The present application claims priority from Japanese applications JP2003-090212 filed on Mar. 28, 2003, JP2003-172486 filed on Jun. 17, 2003, and JP2004-029033 filed on Feb. 5, 2004 the contents of which are hereby incorporated by reference into this application.
The present invention relates to a semiconductor integrated-circuit device and a method for speeding up CMOS circuit operation, and more particularly to a technology advantageously used for speeding up the operation of semiconductor integrated-circuit devices comprising CMOS circuits.
In Japanese Patent Laid-open No. 11-195976, the present applicant has already proposed a MOSFET-constructed semiconductor integrated-circuit device in which there is attained a preferred harmony between operating speed and increases in power consumption due to leakage currents. According to the above-mentioned publication, among the plurality of signal paths in the semiconductor integrated-circuit device, a signal path having a margin for delay with which a signal is transferred along the signal path is constituted by MOSFETs with high threshold voltage. Conversely, a path not having a margin for delay is constituted by low-threshold-voltage MOSFETs which, although large in sub-threshold leakage current, operate at high speed. Means for achieving a high threshold voltage and a low threshold voltage in the MOSFETs as mentioned above can be obtained by changing the density of the impurities under the gate oxide film of the semiconductor substrate, by changing the thickness of the gate oxide film, by changing the substrate bias voltage applied to a well region, by changing the gate length, and by combining these methods. Also, a semiconductor integrated-circuit device that uses high-withstand-voltage MOSFETs and high-threshold-voltage MOSFETs in its input/output circuits is described in Japanese Patent Laid-open No. 2001-015704.
The description in the above-mentioned publications is based only on the recognition that a harmony between operating speed and minimum power consumption is attained by utilizing the characteristics and features of CMOS circuits. Accordingly, speeding-up has its limits, and to achieve further speeding-up, bipolar transistors need to be used. Using bipolar transistors to construct a circuit, however, poses major problems in terms of power consumption and integration density.
An object of the present invention is to provide semiconductor integrated-circuit devices capable of achieving both higher-density circuit integration and faster operation. Another object of the present invention is to provide a CMOS circuit operational speeding-up method by which the operating speeds of CMOS circuits, including existing ones, can be easily increased. The above objects, other objects, and new features of the present invention will be more fully understood from the description of this specification when reference is made to the accompanying drawings.
Among all aspects of the invention disclosed in the present application, a typical one is briefly described below. That is to say, a signal transferring path includes a plurality of CMOS-constructed logic gate circuits provided between one pair of flip-flop circuits for acquiring and holding signals by use of clock signals. The signal transferring path includes a first and a second signal transferring path. The first signal transferring path is constituted by enhancement-type MOSFETs and has a signal transferring delay time equal to, or less than, a permissible signal transferring delay time. The second signal transferring path is configured such that, among the above-mentioned plurality of logic gate circuits, a logic gate circuit having a delay time longer than the above-mentioned permissible signal transferring delay time when constituted using enhancement-type MOSFETs is replaced with a depletion-type MOSFET so that the second signal transferring path may provide a signal transferring delay time equal to or less than the permissible signal transferring delay time mentioned above.
Among all aspects of the invention disclosed in the present Application, another typical one is briefly described below. That is to say, a designing step is repeated so that the signal transferring delay time of all the signal transferring paths may stay within a permissible signal transferring delay time range. The designing step includes: designing a signal-processing circuit using enhancement type MOSFETs, the signal-processing circuit comprising a plurality of flip-flop circuits for acquiring and holding signals by use of clock signals, and a plurality of CMOS-constructed logic gate circuits provided between one pair of flip-flop circuits within the plurality of flip-flop circuits; extracting, from the plurality of signal transferring paths, a signal transferring path whose signal transferring delay time that exceeds a permissible signal transferring delay time; and replacing, among the plurality of logic gate circuits constituting the signal transferring path that has extracted, a logic gate circuit having a delay time longer than the permissible signal transferring delay time when the logic gate circuit is constituted by an enhancement-type MOSFET with a depletion-type MOSFET so that the signal transferring path may provide a signal transferring delay time equal to or less than the permissible signal transferring delay time.
