WO2014036288A1 - Controlling impedance of a switch using high impedance voltage sources to provide more efficient clocking - Google Patents
Controlling impedance of a switch using high impedance voltage sources to provide more efficient clocking Download PDFInfo
- Publication number
- WO2014036288A1 WO2014036288A1 PCT/US2013/057331 US2013057331W WO2014036288A1 WO 2014036288 A1 WO2014036288 A1 WO 2014036288A1 US 2013057331 W US2013057331 W US 2013057331W WO 2014036288 A1 WO2014036288 A1 WO 2014036288A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- voltage
- switch
- gate
- clock signal
- clock
- Prior art date
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F1/00—Details not covered by groups G06F3/00 - G06F13/00 and G06F21/00
- G06F1/04—Generating or distributing clock signals or signals derived directly therefrom
- G06F1/10—Distribution of clock signals, e.g. skew
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F1/00—Details not covered by groups G06F3/00 - G06F13/00 and G06F21/00
- G06F1/04—Generating or distributing clock signals or signals derived directly therefrom
Definitions
- Disclosed embodiments relate to clocking systems and more particularly to switches used to switch between operational modes of the clocking system.
- Power consumption is a critical area of circuit design. That is particularly true for battery operated mobile systems. Many sources of undesired power loss can be found in integrated circuits such as leakage current in transistors and power loss due to undesirable resistive or capacitive loading. While power consumption is critical in mobile battery operated devices, power consumption is also important in other settings such as server farms where the power savings for each processor is multiplied many-fold. Thus, improving efficiency in integrated circuit operation is desirable to extend battery life and more generally to reduce energy consumption.
- an apparatus includes a switch coupled to receive a clock signal at an input node of the switch and supply the clock signal at an output node of the switch when the switch is closed.
- a first high impedance voltage source is coupled to supply a first voltage to a first gate node of a first transistor of the switch.
- a method includes supplying a first voltage from a first high impedance voltage source that is coupled to a gate of a first transistor of the switch.
- a second voltage is supplied from a second high impedance voltage source that is coupled to a gate of a second transistor of the switch.
- a clock signal is supplied to respective first current carrying nodes of the first and second transistors, and respective second current carrying nodes of the first and second transistors supply the clock signal to an inductor when the switch is closed.
- a clock system of an integrated circuit includes a switch that has a first and second transistor.
- the switch receives a clock and supplies the clock signal to an inductor when the switch is closed.
- the inductor forms a resonant circuit with capacitance of a clock system of the integrated circuit when the switch is closed and is disconnected from the capacitance of the clock system when the switch is open.
- First and second high impedance voltage sources supply respective first and second voltages to the switch such that a first near constant overdrive voltage is maintained for a first gate node of the first transistor and a second near constant gate overdrive voltage is maintained for a second gate node of the second transistor.
- low conducting resistance is achieved using a first circuit to selectively couple an inductor to a clock system.
- the circuit includes a first transistor coupling a first terminal of the inductor to a first supply voltage and a second transistor coupling the first terminal of the inductor to a second supply voltage.
- Fig. 1 illustrates a simplified model of a dual mode clocking system which supports both resonant and conventional clocking, according to some embodiments.
- Fig. 2 shows example clock voltage and mode switch current waveforms, according to some embodiments.
- Fig. 3 illustrates a simplified model of a mode switch, according to some embodiments.
- Fig. 4 illustrates mode switch sizing tradeoffs, according to some
- Fig. 5 illustrates a simplified schematic using high-impedance voltage sources to drive the gate terminals of the mode switch, according to some embodiments.
- Fig. 6A shows nFET and pFET gate voltages with respect to clock voltage when the mode switch is closed, according to some embodiments.
- Fig. 6B shows nFET and pFET gate voltages with respect to clock voltage when the mode switch is open, according to some embodiments.
- Fig. 7 illustrates an exemplary switched-capacitor implementation of a high impedance voltage generator to generate a gate voltage for a switch, according to some embodiments.
- Fig. 8 illustrates an exemplary switched-capacitor implementation of a high impedance voltage generator to generate a gate voltage for a switch, according to some embodiments.
- Fig. 9 illustrates an embodiment to achieve low conducting resistance based on the location of the mode switch.
- Fig. 1 illustrates a simplified model of a resonant clock system 100, according to some embodiments.
- the resonant clock system 100 may be part of a
- a distinct feature of the resonant clock system 100 is the use of an inductance (L) 101 that is connected in parallel with the parasitic network capacitance (C) 103.
- the role of the clock driver 102 is to replenish only the energy lost in the parasitic resistance of the LC system formed by the inductor 101 and parasitic network 103.
- Resonant clocking thus relies on efficient LC resonance to achieve energy efficiency. For an ideal inductor and interconnect, with no parasitic resistance, the clock network would oscillate with zero energy dissipation. Resonant clocking is most efficient at frequencies close to the resonant frequency.
- Mode switch 104 is closed during resonant operation (rclk) to form an LC circuit and open in conventional mode (cclk) so as to disconnect the inductor from the clock system.
- mode switch 104 MSw
- mode switch 104 is located as shown in Fig. 1. Using mode switch 104 provides the required functionality of a dual-mode clock, but increases the resistance of the LC tank, reducing resonant clocking efficiency.
- Fig. 2 shows example clock voltage 201 and current waveforms 203 associated with mode switch 104.
- the mode switch 104 includes a pFET device 107 and an nFET device 109.
- the clock voltage transitions from ground to Vdd during its cycle, and as a result, the nFET (pFET) device sees a substantially reduced gate overdrive voltage when the clock signal approaches Vdd (ground), contributing to the increased resistance.
- the gate overdrive voltage is the voltage between the gate and source (Vgs) in excess of the threshold voltage.
- Fig. 3 shows a simplified diagram of the mode switch.
