JP5990306B2 - Semiconductor device - Google Patents

Description

  The present invention relates to a semiconductor device, and more particularly to a technique effective when applied to a semiconductor device including a memory unit such as an SRAM.

  For example, Patent Document 1 shows a configuration in which the voltage level of a cell power supply line is reduced when data is written in an SRAM. Japanese Patent Application Laid-Open No. 2004-228561 shows a configuration for reducing the voltage level of a selected word line in an SRAM. In Patent Document 3, in the SRAM, the power supply voltage level of the memory cell is supplied to the power supply node of the word line driver when the word line rises, and the power supply voltage level of the memory cell is supplied to the power supply node of the word line driver after the word line rises. A configuration for providing a lower voltage level is shown.

JP 2007-4960 A JP 2009-252256 A JP 2008-210443 A

  For example, in a semiconductor device on which an SRAM (Static Random Access Memory) memory module (static memory module) or the like is mounted, voltage scaling is usually performed from the viewpoint of reliability, power consumption, and the like with miniaturization. However, as miniaturization progresses, a decrease in the operating margin of the SRAM memory cell becomes a problem with an increase in manufacturing variation. For this reason, various devices for ensuring a stable operation margin even at a low voltage are required.

  FIG. 24A is an explanatory diagram showing a schematic configuration example and an operation example of the main part of the static memory module in the semiconductor device studied as a premise of the present invention, and FIG. It is explanatory drawing which shows the schematic structural example and operation example different from a). The static memory module shown in FIGS. 24A and 24B includes a memory array MARY, a word driver block WLD, and a write assist circuit WAST ′. In MARY, a word line WL driven by WLD and extending in the X-axis direction, a memory cell (SRAM memory cell) MC selected by the WL, and driven by WAST ′, extending in the Y-axis direction, A memory cell power supply line for supplying the memory cell power supply voltage ARVDD to the MC is included. WAST 'has a function of reducing the ARVDD of the selected memory cell MC for a predetermined period during a write operation. As a result, in the selected MC, the information retention capability (latch capability) is reduced (in other words, the static noise margin (SNM) is reduced), and as a result, the MC can be easily rewritten (the write margin is improved). .

  Here, in the MARY shown in FIG. 24A, the Y-axis direction (the extending direction of the memory cell power supply line (ARVDD) or the extending direction of the bit line not shown) is the vertical direction, and the X-axis direction (the extending direction of WL). It has a horizontally long shape in the horizontal direction. On the other hand, the MARY shown in FIG. 24 (b) has a vertically long shape unlike the case of FIG. 24 (a). Assume that WAST 'is designed so that the voltage level of ARVDD is lowered under the optimum conditions with respect to MARY shown in FIG. In this case, if the WAST ′ is applied to the MARY shown in FIG. 24B, the load of the memory cell power supply line (ARVDD) is larger in the MARY of FIG. 24B than in FIG. It may take a certain period of time for the voltage level to reach the desired level. During this period, the selected MC has a relatively high information holding capability (latching capability) (in other words, a large SNM), and as a result, the MC may not be easily rewritten (the write margin is reduced). is there.

  FIG. 25A is an explanatory diagram showing a schematic configuration example and an operation example of the main part of the static memory module in the semiconductor device studied as a premise of the present invention, and FIG. It is explanatory drawing which shows the schematic structural example and operation example different from a). The static memory module shown in FIGS. 25A and 25B includes a memory array MARY, a word driver block WLD, and a word driver power supply circuit block VGEN ′. In MARY, a word line WL driven by WLD and extending in the X-axis direction, a memory cell (SRAM memory cell) MC selected by the WL, and VGEN ′ are driven and extended in the Y-axis direction. A word driver power supply line for supplying the word driver power supply voltage WLVDD to each word driver in the WLD is included.

  VGEN 'has a function of lowering the power supply voltage WLVDD of the WLD (the word driver therein) for a predetermined period when the predetermined WL is activated by the WLD. As a result, the information holding capability (latch capability) of the MC holding information on the WL can be improved, and the read margin can be improved. That is, since the drive capability of the access NMOS transistor in the SRAM memory cell is equivalently reduced, the so-called β ratio, which is the ratio of the drive capability of the driver NMOS transistor and the access NMOS transistor in the SRAM memory cell, increases. The static noise margin (SNM) can be improved.

  Here, in MARY shown in FIG. 25A, the Y-axis direction (the extending direction of the word driver power supply line (WLVDD) or the extending direction of the bit line (not shown)) is the vertical direction, and the X-axis direction (the extending direction of WL). It has a horizontally long shape in the horizontal direction. On the other hand, the MARY shown in FIG. 25 (b) has a vertically long shape unlike the case of FIG. 25 (a). In MARY of FIG. 25A, the load level of the word driver power supply line (WLVDD) is smaller than that of FIG. 25B, so that the voltage level of WLVDD rapidly decreases. As a result, there may be a case where the voltage level of WLVDD required for starting WL at high speed cannot be sufficiently secured. In addition, MARY in FIG. 25A has a larger WL load than in FIG. 25B, so it is difficult to increase the rising speed. The synergistic effect with WLVDD described above increases the rising speed. May become even more difficult. Therefore, in the MARY shown in FIG. 25A, there may be a case where the required access time cannot be sufficiently secured due to the delay of the rising speed of WL.

  On the other hand, the MARY of FIG. 25B has a larger load on the word driver power supply line (WLVDD) than that of FIG. 25A, so that the voltage level of WLVDD decreases to a certain level. It may take a long time. During this period, the voltage level of the selected WL becomes relatively high, and the MC on the WL has a relatively low static noise margin (SNM). As a result, the read margin of the MC may be reduced. is there. In addition, the MARY of FIG. 25B has a smaller WL load than that of FIG. 25A, so that its rising speed is likely to increase (that is, the WL voltage tends to increase due to overshoot). The synergistic effect with WLVDD may further accelerate the decrease in static noise margin (SNM). Therefore, in the case of MARY in FIG. 25B, there may be a case where a sufficient read margin cannot be secured.

  As described above, when memory arrays having different shapes such as a vertically long shape and a horizontally long shape exist in the semiconductor device, an operation margin (read margin, write margin) may be reduced or an access time may be delayed depending on the shape. In particular, when a compiled SRAM or the like is mounted on a semiconductor device such as an SOC (System On a Chip), many SRAM memory modules having different shapes are mounted on the SOC in accordance with product requirements from the market. There is. In such a case, it has been found that the problem of the operation margin and the access time as described above can be particularly manifested. The compiled SRAM is a macro cell that is automatically generated by an automatic design tool by specifying the number of bit lines and word lines, for example. In a compiled SRAM, the memory array and its peripheral circuits are usually automatically arranged by the automatic design tool according to the number of bit lines / word lines according to the layout of various element circuits (for example, word drivers, etc.) that are defined in advance. To be built.

  The present invention has been made in view of the above, and an object of the present invention is to realize an improvement in the operation margin of a semiconductor device including a plurality of static memory modules. The above and other objects and novel features of the present invention will be apparent from the description of this specification and the accompanying drawings.

  Of the inventions disclosed in the present application, the outline of a typical embodiment will be briefly described as follows.

  The semiconductor device according to the present embodiment includes a first memory module and a second memory module. The first memory module extends in the same direction as the first bit line in addition to the plurality of first word lines, the plurality of first bit lines, and the plurality of first SRAM memory cells arranged at the intersections thereof. A plurality of first memory cell power supply lines for supplying power to the first SRAM memory cell, and a first write assist circuit. The first write auxiliary circuit discharges the charge of the first memory cell power supply line corresponding to the first SRAM memory cell to be written in the first period during the write operation. Similarly, the second memory module extends in the same direction as the second bit lines in addition to the plurality of second word lines, the plurality of second bit lines, and the plurality of second SRAM memory cells arranged at the intersections thereof. And a plurality of second memory cell power supply lines for supplying power to the plurality of second SRAM memory cells, and a second write assist circuit. The second write auxiliary circuit discharges the charge of the second memory cell power supply line corresponding to the second SRAM memory cell to be written in the second period during the write operation. Here, when the number of the plurality of first word lines is larger than the number of the plurality of second word lines, the first period is set longer than the second period.

  When such a configuration is used, the power supply voltage of the SRAM memory cell to be written can be lowered during the write operation, so that the write margin can be improved. In addition, since the power supply voltage decrease rate (and the decrease width) can be controlled by the first and second periods, the write margin can be improved regardless of the number of word lines in each memory module.

  In the semiconductor device according to the present embodiment, the first memory module as described above includes a plurality of first word drivers, a first word driver power supply line, and a first power supply circuit block. The second memory module includes a plurality of second word drivers, a second word driver power supply line, and a second power supply circuit block. The plurality of first word drivers drive the plurality of first word lines, and the first word driver power supply line extends in the same direction as the first bit line and supplies power to the plurality of first word drivers. The first power supply circuit block lowers the voltage level of the first word driver power supply line with the first drive capability during the read operation. Similarly, the plurality of second word drivers drive the plurality of second word lines, and the second word driver power supply line extends in the same direction as the second bit line and supplies power to the plurality of second word drivers. . The second power supply circuit block lowers the voltage level of the second word driver power supply line with the second drive capability during the read operation. Here, the first driving capability and the second driving capability are set to be larger as the number of word lines is larger and as the number of bit lines is smaller. That is, when the number of the plurality of first word lines and the number of the plurality of second word lines are substantially equal and the number of the plurality of first bit lines is larger than the number of the plurality of second bit lines, One driving capability is set smaller than the second driving capability. Further, when the number of the plurality of first bit lines and the number of the plurality of second bit lines are substantially equal, and the number of the plurality of first word lines is larger than the number of the plurality of second word lines, One driving capability is set larger than the second driving capability.

  When such a configuration is used, the voltage level of the selected word line can be lowered during the read operation, so that the read margin of the SRAM memory cell on the word line can be improved. In addition, since the voltage level drop rate (and width) can be controlled by the first drive capability and the second drive capability, the prescribed access performance is maintained regardless of the number of word lines and bit lines of each memory module. In this state, the reading margin can be improved.

  The effects obtained by the representative embodiments of the invention disclosed in this application will be briefly described. In the semiconductor device including a plurality of static memory modules, the operation margin can be improved.

