Leakage Power Reduction Techniques - PowerPoint PPT Presentation

About This Presentation
Title:

Leakage Power Reduction Techniques

Description:

Leakage Power Reduction Techniques Yuanlin Lu ECE Dept. Auburn University ELEC 6970 Outline Transistor Leakage Mechanisms Leakage Reduction techniques - Mutli-, dual ... – PowerPoint PPT presentation

Number of Views:380
Avg rating:3.0/5.0
Slides: 39
Provided by: Grac81
Category:

less

Transcript and Presenter's Notes

Title: Leakage Power Reduction Techniques


1
Leakage Power Reduction Techniques
  • Yuanlin Lu
  • ECE Dept. Auburn University
  • ELEC 6970

2
Outline
  • Transistor Leakage Mechanisms
  • Leakage Reduction techniques
  • - Mutli-, dual-, Variable Vth
  • - Dual Power Supply
  • - Transistor Sizing
  • - Transistor Stacking
  • - Optimal Input Vector Selection
  • Proposed Technique
  • - Using ILP to Minimize leakage
  • - Extend ILP to Minimize leakage and Glitch
    Power together

3
Transistor Leakage Mechanisms
  • I1 - the reverse-bias pn junction leakage
  • I2 - the subthreshold leakage weak inversion
    conduction current between source and drain in an
    MOS transistor occurs when gate voltage is below
    Vth.
  • I3 - the oxide tunneling current due to the low
    oxide thickness and the high electric field
  • I4 - the gate current due to hot-carrier
    injection
  • I5 - the GIDL (Gate-Induced Drain Leakage) due
    to high field effect in the drain junction
  • I6 - the channel punchthrough current due to the
    proximity of the depletion regions of the drain
    and the source.
  • I2, I5, I6 and are off-state leakage mechanisms
  • I1 and I3 occur in both ON and OFF states
  • I4 can occur in the off state, but more
    typically occurs during the transistor bias
    states in transition.

4
Subthreshold Leakage current
  • u0 is the zero bias electron mobility, n is the
    subthreshold slope coefficient.
  • To decrease Subthreshold current
  • Cox eox/Tox ? Determined by foundry
  • Vgs Vds ? Vdd ? dual power supply
  • Vth ? dual-Vth, Multi-Vth, Variable Vth
  • W or L ? gate sizing
  • Temperature ( VT KT/q)

5
Outline
  • Transistor Leakage Mechanisms
  • Leakage Reduction techniques
  • - Mutli-, dual-, Variable Vth
  • - Dual Power Supply
  • - Transistor Sizing
  • - Transistor Stacking
  • - Optimal Input Vector Selection
  • Proposed Technique
  • - Using ILP to Minimize leakage
  • - Extend ILP to Minimize leakage and Glitch
    Power together

6
Leakage Delay
  • Increasing Vth can decrease Isub exponentially
  • But, gate delay increase at the same time
  • where a models short channel effects (1.3)
  • When using Vth changing techniques, must
  • consider the tradeoff between leakage
  • reduction and performance reduction

7
MTCMOS (Multi-Threshold CMOS)
8
MTCMOS (cont.)
  • Advantage
  • - Circuit can be modified easily
  • Disadvantages
  • - Affect delay, area
  • - Can only reduce leakage power in standby mode
  • - Not suitable for sequential circuit

9
VTMOS (Variable Threshold CMOS)
  • Vth0 - zero-substrate-bias value for Vth
  • r - body effect parameter
  • 2?F - surface potential parameter
  • ? - Drain-induced barrier lowering (DIBL)
  • coefficient (0.02-0.1)
  • Using body effect, change Vth
  • In active mode, a zero body bias
  • In standby mode, a deeper reverse body bias, Vth
    increase
  • Can only reduce leakage power in standby mode

10
VTMOS (cont.)
11
Dual Threshold CMOS
  • To maintain performance, all gates on the
    critical path are assigned low Vth
  • Part of the gates on the non-critical paths are
    assigned high Vth
  • Disadvantage Circuit structure sensitive
  • Advantage Can reduce leakage power in both
    standby mode and active mode !