- Mode switch 104 includes the gate to drain capacitances (Cgd) 301 and 302 and the gate to source capacitances (Cgs) 303 and 304.
- the switch 104 includes resistance shown as resistance 305.
- a larger switch increases the capacitive loading on the clock network, increasing not only the dissipation in the conventional mode, but also in the resonant mode, due to the higher current flow through the resistance of mode switch 104 and the inductor 101.
- Fig. 4 illustrates the tradeoff between the switch resistance and the capacitive loading of the switch, which results in significant inefficiency of the overall system.
- the tradeoff relates to switch size. Smaller switches offer higher conventional clock efficiency due to reduced loading, e.g., at 401 as compared to 403. Larger switches offer improved resonant clocking performance up to a certain point, e.g., at 407 as compared to 405. Thus, trying to achieve efficiency for one mode of operation, e.g. resonant clocking, sacrifices efficiency in the non-resonant mode of operation, and vice versa.
- One existing solution to the inefficiency resulting from the tradeoff between the mode switch resistance and the capacitive loading offered to the clock network is to use low threshold voltage (V th ) devices. That technique, however, does not offer significant resistance reduction, especially at higher voltages where power efficiency is particular important to obtain improved performance.
- Another problem associated with the use of the modes switch is leakage current.
- Using a low on-resistance mode switch implies a higher leakage current through the mode switch when it is off.
- the use of low V th devices reduces the on/off current ratio, increasing the leakage through the mode switch when the switch is open.
- the higher leakage through the switch results in additional undesirable power dissipation.
- some embodiments address the disadvantages of low Vth devices by improving the pareto-efficiency of the mode switch 104 by significantly reducing the on-resistance of the mode switch 104 for a given switch width.
- the reduced resistance can be used for improved efficiency in resonant mode, or be traded off for improved conventional mode clocking through mode switch width reduction. It also allows for reduced leakage through the mode switch when the switch is open.
- Fig. 5 shows a simplified schematic of an embodiment that reduces the on- resistance of the mode switch 104 for a given switch width and provides for reduced leakage through the mode switch when the switch is open.
- high impedance voltage sources Vgn 501 and Vgp 503, are used to allow the gate voltage to transition along with the source/drain voltage of the mode switch 104.
- Use of the high impedance voltage sources allows for better pareto-efficiency with regard to the design of the mode switch 104.
- the use of the high impedance voltage source also allows the added benefit of reduced loading on the clock network while in conventional mode since the gate terminals can be thought of as floating terminals in the AC sense, and in the best- case scenario, the loading on the clock node goes from 2Cdg (as per Fig. 3) down to Cdg.
- the assumption is that Cdg is approximately equal to Csg. If the gate is floating and the mode switch is turned off, there are two parallel paths made up of two series capacitors. Each of these paths is approximately Cdg/2 in capacitance and thus, the resultant capacitance is Cdg.
- Fig. 6A shows waveforms of the applied gate voltage of the nFET and the pFET device of the mode switch and the clock signal when the mode switch 104 is closed in resonant mode. Note how the gate signals 601 and 603 transition along with the clock signal 605 maintaining a near-constant gate overdrive in the conducting transistor, while ensuring that reliability limits are not exceeded.
- Nearly or substantially constant means that the gate overdrive voltage does not vary by more than, e.g., approximately 10% or less in a typical embodiment depending on the needs of the system.
- the amount of change in the gate overdrive depends on the impedance of the supply. If the supply impedance is infinitely high, the gate overdrive will actually be almost constant (e.g., less than 1% variation).
- some embodiments may have gate overdrive that varies by more than 10% but the variation is sufficiently small to improve the conductivity of the mode switch.
- the gate voltage when using the high impedance voltage source, the gate voltage, e.g., shown in gate signal 601, tracks the clock signal 605.
- Fig. 6A shows the gate voltage transitioning in synchronism with the clock signal around the voltage level 602, which is supplied by the power supply Vgn 501.
- the gate signal tracks the clock signal due to the impedance seen by the source and drain into the gate, past the Cgd and Cgs (see Fig. 3). If for example the impedance was low, e.g., a zero impedance, then the gate would just be tied to that voltage, regardless of the gate and source voltage transitions.
- Fig. 6B shows the operation of the mode switch 104 with high impedance voltage sources when the mode switch 104 is open. Note that the gate signals 611 and 613 transition along with the clock signal 615 maintaining a near-constant gate overdrive in the conducting transistor, while ensuring that reliability limits are not exceeded. The use of the high impedance voltage sources dramatically reduces leakage while in the open state shown in Fig. 6B.
- the high impedance voltage source can be generated by several techniques, one of which is the use of charge pumps to achieve the desired voltage levels.
- Fig. 6A shows a DC level of 1.5Vdd for the gate voltage (Vgn) for the nFET and a DC level of -0.5Vdd for the gate voltage (Vgp) for the pFET
- the appropriate voltage settings for the voltage supplies can be selected based on the requirements for a particular design.
- Fig. 6B note that the voltage swing of waveforms 611 and 613 on the gate terminals is lower when the mode switch is open. That is expected since the side of the switch connected to the inductor does not toggle, and instead serves to reduce the amplitude of the coupled voltage.
- V gn and Vgp should be set according to reliability constraints.
- Figs. 7 and 8 show, respectively, embodiments to achieve the voltages of 1.5 Vdd and -0.5 Vdd with a high output impedance.
- the charge pump embodiment of Fig. 7 utilizes a clock signal to alternate the open/closed state of the switches in phase 1 and phase 2 illustrated in Fig. 7.
- switches 701, 703, and 706 are closed to charge capacitors 707 and 709, which are configured in series.
- switches 711, 713, 715, and 717 are closed configuring Vdd in series with the parallel capacitors 707 and 709, allowing a voltage to be supplied that is 1.5 Vdd.
- a lower frequency clock may be used to toggle the switches between the two phases because output impedance is inversely related to switch frequency.
