WO2007135487A1

WO2007135487A1 - Device and method for reducing power consumption

Info

Publication number: WO2007135487A1
Application number: PCT/IB2006/051611
Authority: WO
Inventors: Michael Priel; Dan Kuzmin; Anton Rozen
Original assignee: Freescale Semiconductor, Inc.
Priority date: 2006-05-19
Filing date: 2006-05-19
Publication date: 2007-11-29

Abstract

A method and device for power reduction. The device includes a first group (20) of latches located within at least one normal region (40) of an integrated circuit, and second group (30) of latches that are located within at least one retention region (50) of the integrated circuit. The area of the at least one normal region (40) is larger than an area of the at least one retention region (50). The first group (20) of latches is adapted to latch data signals. The second group of latches is adapted to store information representative of the data signals while the first group (20) of latches is deactivated and is further adapted to send to the first group (20) of latches said information when the first group of latches (20) is reactivated.

Description

DEVICE AND METHOD FOR REDUCING POWER CONSUMPTION

FIELD OF THE INVENTION

The present invention relates to devices that have power reduction capabilities and to methods for reducing power consumption.

BACKGROUND OF THE INVENTION

Mobile devices or devices, such as but not limited to personal data appliances, cellular phones, radios, pagers, lap top computers, and the like are required to operate for relatively long periods before being recharged. These mobile devices usually include one or more processors as well as multiple memory modules and other peripheral devices.

The power consumption of a transistor-based device is highly influenced by leakage currents that flow through the transistor. The leakage current is responsive to various parameters including the threshold voltage (Vt) of the transistor, the temperature of the transistor, supply voltage and the like. Transistors that have higher Vt are relatively slower but have lower leakage currents while transistors that have lower Vt are relatively faster but have higher leakage current. In order to reduce the power consumption of mobile devices various power consumption control techniques were suggested. A first technique includes reducing the clock frequency of the mobile device. A second technique is known as dynamic voltage scaling (DVS) or alternatively is known as dynamic voltage and frequency scaling (DVFS) and includes altering the voltage that is supplied to a processor as well as altering the frequency of a clock signal that is provided to the processor in response to the computational load demands (also referred to as throughput) of the processor. Higher voltage levels are associated with higher operating frequencies and higher computational load but are also associated with higher energy consumption. A third technique uses domino circuits that include both high threshold voltage transistors and low threshold voltage transistors. U.S. patent application number 2004/0008056 of Kursun et al . , which is incorporated herein by reference, discloses a domino circuit that is configured such as to reduce power consumption, for example by limiting the energy consumed during power switching .

Yet another technique is based upon creating a stack effect that involves shutting down multiple transistors of the same type that are serially connected to each other. U.S. patent 6169419 of De et al . , which is incorporated herein by reference, discloses a method and apparatus for reducing standby leakage current using a transistor stack effect. De describes a logic that has both a pull up path and a pull down path.

In some prior art devices data is stored in data retention circuits while other parts of the device are shut down. The following patents and patent applications, all being incorporated herein by reference, provide a brief overview of prior art retention circuits: U. S patent application publication number 2004/0051574 of Ko et al; PCT patent application publication number WO 2004/021351A1 of Garg et al; U.S. patent 5600588 of Kawashima; U.S. patent application 2004/0227542 of Bhavnagarwala et al . and U.S. patent 6755180 of Biyani et al.

Many data retention circuits involve providing multiple power grids. A first power grid provides a "regular" power signal that is shut down during a shut down mode. A second power grid provides a retention power signal that is provided even during the shut down mode. Multiple power grids are size consuming and also complicate the design of the integrated circuit. U. S patent application publication number

2004/0051574 of Ko et al describes a flip-flop that includes two latches and a data retention circuit that are highly complex, include multiple transistors, are area consuming, can enter unstable or non-defined states that can damages the data retention circuit.

There is a growing need to find effective devices and methods for reducing power consumption of integrated circuits while retaining data.

SUMMARY OF THE PRESENT INVENTION

A device and a method for reducing power consumption of a transistor-based circuit, as described in the accompanying claims .