12
Outline
  • Transistor Leakage Mechanisms
  • Leakage Reduction techniques
  • - Mutli-, dual-, Variable Vth
  • - Dual Power Supply
  • - Transistor Sizing
  • - Transistor Stacking
  • - Optimal Input Vector Selection
  • Proposed Technique
  • - Using ILP to Minimize leakage
  • - Extend ILP to Minimize leakage and Glitch
    Power together

13
Dual Power Supply Voltages
  • Vdd ? Isub
  • ? gate delay
  • Assign Low Vdd to the gates on the non-critical
    path, to decrease leakage power
  • Assign High Vdd to the gates on the critical
    path, to maintain performance

14
Outline
  • Transistor Leakage Mechanisms
  • Leakage Reduction techniques
  • - Mutli-, dual-, Variable Vth
  • - Dual Power Supply
  • - Transistor Sizing
  • - Transistor Stacking
  • - Optimal Input Vector Selection
  • Proposed Technique
  • - Using ILP to Minimize leakage
  • - Extend ILP to Minimize leakage and Glitch
    Power together

15
Transistor Stacking
  • Serious connected off Transistors (Transistor
    Stacking) can reduce leakage current greatly
  • When M1 and M2 are turned off, Vm at the
    intermediate node is positive due to small drain
    current.
  • Vgs1 lt 0, reduce the subthreshold current
    substantially.
  • Vbs1 lt 0, increase Vth1 (larger body effect) and
    thus reducing the subthreshold leakage.
  • Vds1 decrease, increase Vth1
  • Vds2 decrease, increase Vth2

Vm
16
Outline
  • Transistor Leakage Mechanisms
  • Leakage Reduction techniques
  • - Mutli-, dual-, Variable Vth
  • - Dual Power Supply
  • - Transistor Sizing
  • - Transistor Stacking
  • - Optimal Input Vector Selection
  • Proposed Technique
  • - Using ILP to Minimize leakage
  • - Extend ILP to Minimize leakage and Glitch
    Power together

17
Leakage Dependence on the Input Vector
  • Different Input vector, different leakage
    current.
  • 00 p1 p2 on, n1 n2 off.
  • Ileak 00 In1 In2 2 Ileak
  • 01 n1 off. n2 is on and can be treated as
    shorted,
  • so leakage current of n1 is
    ignored.
  • p1 is on and p2 is off.
  • Ileak 01 Ip2 Ileak
  • 10 the same as the 01
  • Ileak 10 Ip1 Ileak
  • 11 n1 n2 on. p1 p2 off. Due to the
    stacking
  • effect,
  • Ileak 11 lt Ileak
  • So, when the input vector is 00, the NOR gate
    has the maximal leakage current. When the input
    vector is 11, the NOR gate has the minimum
    leakage current.

18
Optimal Input Vectors Selection
  • There must be optimal primary input vectors which
    lead to the minimum leakage power in the standby
    mode.
  • For smaller ciruits
  • - Exhaustive Search
  • For larger circuits
  • - Random Search
  • - Genetic algorithm ( exploit historical
    information to speculate on new search points
    with expected improved performance to find a
    near-optimal solution )

19
Outline
  • Transistor Leakage Mechanisms
  • Leakage Reduction techniques
  • - Mutli-, dual-, Variable Vth
  • - Dual Power Supply
  • - Transistor Sizing
  • - Transistor Stacking
  • - Optimal Input Vector Selection
  • Proposed Technique
  • - Using ILP to Minimize leakage
  • - Extend ILP to Minimize leakage and Glitch
    Power together

20
Dual Threshold CMOS
  • To maintain performance, all gates on the
    critical path are assigned low Vth
  • Part of the gates on the non-critical paths are
    assigned high Vth, to avoid the change from
    non-critical path to critical path.
  • Disadvantage Circuit structure sensitive
  • Advantage Can reduce leakage power in both
    standby mode and active mode !