- Fig. 8 illustrates a charge pump embodiment that utilizes a clock signal to alternate the open/closed state of the switches in phase 1 and phase 2.
- switches 801, 803 and 805 are closed to charge capacitors 807 and 809 in series.
- switches 813, 815, and 817 are closed allowing a voltage to be supplied from capacitors 807 and 809 configured in parallel that is -0.5 Vdd.
- the particular voltage values in Figs. 7 and 8 are exemplary and other voltage values may be chosen to be supplied according to the needs of the particular system.
- Embodiments described herein offer several advantages including reduced on- resistance of the mode switch 104 while operating within reliability constraints of the device and increased off -resistance (lower leakage) of the mode switch 104 while operating within reliability constraints of the device.
- the dramatically lower leakage also allows for the use of even lower Vth devices for the mode switch while still avoiding substantial power loss through leakage current.
- the use of a high-impedance gate driver for the mode switch reduces the capacitive loading on the clock node, allowing for improved energy efficiency. Reduced on-resistance is achieved through the higher gate overdrive on both the nFET and the pFET devices of the mode switch 104.
- the mode switch 104 When the mode switch 104 is open, the devices operate in deep-cutoff with drastically lower leakage as compared to conventional techniques. That improves energy efficiency of the clock system in conventional mode.
- the high impedance power supplies are shared across a plurality of the mode switches.
- the number of mode switches that share a high impedance power supply is determined according to system requirements such as the ability of the high impedance power supplies to supply a particular number of switches. Note also than when switching from resonant mode to non- resonant mode, the power supplies are switched by a switch (not shown in Fig. 5) so -0.5 Vdd goes to the nFET and 1.5 Vdd goes to pFET.
- Some embodiments address the problem of increased I 2 R losses due to the lower gate overdrive in the mode switch by relocating the mode switch 104 so that the nFET and pFET devices of the mode switch 104 conduct with a higher gate overdrive, and in doing so offer a lower conducting resistance for a given mode switch width.
- the higher overdrive is present when the pFET 901 is on in resonant mode since the drain and source voltages are near Vdd while the gate voltage is low.
- the gate voltage is high while the source and drain are low.
- Fig. 9 shows a simplified representation of such an embodiment.
- the mode switch is implemented as a header-footer combination using the pFET 901 and nFET 903.
- the FETs 901 and 903 are turned on in resonant mode and turned off in non- resonant mode. Note that the FETs 901 and 903 use with-in supply rail voltages in contrast to the FETS 502 and 504 shown in Fig. 5, which have boosted supply voltages from the charge pumps shown in Figs. 7 and 8.
- Capacitors 907 and 909 may be included to ensure that the n2 node of inductor 101 is connected to ground in an AC sense.
- circuits and physical structures have been generally presumed in describing embodiments herein, it is well recognized that in modern semiconductor design and fabrication, physical structures and circuits may be embodied in a computer readable medium as data structures for use in subsequent design, simulation, test, or fabrication stages.
- data structures may encode a functional description of circuits or systems of circuits.
- the functionally descriptive data structures may be, e.g., encoded in a register transfer language (RTL), a hardware description language (HDL), in Verilog, or some other language used for design, simulation, and/or test.
- Data structures corresponding to embodiments described herein may also be encoded in, e.g., Graphic Database System II (GDSII) data, and functionally describe integrated circuit layout and/or information for photo-mask generation used to manufacture the integrated circuits.
- GDSII Graphic Database System II
- Other data structures, containing functionally descriptive aspects of embodiments described herein, may be used for one or more steps of the manufacturing process.
- Computer-readable media include tangible computer readable media, e.g., a disk, tape, or other magnetic, optical, or electronic storage medium.
- the computer readable media may store instructions as well as data that can be used to implement embodiments described herein or portions thereof.
- the data structures may be utilized by software executing on one or more processors, firmware executing on hardware, or by a combination of software, firmware, and hardware, as part of the design, simulation, test, or fabrication stages.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Dc-Dc Converters (AREA)
- Electronic Switches (AREA)
- Nonlinear Science (AREA)
- Logic Circuits (AREA)
- Power Sources (AREA)
Abstract
A clock system of an integrated circuit includes first (502) and second (504) transistors forming a switch that is used when switching the clock system (Clk) between a resonant mode of operation and a non-resonant mode of operation. An inductor forms a resonant circuit with capacitance of the clock system in resonant mode. The switch receives a clock signal (Clk) and supplies the clock signal to the inductor when the switch is closed and disconnects the inductor from the clock system when the switch is open. First (501) and second (503) high impedance voltage sources supply respective first and second voltages to the switch and a gate voltage of the first transistor transitions with the clock signal around the first voltage and a gate voltage of the second transistor transitions with the clock signal around the second voltage such that near constant overdrive voltages are maintained for the first and second transistors.
Description
CONTROLLING IMPEDANCE OF A SWITCH USING HIGH IMPEDANCE VOLTAGE SOURCES TO PROVIDE MORE EFFICIENT CLOCKING
TECHNICAL FIELD
Disclosed embodiments relate to clocking systems and more particularly to switches used to switch between operational modes of the clocking system.
BACKGROUND ART
Power consumption is a critical area of circuit design. That is particularly true for battery operated mobile systems. Many sources of undesired power loss can be found in integrated circuits such as leakage current in transistors and power loss due to undesirable resistive or capacitive loading. While power consumption is critical in mobile battery operated devices, power consumption is also important in other settings such as server farms where the power savings for each processor is multiplied many-fold. Thus, improving efficiency in integrated circuit operation is desirable to extend battery life and more generally to reduce energy consumption. DISCLOSURE OF THE INVENTION
In some embodiments, an apparatus includes a switch coupled to receive a clock signal at an input node of the switch and supply the clock signal at an output node of the switch when the switch is closed. A first high impedance voltage source is coupled to supply a first voltage to a first gate node of a first transistor of the switch.