BRIEF DESCRIPTION OF THE DRAWINGS

The present invention will be understood and appreciated more fully from the following detailed description taken in conjunction with the drawings in which : FIG. 1 illustrates a device, according to an embodiment of the invention;

FIG. 2 illustrates three latches, according to an embodiment of the invention;

FIG. 3 is a timing diagram illustrating the voltage supply signals provided to the three latches, according to an embodiment of the invention;

FIG. 4 illustrates an integrated circuit that includes multiple regions and a single other region, according to an embodiment of the invention; FIG. 5 illustrates an integrated circuit that includes a single regions and a single other region, according to an embodiment of the invention;

FIG. 6 illustrates an integrated circuit that includes multiple regions and multiple other regions, according to an embodiment of the invention;

FIG. 7 is a flow chart of a method for reducing power consumption of a device, according to various embodiments of the invention.

DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS

According to various embodiments illustrated below a device can operate in various modes that differ from each other by the power consumed by the device. Retention latches operate during a low power mode (such as idle, standby or shut down modes) while normal latches operate during other operational modes (referred to as normal power mode) .

Conveniently, the normal latches form a first group of latches while the retention latches form a second group of latches.

Conveniently, the retention latches are concentrated in one or more relatively small retention regions while the normal latches can be spread over at least one larger normal region. A normal region is a region that includes one or more latches that are deactivated during a low power mode while a retention region is a region that includes at least one or more latches that are active during the low power mode. Conveniently, the retention latches are characterized by smaller leakage current but can be slower than the normal latches. This can be achieved by using higher threshold voltages, thicker transistors, applying different manufacturing processes, and the like. Conveniently, the state of a normal latch can affect a state of a transistor switch that can selectively connect a corresponding retention latch to the ground or to another predefined voltage. If the transistor switch remains open the retention latch converges to a certain state that differs from the state to which the retention latch converges when grounded. This convergence can be achieved by using an asymmetrical retention latch in which the pull down transistors differ from the pull up transistors. For example, NMOS transistors can be stronger than corresponding PMOS transistors.

Conveniently, after the idle mode ends the state of a retention latch can affect a state of another transistor switch that can selectively connect the corresponding normal latch to the ground or to another predefined voltage. If the other transistor switch remains open the normal latch converges to a certain state. This convergence can be achieved by using an asymmetrical latch in which the pull down circuitry differs from the pull up circuitry. For example, NMOS transistors can be stronger than corresponding PMOS transistors .

The following examples refer to a shut down mode as well as a "normal" operational mode of multiple latches. It is noted that the method can be applied, mutatis mutandis, to devices in which there are multiple operational modes that are characterized by different power consumption level. For example, the "normal" operational mode can include a medium power consuming mode, a high power consuming mode and the like while the shut down mode can include idle modes or other low power modes .

It is further noted that all the drawings are out of scale . FIG. 1 is a schematic illustration of a device, such as an integrated circuit, 10 that includes various devices. Each device may include a large number of power reduction circuitries, according to an embodiment of the invention.

Device 10 includes various units such as but not limited to general purpose processor 12, I/O module 14, memory unit 16, peripheral 18, and digital signal processor (DSP) 21. These units are linked to each other by various lines and buses and receive clock signals and power supply from one or more sources, such as clock signal source 17 and voltage source 15. It is noted that device 10 can include other units, that some of these unit are optional and that device 10 can include multiple units of the same kind.

Optionally, the clock signal source 11 and the voltage source 15 are also connected to a synchronization control unit 13 that matches between the clock frequency and supplied voltages, such as to prevent a case in which the voltage supplied to one or more units is too low to support the clock frequency of the clock signal. This matching is useful when altering the operational mode of the integrated circuit and applying DVFS techniques. It is noted that this configuration is optional and that device 10 does not necessarily apply DVFS techniques.

Typically, device 10 includes multiple busses and lines and the various units of device 10 can be connected to the same bus, but this is not necessarily so. For convenience of explanation FIG. 1 illustrates a device bus that is shared by units 12, 14, 16, 18 and 21.

It is noted that device 10 can have various configurations and that the units illustrated in FIG. 1 represent only a single exemplary configuration of a device that applies the power reduction technique. Typically, device 10 is included within a mobile device such as a cellular phone, a music player, a video player, a personal data accessory, and the like.

Modern cores such as processor 12 and DSP 21 can include millions of transistors. Device 10, or at least some of its units (such as but not limited to processor 12 and DSP 21) can operate in various operational modes, including low power modes such as but not limited to an idle (also being referred to a shut down or standby) mode. During an idle mode it is desired to reduce the power consumption of a device, especially in view of the low computational load imposed on said device during the idle mode .

