Self regulating body bias generator

Information

  • Patent Grant
  • 6731158
  • Patent Number
    6,731,158
  • Date Filed
    Thursday, June 13, 2002
    22 years ago
  • Date Issued
    Tuesday, May 4, 2004
    20 years ago
Abstract
The back bias voltage on a functional circuit is controlled through a closed loop process. A delay element receives a clock pulse and produces a delay output. The delay element is advantageously constructed of the same materials as the functional circuit so that the aging and degradation of the delay element parallels the degradation of the functional circuit. As the delay element degrades, the transistor switching time increases, increasing the time delay of the delay output. An AND gate compares a clock pulse to an output pulse of the delay element, the AND output forming a control pulse. A duty cycle of the control pulse is determined by the delay time between the clock pulse and the delay element output. The control pulse is received at the input of a charge pump. The charge pump produces a back bias voltage which is then applied to the delay element and to the functional circuit. If the time delay produced by the delay element exceeds the optimal delay, the duty cycle of the control pulse is shortened, and the back bias voltage is lowered, thereby increasing the switching speed of the transistors in the delay element and reducing the time delay. If the throughput of the delay element is too fast, the duty cycle of the control pulse is lengthened, raising the back bias voltage produced by the charge pump. This, in turn, lowers the switching speed of the transistors in both the delay element and the functional circuit. The slower switching speed in the delay element increases time delay. In this manner, the switching speed of the delay element, and of the functional circuit, is maintained at a constant level over the life of the circuit.
Description




FIELD OF THE INVENTION




The present invention relates to controlling the back bias on a functional circuit to reduce power consumption by the functional circuit while maintaining switching speed of the transistors at levels sufficient to satisfy system requirements. More particularly, the present invention relates to biasing a MOS integrated circuit through a closed loop circuit configured to measure a time delay created by a delay circuit, and to generate a back bias voltage dependent on the time delay.




BACKGROUND OF THE INVENTION




Although reduction in power consumption has been an ongoing goal in the field of electronics, the proliferation of portable and space based CMOS devices and microprocessor driven apparatuses has created an ever growing need for power conservation, giving rise to a variety of processes and apparatuses for reducing power consumption within personal computers, including a reduction of power within integrated MOS circuits. Power consumption of a digital circuit equals dynamic power consumption plus static power consumption. This relationship can be represented by the formula:






power≈


fCV




DD




2


+(


V




DD


)(


I




DC


)  1)






where f is the operating frequency, C is the equivalent capacitance of the circuit, and I


DC


is the static current. Because it has long been recognized that dynamic power consumption is proportional to V


DD




2


, it has been similarly understood that a reduction in supply voltage can dramatically reduce power consumption, as illustrated in a paper by von Kaenel, Macken and Degrauwe entitled “A Voltage Reduction Technique for Battery-Operated Systems” in the IEEE Journal of Solid-State Circuits, Vol. 25, No. 5, October 1990, pages 1136-1140. A reduction of the supply voltage, however, has the unwanted effect of delaying the switching of the transistors within an integrated circuit, thereby slowing down the critical path of the circuit. If the circuit through-put is slowed below a certain minimum threshold, an integrated circuit can no longer function reliably in the environment for which it was designed.




An alternative approach for reducing power consumption is to raise the threshold voltage of a transistor, typically by a process known as “back biasing” the transistor substrate.

FIG. 1

illustrates an n-channel MOS field effect transistor


100


with a “body bias” or “back bias” voltage


104


applied to the substrate


102


. Increasing the threshold voltage has the desirable effect of decreasing the transistor's leakage current, thereby reducing power consumption. As the threshold voltage is raised, power consumption is reduced, but the switching speed of the transistor is also slowed. Once again, if the switching time of a transistor becomes excessive, the delay will render the circuit inoperative for the environment in which it is intended to function. To maximize the efficiency in terms of power consumption, designs have been proposed wherein the threshold voltage could be raised such that the circuit speed is just fast enough to satisfy system requirements. A fundamental challenge to this design goal is the fact that, although a transistor threshold is typically fixed when the circuit is manufactured, the threshold voltage will typically vary from chip to chip as a result of manufacturing tolerances. More significantly, a circuit degrades over time, typically from heat and radiation. This degradation affects a variety of functional parameters such as the threshold switching voltage and the switching speed of the transistor. Accordingly, if the threshold voltage is set as high as allowable at the time of manufacture to minimize power consumption, as soon as the transistor speed degrades at all, the transistor will be too slow to satisfy system requirements. Alternatively, if the bias is set such that it allows for a certain degradation of switching speed before the transistor falls below the acceptable lower limit, the power consumption is higher than necessary for much of the life of the transistor.




One solution has been to set the body bias as high as allowable, and lower the bias voltage over the life of the transistor as the transistor slows. To achieve this, however, the circuit must include some means of monitoring or estimating the switching speed of the transistors within a circuit to ensure that the speed remains within system requirements. Earlier methods utilizing a back biasing voltage for power reduction have typically used a single reference transistor selected from among the many transistors comprising a MOS integrated circuit. However, if the gradual degradation of the reference transistor did not exactly match the average degradation of the circuit, the technique was inaccurate. Additionally, earlier methods taking advantage of the body bias effect have been component intensive, variously requiring a fixed reference voltage, a fixed reference current, fixed resistors creating a voltage divider for use as a reference voltage, variable frequency clocks, op-amps, and multiple memory registers for storing data defining various voltage levels for application to a back-bias voltage. These various approaches can be seen in a variety of works, including U.S. Pat. No. 3,609,414 to Pleshko et al. entitled “Apparatus for Stabilizing Field Effect Transistor Thresholds,” U.S. Pat. No. 4,142,114 to Green, U.S. Pat. No. 4,670,670 to Shoji, U.S. Pat. No. 5,682,118 to Kaenel et al., U.S. Pat. No. 5,744,996 to Kötzle et al., the above referenced paper by von Kaenel, Macken and Degrauwe, a related paper by Gutnik and Chandraksan entitled “Embedded Power Supply for Low-Power DSP,” and a paper entitled “A Delay Distribution Squeezing Scheme with Speed-Adaptive Threshold-Voltage CMOS (SA-Vt CMOS) for Low Voltage LSIs” by Mizuno and Ishibashi, which was presented at the 1998 International Symposium on Low Power Electronics and Design.




