a high column reference voltage; the device (901) including:
    a circuit to generate sensing reference voltages as a function of the high and low column reference voltages; and
    wherein the voltage sensing circuits (980) include comparators configured to compare voltages on selected columns of cells with the sensing reference voltages to generate outputs indicating voltage levels on the selected columns.
     
    12. A method for operating an array of variable resistance cells (960), the array (960) including n columns of cells including strings of series connected cells and m rows of cells, wherein n and m are non-zero integers, to produce sum-of-products data, comprising:

    selectively applying inputs Xm to rows m;

    applying currents In to corresponding ones of the n columns of cells; and

    sensing voltages on one or more of the n columns of cells to sense the voltage representing the sum-of-products of the inputs Xm by weight factors Wmn,

    characterized in that

    each variable resistance cell comprises a programmable threshold transistor (12; 35) and a resistor (14; 36) connected in parallel, wherein

    the method further comprises programming the programmable threshold transistors (12; 35) in the array (960) with thresholds corresponding to values of the weight factor Wmn for the corresponding cell.


     
    13. The method of claim 12, including configured cells in the array (960) into functional sets of cells including one or more members, implementing respective terms XiWi of a sum-of-products function, each functional set receiving a corresponding term input Xi and term weight Wi, the term weight being a function of the programmable thresholds of the one or more members of the functions set.
     
    14. The method of any of claims 12 and 13, including configured cells in the array (960) into a functional set including a plurality of members connected in series on column n, wherein the resistors (14; 36) in the cells in the functional set have different values of resistance, the functional set implementing a term XiWi of a sum-of-products function, receiving a corresponding term input Xi and having term weight Wi, the term weight being a function of the programmable thresholds of the plurality of members of the functions set.
     
    15. The method of any of claims 12 to 14, including
    using a reference column of cells to generate a column reference voltage;
    generating sensing reference voltages as a function of the column reference voltage; and
    comparing voltages on selected columns of cells with the sensing reference voltages to generate outputs indicating voltage levels on the selected columns.
     


    Ansprüche

    1. Vorrichtung zum Erzeugen disjunktiver Normalformen (901), umfassend
    ein Array von Zellen mit variablem Widerstand (960), wobei das Array (960) n Spalten von Zellen einschließlich Strings von in Reihe geschalteten Zellen und m Reihen von Zellen enthält, wobei n und m ganze Zahlen ungleich Null sind;
    m Eingangstreiber (991), die an entsprechenden der m Reihen von Zellen gekoppelt sind, wobei die Eingangstreiber so konfiguriert sind, dass sie selektiv Eingänge Xm an die Reihen m anlegen;
    n Spaltentreiber (970), die so konfiguriert sind, dass sie Ströme In an entsprechende der n Spalten von Zellen anlegen; und
    Spannungsmessschaltungen (980), die operativ mit den n Spalten der Zellen gekoppelt sind, um die Spannung zu erfassen, die die disjunktive Normalform der Eingänge Xm durch Gewichtsfaktoren Wmn darstellt,
    dadurch gekennzeichnet, dass
    jede Zelle mit variablem Widerstand einen programmierbaren Schwellenwerttransistor (12; 35) und einen Widerstand (14; 36) umfasst, die parallel geschaltet sind, wobei
    die Vorrichtung zur Erzeugung disjunktiver Normalformen (901) ferner eine mit dem Array (960) gekoppelte Steuer- und Vorspannungsschaltung (910) umfasst, die Logik zur Programmierung der programmierbaren Schwellenwerttransistoren in dem Array mit Schwellenwerten enthält, die Werten des Gewichtsfaktors Wmn für die entsprechende Zelle entsprechen.
     
    2. Vorrichtung (901) nach Anspruch 1, bei der die programmierbaren Schwellenwerttransistoren (12; 35) Ladungsfangspeichertransistoren sind und die Steuer- und Vorspannungsschaltung (910) so konfiguriert ist, dass sie Programmierungsvorspannungen an die programmierbaren Zellen mit Widerstand in dem Array (960) anlegt, um den Schwellenwert in den programmierbaren Schwellenwerttransistoren (12; 35) in den Zellen selektiv einzustellen.
     
    3. Vorrichtung (901) nach einem der vorhergehenden Ansprüche, wobei der Widerstand (14; 36) in der Zelle mit variablem Widerstand einen verdeckten Implantatwiderstand in dem Ladungsfangspeichertransistor umfasst.
     
