Method and apparatus for adaptive clock phase control for LSI power reduction
View Patent ↗Methods and apparatus for distributing a clock signal to a digital circuit provide for: producing a clock signal; and delaying, advanced, or leaving the clock signal unchanged to produce an output clock signal as a function of a control signal, wherein an amount of delay or advancement between the clock signal and the output clock signal (phase difference) is a function of time variant changes in a magnitude of a power supply voltage to the digital circuit.
1. A clock distribution circuit, comprising:
a clock source that operates to produce a clock signal;
a phase adjustment circuit that operates to receive the clock signal and produce a delayed, advanced, or unchanged output clock signal as a function of a control signal, and which operates to vary the phase difference between the clock signal and the output clock in a way that adheres to the following equation:
φ
(
t
)
=
-
β
T
0
∫
(
v
(
t
)
-
v
0
)
ⅆ
t
where φ(t) is the phase difference in radians as a function of time variant changes in a magnitude of the power supply voltage v(t), T 0 is a cycle time of the clock signal and β is a constrained constant; and
a clock distribution tree that operates to distribute the output clock signal to a plurality of areas of a digital circuit,
wherein the phase adjustment circuit is that operates to vary an amount of delay or advancement between the clock signal and the output clock signal (phase difference) as a function of time variant changes in a magnitude of a power supply voltage to the digital circuit.
2. The clock distribution circuit of claim 1 , wherein β is constrained limit the permissible phase difference in accordance with the following expression:
ⅆ
max
≥
β
2
π
max
∫
(
v
(
t
)
-
v
0
)
ⅆ
t
β
≤
2
π
ⅆ
max
max
∫
(
v
(
t
)
-
v
0
)
ⅆ
t
where dmax is a maximum quantity of time permitted to be added to or subtracted from the clock signal.
3. The clock distribution circuit of claim 2 , wherein the phase adjustment circuit includes:
a sensing circuit that operates to produce a noise signal indicative of the time variant changes in the magnitude of the power supply voltage;
a low frequency cut-off circuit that operates to produce a filtered version of the noise signal;
an integrator circuit that operates to produce an integral of the filtered noise signal as a function of time; and
a gain circuit that operates to produce the control signal from the integral of the filtered noise signal.
4. The clock distribution circuit of claim 3 , wherein:
the power supply is implemented with a circuit having a chip, package, and/or board parasitic LCR resonance frequency as a function of stray inductances and capacitances within the power supply circuit; and
a cut-off frequency of the low frequency cut-off circuit is set below the LCR resonance frequency such that β is larger than would otherwise be possible when the cut-off frequency of the low frequency cut-off circuit is set above the LCR resonance frequency.
5. A phase adjustment circuit that operates to receive a clock signal for distribution in an digital circuit and produce a delayed, advanced, or unchanged output clock signal as a function of a control signal, wherein the phase adjustment circuit operates to vary an amount of delay or advancement between the clock signal and the output clock signal as a function of time variant changes in a magnitude of a power supply voltage to the digital circuit, wherein the phase adjustment circuit is operable to vary the phase difference between the clock signal and the output clock as expressed by the following equation:
φ
(
t
)
=
-
β
T
0
∫
(
v
(
t
)
-
v
0
)
ⅆ
t
,
wherein φ(t) is the phase difference in radians as a function of time variant changes in a magnitude of the power supply voltage v(t), T 0 is a cycle time of the clock signal and β is a constrained constant.
6. The phase adjustment circuit of claim 5 , wherein β is constrained limit the permissible phase difference in accordance with the following expression:
ⅆ
max
≥
β
2
π
max
∫
(
v
(
t
)
-
v
0
)
ⅆ
t
β
≤
2
π
ⅆ
max
max
∫
(
v
(
t
)
-
v
0
)
ⅆ
t
where dmax is a maximum quantity of time permitted to be added to or subtracted from the clock signal.
7. A method of distributing a clock signal to a digital circuit, comprising:
producing a clock signal; and
delaying, advanced, or leaving the clock signal unchanged to produce an output clock signal as a function of a control signal, wherein an amount of delay or advancement between the clock signal and the output clock signal (phase difference) is a function of time variant changes in a magnitude of a power supply voltage to the digital circuit, the phase difference between the clock signal and the output clock being expressed by the following equation:
φ
(
t
)
=
-
β
T
0
∫
(
v
(
t
)
-
v
0
)
ⅆ
t
,
wherein φ(t) is the phase difference in radians as a function of time variant changes in a magnitude of the power supply voltage v(t), T 0 is a cycle time of the clock signal and β is a constrained constant.
8. The method of claim 7 , further comprising distributing the output clock signal to a plurality of areas of a digital circuit using a clock distribution tree.
9. The method of claim 7 , wherein β is constrained limit the permissible phase difference in accordance with the following expression:
ⅆ
max
≥
β
2
π
max
∫
(
v
(
t
)
-
v
0
)
ⅆ
t
β
≤
2
π
ⅆ
max
max
∫
(
v
(
t
)
-
v
0
)
ⅆ
t
where dmax is a maximum quantity of time permitted to be added to or subtracted from the clock signal.
10. The method of claim 9 , further comprising:
producing a noise signal indicative of the time variant changes in the magnitude of the power supply voltage;
producing a filtered version of the noise signal; and
integrating the filtered noise signal as a function of time to produce a signal indicative of the control signal.
11. The method of claim 10 , wherein:
the power supply is implemented with a circuit having a chip, package, and/or board parasitic LCR resonance frequency as a function of stray inductances and capacitances within the power supply circuit; and
the method further comprises filtering the noise signal below the LCR resonance frequency such that β is larger than would otherwise be possible when filtering the noise signal above the LCR resonance frequency.