It is noted that although the following figures illustrate a first (normal) power supply such as VDD 162 as well as a second (retention) power supply VERT 164 that multiple power supplies can be used for powering different regions of the device. For example, the device 10 can include multiple voltage/frequency regions. Each voltage/frequency region can be powered by its own power supply. Accordingly, different retention latches (latches that belong to a second group of latches) can receive different retention power signals.

Typically, once device 10, or one of its units (it is assumed for convenience of explanation that processor 12 makes this decision) decides to enter an idle mode. Once this decision is made the retention power supply is asserted and after a short period (required for storing data at the retaining latches) the normal power supply is negated. This idle mode ends by asserting the normal power supply and after a short period (required for restoring the data to the normal latches) the retention power supply is negated. FIG. 2 illustrates three latches, according to an embodiment of the invention.

The three latches include a retention latch 30-1 and two normal latches 20-1 and 20-2. Latches 20-1 and 20-2 are serially connected to each other to provide flip-flip 111. Retention latch 30-1 is positioned within other area 50 while normal latches 20-1 and 20-2 can be positioned close to retention latch 30-1 or even relatively far away from retention latch 30-1. Normal latches 20-1 and 20-2 are powered by normal power supply VDD 162. Retention latches 30-1 is powered by retention power supply VRET 164.

First normal latch 20-1 includes a first transfer gate 101, two inversely connected inverters 112 and 114, and a third inverter 116.

An input node of first transfer gate 101 forms the input node of flip-flop 111. The output of the first transfer gate 101 is connected to an input of inverter 112 and to an output of inverter 114. The input of inverter 114 is connected to the output of inverter 112 and to an input mode of third inverter 116.

Second normal latch 20-2 includes a second transfer gate 118, two inversely connected inverters 124 and 126, a sixth inverter 116 and transistor switch Tl 132. An input node of second transfer gate 118 is connected to an output of third inverter 118. The output of transfer gate 118 is connected (at node Nl) to an input of inverter 122, to an output of inverter 124 and to a source of transistor switch Tl 132. The input of inverter 124 is connected to the output of inverter 122 and to an input mode of sixth inverter 126.

The drain of transistor switch Tl 132 is connected to the ground and the gate of transistor switch Tl 132 is connected to an output node (N3) of the retention latch, for receiving output signal DLOUT 108, when the retention latch 30-1 is active. If DLOUT 108 is asserted than transistor switch 132 is opened and it grounds node Nl. If DLOUT 108 is negated transistor switch 132 is closed and does not affect the state of second normal latch 20- 2. Conveniently, the second normal latch 20-2 is an asymmetrical latch thus it converges to a state in which the voltage at Nl is "1" and the output signal (DOUT 106) of the second normal latch 20-2 is also "1". Retention latch 30-1 includes two inversely connected inverters 142 and 144 and retention transistor switch T2 134.

The source of retention transistor switch T2 134, the input of inverter 142 and the output of inverter 144 are connected to each other (at node N2) to form an input of retention latch 30-1. The input of inverter 144 and the output of inverter 142 are connected to each other (at node N3) to form an output of retention latch 30-1. This output provides (when retention latch 30-1 is powered by VERT 164) a retention latch output signal DLOUT 108.

The drain of retention transistor switch T2 134 is connected to the ground and the gate of retention transistor switch T2 134 is connected to an output node of second latch 20-2 for receiving output signal DOUT 106, when the second normal latch 20-2 is active. If DOUT 106 is asserted than retention transistor switch 134 is opened and it ground node N2 thus causing DLOUT to be asserted. If DOUT 106 is negated retention transistor switch T2 134 is closed and does not affect the state of retention latch 30-1. Conveniently, the retention latch 30-1 is an asymmetrical latch thus is converges to a state in which the voltage at N2 is "1" and the output signal (DOUT 108) of the retention latch 30-1 is "0". It is noted that other latches can be used. For example, a latch can include a logic gate that differs from an inverter and that other logic gate can be responsive to additional signals (such as reset, and the like) .

FIG. 3 is a timing diagram 160 illustrating the voltage supply signals provided to the three latches, according to an embodiment of the invention.

