The present invention relates to field programmable gate array (FPGA) technology. More particularly, the present invention relates to programmable elements for use in FPGA devices, and specifically to programmable elements configured from resistive random access memories (ReRAMs) formed from individual ReRAM devices.
FPGA integrated circuits are versatile, but are large therefore are cost sensitive and consume considerable amounts of power. Making them area efficient has been a major goal.
ReRAMs have been proposed for fabricating multiplexers in FPGA devices. A ReRAM device is a two-terminal device including an ion source layer and a solid electrolyte layer. To program a ReRAM device a voltage potential placed across the two terminals of the device causes metal ions from the ion source layer to migrate into the solid electrolyte layer to form a conductive path across the entire device. The ReRAM device is erased by applying a voltage potential across the two terminals of the device opposite in polarity to the potential that was used to program the device. This causes the metal ions to migrate back into the ion source layer from the solid electrolyte layer to eliminate the conductive path across the entire device. Most proposals suggest using a pair of ReRAM devices in back-to-back orientation (in which either the ion source layers or the solid electrolyte layers of the two ReRAM devices are connected together) so that one device is always in the reverse bias condition to avoid disturb (unintentional programming of the ReRAM memory device to its on state) during logic switching.
During normal circuit operation, a first end 24 of the memory cell 10 is connected to a first circuit node of the circuit in which the memory cell 10 is used and a second end 26 of the memory cell 10 is connected to a second circuit node of the circuit in which the memory cell 10 is used. When the ReRAM devices 12 and 14 are erased, the first node 24 remains unconnected to the second node 26 and when the ReRAM devices 12 and 14 are programmed, the first node 24 is connected to the second node 26 through the memory cell 10. As will be appreciated by persons of ordinary skill in the art, the first node 24 and the second node 26 may be any nodes in the integrated circuit that the ReRAM cell 10 will programmably connect to one another. Non-limiting examples include inputs and outputs of clocked or static logic function circuits, or interconnect conductors in a circuit routing architecture of an integrated circuit that has user-programmable connections.
To program memory cell 10, the ReRAM devices are individually programmed. A first programming potential is placed on Y-decode line 20, a second programming potential is placed on first and second ends 24 and 26 of the ReRAM cell 10, and programming transistor 16 is turned on by applying an appropriate voltage to its gate from X-decode line 20. To erase memory cell 10, the ReRAM devices are individually erased. A first erase potential is placed on Y-decode line 20, a second erase potential is placed on first and second ends 24 and 26 of the ReRAM cell 10, and programming transistor 16 is turned on by applying an appropriate voltage to its gate from X-decode line 20. ReRAM cells 12 and 14 can be programmed and erased either simultaneously, as described above, or individually by controlling the voltage potentials delivered to first and second ends 24 and 26 of ReRAM cell 10 and to X-decode line 22 and Y-decode line 20. The design of circuits for supplying such program and erase voltages at appropriate voltage levels, polarities, and timings for resistive random access memory devices formed using particular materials and having specific geometries are well within the level of ordinary skill in the art.
A contact 38 connects the drain region 32 of the transistor 16 to a segment 40 of a first metal interconnect line over which the ReRAM devices 12 and 14 of
Referring now to
The small squares in dashed lines represent the ReRAM devices 12 and 14 for each ReRAM cell used in the multiplexers 70 of
A major issue that arises when using ReRAM memory cells formed from pairs of ReRAM devices in back-to-back orientation is the possible failure of a memory cell due to one of the ReRAM devices either becoming short circuited or losing its ability to be switched off once it is programmed. With one of the devices stuck in the on state, the probability that disturb is likely to eventually cause an erased ReRAM memory cell, in which both ReRAM devices are supposed to be switched to their erased state, to fail by disturbing the working erased ReRAM device to its programmed state during normal use of the FPGA device creates a significant endurance issue for integrated circuits incorporating these memory cells, particularly in circuits such as, but not limited to, multiplexers and look-up tables (LUTs) where unpredictable logic level voltages will appear at one end of ReRAM cells disposed in unselected circuit paths.
In the exemplary prior-art multiplexer circuit shown in
In accordance with the present invention, a third ReRAM device is connected in series with two back-to-back ReRAM devices. This third device becomes a redundant element in case one of the three devices fails to erase (become open circuited). The remaining two functioning ReRAM devices are still able to provide the function. The orientation of the third device is not important because the logic voltage is so low (˜0.8V) that dividing it across any two ReRAM devices reduces the stress on each of the devices to 0.4V, a value below which the stress threatens the integrity of the programmed device.
