The present invention, in some embodiments thereof, relates to a gain cell with internal feedback and, more particularly, but not exclusively, to a five-transistor gain cell with internal feedback.
Modern microprocessors and other VLSI systems-on-chip (SoCs) implemented in aggressively scaled CMOS technologies are characterized by high leakage currents, and require an increasing amount of embedded memory [ref. 1]. Such embedded memory, typically implemented as 6-transistor (6T)-bitcell SRAM macrocells, not only consume an ever growing share of the total silicon area but also significantly contributes to the leakage power of the system. (The leakage power is a large share of the total power budget in deeply scaled CMOS nodes.) Unfortunately, besides several advantages like fast access speed and robust, static data retention, the 6T SRAM bitcell is relatively large, exhibits several leakage paths, and has dramatically increased failure rates under voltage scaling.
Additionally, in ultra-low power (ULP) applications the silicon area is often dominated by embedded memories, which are the main consumers of both the static and dynamic power in these applications. Supply voltage scaling down to the sub-threshold region is widely used to significantly reduce both the static and dynamic power dissipation of ULP applications. However, embedded memories, typically implemented with SRAM, have been the limiting factor for aggressive voltage scaling, since the conventional 6T SRAM bitcell becomes unreliable at near-threshold operating voltages.
Gain-cell embedded DRAM (GC-eDRAM) is an alternative to conventional SRAM memories, as it is fully logic-compatible, offers smaller area, and can consume lower standby power than SRAM. GC-eDRAM may be compatible with sub-threshold operation in mature process nodes, such as 180 nm technology. However, GC-eDRAM requires periodic power-hungry refresh cycles to retain its data.
The Data Retention Time (DRT) of GC-eDRAMs is the maximum time interval from writing a data level into the bitcell to still being able to correctly read out the written level. The DRT is primarily limited by the level set by the initial charge stored in the bitcell and the leakage currents that degrade this level. Gain cell implementations in mature technology nodes, such as 180 nm, have been shown to exhibit high DRTs of tens to hundreds of milliseconds [ref. 4,5]. However conventional 2T gain cells in newer technology nodes, such as 65 nm, display much lower DRTs of only tens of microseconds [ref. 6]. The lower DRT is a direct consequence of the substantially higher leakage currents which result in a much faster deterioration of the stored levels [ref. 5]. Depending on the type of write transistor (WT), one of the data levels has a much higher retention time than the other (e.g. data ‘1’ for a PMOS WT and data ‘0’ for a NMOS WT) [ref. 6]. However, when determining the refresh frequency, one must consider the deterioration of the weaker data level under worst-case conditions, i.e. when the write bit line (WBL) is driven to the opposite level of the stored data during retention periods. While mature technology nodes offered sufficient retention times at low operating voltages, process scaling has led to a substantial decrease in retention time, limiting the operating voltage of GC-eDRAM in sub-100 nm technologies to the near-threshold or even above-threshold domain.
Embodiments herein present gain cells, optionally for use in scaled CMOS nodes characterized by high leakage currents. The gain cell protects the “weak” data level with relatively fast decay by a conditional, cell-internal feedback path. The feedback is disabled for the “strong” data level with relatively slow decay.
The feedback path is optionally implemented by a retention element which includes two switching elements. One switching element is controlled by the write line to provide a buffer effect between the write element and the read element during data retention (standby). The second switching element opens a path to a constant voltage (or ground) when the “weak” data level is retained at the storage node.
Embodiments are presented herein of:
1) Four-transistor (4T) gain cells in which the read element is a single transistor, the retention element includes two transistors and the write element is a single transistor;
2) Five-transistor (5T) gain cells in which the read element includes two transistors, the retention element includes two transistors and the write element is a single transistor;
3) Five-transistor (5T) gain cells in which the read element is a single transistor, the retention element includes two transistors and the write element includes two transistors; and
4) Six-transistor (6T) gain cells in which the read element includes two transistors, the retention element includes two transistors and the write element includes two transistors.
According to an aspect of some embodiments of the present invention there is provided a gain cell which includes:
The write element writes a data level from the write bit line input to the gain cell when triggered by the write trigger input. The retention element buffers between the buffer node and an internal storage node during data retention, connects the buffer node to a first constant voltage during retention of a first data level at the storage node and disconnects the buffer node from the first constant voltage during retention of a second data level at the storage node. The read element decouples the storage node from the read bit line output during data read. The read element connects the read bit line output to a second constant voltage when a data level associated with the second constant voltage is read from the gain cell and disconnects the read bit line output from the second constant voltage when the opposite data level is read from the gain cell.
According to some embodiments of the invention, the read element includes:
a first read switch comprising a first read data input, a first read control input and a first read data output; and
a second read switch comprising a second read data input, a second read control input and a second read data output.
The first read data input is connected to the second constant voltage level, the first read control input is connected to the storage node, the first read data output is connected to the second read data input, the second read control input is connected to the read trigger input, and the second read data output is connected to the read bit line output.
According to some embodiments of the invention, the write element and the retention element are formed of n-type transistors, the read element is formed of p-type transistors and the first voltage level equals the second voltage level.
According to some embodiments of the invention, the write element and the retention element are formed of p-type transistors, the read element is formed of n-type transistors and the first voltage level equals the second voltage level.
According to some embodiments of the invention, the write element, the retention element and the read element are formed of the same type of transistor, and the first voltage level differs from the second voltage level.
