Static random access memory device with a power dissipation reduction function.

0001The present invention relates to a static random access memory device, and more particularly to a static random access memory device with a power down function to reduce the consume current during a write cycle.

0002Generally, in a static random access memory device (SRAM), a DC current flow between the terminals of a power source of the device continues during a write cycle. It is for this reason that the consume power of the SRAM is great. How the DC current flows inside the SRAM will be described with reference to Fig. 1.

0003A circuit configuration of only one column unit of an ordinary SRAM, viz., a configuration concerning a pair of bit lines BL1 and <o ostyle="single">BL1</o>, is illustrated in Fig. 1. As shown, a memory cell MC1 is made up of a flip-flop including a couple of resistors R1 and R2 and a couple of N channel MOSFETs Q11 and Q12, and a couple of N channel MOSFETs Q13 and Q14 for data transfer. When data "0" is loaded into this memory cell MC1, viz., the potential at node A is placed at "L" level and the potential at node B is placed "H" level, data Din in "L" level is produced from a buffer 101, and data <o ostyle="single">Din</o> in "H" level, from a buffer 102. Under this condition, a word line WL1 and a column select line CSL1 are energized by a row decoder and a column decoder (both decoders are not shown), respectively. In turn, N channel MOSFETs Q1, Q2, Q13 and Q14 are turned on. Consequently, the node A is placed in "L" level by the data Din, while the node B, in "H" level by the data <o ostyle="single">Din</o>. In this way, data "0" is written into the memory cell MC1.

0004In the SRAM thus configured, the potential of "L" level at the node A turns off FET Q12 and the potential of "H" level at the node B turn on FET Q11. According­ly, a current I1 indicated by a broken line flows from a power source terminal VDD to a ground VSS terminal, through N channel MOSFETs Q9, Q13 and Q11. Another current I2 also flows from a power source terminal VDD to a ground VSS terminal, through N channel MOSFETs Q9 and Q1, and the MOSFET Q8 in the buffer 101. During a write cycle, the word line WL1 and the column select line CSL1 are kept in an active state, so that during this period, the flow of the currents I1 and I2 con­tinues. This leads to great consume current during the write cycle of the conventional SRAM.

0005Accordingly, an object of the present invention is to provide a static random access memory device capable of reducing the consume current during a write cycle.

0006According to the present invention, there is pro­vided a static random access memory device comprising: a memory cell array having a plurality of static memory cells; a pair of bit lines for transferring data to and from the memory cells in the memory cell array; a write circuit for supplying write data to the bit line pair; memory cell drive section for selectively driving one of the memory cells in accordance with a row address signal supplied from exterior; write mode detecting section for detecting a signal state transition of a write enable signal changing to an active state; input data tran­sition detecting section for detecting a transition of the input data supplied from exterior, during a con­tinuation of the active state of the write enable signal; address signal transition detecting section for detecting a transition of an externally applied address signal during an active state of the write enable signal; pulse signal generating section for generating a pulse signal with a predetermined pulse width in response to any of the detecting signals outputted from the write mode detecting section, the input data tran­sition detecting section, and the address signal tran­sition detecting section; drive control section for permitting the memory cell drive section to operate during a period that the pulse signal generating means generates a pulse signal, and for inhibiting the memory cell drive section from operating during a period that the pulse signal generating section rests; and write circuit control section for supplying write data based on the input data to the write circuit during a period that the pulse signal generating section generates a pulse signal, and for supplying to the write circuit initial data to set the bit line pair at the same poten­tials during the period that the pulse signal generating section rests.

0007In the static random access memory device thus arranged, the period to execute an actual write opera­tion is determined not by the period of a write cycle which is determined by an external control signal but by the period that the pulse signal generating means generates a pulse signal. The reduction of the consume current in a write mode of the SRAM is noticeable par­ticularly when the write cycle is long. Further, a plurality of memory cells of different addresses may be accessed for data write in a sequential order due to the address signal transition detecting section. In this case, the consume current during each write cycle will be reduced.