At time Tl device 10 operates at a normal operational mode, thus the normal latches are active while retention latches are not active. VDD 162 is high (VDD="1") and VRET 164 is low (VRET="0") .

At a time between Tl and T2 device 10 decides to enter an idle mode. Accordingly, between T2 and T3 VRET 164 is asserted and the retention latches to allow these latches to store data representative of the state of the normal latches before entering the idle mode.

Referring to the example set forth in FIG. 2, if, for example, second normal latch 20-2 outputs a high level output signal (DOUT 106 = "1") then the retention transistor switch T2 is opened, thus causing node N2 to be grounded and cause node N3 to at a high level (DLOUT 108="l") . Before DLOUT 108 can affect switch transistor Tl 132 VDD 162 is negated. As illustrated by timing diagram 160 VDD 162 is negated between T4 and T5.

Referring to the example set forth in FIG. 2, if, for example, second normal latch 20-2 outputs a low level output signal (DOUT 106 = "0") then the retention transistor switch T2 remains closed, and the retention latch 30-1 converges to a state in which the voltage at node N2 is "1" and the voltage at node N3 is low (DLOUT 108="0") .

Between T5 and T6 device 10 operates in an idle mode. It is assumed that prior to T6 device 10 decides to exit the idle mode thus it asserts (between T6 and T7) VDD 162. VRET 164 remains high a short while after VDD 162 is asserted, in order to allow the data stored in the retention latch to affect the normal latches. After this short while VRTE 164 is negated. (Negation occurs between T8 and T9) .

According to an embodiment of the invention the normal latches are spread in one or more normal regions while the retention latches are concentrated in one or more other regions of device 10. The separation between normal latches and retention latches allows using different retention latches and even use different manufacturing processes (for example applying different lithographic masks) for manufacturing the normal latches and the retention latches.

When multiple processes are applied a certain guard area (such as guard area 70) should be provided between transistors that are manufactured in different processes. By placing many retention latches together the overall (aggregate) size of the guard zone can be reduced.

Various configurations, illustrating one or more normal regions and one or more retention regions are illustrated in FIG. 4 - FIG. 6. The size of the retention region is usually much smaller than the size of the normal regions .

FIG. 4 illustrates integrated circuit 60 that includes multiple regions 40 and a single other region 50, according to an embodiment of the invention.

A region such as region 40 includes at least a portion of a first group of latches 20. Latches 20-1 and 20-2 belong to this group. It is noted that the number of latches that belong to this first group can be vary large . A region such as other region 50 includes at least a portion of a second group of latches 30. Latch 30-1 belongs to this group. It is noted that the number of latches that belong to this second group can be vary large.

FIG. 5 illustrates integrated circuit 60' that includes a single region 40 and a single other region 50, according to an embodiment of the invention. FIG. 6 illustrates integrated circuit 60" that includes multiple regions 40 and multiple other regions 50, according to an embodiment of the invention.

FIG. 7 is a flow chart of method 200 for reducing power consumption of a device, according to various embodiments of the invention. Method 200 starts by stage 210 of latching data signals at a first group of latches, wherein the first group of latches is located within at least one normal region of an integrated circuit.

Stage 210 is followed by stage 220 storing at a second group of latches (located within at least one retention region) information representative of these data signals. The area of the at least one normal region is larger than an area of the at least one retention region. If there are multiple regions of a certain type then their area is the aggregate area of the regions of that same type. A normal region does not overlap a retention regions

Stage 220 is followed by stage 230 of shutting down the second group of latches. Stage 230 may include entering an idle mode.

Stage 230 is followed by stage 240 of reactivating the first group of latches and sending from the second group of latches the information and shutting down the second group of latches. Conveniently, the latching includes latching data signals at a first group of latches that are positioned outside the at least one retention region.

Conveniently, the latching includes latching data signals at a first group of latches fabricated by using a first fabrication process and the storing includes storing information at a second group of latches that are fabricated by a second fabrication process that differs from the first fabrication process. Conveniently, the latching includes latching data signals at a first group of latches that include field effect transistors while the storing includes storing information at a second group of latches that include other types of transistors. The other type of transistors can be manufactured by different manufacturing process, be produced by different lithographic masks and the like. For example, the other type of transistors can include DGO transistors, but this is not necessarily so.