According to one aspect of the present invention, a resistive random access memory cell includes three resistive random access memory devices, each resistive random access memory device having an ion source layer and a solid electrolyte layer. The first and second resistive random access memory devices are connected in series such that either both ion source layers or both solid electrolyte layers are adjacent to one another. A third resistive random access memory device is connected in series with the first and second resistive random access memory devices.
According to another aspect of the present invention, the ReRAM cell of the present invention is connected between a first circuit node and a second circuit node and may be used to programmably connect the first circuit node to the second circuit node.
According to other aspects of the present invention, methods are disclosed for forming integrated circuit including the resistive random access memory devices of the present invention.
Employing a ReRAM cell formed from three ReRAM devices connected in series to configure programmable circuits is a significant advantage of the present invention because it is a good solution to the endurance issue raised by the disturb phenomenon to which ReRAM memory cells configured from a pair of back-to-back ReRAM devices are susceptible and still represents a major improvement in density over prior solutions employing other programmable device technologies.
The invention will be explained in more detail in the following with reference to embodiments and to the drawing in which are shown:
Persons of ordinary skill in the art will realize that the following description of the present invention is illustrative only and not in any way limiting. Other embodiments of the invention will readily suggest themselves to such skilled persons.
Referring first of all to
The series string of ReRAM devices 102, 104, and 106 is connected to a conductor 108. Conductor 108 may be any circuit node in an integrated circuit that can be programmably connected to another circuit node in the integrated circuit. In the non-limiting example of
A first programming transistor 110 has its drain connected to conductor 112, which represents any conductor, such as a programmable routing resource or the output node of some circuit element in the FPGA or other integrated circuit that will be connected to the circuit node represented by conductor 108 using the ReRAM cell 100. The ion source end of ReRAM device 102 is connected to the conductor 112. The source of the first programming transistor 110 is connected to a programming voltage node 114, and its gate is connected to a word line 116.
A second programming transistor 118 has its drain connected to the common solid electrolyte ends of ReRAM devices 102 and 104, its source connected to a programming voltage at bit line node 120, and its gate connected to a word line 122. A third programming transistor 124 has its drain connected to the ion source end of ReRAM device 104 and to the ion source end of ReRAM device 106, its source connected to a programming voltage bit line node 126, and its gate connected to a word line 128. A fourth programming transistor 130 has its drain connected to the common connection between the solid electrolyte end of ReRAM device 106 and the gate of the input buffer 108, its source connected to a programming voltage at bit line node 132, and its gate connected to a word line 134. Word lines 122, 128, and 134 may be connected in common with one another and configured from, for example, a single strip of polysilicon gate material.
Referring now to
A contact 148 connects the drain region 144 of the transistor 118 to a segment 150 of a first metal interconnect line over which the ReRAM devices 102 and 104 of
Diffused regions 172 and 174 serve as the source and drain, respectively, of transistor 124 of
The drain 174 of third programming transistor 124 is connected by contact 178 to segment 180 of the first metal interconnect line. Contact 182 connects the segment 180 of the first metal interconnect line to segment 170 of the second metal interconnect line. Contact 184 connects segment 170 of the second metal interconnect line to the ion source layer 186 of third ReRAM device 106. The solid electrolyte layer 188 of ReRAM device 106 is connected by contact 190 to segment 192 of the first metal interconnect layer. Contact 194 connects segment 192 of the first metal interconnect layer to diffused region 196 which acts as the drain of fourth programming transistor 130. The source 198 of fourth programming transistor 130 is connected to the bit line 132 of
In the particular embodiment shown in
First programming transistor 110 of
Methods for fabricating the ReRAM cells of the present invention are readily apparent to integrated circuit designers from an examination of
Initially, all of the ReRAM devices 102, 104 and 106 in ReRAM memory cell 100 will be in an erased (i.e., non-conducting) state. ReRAM devices 104 and 106 are preferably programmed first. ReRAM device 104 is programmed by applying a programming potential (e.g., about 4V) between bit lines 120 and 126 and turning on second and third programming transistors 118 and 124 by applying appropriate voltages to the word lines 122 and 128. Similarly, ReRAM device 106 is programmed by applying a programming potential (e.g., about 4V) between bit lines 126 and 132 and turning on third and fourth programming transistors 124 and 130 by applying appropriate voltages to the word lines 128 and 134. ReRAM device 102 may be programmed by applying a programming potential between bit lines 114 and 120 and applying it to ReRAM device 102 by turning on first and second programming transistors 110 and 118 by applying appropriate voltages to the word lines 116 and 122. Erasing of the ReRAM devices 102, 104, and 106 is performed in the same manner as programming of these devices, except that the polarities of the programming potentials are reversed. The design of circuits for supplying such program and erase voltages at appropriate voltage levels, polarities, and timings for resistive random access memory devices formed using particular materials and having specific geometries are well within the level of ordinary skill in the art.