0008This invention can be more fully understood from the following detailed description when taken in con­junction with the accompanying drawings, in which: <ul id="ul0001" list-style="none"><li>Fig. 1 is a circuit diagram showing a part of a conventional static random access memory device;</li><li>Fig. 2 is a block diagram showing a static random access memory device according to an embodiment of the present invention;</li><li>Fig. 3 is a circuit diagram showing the details of a data transition detector used in the memory device of Fig. 2;</li><li>Fig. 4 is a circuit diagram showing the details of a write mode detector used in the memory device of Fig. 2;</li><li>Fig. 5 is a circuit diagram showing the details of a write power down timer used in the memory device of Fig. 2;</li><li>Figs. 6 to 8 are timing charts useful in explaining the operation of the memory device of Fig. 2; and</li><li>Fig. 9 is a graph showing a consume current vs. write cycle characteristic of the memory device of Fig. 2.</li></ul>

0009Referring to Fig. 2, there is shown a circuit con­figuration of a static random access memory device (SRAM) according to an embodiment of the present inven­tion. In Fig. 2, only one column unit associated with a pair of bit lines BL1 and <o ostyle="single">BL1</o> is illustrated for simpli­city. The SRAM is actually composed of a number of column units.

0010In Fig. 2, static memory cells MC1 to MCn are con­nected to the cross-points of paired bit lines BL1 and <o ostyle="single">BL1</o> and word lines WL1 to WLn. As shown, the memory cell MC1 is made up of a flip-flop including a couple of resistors R11 and R12, a couple of N channel MOSFETs Q21 and Q22, and a couple of N channel MOSFETs Q23 and Q24 for data transfer. The circuit configuration of each of the remaining memory cells is the same as that of the memory cell MC1. The paired bit lines BL1 and <o ostyle="single">BL1</o> are coupled at one side with a load circuit 10. The load circuit 10 is also coupled at the other side with a power source VDD. The circuit 10 includes normally-on transistors Q9 and Q10 as N channel MOSFETs.

0011The paired bit lines BL1 and <o ostyle="single">BL1</o> are coupled with a write circuit 17 through N channel MOSFETs Q1 and Q2. N channel MOSFETs Q1 and Q2 are used as switches which respond to an output signal of a column decoder 14 to place paired bit lines BL1 and <o ostyle="single">BL1</o> in a select mode or a nonselect mode. The gates of those FETs Q1 and Q2 are connected together to the column select line CSL1. The write circuit 17 for supplying write data to the paired bit lines BL1 and <o ostyle="single">BL1</o> is made up of a pair of buffers 101 and 102, and a pair of N channel MOSFETs Q3 and Q4 inserted respectively between the buffers 101 and 102 and the bit line pairs BL1 and <o ostyle="single">BL1</o>.

0012An address circuit 12 receives an address signal through an address input terminal 11 from outside, and applies a row address signal contained in the address signal to a row decoder 13, and a column address signal to a column decoder 14. The row decoder 13 decodes a row address signal, and produces a signal to select one of the word lines WL1 to WLn. The output signal of the row decoder 13 is transferred to the word lines WL1 to WLn by way of a gate G1. The gate G1 includes NOR gates 421 to 42n and inverters 431 to 43n. The column decoder 14 decodes a column address signal and produces a signal to select one of column select lines CSL to CSLm.

0013A write mode detector 31 is for detecting a level transition of a write enable signal <o ostyle="single">WE</o> going active, i.e., low "L" in level. The write enable signal <o ostyle="single">WE</o> and a chip enable signal <o ostyle="single">CE</o> are applied to NOR gate 41, from exterior. The NOR gate 41, after receiving these signals, produces a signal WE′, which in turn is applied to the write mode detector 31. During an active state of the chip enable signal <o ostyle="single">CE</o>, if the write enable signal <o ostyle="single">WE</o> changes its signal state from a nonactive state to an active state, the write mode detector 31 produces a one-shot pulse φwe as a write mode detect signal.