1 is a block diagram illustrating a schematic configuration example of a static memory module included in a semiconductor device according to a first embodiment of the present invention. FIG. FIG. 2 is a circuit diagram illustrating a configuration example of each memory cell in the memory module of FIG. 1. FIG. 2 is a waveform diagram illustrating a schematic operation example of the memory module of FIG. 1. 1 is a block diagram showing an example of the overall schematic configuration of a semiconductor device according to a first embodiment of the present invention. FIG. 2 is a schematic diagram illustrating an example of functions around the write assist circuit in the memory module of FIG. 1. FIG. 6 is a schematic diagram showing an example of an effect when the write assist circuit of FIG. 5 is applied to each static memory module in the memory unit of FIG. 4. (A) is a circuit diagram showing a detailed configuration example of the write assist circuit in FIG. 5, (b) is a circuit diagram showing a configuration example different from (a). FIG. 8 is a waveform diagram showing an operation example of the write assist circuit in FIGS. 7 (a) and 7 (b). FIG. 6 is a circuit diagram showing a detailed configuration example of a write assist timing generation circuit in FIG. 5. FIG. 6 is a schematic diagram showing an example of functions around a write assist circuit that is partially different from FIG. 5. FIG. 11 is a schematic diagram illustrating an example of an effect when the write assist circuit of FIG. 10 is applied to each static memory module in the memory unit of FIG. 4. FIG. 6 is a schematic diagram showing a configuration example around a write assist timing generation circuit different from FIG. 5 in the semiconductor device according to the second embodiment of the present invention. FIG. 13 is a circuit diagram illustrating a detailed configuration example of a row number dummy load circuit and a write assist timing generation circuit in FIG. 12. FIG. 5 is a schematic diagram illustrating an example of characteristics of a word driver power supply circuit block included in each static memory module in the memory unit of FIG. 4 in a semiconductor device according to a third embodiment of the present invention. (A)-(c) is the schematic which illustrates the relationship of the size of each power circuit block for word drivers, when the array structure of each static type memory module in a memory unit differs from FIG. FIG. 15 is a circuit diagram showing a detailed configuration example of a word driver power supply circuit block, a word driver block, and a memory array in each static memory module of FIG. 14; FIG. 17 is a waveform diagram showing an operation example of the word driver power supply circuit block in FIG. 16; FIG. 15 is a circuit diagram illustrating a detailed configuration example of a word driver power supply circuit block, a word driver block, and a memory array in each static memory module of FIG. 14 in a semiconductor device according to a fourth embodiment of the present invention; FIG. 15 is a circuit diagram illustrating a detailed configuration example of a word driver power supply circuit block, a word driver block, and a memory array in each static memory module of FIG. 14 in a semiconductor device according to a fifth embodiment of the present invention; FIG. 15 is a circuit diagram illustrating a detailed configuration example of a word driver power supply circuit block, a word driver block, and a memory array in each static memory module of FIG. 14 in a semiconductor device according to a sixth embodiment of the present invention; (A), (b) is a top view which shows the rough example of arrangement | positioning of each power supply circuit for word drivers in the static type memory module in the semiconductor device by Embodiment 7 of this invention. FIG. 22 is a plan view showing a schematic layout configuration example of a partial region in the static memory module of FIG. In the semiconductor device by Embodiment 8 of this invention, it is the schematic which shows the structural example of the memory unit contained in it. (A) is explanatory drawing which shows the schematic structural example and operation example of the principal part of the static memory module in the semiconductor device examined as a premise of this invention, (b) is a different outline from (a). It is explanatory drawing which shows a structural example and an operation example. (A) is explanatory drawing which shows the schematic structural example and operation example of the principal part of the static memory module in the semiconductor device examined as a premise of this invention, (b) is a different outline from (a). It is explanatory drawing which shows a structural example and an operation example.

  In the following embodiment, when it is necessary for the sake of convenience, the description will be divided into a plurality of sections or embodiments. However, unless otherwise specified, they are not irrelevant, and one is the other. Some or all of the modifications, details, supplementary explanations, and the like are related. Further, in the following embodiments, when referring to the number of elements (including the number, numerical value, quantity, range, etc.), especially when clearly indicated and when clearly limited to a specific number in principle, etc. Except, it is not limited to the specific number, and may be more or less than the specific number.

  Further, in the following embodiments, the constituent elements (including element steps and the like) are not necessarily indispensable unless otherwise specified and apparently essential in principle. Needless to say. Similarly, in the following embodiments, when referring to the shapes, positional relationships, etc. of the components, etc., the shapes are substantially the same unless otherwise specified, or otherwise apparent in principle. And the like are included. The same applies to the above numerical values and ranges.

  The circuit elements constituting each functional block of the embodiment are not particularly limited, but are formed on a semiconductor substrate such as single crystal silicon by a known integrated circuit technology such as a CMOS (complementary MOS transistor). . In the embodiment, MOSFET (Metal Oxide Semiconductor Field Effect Transistor) (abbreviated as MOS transistor) is used as an example of MISFET (Metal Insulator Semiconductor Field Effect Transistor) (abbreviated as MIS transistor), but non-oxidized as a gate insulating film. It does not exclude membranes. In the drawing, a p-channel MOS transistor (PMOS transistor) is distinguished from an n-channel MOS transistor (NMOS transistor) by adding a circle symbol to the gate. Although the connection of the substrate potential of the MOS transistor is not particularly specified in the drawing, the connection method is not particularly limited as long as the MOS transistor can operate normally.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. Note that components having the same function are denoted by the same reference symbols throughout the drawings for describing the embodiment, and the repetitive description thereof will be omitted.

(Embodiment 1)
<Schematic configuration of the entire memory module>
FIG. 1 is a block diagram showing a schematic configuration example of a static memory module included in a semiconductor device according to the first embodiment of the present invention. The static memory module SRMD shown in FIG. 1 includes a control circuit block CTLBK, a word driver block WLD, a word driver power supply circuit block VGEN, a memory array MARY, a column selection circuit YSW, a write driver WTD, a write assist circuit WAST, and a sense amplifier SA. A write auxiliary timing generation circuit TDG and an input / output buffer circuit IOB. The CTLBK includes an address control circuit ADRCTL and a read / write control circuit RWCTL.

  MARY extends (m + 1) word lines WL [0] to WL [m] extending side by side in the first direction and side by side in the second direction intersecting the first direction (n + 1). The bit line pairs (BL [0], ZBL [0]) to (BL [n], ZBL [n]) are arranged at the intersections of (m + 1) word lines and (n + 1) bit line pairs. A plurality of memory cells MC. Each bit line pair is composed of two bit lines (for example, BL [0] and ZBL [0]) that transmit complementary signals. Further, MARY includes (n + 1) memory cell power supply lines (memory cell power supply voltages) ARVDD [0] to ARVDD [n] extending side by side in the second direction, and a bit line pair (BL [s ], ZBL [s]) (s is an integer from 0 to n) is connected to the corresponding ARVDD [s].

  The address control circuit ADRCTL uses the decode activation signal TDEC as a trigger to decode (or predecode) the address signals A [0] to A [j] from the external address terminals of the SRMD, and to select the row selection signals X [0] to X [ k] and column selection signals Y [0] to Y [i] are output. The word driver block WLD selects (activates) one of (m + 1) word lines in accordance with X [0] to X [k]. The column selection circuit YSW selects any one of (n + 1) bit line pairs according to Y [0] to Y [i]. The word driver power supply circuit block VGEN supplies the word driver power supply voltage WLVDD to each word driver (not shown) in the WLD.

  The read / write control circuit RWCTL receives a decode start signal TDEC, an internal write enable signal WE, a write auxiliary enable signal WTE, and a sense amplifier enable signal SE in accordance with various control signals (WEN, CLK, CEN) from the external control terminal of the SRMD. Generate. WEN is a write enable signal for identifying a read command and a write command, CLK is a clock signal serving as a reference for a read / write operation, and CEN is a clock enable signal for controlling validity / invalidity of the clock signal. The input / output buffer circuit IOB takes in the data input signal Di from the external data terminal of the SRMD and transmits it to the write driver WTD, and takes in the output signal from the sense amplifier SA and outputs it as the data output signal Do to the external data terminal. To do.

  The WTD differentially amplifies data from the IOB according to the internal write enable signal WE and transmits the data to a predetermined bit line pair via the column selection circuit YSW described above. The write auxiliary timing generation circuit TDG receives the write auxiliary enable signal WTE and outputs a control signal to the write auxiliary circuit WAST. The WAST controls the memory cell power supply voltage ARVDD of the selected memory cell MC using a control signal from the TDG during a write operation. Although details will be described later, the TDG and WAST portions are one of the main features of the first embodiment. The sense amplifier SA differentially amplifies the signal pair transmitted from the predetermined bit line pair via the YSW using the sense amplifier enable signal SE as a trigger, and outputs the signal pair to the IOB.

  FIG. 2 is a circuit diagram showing a configuration example of each memory cell in the memory module of FIG. The memory cell MC shown in FIG. 2 is an SRAM memory cell including four NMOS transistors MN1 to MN4 and two PMOS transistors MP1 and MP2. MN1 and MN2 are driver transistors, MN3 and MN4 are access transistors, and MP1 and MP2 are load transistors. The gate of MN3 is connected to the word line WL, and one of the source and the drain is connected to the positive bit line BL. The gate of MN4 is connected to WL, and one of the source and the drain is connected to the negative bit line ZBL.

  MN1, MP1, MN2, and MP2 constitute a CMOS inverter circuit between the memory cell power supply voltage ARVDD and the ground power supply voltage VSS, respectively. These two CMOS inverter circuits constitute a latch circuit by connecting one input to the other output. The other of the source and drain of MN4 is connected to the input of the CMOS inverter circuit (MN1, MP1) (the output of the CMOS inverter circuit (MN2, MP2)), and the other of the source and drain of MN3 is connected to the CMOS inverter circuit (MN2, MP2). MP2) input (output of the CMOS inverter circuit (MN1, MP1)).

<< Overall Operation of Memory Module >>
FIG. 3 is a waveform diagram showing a schematic operation example of the memory module of FIG. In the example of FIG. 3, when the clock signal CLK rises, if the clock enable signal CEN is at the “L” level and the write enable signal WEN is at the “H” level, a read (read) cycle (T0) is executed. When CEN is at “L” level and WEN is at “L” level, a write (write) cycle (T1) is executed. Further, when CEN rises and CEN is at “H” level, it becomes a no operation cycle (T2), and neither a read operation nor a write operation is executed.

  In the read cycle (T0), first, the read / write control circuit RWCTL changes the decode activation signal TDEC from the ‘L’ level to the ‘H’ level in response to the rising edge of the clock signal CLK. RWCTL outputs 'L' level as the internal write enable signal WE and the write assist enable signal WTE. In response to the transition of TDEC to the “H” level, the address control circuit ADRCTL receives the row selection signals X [0] to X [k] and the column selection signal Y according to the address signals A [0] to A [j]. [0] to Y [i] (Y [0] is displayed in FIG. 3) are generated. In the example of FIG. 3, the word line WL [0] is selected by X [0] to X [k], and the bit line pair (BL [0], ZBL [0]) is selected by Y [0] to Y [i]. Shall be selected.

  The word driver block WLD raises WL [0] in accordance with X [0] to X [k], and in response thereto, the bit line to which the storage data of each memory cell MC connected to WL [0] corresponds Read in pairs. In this example, read signals in BL [0] and ZBL [0] are transmitted to the sense amplifier SA via the column selection circuit YSW. The read / write control circuit RWCTL shifts the sense amplifier enable signal SE to the valid state ('H' level) after a predetermined delay time from the transition of TDEC to 'H' level. The SA amplifies the read signals of BL [0] and ZBL [0] transmitted through the above-described YSW using the “H” level of the SE as a trigger. The amplified signal is output to the external terminal as the data output signal Do via the input / output buffer circuit IOB. In addition, the raised word line WL [0] is lowered in response to the transition of the TDEC from the “H” level to the “L” level.

