Write-assisted SRAM bit cell

[0002] This invention facilitates efficient writing of data into a static random access memory (SRAM) bit cell, particularly if a logic 0 value stored in the bit cell is to be overwritten by a logic 1 value.

[0003] FIGS. 1 and 2 schematically depict a typical prior art single binary digit (i.e. 1-bit) SRAM bit cell incorporating cross-coupled inverters 10, 12. Inverter 10 is formed by PMOS “pull-up” transistor 14 and NMOS “pull-down” transistor 16. PMOS transistor 14's source is connected to a logic “high” voltage reference (V_DD), PMOS transistor 14's drain is connected in series with NMOS transistor 16's drain, NMOS transistor 16's source is connected to a logic “low” voltage reference (ground), and PMOS transistor 14's gate is connected to NMOS transistor 16's gate. Inverter 12 is formed by PMOS pull-up transistor 18 and NMOS pull-down transistor 20. PMOS transistor 18's source is connected to V_DD, PMOS transistor 18's drain is connected in series with NMOS transistor 20's drain, NMOS transistor 20's source is connected to ground, and PMOS transistor 18's gate is connected to NMOS transistor 20's gate. Inverters 10, 12 are cross-coupled by connecting the gates of PMOS transistor 14 and NMOS transistor 16 to the drains of PMOS transistor 18 and NMOS transistor 20 to define a first storage node S1; and, by connecting the gates of PMOS transistor 18 and NMOS transistor 20 to the drains of PMOS transistor 14 and NMOS transistor 16 to define a second storage node S2. The source-to-drain path of NMOS pass transistor 22 is connected between first storage node S1 and first bit line BIT, and the gate of NMOS transistor 22 is connected to word line WL. The source-to-drain path of NMOS pass transistor 24 is connected between second storage node S2 and second bit line {overscore (BIT)}, and the gate of NMOS transistor 24 is connected to word line WL.

[0004] Pass transistors 22, 24 are selectively turned on or off via word line WL to read or write data from the bit cell via bit lines BIT, {overscore (BIT)}. The cross-coupled structure of inverters 10, 12 ensures that logically opposite voltages are maintained at first and second storage nodes S1, S2 respectively. To read the single bit value stored in the bit cell, a logic high voltage signal is applied to word line WL, turning pass transistors 22, 24 on, thereby coupling nodes S1, S2 to bit lines BIT, {overscore (BIT)} respectively and allowing the bit cell to apply a differential voltage signal to bit lines BIT, {overscore (BIT)} which is in turn amplified by sense amplifiers (not shown). The sizes of transistors 14 through 24 are selected to accommodate writing of a single bit value into the bit cell by either pulling bit line BIT and node S1 low to drive node S2 high; or, pulling bit line {overscore (BIT)} and node S2 low to drive node S1 high when pass transistors 22, 24 are turned on. The cross coupled inverters latch the new data. Large numbers of such bit cells are combined to form memory arrays.

[0005] FIGS. 3 and 4 schematically depict another prior art SRAM bit cell incorporating cross-coupled inverters, as described in U.S. Pat. No. 5,754,468. The FIG. 3 bit cell is similar to the FIG. 1 bit cell, except:

[0006] NMOS pass transistor 24 is replaced with PMOS pass transistor 26;

[0007] word line WL is replaced with a dedicated write enable line and a complementary dedicated read enable line;

[0008] NMOS transistor 22 and PMOS transistor 26 are decoupled from one another—NMOS transistor 22's gate is instead connected to the write enable line and PMOS transistor 26's gate is connected to the complementary read enable line; and,

[0009] bit lines BIT, {overscore (BIT)} are replaced with single-ended (true or complement) write and read buses respectively.

[0010] In the FIGS. 3 and 4 embodiment, data is written into the bit cell by applying a logic high voltage signal to the write enable line, thereby turning NMOS transistor 22 on and coupling node S1 to the write bus. Data is read from the bit cell by applying a logic low voltage signal to the complementary read enable line, thereby turning PMOS transistor 26 on and coupling node S2 to the read bus, which is preferably pre-charged to a logic low state, since PMOS transistor 26 pulls up better than it pulls down.