0014During an active state of the chip enable signal <o ostyle="single">WE</o>, if input data IN which is supplied to a data input circuit 19 from exterior, changes from "0" to "1" or "1" to "0", the data transition detector 32 produces a one-­shot pulse <o ostyle="single">φdt</o> as a data transition detect signal

0015The address signal transition detector 33 produces a one-shot pulse <o ostyle="single">φat</o> as an address signal transition detect signal, when either of a row address signal and a column address signal as supplied to the address input circuit 12, changes its signal state during an active period of the write enable signal <o ostyle="single">WE</o>.

0016The write mode detect signal φwe from the write mode detector 31, the data transition detect signal φdt from the data transition detector 32, and the address transition detect signal <o ostyle="single">φat</o> from the address signal transition detector 33 are supplied to a write power down timer 34.

0017The write power down timer 34, when receiving either the detect signal φwe, <o ostyle="single">φdt</o> or <o ostyle="single">φat</o>, produces simultaneously two signals <o ostyle="single">WE˝</o> and <o ostyle="single">φact</o> that are in "0" level during a predetermined period. Those pulses <o ostyle="single">WE˝</o> and <o ostyle="single">φwact</o> have the same pulse widths or the latter is slightly longer than the faster.

0018The pulse signal <o ostyle="single">WE˝</o> is for controlling the gating operation of the gate circuit G1 and the write circuit 17. The pulse signal <o ostyle="single">WE˝</o> is applied to the first input terminals of NOR gates 421 to 42n, which are respec­tively coupled with the word lines WL1 to WLn. The second input terminals of these NOR gates 421 to 42n are coupled for reception with the output signal of the row decoder 13, through inverters 431 to 43n. Accordingly, when the row decoder 13 produces a drive signal to drive the word line WL1, during a period that the pulse signal <o ostyle="single">WE˝</o> is generated, viz., during an "L" level period, the gate G1 allows the drive signal to go to the word line WL1. During a period that no pulse signal <o ostyle="single">WE˝</o> is gener­ated, viz., during an "H" level period, the gate G1 pro­hibits the drive signal from going to the word line WL1.

0019The pulse signal <o ostyle="single">WE˝</o> is also applied to the first input of a NOR gate 44, which is connected to the gates of the N channel MOSFETs Q3 and Q4 in the write circuit 17. The output signal WE′ of the NOR gate 41 is applied through an inverter 45 to the second input of the NOR gate 44. Accordingly, the write buffers 101 and 102 are allowed to be coupled with the paired bit lines BL1 and <o ostyle="single">BL1</o> only when the SRAM is in a write mode and during a period that the pulse signal <o ostyle="single">WE˝</o> is being generated. During a period that the pulse signal <o ostyle="single">WE˝</o> generation rests, the write buffers 101 and 102 are separated from the paired bit lines BL1 and <o ostyle="single">BL1</o>.

0020The pulse signal φwact, together with the output signal WE′ of the NOR gate 41, is supplied to the write control circuit 35. The write control circuit 35 controls a data input circuit 19 in accordance with these signals <o ostyle="single">φwact</o> and WE′. During a period that the pulse signal <o ostyle="single">φwact</o> is generated, viz., when it is in "L" level, the write control circuit 35 controls the data input circuit 19 so that the circuit 19 transfers the write data corresponding to the externally applied input data IN to the write circuit 17. When the input data IN is "0" in logical level, the output signals of the write buffers 101 and 102 in the write circuit 17 are set to "L" and "H" levels by the write control circuit 34 and the data input/output circuit 19, respectively. When the input data IN is "1", the write buffers 101 and 102 are set to "H" and "L" levels. During a period of generating no pulse signal <o ostyle="single">φwact</o>, that is, when the pulse signal is in "H" level, the write control circuit 34 controls the data input circuit 19 so that the circuit 19 transfers write initial setting data to the write circuit 17. The initial setting data places both the output signals of the buffers 101 and 102 in "H" level.

0021It is noted that in the SRAM of the present embodi­ment, an actual write operation period is determined by the pulse signals <o ostyle="single">WE˝</o> and <o ostyle="single">φwact</o>. Therefore, the periods that the DC currents I1 and I2 flow (see Fig. 1) are reduced. In other words, the power dissipation during the write cycle can be reduced. Additionally, when the address signal sequentially changes, the pulse signals <o ostyle="single">WE˝</o> and <o ostyle="single">φwact</o> are generated every time it varies. Therefore, it is possible to sequentially access a plurality of memory cells of different addresses and write data therein. Also in this case, the period of the actual data writing in each write cycle is limited, so that the power dissipation each write cycle can be reduced.