There exists therefore a need for a method and apparatus for reducing the power consumption of a CMOS circuit by controlling the back bias voltage applied to a CMOS circuit. There is a further need for a method and apparatus for reducing the power consumption of a CMOS circuit while monitoring the switching speed of the transistors. There is a further need for a method and apparatus for automatically adjusting the back-bias voltage to insure that the CMOS switching speed does not fall below a certain threshold level. There is also a need for a method and apparatus for controlling the back bias voltage over the life of a circuit that is not dependent on a single reference transistor. There is a further need for a method and apparatus for adjusting a back-bias voltage to reduce power consumption which avoids the addition of excessive ancillary components such as fixed reference voltage sources, fixed current sources, fixed resistors forming voltage dividers for reference voltages, variable speed clocks, op-amps, divide-by-N clock-counters, or multiple memory locations storing data defining various voltage levels for application to a back-bias voltage.




BRIEF SUMMARY OF THE INVENTION




The present invention is a method of and apparatus for reducing the power consumption of a CMOS circuit by controlling the back bias voltage applied to a CMOS circuit. The present invention further provides a method of and apparatus for reducing the power consumption of a CMOS circuit while monitoring the switching speed of the transistors. The present invention further discloses a method of and apparatus for automatically adjusting the back-bias voltage to insure that the CMOS switching speed does not fall below system requirements. The present invention further discloses a method of and apparatus for adjusting a back-bias voltage to reduce power consumption while limiting the addition of excessive ancillary components such as fixed reference voltage sources, fixed current sources, fixed resistors forming voltage dividers for reference voltages, variable speed clocks, op-amps, divide-by-N counters and multiple memory locations for storing data defining various voltage levels for application to a back-bias voltage.




An apparatus for regulating a back bias voltage in a functional circuit comprises a closed loop voltage regulator having a charge pump including a charge pump input and a charge pump output, and a delay element including a delay element input, a delay element output, and a delay element body, wherein the delay element body is coupled to the charge pump output. A functional circuit with a body is coupled to the charge pump output such that a charge pump output voltage produced at the charge pump output controls a body bias on the delay element and on the functional circuit. The charge pump output voltage is dependent upon a duty cycle of a control pulse received at the charge pump input. A logic element having a logic element output is coupled to the charge pump input. The logic element further comprises a first logic input coupled to a clock pulse, and a second logic input coupled to the delay element output. The functional circuit comprises a functional circuit input. The clock pulse is advantageously coupled to the delay element input and to the functional circuit input. According to an embodiment of the invention, the logic element is an AND gate. The delay element advantageously comprises a plurality of logical inverters coupled in series. The functional circuit advantageously comprises a CMOS circuit comprising a substrate, wherein the substrate comprises the body of the functional circuit. According to one embodiment, the charge pump output is coupled directly to the substrate of the functional circuit. According to an alternative embodiment, the charge pump output is coupled to the substrate through a proportional element, wherein the back bias voltage applied to the substrate is proportional to the charge pump output voltage.




The delay element and the functional circuit are comprised of materials configured to degrade in performance at a substantially identical rate over time. The charge pump is advantageously comprised of a plurality of pumping stages, wherein a pumping stage comprises a transistor having a gate coupled to a capacitor. According to an embodiment of the present invention, a transistor within a pumping stage exhibits a threshold voltage of less than 110 mV.




A method for regulating a back bias voltage on a functional circuit comprises controlling a duty cycle of a control signal, adjusting an output voltage of a charge pump according to the duty cycle of the control signal, and applying the back bias voltage to the functional circuit, wherein the back bias voltage is related to the output voltage of the charge pump. The step of controlling a duty cycle advantageously comprises the steps of receiving the clock pulse into an input of a delay element, receiving the clock pulse into a first input of a logic element, and receiving an output signal generated at an output of the delay element into a second input of the logic element. According to a preferred embodiment, the logic element is an AND gate. The control signal is generated by the logic element. An output of the logic element is coupled to an input of a charge pump. The delay element and the functional circuit are comprised of materials configured to degrade at a substantially identical rate over time. The charge pump is advantageously comprised of a plurality of pumping stages, a pumping stage comprising a transistor having a gate coupled to a capacitor. According to an embodiment, the delay element comprises a plurality of logical inverters coupled in series.




A functional CMOS circuit comprises a plurality of transistors while maintaining a switching time of the plurality of transistors within an optimum range. A method of reducing power consumption within the functional CMOS circuit comprises the steps of altering a duty cycle of a control-pulse, regulating a back-bias voltage according to the duty-cycle of the control-pulse, and applying the back-bias voltage to the functional CMOS circuit. The control-pulse is advantageously formed at an output of an AND gate. The step of altering the duty cycle of the control-pulse comprises the steps of transmitting a clock input signal into a first input of the AND gate, delaying the clock input signal through a delay element to form a delayed clock signal, and transmitting the delayed clock signal to a second input of the AND gate. The output of the AND gate is coupled to an input of a charge pump. The back bias voltage is proportional to an output voltage of the charge pump. The back-bias voltage is applied to the delay element.