In some embodiments, a method includes supplying a first voltage from a first high impedance voltage source that is coupled to a gate of a first transistor of the switch. A second voltage is supplied from a second high impedance voltage source that is coupled to a gate of a second transistor of the switch. A clock signal is supplied to respective first current carrying nodes of the first and second transistors, and respective second current carrying nodes of the first and second transistors supply the clock signal to an inductor when the switch is closed.
In some embodiments, a clock system of an integrated circuit includes a switch that has a first and second transistor. The switch receives a clock and supplies
the clock signal to an inductor when the switch is closed. The inductor forms a resonant circuit with capacitance of a clock system of the integrated circuit when the switch is closed and is disconnected from the capacitance of the clock system when the switch is open. First and second high impedance voltage sources supply respective first and second voltages to the switch such that a first near constant overdrive voltage is maintained for a first gate node of the first transistor and a second near constant gate overdrive voltage is maintained for a second gate node of the second transistor.
In some embodiments, low conducting resistance is achieved using a first circuit to selectively couple an inductor to a clock system. The circuit includes a first transistor coupling a first terminal of the inductor to a first supply voltage and a second transistor coupling the first terminal of the inductor to a second supply voltage.
BRIEF DESCRIPTION OF THE DRAWINGS
The embodiments disclosed herein may be better understood, and its numerous objects, features, and advantages made apparent to those skilled in the art by referencing the accompanying drawings.
Fig. 1 illustrates a simplified model of a dual mode clocking system which supports both resonant and conventional clocking, according to some embodiments. Fig. 2 shows example clock voltage and mode switch current waveforms, according to some embodiments.
Fig. 3 illustrates a simplified model of a mode switch, according to some embodiments.
Fig. 4 illustrates mode switch sizing tradeoffs, according to some
embodiments.
Fig. 5 illustrates a simplified schematic using high-impedance voltage sources to drive the gate terminals of the mode switch, according to some embodiments.
Fig. 6A shows nFET and pFET gate voltages with respect to clock voltage when the mode switch is closed, according to some embodiments.
Fig. 6B shows nFET and pFET gate voltages with respect to clock voltage when the mode switch is open, according to some embodiments. Fig. 7 illustrates an exemplary switched-capacitor implementation of a high impedance voltage generator to generate a gate voltage for a switch, according to some embodiments.
Fig. 8 illustrates an exemplary switched-capacitor implementation of a high impedance voltage generator to generate a gate voltage for a switch, according to some embodiments.
Fig. 9 illustrates an embodiment to achieve low conducting resistance based on the location of the mode switch.
The use of the same reference symbols in different drawings indicates similar or identical items. MODE(S) FOR CARRYING OUT THE INVENTION
Use of high impedance voltage sources for the gate terminals of the switch used in a resonant clocking system reduces the on-resistance of the switch for a given switch width and provides reduced leakage through the switch when the switch is open, thereby improving clocking efficiency across multiple frequencies. Clock distribution networks account for a significant portion of overall power consumption in most high performance digital circuits today due to the vast amounts of parasitic capacitance that is connected to the clock network. One technique to implement a more energy-efficient clock distribution is resonant clocking.
Fig. 1 illustrates a simplified model of a resonant clock system 100, according to some embodiments. The resonant clock system 100 may be part of a
microprocessor, graphics processor, or other integrated circuit with a clock system. A distinct feature of the resonant clock system 100 is the use of an inductance (L) 101 that is connected in parallel with the parasitic network capacitance (C) 103. The role
of the clock driver 102 is to replenish only the energy lost in the parasitic resistance of the LC system formed by the inductor 101 and parasitic network 103. Resonant clocking thus relies on efficient LC resonance to achieve energy efficiency. For an ideal inductor and interconnect, with no parasitic resistance, the clock network would oscillate with zero energy dissipation. Resonant clocking is most efficient at frequencies close to the resonant frequency.
Driving the resonant clock network at frequencies much lower than the resonant frequencies results in malformed clock waveforms that prevent correct operation of the processor. Modern microprocessors operate over a frequency range that cannot be supported by a purely resonant clock network designed to resonate at a single frequency. One way to address the wide frequency range uses dual clock modes of operation. At frequencies near the resonant frequency, the processor operates in the resonant clock mode. At frequencies outside the range where resonant clocking can be robustly operated, the clock system works in a conventional mode or non- resonant mode. Switching between resonant mode and non-resonant mode uses mode switch 104 that is in series with the inductor 101. Mode switch 104 is closed during resonant operation (rclk) to form an LC circuit and open in conventional mode (cclk) so as to disconnect the inductor from the clock system. For the purposes herein, mode switch 104 (MSw) is located as shown in Fig. 1. Using mode switch 104 provides the required functionality of a dual-mode clock, but increases the resistance of the LC tank, reducing resonant clocking efficiency.
Fig. 2 shows example clock voltage 201 and current waveforms 203 associated with mode switch 104. In the embodiment of Fig. 1, the mode switch 104 includes a pFET device 107 and an nFET device 109. The clock voltage transitions from ground to Vdd during its cycle, and as a result, the nFET (pFET) device sees a substantially reduced gate overdrive voltage when the clock signal approaches Vdd (ground), contributing to the increased resistance. The gate overdrive voltage is the voltage between the gate and source (Vgs) in excess of the threshold voltage. Thus, when the clock signal is at its highest voltage, approximately Vdd, the difference between the source and the gate voltage (when the nFET is on) is at a minimum. The increased resistance contributed by the reduced gate overdrive voltage can be reduced by using a larger switch.