Conveniently, the storing includes storing information at a second group of latches that comprises asymmetrical latches.

Conveniently, the storing includes sending a data signal from a first latch towards a retention transistor switch that is coupled between a ground and two inversely coupled inverters that belong to a retention latch.

Conveniently, the reactivation includes sending a data signal from a retention latch towards a transistor switch that is coupled between a ground and two inversely coupled inverters that belong to a latch of the first group of latches.

Variations, modifications, and other implementations of what is described herein will occur to those of ordinary skill in the art without departing from the spirit and the scope of the invention as claimed. Accordingly, the invention is to be defined not by the preceding illustrative description but instead by the spirit and scope of the following claims.

Claims

WE CLAIM

1. A device (10) comprising a first group (20) of latches located within at least one normal region (40) of an integrated circuit; wherein the device (10) is characterized by comprising a second group (30) of latches that are located within at least one retention region (50) of the integrated circuit; wherein an area of the at least one normal region (40) is larger than an area of the at least one retention region (50); wherein a normal region does not overlap a retention region; wherein the first group (20) of latches is adapted to latch data signals; wherein the second group of latches is adapted to store information representative of the data signals while the first group (20) of latches is deactivated and is further adapted to send to the first group (20) of latches said information when the first group of latches (20) is reactivated.

2. The device (10) according to claim 1 wherein the latches of first group of latches are positioned outside the at least one retention region (50) .

3. The device (10) according to any claim of claims 1-2 wherein the second group (30) are fabricated by a second fabrication process that differs from a first fabrication process of the first group (20) of latches.

4. The device (10) according to any claim of claim 1-3 wherein the second group of latches comprise DGO transistors while the first group of latches comprise field effect transistors.

5. The device (10) according to any claim of claims 1-4 wherein the second group (30) of latches comprises asymmetrical latches.

6. The device (10) according to any claim of claims 1-5 wherein a first latch of the first group of latches (20) comprises two inversely coupled inverters and a transistor switch that is adapted to selectively couple the first latch to a ground in response to a signal provided from a corresponding retention latch that belongs to the second group of latches .

7. The device (10) according to claim 6 wherein the corresponding retention latch comprises two inversely coupled inverters and a transistor switch that is adapted to selectively couple the corresponding retention latch to a ground in response to a data signal provided from the first latch.

8. The device (10) according to any claim of claims 5-6 wherein the first latch comprises asymmetric inverters.

9. The device (10) according to any claim of claims 1-8 wherein the second group (30) of latches is located in multiple retention regions (50) and wherein an aggregate area of the multiple retention regions (50) is smaller than an area of the at least one normal region (40) .

10. A method (200) comprising: latching (210) data signals at a first group of latches, wherein the first group of latches is located within at least one normal region of an integrated circuit; the method (200) is characterized by further including: storing (220) at a second group of latches located within at least one retention region (50) information representative of these data signals; wherein an area of the at least one normal region (40) is larger than an area of the at least one retention region (50) and normal regions do not overlap retention regions; shutting down (230) the second group of latches; and reactivating (240) the first group of latches and sending from the second group of latches the information and shutting down the second group of latches .

11. The method (200) according to claim 10 wherein the latching (210) comprises latching data signals at a first group of latches that are positioned outside the at least one retention region.

12. The method (200) according to any claim of claims 10-11 wherein the latching (210) comprises latching data signals at a first group of latches fabricated by using a first fabrication process and wherein the storing (220) comprises storing information at a second group of latches that are fabricated by a second fabrication process that differs from the first fabrication process.

13. The method (200) according to any claim of claims 10-12 wherein the latching (210) comprises latching data signals at a first group of latches that comprise field effect transistors while the storing (220) comprises storing information at a second group of latches that comprises DJO transistors .

14. The method (200) according to any claim of claims 10-13 wherein the storing (220) comprises storing information at a second group of latches that comprises asymmetrical latches.

15. The method (200) according to any claim of claims 10-14 wherein the storing (220) comprises sending a data signal from a first latch towards a retention transistor switch that is coupled between a ground and two inversely coupled inverters that belong to a retention latch.

16. The method (200) according to any claim of claims 10-14 wherein the reactivation (240) comprises sending a data signal from a retention latch towards a transistor switch that is coupled between a ground and two inversely coupled inverters that belong to a latch of the first group of latches.