According to some embodiments of the invention, the write element includes:
a first write switch comprising a first write data input, a first write switch input and a first write data output; and
a second write switch comprising a second write data input, a second write control input and a second write data output,
The first write switch and the second write switch are of complementary types. The first write data input and the second write data input are connected to the write bit line input, the first write control input and the second write control input are respectively connected to the write trigger input and to a complementary write trigger line, and the first write data output and the second write data output are connected to the buffer node. According to further embodiments of the invention, the first write switch is a p-type transistor and the second write switch is an n-type transistor.
According to some embodiments of the invention, the retention element connects the buffer node to the storage node during a write bit operation and disconnects the buffer node from the storage node during data retention.
According to some embodiments of the invention, the retention element includes:
a buffer switch, having a buffer input, a buffer output and a buffer control input; and
a feedback switch associated with the buffer switch, having a feedback input, a feedback output and a feedback control input, wherein the feedback input is connected to the first constant voltage;
The buffer switch input is connected to the buffer node and to the feedback switch output, and the buffer switch output is connected to the storage node and to the feedback control input and the buffer control input is connected to the write trigger input. The buffer switch connects the buffer input to the buffer output when the write trigger is on and disconnects the buffer input from the buffer output when the write trigger is off. The feedback switch connects the feedback input to the feedback output when the feedback control input is at the first data level and disconnects the feedback input from the feedback output when the feedback control input is at the second data level.
According to some embodiments of the invention, the gain cell includes multiple transistors, and at least two of the transistors operate at different threshold voltages.
According to an aspect of some embodiments of the present invention there is provided a memory array which includes multiple gain cells as embodied herein. Respective write bit line inputs of the gain cells are connected to form a common write bit line, and respective read bit line outputs of the gain cells are connected to form a common read bit line output. At least some of the gain cells contain a respective: write bit line input, read bit line output, write trigger input, read trigger input, write element, retention element and read element. The gain cell's respective write element includes a write bit line connection connected to the write bit line input, a write trigger connection connected to the write trigger line input and a buffer node connection connected to an internal buffer node. The respective write element writes a data level from the write bit line input to the gain cell when triggered by the write trigger input. The respective retention element buffers between the buffer node and an internal storage node during data retention, connects the buffer node to a first constant voltage during retention of a first data level at the storage node disconnects the buffer node from the first constant voltage during retention of a second data level at the storage node. The respective read element decouples the storage node from the read bit line output during data read. The read element connects the read bit line output to a second constant voltage when a data level associated with the second constant voltage is read from the gain cell and disconnects the read bit line output from the second constant voltage when the opposite data level is read from the gain cell.
According to an aspect of some embodiments of the present invention there is provided a gain cell which includes:
A) a write bit line input;
B) a read bit line output;
C) a write trigger input;
D) a read trigger input;
E) a write element, which includes a first write switch with a first write data input, a first write control input and a first write data output and a second write switch with a second write data input, a second write control input and a second write data output. The first write switch and the second write switch are of complementary types. The first write data input and the second write data input are connected to the write bit line input; the first write control input and the second write control input are respectively connected to the write trigger input and a complementary write trigger line; and the first write data output and the second write data output are connected to an internal buffer node;
F) a retention element; and
G) a read element which includes a read bit line connection connected to the read bit line output, a read trigger connection connected to the read trigger input and a storage node connection connected to the storage node.
The retention element buffers between the buffer node and an internal storage node during data retention, connects the buffer node to a first constant voltage during retention of a first data level at the storage node and disconnects the buffer node from the first constant voltage during retention of a second data level at the storage node. The read element retains the read bit line output at its present voltage level when a first data level is read from the gain cell and drives the read bit line output to a different voltage level when the opposite data level is read from the gain cell.
According to some embodiments of the invention, the read element decouples the storage node from the read bit line output during data read, connects the read bit line output to a second constant voltage when a data level associated with the second constant voltage is read from the gain cell and disconnects the read bit line output from the second constant voltage when the opposite data level is read from the gain cell.
According to some embodiments of the invention, the read element includes:
a first read switch comprising a first read data input, a first read control input and a first read data output; and
a second read switch comprising a second read data input, a second read control input and a second read data output.
The first read data input is connected to a second constant voltage level, the first read control input is connected to the storage node, the first read data output is connected to the second read data input, the second read control input is connected to the read trigger input and the second read data output is connected to the read bit line output.
According to some embodiments of the invention, the read element the retention element includes:
The buffer switch input is connected to the buffer node and to the feedback switch output, and the buffer switch output is connected to the storage node and to the feedback control input and the buffer control input is connected to the write line input. The buffer switch connects the buffer input to the buffer output when the write trigger is on and disconnects the buffer input from the buffer output when the write trigger is off. The feedback switch connects the feedback input to the feedback output when the feedback control input is at the first data level and disconnects the feedback input from the feedback output when the feedback control input is at the opposite data level.
According to some embodiments of the invention, the gain cell includes multiple transistors, and at least two of the transistors operate at different threshold voltages.
According to an aspect of some embodiments of the present invention there is provided a memory array which includes multiple gain cells as embodied herein. Respective write bit line inputs of the gain cells are connected to form a common write bit line, and respective read bit line outputs of the gain cells are connected to form a common read bit line output. At least some of the gain cells contain a respective: write bit line input, read bit line output, write trigger input, read trigger input, write element, retention element and read element. The gain cell's respective write element includes a first write switch with a first write data input, a first write control input and a first write data output and a second write switch with a second write data input, a second write control input and a second write data output. The first write switch and the second write switch are of complementary types. The first write data input and the second write data input are connected to the write bit line input; the first write control input and the second write control input are respectively connected to the write trigger input and a complementary write trigger line; and the first write data output and the second write data output are connected to an internal buffer node. The read element has a read bit line connection connected to the read bit line output, a read trigger connection connected to the read trigger input and a storage node connection connected to the storage node. The retention element buffers between the buffer node and an internal storage node during data retention, connects the buffer node to a first constant voltage during retention of a first data level at the storage node and disconnects the buffer node from the first constant voltage during retention of a second data level at the storage node. The read element retains the read bit line output at its present voltage level when a first data level is read from the gain cell and drives the read bit line output to a different voltage level when the opposite data level is read from the gain cell.
