See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/333116907

Design Principles of SRAM Memory in Nano-CMOS Technologies

Article  in  International Journal of Computer Applications · May 2019

DOI: 10.5120/ijca2019918395

CITATIONS READS

5 816

1 author:

Apollos Chinonso Ezeogu

Kings Business Institute of Information Technology; National Information Technology Development Agency; Nile University of Nigeria.
6 PUBLICATIONS   7 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Process Variation Aware Non-Volatile(Memristive) Embedded 9T SRAM Memory Design in Nano-CMOS Technologies View project

Computing and Physics View project

All content following this page was uploaded by Apollos Chinonso Ezeogu on 19 June 2019.

The user has requested enhancement of the downloaded file.

 International Journal of Computer Applications (0975 – 8887)
Volume 178 – No. 11, May 2019

Design Principles of SRAM Memory in Nano-CMOS

Technologies

Ezeogu Chinonso Apollos

Scholar, National Information Technology Development Agency (NITDA)
Nigeria

ABSTRACT believed to be the first fully functional SRAM using 45nm

Static Random Access Memory (SRAM) is a volatile memory process technology manufactured on January 26, 2006 by
that is widely used in every embedded system – Silicon on Intel Corporation [1].
Chip (SoC), Digital Signal Processing (DSP),
Microcontroller, Field Programmable Gate Array (FPGA) and
Video applications. It is also used in register, cache and
cache-less applications due to large storage *density, reduced
read-write access time, low power consumption and stability.
Thus, this paper presents the design principle of SRAM at the
45nm technology node, the peripheral building blocks and
functionalities, operations, transistor scalene challenges, and
process variation effects of SRAM designs. A clear detail
schematic diagrams using Cadence Virtuoso design tool for
IC design was used for designing the peripheral circuitry and
the SRAM cell.

General Terms
SRAM, CMOS, Design Principles, Nano Technology

Keywords
Memory cell, Embedded System, Read, Write, Process
variation, Leakage current, Power consumption.

1. INTRODUCTION
SRAM memory is still currently the main memory block of
today’s embedded systems and computing devices cache and
register designs. It is used in portable devices and in
embedded systems due to high demand for data storage, Figure 1a: Intel 45nm SRAM chip [2]
computing speed, data stability, and low power consumption
[1] and plays essential role in all Intel products in achieving
power performance goals; process defect sensitivity and
detection [2]. In high performance computing systems, the use
of SRAM for cache design helps speed up data
communication between the Central Processing Unit (CPU)
and the memory block; this ensures that frequently accessed
data are retrieved from the cache rather than the CPU this
technique is called the John Von Neumann stored program
computation concept that is widely used. However, this multi-
core system computational efficiency is getting saturated in
high data computation because of decreased efficiency, power
consumption, and stability as CMOS scaling continues to get
smaller below the 90nm technology node. Figure 1b: Intel 45nm SRAM in CMOS technology [11]
The SRAM cell is classified into different configurations
From figure 1a, the Intel’s SRAM bits are arranged in
which are named according to the number of transistors used
subarrays with rectangular matrix-like structure with rows and
in designing the memory cell. We have 4T, 5T, 6T, 7T, 8T,
columns such that in a read or write operation from the
9T, 10T and higher order SRAM configurations [1]. SRAM is
memory subarray, a specific row and column are activated
used in main memory for cache-less embedded processors and
depending on the address, and a group of bits called a word
hence must be optimized in terms of power, density, area and
are read or written [2]. The subarray is designed to be very
delay [4]. The design factors that are paramount to be
compact since it is tiled many times, to form an on-die cache
considered when designing SRAM are power consumption,
in a real product. After completion of the subarray design,
leakage current and stability under process variation. As
then large portions of the X-chip area can be tiled with these
CMOS scaling down is reaching its physical limit this poses
subarrays with minimal additional effort [2].
other challenges in leakage power, reliability, test complexity,
cost for mask and design, yield and fabrication processes [7].
Figure 1 shows 45nm 6T SRAM logic technology which was