An apparatus for regulating a back bias voltage in a functional circuit comprises means for comparing a clock signal and a delay clock signal, wherein the means for comparing produces a control pulse output, means for producing a controllable DC bias voltage according to a duty cycle of the control pulse output, means for delaying having an input coupled to the clock signal, wherein an output of the means for delaying forms the delay clock signal, and wherein the DC bias voltage is coupled to a body of the means for delaying, and means for performing a digital function, wherein a body of the means for performing is coupled to the DC bias voltage. According to an embodiment, the means for comparing is an AND gate. The same clock signal compared in the means for comparing is advantageously coupled to the means for performing a digital function. The means for delaying comprises a plurality of logical inverters coupled in series, and the means for performing a digital function comprises a CMOS circuit. The means for delaying and the means for performing a digital function are comprised of materials configured to degrade in performance at a substantially identical rate over time. According to an embodiment, the means for producing a controllable DC bias voltage comprises a plurality of pumping stages, a pumping stage comprising a transistor having a gate coupled to a capacitor.











BRIEF DESCRIPTION OF SEVERAL VIEWS OF THE DRAWINGS





FIG. 1

is an n-channel MOSFET transistor with a connection for applying a body bias voltage.





FIG. 2

is a closed loop voltage regulator for regulating the back bias voltage used to regulate the threshold voltage of the functional circuit.





FIG. 3

is a delay element comprising a plurality of inverters.





FIG. 4

is a timing diagram during steady state operation.





FIG. 5

is a timing diagram when the delay element is too slow.





FIG. 6

is a timing diagram when the delay element is too fast.





FIG. 7

is an embodiment of a charge pump using p-channel MOSFETs.











DETAILED DESCRIPTION OF THE INVENTION




Reference will now be made in detail to the preferred embodiments of the invention, examples of which are illustrated in the accompanying drawings. Although the following detailed description of the present invention includes numerous specific details set forth in order to provide a thorough understanding of the present invention, it will be readily apparent to one of ordinary skill in the prior art that the present invention may be practiced without these specific details. Minor variations of the disclosed invention will become apparent to those skilled in the art after studying the teachings disclosed herein. For example, specific reference is frequently made to a specific transistor type, p-channel, n-channel, CMOS integrated circuits, etc. Those skilled in the art will understand that the present invention may be applied to a wide variety of solid state and transistor structures. The recitation of these and other specific structures are offered to more fully explain and enable the present invention, not to limit it to one particular embodiment. The invention is intended to cover alternatives, modifications and equivalents, which may be included within the spirit and scope of the invention as defined by the appended claims. In other instances, numerous details which are commonly known and understood by those skilled in the art have not been recited within the present disclosure so as not to unnecessarily obscure aspects of the present invention.





FIG. 2

illustrates a closed loop voltage control circuit for regulating a back bias voltage


236


to a functional circuit


228


, preferably a CMOS integrated circuit. The closed loop comprises a delay element


220


advantageously comprised of materials similar to the functional circuit


228


which it serves to regulate. As the delay element


220


degrades with time due to heat, radiation, and other factors, the switching speed of the circuits within the delay element


220


slow, increasing the delay T


DELAY


as discussed further in

FIGS. 4-6

. A closed loop monitors the delay T


DELAY


within the delay element


220


, and continually adjusts the back bias voltage


236


applied to the delay element


220


, maintaining the delay time T


DELAY


generated by the delay element


220


at an optimal predetermined value. The same back bias voltage


236


, or a ratio thereof, is then applied to the functional circuit


228


. Because the materials and manufacturing processes of the delay element


220


are selected such that the operation and degradation of the delay element approximates that of the functional circuit


228


, the back bias voltage


236


produced through the closed loop process also adjusts to the progressive deterioration of the functional circuit


228


over the life of the circuit


228


.




According to the embodiment disclosed in

FIG. 2

, a clock


204


is coupled to a functional circuit


228


. In order to maintain an optimal back-bias voltage


236


on the functional circuit


228


a closed loop is formed by coupling the clock


204


to a first input


208


of an AND gate


206


and to an input of the delay element


220


. A delay clock pulse


230


is formed on the output of the delay element


220


. The output of the delay element


220


is coupled to the second input


210


of the AND gate


206


. The input and output signals associated with the AND gate


206


are discussed in greater detail in conjunction with

FIGS. 4-6

.




The clock pulse


234


and the delay clock pulse


230


oscillate from a binary high logic state to a binary low logic state, respectively forming square waves. Preferably, the high states and low states are of equal duration. Accordingly, as discussed in greater detail below in conjunction with

FIGS. 4-6

, if the clock pulse


234


and the delay clock pulse


230


are in phase, the control pulse


232


of the AND


206


output will be in a high voltage state approximately fifty percent of the time. If the clock pulse


234


and the delay clock pulse


230


are out of phase, at least one input


208


,


210


of the AND will always be in a low logic state, and the control pulse


232


produced at the AND output will be in a perpetual low logic state. Accordingly, the duty cycle of the control pulse


232


produced by the AND output can range from fifty percent to zero percent.