21
Using ILP (Integer Linear Programming) to Reduce
Leakage Power
  • In dual-threshold CMOS process
  • Firstly, assign all gates low Vth
  • Use ILP model 1 to find the delay of the critical
    path (Tc)
  • Use ILP model 2 to find the optimal Vth
    assignment as well as the leakage reduction of
    all gates without increasing Tc
  • Further reduce leakage power by increasing Tc

22
ILP
  • Raja et al. 16 proposed a technique to reduce
    dynamic glitch power by a reduced constraint set
    linear program.
  • We modify their formulation into an integer
    linear program (ILP) to reduce leakage power.
  • ILP is a mixed ( integer value and continuous
    values combined together) linear programming

23
ILP -Variables
  • Each gate has two variables.
  • Ti the latest time at which the output of gate i
    can produce an event after the occurrence of an
    input event at primary inputs of the circuit.
    Continuous value
  • Xi the assignment of low or high Vth to gate i
    Xi is an integer which can only be 0 or 1.
  • 1 ? gate i is assigned low Vth
  • 0 ? gate i is assigned high Vth.

24
ILP - objective function
  • objective function
  • - minimize the sum of all gates leakage
    currents, which is given by
  • ILi is the leakage current of gate i with low
    Vth
  • IHi is the leakage current of gate i with high
    Vth
  • Each gates leakage current can be either ILi or
    IHi
  • Using SPICE simulation results, we constructed a
    leakage current look up table, which is indexed
    by the gate type and the input vector.

25
ILP - Constraints
  • Constraints for each gate
  • (1)
  • gate j s output is gate i s fan in
  • (2)
  • Max delay constraints for primary outputs (PO)
  • (3)
  • Tmax can be spec. or the delay of the critical
    path

26
ILP Constraints 1
  • assume all primary input (PI) signals on the left
    arrive at the same time.
  • For gate 2, constraints can be given by

27
ILP Constraints 1 (cont.)
  • DHi is the delay of gate i with high Vth
  • DLi is the delay of gate i with low Vth.
  • A second look-up table is constructed and
    specifies the delay for given gate type and
    fanout number.

28
ILP Constraints 3
  • Tmax can be spec. or the delay of the critical
    path (Tc).
  • To find Tc, we change constraints 2 to a
    equation, which means all gates are assigned low
    Vth.
  • The maximum Ti given by AMPL CPLEX, is equal to
    Tc.
  • If we replace Tmax with Tc, the real objection
    function becomes minimize leakage power without
    sacrificing any performance.

29
ILP Constraints 3 (cont.)
  • If we gradually increase Tmax from the smallest
    value Tc, more leakage power can be reduced,
    because more gates on the non-critical path can
    be assigned high Vth.
  • But, the reduction trend becomes slower.
  • When Tmax (130) Tc, the reduction is
    saturated, because almost all the gates are
    assigned high Vth, and there is no more
    optimization space.
  • The maximum leakage reduction can be 98.
  • Tradeoff between Leakage and Performance