0022The data transition detector 32 may be formed by using inverters 51 to 54, NOR gates 55 and 56, and N channel MOSFETs 57 and 58, as shown in Fig. 3. When the input data changes its logical state from "1" to "0", a pulse signal that is set in "1" level during a period equal to a total sum of delay times by three inverters 51 to 53, is produced from the NOR gate 55, and applied to the gate of the FET 57. When it is changed from "0" to "1", the NOR gate 56 produces a pulse signal that is set in "1" level during a period equal to a total sum of delay times by three inverters 52 to 54, and applies it to the gate of the FET 58. At both the transient of the input data from "0" to "1" and "1" to "0", the detect signal <o ostyle="single">φdt</o> is maintained to "L" level during a predeter­mined period from the transition time.

0023The details of the write mode detector 31 is shown in Fig. 4. The detector 31 is made up of inverters 60 and 61, a delay circuit 62, and a NOR gate 63. The first input of the NOR gate 63 is coupled for reception with the output signal WE′ of the NOR gate 41 (Fig. 2) through the inverter 60, while the second input of the same receives the output signal WE′ through the inver­ters 60 and 61, and the delay circuit 62. Accordingly, when the signal WE′ changes from "L" to "H", the NOR gate 63 produces a detect signal φwe in the form of a pulse signal that is placed in "H" level during a period corresponding to the delay time by the delay circuit 62.

0024The address signal transition detector 33 may be constructed like the data transition detector shown in Fig. 3. In this case, an address signal is applied to the first input of the NOR gate 55, and the input of the inverter 51 in Fig. 3. The circuit of Fig. 3 is required for each bit of the address signal. Therefore, if the address signal consists of 8 bits, eight circuits of Fig. 3 must be used for the address signal transistor detector 33.

0025The details of the write power down timer 34 is shown in Fig. 5. A series circuit including a P channel MOSFET 71 and an N channel MOSFETs 72 and 73 is inserted between a power source VDD terminal and a ground VSS terminal. The FETs 71 and 72 are normally in an on state, and a node N1 of these transistors 71 and 72 is coupled for reception with a detect signal <o ostyle="single">φdt</o> supplied from the data transition detector 32 and a detect signal <o ostyle="single">φat</o> from the address signal transition detector 33. The gate of the FET 73 receives a detect signal φwe from the the write mode detect circuit 31. The potential at the node N1 continues an "L" level during a predetermined period when it receives the signal <o ostyle="single">φdt</o>, <o ostyle="single">φat</o> or φwe. The inverter 74 produces a pulse signal φD1 of a predeter­mined "H" level duration. This pulse signal φD1 is directly applied to the first input terminals of the NOR gates 75 and 78, and is applied to the second input terminals of the NOR gates 75 and 78 through a pulse elongation circuit 81. The pulse elongation circuit 81 functioning to elongate the pulse width of the pulse signal φD1 outputted from the inverter 74, may be constructed by the pulse generating circuit shown in Fig. 4. The output signals of the NOR gates 75 and 78 are placed in "L" level during a period corresponding to the pulse width obtained by the pulse elongation circuit 81. The output signal of the NOR gate 75 is supplied to the second input of a NOR gate 76 whose first input is connected to the ground VSS terminal. The output signal of this NOR gate 76 is supplied as a signal φwact, by way of the inverter 77. The output of the NOR gate 78 is coupled with the second input of a NOR gate 79 that is connected at the first input to the ground VSS ter­minal. The output signal of the NOR gate 79 is out­putted as a signal <o ostyle="single">WE˝</o>, by way of an inverter 80.

0026The write power down timer 34 thus arranged, when receiving either detect signal <o ostyle="single">φdt</o>, <o ostyle="single">φat</o> or φwe, produces a pulse signals <o ostyle="single">φwact</o> and <o ostyle="single">WE˝</o> each having a pulse width as obtained by the pulse elongation circuit 81.