In step (1) of
In step (2) of
In step (3) of
In step (4) of
In step (5) of
In step (6) of
In step (3) of
In step (4) of
In step (5) of
In step (5) mentioned above, if the delay of the above single path is judged to be equal to,.or smaller than, the target value, all the remaining paths are also judged in step (7) whether they are equal to or smaller than the target value. If there still exist any paths whose delays are greater than the target value, processing returns to step (2) and the same steps are repeated.
In step (7), if the delays of all paths are judged to be equal to or smaller than the target value, the delays in each path are re-simulated in step (8) for confirmation.
As shown in
In step (1), the signal-processing circuits that conduct desired digital signal processing do not need to be newly designed ones and can be existing CMOS circuits. For example, in the case of a currently operating circuit or of, as with a microprocessor that has already been developed as an old-generation one, a macro-structured CMOS circuit, if the operating speed of the circuit is too slow and the design and development of a new circuit are required only for this reason, circuit operation can be easily speeded up by using the design data intact and applying a CMOS circuit operational speeding-up method pertaining to the present invention. With such a CMOS circuit operational speeding-up method, circuit operation can be speeded up without special circuit debugging being required, since the circuit function itself for providing digital signal processing, such as the function itself of a microprocessor circuit, has already been proved to operate properly at low speed.
Here, the MOSFETs that are ultra-low-Vth cells are constructed of depletion-type MOSFETs. The MOSFETs generally called “depletion-type MOSFETs”, for example, N-channel MOSFETs refer to MOSFETs having a negative threshold voltage Vthn, and P-channel MOSFETs refer to MOSFETs having a positive threshold voltage Vthp. As shown in
In general, a greater leakage current occurs in MOSFETs lower in Vth value, and the respective leakage currents of ultra-low-Vth and low-Vth MOSFETs per unit gate width are, respectively, about 100 times and 10 times the leakage currents of standard-Vth MOSFETs. Accordingly, if a multitude of low-Vth MOSFETs are used, their leakage currents exceed a permissible value. Also, if the leakage current value of the entire chip is suppressed to a certain permissible value, the number of MOSFETs which can be used decreases as the Vth value becomes smaller. Conversely, however, the rate of contribution to speeding-up is increased since the drain current increases with decreases in Vth value.
In step (3) of
The delay time of the cells that are changed from the above-mentioned standard-Vth cells to ultra-low-Vth cells is reduced to 0.6 times the delay time of the standard-Vth cells. This value is selected from the relative relationship between the standard-Vth cells and ultra-low-Vth cells that are set as described above. This relationship can therefore be modified by varying the respective threshold voltages of the ultra-low-Vth cells and the standard-Vth cells. Also, at such threshold voltages as shown in the above example, if the standard-Vth cells are replaced with low-Vth cells, the delay time may be reduced to about 0.8 times that of the standard-Vth cells.
Table 1 shown below indicates the relationships between the respective operating frequencies and standby currents existing when MOSFETs including only standard-Vth cells, MOSFETs including only low-Vth cells, MOSFETs including only ultra-low-Vth cells, MOSFETs including standard-Vth cells and low-Vth cells (30%), and MOSFETs including standard-Vth cells and ultra-low-Vth cells (2%) are mounted in specific independent digital logic circuits studied by the present inventor. As shown in Table 1, even if the standard-Vth MOSFETs are replaced with low-Vth ones, the operating frequency ratio can only be increased to 1.25 and cannot be improved too significantly. Also, although replacement of all standard-Vth MOSFETs with ultra-low-Vth ones greatly increases the operating frequency ratio to 1.75, the leakage current ratio increases to as excessively high as 220, and therefore, this method is not realistic.
Therefore, when standard-Vth MOSFETs and low-Vth MOSFETs are combined, although the leakage current ratio can be improved to a certain extent over that of low-Vth-only MOSFETs, the operating frequency ratio that is the more important of the two factors can only be improved to almost the same extent as that of low-Vth-only MOSFETs. However, combination of standard-Vth MOSFETs and ultra-low-Vth MOSFETs makes it possible not only for the operating frequency ratio to be improved to a level almost comparable to that of ultra-low-Vth-only MOSFETs, but also for the leakage current ratio to be decreased to a level slightly higher than in the case of the above-mentioned combination of standard-Vth MOSFETs and ultra-low-Vth MOSFETs. Increases in the leakage current ratio can be reduced by changing the substrate bias described later.