[0011] The writing of a logic 0 value from the write bus through NMOS transistor 22 into the FIGS. 3 and 4 bit cell is a relatively “strong” event, in the sense that NMOS transistor 22 inherently pulls down to ground, so if a logic 1 value is already stored at node S1 that value is easily overwritten by the logic 0 value. However, the writing of a logic 1 value through NMOS transistor 22 into the FIGS. 3 and 4 bit cell is a relatively “weak” event, in the sense that if a logic 0 value is already stored at node S1, NMOS transistor 22 tends to shut off before the stored logic 0 is overwritten by the logic 1 value. Consequently, the operation of writing of a logic 1 value into the FIGS. 3 and 4 bit cell may fail or require an unacceptably long period of time. One way to compensate for this is to preferentially increase the sizes of the transistors in inverters 10, 12 so that a “weak” logic 1 value driven onto inverter 10's node S2 cross-couples through inverter 12 to complete the writing event at node S1. While this technique can be made to work, the drawback is that larger transistors consume additional integrated circuit silicon area and power. In a memory array consisting of many cells, the increased silicon area can be considerable.

[0012] This invention addresses the foregoing drawbacks of the FIGS. 3 and 4 bit cell, addresses further drawbacks inherent to single-ended bit line cells, and provides novel techniques for splitting rows of cells into 2 or more separate groups.

[0013] The invention allows efficient writing of data into an SRAM bit cell. In cells configured with separate read and write buses, whenever a write operation is performed, the read bus is forcibly held in a pre-charge state (pulled to ground) during the entire write operation. In cells configured with a common (single-ended) read/write bus, the pass transistor on the side of the cell opposite to the read/write bus is turned on during the entire write operation, reducing the voltage on that side of the cell such that the inverters cross-couple more quickly In either case, this “write assist” feature facilitates the writing of a logic 1 value into the cell without hindering writing of a logic 0 value into the cell. Bits cells formed in accordance with the invention can be replicated in various row arrangements to share write assist circuitry amongst pairs of cells while providing separate read/write mechanisms for each cell in a pair of cells.

[0014] FIG. 1 is a circuit schematic diagram of a first prior art SRAM bit cell.

[0015] FIG. 2 is a simplified schematic depiction of the FIG. 1 prior art SRAM bit cell.

[0016] FIG. 3 is a circuit schematic diagram of a second prior art SRAM bit cell.

[0017] FIG. 4 is a simplified schematic depiction of the FIG. 3 prior art SRAM bit cell.

[0018] FIG. 5 schematically depicts an SRAM bit cell having a read bus pre-charge transistor in accordance with the invention.

[0019] FIG. 6 is a logic diagram of a circuit for generating complementary read and write enable signals from a fully decoded memory read signal gated with a write strobe signal.

[0020] FIG. 7 is a logic diagram of a circuit for generating complementary read, write enable high and write enable low signals from a fully decoded memory read signal gated with a write strobe signal.

[0021] FIG. 8 schematically depicts a single-ended SRAM bit cell having a write-assist feature in accordance with an alternate embodiment of the invention.

[0022] FIG. 9 is a logic diagram of a circuit for generating read/write enable and write assist signals for the FIG. 8 bit cell.

[0023] FIG. 10 schematically depicts a pair of FIG. 8 type bit cells connected to share a common read/write bus and a common write-assist signal.

[0024] FIG. 11 schematically depicts a pair of FIG. 8 type bit cells connected to separate read/write buses and sharing a common write-assist signal.

[0025] FIG. 12 schematically depicts interleaved pairs of FIG. 8 type bit cells sharing a common write-assist signal.

[0026] Throughout the following description, specific details are set forth in order to provide a more thorough understanding of the invention. However, the invention may be practiced without these particulars. In other instances, well known elements have not been shown or described in detail to avoid unnecessarily obscuring the invention. Accordingly, the specification and drawings are to be regarded in an illustrative, rather than a restrictive, sense.

[0027] FIG. 5 depicts addition of read bus pre-charge NMOS transistor 28 to the FIGS. 3 and 4 bit cell. Several (i.e. 16-64, or more) SRAM cells are typically connected in parallel along the write and read buses. During a read operation, NMOS transistor 28 is first turned on by applying a logic high voltage signal to the pre-charge enable line, to initialize the read bus by pulling it down to ground. NMOS transistor 28 is then turned off for a brief period of time, allowing the read-selected cell (i.e. selected by turning PMOS transistor 26 on via the read enable line) to pull the read bus towards V_DDif a logic 1 value is being held at node S2 (or leave it dynamically at ground potential if a logic 0 value is being held at node S2).