    4. Vorrichtung (901) nach einem der Ansprüche 1 und 2, wobei die programmierbaren Schwellenwerttransistoren (12; 35) in den Zellen mit variablem Widerstand Ladungsfangtransistoren mit schwebendem Gate oder dielektrische Ladungsfangtransistoren sind und der Widerstand (14; 36) in der Zelle mit variablem Widerstand einen verdeckten Implantatwiderstand in dem Ladungsfangtransistor mit schwebendem Gate bzw. dem dielektrischen Ladungsfangtransistor umfasst.
     
    5. Vorrichtung (901) nach einem der vorhergehenden Ansprüche, wobei Zellen in dem Array (960) in funktionale Sätze konfiguriert sind, die ein oder mehrere Elemente enthalten, die jeweilige Terme XiWi einer Disjunktiv-Normalform-Funktion implementieren, wobei jeder funktionale Satz einen entsprechenden Term-Eingang Xi und ein Term-Gewicht Wi erhält, wobei das Term-Gewicht eine Funktion der programmierbaren Schwellenwerte des einen oder der mehreren Elemente des funktionalen Satzes ist.
     
    6. Vorrichtung (901) nach einem der Ansprüche 1 bis 4, wobei Zellen in dem Array (960) in einen funktionalen Satz konfiguriert sind, der eine Vielzahl von Elementen enthält, die in Reihe auf Spalte n geschaltet sind, und die Widerstände (14; 36) in den Zellen in dem funktionalen Satz unterschiedliche Widerstandswerte (festes Verhältnis von R) haben, wobei der funktionale Satz einen Term XiWi einer Disjunktiv-Normalform-Funktion implementiert, einen entsprechenden Term-Eingang Xi erhält und ein Term-Gewicht Wi hat, wobei das Term-Gewicht eine Funktion der programmierbaren Schwellenwerte der Vielzahl von Elementen des funktionalen Satzes ist.
     
    7. Vorrichtung (901) nach einem der Ansprüche 1 bis 4, wobei Zellen in dem Array (960) in einen funktionalen Satz konfiguriert sind, der eine Vielzahl von Elementen enthält, die auf verschiedenen Spalten in dem Array (960) angeordnet sind, und die Widerstände (14; 36) in den Zellen in dem funktionalen Satz verschiedene Widerstandswerte haben, wobei der funktionale Satz einen Term XiWi einer Disjunktiv-Normalform-Funktion implementiert, einen entsprechenden Term-Eingang Xi empfängt und ein Term-Gewicht Wi hat, wobei das Term-Gewicht eine Funktion der programmierbaren Schwellenwerte der Vielzahl von Elementen des funktionalen Satzes ist; und
    die Spannungsmessschaltungen (980) Spannungssummationsschaltungen enthalten, die operativ mit den Summenspannungen verbunden sind, die an Spalten von Zellen in dem funktionalen Satz erzeugt werden, wobei optional
    die Spannungssummationsschaltungen operativ mit den Summenspannungen verbunden sind, die an den Spalten der Zellen in dem funktionalen Satz erzeugt werden, und bei der Summierung feste Verhältnisse von Gewichtungsfaktoren auf die Spannungen an den verschiedenen Spalten anlegen, wobei die Gewichtung eine Funktion der programmierbaren Schwellenwerte der Vielzahl von Elementen des funktionalen Satzes und der festen Verhältnisse ist.
     
    8. Vorrichtung (901) nach einem der Ansprüche 1 bis 4, wobei die Zellen in dem Array (960) in einen funktionalen Satz konfiguriert sind, der mehrere Strings aus mehreren Zellen enthält, wobei die mehreren Strings auf verschiedenen Spalten in dem Array (960) angeordnet sind, und die n Spaltentreiber (970) so konfiguriert sind, dass sie Ströme In an entsprechende der n Spalten von Zellen anlegen und verschiedene Stromstärken an die verschiedenen Spalten anlegen, der funktionale Satz, der einen Term XiWi einer Disjunktiv-Normalform-Funktion implementiert, einen entsprechenden Term-Eingang Xi erhält und ein Term-Gewicht Wi hat, wobei das Term-Gewicht eine Funktion der programmierbaren Schwellenwerte der Vielzahl von Elementen des funktionalen Satzes und der verschiedenen Stromstärken ist; und
    die Spannungsmessschaltungen (980) Spannungssummationsschaltungen enthalten, die mit den Summenspannungen, die an den Spalten der Zellen des funktionalen Satzes erzeugt werden, operativ verbunden sind.
     