0027The operation of the SRAM according to the present invention in a write mode will be described with refer­ence to Figs. 6 through 8. The timing charts shown in these figures are for the SRAM using the write mode detector 31, data transition detector 32, write power down timer 34, which are illustrated in Figs. 3 through 5, respectively.

0028A timing chart of Fig. 6 is obtained under con­ditions that an externally applied write enable signal <o ostyle="single">WE</o> is in an active state, i.e., in "0" level, and a logical state of input data IN changes from a "1" state to "0" state. When the input data IN changes its logi­cal state from "1" to "0", the data transition detector 32 detects the transition and produces a detect signal <o ostyle="single">φdt</o>. When this detect signal φdt is applied to the power down timer 34, the write power down timer 34 produces a pulse signal <o ostyle="single">φwact</o>, <o ostyle="single">WE˝</o> each having a prede­termined "L" duration. During the period of generating the pulse signal <o ostyle="single">WE˝</o>, the energization of the word line is allowed by the gate circuit G1. Hence, when the word line WL1 is selected, its potential rises as shown, and falls at the trailing edge of the pulse signal <o ostyle="single">WE˝</o>. During a period that the pulse signal <o ostyle="single">φwact</o>, <o ostyle="single">WE˝</o> is in "L", the write data corresponding to the input data IN is inputted to the write circuit 17, and the FETs Q3 and Q4 in the write circuit 17 are in an on state. During this period, therefore, the data lines Din and <o ostyle="single">Din</o> and the bit lines BL1 and <o ostyle="single">BL1</o> are set at a potential based on the the input data. Under this condition, the write data is written into the memory cell MC1. The write operation is executed only during the "L" level period of the pulse signal <o ostyle="single">φwact</o>, <o ostyle="single">WE˝</o>. When the pulse signal <o ostyle="single">φwact</o>, <o ostyle="single">WE˝</o> goes high ("H") in level, the energization of the word line WL1 is inhibited. At the same time, the supply of the write data is prohibited. No data write operation will be performed. Therefore, if the pulse width of the pulse signal <o ostyle="single">φwact</o>, <o ostyle="single">WE˝</o> is selected to be short as possible, under a condition that the data set­up time (Tds) proper to the SRAM is satisfied, the reduction of the DC currents can be realized without erroneous writing of data.

0029The timing chart shown in Fig. 7 is depicted when the input data is "1" and the write enable signal <o ostyle="single">WE</o> changes from "H" to "L", viz., changes to an active state. During a period that the chip enable signal <o ostyle="single">CE</o> is in "L" level, when the write enable signal <o ostyle="single">WE</o> becomes "L" in level, the output signal WE′ of the NOR gate 41 rises to "H" level. The write mode detector 31 detects the signal WE′ of "H" level and produces a detect signal φwe. When receiving the detect signal φwe, the write power down timer 34 produces a pulse signal <o ostyle="single">φwact</o>, <o ostyle="single">WE˝</o> each having a predetermined "L" duration. As described earlier, during the period of generating the pulse signal <o ostyle="single">φwact</o>, <o ostyle="single">WE˝</o>, the word line WL1 is driven and the write data based on the input data is supplied to the write circuit 17. When the input data IN is "1", the data lines Din and <o ostyle="single">Din</o> are respectively set at "H" and "L" levels, so that the bit lines BL1 and <o ostyle="single">BL1</o> are also set at "H" and "L", respectively. As a result, data "1" is written into the memory cell MC1. When the pulse signal <o ostyle="single">φwact</o>, <o ostyle="single">WE˝</o> goes high ("H") in level, the energi­zation of the word line WL1 is inhibited. At the same time, the supply of the write data is prohibited. No data write operation will be performed. Therefore, if the pulse width of the pulse signal <o ostyle="single">φwact</o>, <o ostyle="single">WE˝</o> is selected to be short as possible, under a condition that the data set-up time (Tds) proper to the SRAM is satis­fied, the reduction of the DC currents can be realized without erroneous writing of data.