Fig. 3 shows a simplified diagram of the mode switch. Mode switch 104 includes the gate to drain capacitances (Cgd) 301 and 302 and the gate to source capacitances (Cgs) 303 and 304. In addition, the switch 104 includes resistance shown as resistance 305. A larger switch increases the capacitive loading on the clock network, increasing not only the dissipation in the conventional mode, but also in the resonant mode, due to the higher current flow through the resistance of mode switch 104 and the inductor 101. Fig. 4 illustrates the tradeoff between the switch resistance and the capacitive loading of the switch, which results in significant inefficiency of the overall system. The tradeoff relates to switch size. Smaller switches offer higher conventional clock efficiency due to reduced loading, e.g., at 401 as compared to 403. Larger switches offer improved resonant clocking performance up to a certain point, e.g., at 407 as compared to 405. Thus, trying to achieve efficiency for one mode of operation, e.g. resonant clocking, sacrifices efficiency in the non-resonant mode of operation, and vice versa. One existing solution to the inefficiency resulting from the tradeoff between the mode switch resistance and the capacitive loading offered to the clock network is to use low threshold voltage (Vth) devices. That technique, however, does not offer significant resistance reduction, especially at higher voltages where power efficiency is particular important to obtain improved performance. Another problem associated with the use of the modes switch is leakage current. Using a low on-resistance mode switch implies a higher leakage current through the mode switch when it is off. The use of low Vth devices reduces the on/off current ratio, increasing the leakage through the mode switch when the switch is open. The higher leakage through the switch results in additional undesirable power dissipation. Accordingly, some embodiments address the disadvantages of low Vth devices by improving the pareto-efficiency of the mode switch 104 by significantly reducing the on-resistance of the mode switch 104 for a given switch width. The reduced resistance can be used for improved efficiency in resonant mode, or be traded off for improved conventional mode clocking through mode switch width reduction. It also allows for reduced leakage through the mode switch when the switch is open.
Fig. 5 shows a simplified schematic of an embodiment that reduces the on- resistance of the mode switch 104 for a given switch width and provides for reduced
leakage through the mode switch when the switch is open. Instead of within-rail signals driving the gate of the mode switch 104, high impedance voltage sources, Vgn 501 and Vgp 503, are used to allow the gate voltage to transition along with the source/drain voltage of the mode switch 104. Use of the high impedance voltage sources allows for better pareto-efficiency with regard to the design of the mode switch 104. The use of the high impedance voltage source also allows the added benefit of reduced loading on the clock network while in conventional mode since the gate terminals can be thought of as floating terminals in the AC sense, and in the best- case scenario, the loading on the clock node goes from 2Cdg (as per Fig. 3) down to Cdg. The assumption is that Cdg is approximately equal to Csg. If the gate is floating and the mode switch is turned off, there are two parallel paths made up of two series capacitors. Each of these paths is approximately Cdg/2 in capacitance and thus, the resultant capacitance is Cdg.
Fig. 6A shows waveforms of the applied gate voltage of the nFET and the pFET device of the mode switch and the clock signal when the mode switch 104 is closed in resonant mode. Note how the gate signals 601 and 603 transition along with the clock signal 605 maintaining a near-constant gate overdrive in the conducting transistor, while ensuring that reliability limits are not exceeded. Nearly or substantially constant means that the gate overdrive voltage does not vary by more than, e.g., approximately 10% or less in a typical embodiment depending on the needs of the system. The amount of change in the gate overdrive depends on the impedance of the supply. If the supply impedance is infinitely high, the gate overdrive will actually be almost constant (e.g., less than 1% variation). As impedance gets lower, the gate voltage tracks the source/drain voltage less until ultimately, the gate voltage does not track the source drain voltage at all when the power supply has zero impedance. Thus, depending on the impedance, some embodiments may have gate overdrive that varies by more than 10% but the variation is sufficiently small to improve the conductivity of the mode switch.
Referring still to Fig. 6A, when using the high impedance voltage source, the gate voltage, e.g., shown in gate signal 601, tracks the clock signal 605. Fig. 6A shows the gate voltage transitioning in synchronism with the clock signal around the voltage level 602, which is supplied by the power supply Vgn 501. The gate signal
tracks the clock signal due to the impedance seen by the source and drain into the gate, past the Cgd and Cgs (see Fig. 3). If for example the impedance was low, e.g., a zero impedance, then the gate would just be tied to that voltage, regardless of the gate and source voltage transitions. On the other hand, if the impedance is high, then a high pass filter is seen with Cgd and Cgs in series with the resistance of the supply. For high frequency clocks, nearly all the voltage is transferred to the gate node. Of course, with a high pass filter, if the clocks were really slow, a very large impedance would be needed for the supply voltage to allow the gate voltage the track.
Fig. 6B shows the operation of the mode switch 104 with high impedance voltage sources when the mode switch 104 is open. Note that the gate signals 611 and 613 transition along with the clock signal 615 maintaining a near-constant gate overdrive in the conducting transistor, while ensuring that reliability limits are not exceeded. The use of the high impedance voltage sources dramatically reduces leakage while in the open state shown in Fig. 6B.
The high impedance voltage source can be generated by several techniques, one of which is the use of charge pumps to achieve the desired voltage levels. In addition, although Fig. 6A shows a DC level of 1.5Vdd for the gate voltage (Vgn) for the nFET and a DC level of -0.5Vdd for the gate voltage (Vgp) for the pFET, the appropriate voltage settings for the voltage supplies can be selected based on the requirements for a particular design. As shown in Fig. 6B, note that the voltage swing of waveforms 611 and 613 on the gate terminals is lower when the mode switch is open. That is expected since the side of the switch connected to the inductor does not toggle, and instead serves to reduce the amplitude of the coupled voltage.
Consequently, the voltages Vgn and Vgp should be set according to reliability constraints.