The AND gate


206


output is coupled to the charge pump


214


input. Accordingly, the control pulse


232


formed by the AND gate


206


controls the activity of the charge pump


214


, which determines the back bias voltage


236


. As discussed in greater detail below in conjunction with

FIG. 7

, the output voltage produced by the charge pump


214


, which forms the back bias voltage


236


applied to the delay element


220


and the functional circuit


228


, is highest when the control pulse


232


input into the charge pump has a duty cycle of fifty percent. As the duty cycle of the control pulse decreases below fifty percent, or, in alternative embodiments of the present invention, increases above fifty percent, the back bias voltage


236


produced by the charge pump


214


decreases from its maximum output.




The output of the charge pump


214


is coupled to the delay element


220


and the functional circuit


228


, thereby applying the back bias voltage


236


produced by the charge pump


214


to the substrate of both the functional circuit


228


and the delay element


220


. As the back bias voltage


236


decreases, the circuit speed of both the delay element


220


and the functional circuit


228


increase. Conversely, as the back bias voltage


236


increases, the speed of the delay element


220


and the functional circuit


228


decreases. As will be better appreciated in conjunction with

FIGS. 3-6

, the delay element


220


can be constructed such that as the delay element


220


slows down below the predetermined circuit speed, the phase difference between the clock pulse


234


and the delay clock pulse


230


increases, thereby decreasing the duty cycle of the control pulse


232


, thereby reducing the back bias voltage


236


produced by the charge pump


214


. This, in turn, speeds delay element


220


back up to its desired speed. Conversely, if the delay element


220


operates faster than a predetermined speed, the closed loop process described above acts to slow down the delay element


220


to a predetermined speed. Since the back bias voltage


236


is simultaneously applied to both the delay element


220


and the functional circuit


228


, the speed of the functional circuit


228


is likewise maintained at a desired operational speed by this process.





FIG. 3

discloses one embodiment of a delay element


302


comprising a plurality of inverters


307


,


308


,


309


arranged in series. Because an inverter effectively creates a 180 degree phase shift, an even number of inverters arranged in series will restore an output signal of the delay element


220


to the same phase as an incoming signal. When a delay element is comprised of inverters as illustrated in

FIG. 3

, according to the preferred embodiment, an even number of inverters is used, thereby maintaining an “in phase” relationship between an input signal and an output signal. As the threshold voltage within an inverter is raised, the switching time is slowed, creating a phase shift as a result of the critical path delay. The input terminal on the first inverter


307


forms the delay element


220


input of

FIG. 2

, and the output terminal on the last inverter


309


forms the delay element


220


output of FIG.


2


. Because each inverter must reach a predetermined voltage before it switches, the delay created by series of inverters


307


-


309


depicted in

FIG. 3

is proportional to the number of inverters. The specific number of inverters is therefore preselected to regulate the functional circuit


228


at a predetermined speed. Different numbers of inverters will therefore be appropriate for operation in conjunction with different functional circuits


228


. The multi-inverter delay element depicted in

FIG. 3

is not intended to limit alternative embodiments of a delay element, but is illustrative of one way in which a delay element can be constructed. The invention described herein can operate with any collection of circuitry that delays the clock signal and experiences degradation similar to that of the functional circuit


228


. Similarly, although the invention disclosed herein is generally described in terms of a non-inverted delay processed by an AND gate, the present invention envisions equivalent embodiments which utilize an inverted delay pulse processed through comparators comprising alternative logic gating, such as an OR gate or a NOR gate. Accordingly, the pulse samples illustrated in the accompanying drawings are not intended to limit the present invention. To ensure that the delay of the delay element


220


approximates the delay experienced in the functional circuit


228


over the life of the circuit, the delay element will most advantageously comprise semiconductor materials with properties similar or identical to the functional circuit


228


.




As noted, a slowing of the switching time of the inverters


307


,


308


,


309


increases the time delay T


DELAY


(

FIGS. 4-6

) between the clock pulse


234


and the delay clock pulse


230


respectively occurring at the input and the output of the delay element


220


. By comparing the delay clock pulse


230


to the original clock pulse


234


, the extent of the delay can be determined. To avoid the use of expensive components such as elaborate phase locked loops and pulse counters, the preferred embodiment of the present invention utilizes a basic AND gate


206


to compare the clock pulse


234


and the delay clock pulse


230


. Accordingly, the clock pulse


234


is coupled to the first input


208


of the AND gate


206


, and the delay clock pulse


230


is coupled to the second input


210


of the AND gate


206


. According to this configuration, the AND gate


206


output is only in a high logic state when both clock pulse


234


and the delay clock pulse


230


are in a high logic state.





FIG. 4

is an exemplary timing diagram illustrating the relationship of the clock pulse


234


, and the delay clock pulse


230


entering the input terminals


208


,


210


of the AND gate


206


, and the control pulse


232


produced at the output of the AND gate


206


. The delay clock pulse


230


lags the clock pulse


234


by a delay time T


DELAY


. During the delay time T


DELAY


, the clock pulse


234


at the first AND input


208


is in a high voltage state, and the delay clock pulse


230


at the second AND input


210


is in a low voltage state, resulting in a control pulse


232


in a low voltage state. As

FIG. 4

further illustrates, when the delay clock pulse


230


and the clock pulse


234


are simultaneously in a high state, the control pulse


232


formed at the AND gate


206


output is in a high state. Accordingly, as the delay time T


DELAY


increases, the pulse width of the control pulse


232


decreases.




As the delay element degrades through temperature, radiation, or any other cause, the switching time of the delay element


220


,


302


is slowed, thereby increasing the delay time T


DELAY


through the delay element


220


. It is recalled that the present invention is advantageously to slow the operation of the functional circuit


228


to the slowest possible speed that is capable of reliably maintaining circuit performance, thereby lowering power consumption. Defining the duty cycle of

FIG. 4

as optimum for a particular circuit, giving rise to a steady state operation,

FIGS. 5 and 6

can be contrasted against it to illustrate the operation of the present invention.