A schematic block diagram of an embodiment of a semiconductor integrated-circuit device to which the present invention is applied is shown in
In this embodiment, the semiconductor integrated-circuit device, although not limited specifically, is constructed of five types of MOSFETs different in threshold voltage Vth. In addition to the ultra-low-Vth, standard-Vth and low-Vth MOSFETs used for the digital logic circuit block, there are provided two types: low-Vth I/O cells and standard-Vth I/O cells. The low-Vth I/O cells and the standard-Vth I/O cells are used as high-withstand-voltage MOSFETs since they have a thick gate-insulation film so that a high voltage can be applied to the gate.
The digital logic circuit in this embodiment is shown with focus being placed on the output signal “x”. Therefore, although the number of fan-outs in each gate circuit is one, the output signals of each logic stage, including the above-mentioned input signals “a”, “b”, “c”, “d”, “e”, and “f”, may actually be output to other logic gate circuits not shown in the figure. The inverter circuits and the logic gate circuits, both large in the number of fan-outs, increase in load capacity, thus prolonging the signal transferring delay time in these circuits. These gate circuits with a prolonged signal transferring delay time are changed from standard-Vth cells to ultra-low-Vth cells as described earlier, and hereby, the signal transferring delay time to the acquisition of the output signal “x” is controlled to or below a target value.
In this way, the digital logic circuit is constituted by the above-mentioned combination of standard-Vth cells and ultra-low-Vth cells. That is to say, basically, this circuit is constituted by combining, as described earlier, enhancement-type MOSFETs having standard-Vth cells with depletion-type MOSFETs having ultra-low-Vth cells. The above-mentioned enhancement-type MOSFETs, however, may also be able to include low-Vth cells, provided that the target delay value of the path can be attained by combining standard-Vth MOSFETs with low-Vth MOSFETs.
As described above, ultra-low-Vth MOSFETs (depletion-type) and standard-Vth MOSFETs are used in the digital logic circuit block. These two types of MOSFETs are effectively used for each logic cell. Since the leakage current in a MOSFET exponentially increases with respect to its Vth value, a large leakage current occurs in a MOSFET whose Vth value is reduced to provide the depletion type. Accordingly, using depletion-type MOSFETs may undesirably increase a standby current or create a thermally uncontrollable state, and therefore, ultra-low-Vth MOSFETs of the depletion type have not been used in conventional CMOS circuits. However, since the cells that use ultra-low-Vth MOSFETs can be speeded up more significantly than in the case of standard-Vth or low-Vth MOSFETs, the speeds of critical paths can be sufficiently increased. For this reason, leakage currents can be minimized for faster operation by limiting the application of depletion-type ultra-low-Vth MOSFETs only to critical paths. Accordingly, the digital logic circuit block can use ultra-low-Vth and standard-Vth MOSFETs to speed up circuit operation and to minimize leakage currents.
Analog circuits include low-Vth and standard-Vth cells. For example, in such a differential circuit as shown in
When analog circuits are operated at low voltage, if Vth is too high, cascade-connected circuits are liable to become inoperative. Therefore, for such a cascade-connected MOSFET circuit as shown in
As shown in the block diagram of
Accordingly, the memory array block uses standard-Vth MOSFETs to minimize non-operating power consumption and to ensure sufficient operating margins and a higher yield. For memory peripheral circuits, more particularly for the address-decoding circuit and other memory peripheral circuits shown in
Here, the above main amplifier and the output drivers use such a tri-state buffer as shown in
For such a memory section, it takes a very long designing period if that circuit is closely designed taking the signal transferring speed and power consumption into consideration. In addition, in spite of memories also usually having their use diverted to other sections as a design asset, this is made impossible by such a designing task. Accordingly, closely designing with the signal transferring speed and power consumption being taken into consideration is usually not realistic. Therefore, it becomes possible, by unifying the threshold voltages Vth of the MOSFETs for each block such as the memory array or X-driver, to reduce the designing period and to facilitate diverted use as a design asset.