    9. Vorrichtung (901) nach einem der Ansprüche 1 bis 4, wobei die Zellen in dem Array (960) in einen funktionalen Satz konfiguriert sind, der Zellen enthält, die auf verschiedenen Spalten in dem Array (960) angeordnet sind, die Widerstände (14; 36) in den Zellen in dem funktionalen Satz unterschiedliche Widerstandswerte haben und die n Spaltentreiber (970) so konfiguriert sind, dass sie Ströme In an entsprechende der n Spalten der Zellen anlegen, unterschiedliche Stromstärken an die verschiedenen Spalten anlegen, wobei der funktionale Satz einen Term XiWi einer Disjunktiv-Normalform-Funktion implementiert, einen entsprechenden Term-Eingang Xi erhält und ein Term-Gewicht Wi hat, wobei das Term-Gewicht eine Funktion der programmierbaren Schwellenwerte der Vielzahl von Elementen des funktionalen Satzes, der unterschiedlichen Widerstandswerte und der unterschiedlichen Stromstärken ist; und
    die Spannungsmessschaltungen (980) Spannungssummationsschaltungen enthalten, die mit den Summenspannungen, die an Spalten der Zellen in dem funktionalen Satz erzeugt werden, operativ verbunden sind.
     
    10. Vorrichtung (901) nach einem der vorhergehenden Ansprüche, wobei das Array (960) eine Referenzspalte von Zellen enthält, die so konfiguriert ist, dass sie eine Spaltenreferenzspannung erzeugt; wobei die Vorrichtung (901) umfasst:

    eine Schaltung zur Erzeugung von Abtastreferenzspannungen als eine Funktion der Spaltenreferenzspannung; und

    wobei die Spannungsmessschaltungen (980) Komparatoren enthalten, die so konfiguriert sind, dass sie Spannungen an ausgewählten Spalten von Zellen mit den Messreferenzspannungen vergleichen, um Ausgaben zu erzeugen, die Spannungspegel an den ausgewählten Spalten anzeigen.


     
    11. Vorrichtung (901) nach einem der Ansprüche 1 bis 9, wobei das Array (960) eine erste Bezugsspalte von Zellen, die so konfiguriert ist, dass sie eine niedrige Spaltenbezugsspannung erzeugt, und eine zweite Bezugsspalte von Zellen, die so konfiguriert ist, dass sie eine hohe Spaltenbezugsspannung erzeugt, enthält; wobei die Vorrichtung (901) umfasst:
    eine Schaltung zur Erzeugung von Abtastreferenzspannungen als Funktion der hohen und niedrigen Spaltenreferenzspannungen; und
    wobei die Spannungsmessschaltungen (980) Komparatoren enthalten, die so konfiguriert sind, dass sie Spannungen an ausgewählten Spalten von Zellen mit den Messreferenzspannungen vergleichen, um Ausgaben zu erzeugen, die Spannungspegel an den ausgewählten Spalten anzeigen.
     
    12. Verfahren zum Betreiben eines Arrays von Zellen (960) mit variablem Widerstand, wobei das Array (960) n Spalten von Zellen einschließlich Strings von in Reihe geschalteten Zellen und m Reihen von Zellen enthält, wobei n und m ganze Zahlen ungleich Null sind, um disjunktive Normalformen zu erzeugen, umfassend:

    selektives Anwenden von Eingaben Xm auf Reihen m;

    Anlegen von Strömen In an entsprechende der n Spalten von Zellen; und

    Abtasten von Spannungen an einer oder mehreren der n Spalten der Zellen, um die Spannung abzutasten, die die disjunktive Normalform der Eingänge Xm durch Gewichtsfaktoren Wmn darstellt,

    dadurch gekennzeichnet, dass

    jede Zelle mit variablem Widerstand einen programmierbaren Schwellenwerttransistor (12; 35) und einen Widerstand (14; 36), die parallel geschaltet sind, umfasst, wobei

    das Verfahren ferner umfasst

    Programmieren der programmierbaren Schwellentransistoren (12; 35) im Array (960) mit Schwellenwerten, die den Werten des Gewichtsfaktors Wmn für die entsprechende Zelle entsprechen.