According to an aspect of some embodiments of the present invention there is provided a gain cell which includes:
A) a write bit line input;
B) a read bit line output;
C) a write trigger input;
D) a read trigger input;
E) a write transistor, which includes a first diffusion connection, a gate connection, and a second diffusion connection. The first diffusion connection is connected to the write bit line input and the gate connection is connected to the write trigger input;
F) a read transistor, which includes a first diffusion connection, a gate connection and a second diffusion connection. The first diffusion connection is connected to the read bit line output and the second diffusion connection is connected to the read trigger input; and
G) a retention element associated with the write transistor and the read transistor.
The retention element buffers between the second diffusion connection of the write transistor and the gate connection of the read transistor during data retention, connects the second diffusion connection of the write transistor to a constant voltage during retention of a first data level at the gate connection of the read transistor, and disconnects the second diffusion connection of the write transistor from the constant voltage during retention of a second data level at the gate connection of the read transistor.
According to some embodiments of the invention, the retention element connects the second diffusion connection of the write transistor to the gate connection of the read transistor during a write bit operation, and disconnects the second diffusion connection of the write transistor from the gate connection of the read transistor during data retention.
According to some embodiments of the invention, the retention element includes:
A) a buffer switch, having a buffer input, a buffer output and a buffer control input; and
B) a feedback switch associated with the buffer switch, having a feedback input, a feedback output and a feedback control input, wherein the feedback input is connected to a constant voltage.
The buffer input is connected to the second diffusion connection of the write transistor and to the feedback switch output. The buffer switch output is connected to the gate connection of the read transistor and to the feedback control input, and the write control input is connected to the write trigger input. The buffer switch connects the buffer input to the buffer output when the write trigger is on and disconnects the buffer input from the buffer output when the write trigger is off. The feedback switch connects the feedback input to the feedback output when the feedback control input is at the first data level and disconnects the feedback input from the feedback output when the feedback control input is at the second data level.
According to some embodiments of the invention, the retention element includes:
A) a buffer transistor, having a first diffusion connection, a gate connection connected to the write trigger input, and a second diffusion connection; and
B) a feedback transistor, having a first diffusion connection, a gate connection, and a second diffusion connection, wherein the first diffusion connection is connected to a constant voltage.
The first diffusion connection of the buffer transistor is connected to the second diffusion connection of the write transistor and to the second diffusion connection of the feedback transistor. The second diffusion connection of the buffer transistor is connected to the gate connection of the read transistor and to the gate connection of the feedback transistor.
According to some embodiments of the invention, the feedback transistor is a p-type transistor and the constant voltage is a low data voltage level.
According to some embodiments of the invention, the feedback transistor is an n-type transistor and the constant voltage is a high data voltage level.
According to some embodiments of the invention, the gain cell further includes a capacitor between the gate connection of the read transistor and ground.
According to an aspect of some embodiments of the present invention there is provided a memory array which includes a plurality of gain cells as embodied herein. Respective write bit lines inputs of the gain cells are connected to form a common write bit line, and respective read bit line outputs of the gain cells are connected to form a common read bit line output.
According to an aspect of some embodiments of the present invention there is provided a gain cell which includes:
A) a write bit line input;
B) a read bit line output;
C) a write trigger input;
D) a read trigger input;
E) a write transistor, comprising a first diffusion connection, a gate connection, and a second diffusion connection. The first diffusion connection is connected to the write bit line input and the gate connection is connected to the write trigger input;
F) a read transistor, comprising a first diffusion connection, a gate connection and a second diffusion connection, the first diffusion connection is connected to the read bit line output and the second diffusion connection is connected to the read trigger input;
G) a buffer transistor, having a first diffusion connection, a gate connection connected to the write trigger input, and a second diffusion connection; and
H) a feedback transistor, having a first diffusion connection, a gate connection, and a second diffusion connection, wherein the first diffusion connection is connected to a constant voltage.
The first diffusion connection of the buffer transistor is connected to the second diffusion connection of the write transistor and to the second diffusion connection of the feedback transistor. The second diffusion connection of the buffer transistor is connected to the gate connection of the read transistor and to the gate connection of the feedback transistor.
According to some embodiments of the invention, the write transistor, the buffer transistor the feedback transistor and the read transistor are p-type transistors.
According to some embodiments of the invention, the write transistor, the buffer transistor, the feedback transistor and the read transistor are n-type transistors.
According to some embodiments of the invention, the write transistor, the buffer transistor and the feedback transistor are p-type transistors and the read transistor is an n-type transistor.
According to some embodiments of the invention, the write transistor, the buffer transistor and the feedback transistor are n-type transistors and the read transistor is a p-type transistor.
According to an aspect of some embodiments of the present invention there is provided a memory array which includes a plurality of gain cells as embodied herein. Respective write bit line inputs of the gain cells are connected to form a common write bit line input, and respective read bit line outputs of the gain cells are connected to form a common read bit line output.
Unless otherwise defined, all technical and/or scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the invention pertains. Although methods and materials similar or equivalent to those described herein can be used in the practice or testing of embodiments of the invention, exemplary methods and/or materials are described below. In case of conflict, the patent specification, including definitions, will control. In addition, the materials, methods, and examples are illustrative only and are not intended to be necessarily limiting.