FIG. 5

displays signals associated with a delay element


220


which has been slowed beyond the acceptable limit. It is noted that the delay time T


DELAY


in

FIG. 5

has become so great that by the time the delay clock pulse


230


rises to a high voltage state, the clock pulse


234


has already transitioned to a low voltage state. As a result, the control pulse


232


remains in a low voltage state continually. Because the closed loop circuit of

FIG. 2

will act to correct the speed of delay element as it drifts from its ideal speed, in practice, the delay time T


DELAY


will not reach the state of

FIG. 5

under normal operating conditions. However, as will be understood in conjunction with

FIG. 7

, if a duty cycle of zero as depicted in

FIG. 5

were present, the back bias voltage


236


produced by the charge pump


214


would be zero. As the back bias voltage applied to a CMOS circuit is reduced, the circuit runs faster. Accordingly, the excessive delay of the delay clock pulse


230


of

FIG. 5

would give rise to a self correcting back bias voltage, speeding up the delay element


220


and reducing the delay T


DELAY


.




Alternatively, as illustrated in

FIG. 6

, if the delay time T


DELAY


through the delay element


220


is shorter than the optimum delay T


DELAY


depicted in

FIG. 4

, the duty cycle of the control pulse


232


in

FIG. 4

is seen to be closer to 50% than the optimum duty cycle of FIG.


4


. As the duty cycle of the control pulse


232


approaches 50%, the charge pump of

FIG. 7

will increase the back bias voltage


236


being applied to the functional circuit


228


and the delay element


220


,


302


.




The charge pump generates a voltage that is above the power supply voltage for use as the body bias. An example of a charge pump using diode-connected MOSFETs is shown in FIG.


7


. The charge pump consists of two inverters, U


1


and U


2


, a number of pumping stages (M


1


/C


1


, M


2


/C


2


, M


3


/C


3


, etc.) and an output filter composed of transistor MX and capacitor CX.




To illustrate the functionality of the charge pump


700


, it is assumed that the PULSE input to the charge pump is initially a binary high voltage state 1, such that the PLSNOT signal is a binary low voltage state 0. It is further assumed that all capacitors are discharged in their initial state. From this, the node V


2


X is also in a binary low state 0. The diode-connected PMOS transistor M


1


will only conduct current when the common gate/drain node is at a lower voltage than the source node. Since M


1


's source node is connected to V


DD


, and its gate/drain node is in a low voltage state 0, current will flow through M


1


to charge C


1


and bring node V


2


X to V


DD


or a binary 1. Similarly, node V


3


X must be a 1 because PLS is a 1 and C


2


is discharged, so capacitor C


3


is charged by current flowing through M


3


. Since V


3


X is at a higher voltage than V


2


X, no current flows through M


2


.




If the PULSE input then falls to a low voltage 0, PLSNOT will rise to a 1 and PLS will fall to a 0. Since C


1


has been charged, node V


2


X will rise to approximately 2×V


DD


. Node V


2


X is now at a higher voltage than node V


3


X so transistor M


2


conducts current into capacitor C


2


, bringing the voltage across this capacitor to approximately 2×V


DD


, owing to the relationship V


C1


+V


DD


≈2×V


DD


. No current flows through M


1


because V


2


X is at a higher voltage than V


DD


.




As the PULSE input then rises to a 1, the voltage at node V


3


X will be increased to 3×V


DD


because node PLS will rise from a 0 to a 1 and capacitor C


2


has been charged to 2×V


DD


. Since node V


3


X is now at a higher voltage than node V


4


X, transistor M


3


will conduct current into capacitor C


3


, bringing the voltage across this capacitor to approximately 3×V


DD


.




This process repeats indefinitely, with capacitor C


1


gathering charge from V


DD


when PLSNOT is low and then passing that charge to C


2


when PLSNOT is high. Likewise, capacitor C


2


gathers charge from node V


2


X when PLS is 0 and then passes that charge to C


3


when PLS is 1. The diode-connected transistors allow current to flow only in one direction: from V


DD


toward the output, V


BIAS


. Each pumping stage M


1


/C


1


, M


2


,C


2


, M


3


/C


3


, etc., increases the peak output voltage by approximately the value of V


DD


.




The voltage at the output of any pumping stage N alternates between N×V


DD


and (N+1)×V


DD


. A voltage fluctuation of V


DD


, however, would be an unacceptably large a.c. component superimposed on the bias voltage. The final output filter stage MX/CX comprising transistor MX and capacitor CX acts to remove most of the a.c. component and provide a stable voltage for V


BIAS


. In practice, capacitor CX will typically consist of the large parasitic capacitance that is inherent in the construction of the integrated circuit. The average current drawn from this capacitance is the leakage current through reverse biased P-N junctions and is therefore quite small. As a result, relatively small transistors can be used in the pumping stages and the output filter, and V


BIAS


will be substantially free of switching noise.