     
    13. Verfahren nach Anspruch 12, bei dem konfigurierte Zellen im Array (960) in funktionale Sätze von Zellen mit einem oder mehreren Elementen einbezogen werden, wobei die jeweiligen Terme XiWi einer Disjunktiv-Normalform-Funktion implementiert werden, wobei jeder funktionale Satz einen entsprechenden Term-Eingang Xi und ein Term-Gewicht Wi erhält, wobei das Term-Gewicht eine Funktion der programmierbaren Schwellenwerte des einen oder der mehreren Elemente des funktionalen Satzes ist.
     
    14. Verfahren nach einem der Ansprüche 12 und 13, bei dem konfigurierte Zellen im Array (960) in einen funktionalen Satz von Zellen mit einer Vielzahl von Elementen in Reihe auf Spalte n einbezogen werden, wobei die Widerstände (14; 36) in den Zellen im funktionalen Satz unterschiedliche Widerstandswerte haben, wobei der funktionale Satz einen Term XiWi einer Disjunktiv-Normalform-Funktion implementiert, einen entsprechenden Term-Eingang Xi empfängt und ein Term-Gewicht Wi hat, wobei das Term-Gewicht eine Funktion der programmierbaren Schwellenwerte der Vielzahl von Elementen des funktionalen Satzes ist.
     
    15. Verfahren nach einem der Ansprüche 12 bis 14, enthaltend
    Verwenden einer Referenz-Zellenspalte zur Erzeugung einer Spaltenreferenzspannung;
    Erzeugen von Abtastreferenzspannungen als Funktion der Spaltenreferenzspannung; und
    Vergleichen der Spannungen an ausgewählten Zellenspalten mit den Abtastreferenzspannungen, um Ausgaben zu erzeugen, die Spannungspegel an den ausgewählten Spalten anzeigen.
     


    Revendications

    1. Dispositif pour générer une somme de produits (901), comprenant :

    un réseau de cellules à résistance variable (960), le réseau (960) comprenant n colonnes de cellules comprenant des chaînes de cellules connectées en série et m rangées de cellules, dans lequel n et m sont des nombres entiers non nuls ;

    m pilotes d'entrée (991) couplés aux rangées correspondantes des m rangées de cellules, les pilotes d'entrée étant conçus pour appliquer sélectivement des entrées Xm aux rangées m ;

    n pilotes de colonne (970) conçus pour appliquer des courants In aux colonnes correspondantes des n colonnes de cellules ; et

    des circuits de détection de tension (980) couplés fonctionnellement aux n colonnes de cellules pour détecter la tension représentant la somme de produits des entrées Xm par des facteurs de pondération Wmn'

    caractérisé en ce que

    chaque cellule à résistance variable comprend un transistor à seuil programmable (12 ; 35) et une résistance (14 ; 36) connectés en parallèle, dans lequel

    le dispositif pour générer une somme de produits (901) comprend en outre des circuits de commande et de polarisation (910) couplés au réseau (960), y compris une logique pour programmer les transistors à seuil programmable dans le réseau avec des seuils correspondant aux valeurs du facteur de pondération Wmn pour la cellule correspondante.


     
    2. Dispositif (901) selon la revendication 1, dans lequel les transistors à seuil programmable (12 ; 35) sont des transistors à mémoire de piégeage de charge, et les circuits de commande et de polarisation (910) sont conçus pour appliquer des tensions de polarisation de programme aux cellules à résistance programmables dans le réseau (960) pour définir sélectivement le seuil dans les transistors à seuil programmable (12 ; 35) dans les cellules.
     
    3. Dispositif (901) selon l'une quelconque des revendications précédentes, dans lequel la résistance (14 ; 36) dans la cellule à résistance variable comprend une résistance d'implant enterrée dans le transistor à mémoire de piégeage de charge.
     
    4. Dispositif (901) selon l'une quelconque des revendications 1 et 2, dans lequel les transistors à seuil programmable (12 ; 35) dans les cellules à résistance variable sont des transistors à piégeage de charge à grille flottante ou des transistors à piégeage de charge diélectrique, et la résistance (14 ; 36) dans la cellule à résistance variable comprend une résistance d'implant enterrée dans le transistor à piégeage de charge à grille flottante ou le transistor à piégeage de charge diélectrique, respectivement.
     