[0028] Circuitry (not shown) can be provided to weakly hold the read bus at ground potential. Typically for example, the read bus is connected to sensing circuitry (not shown) having a weak feedback characteristic. If many bit cells are connected to the read bus, when pre-charge NMOS transistor 28 is not turned on leakage currents associated with PMOS transistors may develop, tending to pull the read bus up toward V_DD. Such leakage can be compensated for in well known fashion to hold the read bus at ground potential. Alternatively, the bit cell can be self-timed as explained below.

[0029] Thus, if the read bus is being pulled up towards V_DD, the voltage at node S2 is pulled down from V_DDby a small amount, ΔV, due to current flow through PMOS transistor 26. ΔV is made sufficiently small that the value stored in the bit cell is not disturbed.

[0030] Data is written into the FIG. 5 bit cell by turning NMOS transistor 22, PMOS transistor 26 and NMOS transistor 28 on (i.e. the write cycle is initiated simultaneously with the read cycle). If a logic 1 value is to be written into the bit cell no assistance is required if node S2 is initially at ground potential (i.e. if a logic 1 value is already stored in the bit cell). But, if node S2 is initially at V_DD(i.e. if a logic 0 value stored in the bit cell at node S1 is to be overwritten by a logic 1 value), the voltage at node S2 is again reduced by a small (non-constant) amount, ΔV, due to current flow through PMOS transistor 26. This slight voltage reduction at node S2 weakens inverter 12's NMOS transistor 20, such that if a “weak” logic 1 value is written into the bit cell via NMOS transistor 22, the cross-coupling of inverters 10, 12 completes more quickly. An impending transition of node S2 to ground potential is further sped-up because of its lower starting state. Modern integrated circuit fabrication techniques facilitate careful selection of a variety of PMOS transistor 26's operating characteristics (i.e. turn-on voltage threshold, width, etc.) thus facilitating careful control of the voltage at node S2.

[0031] It will be noted that, during a read cycle, the pre-charge enable signal is normally already on to initialize the read bus, then turned off for a brief period of time, then turned on again. The pre-charge enable signal remains on at all other times, unless another read cycle is initiated for the bit cell. Thus, the pre-charge enable signal remains on if a write cycle is initiated simultaneously with a read cycle, and is turned off only during a “pure read” cycle (i.e. a read cycle during which no simultaneous write cycle occurs). An important consequence of this pre-charge-assisted write technique is that the sizes of the transistors in inverters 10, 12 do not have to be increased in such a way that the memory cell area is also significantly increased.

[0032] Alternatively, instead of turning on both NMOS transistor 22 and PMOS transistor 26 to speed up the write operation, NMOS transistor 22 alone can be turned on and its voltage level boosted above V_DDby a small amount. However, this may damage NMOS transistor 22 over long periods of time. This technique also requires more power than the previously described preferred embodiment—which can also be self-timed as explained below. Voltage boosting also requires additional circuitry—it is traditionally used to improve the write performance of dynamic random access memory (DRAM) cells (see for example U.S. Pat. No. 5,103,113) or prior art SRAM cells utilizing single-ended bit line operation.

[0033] FIG. 6 depicts one way of producing properly synchronized, complementary read enable and write enable signals for operating the FIG. 5 bit cell. NOR gate 30 combines pre-decoded read_X and read_Y signals, producing the read enable signal at the output of NOR gate 30. Binary inverter 32 inverts the signal output by NOR gate 30, providing the desired complementary read enable signal. The read enable signal output by NOR gate 30 is also applied to one of AND gate 34's inputs. The “strobed” write enable signal is applied to AND gate 34's other input. The desired write enable signal is consequently output by AND gate 34 in synchronization with the complementary read enable signal output by binary inverter 32. This saves decoder logic in single port memory applications.

[0034] A plurality of FIG. 5 type bit cells can be configured to form a 2-part memory. The 2 parts are typically equal in size, and each part has a separate write enable line. One part is labeled “high” and the other part is labeled “low”. FIG. 7 shows how to produce properly synchronized, complementary read enable, and write enable high/write enable low signals for operating such a 2-part memory. NOR gate 36 combines pre-decoded read_X and read_Y signals, producing the read enable signal at the output of NOR gate 36. Binary inverter 38 inverts the signal output by NOR gate 36, providing the desired complementary read enable signal. The complementary read enable signal output by inverter 38 is applied to one input of each of NOR gates 40, 42. The “strobed” write enable high signal is applied to the other input of NOR gate 40; and, the “strobed” write enable low signal is applied to the other input of NOR gate 42. The desired write enable high and write enable low signals are consequently output by NOR gates 40, 42 respectively in synchronization with the complementary read enable signal output by inverter 38. Persons skilled in the art will realize that NOR gates 40, 42 can be fabricated to share one of their PMOS transistors, thereby possibly saving silicon area.