Accordingly, a switch composed of MOSFETs Q42 and Q43, and a back-bias switch composed of MOSFETs Q44 and Q45 are provided for the well regions in which the MOSFETs Q40 and Q41 in the logic circuit block are formed. That is, during operation, the MOSFETs Q42 and Q43 are turned on, and a voltage VD1 is supplied to the well region in which the P-channel MOSFET Q40 is formed, and a grounding potential GND of the circuit is supplied to the well region in which the N-channel MOSFET Q41 is formed. The above voltage VD1 is the same as the operating voltage applied to the source electrode of the P-channel MOSFET Q40 in the inverter circuit.
Under the standby state in which the above logic circuit does not operate at all, the MOSFETs Q42 and Q43 are turned off, and a voltage VD2 is supplied to the well region in which the P-channel MOSFET Q40 is formed, and a negative voltage VB of the circuit is supplied to the well region in which the N-channel MOSFET Q41 is formed. Since the above-mentioned VD1 and VD2 are maintained in the relationship of VD2>VD1, a back-bias relationship is established between the source electrode of the P-channel MOSFET Q40 and the well region and likewise, a back-bias relationship is established between the source electrode of the N-channel MOSFET Q41 and the well region. Consequently, the effective threshold voltages of these MOSFET Q40 and MOSFET Q42 are increased by a substrate effect, thereby allowing the above DC current to be reduced significantly.
Referring to
As a result of the design using the method of
PLC-C is a PCI control unit that controls data exchange with a PCI path. The VLIW-core is a core CPU that uses a programmed control scheme to execute required arithmetic processing and control the entire functional block. The Ib section in the VLIW-core is an instruction control unit having an instruction cache and controlling instructions, Db is a data control unit having a data cache and controlling data, and Eb executes arithmetic processing based on the instruction commands stored within the instruction cache of Ib. Jtag is a circuit for a Jtag interface, and the PLLs section constitutes one or more circuit blocks for supplying a progressively multiplied reference clock to the entire functional block. Vf0 and Vf1 are circuits that provide image data scaling (enlarging/reducing).
IIS-C is a control unit for an interface based on JIS standards. IIC-C is a control unit circuit for an interface based on IIC standards. IEC-C is a control unit circuit for an interface based on IEC Standard 958. ROM-C is a control unit circuit for an external ROM flash interface. SC is a control unit circuit for a serial interface. General-purpose I/O is a general-purpose input/output unit circuit. DES and Multi2 are encryption circuits. TCIIN1 and TCIIN0 are input control circuits for interfacing with data based on TCI (Transport Channel Interface) standards.
NTSCIN1 and NTSCIN0 are input control circuits for interfacing with data based on ITU Standard 656. GPDP is a general-purpose communications unit circuit. TCIOUT is an output control circuit for interfacing with data based on TCI (Transport Channel Interface) standards. NTSCOUT1 and NTSCOUT0 are output control circuits for interfacing with data based on ITU Standard 656. VLx is a variable-length code-processing circuit. DRC is a circuit for data display on an external display device.