Figs. 7 and 8 show, respectively, embodiments to achieve the voltages of 1.5 Vdd and -0.5 Vdd with a high output impedance. The charge pump embodiment of Fig. 7 utilizes a clock signal to alternate the open/closed state of the switches in phase 1 and phase 2 illustrated in Fig. 7. In phase one, switches 701, 703, and 706 are closed to charge capacitors 707 and 709, which are configured in series. In phase 2, switches 711, 713, 715, and 717 are closed configuring Vdd in series with the parallel capacitors 707 and 709, allowing a voltage to be supplied that is 1.5 Vdd. To achieve
a higher output impedance, a lower frequency clock may be used to toggle the switches between the two phases because output impedance is inversely related to switch frequency.
Fig. 8 illustrates a charge pump embodiment that utilizes a clock signal to alternate the open/closed state of the switches in phase 1 and phase 2. In phase 1, switches 801, 803 and 805 are closed to charge capacitors 807 and 809 in series. In phase 2, switches 813, 815, and 817 are closed allowing a voltage to be supplied from capacitors 807 and 809 configured in parallel that is -0.5 Vdd. Of course, the particular voltage values in Figs. 7 and 8 are exemplary and other voltage values may be chosen to be supplied according to the needs of the particular system.
Embodiments described herein offer several advantages including reduced on- resistance of the mode switch 104 while operating within reliability constraints of the device and increased off -resistance (lower leakage) of the mode switch 104 while operating within reliability constraints of the device. The dramatically lower leakage also allows for the use of even lower Vth devices for the mode switch while still avoiding substantial power loss through leakage current. Further, in conventional (non-resonant) mode, the use of a high-impedance gate driver for the mode switch reduces the capacitive loading on the clock node, allowing for improved energy efficiency. Reduced on-resistance is achieved through the higher gate overdrive on both the nFET and the pFET devices of the mode switch 104. Reliability constraints are satisfied through the use of the high impedance voltage source, which allows the voltage of the source and drain to couple into the gate terminal, allowing it to transition along with the source/drain, thereby maintaining a substantially constant gate overdrive regardless of the clock voltage level.
When the mode switch 104 is open, the devices operate in deep-cutoff with drastically lower leakage as compared to conventional techniques. That improves energy efficiency of the clock system in conventional mode. In some embodiments, the high impedance power supplies are shared across a plurality of the mode switches. In some embodiments, the number of mode switches that share a high impedance power supply is determined according to system requirements such as the ability of
the high impedance power supplies to supply a particular number of switches. Note also than when switching from resonant mode to non- resonant mode, the power supplies are switched by a switch (not shown in Fig. 5) so -0.5 Vdd goes to the nFET and 1.5 Vdd goes to pFET. Some embodiments address the problem of increased I2R losses due to the lower gate overdrive in the mode switch by relocating the mode switch 104 so that the nFET and pFET devices of the mode switch 104 conduct with a higher gate overdrive, and in doing so offer a lower conducting resistance for a given mode switch width. The higher overdrive is present when the pFET 901 is on in resonant mode since the drain and source voltages are near Vdd while the gate voltage is low. Similarly for pFET 903, the gate voltage is high while the source and drain are low.
Fig. 9 shows a simplified representation of such an embodiment. The mode switch is implemented as a header-footer combination using the pFET 901 and nFET 903. The FETs 901 and 903 are turned on in resonant mode and turned off in non- resonant mode. Note that the FETs 901 and 903 use with-in supply rail voltages in contrast to the FETS 502 and 504 shown in Fig. 5, which have boosted supply voltages from the charge pumps shown in Figs. 7 and 8. Capacitors 907 and 909 may be included to ensure that the n2 node of inductor 101 is connected to ground in an AC sense. That avoids a DC drop across the inductor (which would lead to current buildup) so that node n2 takes on the natural DC value of the clock signal, and at the same time provides a very low impedance at the resonant frequency so that the inductor 101 and the clock network load look like a parallel LC tank circuit, with the capacitors 907 and 909 appearing like ground connections. The gate higher overdrive and lower conducting resistance achieved in the configuration illustrated in Fig. 9 is significantly better for a circuit using within supply rail voltages than the switch configuration shown in Fig. 1. The reduced conducting resistance may then be used to arrive at a more pareto-optimal tradeoff between conducting resistance of the mode switch and parasitic loading due to the mode switch.
While circuits and physical structures have been generally presumed in describing embodiments herein, it is well recognized that in modern semiconductor design and fabrication, physical structures and circuits may be embodied in a computer readable medium as data structures for use in subsequent design,
simulation, test, or fabrication stages. For example, such data structures may encode a functional description of circuits or systems of circuits. The functionally descriptive data structures may be, e.g., encoded in a register transfer language (RTL), a hardware description language (HDL), in Verilog, or some other language used for design, simulation, and/or test. Data structures corresponding to embodiments described herein may also be encoded in, e.g., Graphic Database System II (GDSII) data, and functionally describe integrated circuit layout and/or information for photo-mask generation used to manufacture the integrated circuits. Other data structures, containing functionally descriptive aspects of embodiments described herein, may be used for one or more steps of the manufacturing process.
Computer-readable media include tangible computer readable media, e.g., a disk, tape, or other magnetic, optical, or electronic storage medium. In addition to computer-readable medium having encodings thereon of circuits, systems, and methods, the computer readable media may store instructions as well as data that can be used to implement embodiments described herein or portions thereof. The data structures may be utilized by software executing on one or more processors, firmware executing on hardware, or by a combination of software, firmware, and hardware, as part of the design, simulation, test, or fabrication stages.
The description of the embodiments set forth herein is illustrative, and is not intended to limit the scope of the invention as set forth in the following claims. For example, embodiments of the invention are not limited in scope to microprocessors. Rather, the solution described herein applies to integrated circuits in general. Other variations and modifications of the embodiments disclosed herein may be made based on the description set forth herein, without departing from the scope of the invention as set forth in the following claims.
