Method and apparatus for writing operation in SRAM cells employing PFETS pass gates

[0002] The present invention relates generally to integrated circuit memory devices and, more particularly, to a method and apparatus for implementing a writing operation in Static Random Access Memory (SRAM) cells employing PFET pass gates.

[0003] A typical static random access memory (SRAM) cell includes an array of individual SRAM cells. Each SRAM cell is capable of storing a binary voltage value therein, which voltage value represents a logical data bit (e.g., “0” or “1”). One existing configuration for an SRAM cell includes a pair of cross-coupled devices such as inverters. With CMOS (complementary metal oxide semiconductor) technology, the inverters further include a pull-up PFET (p-channel) transistor connected to a complementary pull-down NFET (n-channel) transistor. The inverters, connected in a cross-coupled configuration, act as a latch which stores the data bit therein so long as power is supplied to the memory array. In a conventional six-transistor cell, a pair of access transistors or pass gates (when activated by a word line) selectively couple the inverters to a pair of complementary bit lines.

[0004] FIG. 1 illustrates a conventional SRAM cell structure 100 which represents a single memory cell included within a memory array arranged in rows and columns. The SRAM cell structure 100 includes a six-transistor memory cell 102 which is capable of storing a binary bit of information. Specifically, the memory cell includes a pair of cross-coupled, complementary metal oxide semiconductor (CMOS) inverters. One inverter includes an NFET storage transistor N1 and a PFET load transistor P1. Similarly, a second inverter includes an NFET storage transistor N2 and a PFET load transistor P2. Transistors P1 and P2 are often referred to as pull-up transistors because of their coupling to the voltage source V_DD. Transistors N1 and N2 are similarly referred to as pull-down transistors because of their coupling to ground. The memory cell 102 further contains NMOS access transistors (i.e., pass gates) NL and NR serving as switches, each of which are coupled between the bistable circuit (P1, N1, P2 and N2) and a pair of complementary bit lines BL and BR, respectively. Pass gates NL and NR are activated by an appropriate signal generated on a wordline WL.

[0005] While conventional SRAM cells typically employ NFETs for the pass gates, PFET pass gates have also been proposed for better stability, lower power and a substantially higher density for multi-port layouts. As a consequence of better stability, however, some difficulties may be experienced in writing through a PFET port. With decreased conductivity of smaller PFET pass gates, it takes longer for the node voltage within the cell to be pulled high or low. For a specific type of CMOS technology where the threshold voltage (V_T) is about the same for both PFETs and NFETs, a reverse beta ratio of at least 1.8 in the layout is desired in order to provide the SRAM cell with acceptable write performance under worst case conditions. The reverse beta ratio for a PFET pass gate is defined as the ratio of the (W/L) of the PFET pass gate over the (W/L) of the pull-down NFET. Moreover, a beta ratio for a conventional SRAM cell (the beta ratio being defined as the ratio of the (W/L) of the pull-down NFET over the (W/L) of the NFET pass gate) of at least 1.8 is needed for stability.

[0006] In an exemplary layout for an SRAM cell having PFET pass gates, a PFET pass gate may have a width W≈0.16 μm and a length L≈0.08 μm. For low power applications, a more desired PFET pass gate configuration would be to have devices with a narrower width and a longer channel (e.g., W≦0.11 μm and L≧0.10 μm) so that off currents in a standby mode may be minimized. However, if such PFET pass gate dimensions were so implemented, the pull-down NFET would be much longer in order to maintain a reverse beta ratio ≧2, and thus the corresponding overall cell size would be increased, perhaps by as much as 20%. In the meantime, the read performance of the cell would be degraded.

[0007] The above discussed and other drawbacks and deficiencies of the prior art are overcome or alleviated by a method for preparing a computer memory cell for a data write operation thereto. The memory cell has a cell supply voltage source which is connected at one end to pull-up devices within the memory cell, and is connected at an opposite end to pull-down devices within the memory cell. The memory cell further has a pair of access transistors for selectively coupling the memory cell to a pair of complementary bitlines. In an exemplary embodiment, the method includes adjusting the voltage of the cell supply voltage source from a first voltage value to a second voltage value, the second voltage value being less than the first voltage value. The memory cell is then coupled to the pair of complementary bitlines, thereby facilitating the data write operation. Following the data write operation, the cell supply voltage is restored from the second voltage value back to the first voltage value.

[0008] In one embodiment, the first voltage value corresponds to the pull-up devices within the memory cell being coupled to a logic supply voltage V_DD, and the pull-down devices within the memory cell being removably coupled to ground. The second voltage value corresponds to the pull-down devices being floated up to a potential above ground, thereby reducing the magnitude of the voltage of the cell supply voltage source. In an alternative embodiment, the first voltage value corresponds to the pull-up devices within the memory cell being removably coupled to a logic supply voltage V_DD, and the pull-down devices within the memory cell being coupled to ground. The second voltage value corresponds to the pull-up devices being floated down to a potential below V_DD, thereby reducing the magnitude of the voltage of the cell supply voltage source.