  Here, during such a read operation, the word driver power supply circuit block VGEN receives the ‘H’ level of the decode activation signal TDEC and lowers the word driver power supply voltage WLVDD to a predetermined voltage level. For example, VGEN decreases the voltage level of WLVDD from the voltage level of the memory cell power supply voltage ARVDD to a lower voltage level. The voltage level of the selected word line (WL [0]) is determined according to the voltage level of WLVDD. As a result, in each MC connected on WL [0], the static noise margin (SNM) is improved as the β ratio of the driver transistor and the access transistor is improved, and the read margin is improved. After that, VGEN receives the “L” level of TDEC and returns the voltage level of WLVDD to the original voltage level (for example, the voltage level of ARVDD).

  Next, in the write cycle (T1), the read / write control circuit RWCTL first changes the decode activation signal TDEC from the ‘L’ level to the ‘H’ level in response to the rising edge of the clock signal CLK. Further, RWCTL outputs an 'H' level as the internal write enable signal WE and the write auxiliary enable signal WTE. The address control circuit ADRCTL generates row selection signals X [0] to X [k] and column selection signals Y [0] to Y [i] in response to the transition of TDEC to the “H” level, and the word driver block WLD starts up a word line (in this case, WL [0]) corresponding to X [0] to X [k].

  On the other hand, in parallel with this, the data input signal Di from the external terminal is input to the write driver WTD via the input / output buffer circuit IOB. The WTD amplifies the input signal from the IOB in response to the above-mentioned WE “H” level, and the column selection circuit YSW outputs the output of the WTD to a bit line pair (Y [0] to Y [i]) Here, it connects to BL [0], ZBL [0]). As a result, Di information is written into the selected memory cell MC. Thereafter, the raised word line (WL [0]) is lowered in response to the transition from the ‘H’ level to the ‘L’ level of the TDEC. As a result, the selected memory cell MC holds Di information.

  Here, during such a write operation, the write auxiliary circuit WAST receives the “H” level of the write auxiliary enable signal WTE via the write auxiliary timing generation circuit TDG, and the memory cell power supply voltage (here, the write target) ARVDD [0]) is reduced to a predetermined voltage level. As a result, in the memory cell to be written (here, MC at the intersection of WL [0], BL [0], and ZBL [0]), the static noise margin (SNM) is reduced as the driving capability of the driver transistor is reduced. As a result, the write margin is improved. Thereafter, the WAST receives the ‘L’ level of the WTE and returns the voltage level of the WLVDD to the original voltage level (for example, the voltage level of the ARVDD). In such a write operation, the word driver power supply circuit block VGEN reduces the word driver power supply voltage WLVDD to a predetermined voltage level as in the case of the read operation described above. As a result, the non-write target memory cell on the selected word line (WL [0]) is improved in SNM (read margin), so that the stored data can be reliably held.

<< Schematic configuration of the entire semiconductor device >>
FIG. 4 is a block diagram showing a schematic configuration example of the entire semiconductor device according to the first embodiment of the present invention. FIG. 4 shows a semiconductor device (LSI) called SOC (System On a Chip) in which various logic circuits and memory circuits are formed in one semiconductor chip. The semiconductor device of FIG. 4 is, for example, a mobile phone LSI, and includes two processor units CPU1, CPU2, an application unit APPU, a memory unit MEMU, a baseband unit BBU, and an input / output unit IOU. The MEMU includes a plurality (three in this case) of static memory modules SRMD1 to SRMD3 each having a different array configuration (number of rows (number of word lines) and number of columns (number of bit line pairs)). 1 is applied.

  CPU1 and CPU2 perform predetermined calculation processing based on the program, APPU performs predetermined application processing required by the mobile phone, BBU performs predetermined baseband processing associated with wireless communication, and IOU between the outside I / O interface. Here, the SRMD1 to SRMD3 in the MEMU are accessed as cache memories, for example, when processing such various circuit blocks. Since the optimal cache memory configuration (number of lines and bit width) can be changed as appropriate according to the configuration of various circuit blocks, processing contents, and the like, the array configuration of each memory module can be different accordingly. As a result, in the semiconductor device, as shown in FIG. 4, memory modules having various array configurations such as a vertically long configuration (SRMD1), a horizontally long configuration (SRMD2), and a substantially square configuration (SRMD3) may be mounted. . Although not particularly limited, each memory module is appropriately determined to have an optimal array configuration from, for example, the number of rows of 8 to 512 and the number of columns of 16 to 512.

  Each memory module in such a semiconductor device is automatically generated by specifying the number of rows and the number of columns for an automatic design tool called a memory compiler or the like, for example. Each memory module generated in this way is called a compiled SRAM or the like as described above. Since the compiled SRAM is automatically generated using a layout of various element circuits (for example, word drivers) defined in advance, the drive capability (transistor size) of the various element circuits (for example, word drivers) is provided for each memory module. It is difficult to perform individual optimization according to the array configuration. In some cases, more than ten compiled SRAMs may be mounted in the semiconductor device, and optimization for each memory module described above may be more difficult particularly in such a case. As a result, as described with reference to FIG. 24 and FIG. 25, there is a possibility that the operation margin (read margin, write margin) may be reduced and the access time may be delayed depending on the array configuration.

<< Overview of Write Auxiliary Circuit (Main Features of First Embodiment) >>
FIG. 5 is a schematic diagram showing an example of functions around the write assist circuit in the memory module of FIG. In the static memory module SRMDa of FIG. 5, the word driver block WLD, the control circuit block CTLBK, the write assist timing generation circuit TDG1, the input / output buffer circuit IOB, and the write assist circuit WAST1 [0] are representatively selected from the configuration example of FIG. ] To WAST1 [q] and a plurality of memory cells MC are shown. WAST1 [0] controls the memory cell power supply voltage ARVDD [0] described above, and WAST1 [q] controls the memory cell power supply voltage ARVDD [n] described above. Row number information XSET is set in advance in the write assist timing generation circuit TDG1. XSET is a digital code representing the number of word lines included in the SRMDi, and is not particularly limited. However, XSET is stored in a circuit or a register in advance, and is non-volatile when the semiconductor device is initialized. It is determined by a method of loading from a memory or the like. The TDG 1 outputs a write assist pulse signal WPT having a wider pulse width as the number of rows set by XSET increases.

  When the write operation is performed on the memory cell MC connected to ARVDD [0], WAST1 [0] drives the switch SWm according to the write assist enable signal WTE to turn on the ARVDD [0]. The voltage level is lowered to a predetermined voltage level VM1. Further, when WAST1 [0] decreases the voltage level of ARVDD [0], WST1 [0] drives switch SWs to ON during the WPT pulse period, and temporarily charges ARVDD [0] to voltage level VM2 (for example, The voltage level lowering speed is controlled by discharging toward the voltage level of VM1 or lower. Similarly, WAST1 [q] is a voltage level of ARVDD [n] by driving SWm on according to WTE when a write operation is performed on a memory cell MC connected to ARVDD [n]. Is reduced to VM1. Furthermore, WAST1 [q] controls the rate of voltage level reduction by driving SWs on during the WPT pulse period when the voltage level of ARVDD [n] is reduced.

  FIG. 6 is a schematic diagram showing an example of the effect when the write assist circuit of FIG. 5 is applied to each static memory module in the memory unit of FIG. The memory unit MEMU shown in FIG. 6 has a word line (not shown) extending in the horizontal direction (X-axis direction) and a bit line (not shown) and the memory cell power supply line ARVDD extending in the vertical direction (Y-axis). As a direction), a vertically long static memory module SRMD1 and a horizontally long static memory module SRMD2 are included. Since SRMD1 has more rows than SRMD2 (long in the Y-axis direction), write assist pulse signal WPT having a wide pulse width is applied to write assist circuit WAST1_1 of SRMD1, and write assist circuit WAST1_2 of SRMD2 is applied to SRMD2. WPT with a narrow pulse width is applied.

  Here, it is assumed that the drive capability of the switch SWm in FIG. 5 is set to the drive capability according to the minimum row value that the memory module can take, and the number of rows of SRMD2 is slightly larger than the minimum row value. Shall. In SRMD1 and SRMD2, if the memory cell power supply voltage ARVDD is lowered only by WTE control without providing the WPT (switch SWs in FIG. 5), the length of the memory cell power supply line as shown in the comparative example of FIG. The time required to reach a predetermined voltage level varies depending on the size (the size of the load). Here, a longer time is required for SRMD1 than for SRMD2. As described above, when the time until the predetermined voltage level is reached becomes long, the write margin may decrease as described in FIG.

  Therefore, in the case of SRMD1, the time to reach a predetermined voltage level is greatly shortened by promoting the falling speed of ARVDD by using a WPT having a wide pulse width. In the case of SRMD2, a narrow pulse is used. The time is slightly shortened by slightly promoting the falling speed using a WPT having a width. This makes it possible to improve the write margin regardless of the array configuration of the memory module. Note that the drive capability of the switch SWm in FIG. 5 is set to be, for example, a drive capability corresponding to the minimum row value that can be taken by the memory module or lower. In the former case, when the memory module has the minimum row value, for example, the design specification is such that no pulse is input to the WPT. In the latter case, the memory module has the minimum row value, for example. The design specification is such that a narrow pulse is input to the WPT.

  Here, the variation in the write margin due to the array configuration is compensated by changing the pulse width of the WPT. However, in some cases, a circuit configuration with variable drive capability is applied to the switches SWm and SWs in FIG. It is also possible to perform such compensation. That is, for example, it is also possible to use a system in which SWm and SWs are configured from a plurality of switches connected in parallel and the number of switches actually used is selected according to the array configuration. However, in this case, in order to vary the driving capability over a wide range, a large number of switches are required, which may increase the circuit area. From this point of view, it is desirable to use a method of adjusting according to the WPT pulse width as shown in FIG.

<Details of write assist circuit>
FIG. 7A is a circuit diagram showing a detailed configuration example of the write assist circuit in FIG. 5, and FIG. 7B is a circuit diagram showing a configuration example different from FIG. 7A. FIG. 8 is a waveform diagram showing an operation example of the write assist circuit in FIGS. 7 (a) and 7 (b). First, the write assist circuit WAST1a shown in FIG. 7A includes a static part VSBK composed of PMOS transistors MP10 to MP12 and NMOS transistors MN10 and MN11, and a dynamic part VDBK1a composed of NMOS transistor MN12. The VSBK is a circuit that switches the memory cell power supply voltage from a certain voltage level to a predetermined voltage level lower than that at the time of a write operation, and mainly performs setting and stable supply of the predetermined voltage level. On the other hand, the VDBK1a is a circuit that operates only when the voltage level is switched and controls the switching speed. Conceptually, VSBK corresponds to the part of the switch SWm in FIG. 5, and VDBK1a corresponds to the part of the switch SWs in FIG.

