Beruflich Dokumente
Kultur Dokumente
GROUP NUMBER: 22
BRUNDA V (1PE15EC039)
JIGYASA N. (1PE15EC060)
MAHESHWARI
PAVITRA M (1PE15EC097)
NIKHIL H V (1PE15EC418)
VISVESVARAYA TECHNOLOGICAL UNIVERSITY
Belgavi-590014
Project Report
On
“DESIGN OF MEMORIES USING ADIABATIC LOGIC”
Submitted in partial fulfillment of the requirements for the VII Semester
Bachelor of Engineering
IN
ELECTRONICS AND COMMUNICATION ENGINEERING
CERTIFICATE
Signatures:
Examiners:
1.
2.
ACKNOWLEDGEMENTS
The success of any venture depends substantially on the encouragement and the directions of
many others. We take this opportunity to express our gratitude to the people who have been
influential in the successful completion of this project.
We would like to express deepest appreciation towards Dr. J Surya Prasad, Principal,
PESIT-BSC for providing us with excellent facilities.
We would like to thank our Head of Department, ECE, Dr. Subhash Kulkarni for all the help
and kindness.
We would also like to show our greatest appreciation to Mr. Vinay Reddy N for his expert
guidance and continuous encouragement throughout to see that this project rights its target
since its commencement to its completion.
Finally, we like to thank Mrs. B Hema and Dr. Shashidhar Tantry, Project Coordinators and
the entire faculty whose invaluable guidance supported us in completing this project.
Brunda V (1PE15EC039)
Jigyasa N. Maheshwari (1PE15EC060)
Pavitra M (1PE15EC097)
Nikhil H V (1PE15EC418)
ABSTRACT
Today in all high-performance and computational systems, energy efficiency has become a
major concern in design. Excessive power dissipation requires expensive machines like heat
sinks, heat pipes, fans etc. Battery life is becoming a product differentiator in many portable
devices.
Several effective power management design techniques have been developed over the past
few years, including lowering the supply voltage. As process scaling continues below 90nm,
however, it becomes more difficult to scale the supply voltage for several reasons. As a
result, there is a demand on novel circuits whose power saving mechanism is not heavily
dependent supply voltage scaling.
Adiabatic logic is a new approach to the VLSI circuit design to achieve low power. Adiabatic
logic circuits are low power circuits. They restrict the current to flow across the device with
low voltage drop. The success of adiabatic logic depends on efficient implementation of
memories as well as any random logic.
CONTENTS
1 Introduction
1.1 Adiabatic logic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Comparison between various adiabatic logic families . . . . . . . . . . 1
2 Literature survey
2.1 comparison between various adiabatic logic families . . . . . . . . . . . . 3
3 memories
3.1 memory cell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3.2 Types of memory
3.2.1 volatile memory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.2.2 non-volatile memory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
4 SRAM
4.1 6T SRAM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
4.2 Adiabatic SRAM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
5 CAM
5.1 Overview of CAM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
5.2 Types of architecture of CAM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
7 Sense amplifier
7.1 overview of sense amplifier . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
7.2 Differential voltage mode sense amplifier . . . . . . . . . . . . . . . . . . . . . 18
7.2.1 Working . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .19
9 Project planning 21
10 Simulation results
10.1 Conventional AND/NAND gate . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
10.2 Adiabatic AND/NAND gate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
10.3 Conventional SRAM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
10.4 Adiabatic SRAM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
10.5 Sense amplifier . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
10.6 conventional SRAM with sense amplifier . . . . . . . . . . . . . . . . . . . 27
10.7 Adiabatic SRAM with sense amplifier . . . . . . . . . . . . . . . . . . . . . . .28
10.8 Power analysis results . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . 29
12 References 31
1
1
1
1
CHAPTER 1
INTRODUCTION
Adiabatic logic is used to describe logic families that could theoretically operate
without losses. Adiabatic logic is a new approach to the VLSI circuit design to achieve low
power.
The dynamic energy dissipation in a circuit is due to charging and discharging for its
node capacitance. The charging and discharging paths are different, causing energy to be
dissipated in the form of heat. An adiabatic switching technique reduces the dynamic energy
dissipation by recycling the charge stored in the node capacitance in circuits. This recycling
is achieved by using an AC power source instead of the traditional DC power source. The
energy dissipation is reduced by maintaining a very low potential difference across the two
terminals of turn-on MOS transistors.
1
v. Clocked Adiabatic Logic (CAL).
vi. True Single-Phase Adiabatic Logic (TSEL).
vii. Source-coupled Adiabatic Logic (SCAL).