[0009] Referring to the exemplary drawings wherein like elements are numbered alike in the several Figures:

[0010] FIG. 1 is a schematic of an existing six-transistor SRAM cell structure;

[0011] FIG. 2 is a schematic of a partial SRAM cell array, illustrating various possible circuit configurations for implementing an SRAM write operation, in accordance with an embodiment of the invention;

[0012] FIG. 3 is a graph which illustrates a comparison of cell writing operations between a conventional NFET pass gate SRAM cell and a PFET pass gate SRAM cell configured according to an embodiment of the invention;

[0013] FIG. 4 is an alternative embodiment of the partial SRAM cell array shown in FIG. 2;

[0014] FIG. 5(a) is a schematic of a dual-port SRAM cell structure; and

[0015] FIG. 5(b) is an area layout of the dual-port SRAM cell structure of FIG. 5(a).

[0016] One possible solution to the above discussed drawbacks would be to implement a write scheme for a PFET pass gate SRAM cell having a reverse beta ratio of about 1. A “ratioless” SRAM cell layout may implemented, for example, by providing high voltage threshold (V_TN) implants for the NFET and/or low voltage threshold implants (V_TP) for the PFET, such that V_TNis over three times V_TP. Alternatively, the bias of the N-well and/or the P-well of the devices might be manipulated to produce a PFET pass gate which is twice as conductive as the pull-down NFET. However, such proposals result in lengthy process development and increased manufacturing costs.

[0017] Therefore, in accordance with an embodiment of the invention, there is disclosed a method and apparatus for implementing a write scheme for an SRAM cell having PFET pass gates, which write scheme allows for quick and reliable write performance within a ratioless layout. Thereby, lower off-currents for the pass gates, as well as higher cell density (particularly for multiport designs), are realized without sacrificing write performance.

[0018] Referring initially to FIG. 2, there is shown a partial SRAM cell array 200 which illustrates various possible circuit configurations for implementing the presently disclosed write schemes. Array 200 includes a plurality of individual SRAM cells 202 which are each capable of storing a bit of information therein. The SRAM cells 202 are similar to that shown in FIG. 1, with the exception that the pass gates in cells 202 comprise PFETs. For purposes of illustration only, cell array 200 shows nine individual cells 202 arranged in three rows and three columns. Each cell is connected to a wordline (WL1, WL2 or WL3), as well as a pair of complementary bitlines (BL1-BR1, BL2-BR2 or BL3-BR3). In order to access a particular cell, the particular wordline associated therewith is activated, and a read or write or operation may be executed.

[0019] In each of the following described writing scheme embodiments, the magnitude of the cell voltage supply is manipulated (i.e., reduced) in some manner from the normal logic supply voltage (e.g., V_DDand ground), such that a smaller PFET pass gate device size used in a ratioless layout will still allow for an effective write operation. Further, it will be noted that the following writing scheme embodiments are each described with reference to FIG. 2:

[0020] In this embodiment, the pull-down NFETs in each cell are floated upward from ground potential during a write operation thereto. Rather than connecting the pull-down NFETs in each cell directly to ground, they are connected to a common line V_L. In turn, V_Lis removably coupled to ground, with V_Lfloating above ground only during a write operation, and connected to ground at all other times. The pull-up PFETs within the cell remain connected to V_DD. This embodiment is shown implemented in the bottom row (along WL3) of the array 200 in FIG. 2.

[0021] More specifically, a switching mechanism 204 includes a NFET switch 206 connected between V_Land ground. In addition, another NFET 208 is configured as a diode, having both its drain terminal and gate terminal connected to V_L. During a non-writing mode, an input signal on the gate terminal of NFET switch 206 is equal to V_DD, thereby causing NFET switch 206 to be conductive and thus pulling V_Ldown to ground. The magnitude of cell voltage in this instance is equivalent to V_DD(e.g., 1.2 volts) However, during a write operation, the input signal goes to low or ground, thereby turning NFET switch 206 off. As a result, NFET 208, configured as a diode, causes the potential at V_Lto rise above ground by a voltage equivalent to the threshold voltage of the NFET device (e.g., 0.36 volts).