  In VSBK, MP10 and MP12 have source / drain paths connected in parallel between the power supply voltage VDDM and the common power supply node CWSRC [0]. MP11, MN11, and MN10 are serially connected in series between CWSRC [0] and the ground power supply voltage VSS, with MP11 as the CWSRC [0] side and MN10 as the VSS side. The gates of MP10 and MN10 are controlled by a write assist enable signal WTE, and a fixed voltage TE is applied to the gate of MP11. The voltage level of CWSRC [0] is fed back to the gate of MN11, and the voltage level of the common connection node of MP11 and MN11 is fed back to the gate of MP12. On the other hand, in VDBK1a, the source / drain path of MN12 is connected between CWSRC [0] and VSS, and the gate is controlled by a write assist pulse signal WPT.

  The common power supply node CWSRC [0] is connected to the memory cell power supply lines ARVDD [0] to ARVDD [3] via the source / drain paths of four PMOS transistors here. Here, as one of the four PMOS transistors, a PMOS transistor MP21 corresponding to ARVDD [0] is representatively shown. The source / drain path of the PMOS transistor is also connected between each of ARVDD [0] to ARVDD [3] and the power supply voltage VDDM. Here, a PMOS transistor MP20 corresponding to ARVDD [0] is shown as a representative.

  The gate of MP20 is controlled by a write column selection signal CWSE [0], and the gate of MP21 is controlled by a read column selection signal CRSE [0]. MP20 and MP21 correspond to a part of the column selection circuit YSW in FIG. Here, in the configuration example of FIG. 1, one I / O is assigned to four bit line pairs during the write operation (that is, the write operation is performed on one of the four bit line pairs). In the configuration example of FIG. 7A, four memory cell power supply lines are assigned to one write auxiliary circuit.

  Here, for example, when writing to a memory cell connected to ARVDD [0], as shown in the write cycle (T1) of FIG. 8, the write assist enable signal WTE and the write column select signal CWSE [0] are ' The read column selection signal CRSE [0] changes from the “H” level to the “L” level. At this time, the other column selection signals for writing (here, CWSE [1] corresponding to ARVDD [1] is exemplified) hold the “L” level, and other column selection signals for reading (here, ARVDD [1]). ], For example, CRSE [1] corresponding to] holds the “H” level. As a result, MP20 is controlled to be off and ARVDD [0] is connected to CWSRC [0] via MP21, and ARVDD [1] to ARVDD [3] include PMOSs corresponding to MP20 of ARVDD [0]. VDDM is applied through the transistor.

  Also, VDDM is applied to CWSRC [0] in VSBK through MP10 when WTE is at the “L” level. At this time, MN10 and MP12 are turned off. On the other hand, when WTE transitions to the “H” level, MP10 is turned off and MN10 is turned on. As a result, MN11 is turned on. TE is applied to the gate of MP11 so as to have an appropriate on-resistance. As a result, the charge of CWSRC [0] is discharged through MP11, MN11, and MN10, the voltage level of CWSRC [0] is lowered, and MP12 is turned on. Here, if the voltage level of CWSRC [0] is too low, MP12 is turned on strongly, and MN11 is weakened so that the voltage level rises. Conversely, if the voltage level is too high, MP12 is turned on. The voltage level decreases because the MN11 is turned on and becomes weaker. As a result, the voltage level of CWSRC [0] converges to a predetermined voltage level determined by the ratio of the on resistance when MP12, MP11, MN11, and MN10 are all balanced in the on state. The voltage level of CWSRC [0] becomes the voltage level of ARVDD [0] via MP21.

  Further, in the write cycle (T1), the “H” pulse is applied to the write assist pulse signal WPT together with the transition of the WTE to the “H” level. As a result, the MN12 in the VDBK1a is turned on, and the charge of CWSRC [0] is rapidly discharged toward VSS in the WPT 'H' pulse period, and the voltage level of CWSRC [0] rapidly decreases. Therefore, by controlling the “H” pulse period of the WPT, it is possible to control the rate of decrease in the voltage level of CWSRC [0] (ARVDD [0]). Thereafter, when the write operation is completed, WTE and CWSE [0] transition to the 'L' level, and CRSE [0] transitions to the 'H' level. As a result, the voltage levels of both CWSRC [0] and ARVDD [0] return to VDDM.

  Next, the write assist circuit WAST1b shown in FIG. 7B is different in the circuit configuration of the dynamic part from the WAST1a shown in FIG. The dynamic part VDBK1b in the WAST1b of FIG. 7B includes a PMOS transistor MP13 having a source / drain path connected between a common connection node of MP11 and MN11 in the static part VSBK and a common power supply node CWSRC [0]. Yes. The gate of MP13 is controlled by the inverted signal (/ WPT) of the write assist pulse signal WPT.

  When VDBK1b of FIG. 7B is used, unlike the case of using VDBK1a of FIG. 7A, it is possible to easily prevent a situation where the voltage level of CWSRC [0] is excessively lowered. That is, if the voltage level of CWSRC [0] is lowered too much, the MN11 in VSBK is driven off, so that the voltage level drop can be automatically stopped. As a result, the timing design of the write assist pulse signal WPT (/ WPT) can be facilitated. Further, by using a feedback circuit type static unit VSBK as shown in FIGS. 7A and 7B, for example, compared with a case where a predetermined voltage level is generated by simple resistance voltage division or the like, A stable voltage level can be generated. Note that, in the write cycle (T1), each transistor in the VSBK mainly has a function of determining a DC voltage level, so that the transistor size may be small. On the other hand, the transistors in VDBK1a and VDBK1b desirably have a relatively large driving capability in order to extract charges at a high speed, and the transistor size is preferably larger than each transistor in VSBK.

<Details of write assist timing generation circuit>
FIG. 9 is a circuit diagram showing a detailed configuration example of the write assist timing generation circuit in FIG. The write assist timing generation circuit TDG1 shown in FIG. 9 includes an inverter circuit IV1, a plurality (three in this case) of delay circuit blocks DLYBK1 to DLYBK3, a NAND operation circuit ND1, and a buffer circuit BF. A write assist enable signal WTE is input to one of the two inputs of ND1, and an inverted signal of WTE via IV1 is sequentially input to DLYBK1 through DLYBK3 to the other of the two inputs of ND1. The BF buffers the output of ND1, and outputs an inverted signal (/ WPT) of the write assist pulse signal WPT. The inverted WPT signal (/ WPT) is input to the write auxiliary circuit WAST1b ([0], [1], [2],...) Having the circuit configuration shown in FIG.

  Each of DLYBK1 to DLYBK3 has two paths, one end of which is commonly connected to the input node of the delay circuit block, the other end of the two paths having two inputs, and an output connected to the output node of the delay circuit block. The selector circuit SEL is provided. A delay element DLY (for example, an inverter circuit having a plurality of stages) having a predetermined delay amount is inserted into one of the two paths. The output node of DLYBK1 is connected to the input node of DLYBK2, and the output node of DLYBK2 is connected to the input node of DLYBK3. Here, selection of SEL included in each of DLYBK1 to DLYBK3 (that is, whether or not via DLY) is performed based on the above-described number-of-rows information XSET.

  As a result, the WTE 'H' pulse is input to one of the two inputs of ND1, and the 'L' pulse that is an inverted signal of the 'H' pulse is delayed based on XSET to the other of the two inputs of ND1. A signal is input. As a result, ND1 outputs an 'L' pulse signal having a delay time based on XSET as a pulse width. It is desirable that each delay element DLY included in DLYBK1 to DLYBK3 is weighted. For example, by setting the delay amount of each DLY in DLYBK1: DLYBK2: DLYBK3 to 1: 2: 4, etc., the pulse width can be adjusted in the range of 0 to 7 according to the value of XSET. Become.

<< Outline of Write Auxiliary Circuit (Modification) >>
FIG. 10 is a schematic diagram illustrating an example of functions around the write assist circuit that is partially different from FIG. The static memory module SRMDa of FIG. 10 includes write assist circuits WAST2 [0] to WAST2 [q] instead of WAST1 [0] to WAST1 [q] shown in FIG. Each of the write assist circuits WAST2 [0] to WAST2 [q] includes only the switch SWs for the write assist pulse signal WPT without including the switch SWm portion for the write assist enable signal WTE in FIG. It has become. Specifically, each of WAST2 [0] to WAST2 [q] has a configuration including only the dynamic portion VDBK1a without including the static portion VSBK in FIG. 7A, for example.

  An SRAM memory cell normally has a very low current consumption. Therefore, in some cases, a static part is not provided, and a memory part power supply voltage is lowered to a predetermined voltage level by a dynamic part (switch SWs). The voltage level can also be maintained to some extent by turning off and setting the memory cell power line to a high impedance state. Therefore, it is possible to realize the write assist circuit with a configuration as shown in FIG. 10, thereby reducing the circuit area and the like. However, if the memory cell power supply line is set to a high impedance state, for example, a malfunction due to mixing of external noise or the like may occur, so it is desirable to use the configuration example shown in FIG. 5 from this viewpoint.

  FIG. 11 is a schematic diagram showing an example of the effect when the write assist circuit of FIG. 10 is applied to each static memory module in the memory unit of FIG. The memory unit MEMU shown in FIG. 11 includes a vertically long static memory module SRMD1 and a horizontally long static memory module SRMD2, as in FIG. Since SRMD1 has more rows than SRMD2 (long in the Y-axis direction), a write assist pulse signal WPT having a wide pulse width is applied to the write assist circuit WAST2_1 of SRMD1, and the write assist circuit WAST2_2 of SRMD2 is applied to the write assist circuit WAST2_2 of SRMD2. WPT with a narrow pulse width is applied.

  In SRMD1 and SRMD2, if the drive capability of the switch SWs and the pulse width of WPT in FIG. 10 are the same, the length of the memory cell power line (the size of the load) as shown in the comparative example of FIG. The voltage level of the memory cell power supply voltage ARVDD after the decrease can be different. Here, a situation occurs in which the voltage level of ARVDD is too high in SRMD1, and a situation in which the voltage level of ARVDD is too low in SRMD2. If the voltage level of ARVDD is too high, the write margin may be lowered as described above. If the voltage level of ARVDD is too low, for example, the latch operation at the end of writing may be insufficient, or the ARVDD may be reduced. There may occur a situation in which the latch capability is insufficient in the non-write target memory cell to be connected. Therefore, as shown in FIG. 11, by changing the WPT pulse width according to the length of the memory cell power supply line (number of word lines (number of rows)), the voltage level of ARVDD is constant regardless of the array configuration. It is possible to avoid the situation as described above.

  As described above, by using the semiconductor device according to the first embodiment, typically, the operation margin of a plurality of static memory modules included in the semiconductor device can be improved.

<< Other modifications >>
In the description so far, the method of pulling out charges from the memory cell power supply line by adjusting the pulse width has been shown. However, instead of adjusting the pulse width, for example, the adjustment may be made according to the size (for example, gate width) of the transistors MN12 and MP13 in FIG. That is, in the two memory modules, the size of the transistors MN12 and MP13 is set larger in the memory module having a larger number of word lines than in the memory module having a smaller number of word lines. A plurality of transistors may be provided, and the sum of their sizes (gate widths) (in other words, driving capability) is larger when the number of word lines is larger. As a result, the ability to extract charges from the memory cell power supply line of the memory module having a large number of word lines can be increased.