2
Chapter 2 : Literature Survey
Comparison between Various Adiabatic logics
Name Application Advantage Disadvantage
(Operating
Frequency )
3
CHAPTER 3
Memories
4
DRAM, but faster and does not require memory refresh. It is used for smaller cache
memories in computers.
c) Content-addressable memory: This is a specialized type in which, instead of
accessing data using an address, a data word is applied and the memory returns the
location if the word is stored in the memory. It is mostly incorporated in other chips
such as microprocessors where it is used for cache memory.
b) NVRAM (Flash memory): In this type the writing process is intermediate in speed between
EEPROMS and RAM memory; it can be written to, but not fast enough to serve as main
memory. It is often used as a semiconductor version of a hard disk, to store files. It is used in
portable devices such as PDAs, USB flash drives, and removable memory cards used
in digital cameras and cellphones.
5
CHAPTER 4
SRAM
4.1 6T SRAM
Figure 1 : 6T SRAM
The most commonly used SRAM cell is the full CMOS 6-transistor memory cell. The SRAM
cell consists of two inverters and two access MOSFETs which are connected to a pair of bit
lines. The gates of two access MOSFETs are also connected to a word line (WL). To form a
cross coupled inverter, the output of one inverter is connected to the input of the other and its
output is connected to the input of the previous inverter.
The access transistors enable access to the cell during read and write operations.
OPERATION OF SRAM
Read operation: In SRAM, for any operation to be performed, the word line should be high.
To perform read operation, initially memory should have some value. Let Q = 1 and ~Q = 0.
Raise the word line to high, to perform the read operation, these bit lines are initially pre-
charged to Vdd. As Q and BL are high, there will be no discharge in the circuit. But there will
be potential difference between ~Q and ~BL and this leads to discharge and current flows.
6
Write operation: If the memory bits consist of Q=0 and ~Q=1.
Initially word line is high and hence write operation can be performed. In write operation BL
and ~BL are input lines. As there is control on the input bit lines, initially making ~BL as low
would create the potential difference between the ~Q and ~BL.
If BL is made high the output Q obtained is high. Hence write operation is accomplished
successfully into the memory.
7
The implemented Adiabatic SRAM cell has two access transistors and the latch similar to the
conventional 6T SRAM cell.
The latch part has been implemented using the SCRL technique. MP1 and MN1 forms the
conventional NOT gate MP2 and MN2 forms the NOT gate.
T1 and T2 are the transmission gates.
Clock 1 and Clock 3 has half rail swing that is either from Vdd/2 to 0 or from Vdd/2 to Vdd.
Clock2 and Clock4 swings from 0 to Vdd or from Vdd to 0. Let us assume that all the internal
nodes are at Vdd/2.
Consider the clock diagram as shown above for the implemented SRAM cell.
Let us assume that the initial value stored in the SRAM cell is ‘0’.
Now WL = 1, and the logic 1 will be written through the bit-line.
At t1, the potential of node D slowly increased from 0 to Vdd which makes the transistor
MN1 to be in ON state. The node n1 will be at Vdd/2. The transistor MN1 is turned ON
without any non-adiabatic loss since there is no potential difference between source and drain
of MN1.
The node n1 will be Vdd/2. ~D slowly decreases from Vdd to 0. The node n2 is maintained at
Vdd/2.
At t2, D is at Vdd and clock1 is slowly increasing from Vdd/2 to Vdd. Since the transistor MN1
is in ON state, the node n1 will follows the ~clock1.
At t3, Clock2 slowly increases from 0 to Vdd by making the transmission gate T1 be ON and
connecting the nodes n1 and ~D. Now the transistor MP2 will be switched ON adiabatically
and the node n2 will follow the Clock3.
Clock3 slowly increases from Vdd/2 to Vdd.
At t4, Clock4 rises from 0 to Vdd there by turning ON transmission gates and
connecting node n2 and D.
At t5, Clock2 is switched off adiabatically since the potentials of node n2 and ~D are same.
8
At t6, Clock1 is reduced slowly from Vdd to Vdd/2 which makes node n1 to rise from 0 to
Vdd/2.
At t7, the transmission gate T2 is switched OFF adiabatically by reducing the Clock4 from
Vdd to zero.
At t8, Clock3 is reduced from Vdd to Vdd/2 there by making nodes n2 to be Vdd/2. The values
of D and ~D stored in the cell will be used for the next stage of the cycle.
9
CHAPTER 5
CAM(Content-Addressable Memory)
CMOS devices have been scaled down in order to achieve higher speed. Leakage current
reduction is the major concern in CMOS technology. Power consumption and the speed are
the major factors of concern for designing a chip along with the leakage power. The
consumption of power and speed of SRAMs are some important issues among a number of
factors that provides a solution with multiple designs that minimize the consumption of
power. The main CAM-design challenge is to reduce power consumption associated with the
large amount of parallel active circuitry, without sacrificing speed or memory density. At the
circuit level, the low power match line sensing techniques and search line driving approaches
are major concerns.