Ds is a circuit for controlling data transfer within the chip, and it is one type of DMAC (Direct Memory Access Controller) for transferring data to the required location automatically and continuously. Mb is a control circuit for a memory interface, and Ma is a circuit for performing motion-compensating and motion-detecting processes during decoding and encoding. DAC is a digital-analog converter. There also exist memories other than the Ib and Db described above. A memory is present in Vf0, Vf1, DES, Multi2, VLx, Ds, Mb, and Ma each. Each of these circuits has a coprocessor for providing arithmetic control in order to perform the above-mentioned functions, and each memory exists as a cache for the coprocessor. These memories correspond to
M1 is a constant-current source, M2, M3, M7, M11 are differential amplifiers, and the M2 and the M3 constitute a differential amplification circuit. MB1, MB3, and M1 are current mirrors with respect to a potential of nb1, MB4, MB7, M6, and M10 are current mirrors with respect to a potential of nb2, MB5, MB8, M7, and M11 are current mirrors with respect to a potential of nb3, MB6, M5, and M9 are current mirrors with respect to a potential of nb4, and M4 and M8 are current mirrors with respect to a potential of nb5. The MB3, the MB4, the MB7, MB5, and the MB8 form such a cascade-connected circuit as in
Since these current mirrors have their MOSFETs multi-stage-stacked at the potential between ADSS and AVSS, if the ADSS-AVSS potential is too low, using MOSFETs with high Vth makes the MOSFETs inoperative. Conversely, using MOSFETs with too low Vth results in a too small gain. For these reasons, low-Vth MOSFETs are used. At the differential amplifiers, the Vth of the MOSFETs as shown in
Incidentally,
Although the invention that was made by the present applicant has been heretofore described in detail on the basis of embodiments, the present invention is not limited by these embodiments, and needless to say, a variety of changes may be made without departing from the scope of the invention. For example, the digital integrated circuits may be composed of random logic circuits, other gate arrays, or the like. The substrate bias voltage may be formed inside the semiconductor integrated-circuit device via a charge pump circuit or may be a voltage supplied from external terminals. The present invention can be widely used for semiconductor integrated-circuit devices composed of CMOS circuits, and for a method of speeding up the operation of these circuits.
A signal transferring path includes a plurality of CMOS-constructed logic gate circuits provided between one pair of flip-flop circuits for acquiring and holding signals by use of clock signals. The signal transferring path includes a first and a second signal transferring path. The first signal transferring path is constituted by enhancement-type MOSFETs and has a signal transferring delay time equal to, or less than, a permissible signal transferring delay time. The second signal transferring path is configured such that, among the above-mentioned plurality of logic gate circuits, a logic gate circuit having a delay time longer than the above-mentioned permissible signal transferring delay time when constituted using enhancement-type MOSFETs is replaced with a depletion-type MOSFET so that the second signal transferring path may provide a signal transferring delay time equal to or less than the permissible signal transferring delay time mentioned above. Thus, both higher-density circuit integration and operational speeding-up can be achieved.
A designing step is repeated so that the signal transferring delay time of all the signal transferring paths may stay within a permissible signal transferring delay time range. The designing step includes: designing a signal-processing circuit using enhancement type MOSFETs, the signal-processing circuit comprising a plurality of flip-flop circuits for acquiring and holding signals by use of clock signals, and a plurality of CMOS-constructed logic gate circuits provided between one pair of flip-flop circuits within the plurality of flip-flop circuits; extracting, from the plurality of signal transferring paths, a signal transferring path whose signal transferring delay time that exceeds a permissible signal transferring delay time; and replacing, among the plurality of logic gate circuits constituting the signal transferring path that has extracted, a logic gate circuit having a delay time longer than the permissible signal transferring delay time when the logic gate circuit is constituted by an enhancement-type MOSFET with a depletion-type MOSFET so that the signal transferring path may provide a signal transferring delay time equal to or less than the permissible signal transferring delay time. Thus, CMOS circuit speeding-up can be accomplished.
The application of the present invention allows higher-density integration of semiconductor integrated circuits and the speeding-up of their operation to be implemented by using depletion-type MOSFETs in respective digital logic circuits.
Reference symbols are as follows:
(1) to (8) . . . Design steps, FF1, FF2 . . . Flip-flop circuits, a to f . . . Input signals, x . . . Output signal, Q1 to Q59 . . . MOSFETs.
Number | Date | Country | Kind |
---|---|---|---|
2003-090212 | Mar 2003 | JP | national |
2003-172486 | Jun 2003 | JP | national |
2004-029033 | Feb 2004 | JP | national |
Number | Name | Date | Kind |
---|---|---|---|
6222410 | Seno | Apr 2001 | B1 |
6380764 | Katoh et al. | Apr 2002 | B1 |
6577153 | Kodama | Jun 2003 | B1 |
6769110 | Katoh et al. | Jul 2004 | B1 |
20020027256 | Ishibashi et al. | Mar 2002 | A1 |
Number | Date | Country |
---|---|---|
11-195976 | Jul 1999 | JP |
2001-15704 | Jan 2001 | JP |
Number | Date | Country | |
---|---|---|---|
20040223401 A1 | Nov 2004 | US |