(Embodiment 2)
<< Summary around the write assist timing generation circuit (modification) >>
FIG. 12 is a schematic diagram showing a configuration example around a write assist timing generation circuit different from FIG. 5 in the semiconductor device according to the second embodiment of the present invention. In the static memory module SRMDb of FIG. 12, the word driver block WLD, the control circuit block CTLBK, the input / output buffer circuit IOB, the write auxiliary circuits WAST1 [0] to WAST1 [q] and a plurality of memories are provided as in the case of FIG. A cell MC is representatively shown. Furthermore, the SRMDb in FIG. 12 includes a write assist timing generation circuit TDG2 different from that in FIG. 5, and a row number dummy load circuit XDMY is newly added.

  The number-of-rows dummy load circuit XDMY has a size in the Y-axis direction proportional to the size of the word driver block WLD in the Y-axis direction (number of word lines (number of rows)), and the delay amount increases as the size in the Y-axis direction increases. It has a function to generate. The size of XDMY in the Y-axis direction is typically set equal to the size of WLD in the Y-axis direction. Unlike the TDG 1 in FIG. 5, the write assist timing generation circuit TDG 2 does not receive the row number information XSET. Instead, the write assist timing generation circuit TDG 2 obtains the number of rows based on the delay amount generated by XDMY, and sets the pulse width corresponding to this delay amount. A write assist pulse signal WPT is output. Each of WAST1 [0] to WAST1 [q] controls the falling speed of the memory cell power supply voltages ARVDD [0] to ARVDD [n] using the WPT in the write operation as in the case of FIG.

  When such a configuration example is used, the delay amount corresponding to the number of rows can be generated easily or with high precision by XDMY, and as a result, the fall of the memory cell power supply voltage using the write assist circuit is achieved. Speed control can be facilitated or made highly accurate. That is, for example, when the pulse width of the WPT is adjusted using the TDG 1 of FIG. 9 described above, the pulse width is digitally controlled at a predetermined step size, so that the number of rows is reflected in the pulse width of the WPT with high accuracy. For this purpose, it is necessary to reduce the delay amount of each delay element DLY and to provide many delay circuit blocks (DLYBK). In this case, the circuit area may increase, the circuit may become complicated, and the like. On the other hand, as shown in FIG. 12, when the pulse width is controlled using XDMY, the larger the XDMY size (that is, the number of rows), the larger the delay amount using the parasitic components (parasitic capacitance, parasitic resistance). Therefore, analog control of the pulse width can be easily realized. Furthermore, since the number-of-rows information XSET becomes unnecessary, the complexity associated with this setting can be eliminated.

<< Details around the write assist timing generation circuit (modification) >>
FIG. 13 is a circuit diagram showing a detailed configuration example of the row number dummy load circuit and the write assist timing generation circuit in FIG. In FIG. 13, the row number dummy load circuit XDMY includes two inverter circuits IV10 to IV12 that extend side by side in the Y-axis direction (bit line (not shown), extending direction of the memory cell power supply line ARVDD). Dummy bit lines DBL1, DBL2 and capacitors C1, C2 are provided. IV10 receives the write assist enable signal WTE and outputs an inverted signal toward one end of DBL1. IV11 receives the other end of DBL1 and outputs an inverted signal toward one end of DBL2. IV12 receives the other end of DBL2 and outputs an inverted signal toward the write assist timing generation circuit TDG2. Here, DBL1 is a forward wiring and DBL2 is a backward wiring.

  As described above, the wiring lengths of DBL1 and DBL2 are determined according to the size of the word driver block WLD in the Y-axis direction. The capacitor C1 is connected between DBL1 and the ground power supply voltage VSS, and the capacitor C2 is connected between DBL2 and VSS. C1 includes the parasitic capacitance of DBL1, and C2 includes the parasitic capacitance of DBL2. Thereby, the capacitance values of C1 and C2 increase as DBL1 and DBL2 become longer. Further, C1 and C2 may include a separately formed capacitor. Specifically, for example, a circuit configuration and a layout configuration in which a capacitive element (for example, a diffusion layer capacitance or a MOS capacitance) is added to DBL1 and DBL2 at a certain length can be used. Also in this case, the capacitance values of C1 and C2 increase as DBL1 and DBL2 become longer.

  XDMY delays the WTE 'H' pulse input from IV10 by a time corresponding mainly to the parasitic resistance values of DBL1 and DBL2 and the capacitance values of C1 and C2, and outputs the 'L' pulse via IV12. . On the other hand, the write assist timing generation circuit TDG2 of FIG. 13 deletes the delay path including the inverter circuit IV1 and the delay circuit blocks DLYBK1 to DLYBK3 from the TDG1 shown in FIG. The configuration is replaced with a route.

  That is, WTE is input to one of the two inputs of the NAND operation circuit ND1, and a signal obtained by delaying and inverting WTE via XDMY (an output signal of IV12) is input to the other of the two inputs of ND1. As a result, the ND1 outputs an 'L' pulse signal having a delay time based on XDMY as a pulse width, as in the case of TDG1. This 'L' pulse signal becomes an inverted signal (/ WPT) of the write assist pulse signal WPT via the buffer circuit BF, and the write assist circuit WAST1b ([0], [1], [2],...) Signal (/ WPT) is used to control the falling speed of memory cell power supply voltage ARVDD. As a result, it is possible to improve the write margin regardless of the array configuration. In the XDMY of FIG. 13, the delay time is set by one reciprocal wiring (DBL1, DBL2). However, in some cases, the delay time can be set by providing two or more reciprocal wirings. .

  As described above, by using the semiconductor device according to the second embodiment, typically, the operation margin of a plurality of static memory modules included in the semiconductor device can be improved. Note that the difference in the number of rows of the memory modules shown in the first and second embodiments usually differs by a power of two. For example, when the number of word lines (number of rows) of one memory module is 256 (2 to the 8th power), for example, it is 128 (2 to the 7th power), for example, and when it is more than this, for example, 512 ( 2 to the 9th power).

(Embodiment 3)
<< Outline of Power Circuit Block for Word Driver (Main Features of Embodiment 3) >>
FIG. 14 is a schematic diagram showing an example of characteristics of the word driver power supply circuit block included in each static memory module in the memory unit of FIG. 4 in the semiconductor device according to the third embodiment of the present invention. In the memory unit MEMU shown in FIG. 14, the extending direction of the word line WL is the horizontal direction (X-axis direction), and the extending direction of the bit line (not shown) and the word driver power supply line WLVDD is the vertical direction (Y-axis direction). A vertically long static memory module SRMD1 and a horizontally long static memory module SRMD2 are included.

  SRMD1 is a word driver block WLD1 including word drivers WD corresponding to the number of word lines WL in the memory array MARY1, and a word driver for supplying a word driver power supply voltage to each WD in WLD1 via WLVDD. A power supply circuit block VGEN1 is provided. Similarly, the SRMD 2 includes a word driver block WLD2 including a number of WD corresponding to the number of WLs in the memory array MARY2, and a word driver power supply that supplies a word driver power supply voltage to each WD in the WLD2 via WLVDD. A circuit block VGEN2 is provided.

  The configuration example of FIG. 14 is characterized in that VGEN1 has a larger size (drive capability) than VGEN2. Specifically, the gate width of the transistors in the word driver power supply circuit block is large (the gate widths of the transistors MP30 to MP32 and MN30 in FIG. 16). VGEN1 and VGEN2 have a function of lowering the voltage level of the word driver power supply line (word driver power supply voltage) WLVDD during the read operation (write operation) as described in FIG. As a result, the static noise margin (SNM) is improved and the read margin can be improved. However, if the size (driving capability) of the word driver power supply circuit blocks is the same in SRMD1 and SRMD2, the following may be a concern.

  First, since the length of the word driver power supply line WLVDD is long (the load is large) in SRMD1, it may take time to reduce the word driver power supply voltage WLVDD to a predetermined voltage level as shown in the comparative example of FIG. There is. Furthermore, in SRMD1, since the length of the word line WL is short (load is small), as shown in the comparative example of FIG. 14, the rising speed of WL is fast, and there is a possibility that overshoot occurs in the voltage level of WL. As a result, in SRMD1, as described with reference to FIG. 25B, the voltage level of WL tends to become excessively high, and there is a possibility that a sufficient read margin cannot be secured.

  On the other hand, in SRMD2, since the length of WLVDD is short (load is small), as shown in the comparative example of FIG. 14, WLVDD rapidly decreases toward a predetermined voltage level, and undershoot may occur in some cases. is there. Furthermore, in SRMD2, since the length of WL is long (the load is large), as shown in the comparative example of FIG. 14, the rise of WL can be delayed. As a result, in SRMD2, as described with reference to FIG. 25A, the rising speed of WL tends to be excessively slow, and the access time may not be increased.

  Therefore, in the semiconductor device of the third embodiment, the larger the number of rows (number of word lines) and the smaller the number of columns (number of bit line pairs), the size (driving capability) of the word driver power supply circuit block VGEN. ) Is one of the main features. That is, when the number of rows is large, by increasing the drive capability of VGEN, the falling speed of WLVDD is increased to ensure a read margin. Conversely, when the number of rows is small, the VGEN drive capability is reduced to suppress an excessive voltage drop of WLVDD and to ensure a sufficient word line rise speed (access time). Also, when the number of columns is small, increasing the VGEN drive capability increases the falling speed of WLVDD, suppresses the word line voltage level from becoming excessively high, and ensures a read margin. On the other hand, when the number of columns is large, the VGEN drive capability is reduced to secure a sufficiently high voltage level necessary for WLVDD, and the rise speed (access time) of the word line is suppressed from slowing down.

  In the configuration example of FIG. 14, in SRMD1 with a large number of rows and a small number of columns, the drive capability (size) of VGEN1 is set to be large, and conversely, in SRMD2 with a small number of rows and a large number of columns, VGEN2 The driving ability (size) is set small. As a result, as shown in FIG. 14, the word driver power supply voltage WLVDD is lowered to a suitable voltage level at a suitable falling speed, and the rising speed of the word line WL is also a suitable speed. Regardless, it is possible to secure a sufficient read margin and a sufficient access time.

  FIGS. 15A to 15C are schematic views illustrating the relationship of the sizes of the power supply circuit blocks for each word driver when the array configuration of each static memory module in the memory unit is different from that in FIG. . First, in FIG. 15A, two static memory modules SRMD4 and SRMD5 having the same number of columns are provided in the memory unit MEMU. The SRMD4 has a memory array MARY4 whose size (number of columns) in the X-axis direction is X4 and whose size (number of rows) in the Y-axis direction is Y4. The SRMD5 has the same size (number of columns) in the X-axis direction as X4 and Y It has a memory array MARY5 whose axial size (number of rows) is Y5. Here, since Y4> Y5, the size (driving capability) of the word driver power supply circuit block VGEN4 of SRMD4 is set larger than the size (driving capability) of the word driver power supply circuit block VGEN5 of SRMD5.