Implementation of the method and/or system of embodiments of the invention can involve performing or completing selected tasks manually, automatically, or a combination thereof. Moreover, according to actual instrumentation and equipment of embodiments of the method and/or system of the invention, several selected tasks could be implemented by hardware, by software or by firmware or by a combination thereof using an operating system.
For example, hardware for performing selected tasks according to embodiments of the invention could be implemented as a chip or a circuit. As software, selected tasks according to embodiments of the invention could be implemented as a plurality of software instructions being executed by a computer using any suitable operating system. In an exemplary embodiment of the invention, one or more tasks according to exemplary embodiments of method and/or system as described herein are performed by a data processor, such as a computing platform for executing a plurality of instructions. Optionally, the data processor includes a volatile memory for storing instructions and/or data and/or a non-volatile storage, for example, a magnetic hard-disk and/or removable media, for storing instructions and/or data. Optionally, a network connection is provided as well. A display and/or a user input device such as a keyboard or mouse are optionally provided as well.
Some embodiments of the invention are herein described, by way of example only, with reference to the accompanying drawings. With specific reference now to the drawings in detail, it is stressed that the particulars shown are by way of example and for purposes of illustrative discussion of embodiments of the invention. In this regard, the description taken with the drawings makes apparent to those skilled in the art how embodiments of the invention may be practiced.
In the drawings:
The present invention, in some embodiments thereof, relates to a gain cell with internal feedback and, more particularly, but not exclusively, to a five-transistor gain cell with internal feedback.
Embodiments herein present a gain cell that selectively protects a weaker data level by means of a feedback loop. The feedback loop decreases the required refresh frequency and reduces refresh power consumption.
Before explaining at least one embodiment of the invention in detail, it is to be understood that the invention is not necessarily limited in its application to the details of construction and the arrangement of the components and/or methods set forth in the following description and/or illustrated in the drawings and/or the Examples. The invention is capable of other embodiments or of being practiced or carried out in various ways.
Referring now to the drawings,
Reference is now made to
Reference is now made to
The most significant drawback of both the 2T and 3T cells is their short DRT in sub-100 nm technologies, which is caused by the ever-growing leakage currents at deeply scaled nodes, dominated by sub-threshold leakage from the SN to WBL. This results in decreased memory availability and significantly higher refresh power consumption, making it unsuitable for ULP applications.
Reference is now made to
Optionally, the read element is a two-switch (2-SW) read element, embodiments of which are presented below (e.g.
As used herein the term “2-SW read element” means a read element which is implemented by two switches (as described in more detail below). Exemplary embodiments of gain cells in which the 2-SW read element switches are transistors (denoted herein a 2T read element) are also described below (see
As used herein the term “2-SW write element” means a write element which is implemented by two switches (as described in more detail below). Exemplary embodiments of gain cells in which the 2-SW write element switches are transistors (denoted herein a 2T write element) are also described below (see
As shown in
Retention element 320 is connected between write element 310 and read element 330. Retention element 320 forms a buffer node (BN) within the gain cell, and provides a feedback effect which maintains the voltage level at BN when the “weak” data level is being retained at SN. The buffer node (BN) is formed between retention element 320 and the output of write element 310. Retention element 320 also connects to a constant voltage, which may be a low data level (or ground) or a high data level depending on the specific implementation as illustrated by the embodiments presented below.
Retention element 320 serves a dual purpose:
A) Retention element 320 connects and disconnects the buffer node (BN) and the storage node (SN), based on the state of WWL. During write operations, retention element 310 connects BN and SN so that the voltage levels at both nodes are equal. During voltage retention (e.g. standby), retention element 310 disconnects BN from SN.
B) Retention element 320 also provides a feedback mechanism which is controlled by the voltage level at SN during voltage retention. During one logic state (the “weak” state) the output of write element 310 is connected to the constant voltage, which slows the decay of the voltage level at BN. During the opposite logic state (the “strong” state) the output of write element 310 is disconnected from the constant voltage, and the voltage at BN decays relatively slowly due to self-limiting effects (similarly to the 2T gain cell).
Optionally, gain cell 300 further includes a capacitor between SN and ground.
2-SW Retention Element
Reference is now made to
Optionally, one or both of switching elements 410 and 420 are transistors, with the control signal input (e.g. WWL and/or SN level) being input to the respective transistor gate.
Read Element
Optionally, gain cell 300 includes a read element that decouples the storage node from the read bit line output during data read. The read element also connects the read bit line output to a constant voltage when one data level is read from the gain cell and disconnects the read bit line output from the constant voltage when the opposite data level is read from the gain cell. Thus, for the data level at which the read element connects RBL to the constant voltage, the voltage read at RLB is transferred from a constant voltage source rather than by a connection to SN. Otherwise, RLB is unconnected and remains pre-charged or pre-discharged according to the specific gain cell layout.
For example, for the read element of
Since, for the appropriate data level, RBL is decoupled from SN (e.g. by switch 413 in
Reference is now made to
Retention element 320 is connected between the write element output and a control input of 2-SW read element 331. A buffer node (BN) is formed between retention element 320 and the output of write element 305. Retention element 320 also connects to a constant voltage, which may be a low data level (or ground) or a high data level depending on the specific implementation as described below. SN and RWL are control inputs to the switches forming the 2-SW read element 331.
Reference is now made to
In order to avoid a limited swing on the RBL, read element 430 uses switch 433 to decouple SN from the RBL. For read, the RWL opens switch 433, enabling a current path from the constant voltage input into switch 432 to the RBL when the cell stores a first data level which closes switch 432 (e.g. ‘1’). This enables a full swing to the constant voltage on the RBL during read. When the cell stores the opposite data level (e.g. ‘0’), switch 432 is open and there is no current path to the constant voltage. RBL remains at its precharged (or pre-discharged) level. This reduces power consumption, as well as reducing the possibility of a read failure, especially when operated under a sub-VT supply voltage due to a very small RBL voltage swing.