5
 International Journal of Computer Applications (0975 – 8887)
Volume 178 – No. 11, May 2019

2. SRAM DESIGN ARCHITECTURE number of transistors include 4T, 5T, 6T, 7T, 8T, 9T, 10T,
11T, and 12T. The cell name is given base on the number of
2.1 SRAM Block Diagram transistors it contains, where “T” stands for “transistor”. The
SRAM complete block structure is shown in figure 2, with fundamental building block of a Static Random Access
other peripheral circuitry such as sense amplifier, row and Memory (SRAM) is the SRAM memory cell. The cell is
column decoders, read and write drivers, and timing control activated by raising the word line and is read or written
logic required for the complete implementations and through the bit lines. There are three different states of an
simulation of the SRAM cell for read, write and hold states. SRAM cell, namely: standby, reading and writing states to be
The SRAM architecture is arranged in cores for a larger discussed later in section 3.
system then in blocks and arrays depending on the design
specifications. The memory arrays are arranged in rows (word wordline
lines) and column (bit lines) of memory cells and have unique Vdd
location defined by the intersection of the rows and column of

Bitline

Bitline
the array. And each address or memory cell has its own pre-
charged circuit, read buffer and write driver, sense amplifier, R1 R2
and activation word lines and bit lines for active cell selection.
M3 M4

Pre-charge Circuitry
RD_Enable
M1 M2
BL BLB
WL[0]
Row Decoder

A0 SRAM Cell

A1

A2 Figure 3a: 4T SRAM Cell

wordline
WL[N-1] Vdd
Bitline

Bitline
M1 M3
Column Decoder
M5 M6
CD_Enable

M2 M4
SA_Enable Sense Amplifier Data out

Timing Scheme Write Driver

Write_Enable Data in
Figure 3b: 6T SRAM Cell

Figure 2: SRAM Block Diagram [1]. Writeline

Readline
wordline
2.2 SRAM Memory Cell
Vdd
Static random access memory (SRAM) is a type of memory
that uses bi-stable flip-flop (two cross-coupled inverters)
Bitline

made up of at least four transistors; the flip-flop may be in

Bitline

M1 M3
either of two stable states, 1 or 0, and also the access
transistors grant access to stored data for read and write. Thus, M5
M7
the term static means that it holds data as long as power is M6

applied which differentiates it from dynamic RAM which

M2 M4
must be periodically refreshed. There are different types of
SRAM cell based on their functionality, transistors number
used and memory size. Based on functionality we have
synchronous and asynchronous SRAMs. The synchronous
SRAM uses one or more clock to time the operations of the
SRAM and it is the most used design nowadays. While the
asynchronous SRAM is independent of clock frequency with Figure 3c: 7T SRAM Cell
a sequential pattern of read and write operations and has some
limitations when it comes to performance as compared to
synchronous SRAMs.
Transistor number is now used to classify the operation,
performance and improvement in SRAM’s operations; the

6
 International Journal of Computer Applications (0975 – 8887)
Volume 178 – No. 11, May 2019

Write_wordline
equalization transistor which ensure that asymmetric defect is
Vdd
eliminated for correct read operation than if it was not
included; in order words it helps to minimize the voltage
difference between the bit lines, reduce pre-charge time by

Bitline
M1 M3 making sure that the bit lines (BL and BLB) are at nearly at
equal potential. A good pre-charge circuit should not have bit
M5 line voltage difference greater than 80mV for correct read
operation of the SRAM cell [8]. Consequently, M2 and M3
are the load transistors that connect the bit lines to the Vdd for
M2 M4
the pull up. It is also possible to use NMOS transistors which
will pre-charge the bit lines to Vdd - Vth; this gives a faster
single-ended bit line sensing because the bit lines do not
Vdd
swing as much to the Vdd, but the disadvantage is that it
reduces the noise margins and require more pre-charge time.
M8

M6

M7

Read_wordline

Figure 3d: 8T SRAM Cell

wordline
Vdd
Bitline

Bitline

M1 M3

M5 M6

M2 M4 Figure 4: Pre-charged Circuit [1].

