Вы находитесь на странице: 1из 6

Chetan Sharma, Int. J. Comp. Tech. Appl.

, Vol 2 (1), 88-93 ISSN : 2229-6093

LOW POWER AT DIFFERENT LEVELS OF VLSI DESIGN


AND CLOCK DISRIBUTION SCHEMES

Chetan Sharma
(M. Tech-VLSI Department, JSS Academy of Technical Education, Noida)
Chetan2042@gmail.com

Abstract- Low power chip requirement month.Althougy after few year VLSI
in the VLSI industry is main considerable industry will reach to its saturation
field due to the reduction of chip condition i.e. growth rate of density of
dimension day by day and environmental component will not increase as recent
factors. In this paper various low power years but demands of portable, long
techniques at Gate level, Architecture durable, lighter battery will never
level and different tradeoffs between decrease. Specific weight (stored energy
different clock distribution schemes like per unit weight) of batteries barely
as single driver clock scheme and double in several years.Besides
distributed buffers clock scheme are technological issues, further increase in
reviewed. Here it is also tried to showing specific weight will soon draw safety
various effects of particular clock concern because the energy density is
distribution scheme such as clock skew, approaching that of explosive chemicals.
clock jitter etc. So battery technology alone will not
solve the low power problem in near
Keywords: Algorithm level techniques, future.
Circuit Level Aspects, Local restructing, CMOS digital systems are approaching
Clock jitter, Clock skew. to gigahertz frequency range. It is the
result of submicron technology. The
INTRODUCTION: Intel co-founder total power dissipation consists of two
Moore’s low describe the growth of VLSI components: (a)The static power
Industries that is double of number of dissipation, which is due to a leakage
transistors in every eighteen current of transistors during steady state.

88
Chetan Sharma, Int. J. Comp. Tech. Appl., Vol 2 (1), 88-93 ISSN : 2229-6093

formation occurs in a metal–oxide–


semiconductor field- effect transistor
(MOSFET).
Power dissipation is given by
P=α.CV2f…………. (2)

Where: V yields to supply voltage and


Voltage swing. It is tried to reduce both
supply voltage and Voltage swing
because voltage have highest impact on
total power dissipation as show in
equation (2). α is switching activity, C is

Fig (a)-Power dissipation in 80C51 capacitance and f is operating frequency.

(b)The dynamic power dissipation,


which has two components: short circuit
and charge/discharge of capacitance
power dissipation. The short circuit
power dissipation is a function of the
slew rate of the input voltage; the sharper
Fig (b)-Switching energy
the clock edge, the lower the short circuit
power dissipation.
However, one major drawback
Short circuit power is result when both p-
associated with clock networks is their
transistor and n-transistor is on for short
power dissipation. Studies have shown
duration of time. Mathematicall,
that the clock network can dissipate 20-
Vdd < |Vtp | + Vtn …………… (1)
50% of the total power on a chip. In the
context of the growing importance of
Where: Vtp and Vtn are threshold
low power designs for portable
voltages for PMOS and NMOS
electronics, it is necessary to develop
transistors, respectively. Threshold
strategies to significantly reduce the
voltage is a voltage at which channel

89
Chetan Sharma, Int. J. Comp. Tech. Appl., Vol 2 (1), 88-93 ISSN : 2229-6093

power dissipation of the clock network, minimum number of switching


since this will lead to a major reduction in requirements. That algorithm is more
the overall power dissipation of the chip. useful which have minimum number of
The power is reduced at different levels operation because it will require less
of VLSI design. Levels of optimization hardware. By increasing concurrency we
are shown in fig(c). can increase efficiency of that device.

2. Architectural Level Techniques:


The basic building block at this level are
registers, busses multipliers, memories,
state machine etc.Each block perform
high level function. At architectural level
it is important to power analysis because
now a days chips become complex & it
is not easy to analyses each and every
gate. The architectural level is the design
entry point for the large majority of
Fig(c)-Power optimization at different digital designs and design decisions at
level of design. this level can have dramatic impact on
Here we see that as the level of design the power budget design.
goes lower level to upper level as layout Power is as the function of their
level to system level number of counted operating frequency and number of bits
error increase means power saving of components. For example power in
possibilities are more at higher levels. the adder is given by
1. Algorithm level reduction: P= (n.k1+k2).f ……………. (4)
Power consumption at algorithm level (a) Power and performance
related to properties of that particular management:
algorithm techniques. So it should Firstly power management is done by
carefully selected for lowering the power different saving modes such as DOZE,
consumption. For lowering the power, NAP and SLEEP modes by deactivating
algorithm should be such that it should different levels, function levels. At Doze