Optionally, one or both of switching elements 440 and 450 are transistors (denoted herein a 2T read element). Exemplary embodiments of a gain cell with a 2T read element are described below (see
2-SW Write Element
Optionally, the write element is a two-switch element, formed of two complementary switches connected in parallel, and triggered by complementary write line signals (WWL and WWLcomp). The full-swing passing capability of the write element enables the propagation of strong levels to the SN without the need for a boosted word line. Read is performed by precharging the read bit line (RBL) and subsequently driving the read word line (RWL) to GND, thereby conditionally discharging the RBL capacitance if the SN is high (data ‘1’) or blocking the discharge path if the SN is low (data ‘0’).
Reference is now made to
Reference is now made to
Reference is now made to
Retention element 320 is connected between the write transistor D2 diffusion connection and the gate of read transistor 330. A buffer node (BN) is formed between retention element 320 and the write transistor D2 diffusion connection. Retention element 320 also connects to a constant voltage, which may be a low data level (or ground) or a high data level depending on the transistor type as described below.
Gain cell 301 further includes capacitor CSN 340, between SN and ground.
Embodiments of gain cells formed of four transistors (4T), five transistors (5T) and six transistors (6T) are presented below.
Optionally, all of the transistors in the gain cell are of the same type. In other embodiments, not all of the transistors in gain cell are of the same type, but rather each transistor is implemented in a respective type which may vary.
A non-limiting list of transistor types which may be included in a gain cell includes:
A) High-VT;
B) Low-VT;
C) Standard-VT; and
D) “IO” transistors.
Optionally, the gain cell includes different threshold transistors (for example a combination of low-VT transistors and regular-VT transistors).
Four-Transistor (4T) Gain Cell
Optionally, the read element, the write element and both switching elements forming the retention element (i.e. 322 and 323 of
Embodiments herein present a four-transistor gain cell that utilizes an internal feedback mechanism to significantly increase the data retention time in scaled CMOS technologies. The resulting gain cell displays a large reduction in retention power, with a reduction in bitcell area (as compared to a standard 6T SRAM).
Reference is now made to
In gain cell 500, WBL connects to a diffusion connection write transistor 510 (PW). Feedback transistor 522 (PF) and buffer transistor 521 (PB) together serve as a retention element 520 between storage node SN and write transistor 510 (PW). Gain cell 500 includes two nodes, buffer node BN (at the junction of PW 510, PB 521 and PF 522) and storage node SN (at the junction of PB 520, PF 522, CSN 540 and PR 530). Feedback transistor 522 (PF) conditionally discharges the BN node according to level at the SN node. Buffer transistor 522 separates the stored data level at SN from the BN level to ensure extended retention time. The gate of read transistor 530 (PR) connects to SN. Stored data is read at the read bit line output (RBL).
In the embodiment of a 4T gain cell of
A) Write transistor, feedback transistor, buffer transistor and read transistor are all p-type transistors;
B) Write transistor, feedback transistor, buffer transistor and read transistor are all n-type transistors (e.g.
C) Write transistor, feedback transistor, buffer transistor are p-type transistors, and the read transistor is n-type (e.g.
D) Write transistor, feedback transistor and buffer transistor are n-type transistors, and the read transistor is p-type (see
When the 4T gain cell includes n-type transistors and the voltage levels and cell operation are adapted to the transistor type, as known in the art.
Optionally, gain cell 500 includes four standard threshold-voltage (VT) transistors and is fully compatible with standard CMOS processes. In some embodiments, PMOS transistors are used. PMOS transistors have lower sub-VT and gate leakages relative to NMOS transistor, which may provide longer retention times while maintaining a small cell area. Detailed cell operation is explained hereafter.
In some embodiments, all of the transistors in gain cell 500 are of the same type. In other embodiments, not all of the transistors in gain cell 500 are of the same type, but rather each transistor is implemented in a respective type which may vary.
A non-limiting list of transistor types which may be included in gain cell 500 includes:
A) High-VT;
B) Low-VT;
C) Standard-VT; and
D) “IO” transistors.
Optionally, the gain cell includes different threshold transistors (for example a combination of low-VT transistors and regular-VT transistors as illustrated in
Reference is now made to
Reference is now made to
Reference is now made to
PMOS Four-Transistor (4T) Gain Cell
The operation of a 4T gain cell is described for a non-limiting all-PMOS gain cell configured as shown in
Cell access is achieved in a similar fashion as with the PMOS 2T gain cell of
The increased retention time of 4T gain cell 500 occurs during standby periods, when the internal feedback mechanisms come into play. During hold, PW 510 and PB 521 are off (WWL=VDD), and worst-case retention conditions are assumed, i.e., that WBL is driven to the opposite voltage of the stored data level. For a stored data ‘1’ (i.e. high data level), a self-limiting mechanism, similar to that of the 2T gain cell of
In summary, the feedback path protects the weak ‘0’ level on the SN by pulling BN to ground, while the worst-case VDD drop across PW 510 and the corresponding sub-VT leakage do not affect the retention time of the cell; the feedback path is disabled for the strong ‘1’ level. Note that the proposed technique does not fully avoid the decay of a ‘0’ level. Gate tunneling through PR 530, as well as the gate-induced drain leakage (GIDL) and junction leakage of PB 521, still charge SN, while sub-VT leakage of the turned-off PB 521 counteracts (but does not avoid) the SN charging process.