NAND CAM: The CAM cell consists of the memory cell of either 6T SRAM or the single bit
line SRAM. The comparison circuit used is the 3 transistor circuit to compare the data stored
in the memory and the search data. It consumes least power but is slow.
NOR CAM: The CAM cell consists of the memory cell of either 6T SRAM or the single bit
line SRAM. The comparison circuit used is the 4 transistor circuit to compare the data stored
in the memory and the search data. It is fast but consumes more power.
10
CHAPTER 6
Power consumption in CMOS and Adiabatic circuits
Power dissipation in conventional CMOS circuits primarily occurs during device switching.
As shown in Fig. 4, both PMOS and NMOS transistors can be modeled by including an ideal
switch in series with a resistor in order to represent the effective channel resistance of the
switch and the interconnect resistance.
The pull-up and pull-down networks are connected to the node capacitance C L, which is also
referred to as the load capacitance.
When the logic level in the system is “1” there is a sudden flow of current through R.Q. =
CLVdd is the charge supplied by the positive power supply rail for charging C L to Vdd. Hence,
the energy drawn from the power supply is Q .Vdd = CLVdd2. If it is assumed that the
energy drawn from the power supply is equal to that supplied to CL, the energy stored in CL
becomes one-half the supplied energy, i.e.
Estored = 0.5 CLVdd2
The remaining energy is dissipated in R. The same amount of energy is dissipated during
discharging in the NMOS pull-down network when the logic level in the system is “0.”
Therefore, the total amount of energy dissipated as heat during charging and discharging is
11
Etotal = Echarge+ Edischarge
= 0.5 CLVdd2+ 0.5 CLVdd2
= CLVdd2
From the above equation, it is apparent that the energy consumption in a conventional CMOS
circuit can be reduced by reducing Vdd. By decreasing the switching activity in the circuit, the
power consumption (P = dE/dt) can also be proportionally suppressed.
12
Consider the capacitor C. This capacitor has to be charged. The simplest circuit by
which it can be done is through a power supply Vdd, a switch and a resistor as shown in figure
3. As switch closes at t = 0, as the initial charge on the capacitor is zero, the complete Vdd
drops across resistor R. The capacitor starts to charge slowly when t >0 and the voltage
across R starts reducing.
Hence the current that was maximum initially i.e I=Vdd/R starts to reduce as shown in the
figure 3. This current will keep on decreasing with time and asymptotically it will approach
zero and the corresponding equation of current is
Vdd −t
I= exp (RC)
R
On applying KVL,
Vdd = IR + Q/C
Where,
Q = C . Vc
−t
VC = Vdd (1- exp (RC))
Consider the circuit as shown in figure 4.A time varying current I(t) is applied across
RC circuit through a switch. Hence the voltage across C is also time-dependent given by the
equation (assuming Vc=0 at t=0)
1
Vc = 𝑐 . 𝐼(𝑡). 𝑡
13
The energy dissipation in R from 0 to t=T is given by:
𝑅𝐶
Ediss = RT∫0 I2 dt = RI2(T)T = . 𝐶 Vc2(T)
𝑇
The following observations can be done from the energy dissipation equation:
For T>RC, the dissipated energy is smaller than the conventional case.
The dissipation can be made arbitrarily small by further extending the charging time
T.
The dissipated energy is proportional to R. A smaller R results in a lower dissipation
unlike conventional case.
14
Adiabatic NAND/AND gate
Working
AND gate: The output f goes high only when both x and y are high. The path from Va to f
closes and f follows Va.
NAND gate: The output f goes high even if one of the inputs x or y go high and hence the
path from Va to ~f closes and ~f follows Va.
15
Case (2): In figure b, the capacitor is charged initially to Vdd/2 by closing the switch S1 .The
energy dissipated is
1 1
Ediss=2.CL.(Vdd/2)2=8.CL.Vdd2
and later the switch S1 is opened and S2 is closed to charge the capacitor from Vdd/2 to Vdd.
The energy dissipated is
1 1
Ediss = .CL.(Vdd/2)2 = .CL.Vdd2
2 8
16
CHAPTER 7
Sense Amplifier
Differential sense amplifiers, also known as voltage mode sense amplifiers, are
connected directly to the bitlines. At the beginning of a read cycle, the bitlines are pre-
charged. As one bitline begins to discharge, the difference between the bitline voltages
determines the output logic value that is seen at the output. The performance of voltage mode
sense amplifiers depends on the bitline capacitances.
17
2) Current Mode sense amplifier
Non-differential amplifiers are also known as current mode sense amplifiers. Current
mode sense amplifiers may also be used independently .The amplifier inputs are connected to
the bitlines, and after precharge, the discharging bitline causes a decrease in current flow
which is used to determine the output voltage level.