  Next, in FIG. 15B, two static memory modules SRMD4 and SRMD6 having the same number of rows are provided in the MEMU. As described above, SRMD4 has MARY4 of X4 and Y4, and SRMD6 has a memory array MARY6 in which the size (number of columns) in the X-axis direction is X6 and the size (number of rows) in the Y-axis direction is Y4. Here, since X4> X6, the size (drive capability) of the word driver power supply circuit block VGEN6 of SRMD6 is set larger than the size (drive capability) of VGEN4 of SRMD4. Subsequently, in FIG. 15C, two static memory modules SRMD4 and SRMD7 each having different numbers of rows and columns are provided in the MEMU. As described above, SRMD4 has X4 and Y4 MARY4, and SRMD7 has a memory array MARY7 whose X-axis direction size (column number) is X7 and Y-axis direction size (row number) is Y7. Here, Y4> Y7, but since X4> X7, the size (driving capability) of the word driver power supply circuit block VGEN7 of SRMD7 and the size (driving capability) of VGEN4 of SRMD4 are sometimes equal. obtain.

  Note that, here, it has been described that the size is the same between the two static memory modules, specifically, the number of rows is the same, or the number of columns is the same. It only has to be equal. The number of rows and the number of columns are usually composed of powers of 2. If the number of rows is 512 (2 to the 9th power) in the first static memory module, there may be a difference of, for example, about 10 in 512 in the second static memory module. This difference may include redundant rows. On the other hand, in the case of 256 (2 to the 8th power) and 1024 (2 to the 10th power) in the second static memory module, that is, when they differ by 2 multipliers, they are not within the same range.

《Details around the power circuit block for word driver》
FIG. 16 is a circuit diagram showing a detailed configuration example of the word driver power supply circuit block, the word driver block, and the memory array in each static memory module of FIG. FIG. 17 is a waveform diagram showing an operation example of the word driver power supply circuit block in FIG. In FIG. 16, the word driver power supply circuit block VGENa includes PMOS transistors MP30 to MP32 and an NMOS transistor MN30. The source / drain path of the MP30 is connected between the power supply voltage VDDM and the word driver power supply line (word driver power supply voltage) WLVDD. MP31 and MP32 have source / drain paths connected in parallel between WLVDD and the drain of MN30, and the source of MN30 is connected to the ground power supply voltage VSS. The gates of MP30 to MP32 are controlled by enable signals EN1 to EN3, respectively, and the gate of MN30 is controlled by an enable signal VDDEN.

  The word driver block WLDa includes (m + 1) word drivers WD [0] to WD [m]. Here, each of WD [0] to WD [m] is a CMOS inverter circuit including a PMOS transistor MP40 and an NMOS transistor MN40. The power supply voltage of the CMOS inverter circuit is supplied in common via the word driver power supply line WLVDD from VGENa. The memory array MARY includes (m + 1) word lines WL [0] to WL [m] and (n + 1) bit line pairs (BL [0], ZBL [0]) to (BL [n], ZBL). [N]) and a plurality of (here, (m + 1) × (n + 1)) memory cells MC arranged at the intersections of the word line and the bit line pair. WL [0] to WL [m] are driven by WD [0] to WD [m] in WLDa, respectively.

  VGENa performs an operation as shown in FIG. 17 during a read operation (write operation). First, when the decode activation signal TDEC shown in FIGS. 1 and 3 is at the ‘L’ level, EN1 and EN2 are at the ‘L’ level and VDDEN is at the ‘L’ level. As a result, MP30 and MP31 in VGENa are turned on, MN30 is turned off, and the word driver power supply voltage WLVDD becomes VDDM. Thereafter, when TDEC transitions to the “H” level in accordance with the read operation (write operation), VDDEN transitions to the “H” level accordingly. As a result, the voltage level of WLVDD decreases from VDDM to a voltage level determined by the on-resistance ratio of MP30, MP31, and MN30.

  At this time, EN3 is preset to either the 'H' level or the 'L' level. If EN3 is set to the “L” level, MP32 is turned on, and the ON resistance in the parallel circuit of MP31 and MP32 is lowered. The decrease level of the voltage level increases. The setting of EN3 is performed, for example, according to the magnitude of the power supply voltage VDDM (corresponding to the power supply voltage of VGENa and the power supply voltage of the memory cell MC) used during the read operation.

  For example, when the static memory module has a normal operation mode and a high-speed operation mode, the voltage level of VDDM is set higher in the high-speed operation mode than in the normal operation mode. In this case, the static noise margin (SNM) (reading margin) may be lower than that in the normal operation mode because of the variation in the threshold voltage of each transistor in the MC. In view of this, in the high-speed operation mode, it is possible to compensate for the decrease in the read margin by increasing the decrease width of the voltage level of WLVDD compared to that in the normal operation mode. Of course, the functions associated with EN3 and MP32 can be omitted.

  On the other hand, in parallel with the operation of VGENa, the word driver WD [s] (s is an integer of 0 to m) in WLDa is moved to the 'H' level of TDEC as shown in FIG. The corresponding word line WL [s] is activated according to the transition of. At this time, the voltage level of WL [s] is determined by the voltage level of WLVDD described above. Thereafter, when TDEC transitions to the “L” level, WL [s] is deactivated via WD [s], and VDDEN returns to the “L” level, and the voltage level of WLVDD is accordingly changed to VDDM. Return to.

  In the configuration example of FIG. 16, the transistor size of each MOS transistor (MP30 to MP32, MN30) in VGENa increases as the number of rows (number of word lines) (m + 1) increases and the number of columns (number of bit lines). It is characterized in that it is set larger as (n + 1) decreases. As a result, as described with reference to FIG. 14, it is possible to ensure a sufficient read margin and a sufficient access time regardless of the memory array configuration.

  As described above, by using the semiconductor device according to the third embodiment, typically, the operation margin of a plurality of static memory modules included in the semiconductor device can be improved. It is also possible to increase the speed of a plurality of static memory modules.

(Embodiment 4)
<< Details of Power Supply Circuit Block for Word Driver (Modification [1]) >>
18 is a circuit diagram showing a detailed configuration example of the word driver power supply circuit block, the word driver block, and the memory array in each static memory module of FIG. 14 in the semiconductor device according to the fourth embodiment of the present invention. The configuration example shown in FIG. 18 differs from the configuration example of FIG. 16 described above in the internal configuration of the word driver power supply circuit block. Since other configurations are the same as those in FIG. 16, a detailed description thereof will be omitted. The word driver power supply circuit block VGENb in FIG. 18 includes (p + 1) word driver power supply circuits VG [0] to VG [p].

  Each of VG [0] to VG [p] includes PMOS transistors MP30 to MP32 and an NMOS transistor MN30, similarly to VGENa shown in FIG. The gates of MP30 to MP32 included in each of VG [0] to VG [p] are commonly controlled by enable signals EN1 to EN3, respectively. Similarly, the gates of the MNs 30 included in the VG [0] to VG [p] are commonly controlled by the enable signal VDDEN. The drains of MP30 (the sources of MP31 and MP32) included in each of VG [0] to VG [p] are connected in common, and the word driver power supply voltage WLVDD is output from the common connection node.

  Here, in the configuration example of FIG. 18, the number of word driver power supply circuits (the value of “p” in VG [0] to VG [p]) increases as the number of rows (number of word lines) increases and the number of columns increases. The feature is that the smaller the (number of bit line pairs), the larger. That is, the sizes of the MOS transistors included in VG [0] to VG [p] are assumed to be the same. In the configuration example of FIG. 16 described above, the WLVDD driving capability is adjusted according to the size of each MOS transistor itself. On the other hand, in the configuration example of FIG. 18, the driving capability is adjusted according to the number of word driver power supply circuits. From a circuit perspective, the MOS transistors in the configuration example of FIG. 16 are connected in parallel, and the driving capability is adjusted according to the number of parallel connections. As a result, as described with reference to FIG. 14, it is possible to ensure a sufficient read margin and a sufficient access time regardless of the memory array configuration.

  Further, the method of FIG. 18 can be said to be a method more suitable for a compiled SRAM than the method of FIG. For example, when the method of FIG. 16 is used, preparation of a plurality of layout cells having different transistor sizes may be required, but when the method of FIG. 18 is used, one layout cell may be prepared. In the configuration example of FIG. 18, MP31 and MP32 included in each word driver power supply circuit are, for example, the same threshold voltage as the load transistors (MP1 and MP2 in FIG. 2) included in the memory cell MC. It can be configured to have characteristics. In this case, variations in the threshold voltages of MP1 and MP2 in MC are also reflected in MP31 and MP32 in each word driver power supply circuit, and the voltage level of WLVDD is corrected in accordance with the variations in threshold voltages of MP1 and MP2. can do.

  As described above, by using the semiconductor device of the fourth embodiment, it is possible to improve the operation margins of a plurality of static memory modules included in the semiconductor device, as in the third embodiment. It is also possible to increase the speed of a plurality of static memory modules.

(Embodiment 5)
<< Details of power supply circuit block for word driver (variation [2]) >>
FIG. 19 is a circuit diagram showing a detailed configuration example of the word driver power supply circuit block, the word driver block, and the memory array in each static memory module of FIG. 14 in the semiconductor device according to the fifth embodiment of the present invention. The configuration example shown in FIG. 19 differs from the configuration example of FIG. 18 described above mainly in the output destination of each word driver power supply circuit in the word driver power supply circuit block. Here, the description will be made focusing on this difference. The word driver power supply circuit block VGENb ′ shown in FIG. 19 has (p + 1) word driver power supply circuits VG [0] to VG [0] controlled in common by enable signals EN1 to EN3 and VDDEN, similarly to VGENb of FIG. VG [p] is provided.

  In addition, in the word driver block WLDa ′ shown in FIG. 19, (m + 1) word drivers WD ([0],..., [D], [d + 1],... [2d + 1) are sequentially arranged in the extending direction of the bit line pairs. ], ..., ..., [md], ..., [m]). Power is supplied to (m + 1) WDs by one word driver power supply line WLVDD extending in the extending direction of the bit line pair. Here, there is a connection node for every (d + 1) WDs on WLVDD, and the word driver power supply circuits VG [0] to VG [p] are connected to different nodes in the connection node. Output. That is, VG [0] outputs to a connection node near WD [0], VG [1] outputs to a connection node near WD [d + 1], and thereafter VG [p ] Outputs to a connection node in the vicinity of WD [md].

  In this way, the word driver power supply circuits VG [0] to VG [p] supply power to the nodes distributed at predetermined intervals in the word driver power supply line WLVDD, respectively, thereby supplying power from only one end of WLVDD, for example. Compared with the case of performing the so-called, the so-called far-end difference on WLVDD can be reduced. That is, for example, when the voltage level of WLVDD is lowered using the word driver power supply circuit during a read operation, the arrival time of the voltage level is arranged far from the word driver arranged near the word driver power supply circuit. It can be different for word drivers. In this case, there is a possibility that a difference occurs in the read margin or the like in each memory cell MC in the memory array MARY. Therefore, as described above, such a difference can be reduced by supplying power to distributed nodes.