    5. Dispositif (901) selon l'une quelconque des revendications précédentes, dans lequel les cellules dans le réseau (960) sont configurées en ensembles fonctionnels comprenant un ou plusieurs éléments, mettant en œuvre les termes respectifs XiWi d'une fonction de somme de produits, chaque ensemble fonctionnel recevant une entrée de terme correspondante Xi et un poids de terme Wi, le poids de terme étant une fonction des seuils programmables du ou des éléments de l'ensemble de fonctions.
     
    6. Dispositif (901) selon l'une quelconque des revendications 1 à 4, dans lequel les cellules dans le réseau (960) sont configurées en un ensemble fonctionnel comprenant une pluralité d'éléments connectés en série sur la colonne n, et les résistances (14 ; 36) dans les cellules dans l'ensemble fonctionnel ont des valeurs de résistance différentes (rapport fixe de R), l'ensemble fonctionnel mettant en œuvre un terme XiWi d'une fonction de somme de produits, recevant une entrée de terme correspondante Xi et ayant un poids de terme Wi, le poids de terme étant une fonction des seuils programmables de la pluralité d'éléments de l'ensemble de fonctions.
     
    7. Dispositif (901) selon l'une quelconque des revendications 1 à 4, dans lequel les cellules dans le réseau (960) sont configurées en un ensemble fonctionnel comprenant une pluralité d'éléments disposés sur différentes colonnes dans le réseau (960), et les résistances (14 ; 36) dans les cellules dans l'ensemble fonctionnel ont des valeurs de résistance différentes, l'ensemble fonctionnel mettant en œuvre un terme XiWi d'une fonction de somme de produits, recevant une entrée de terme correspondante Xi et ayant un poids de terme Wi, le poids de terme étant une fonction des seuils programmables de la pluralité d'éléments de l'ensemble de fonctions ; et
    les circuits de détection de tension (980) comprennent des circuits de sommation de tension connectés fonctionnellement aux tensions de somme générées sur des colonnes de cellules dans l'ensemble fonctionnel, dans lequel éventuellement,
    les circuits de sommation de tension sont fonctionnellement connectés aux tensions de somme générées sur des colonnes de cellules dans l'ensemble fonctionnel, et appliquent des rapports fixes de facteurs de pondération dans la sommation aux tensions sur les différentes colonnes, le poids de terme étant une fonction des seuils programmables de la pluralité d'éléments de l'ensemble de fonctions et des rapports fixes.
     
    8. Dispositif (901) selon l'une quelconque des revendications 1 à 4, dans lequel les cellules dans le réseau (960) sont configurées en un ensemble fonctionnel comprenant plusieurs chaînes de cellules multiples, les multiples chaînes étant disposées sur différentes colonnes dans le réseau (960), et les n pilotes de colonne (970) sont conçus pour appliquer des courants In aux colonnes correspondantes des n colonnes de cellules et appliquer des amplitudes de courant différentes aux différentes colonnes, l'ensemble fonctionnel mettant en œuvre un terme XiWi d'une fonction de somme de produits, recevant une entrée de terme correspondante Xi et ayant un poids de terme Wi, le poids de terme étant une fonction des seuils programmables de la pluralité d'éléments de l'ensemble de fonctions et des amplitudes de courant différentes ; et
    les circuits de détection de tension (980) comprennent des circuits de sommation de tension connectés fonctionnellement aux tensions de somme générées sur des colonnes de cellules dans l'ensemble fonctionnel.
     
    9. Dispositif (901) selon l'une quelconque des revendications 1 à 4, dans lequel les cellules dans le réseau (960) sont configurées en un ensemble fonctionnel comprenant des cellules disposées sur différentes colonnes dans le réseau (960), les résistances (14 ; 36) dans les cellules dans l'ensemble fonctionnel ont des valeurs de résistance différentes, et les n pilotes de colonne (970) sont conçus pour appliquer des courants In aux colonnes correspondantes des n colonnes de cellules appliquent des amplitudes de courant différentes aux différentes colonnes, l'ensemble fonctionnel mettant en œuvre un terme XiWi d'une fonction de somme de produits, recevant une entrée de terme correspondante Xi et ayant un poids de terme Wi, le poids de terme étant une fonction des seuils programmables de la pluralité d'éléments de l'ensemble de fonctions, des différentes valeurs de résistance, et des différentes amplitudes de courant ; et
    les circuits de détection de tension (980) comprennent des circuits de sommation de tension connectés fonctionnellement aux tensions de somme générées sur des colonnes de cellules dans l'ensemble fonctionnel.
     