Claims
1. An apparatus comprising:
a switch coupled to receive a clock signal at an input node of the switch and supply the clock signal at an output node of the switch when the switch is closed;
an inductor coupled in series with the switch, wherein when the switch is closed, a clock system is operable in a resonant mode and when the switch is open the clock system operable in a non-resonant mode; a first high impedance voltage source coupled to supply a first voltage to a first gate node of a first transistor of the switch; and
a second high impedance voltage source coupled to supply a second voltage to a second gate node of a second transistor of the switch.
2. The apparatus as recited in claim 1 wherein a first gate voltage at the first gate node transitions around the first voltage in accordance with the clock signal.
3. The apparatus as recited in claim 2 wherein the transitions of the first gate voltage around the first voltage in accordance with the clock signal are greater when the switch is closed.
4. The apparatus as recited in any of claims 1 to 3 wherein a second gate voltage at the second gate node transitions around the second voltage in accordance with the clock signal.
5. The apparatus as recited in claim 4 wherein the transitions of the second gate voltage around the second voltage in accordance with the clock signal are greater when the switch is closed.
6. The apparatus as recited in any of claims 1 to 5 wherein a near constant gate overdrive voltage is maintained for the first gate node and a second near constant gate overdrive voltage is maintained for the second gate node.
7. The apparatus as recited in any of claims 1 to 6 wherein the first high impedance voltage source comprises a switched capacitor circuit and wherein the switch comprises a PFET transistor and an NFET transistor.
8. The apparatus as recited in claim 7 wherein the switched capacitor circuit is clocked by another clock signal slower than the clock signal supplied to the switch in the resonant mode of operation.
9. A method comprising:
supplying a first voltage from a first high impedance voltage source to a gate of a first transistor of a switch;
supplying a second voltage from a second high impedance voltage source to a gate of a second transistor of the switch;
supplying a clock signal to respective first current carrying nodes of the first and second transistors and supplying the clock signal from respective second current carrying nodes of the first and second transistors to an inductor when the switch is closed; and
operating a clock system using the clock signal in a resonant mode with the switch closed and operating the clock system in a non-resonant mode with the switch open.
10. The method as recited in claim 9 wherein a first gate voltage at the gate of the first transistor transitions around the first voltage in accordance with transitions of the clock signal and wherein a second gate voltage at the gate of the second transistor transitions around the second voltage in accordance with the transitions of the clock signal.
11. The method as recited in claim 10 wherein a voltage swing of the first gate voltage around the first voltage is greater with the switch closed than with the switch open.
12. The method as recited in any of claims 9 to 11 further comprising maintaining a first substantially constant gate overdrive voltage at the gate of the first
transistor and maintaining a second substantially constant gate overdrive voltage at the gate of the second transistor.
13. The method as recited in any of claims 9 to 12 further comprising:
supplying the first voltage from a switched capacitor circuit; and
supplying a second clock signal to the switched capacitor circuit that is slower than the clock signal supplied to the switch in resonant mode of operation.
Priority Applications (5)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2015530050A JP6063047B2 (en) | 2012-08-31 | 2013-08-29 | Switch impedance control using a high impedance voltage source to provide more efficient clocking |
IN2272DEN2015 IN2015DN02272A (en) | 2012-08-31 | 2013-08-29 | |
CN201380042914.8A CN104541223B (en) | 2012-08-31 | 2013-08-29 | Using the impedance of high impedance voltage source controlling switch to provide more effective clock |
KR1020157007588A KR102141675B1 (en) | 2012-08-31 | 2013-08-29 | Controlling impedance of a switch using high impedance voltage sources to provide more efficient clocking |
EP13759412.3A EP2891025B1 (en) | 2012-08-31 | 2013-08-29 | Controlling impedance of a switch using high impedance voltage sources to provide more efficient clocking |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US13/601,155 | 2012-08-31 | ||
US13/601,155 US8742817B2 (en) | 2012-08-31 | 2012-08-31 | Controlling impedance of a switch using high impedance voltage sources to provide more efficient clocking |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2014036288A1 true WO2014036288A1 (en) | 2014-03-06 |
Family
ID=49118840
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/US2013/057331 WO2014036288A1 (en) | 2012-08-31 | 2013-08-29 | Controlling impedance of a switch using high impedance voltage sources to provide more efficient clocking |
Country Status (7)
Country | Link |
---|---|
US (1) | US8742817B2 (en) |
EP (1) | EP2891025B1 (en) |
JP (1) | JP6063047B2 (en) |
KR (1) | KR102141675B1 (en) |
CN (1) | CN104541223B (en) |
IN (1) | IN2015DN02272A (en) |
WO (1) | WO2014036288A1 (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9429982B2 (en) | 2014-09-27 | 2016-08-30 | Qualcomm Incorporated | Configurable last level clock driver for improved energy efficiency of a resonant clock |
US10003337B1 (en) * | 2017-05-17 | 2018-06-19 | International Business Machines Corporation | Resonant virtual supply booster for synchronous logic circuits and other circuits with use of on-chip integrated magnetic inductor |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6882182B1 (en) * | 2003-09-23 | 2005-04-19 | Xilinx, Inc. | Tunable clock distribution system for reducing power dissipation |