[0035] In a 2-part memory operated in accordance with FIG. 7, all bit lines are affected by a read, while writing options include: (i) write to both the high and low parts of the memory, (ii) write to only the high part of the memory, or (iii) write to only the low part of the memory. A prior art SRAM “column muxing” technique permits more bits (multiple words) than necessary to be read simultaneously from a memory block, followed by selection (column muxing) of only those bits (i.e. a single word) that are desired. Column muxing can be combined with a “write mask” technique to permit sub-portions of a data word to be modified. This invention does not permit column muxing, but the aforementioned write enable high/write enable low lines facilitate a form of write masking at the primary decoder level.

[0036] FIG. 8 depicts an alternative embodiment of the invention. Unlike the FIG. 5 embodiment, the FIG. 8 embodiment has neither a separate read bus nor a read bus pre-charge transistor. Instead, a dual-purpose (single-ended) read/write bus is provided, and PMOS transistor 26's source is grounded. In a single-ended bit cell like that of FIG. 8, NMOS transistor 22 tends to turn off if a logic 1 value is driven onto the read/write bus for storage in the bit cell. This allows the voltage at node S1 to rise, preventing the bit cell's inverters from crosscoupling to “flip” the cell from a logic 0 value to a logic 1 value.

[0037] Consequently, after an attempt to overwrite a previously stored logic 0 value with a logic 1 value, the bit cell may flip back from a logic 1 value to a logic 0 value after completion of the write cycle. The FIG. 8 embodiment circumvents this problem.

[0038] Specifically, during a write operation, in which it is desired to change the logic state at node S1 to a specific (high or low) logic level, the desired logic level is driven onto the read/write bus. PMOS transistor 26 is turned on during the write operation by applying a write assist signal to PMOS transistor 26's gate. As previously explained in relation to the FIG. 5 embodiment, if a logic 1 value is to be written into the bit cell no assistance is required if node S2 is initially at ground potential (i.e. if a logic 1 value is already stored in the bit cell). But, if node S2 is initially at V_DD(i.e. if a logic 0 value stored in the bit cell at node S1 is to be overwritten by a logic 1 value), current flow through PMOS transistor 26 reduces the voltage at node S2 such that the inverters cross-couple more quickly.

[0039] During a read operation, the read/write bus is initially pre-charged (through a gated pull-up transistor—not shown) to a voltage equal or approximately equal to V_DDand NMOS transistor 22 is subsequently turned on by applying a read/write enable signal to NMOS transistor 22's gate. Thereafter, the logic level on the read/write bus is determined by the logic level of node S1. That is, the read/write bus remains at V_DDif node S1 is at V_DD; or, the read/write bus is pulled down to ground if node S1 is at ground potential. As in the case of the FIG. 5 embodiment, turning on PMOS transistor 26 does not in itself cause any storage state changes in the cell.

[0040] An advantage of the FIG. 8 embodiment over the FIG. 5 embodiment is that it is unnecessary to boost the voltage of read/write enable signal above V_DDto speed up the write operation, because both NMOS transistor 22 and PMOS transistor 26 are turned on during the write operation.

[0041] FIG. 9 depicts one way of adapting the FIG. 6 circuit to produce properly synchronized, complementary read/write enable and write assist signals for operating the FIG. 8 bit cell. AND gate 44 combines pre-decoded read_X and read_Y signals, producing the read/write enable signal at the output of AND gate 44. The read/write enable signal output by AND gate 44 is also applied to one of NAND gate 46's inputs. The write “strobe” signal is applied to NAND gate 46's other input. The desired complementary write assist signal is output by NAND gate 46 in synchronization with the read/write enable signal output by AND gate 44. The FIG. 7 circuit can be similarly adapted to operate a plurality of FIG. 8 type bit cells configured to form a 2-part memory.