[0022] The effect of raising the potential at V_Lis to reduce the magnitude of the voltage swing needed (and hence the amount of time needed) within the cell node to flip the bit from “0” to “1” or vice versa. FIG. 3 is a graph which illustrates a comparison of cell writing operations between a conventional NFET pass gate SRAM cell (with a beta ratio of about 1.7) and a PFET pass gate SRAM cell configured according to embodiment (A) (with a reverse beta ratio of about 1.0). As seen from the graph, the transition time for flipping the bit for the conventionally configured cell with NFET pass gates is about 1.18 nanoseconds (ns). That is, the curves illustrating a write operation from logic 0 (at ground) to logic 1 (at V_DD), and vice versa, intersect at about 1.18 ns. The write operation is initiated with the wordline transition at about 0.11 ns, with the voltage on the wordline switching from V_DDat about 1.0 ns to ground at about 1.2 ns.

[0023] In contrast, an SRAM cell with PFET pass gates and a reverse beta ratio smaller than 1.7 becomes more difficult to write thereto, due to the relatively decreased conductivity of the pass gates as their dimensions get smaller. However, by floating the value of V_Lup above ground by about one V_T, the lesser the voltage swing is needed to flip the stored bit within the cell node. This is also shown in FIG. 3. By biasing the cell voltage such that a “0 write” operation pulls the node voltage from V_DDat about 1.2 volts down to V_Lat about 0.36 volts, the transition time is once again about 1.18 ns. While it is true that by manipulating (reducing) the magnitude of the cell voltage supply with respect to the logic supply decreases the stability of the cell, switching mechanism 204 ensures the “floating” operation occurs only during a write operation. Thus, during no operation or a read operation, the cell voltage supply is equivalent to the logic supply voltage, and adequate cell stability is maintained.

[0024] In an alternative embodiment, the pull-down NFETs in each cell are connected to V_Lwhich, in turn, may be biased at ground or above ground. In either case, the writing speed to the cell may be increased by pulling the bitline level below the value of V_L. This scheme is shown implemented for the cell occupying the first row, second column of the array 200 in FIG. 2 (WL1, BL2/BR2). As in scheme (A), the pull-up PFETs in the cell are connected to V_DD. The pull-down NFETs are connected to V_Lwhich, in this embodiment, is a fixed voltage. One possibility is to have V_Lset at a potential slightly higher than ground, such as about 0.36 volts (one V_T). The bitline voltage swing placed on bitlines BL2 and BR2 in this instance can therefore be between V_DDand ground.

[0025] Another possibility for this embodiment is to connect V_Lto ground. In this instance, the bitline swing would be between V_DDand a potential below ground, such as −0.36 volts. In either case, the bitline or its complement (depending on which value is written into the cell) is pulled approximately one threshold voltage value (V_T) below the value of V_L. This has the effect of changing the internal cell node voltage at a faster rate and thereby quickening the write time to the cell. With this embodiment, the particular bias setting of V_L(depending on the overall voltage budget) has its own advantages. For example, setting V_Lequal to ground renders the cell more stable than setting V_Lto one V_Tabove ground. On the other hand, setting V_Lequal to ground may require additional circuitry to set the bitlines at one V_Tbelow ground potential.

[0026] The principles of this embodiment are similar to embodiment (A). Instead of floating the pull-down NFETs in each cell upward from ground potential during a write operation, the pull-up PFETs are floated downward from V_DD. As shown in the second row (WL2) of the array 200 in FIG. 2, the pull-up PFETs in each cell are connected to a common line V_H. In turn, V_His removably coupled to V_DD, with V_Hfloating below ground only during a write operation, and connected to V_DDat all other times. The pull-down NFETs within the cell remain connected to ground.

[0027] A switching mechanism 210 includes a PFET switch 212 connected between V_Hand V_DD. In addition, another PFET 214 is configured as a diode, having both its drain terminal and gate terminal connected to V_H. During a non-writing mode, an input signal on the gate terminal of PFET switch 212 is equal to ground, thereby causing PFET switch 212 to be conductive and thus pulling V_Hup to V_DD. The magnitude of cell voltage in this instance is equivalent to V_DD(e.g., 1.2 volts) However, during a write operation, the input signal goes to high or V_DD, thereby turning PFET switch 212 off. As a result, PFET 214, configured as a diode, causes the potential at V_Hto drop below V_DDby a voltage equivalent to the threshold voltage of the PFET device (e.g., 0.36 volts).

[0028] Once again, the effect of lowering the potential at V_His to reduce the magnitude of the voltage swing needed (and hence the amount of time needed) within the cell node to flip the bit from “0” to “1” or vice versa.

[0029] The principles of this embodiment are similar to embodiment (B). The pull-up PFETs in each cell are connected to V_Hwhich, in turn, may be biased at V_DDor below V_DD. In either case, the writing speed to the cell may be increased by pulling the bitline charge above the value of V_H. This particular scheme is shown implemented for the cell occupying the first row, first column of the array 200 in FIG. 2 (WL1, BL1/BR1). As in scheme (C), the pull-down NFETs in the cell are connected to ground. The pull-up PFETs are connected to V_Hwhich, in this embodiment, is a fixed voltage. One possibility is to have V_Hset at a potential lower than V_DDsuch as about 0.84 volts (one V_Tbelow V_DD). The bitline voltage swing placed on bitlines BL1 and BR1 in this instance can therefore be between V_DDand ground.