18
7.2.1 Working
A read cycle begins when the signal SE is brought high. This causes a constant
biasing current, ISS, to flow through the biasing current source (N3). If both input voltages
are exactly equal, this will cause a current equal to ISS/2 to flow through both halves of the
amplifier. To ensure that the current is divided equally, the PMOS devices P1 and P2 must be
sized identically. The NMOS devices N1 and N2 must also be sized identically.
When the amplifier is first enabled, the input voltages are approximately equal
because the bitlines are pre-charged. After the SRAM cells are enabled, one bitline will begin
to discharge. As the bitline discharges, the voltage supplied to the gate of the transistor on
that side of the amplifier will decrease, causing the drain current to decrease as the transistor
begins to turn off. Because the current source is supplying a total current to the amplifier
equal to ISS, the decrease in drain current on one side will cause the current in the opposite
side of the amplifier to increase. As the transistor continues to turn off, the output voltage will
rise to equal the supply voltage, VDD. A relatively small difference between the bitlines will
produce a large difference in output voltages
19
CHAPTER 8
Software requirements specification
Cadence software is used for the design projects in the graduate level courses Wireless IC
Design. It supports analog and digital designs at the device, cell, and block levels.
Standard device models are used in conjunction with Spectre and SpectreRF simulation in
Cadence to design circuits for Ultra Wide-Band applications.
Baseband and time domain simulations of analog-front end and back-end circuits are
performed.
20
Chapter 9 Project Planning
Aug Sep Oct Nov Dec Jan Feb Mar April
Sl.No Task (2018) (2018) (2018) (2018) (2018) (2019) (2019) (2019) (2019)
1 Literature Survey
Testing Adiabatic
2 circuits
Study of SRAM
and CAM Memory
3 cells
Simulation of
conventional and
4 Adiabatic SRAM
Study of Sense
amplifiers and
Simulation of
Differential
Voltage sense
5 amplifier
Optimization of
6 Adiabatic SRAM
Simulation of
Adiabatic
NAND/NOR CAM
7 Memory cell
Simulation of
Conventional
NAND/NOR CAM
8 Memory cell
Comparison of
Conventional and
9 adiabatic results
Documentation of
10 the project
21
CHAPTER 10
SIMULATION RESULTS
Using Cadence Virtuoso 180nm
22
10.2 Adiabatic AND/NAND gate
23
10.3 Conventional SRAM
24
10.4 Adiabatic SRAM
25
Transient Response for Adiabatic SRAM
26
10.6 Conventional SRAM with Sense Amplifier
27
10.7 Adiabatic SRAM with sense amplifier
28
10.8 Power analysis results
1) AND/NAND gate
Power dissipated
Conventional 28.6uW
Adiabatic 574.3nW
Efficiency 97.99%
Conventional 290.7uW
Adiabatic 171.4uW
Efficiency 41.03%
Conventional 3.628mW
Adiabatic 247.3uW
Efficiency 93.18%
29
Chapter 11
Conclusion and future scope
11.1 Conclusion
As the density and operating speed of CMOS chips increase, power dissipation has become a
critical concern in the design of VLSI circuits, especially in mobile and portable ASIC
systems.
In this work, we have reported the use of Adiabatic logic for low power VLSI circuit design.
Various families of the adiabatic logic are studied, their advantages , disadvantages and the
potential areas of application are discussed. The mathematical proof explaining the power
consumed in conventional CMOS circuits and adiabatic circuits is given. It is seen that
theoretically, adiabatic circuits should minimize total energy dissipation to realize power-
efficient design.
The adiabatic SRAM is constructed and its working is verified using Cadence Virtuoso Tool.
A conventional 6T SRAM is also constructed for the comparison of power results with the
adiabatic SRAM.
Power results of various circuits on comparison showed that power dissipated in adiabatic
circuits is the least.
30
Chapter 12:References
1. “Charge-Recovery Computing on Silicon”, Suhwan Kim, Member, IEEE, Conrad H.
Ziesler, Member, IEEE, and Marios C. Papaefthymiou, Senior Member, IEEE,IEEE
TRANSACTIONS ON COMPUTERS, VOL. 54, NO. 6, JUNE 2005
3. “A Novel Low Power 16X16 Content Addressable Memory using PAL “,G.Josemin
Bala,J.Raja Paul Perinbam , Proceedings of the 18th International Conference on
VLSI Design held jointly with 4th International Conference on Embedded Systems
Design (VLSID’05) 1063-9667/05 $20.00 © 2005 IEEE
5. “A Novel Adiabatic SRAM Cell Implementation using Split Level Charge Recovery
Logic” S Dinesh Kumar M.Des Scholar, Dept. of EDM, IIITD&M Kancheepuram.
31