The diode-connected transistors in the charge pump, however, do not act as ideal diodes. They will not start to conduct current until the voltage at the source node exceeds the voltage at the common gate/drain node by at least the threshold voltage, V


TH


, of the transistor. Given this factor the peak voltage at state N of the charge pump is (N+1)×(V


DD


−V


TH


). It can be understood that if the supply voltage V


DD


were reduced to an amount equal to the threshold voltage V


TH


, the charge pump will cease to pump any charge. According to the preferred embodiment, therefore, the respective threshold voltages V


TH


of transistors M


1


-MX is less than or equal to 100 mV. At these threshold levels, the charge pump


700


is able to operate at ultra low power voltage levels of 500 mV or less. Although the preferred embodiment for charge pump


700


incorporates MOSFETs for transistors M


1


-MX, in applications where the supply voltage V


DD


is above 1 volt, embodiments are envisioned which substitute conventional P-N junctions for the diode-connected MOSFETs.




The example described here used PMOS transistors to generate a V


BIAS


voltage that is above the level of V


DD


. However, those skilled in the art will recognize that the PMOS transistors could be replaced with NMOS transistors and the source of the first transistor connected to ground. This modification would result in a V


BIAS


level that was below ground.




Although the leakage current drawn from VBIAS is small, it is not negligible, and the charge pump must constantly resupply this current. Each toggle of the control pulse


232


, both the on-to-off transition, and the off-to-on transition, equally function to accumulate charge on the respective capacitors, or, alternatively, to pump charge from a capacitor across the next transistor to the next junction. As the duty cycle of the control pulse


232


is shortened from 50% to 0% however, the charging period for one set of capacitors is progressively shortened. Those familiar with the charging characteristics of a capacitor will therefore understand that a maximum voltage output of the charge pump will occur when the control pulse


232


duty cycle is at 50%. According to the preferred embodiment of the present invention, during steady state operation, the propagation time of the delay element


220


,


302


will be slightly less than the CLOCK pulse width, causing narrow pulses on the CONTROL PULSE


232


output, as shown in FIG.


4


. Moreover, by setting a steady state operation such that the duty cycle of the control pulse


232


is between zero-and fifty percent, the back bias voltage


236


produced at the charge pump


214


output can be increased or decreased by increasing or decreasing the duty cycle of the control pulse


232


. According to embodiments of the present invention utilizing an AND gate as a comparator, the duty cycle can have a maximum of 50% and a minimum of zero. However, alternative logic comparators such as a NAND gate are envisioned within the scope of the present invention. In some alternative designs, the duty cycle would vary from a minimum of 50 percent to a maximum of 100 percent. In substituting such equivalent embodiments of the present invention, however, the fundamental operation remains the same. The charge produced by the charge pump reaches a maximum at a 50 percent duty cycle, and decreases toward a zero voltage output as the duty cycle moves away from the fifty percent point. Such equivalent embodiments are envisioned within the scope of the present invention.




Referring again to

FIG. 6

, it can be understood that when the delay element


220


,


302


is too fast, the increased duty cycle of the control pulse


232


signal will cause the charge pump to quickly raise the voltage on V


BIAS


. As V


BIAS


rises it causes the threshold voltages of the transistors in the delay element to increase their switching time and thus increases T


DELAY


. As T


DELAY


is increased, the pulses on PULSE become narrower until the steady state condition shown in

FIG. 4

is reached.




It can similarly be understood that when the delay element is too slow and the duty cycle of the control pulse


232


is shorter than the steady state duration of

FIG. 4

, such as the no-control-pulse condition of

FIG. 5

, leakage currents will discharge CX and cause the voltage on V


BIAS




236


to fall. As V


BIAS




236


falls, the threshold voltages of the transistor in the delay element will decrease, which speeds up the switching time of the inverters


307


-


309


, or any other transistor structure used in the delay element


220


, thereby decreasing T


DELAY


. Once T


DELAY


becomes less than the width of the CLOCK pulse


234


(

FIG. 5

) then pulses will start to appear on the control pulse


232


. When the control pulse


232


becomes wide enough (

FIG. 4

) to replenish the charge on CX lost to leakage currents, the circuit of

FIG. 2

will have again reached the steady state condition.




Since V


BIAS




236


is applied to the functional circuit as well as the delay element, and the propagation delay of the delay element is directly related to the critical path delay in the functional circuit, the critical path delay will also be slaved to the width of the CLOCK pulse


234


. The present invention will adjust V


BIAS


as necessary to maintain this relationship despite the effects of aging, radiation, and temperature. By setting V


BIAS




236


and hence the transistor threshold voltages, at the highest level that allows the circuit to meet its functional timing requirements, the present invention also minimizes power consumption due to transistor leakage current.




The present invention provides a method and apparatus for reducing the power consumption of a CMOS circuit by controlling back bias voltage applied to a CMOS circuit. The back bias is controlled by means of a closed loop circuit that detects change in the operational speed of a delay element. A clock pulse


234


drives a functional circuit


228


and a delay element


220


. Any change in the time delay experienced within the delay element


220


is detected by processing the output of the delay element and the clock pulse


234


through an AND gate, the output of which forms a control pulse


232


. As the duty cycle of the control pulse falls from fifty percent to zero, a charge pump decreases the voltage output. The voltage output serves as a back bias voltage for both the delay element and the functional circuit. The closed loop circuit therefore allows the delay element to operate at a predetermined speed. The predetermined speed can be set to minimize power consumption while maintaining the circuit speed of the functional circuit


228


at a reliable level. This process avoids the addition of excessive ancillary components such as fixed reference voltage sources, fixed current sources, fixed resistors forming voltage dividers for reference voltages, variable speed clocks, op-amps, or multiple memory locations storing data defining various voltage levels for application to a back-bias voltage. By avoiding these components which are largely foreign to CMOS devices, a low cost circuit can be configured that is capable of adjusting the body bias on a CMOS circuit so as to minimize power consumption over the life of the circuit while maintaining the circuit at a requisite operational speed.