Five-Transistor (5T) Gain Cells
Optionally, the gain cell includes five transistors. The 5T gain cell may be obtained by two circuit configurations:
In all of the exemplary embodiments of
In
For write, the WWL is discharged to a negative voltage in order to allow passing a strong ‘0’ level to the cell and the value on WBL is transferred to SN. For read, the RBL is first pre-discharged to GND and RWL is driven to GND. When the gain cell holds ‘0’, a current path is enabled between VDD, connected at the diffusion of MS, and RBL, thereby charging RBL to VDD. On the other hand, when the cell holds ‘1’, MS is in cut-off and therefore RBL remains at ‘0’. The RBLs of each column are connected to a sense inverter, which outputs the logic level of the cell.
During hold, MW and MB are off (WWL=VDD). Assuming worst-case retention conditions, WBL is driven to the opposite voltage of the stored data level. For a stored ‘1’, a self-limiting mechanism ensures that the level decays slowly. In addition, the transistor stack (MW and MB) provides more resistance between SN and WBL and results in leakage and a slower decay compared to the conventional 2T cell. For data ‘1’, MF is in deep cutoff, such that its effect on the circuit is almost negligible. However, following a write ‘0’ operation, VSG of MF is equal to the voltage at BN (VBN). This is much higher than the negative VSG of MB, and therefore any charge that leaks through MW to BN will be discharged through MF and not degrade the ‘0’ level at SN. In this way, the worst-case condition of the 2T cell is eliminated and retention time is significantly increased. In summary, the feedback path protects the weak ‘0’ state on the SN by pulling BN to ground, while the worst-case VDD drop across MW and the corresponding leakage do not affect the retention time of the cell; the feedback path is disabled for the strong ‘1’ level.
In
In
In
Five-Transistor (5T) Gain Cells with 2T Write Element
In 2T write element 930, p-type transistor (PW) connected in parallel to n-type NW. The gates of PW and NW are connected respectively to the complementary word lines, WWLp and WWLn. A common write bit line (WBL) is used to drive data through write element 930 during write operations.
Six-Transistor (6T) Gain Cells
Memory Array
Reference is now made to
Memory array 1000 includes an array of gain cells 1010.1 to 1010.n, with respective write trigger inputs (WWL1 to WWLN) and read trigger outputs (RWL1 to RWLN). The write bit lines inputs (WBLs) of the gain cells in the array are connected together to form a common write bit line (CWBL). Similarly, the read bit line outputs (RBLs) of the gain cells are connected together to form a common read bit line output (CRBL). This creates a memory array with gain cells that may be written and read in unison. The type of gain cells forming the array may be any one or combination of 4T and/or 5T and/or 6T gain cells as described herein. Optionally, memory array 1000 includes one or more additional types of gain cells known in the art which are compatible with the other gain cells 4T and/or 5T and/or 6T gain cells in the array.
Embodiments herein present gain cells which may be used for scaled CMOS nodes characterized by high leakage currents, and which may be embedded in a GC-eDRAM. The gain cell design protects the weak data level by a conditional, cell-internal feedback path, while the feedback is disabled for the strong data level. Optionally, the read element decouples the storage node from the read bit line eliminating leakage currents through non-selected gain cells resulting in a full voltage swing at RBL. The gain cell embodiments require low retention power and improved worst case retention time. This is achieved with a small cell area relative to a 6T SRAM in the same technology, making the embodiments herein an appealing high-density, low-leakage alternative.
It is expected that during the life of a patent maturing from this application many relevant transistor types and transistor technologies will be developed and the scope of the term transistor is intended to include all such new technologies a priori.
The terms “comprises”, “comprising”, “includes”, “including”, “having” and their conjugates mean “including but not limited to”.
The term “consisting of” means “including and limited to”.
The term “consisting essentially of” means that the composition, method or structure may include additional ingredients, steps and/or parts, but only if the additional ingredients, steps and/or parts do not materially alter the basic and novel characteristics of the claimed composition, method or structure.
As used herein, the singular form “a”, “an” and “the” include plural references unless the context clearly dictates otherwise. For example, the term “a compound” or “at least one compound” may include a plurality of compounds, including mixtures thereof.
Throughout this application, various embodiments of this invention may be presented in a range format. It should be understood that the description in range format is merely for convenience and brevity and should not be construed as an inflexible limitation on the scope of the invention. Accordingly, the description of a range should be considered to have specifically disclosed all the possible subranges as well as individual numerical values within that range. For example, description of a range such as from 1 to 6 should be considered to have specifically disclosed subranges such as from 1 to 3, from 1 to 4, from 1 to 5, from 2 to 4, from 2 to 6, from 3 to 6 etc., as well as individual numbers within that range, for example, 1, 2, 3, 4, 5, and 6. This applies regardless of the breadth of the range.
Whenever a numerical range is indicated herein, it is meant to include any cited numeral (fractional or integral) within the indicated range. The phrases “ranging/ranges between” a first indicate number and a second indicate number and “ranging/ranges from” a first indicate number “to” a second indicate number are used herein interchangeably and are meant to include the first and second indicated numbers and all the fractional and integral numerals therebetween.
It is appreciated that certain features of the invention, which are, for clarity, described in the context of separate embodiments, may also be provided in combination in a single embodiment. Conversely, various features of the invention, which are, for brevity, described in the context of a single embodiment, may also be provided separately or in any suitable subcombination or as suitable in any other described embodiment of the invention. Certain features described in the context of various embodiments are not to be considered essential features of those embodiments, unless the embodiment is inoperative without those elements.
Various embodiments and aspects of the present invention as delineated hereinabove and as claimed in the claims section below find simulation support in the following examples.