0030Fig. 8 shows a timing chart describing such a tran­sient operation of the SRAM that the input data main­tains a logical state "1", and the address signal continues a certain state, and under this condition, the write enable signal <o ostyle="single">WE</o> becomes active in a signal state, and then the address signal changes its contents. When the write enable signal <o ostyle="single">WE</o> goes low "L" in logical state, a detect signal φwe derived from the write mode detector 31 causes the write power down timer 34 to pro­duce a pulse signal <o ostyle="single">φwact</o>, <o ostyle="single">WE˝</o> which is placed in "L" level during a predetermined period of time, as already stated. During the period of generating the pulse signal <o ostyle="single">φwact</o>, <o ostyle="single">WE˝</o>, the word line corresponding to the input address, for example, WL1, is driven by an output data of the row decoder 13, while at the same time the write data corresponding to the input data "1" is supplied to the write circuit 17. The paired data lines Din and <o ostyle="single">Din</o> are set in "H" and "L" level, respectively, and consequently the paired bit lines BL1 and <o ostyle="single">BL1</o> are also set in in "H" and "L" level, respectively. This writing operation is executed only when the pulse signal <o ostyle="single">φwact</o>, <o ostyle="single">WE˝</o> is in "L" level. When the pulse signal <o ostyle="single">φwact</o>, <o ostyle="single">WE˝</o> goes high ("H"), the word line is inhibited from being energized, while at the same time the supply of the write data is inhibited, and the writing opera­tion stops.

0031In a state that the write enable signal <o ostyle="single">WE</o> is in an active state of "L" level, when the address signal changes, the address signal transition detector 33 pro­duces a detect signal <o ostyle="single">φat</o>. In response to this signal, the write power down timer 34 produces a pulse signal <o ostyle="single">φwact</o>, <o ostyle="single">WE˝</o> which is placed in "L" level during a prede­termined period of time. Subsequently, the paired data lines Din and <o ostyle="single">Din</o> are set in "H" and "L" level, respec­tively, and then the paired bit lines BL1 and <o ostyle="single">BL1</o> are also set in "H" and "L" level, respectively. There­after, when the address signal changes, a sequence of the operations like the above one is repeated.

0032As described above, the provision of the address signal transition detector 33 enables control to access a plurality of memory cells of different addresses for write operation. In each write access, the period of an actual write operation is limited by the pulse signal <o ostyle="single">φwact</o>, <o ostyle="single">WE˝</o>. Accordingly, the power dissipation at the time of data writing operation is reduced.

0033As seen from the foregoing description, in the SRAM according to the present invention, the period of exe­cuting an actual write operation can be uniquely deter­mined within the SRAM circuit, not using an external control signal. As seen from Fig. 9, accordingly, the consume current of the SRAM according to the present invention is comparable with that of the conventional one with no power down function, when the writing opera­tion is at a high speed more than 25 MHz. In a low speed write operation less than 25 MHz, the consume current is considerably reduced when comparing with that of the conventional SRAM.

0034In the above-mentioned SRAM circuit, the control by the power down timer 34 is applied to the word line, and further the switch circuit (FETs Q3 and Q4) and the write buffers 101, 102 that are provided in the write circuit 17. The DC current saving may be realized by applying that control to the word line and the write buffers except the switch circuit (FETs Q3 and Q4). The reason for this is that when the output signals of the buffers 101 and 102 are in "H", no DC current flows even when the switch circuit (FETs Q3, Q4) is in an on state.

0035In this case, to prevent an erroneous data writing into the memory cell, it is desirable to time the cir­cuit operation so that the setting of the output data of the buffers 101 and 102 in "H" level follows the prohi­bition of the word line drive. To this end, the write power down timer 34 must produce a pulse signal <o ostyle="single">WE˝</o> of a predetermined pulse width and a pulse signal <o ostyle="single">φwact</o> of a pulse width longer than that of the former. This is realized with substituting of two pulse width elongation circuits producing signals of different pulse widths for the pulse width elongation circuit 81 in Fig. 5. Of the pulse width signals, a shorter pulse width signal is coupled with the NOR gate 78 (Fig. 5) and a longer pulse width signal, with the NOR gate 75.