US20110084736A1 (en) * | 2009-10-12 | 2011-04-14 | Cyclos Semiconductor, Inc. | Architecture for controlling clock characteristics |
Family Cites Families (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
AU5382494A (en) | 1994-01-18 | 1995-08-03 | Premlex Pty. Ltd. | A switching circuit |
US6025738A (en) | 1997-08-22 | 2000-02-15 | International Business Machines Corporation | Gain enhanced split drive buffer |
US6088250A (en) | 1998-05-29 | 2000-07-11 | The Aerospace Corporation | Power converters for multiple input power supplies |
US6205571B1 (en) | 1998-12-29 | 2001-03-20 | International Business Machines Corporation | X-Y grid tree tuning method |
US6310499B1 (en) | 2000-07-17 | 2001-10-30 | Hewlett-Packard Company | Methods and apparatus for adjusting the deadtime between non-overlapping clock signals |
US7015765B2 (en) | 2003-01-13 | 2006-03-21 | The Trustees Of Columbia In The City Of New York | Resonant clock distribution for very large scale integrated circuits |
US7082580B2 (en) | 2003-02-10 | 2006-07-25 | Lsi Logic Corporation | Energy recycling in clock distribution networks using on-chip inductors |
TWI261158B (en) | 2003-09-08 | 2006-09-01 | Via Tech Inc | Method and related apparatus for outputting clock through data path |
US7237217B2 (en) | 2003-11-24 | 2007-06-26 | International Business Machines Corporation | Resonant tree driven clock distribution grid |
US7516350B2 (en) | 2004-09-09 | 2009-04-07 | International Business Machines Corporation | Dynamic frequency scaling sequence for multi-gigahertz microprocessors |
US7956664B2 (en) | 2006-12-01 | 2011-06-07 | The Regents Of The University Of Michigan | Clock distribution network architecture with clock skew management |
US7973565B2 (en) | 2007-05-23 | 2011-07-05 | Cyclos Semiconductor, Inc. | Resonant clock and interconnect architecture for digital devices with multiple clock networks |
US7675121B2 (en) | 2007-10-08 | 2010-03-09 | International Business Machines Corporation | SOI substrate contact with extended silicide area |
US8080988B2 (en) * | 2008-06-18 | 2011-12-20 | Active-Semi, Inc. | Switch driver with low impedance initial drive and higher impedance final drive |
US8719748B2 (en) | 2011-06-29 | 2014-05-06 | The Regents Of The University Of California | Distributed resonant clock grid synthesis |
US8576000B2 (en) | 2011-08-25 | 2013-11-05 | International Business Machines Corporation | 3D chip stack skew reduction with resonant clock and inductive coupling |
-
2012
- 2012-08-31 US US13/601,155 patent/US8742817B2/en active Active
-
2013
- 2013-08-29 IN IN2272DEN2015 patent/IN2015DN02272A/en unknown
- 2013-08-29 EP EP13759412.3A patent/EP2891025B1/en active Active
- 2013-08-29 KR KR1020157007588A patent/KR102141675B1/en active IP Right Grant
- 2013-08-29 WO PCT/US2013/057331 patent/WO2014036288A1/en unknown
- 2013-08-29 CN CN201380042914.8A patent/CN104541223B/en active Active
- 2013-08-29 JP JP2015530050A patent/JP6063047B2/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6882182B1 (en) * | 2003-09-23 | 2005-04-19 | Xilinx, Inc. | Tunable clock distribution system for reducing power dissipation |
US20110084736A1 (en) * | 2009-10-12 | 2011-04-14 | Cyclos Semiconductor, Inc. | Architecture for controlling clock characteristics |
Non-Patent Citations (1)
Title |
---|
IKEUCHI K ET AL: "Switched Resonant Clocking (SRC) scheme enabling dynamic frequency scaling and low-speed test", CUSTOM INTEGRATED CIRCUITS CONFERENCE, 2009. CICC '09. IEEE, IEEE, PISCATAWAY, NJ, USA, 13 September 2009 (2009-09-13), pages 33 - 36, XP031542767, ISBN: 978-1-4244-4071-9 * |
Also Published As
Publication number | Publication date |
---|---|
US20140062563A1 (en) | 2014-03-06 |
US8742817B2 (en) | 2014-06-03 |
EP2891025B1 (en) | 2019-06-19 |
CN104541223B (en) | 2017-10-10 |
IN2015DN02272A (en) | 2015-08-21 |
EP2891025A1 (en) | 2015-07-08 |
JP6063047B2 (en) | 2017-01-18 |
CN104541223A (en) | 2015-04-22 |
KR20150052117A (en) | 2015-05-13 |
JP2015534670A (en) | 2015-12-03 |
KR102141675B1 (en) | 2020-09-14 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US8339209B2 (en) | Method for selecting natural frequency in resonant clock distribution networks with no inductor overhead | |
US8941432B2 (en) | Transitioning between resonant clocking mode and conventional clocking mode | |
US8854100B2 (en) | Clock driver for frequency-scalable systems | |
US10348300B2 (en) | Multi-level adiabatic charging methods, devices and systems | |
JP4034781B2 (en) | Clock generator for integrated circuits | |
US20140062562A1 (en) | Constraining clock skew in a resonant clocked system | |
EP2891025B1 (en) | Controlling impedance of a switch using high impedance voltage sources to provide more efficient clocking | |
Salem et al. | 26.4 A 0.4-to-1V 1MHz-to-2GHz switched-capacitor adiabatic clock driver achieving 55.6% clock power reduction | |
Jeanniot et al. | Synchronised 4-phase resonant power clock supply for energy efficient adiabatic logic | |
ur Rahman et al. | Quasi-resonant clocking: Continuous voltage-frequency scalable resonant clocking system for dynamic voltage-frequency scaling systems | |
CN104731161A (en) | Stacked Clock Distribution For Low Power Devices | |
EP2218184A2 (en) | Charge recycling a 1 of n ndl gate with a time varying power supply | |
CN105048998A (en) | Redundant clock transition tolerant latch circuit | |
Ma et al. | Bootstrap capacitorless DCM VOT buck converter with dead‐time‐based off‐time calibration for magnetic resonance wireless power transfer | |
Teichmann et al. | Generation of the power-clock |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 13759412 Country of ref document: EP Kind code of ref document: A1 |
|
ENP | Entry into the national phase |
Ref document number: 2015530050 Country of ref document: JP Kind code of ref document: A |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
ENP | Entry into the national phase |
Ref document number: 20157007588 Country of ref document: KR Kind code of ref document: A |