30
Results-Leakage Reduction
Cir. Number of gates Tc (ns) Unoptimized Ileak (µA) Optimized Ileak (µA) (Tmax Tc) Leakage Reduction Sun OS 5.7 CPU secs. Optimized for Ileak (µA) (Tmax1.25Tc) Leakage Reduction Sun OS 5.7 CPU secs.
C432 160 0.751 2.620 1.022 61.0 0.25 0.132 95.0 0.25
C499 182 0.391 4.293 3.464 19.3 0.31 0.225 94.8 0.30
C880 328 0.672 4.406 0.524 88.1 0.54 0.153 96.5 0.53
C1355 214 0.403 4.388 3.290 25.0 0.33 0.294 93.3 0.36
C1908 319 0.573 6.023 2.023 66.4 0.57 0.204 96.6 0.56
C2670 362 1.263 5.925 0.659 90.4 0.68 0.125 97.9 0.53
C3540 1097 1.748 15.622 0.972 93.8 1.71 0.319 98.0 1.70
C5315 1165 1.589 19.332 2.505 87.1 1.82 0.395 98.0 1.83
C6288 1177 2.177 23.142 6.075 73.8 2.07 0.678 97.1 2.00
C7552 1046 1.915 22.043 0.872 96.0 1.59 0.445 98.0 1.68
31
Results-Dynamic Leakage Comparison
  • VT (thermal voltage, kT/q) and Vth both depend on
    the temperature, so, leakage current also
    strongly depends on the temperature.
  • Spice simulation shows that for a 2-input NAND
    gate
  • - with low Vth, Isub _at_ 90ºC 10 Isub _at_ 27ºC
  • - with high Vth, Isub _at_ 90ºC 20 Isub _at_ 27ºC
  • To manifest the projected contribution of leakage
    to the total power, we compare dynamic and
    leakage power _at_ 90ºC.

32
Results-Dynamic Leakage Comparison (cont.)
  • Without considering glitches, the dynamic power
    is estimated by an event driven simulator, and is
    given by
  • We apply 1000 random test vectors at PIs with the
    test period equal to (120)Tc, and calculate the
    total transition No. in the circuit.

33
Results-Dynamic Leakage Comparison (cont. 2)
Circuit Pdyn (µW) Pleak1 (µW) Pleak1/ Pdyn Pleak2 (µW) Pleak2/ Pdyn
C432 71.17 26.20 36.8 10.22 14.3
C499 149.81 42.93 28.7 34.64 23.1
C880 135.19 44.06 32.6 5.24 3.8
C1355 162.39 43.88 27.0 32.90 20.3
C1908 185.60 60.23 33.4 20.23 10.9
C2670 92.64 59.25 64.0 6.59 7.1
C3540 218.41 156.22 71.5 9.72 4.4
C5315 299.61 193.32 64.6 25.05 8.4
C6288 215.12 231.42 108.0 60.75 28.2
C7552 229.13 220.43 96.2 8.72 3.8
34
Outline
  • Transistor Leakage Mechanisms
  • Leakage Reduction techniques
  • - Mutli-, dual-, Variable Vth
  • - Dual Power Supply
  • - Transistor Sizing
  • - Transistor Stacking
  • - Optimal Input Vector Selection
  • Proposed Technique
  • - Using ILP to Minimize leakage
  • - Extend ILP to Minimize leakage and Glitch
    Power together

35
Extend ILP to Minimize leakage and Glitch Power
together
  • Fig 1. A circuit with potential glitches

Fig 3. Hazard filter effect of high Vth gates.
  • Three black gates are assigned high Vth.
  • Their delays increase accordingly.
  • Only two buffers are needed to eliminate all
    glitches due to the increased gate delay of high
    Vth gates.
  • This hazard filter effect is another advantage
    of dual-Vth reassignment.
  • Fig 2. Inserting buffers in the
  • circuit of Figure 1 to balance the path delays to
    eliminate all glitches.

36
Extend ILP to Minimize leakage and Glitch Power
together (cont.)
  • The inserted buffers for eliminating glitches
    consume additional leakage power, so, we may
    assign high Vth to them.
  • Most of the delay buffers are on non-critical
    paths and can be assigned high Vth.
  • For a larger circuit, the power saving due to
    hazard filtering would be significant while power
    increase due to delay buffers will be small

37
Future Work
  • Using ILP to minimize leakage and dynamic power
    simultaneously.
  • Consider transistor sizing to reduce dynamic
    switching power and leakage power simultaneously.

38
Thank You All !
Write a Comment
User Comments (0)
About PowerShow.com