Reference is now made to the following examples, which together with the above descriptions illustrate some embodiments of the invention in a non limiting fashion.
Simulation Results for 4T PMOS Gain Cell
The simulation results presented below are for a 64×32 bit (2 kb) memory macro based on the 4T PMOS gain cell 500 of
Reference is now made to
Initially, a ‘0’ is written to SN by pulsing WWL to a negative voltage (−700 mV), thereby discharging SN through WBL. Next, a read operation is performed by pre-discharging RBL by pulsing the PC control signal (not shown), and subsequently charging RWL. As required, RBL is driven high through PR. Prior to the next assertion of WWL, WBL is driven high in order to write a ‘1’ to SN. During the next read cycle, the pre-discharged RBL remains low, as the stored ‘1’ level blocks the discharge path through PR 530.
Reference is now made to
The estimated DRT, extracted from the results shown in
Chun et al. [ref. 3] previously showed that a standard 2T GC-eDRAM may exhibit lower retention power than a similarly sized SRAM in 65 nm CMOS. Since the retention time of the presented 4T PMOS gain cell is over 40 times higher than that of a standard 2T cell, the retention power (composed of leakage and refresh power) is even lower. For the simulated 4T PMOS gain cell memory macro, the retention power was found to be 3.86 pW/bit at 27 C and 53.78 pW/bit at 85 C. These retention powers are almost 5 times less than the leakage power of a 6T-bitcell SRAM operated at 0.7 V.
A comparison between the proposed cell and other embedded memories is presented in Table 1. Simplified cell structure diagrams of the 6T SRAM, 2T1C gain cell and 2T gain cell (referred to in Table 1) are shown in
Table 1 clearly emphasizes the benefits of the 4T PMOS gain cell, which achieves much lower power usage due to its increased retention time.
Performance of the proposed 4T cell is summarized in Table 2. At 700 mV, the active refresh energy is 6.89 fJ/bit, composed of 5.88 fJ/bit for read and 1.01 fJ/bit for write. The 4T PMOS gain cell has a read delay of 2.32 ns (using a slow but small sense inverter) and a write delay of 0.4 ns (with and underdrive of −700 mV). A conventional 2T gain-cell was measured to have a 0.29 ns write delay, which is the same order of magnitude as the proposed cell.
Reference is now made to
Reference is now made to
Reference is now made to
In summary, simulations of an exemplary embodiment of a four-transistor (4T) GC-eDRAM memory show a threefold increase in retention time, as compared to the best previously proposed gain cell in the same 65 nanometer node technology, or using the same Production Design Kit (PDK) for simulations and/or manufacturing [ref. 7]. The improved retention time results in a factor of ten decrease in retention power (static plus refresh power) as compared to the static power of a 65 nm 6T SRAM [ref. 8]. The improved performance is achieved with a gain cell that is 40% smaller than a 6T SRAM cell in the same technology. This enables the creation of a high density memory array with low power integration.
Simulation Results for 5T PMOS Gain Cell
Simulation results are now presented for the 5T gain cell illustrated in
The DRT of GC-eDRAM cells is often estimated using their data degradation curves, demonstrating the deterioration of a ‘0’ and ‘1’ levels stored in the SN, when the WBL is charged to the opposite value, resulting in the maximum leakage currents.
In
In contrast, in the 5T gain cell the degradation of a read ‘1’ operation (1510) and the degradation of a read ‘0’ operation (1520) are clearly distinguished.
To verify the readability of the 5T gain cell,
Nevertheless,
Table 3 compares the designed 5T gain cell with other ultra-low power embedded memory options, including an 8T SRAM cell, a 9T SRAM cell, a 10T SRAM cell and a 3T gain cell.
A conventional 6T SRAM gain cell was omitted from the comparison since it is not compatible with sub-VT operating voltages. Nonetheless, the area ratios of the bitcells compared to the area of a conventional 6T SRAM cell are given in the table. While the 8T and 10T bitcells provide increased noise margins and sub-VT operation, their bitcell sizes are 31% and 94% larger than the 6T cell, respectively. On the other hand, the 5T gain cell is only 8% larger than the conventional 6T cell and between 20%-45% smaller than the 8T and 10T bitcells. While the 3T gain cell offers the highest memory density, its DRT is only 6 μs, resulting in a retention power of 812 pW/bit, which is significantly higher than that of the other memory cells. On the other hand the retention time of the 5T gain cell is extended to 1 ms, resulting in the smallest retention power consumption of 11.5 pW/bit.
Although the invention has been described in conjunction with specific embodiments thereof, it is evident that many alternatives, modifications and variations will be apparent to those skilled in the art. Accordingly, it is intended to embrace all such alternatives, modifications and variations that fall within the spirit and broad scope of the appended claims.
All publications, patents and patent applications mentioned in this specification are herein incorporated in their entirety by reference into the specification, to the same extent as if each individual publication, patent or patent application was specifically and individually indicated to be incorporated herein by reference. In addition, citation or identification of any reference in this application shall not be construed as an admission that such reference is available as prior art to the present invention. To the extent that section headings are used, they should not be construed as necessarily limiting.
This application is a Continuation-in-Part (CIP) of U.S. patent application Ser. No. 15/306,796 filed on Oct. 26, 2016, now U.S. Pat. No. 9,691,445, which is a National Phase of PCT Patent Application No. PCT/IL2015/050452 having International filing date of Apr. 30, 2015, which claims the benefit of priority under 35 USC § 119(e) of U.S. Provisional Patent Application No. 61/986,915 filed on May 1, 2014. The contents of the above applications are all incorporated by reference as if fully set forth herein in their entirety.