Claims
  • 1. An apparatus for regulating a back bias voltage in a functional circuit, comprising:a. a closed loop voltage regulator comprising: i. charge pump including a charge pump input and a charge pump output providing the back bias voltage; ii. a delay element including a delay clement input coupled to a clock pulse, a delay element output, and a delay element body, wherein the delay element body is coupled to the charge pump output; and b. the functional circuit is coupled to the clock pulse that is also coupled to the delay element input, wherein the functional circuit has a body that is coupled to the back bias voltage, and wherein the back bias voltage is dependent upon a duty cycle of a control pulse at the charge pump input, and the back bias voltage and a switching speed of the delay element operate in an inverse relationship, such that when the back bias voltage is lowered the switching speed of the delay element is increased, thereby reducing a time delay, and such that when the back bias voltage is increased the switching speed of the delay element is lowered, thereby increasing the time delay.
  • 2. The apparatus according to claim 1, wherein the closed loop voltage regulator further comprising a logic element having a logic element output coupled to the charge pump input, the logic element further comprising a first logic input coupled to the clock pulse, and a second logic input coupled to the delay element output.
  • 3. The apparatus according to claim 2 wherein the logic element is an AND gate.
  • 4. The apparatus according to claim 2 wherein the delay element comprises a plurality of logical inverters coupled in series.
  • 5. The apparatus according to claim 2 wherein the functional circuit is a CMOS circuit comprising a substrate, and wherein the substrate comprises the body of the functional circuit.
  • 6. The apparatus according to claim 5 wherein the charge pump output is coupled to the substrate of the functional circuit through a proportional element, wherein the back bias voltage applied to the substrate is proportional to the charge pump output voltage.
  • 7. The apparatus according to claim 5 wherein the charge pump output is coupled directly to the substrate of the functional circuit.
  • 8. The apparatus according to claim 2 wherein the delay element and the functional circuit are comprised of materials configured to degrade in performance at a substantially identical rate over time.
  • 9. The apparatus according to claim 2 wherein the charge pump is comprised of a plurality of pumping stages, wherein each pumping stage includes a transistor having a gate coupled to a capacitor.
  • 10. The apparatus according to claim 9 wherein the transistor within at least one of the pumping stages exhibits a threshold voltage of less than 110 mV.
  • 11. A method for regulating a back bias voltage on a functional circuit comprising:a. controlling a duty cycle of a control signal controlling a switching speed; b. adjusting an output voltage of a charge pump according to the duty cycle of the control signal; and c. applying the back bias voltage to the functional circuit, wherein the back bias voltage is related to the output voltage of the charge pump and inversely related to the switching speed of a delay element, such that when the back bias voltage is lowered the switching speed of the delay element is increased, thereby reducing a time delay, and such that when the back bias voltage is increased the switching speed of the delay element is lowered, thereby increasing the time delay, and wherein the control signal is generated at an output of a logic element.
  • 12. The method according to claim 11 wherein the step of controlling a duty cycle comprises the steps:a. receiving a clock pulse into an input of the delay element; b. receiving the clock pulse into a first input of the logic element; and c. receiving an output signal generated at an output of the delay element into a second input of the logic element.
  • 13. The method according to claim 12 wherein the logic element is an AND gate.
  • 14. The method according to claim 12 further comprising the step of coupling the output of the logic element to an input of the charge pump.
  • 15. The method according to claim 12 wherein the output voltage of the charge pump is at a maximum voltage when the duty cycle of the control signal is approximately 50%.
  • 16. The method according to claim 12 wherein the delay element and the functional circuit are comprised of materials configured to degrade at a substantially identical rate over time.
  • 17. The method according to claim 12 wherein the charge pump is comprised of a plurality of pumping stages, wherein each pumping stage includes a transistor having a gate coupled to a capacitor.
  • 18. The method according to claim 17 wherein the transistor within at least one of the pumping stages exhibits a threshold voltage of less than 110 mV.
  • 19. The method according to claim 12 wherein the delay element comprises a plurality of logical inverters coupled in series.
  • 20. The method according to claim 13 wherein the functional circuit is a CMOS circuit comprising a substrate, and wherein the substrate comprises a body of the functional circuit.
  • 21. A method or reducing power consumption within a functional CMOS circuit comprising a plurality of transistors while maintaining a switching time of the plurality of transistors within an optimum range, the method comprising the steps:1. altering a duty cycle of a control-pulse; 2. regulating a back-bias voltage according to the duty-cycle of the control-pulse, such that if a time delay produced by a delay element exceeds an optimal delay, the duty cycle of the control-pulse is shortened, and the back bias voltage is lowered, the back-bias voltage and a switching speed of the delay element operating in an inverse relationship; and 3. applying the back-bias voltage to the functional CMOS circuit, wherein the control pulse is formed at an output of an AND gate.
  • 22. The method of reducing power consumption according to claim 21 wherein the step of altering the duty cycle of the control-pulse comprising the steps:a. transmitting a clock input signal into a first input of the AND gate; b. delaying the clock input signal through the delay element to form a delayed clock signal; and c. transmitting the delayed clock signal to a second input of the AND gate.
  • 23. The method of reducing power consumption according to claim 22 further comprising the step of applying the back-bias voltage to the delay element.
  • 24. The method according to claim 23 wherein the output of the AND gate is coupled to an input of a charge pump, and wherein the back bias voltage is proportional to an output voltage of the charge pump.
  • 25. An apparatus for regulating a back bias voltage in a functional circuit, comprising:a. means for comparing a clock signal and a delay clock signal, wherein the means for comparing produces a control pulse output; b. means for producing the back bias voltage according to a duty cycle of the control pulse output; c. means for delaying having an input coupled to the clock signal, wherein an output of the means for delaying forms the delay clock signal, and wherein the back bias voltage is coupled to a body of the means for delaying and is inversely related to a switching speed of the means for delaying, such that when the back bias voltage is lowered the switching speed of the means for delaying is increased, thereby reducing a time delay, and such that when the back bias voltage is increased the switching speed of the means for delaying is lowered, thereby increasing the time delay; and d. means for performing a digital function, wherein a body of the means for performing is coupled to the back bias voltage, the means for performing is also coupled to receive the clock signal.
  • 26. The apparatus according to claim 25 wherein the means for comparing is an AND gate.
  • 27. The apparatus according to claim 25 wherein the means for delaying comprises a plurality or logical inverters coupled in series.
  • 28. The apparatus according to claim 25 wherein the means for performing a digital function comprises a CMOS circuit.
  • 29. The apparatus according to claim 25 wherein the means for delaying and the means for performing a digital function are comprised of materials configured to degrade in performance at a substantially identical rate over time.
  • 30. The apparatus according to claim 25 wherein the means for producing said back bias voltage comprises a plurality of pumping stages, wherein each pumping stage includes a transistor having a gate coupled to a capacitor.
  • 31. The apparatus according to claim 26 wherein the transistor within at least one of the pumping stages exhibits a threshold voltage of less than 110 mV.
  • 32. An apparatus for regulating a back bias voltage in a functional circuit, comprising:a. a closed loop voltage regulator comprising: i. charge pump including a charge pump input and a charge pump output providing the back bias voltage; ii. a delay element including a delay element input coupled to a clock pulse, a delay element output, and a delay element body, wherein the delay element body is coupled to the charge pump output; and b. the functional circuit with a body coupled to the charge pump output, the functional circuit is also coupled to the clock pulse, the functional circuit being manufactured on the same integrated circuit as the closed loop voltage regulator.
  • 33. An apparatus for regulating a back bias voltage in a functional circuit, comprising:a. a closed loop voltage regulator comprising: i. charge pump including a charge pump input and a charge pump output providing the back bias voltage; ii. a delay element including a delay element input coupled to a clock pulse, a delay element output, and a delay element body, wherein the delay element body is coupled to the charge pump output; and b. the functional circuit with a body coupled to the charge pump output, the functional circuit is also coupled to the clock pulse, the functional circuit being manufactured on the same integrated circuit as the closed loop voltage regulator, and wherein the back bias voltage is dependent upon a duty cycle of a control pulse at the charge pump input, and the back bias voltage and a switching speed of the delay element operate in an inverse relationship, such that when the back bias voltage is lowered the switching speed of the delay element is increased, thereby reducing a time delay, and such that when the back bias voltage is increased the switching speed of the delay element is lowered, thereby increasing the time delay.
GOVERNMENT LICENSE RIGHTS