  Note that the transistor sizes Wp [0] (Wn [0]) to Wp [p] (Wn [p]) in VG [0] to VG [p] may be the same, Depending on the case, the value may be slightly different. That is, even when the output destinations of VG [0] to VG [p] are dispersed in this way, the load balance between the transistor size of each word driver power supply circuit and the transistor size of each word driver, etc. A difference similar to the above-described far-end difference may occur. Therefore, such a difference can be further reduced by appropriately adjusting the size of each transistor in VG [0] to VG [p].

  As described above, by using the semiconductor device of the fifth embodiment, typically, as in the third embodiment, it is possible to improve the operation margins of a plurality of static memory modules included therein. It is also possible to increase the speed of a plurality of static memory modules.

(Embodiment 6)
<< Details of word driver power supply circuit block (variation [3]) >>
FIG. 20 is a circuit diagram showing a detailed configuration example of the word driver power supply circuit block, the word driver block, and the memory array in each static memory module of FIG. 14 in the semiconductor device according to the sixth embodiment of the present invention. The configuration example shown in FIG. 20 is different from the configuration example of FIG. 19 described above in that the word driver power supply line WLVDD is divided into (p + 1) word driver power supply lines WLVDD [0] to WLVDD [p]. Is different. Since the configuration other than this is the same as that of FIG. 19, detailed description thereof is omitted.

  WLVDD [0] is connected to the output of the word driver power supply circuit VG [0] in the word driver power supply circuit block VGENb ′, WLVDD [1] is connected to the output of VG [1] in VGENb ′, and so on. Similarly, WLVDD [p] is connected to the output of VG [p] in VGENb ′. Similarly to the case of FIG. 19, the word driver block WLDb includes (m + 1) word drivers WD ([0],..., [D], [d + 1],... [2d + 1],..., [M− d], ..., [m]) are arranged. However, unlike the case of FIG. 19, power is supplied to each (d + 1) word driver via a different word driver power line. That is, WD [0] to WD [d] are supplied with power through WLVDD [0], WD [d + 1] to WD [2d + 1] are supplied with power through WLVDD [1], and so on. , WD [m−d] to WD [m] are supplied with power via WLVDD [p]. By using such a configuration example, the same effect as in the case of FIG. 19 can be obtained. However, since there may be a characteristic variation for each of VG [0] to VG [p], the configuration example of FIG. 19 that can average the characteristic variation is more desirable from this viewpoint.

  As described above, by using the semiconductor device according to the sixth embodiment, typically, as in the third embodiment, it is possible to improve the operation margin of a plurality of static memory modules included therein. It is also possible to increase the speed of a plurality of static memory modules.

(Embodiment 7)
《Example of power supply circuit for word driver》
FIGS. 21A and 21B are plan views showing a schematic arrangement example of each word driver power supply circuit in the static memory module in the semiconductor device according to the seventh embodiment of the present invention. FIGS. 21A and 21B show examples of the arrangement relationship among the memory array MARY, the word driver block WLD, and the word driver power supply circuit VG in the static memory module SRMD. In FIG. 21A, since there are a large number of rows, a plurality of memory arrays MARY (here, three memory arrays MARY [0] to MARY [0] in the Y-axis direction (the extending direction of the bit line (not shown)) are used. 2]).

  Here, tap regions TAP [0] to TAP [3] are provided on both sides of each memory array in the Y-axis direction. Here, MARY [0] is arranged between TAP [0] and TAP [1], MARY [1] is arranged between TAP [1] and TAP [2], and TAP [2] and TAP [ 3] is arranged between [3] and [3]. Here, the tap region is a region for supplying power to the p-type well and the n-type well included in each memory array. If a memory array with a large number of rows is arranged and power is supplied with tap areas on both sides in the Y-axis direction, sufficient power is supplied near the middle of the memory array in the Y-axis direction. There is a fear of not being broken. Thus, it is beneficial to divide the memory array as shown in FIG. 21A and arrange the tap regions between the divided memory arrays.

  In addition, in the X-axis direction (the extending direction of the word line (not shown)), the word driver block WLD [0] is arranged next to MARY [0]. Similarly, a word driver block WLD [1] is arranged next to MARY [1], and a word driver block WLD [2] is arranged next to MARY [2]. In the Y-axis direction, the sizes of MARY [0] to MARY [2] and the sizes of WLD [0] to WLD [2] are equivalent. Further, in the X-axis direction, the sizes of MARY [0] to MARY [2] and the sizes of TAP [0] to TAP [3] are equivalent. Therefore, an empty area can be secured in an area adjacent to the tap area in the X-axis direction and sandwiched between two word driver blocks in the Y-axis direction. Therefore, the word driver power supply circuits VG [0] to VG [3] are arranged in a distributed manner using this empty area. VG [0] to VG [3] are arranged adjacent to TAP [0] to TAP [3] in the X-axis direction, respectively.

  On the other hand, in FIG. 21B, since the number of rows is small, one memory array MARY [0] is arranged in the Y-axis direction (the extending direction of the bit line (not shown)). As in the case of FIG. 21A, tap regions TAP [0] and TAP [1] are arranged on both sides of MARY [0] in the Y-axis direction. In the X-axis direction, a word driver block WLD [0] is arranged next to MARY [0], and word driver power supply circuits VG [0], VG are arranged next to TAP [0], TAP [1]. [1] is arranged.

  When such an arrangement example is used, it is possible to efficiently realize a method of increasing the number of word driver power supply circuits as the number of rows increases (that is, the configuration examples of FIGS. 19 and 20 described above). Specifically, first, it becomes efficient from the viewpoint of the layout area because the empty area can be used. Also, when the design tool automatically generates the layout of the compiled SRAM, as can be seen from FIG. 21A, for example, areas of WLD [0], MARY [0], VG [0], and TAP [0] As a unit, the regular arrangement corresponding to the number of rows may be performed, so that the processing efficiency can be improved. As described above, it is necessary to reflect the influence of the number of columns in the word driver power supply circuit. This reflection is caused by, for example, the transistors VG [0] to VG [3] in FIG. It is possible to adjust the size as appropriate (that is, the method as shown in FIG. 16).

  FIG. 22 is a plan view showing a schematic layout configuration example of a partial area of the static memory module of FIG. FIG. 22 shows a detailed layout configuration example around VG [1] and TAP [1] in FIG. In FIG. 22, n-type wells NW1 to NW3 and p-type wells PW1 to PW3 are alternately arranged in the order of NW1, PW1, NW2, PW2, NW3, and PW3 in the X-axis direction. In practice, n-type wells and p-type wells corresponding to the number of columns are arranged next to PW3, but are omitted here. A word driver block WLD is formed in NW1, PW1, and a memory array MARY is formed in PW1, NW2, PW2, NW3, PW3,.

In the WLD, a plurality of gate layers GT extending along the X-axis direction are arranged above the NW1 and PW1 (in the Z-axis direction) via a gate insulating film. In the NW1, a p-type semiconductor layer (diffusion layer) DFP is formed on both sides (Y-axis direction) of the plurality of GTs, thereby mounting a plurality of PMOS transistors. In PW1, n-type semiconductor layers (diffusion layers) DFN are formed on both sides of the plurality of GTs, thereby mounting a plurality of NMOS transistors. An n ⁺ type semiconductor layer (diffusion layer) N + extending in the X-axis direction is formed in NW1, and a p ⁺ type semiconductor layer (diffusion layer) P + extending in the X-axis direction is formed in PW1. Is formed. N + is a power supply layer for NW1, and P + is a power supply layer for PW1. The n ⁺ type is set to have a higher impurity concentration than the n type, and the p ⁺ type is set to have a higher impurity concentration than the p type.

  Further, NW1 and PW1 are provided with the above-described word driver power supply circuit formation region VG_AREA. For example, when VG_AREA in FIG. 22 is VG [1] in FIG. 21A, WLD [0] in FIG. 21A is formed on one side sandwiching VG_AREA in FIG. 22 in the Y-axis direction. On the other side, WLD [1] shown in FIG. 21A is formed. In FIG. 22, a layout configuration example in VG_AREA is omitted, but a PMOS transistor and an NMOS transistor are mounted in the same manner as in the case of the word driver WLD described above, thereby forming a predetermined circuit.

  In MARY, here, one memory cell MC is formed by two p-type wells (for example, PW1 and PW2) and one n-type well (for example, NW2) between them. In the MC, two gate layers GT extending along the X-axis direction are arranged on the PW1, and two gate layers GT extending along the X-axis direction are also arranged on the PW2. On the NW2, one of the two GTs on the PW1 and one of the two GTs on the PW2 are continuously extended in the X-axis direction so that two gate layers GT are formed. Is placed. Each GT is actually disposed via a gate insulating film.

  In the PW1, an n-type semiconductor layer (diffusion layer) DFN is formed on both sides of the two GTs, whereby one access transistor (MN3) sharing one end of the source / drain with the DFN and the driver A transistor (MN1) is mounted. Also in the PW2, DFNs are formed on both sides of the two GTs, whereby the other access transistor (MN4) and driver transistor (MN2) sharing one end of the source / drain with the DFN are mounted. In the NW2, a p-type semiconductor layer (diffusion layer) DFP is formed on both sides of two GTs, thereby sharing one MN2 and GT with one load transistor (MP1) that shares MN1 and GT. The other load transistor (MP2) is mounted. Similarly, in MARY, MC is formed using PW2, PW3, and NW3 between them in the X-axis direction, and the gate layer GT and semiconductor layers (diffusion layers) DFN and DFP are sequentially arranged also in the Y-axis direction. MC are sequentially formed.

Furthermore, the above-mentioned tap area TAP is provided in MARY. For example, if the TAP in FIG. 22 is TAP [1] in FIG. 21 (a), MARY [0] in FIG. 21 (a) is formed on one side of the TAP in FIG. MARY [1] in FIG. 21A is formed on the other side. 22 includes a p ⁺ type semiconductor layer (diffusion layer) P + sequentially formed in PW1, PW2, PW3,... And an n ⁺ type semiconductor layer (sequentially formed in NW2, NW3,. Diffusion layer) N +. Each well is powered through the corresponding N +, P +.

  As described above, by using the semiconductor device according to the seventh embodiment, typically, as in the third embodiment, it is possible to improve the operation margins of a plurality of static memory modules included therein. It is also possible to increase the speed of a plurality of static memory modules. Furthermore, such an effect can be obtained more efficiently by devising the layout configuration described above.

(Embodiment 8)
<Outline of memory unit>
FIG. 23 is a schematic diagram showing a configuration example of a memory unit included in the semiconductor device according to the eighth embodiment of the present invention. The memory unit shown in FIG. 23 has a configuration that combines the features of the write assist circuit shown in FIG. 6 and the like with the features of the word driver power supply circuit block shown in FIG. The memory unit shown in FIG. 23 includes a vertically long static memory module SRMD1 and a horizontally long static memory module SRMD2, as in the case of FIG. 6 and FIG.