Number | Name | Date | Kind |
---|---|---|---|
5255220 | Filliman | Oct 1993 | A |
6246083 | Noble | Jun 2001 | B1 |
6333892 | Hamamoto | Dec 2001 | B2 |
6747890 | Kirihata et al. | Jun 2004 | B1 |
6831866 | Kirihata | Dec 2004 | B1 |
6845059 | Wordeman et al. | Jan 2005 | B1 |
7120072 | Ye et al. | Oct 2006 | B2 |
8908418 | Yabuuchi | Dec 2014 | B2 |
9691445 | Giterman | Jun 2017 | B2 |
20110103137 | Beat | May 2011 | A1 |
20150003165 | Yamauchi | Jan 2015 | A1 |
20170062024 | Giterman et al. | Mar 2017 | A1 |
Number | Date | Country |
---|---|---|
WO 2005064615 | Jul 2005 | WO |
WO 2013105460 | Jul 2013 | WO |
WO 2015166500 | Nov 2015 | WO |
Entry |
---|
Supplementary European Search Report and the European Search Opinion dated Nov. 8, 2017 From the European Patent Office Re. Application No. 15786178.2. (12 Pages). |
International Preliminary Report on Patentability dated Nov. 10, 2016 From the International Bureau of WIPO Re. Application No. PCT/IL2015/050452. (5 Pages). |
International Search Report and the Written Opinion dated Aug. 12, 2015 From the International Searching Authority Re. Application No. PCT/IL2015/050452. |
Notice of Allowance dated Mar. 2, 2017 From the US Patent and Trademark Office Re. U.S. Appl. No. 15/306,796. (11 pages). |
Notice of Allowance dated Jan. 27, 2017 From the US Patent and Trademark Office Re. U.S. Appl. No. 15/306,796. (8 pages). |
Calhoun et al. “A 256-Kb 65-Nm Sub-Threshold SRAM Design for Ultra-Low-Voltage Operation”, IEEE Journal of Solid-State Circuits, 42(3): 680-688, Mar. 2007. |
Chang et al. “A 32 Kb 10T Sub-Threshold SRAM Array With Bit-Interleaving and Differential Read Scheme in 90 Nm CMOS”, IEEE Journal of Solid-State Circuits, 44(2): 650-658, Published Online Jan. 27, 2009. |
Chun et al. “A Sub-0.9V Logic-Compatible Embedded DRAM with Boosted 3T Gain Cell, Regulated Bit-line Write Scheme and PVT-Tracking Read Reference Bias”, Symposium on VLSI Circuits Digest of Technical Papers, pp. 134-135, 2009. |
Chun et al. “A 2T1C Embedded DRAM Macro with No Boosted Supplies Featuring a 7T SRAM Based Repair and a Cell Storage Monitor,” IEEE Journal of Solid-State Circuits, 47(10): 2517-2526, 2012. |
Chun et al. “A 667 MHz Logic-Compatible Embedded DRAM Featuring an Asymmetric 2T Gain Cell for High Speed On-Die Caches,” IEEE JSSC, 47(2): 547-559, 2012. |
Giterman et al. “4T Gain-Cell with Internal-Feedback for Ultra-Low Retention Power at Scaled CMOS Nodes”, IEEE International Symposium on Circuits and Systems (ISCAS): 2177-2180, Jun. 1-5, 2014. |
Giterman et al. “Single-Supply 3T Gain-Cell for Low-Voltage Low-Power Applications”, IEEE Transactions on Very Large Scale Integration (VLSI) Systems, 24(1): 358-362, Feb. 4, 2015. |
Hanson et al. “A Low-Voltage Processor for Sensing Applications With Picowatt Standby Mode”, IEEE Journal of Solid-State Circuits, 44(4): 1145-1154, Mar. 25, 2009. |
ITRS “International Technology Roadmap for Semiconductors—2012 Update”, 76 pages, 2012. Available online at www.itrs.net. |
Lee et al. “A 5.4nW/kB Retention Power Logic-Compatible Embedded DRAM with 2T Dual-VT Gain Cell for Low Power Sensing Applications,” IEEE Asian Solid-State Circuits Conference, Nov. 8-10, 2010, Beijing, China: 4 pages. |
Meinerzhagen et al. “Exploration of Sub-VT and Near-VT 2T Gain-Cell Memories for Ultra-Low Power Applications Under Technology Scaling,” Journal of Low Power Electronics and Applications, 3(2): 54-72, 2013. |
Somasekhar et al. “2 GHz 2 Mb 2T Gain Cell Memory Macro with 128 GBytes/sec Bandwidth in a 65 nm Logic Process Technology,” IEEE JSSC, 44(1): 174-185, 2009. |
Teman et al. “A 250 mV 8 Kb 40 Nm Ultra-Low Power 9T Supply Feedback SRAM (SF-SRAM)”, IEEE Journal of Solid-State Circuits, 46(11): 2713-2726, Sep. 1, 2011. |
Teman et al. “Review and Classification of Gain Cell eDRAM Implementations”, IEEE 27th Convention of Electrical and Electronics Engineers in Israel, Nov. 17, 2012. |
Zhang et al. “A 3-GHz 70Mb Sram in 65nm Cmos Technology with Integrated Column-Based Dynamic Power Supply,” IEEE Journal of Solid-State Circuits, 41(1): 146-151, 2006. |
Number | Date | Country | |
---|---|---|---|
20170294221 A1 | Oct 2017 | US |
Number | Date | Country | |
---|---|---|---|
61986915 | May 2014 | US |
Number | Date | Country | |
---|---|---|---|
Parent | 15306796 | US | |
Child | 15632555 | US |