[0030] Another possibility for this embodiment is to connect V_Hto V_DD. In this instance, the bitline swing would be between ground and a potential above V_DD, such as 1.56 volts. In either case, the bitline or its complement (depending on which value is written into the cell) is pulled approximately one threshold voltage value (V_T) above the value of V_H. This has the effect of changing the internal cell node voltage at a faster rate and thereby quickening the write time to the cell. With this embodiment, the particular bias setting of V_H(depending on the overall voltage budget) has its own advantages. For example, setting V_Hequal to V_DDrenders the cell more stable than setting V_Hto one V_Tbelow V_DD. On the other hand, setting V_Lequal to V_DDmay require additional circuitry to set the bitlines at one V_Tabove V_DDpotential. Furthermore, with embodiments (C) and (D), additional design care is desirable. In (D), there is a short path between the up bitline voltage level at V_H+V_Tand the cell supply level at V_H, after writing is complete. In the case of embodiment (C), there is a short path between the down bitline voltage level at V_L−V_T, and the cell supply at V_L.

[0031] Referring now to FIG. 4, there is shown another partial SRAM cell array 400 which illustrates an alternative cell power supply arrangement to the array 200 of FIG. 2. In this embodiment, the cell supply lines (V_H, V_Lor both) are arranged vertically along cell columns instead of rows. The choice of the arrangement of cell voltage supply lines may depend upon the availability of metal lines, the specific application of the memory array and/or the designer's preferred style.

[0032] Finally, the above described cell arrangements may also be applied in dual-port SRAM applications. A schematic for an individual, dual-port SRAM cell is shown in FIG. 5(a). Cell 500 comprises an eight transistor cell, wherein there are two sets of pass gates, two wordlines and two pairs of complementary bitlines. Both sets of pass gates, 5 wordlines and bitlines provide independent access to the cell 500. In a dual-port SRAM, one port may be used for read or write operations, and the other port may also be used for read or write operations. For example, in the cell 500, a PFET read port comprises PFET pass gates PL and PR, activated by wordline WLP, which couples cell 500 to bitline pair BLP and BRP. An NFET write port comprises NFET pass gates NL and NR, activated by wordline WLN, which couples cell 500 to bitline pair BLN and BRN.

[0033] In order to enable the PFET port to allow a reliable writing operation, the write acceleration schemes described herein are applied on one form or another. Otherwise, the PFET pass gate would necessarily be made much larger to provide a sufficiently large reverse beta ratio. Consequently, with bigger PFET pass gates, the pull-down NFETs increase in size to prevent access disturb, all at the cost of larger cell area and power dissipation. As shown in FIG. 5(b), dual-port cell 500 has an area layout of about 1.12 μm×1.54 μm≈1.72 μm²(as compared to a conventional, single-port NFET cell with an area of about 1.21 μm²). The cell size is thus minimized without the concerns of reverse beta ratios.

[0034] It will be appreciated that the disclosed embodiments may be combined such both the high and low voltage supply connections within a cell are manipulated below V_DDand above ground, respectively. For example, both embodiments (A) and (C) may be combined such that a write operation to a particular cell causes the power supply thereto to be removably floated from V_DDand ground to V_Hand V_L. Such a controlled “disturbance” is particularly suited to secure SRAM designs, where the cell data is almost non-disturbable. Likewise, a combination of embodiments (B) and (D) allows the cell voltage supply to be fixed at V_Hand V_L, while the bitline swing is from V_DDto ground. Alternatively, the cell voltage may be fixed at V_DDand ground while the bitlines are charged to levels higher than V_DDand lower than ground.

[0035] Through the use of the disclosed method and apparatus embodiments, an SRAM array may take advantage of the benefits of having smaller PFET pass gates (e.g., better stability, lower power, higher density) while not sacrificing write performance. In addition, the reliable write performance is maintained even when the reverse beta ratio of the cell is less than 1.7.

[0036] While the invention has been described with reference to a preferred embodiment, it will be understood by those skilled in the art that various changes may be made and equivalents may be substituted for elements thereof without departing from the scope of the invention. In addition, many modifications may be made to adapt a particular situation or material to the teachings of the invention without departing from the essential scope thereof. Therefore, it is intended that the invention not be limited to the particular embodiment disclosed as the best mode contemplated for carrying out this invention, but that the invention will include all embodiments falling within the scope of the appended claims.