90
Chetan Sharma, Int. J. Comp. Tech. Appl., Vol 2 (1), 88-93 ISSN : 2229-6093

mode cache memory is active except it all 3.3. Gate level techniques:
are off. Thus coherency is made by cache (a) Local restructing: In this technique we
memory. In Nap mode processor walks modified the group of gates based on
up after some interval of time or by different operation such as Combine Gates,
external interrupt. In sleep mode all thing Decomposition of gate, Delete wire, Add
is done by reset. wire, Duplicate a gate etc.
Secondly performance management is ( (b) Signal gating: By using the controller we
done by adaptive technique by sensing can avoid extra switching activity which
the load of input. directly impact on power dissipation.
(c) Parallel and pipelined architecture is (c)(c) Logic Encoding: In the place of binary
another aspect which is also taken into code if gray code is use then switching
consideration. In parallel architecture activity is reduced by one on changing one
frequency is scaled down by factor n, to another using the property of gray code
number of blocks. In parallel architecture which changes only one bit transition on
area required is more but operation is changing in sequence.
become faster. In pipelining mode (d) Bus Inverter Encoding: By comprising the
frequency remain same but voltage is previous and present input it is decided to
scaled down. minimum number of toggling in inverting
code and available code. If there is minimum
number bit toggling in without inverting then
it is directly uses otherwise inverting code is
used. The two inverting machine uses,one at
sending port and other one at receiving port.
According to output of parity, logic inverting
of code is done. At the sending side if
inverting of code is used then at the same way
inverting also will be use at receiving side for
converting code into its original form that’s
Fig (d)-Comparison between parallel and why same parity signal is gone to sending and
pipelined architectures receiving side.

91
Chetan Sharma, Int. J. Comp. Tech. Appl., Vol 2 (1), 88-93 ISSN : 2229-6093

delay from clock source to clock


destination in different clocks. (b)Clock
jitter: It is defined as temporal variation of
clock with respect to reference edge. It is
of two type long jitter and cycle to cycle
jitter.
Single buffer scheme: If the
interconnect resistance of the buffer at the
clock source is small as compared to the
buffer output resistance, it is called as
single driver clock scheme. The single

Fig (e)-Bus Inverter Encoding driver scheme has the advantage of


avoiding the adjustment of intermediate

4. Clock schemes: buffer delay as in distributed buffer

For providing synchronization of the schemes.

digital system one or more reference Distributed buffer scheme: This is the

clocks are used. Fully synchronization is most common and general approach to

done by using common clock. By a single equi-potential clock distribution scheme.

clock all parts of digital system is clocked It leads to an asymmetric structure. In

for different operations. Clock tree used single driver scheme, short circuit power

for globally distribution the clock signal dissipation is more than distributed buffer

to all modules. There are two types of scheme due to the reason of small buffer

clock schemes: Single driver clock used in distributed buffer scheme.

scheme and Distributed buffers clock


scheme. Dynamic power dissipation by
switching of clock is given by:
Pclk = V2dd.f.(CL +CD)……………(1)
Where CL is Total load capacitance on
clock.
There are two problems in clock generation Fig (f) - Distributed buffer scheme
(a) clock skew: This is the variation in

92
Chetan Sharma, Int. J. Comp. Tech. Appl., Vol 2 (1), 88-93 ISSN : 2229-6093

CONCLUSION: [5] Zhang M., Gu J. and Chang C. H., “A


Here we reviewed the different low power novel hybrid pass logic with static
techniques at each level of VLSI Design. CMOS output drive full adder cell,” in
Different techniques have different Proc. IEEE Int. Symp. Circuits Syst.,
tradeoffs. We use particular techniques May 2003, pp. 317–320.
according to specification. Here clock skew [6] Chang C.-H., Gu J. and Zhang M., “A
and clock jitter problem are also reviewed review of 0.18-μm full adder
and these are eliminated by using performances for tree structured
appropriate clock distribution schemes. arithmetic circuits,” IEEE Trans. Very
. Large Scale Integr. (VLSI) Syst., Vol.
REFERENCES:
13, No. 6, pp. 686–695, Jun. 2005.
[1] J. Wang, S. Fang, W. Feng, “New
[7] Goel S., Kumar A. and Bayoumi M.
Efficient Designs for XORand XNOR
A., “Design of Robust, Energy-
Functions on the Transistor Level”,
Efficient Full Adders for Deep-
IEEEJournal of Solid-State Circuits, 29(7),
Submicrometer Design Using Hybrid-
1994, 780–786.
CMOS Logic Style," IEEE Trans. Very
[2] R. Zimmermann, W. Fichtner, “Low-
Large Scale Integr. (VLSI) Syst., Vol.
power Logic Styles:CMOS Versus Pass-
14, No. 12, pp. 1309–1321, Feb. 2006.
transistor Logic”, IEEE Journal of
[8] Giacomotto C. and Oklobdžija V.G.,
Solid-State Circuits, 32 (7), 1997, 1079–
"LogicStyle Comparison for Ultra Low
1089.
PowerApplications,"Techcon
[3] Shams A. M., Darwish T. K. and
(SemiconductorResearch Corp.), Oct.
Bayoumi M. A., “Performance analysis of
2005.
low-power 1-bit CMOS full adder cells,”
[9] "Predictive Technology Model”,
IEEE Trans. Very Large Scale Integr.
Websitecurrently available online at
(VLSI) Syst., Vol. 10, No. 1, pp. 20–29,
www.eas.asu.edu/~ptm.
Feb. 2002.
[4] Radhakrishnan D., “Low-voltage low-
power CMOS full adder,” IEEE Proc.
Circuits Devices Syst., Vol. 148, No. 1, pp.
19–24, Feb. 2001.

93

Вам также может понравиться