    10. Dispositif (901) selon l'une quelconque des revendications précédentes, dans lequel le réseau (960) comprend une colonne de cellules de référence conçue pour générer une tension de référence de colonne ; le dispositif (901) comprenant :

    un circuit pour générer des tensions de référence de détection en fonction de la tension de référence de colonne ; et

    dans lequel les circuits de détection de tension (980) comprennent des comparateurs conçus pour comparer les tensions sur des colonnes de cellules sélectionnées aux tensions de référence de détection afin de générer des sorties indiquant des niveaux de tension sur les colonnes sélectionnées.


     
    11. Dispositif (901) selon l'une quelconque des revendications 1 à 9, dans lequel le réseau (960) comprend une première colonne de cellules de référence conçue pour générer une tension de référence de colonne basse et une seconde colonne de cellules de référence conçue pour générer une tension de référence de colonne haute ; le dispositif (901) comprenant :

    un circuit pour générer des tensions de référence de détection en fonction des tensions de référence de colonne haute et basse ; et

    dans lequel les circuits de détection de tension (980) comprennent des comparateurs conçus pour comparer les tensions sur des colonnes de cellules sélectionnées aux tensions de référence de détection pour générer des sorties indiquant des niveaux de tension sur les colonnes sélectionnées.


     
    12. Procédé pour faire fonctionner un réseau de cellules à résistance variable (960), le réseau (960) comprenant n colonnes de cellules comprenant des chaînes de cellules connectées en série et m rangées de cellules, dans lequel n et m sont des nombres entiers non nuls, pour produire des données de somme de produits, comprenant :

    l'application sélective d'entrées Xm aux lignes m ;

    l'application de courants In aux colonnes correspondantes des n colonnes de cellules ; et

    la détection de tensions sur une ou plusieurs des n colonnes de cellules pour détecter la tension représentant la somme de produits des entrées Xm par des facteurs de pondération Wmn',

    caractérisé en ce que

    chaque cellule à résistance variable comprend un transistor à seuil programmable (12 ; 35) et une résistance (14 ; 36) connectés en parallèle, dans lequel

    le procédé comprend en outre

    la programmation des transistors à seuil programmable (12 ; 35) dans le réseau (960) avec des seuils correspondant aux valeurs du facteur de pondération Wmn pour la cellule correspondante.


     
    13. Procédé selon la revendication 12, comprenant des cellules configurées dans le réseau (960) en ensembles fonctionnels de cellules comprenant un ou plusieurs éléments, mettant en œuvre les termes respectifs XiWi d'une fonction de somme de produits, chaque ensemble fonctionnel recevant une entrée de terme correspondante Xi et un poids de terme Wi, le poids de terme étant une fonction des seuils programmables du ou des éléments de l'ensemble de fonctions.
     
    14. Procédé selon l'une quelconque des revendications 12 et 13, comprenant des cellules configurées dans le réseau (960) en un ensemble fonctionnel comprenant une pluralité d'éléments connectés en série sur la colonne n, dans lequel les résistances (14 ; 36) dans les cellules dans le l'ensemble fonctionnel ont des valeurs de résistance différentes, l'ensemble fonctionnel mettant en oeuvre un terme XiWi d'une fonction de somme de produits, recevant une entrée de terme correspondante Xi et ayant un poids de terme Wi, le poids de terme étant une fonction des seuils programmables de la pluralité d'éléments de l'ensemble de fonctions.
     
    15. Procédé selon l'une quelconque des revendications 12 à 14, comprenant
    l'utilisation d'une colonne de cellules de référence pour générer une tension de référence de colonne ;
    la génération de tensions de référence de détection en fonction de la tension de référence de colonne ; et
    la comparaison des tensions sur les colonnes de cellules sélectionnées aux tensions de référence de détection pour générer des sorties indiquant des niveaux de tension sur les colonnes sélectionnées.
     




    Drawing
















































    REFERENCES CITED IN THE DESCRIPTION



    This list of references cited by the applicant is for the reader's convenience only. It does not form part of the European patent document. Even though great care has been taken in compiling the references, errors or omissions cannot be excluded and the EPO disclaims all liability in this regard.

    Patent documents cited in the description