[0042] FIG. 10 depicts one way of replicating the FIG. 8 bit cell. A pair of FIG. 8 type bit cells 48, 50 are connected in parallel between a common read/write bus and a common write assist line. Reading or writing operations affecting bit cells 48, 50 respectively are controlled via mutually exclusive “read/write enable 0” and “read/write enable 1” lines. Thus, although signals applied to the write assist line affect both cells, only one cell can be written to at any particular time, as determined by the logic values applied to the mutually exclusive read/write enable lines. Similarly, only one cell can be read from at any particular time. Such shared usage of the write assist line simplifies circuit layout, thereby reducing silicon area consumption, at the expense of a small amount of power. The FIG. 9 circuit is readily adapted to produce properly synchronized, mutually exclusive read/write enable and write assist signals for operating the FIG. 10 bit cell.

[0043] More particularly, the FIG. 10 embodiment facilitates construction of a memory array having a plurality of pairs of bit cells 48, 50. Each pair of bit cells has first and second inverters cross-coupled to define first and second storage nodes, a first pass transistor coupled between the read/write bus and the first storage node, and a second pass transistor coupled between a logic low voltage reference signal source and the second storage node. The first pass transistor is coupled to the read/write enable 1 line for controllably switching the first pass transistor on or off. The second pass is transistor coupled to the write assist line for controllably switching the second pass transistor on or off. One of the first and second pass transistors (preferably the first) is an NMOS transistor and the other one of the first and second pass transistors (preferably the second) is a PMOS transistor. Each pair of bit cells also has third and fourth inverters cross-coupled to define third and fourth storage nodes, a third pass transistor coupled between the read/write bus and the third storage node, and a fourth pass transistor coupled between the logic low voltage reference signal source and the fourth storage node. The third pass transistor is coupled to the read/write enable 0 line for controllably switching the third pass transistor on or off. The fourth pass transistor is coupled to the write assist line for controllably switching the fourth pass transistor on or off. One of the third and fourth pass transistors (preferably the third) is an NMOS transistor and the other one of the third and fourth pass transistors (preferably the fourth) is a PMOS transistor.

[0044] FIG. 11 depicts a second way of replicating the FIG. 8 bit cell. A pair of FIG. 8 type bit cells 52, 54 sharing a common write assist line and a common read/write enable line are connected in series between separate read/write buses. Reading or writing operations affecting bit cells 52, 54 are controlled via the common read/write enable line. Although signals applied to the write assist or read/write enable lines affect both cells, the cells operate independently by virtue of their coupling to different read/write buses. Such shared usage of the write assist and read/write enable lines further simplifies circuit layout.

[0045] More particularly, the FIG. 11 embodiment facilitates construction of a memory array having a plurality of pairs of bit cells 52, 54. Each pair of bit cells has first and second inverters cross-coupled to define first and second storage nodes S1₁, S2₁; a first pass transistor coupled between read/write bus 1 and S1₁, and a second pass transistor coupled between a logic low voltage reference signal source and S2₁. The first pass transistor is coupled to the read/write enable line for controllably switching the first pass transistor on or off. The second pass transistor is coupled to the write assist line for controllably switching the second pass transistor on or off. One of the first and second pass transistors (preferably the first) is an NMOS transistor and the other one of the first and second pass transistors (preferably the second) is a PMOS transistor. Each pair of bit cells also has third and fourth inverters cross-coupled to define third and fourth storage nodes S2₀, S1₀; a third pass transistor coupled between read/write bus 0 and S1₀, a fourth pass transistor coupled between the logic low voltage reference signal source and S2₀. The third pass transistor is coupled to the read/write enable line for controllably switching the third pass transistor on or off. The fourth pass transistor is coupled to the write assist line for controllably switching the fourth pass transistor on or off. One of the third and fourth pass transistors (preferably the third) is an NMOS transistor and the other one of the third and fourth pass transistors (preferably the fourth) is a PMOS transistor.

[0046] FIG. 12 depicts a third way of replicating the FIG. 8 bit cell. Four FIG. 8 type bit cells 56, 58, 60, 62 sharing a common write assist line are connected in series to form a 2-part memory. Bit cells 56, 60 constitute the “high” part of the memory; bit cells 58, 62 constitute the “low” part of the memory. Reading or writing operations affecting bit cells 56, 60 are controlled via a common “read/write enable (high)” line. Reading or writing operations affecting bit cells 58, 62 are controlled via a common “read/write enable (low)” line which is separate from the read/write enable (high) line. Note that replication of FIG. 5 type bit cells to form a 2-part memory requires 3 separate control lines (i.e. high and low write enable lines and a read enable line) to be run over top of a possibly narrow cell pitch. By contrast, replication of FIG. 8 type bit cells to form a 2-part memory as shown in FIG. 12 requires only 2½ separate lines per cell, because each write assist line is shared by 2 bit cells (in a memory array consisting of many cells, each write assist line is shared by 2 columns of bit cells).