Referring now to
Circuit 220 includes a 4:1 multiplexer having inputs In 1 (indicated at reference numeral 222), In 2 (indicated at reference numeral 224), In 3 indicated at reference numeral (226), and In 4 (indicated at reference numeral 228). The output of the multiplexer is indicated at reference numeral 230 at the output of buffer 232.
In 1 at reference numeral 222 is connectable to the input of output buffer 232 by programming the ReRAM memory cell indicated within dashed lines 234 and including ReRAM devices 236, 238, and 240. In 2 at reference numeral 224 is connectable to the input of output buffer 232 by programming the ReRAM memory cell indicated within dashed lines 242 and including ReRAM devices 244, 246, and 248. In 3 at reference numeral 226 is connectable to the input of output buffer 232 by programming the ReRAM memory cell indicated within dashed lines 250 and including ReRAM devices 252, 254, and 256. In 4 at reference numeral 228 is connectable to the input of output buffer 232 by programming the ReRAM memory cell indicated within dashed lines 260 and including ReRAM devices 262, 264, and 266. In each case, the three ReRAM devices correspond to ReRAM devices 102, 104, and 106, respectively, shown in
In the embodiment shown in
Referring now to
The path proceeds from segment 160 of the second metal interconnect layer through ReRAM device 102 (contacts 152 and 158 not shown) to segment 150 of the first metal interconnect layer. From there, the path proceeds to segment 170 of the second metal interconnect layer through ReRAM device 104 (contacts 162 and 168 not shown). Segment 150 of the first metal interconnect layer is shown connected to drain diffusion 144 of the second programming transistor 118 through contact 148. ReRAM device 106 is disposed between segment 170 of the second metal interconnect layer and segment 192 of the first metal interconnect layer (contacts 186 and 190 not shown). Contact 194 connects segment 192 of the first metal interconnect layer to drain diffusion 196 of the fourth programming transistor 130. Polysilicon line 200 forms the gate of the fourth programming transistor 130. The source diffusion 198 of the fourth programming transistor is shown connected through a contact to a MuxIn bitline 312 (shown in
Referring now to
The LUT 320 is formed using sets of four CMOS passgates, the first set of which is shown inside dashed lines 338. Each passgate in each set is formed from a pair of n-channel and p-channel transistors as illustrated by n-channel transistor 340 and p-channel transistor 342 connected in parallel. Each set of four passgates can be coupled between an input line and the output of the LUT depending on the states of the four inputs A, B, C, and D.
The gates of the n channel and p-channel transistors in the passgates of each set are uniquely connected to the inputs A, B, C, and D and their complements to decode a one-of-sixteen state arrangement. The ones of the sets of CMOS passgates decoding the inputs (0000), (0001), (1110), and (1111) are shown. Thus when the states of inputs A, B, C, and D are all 0, all four passgates in only the first set 342 of passgates are turned on, connecting input line 344 to output line 346. Input line 344 is programmably connected to either VDD on line 348 or GND on line 350 using a ReRAM memory cell 352 or 354, respectively. From the above discussion, the operation of the other passgate sets (including the ones not shown in
From an examination of
While embodiments and applications of this invention have been shown and described, it would be apparent to those skilled in the art that many more modifications than mentioned above are possible without departing from the inventive concepts herein. The invention, therefore, is not to be restricted except in the spirit of the appended claims.
This patent application claims the benefit of U.S. Provisional Patent Application No. 62/432,047, entitled “Resistive Random Access Memory Cell,” filed Dec. 9, 2016, which applications are incorporated in their entirety here by this reference.
Number | Date | Country | |
---|---|---|---|
62432047 | Dec 2016 | US |