The U.S. Government has a paid-up license in this invention and the right in limited circumstances to require the patent owner to license to others on reasonable terms as provided by terms of the Federal Grant No. NAG5-8392 awarded by NASA for the project entitled “RADIATION TOLERANT VLSI”.

US Referenced Citations (10)
Number Name Date Kind
3609414 Pleshko et al. Sep 1971 A
4142114 Green Feb 1979 A
4670670 Shoji Jun 1987 A
5682118 Kaenel et al. Oct 1997 A
5744996 Kötzle et al. Apr 1998 A
6150858 Sung Nov 2000 A
6166577 Mizuno et al. Dec 2000 A
6466077 Miyazaki et al. Oct 2002 B1
6501313 Boerstler et al. Dec 2002 B2
20020079939 Nair et al. Jun 2002 A1
Non-Patent Literature Citations (8)
Entry
Vadim Gutnik et al., “Embedded Power Supply for Low-Power DSP”, IEEE Trans. on VLSI Systems, vol. 5, No. 4, Dec. 1997, pp. 425-435.
Vincent V. Kaenel et al., “A Voltage Reduction Technique for Battery-Operated Systems”, IEEE Journal of Solid State Circuits, vol. 25, No. 5, Oct. 1990, pp. 1136-1140.
Shih-Wei Sun et al., “Limitation of CMOS Supply-Voltage Scaling by MOSFET Threshold-Voltage Variation”, IEEE Journal of Solid State Circuits, vol. 30, No. 8, Aug. 1995, pp. 947-949.
Koichi Nose et al., “Optimization of VDD and VTH for Low-Power and High-Speed Applications”, Proceedings of Asia and South Pacific Design Automation Conference, Jan. 2000, pp. 469-474.
Ricardo Gonzales et al., “Supply and Threshold Voltage Scaling for Low Power CMOS”, IEEE Journal of Solid State Circuits, vol. 32, No. 8, Aug. 1997, pp. 1210-1216.
Takayasu Sakurai et al., “Low Power Design of Digital Circuits”, International Symposium on Key Technologies for Future VLSI Systems, Jan. 2000, pp. 1-5.
Tadahiro Kuroda et al., “Variable Supply-Voltage Scheme for Low-Power High-Speed CMOS Digital Design”, IEEE Journal of Solid State Circuits, vol. 33, No. 3, Mar. 1998, pp. 454-461.
Masayuki Miyazaki et al., “A Delay Distribution Squeezing Scheme with Speed-Adaptive Threshold-Voltage CMOS (SA-Vt CMOS) for Low Voltage LSIs”, International of Symposium on Low Power Electronics and Design, 1998, pp. 48-53.