[0047] More particularly, the FIG. 12 embodiment facilitates construction of a memory array having a plurality of quadruplets of bit cells 56, 58, 60, 62. Each quadruplet has first and second inverters cross-coupled to define first and second storage nodes, a first pass transistor coupled between read/write bus 1 and the first storage node, and a second pass transistor coupled between a logic low voltage reference signal source and the second storage node. The first pass transistor is coupled to the read/write enable (high) line for controllably switching the first pass transistor on or off. The second pass transistor is coupled to the write assist line for controllably switching the second pass transistor on or off. One of the first and second pass transistors (preferably the first) is an NMOS transistor and the other one of the first and second pass transistors (preferably the second) is a PMOS transistor. Each quadruplet also has third and fourth inverters cross-coupled to define third and fourth storage nodes, a third pass transistor coupled between read/write bus 1 and the third storage node, and a fourth pass transistor coupled between the logic low voltage reference signal source and the fourth storage node. The third pass transistor is coupled to the read/write enable (low) line for controllably switching the third pass transistor on or off. The fourth pass transistor is coupled to the write assist line for controllably switching the fourth pass transistor on or off. One of the third and fourth pass transistors (preferably the third) is an NMOS transistor and the other one of the third and fourth pass transistors (preferably the fourth) is a PMOS transistor. Each quadruplet also has fifth and sixth inverters cross-coupled to define fifth and sixth storage nodes, a fifth pass transistor coupled between read/write bus 0 and the fifth storage node, and a sixth pass transistor coupled between the logic low voltage reference signal source and the sixth storage node. The fifth pass transistor is coupled to the read/write enable (high) line for controllably switching the fifth pass transistor on or off. The sixth pass transistor is coupled to the write assist line for controllably switching the sixth pass transistor on or off. One of the fifth and sixth pass transistors (preferably the fifth) is an NMOS transistor and the other one of the fifth and sixth pass transistors (preferably the sixth) is a PMOS transistor. Each quadruplet also has seventh and eighth inverters cross-coupled to define seventh and eighth storage nodes, a seventh pass transistor coupled between read/write bus 0 and the seventh storage node, and an eighth pass transistor coupled between the logic low voltage reference signal source and the eighth storage node. The seventh pass transistor is coupled to the read/write enable (low) line for controllably switching the seventh pass transistor on or off. The eighth pass transistor is coupled to the write assist line for controllably switching the eighth pass transistor on or off. One of the seventh and eighth pass transistors (preferably the seventh) is an NMOS transistor and the other one of the seventh and eighth pass transistors (preferably the eighth) is a PMOS transistor.

[0048] The FIGS. 11 and 12 embodiments provide a “double density” row arrangement of bit cells, in contrast to the FIG. 10 embodiment's “single density” row arrangement, since the FIG. 11 and 12 embodiments each accommodate twice as many bit cells per read/write bus as the FIG. 10 embodiment. This could be beneficial when larger arrays of cells are desired, and also helps reduce silicon area when the row width is not large (e.g. fewer than 32 bits). Another advantage of the FIG. 12 embodiment is that both reading and writing operations involve the same number of bits, whereas reading a 2-part memory formed of FIG. 5 type bit cells may involve twice as many bits as writing such a memory, (i.e. since writing involves only 1 of the memory's 2 parts).

[0049] As will be apparent to those skilled in the art in the light of the foregoing disclosure, many alterations and modifications are possible in the practice of this invention without departing from the spirit or scope thereof. For example, FIGS. 5 and 6 pertain to a single-port memory embodiment of the invention. Additional ports can be added by duplicating NMOS transistor 22 and PMOS transistor 26 and their respectively associated write and read buses. As another example, the FIGS. 6 and 7 circuits can be adapted to operate various embodiments of the invention. As a further example, the FIG. 8 write assist signal can additionally be applied to PMOS transistor PMOS 20's source terminal (instead of grounding that terminal), since the write assist signal is an “active low” signal. As a still further example, in the FIGS. 5 and 8 embodiments, either one of pass transistors 22, 26 may be an NMOS transistor, with the other one of those transistors being a PMOS transistor, although pass transistor 22 is preferably an NMOS transistor and pass transistor 26 is preferably an PMOS transistor. The scope of the invention is to be construed in accordance with the substance defined by the following claims.