  Each of SRMD1 and SRMD2 includes write assist circuits WAST1_1 and WAST1_2 described with reference to FIG. Since SRMD1 has more rows than SRMD2 (memory cell power supply line ARVDD is longer), write assist pulse signal WPT having a wider pulse width than WAST1_2 is applied to WAST1_1. WAST1_1 and WAST1_2 use this WPT to control the rate of decrease in the voltage level of ARVDD during the write operation. As a result, the write margin can be improved regardless of the array configuration.

  Furthermore, SRMD1 and SRMD2 respectively include word driver power supply circuit blocks VGEN1 and VGEN2 described with reference to FIG. SRMD1 has a larger number of rows (longer word driver power supply line WLVDD) and a smaller number of columns (shorter word lines WL) than SRMD2, and therefore VGEN1 is set to have a larger size (driving capability) than VGEN2. VGEN1 and VGEN2 lower the voltage level of WLVDD during a read operation (write operation). At this time, since the driving capability when the voltage level of WLVDD is lowered is optimized according to the array configuration, it is possible to improve the read margin and shorten the access time regardless of the array configuration. Become.

  For example, a circuit configuration example including a static part (VSBK) and a dynamic part (VDBK) as shown in FIGS. 7A and 7B is applied to the write assist circuits WAST1_1 and WAST1_2. On the other hand, a circuit configuration example having only a static part as shown in FIG. 16 is applied to the power supply circuit blocks VGEN1 and VGEN2 for word drivers. The static part has a function of switching the output voltage from a certain voltage level to a predetermined voltage level lower than that and supplying the predetermined voltage level stably. The dynamic part is only for switching the voltage level. And has a function to control the switching speed.

  Here, the writing auxiliary circuit and the word driver power supply circuit block conceptually perform a substantially similar operation of lowering the voltage level and controlling the speed of the lowering. It is also possible to apply a dynamic part to the power circuit block for word driver. Alternatively, it is also possible to configure the write assist circuit only with a static part as in the word driver power supply circuit block. However, in applying these configurations, the following difference is essentially generated between the write assist circuit and the word driver power supply circuit block.

  First, it is desirable that the power supply circuit block for the word driver continuously supplies power during the read operation (write operation), but the write auxiliary circuit does not necessarily supply power during the write operation as described in FIG. Need not be continuously supplied. Further, the write assist circuit only needs to have a sufficiently low power supply capability (pull-up capability) for supplying small power required for information retention to the CMOS latch type memory cell. On the other hand, the power circuit block for the word driver supplies power to the word driver that drives the gate layer of the MOS transistor, and the pull-up capability is related to the access time. There is.

  As a result, the power circuit block for the word driver needs to include a static part having a sufficiently high pull-up capability. Therefore, the pull-up capability of the static part (and power supply pull-out capability (pull-down capability) as in the write auxiliary circuit) ) Is fixed and the pull-down capability is reinforced by the dynamic part is not suitable. That is, if the pull-up capability of the static part is fixed, it must be fixed to the high side. For example, when a circuit configuration as shown in FIG. It becomes unnecessary.

  Further, depending on the circuit system, it is possible to reinforce the pull-down capability in the dynamic portion on the premise that the static portion has a sufficient pull-up capability and a somewhat lower pull-down capability. However, the dynamic portion requires a transistor having a size larger than that of the static portion having originally a large size transistor, so that the area efficiency may be reduced. For this reason, the power circuit block for the word driver is composed of only the static part, and the overall driving capability (pull-up and pull-down capability) is adjusted to control the voltage level falling speed as a result. It is desirable to use such a method.

  On the other hand, as described above, since the pull-up capability is not so required for the write assist circuit, it is possible to apply a static portion having a fixed capability regardless of the array configuration. Therefore, a method of adjusting the pull-up capability (and pull-down capability) of the static part like the power supply circuit block for the word driver is applicable, but it is difficult to say that it is efficient from the viewpoint of area and layout design. . As described above, the write auxiliary circuit can be provided with a static portion having a fixed pull-up capability (and pull-down capability), while the required pull-down capability is as described above. It can vary depending on the array configuration. Therefore, it is desirable to apply a method of providing a dynamic part that reinforces pull-down capability according to the array configuration in addition to the static part in the write assist circuit.

  As described above, by using the semiconductor device of the eighth embodiment, typically, the operation margins (write margin, read margin) of a plurality of static memory modules included in the semiconductor device can be improved. It is also possible to increase the speed of a plurality of static memory modules.

  As mentioned above, the invention made by the present inventor has been specifically described based on the embodiments. However, the present invention is not limited to the above-described embodiments, and various modifications can be made without departing from the scope of the invention.

  For example, here, the built-in SRAM mounted on a semiconductor device such as an SOC or a microcomputer has been described. However, the present invention is not necessarily limited to this. ). In addition, a single port SRAM is shown here, but a dual port SRAM or the like may of course be used. In addition, the semiconductor device according to the present embodiment is particularly useful when using the advanced process in which the operation margin is likely to be lowered and using the compiled SRAM. However, the semiconductor device is of course limited to this. Instead, the present invention can be applied to a semiconductor device in which a plurality of SRAM memory arrays are mounted using various processes.

A Address signal ADRCTL Address control circuit APPU Application unit ARVDD Memory cell power supply voltage (memory cell power supply line)
BBU Baseband unit BF Buffer circuit BL, ZBL Bit line C Capacity CEN Clock enable signal CLK Clock signal CPU Processor unit CRSE Read column selection signal CTLBK Control circuit block CWSE Write column selection signal CWSRC Common power supply node DBL Dummy bit line DFP, DFN, N +, P + Semiconductor layer (diffusion layer)
DLY delay element DLYBK delay circuit block Di data input signal Do data output signal EN, VDDEN enable signal GT gate layer IOB input / output buffer circuit IOU input / output unit IV inverter circuit MARY memory array MC memory cell MEMU memory unit MN NMOS transistor MP PMOS transistor ND NAND circuit NW n-type well PW p-type well RWCTL read / write control circuit SA sense amplifier SE sense amplifier enable signal SEL selector circuit SRMD static memory module SW switch TAP tap area TDEC decode start signal TDG write assist timing generation circuit TE fixed voltage VDBK Dynamic part VDDM Power supply voltage VG Word driver power supply circuit V _AREA Word driver power supply circuit formation area VGEN, VGEN 'Word driver power supply circuit block VM voltage level VSBK static part VSS Ground power supply voltage WAST, WAST' Write auxiliary circuit WD Word driver WE Internal write enable signal WEN Write enable signal WL Word Line WLD Word driver block WLVDD Word driver power supply voltage (word driver power supply line)
WPT write auxiliary pulse signal WTD write driver WTE write auxiliary enable signal X row selection signal XDMY row number dummy load circuit XSET row number information Y column selection signal YSW column selection circuit

Claims (6)

A plurality of first word lines extending side by side in the first direction, a plurality of first bit lines extending side by side in a second direction intersecting the first direction, and the plurality of first word lines And a plurality of first SRAM memory cells disposed at intersections of the plurality of first bit lines,
A plurality of second word lines extending side by side in the third direction, a plurality of second bit lines extending side by side in a fourth direction intersecting the third direction, and the plurality of second word lines And a second memory module including a plurality of second SRAM memory cells disposed at intersections of the plurality of second bit lines,
The first memory module further includes:
A plurality of first word drivers for driving the plurality of first word lines;
A first word driver power line that extends in the second direction and supplies power to the plurality of first word drivers;
A first power supply circuit block for reducing a voltage level of the first word driver power supply line with a first drive capability during a read operation;
The second memory module further includes:
A plurality of second word drivers for driving the plurality of second word lines;
A second word driver power line extending in the fourth direction and supplying power to the plurality of second word drivers;
A second power supply circuit block that lowers the voltage level of the second word driver power supply line with a second drive capability during a read operation;
The number of the plurality of first word lines and the number of the plurality of second word lines are substantially equal;
The number of the plurality of first bit lines is greater than the number of the plurality of second bit lines,
The first driving capability is a semiconductor device smaller than the second driving capability. The semiconductor device according to claim 1,
The first power supply circuit block includes N (N is an integer of 2 or more) first power supply circuits that commonly drive the first word driver power supply lines,
The second power supply circuit block includes M (M is an integer of 2 or more) second power supply circuits that commonly drive the second word driver power supply lines,
The first driving capability is determined by the number N.
The second driving capability is determined by the number of M,
The N first power supply circuits are respectively coupled to different positions on the first word driver power supply line,
The M second power supply circuits are respectively coupled to different positions on the second word driver power supply line. The semiconductor device according to claim 2,
The first memory module includes:
Each is sequentially arranged in the second direction, and a predetermined number of first word lines in the plurality of first word lines, the plurality of first bit lines, and the predetermined number of first words. A plurality of memory array regions including a first SRAM memory cell disposed at an intersection of a line and the plurality of first bit lines;
One or more tap regions that are arranged between the plurality of memory array regions in the second direction and supply a predetermined substrate potential to the first SRAM memory cells in the memory array region adjacent in the second direction; ,
A plurality of word driver arrangement regions each including a predetermined number of first word drivers in the plurality of first word drivers, each arranged adjacent to the plurality of memory array regions in the first direction;
A single or a plurality of first arrangement areas arranged between the plurality of word driver arrangement areas in the second direction, and arranged close to the one or more tap areas in the first direction;
A semiconductor device in which one of the N first power supply circuits is formed in each of the one or more first arrangement regions. The semiconductor device according to claim 2,
Each of the N first power supply circuits includes:
A first MIS transistor of a first conductivity type having a source / drain path coupled between a first voltage level and the first word driver power line;
Between the first word driver power supply line and a second voltage level lower than the first voltage level, a source / drain path is coupled in series in order from the first word driver power supply line side. A second MIS transistor and a second conductivity type third MIS transistor,
Each of the M second power supply circuits is
A fourth MIS transistor of the first conductivity type having a source / drain path coupled between the first voltage level and the second word driver power line;
A fifth MIS transistor of the first conductivity type, in which a source / drain path is coupled in series from the second word driver power supply line side in order from the second word driver power supply line side between the second word driver power supply line and the second voltage level; A conductive type sixth MIS transistor;
Before the start of the read operation of the first memory module, the first and second MIS transistors are controlled to be on and the third MIS transistor is turned off, and when the read operation of the first memory module is started, The third MIS transistor is controlled from off to on,
Before the read operation of the second memory module is started, the fourth and fifth MIS transistors are controlled to be on and the sixth MIS transistor is turned off, and when the read operation of the second memory module is started, A semiconductor device in which a sixth MIS transistor is controlled from off to on. 3. The semiconductor device according to claim 2, further comprising:
A processor unit that executes predetermined arithmetic processing using the first and / or second memory module;
A semiconductor device in which the first and second memory modules and the processor unit are mounted on one semiconductor chip. 6. The semiconductor device according to claim 5, further comprising:
Each of the first and second memory modules is a semiconductor device that is a compiled SRAM.

Images