2.4 Sense Amplifier

M7 M8 The sense amplifier detects a small differential voltage
difference in the bit lines during read cycle then it amplifies it
R0 M9
to full swing close to the Vdd thereby reducing the time
required for read operation, in order words, the read operation
is very fast. Because SRAM does not require constant
refreshing as DRAM, hence sensing must be non-destructive
unlike the destructive sensing of the DRAM cell [1].
Figure 3e: 9T SRAM Cell

Read_wordline

wordline
Bitline

Vdd
Bitline

M1 M3

M5 M6 R10
M9

M2 M4

M7 M8

Figure 3f: 10T SRAM Cell

2.3 Pre-charged Circuitry Figure 5: Sense Amplifier [1].

The pre-charge circuitry is used for charging the bit lines to Thus, the sense amplifier robustness is an important
Vdd prior a read operation. It is necessary for both bit lines component of the SRAM block; the choice of sense amplifier
(BL and BLB) to be exactly at Vdd and perfectly equalized for design used affects the robustness of the bit line sensing, the
a correct read operation [8]. Thus, the circuit is made up of read speed, reliability and power consumption of the design.
three PMOS transistors, M1, M2, and M3. The M1 is the Sense amplifiers reduce delay in heavily capacitive loaded

7
 International Journal of Computer Applications (0975 – 8887)
Volume 178 – No. 11, May 2019

logic circuits. Furthermore, the sense amplifier in figure 5 is a

clocked control sensed amplifier. The clocking helps save
power because it consumes power only when the clock is
activated, however it requires good timing chain to be
activated at the right time. If it is activated too early then the
bit lines may not be pre-charged enough to operate reliably
while if it activates too late then the SRAM operation will be
slow.
From figure 5, when the sense clock (SAE) is low the
amplifier is inactive and it is said to be in pre-charging mode
and the transistors P1 and P4 are called isolation transistors,
that charge the output nodes to Vdd. Furthermore, when it is
high (SAE=1) then M4 and M8 are activated and the sense
amplifier is said to be in sensing mode. Transistors M4 and
M8 work as common source differential amplifier [9], turning
on the cross-coupled inverter pair, which puts one of the
output low and the other high through regenerative feedback.
And power dissipation can be reduced during read operations Figure 6 Write Driver [1].
by turning off the word lines and the sense amplifier once a
sufficient differential voltage has been reached and end of 2.7 Timing Block Diagram
data sensing at the output on the bit lines so as to save power. The timing block design as shown in figure 7, is the control
The design constraints of the sense amplifier can be circuit for pre-charge clocking, word line (WL), read wordline
summarized as follows: Input differential voltage should be (RWL), sense amplifier, write enable signal activation in
small; less sensitive to process variations (environmental order to make sure that the correct timing sequence is
factors such as noise, temperature and voltage supply); achieved during read and write operations. Timing hazards
Transistors sizing is critical and the amplifier should be made have to be prevented so not to have unintended read/write
symmetrical; and area should be minimum. operation. The timing block presented in this paper was
designed to help improve read and write of the SRAM by
2.5 Row and Column Decoder using wordline boosting scheme for write - that is, using a
The decoders make the number of interconnects to be slightly higher voltage than the Vdd for the wordline [1].
quantified to a factor of log2N where N is the number of
independent address locations. In order words, they help in
reducing the total number of pins such that if there are N + K
address lines then the bit storage will be 2N+K bits. Thus,
supposing we have three address bits (input pins) A0, A1, A2
as shown in our SRAM block design in figure 2; then N+K=3
therefore memory address space is 23 = 8 bits (1byte) wide.
Row decoder is used for the selection of rows (word lines).
While the column decoder is used to select the bit lines. In
addition, there is a multiplexer used to select one or more
columns for input/output of data [1].

2.6 Write Driver

The write driver is used for writing data into the SRAM cell
through the access transistors. The write driver will pull one
of the bit lines low and the other high depending on the data
Figure 7 Timer Schematic Diagra[1].
input. Supposing we have data = 0, then the bit line (BL) is
pulled down to gnd while the BLB is at Vdd. Similarly, if data 3. SRAM OPERATIONAL PRINCIPLE
=1 then one of the bit lines (BL) is pulled to V dd by transistor
M3 while the BLB is pulled to gnd by the transistor M2. Thus,
3.1 Basic Operation
To understand the basic function of SRAM, the 6T SRAM
to write data to a memory cell, the data is applied to the data
will be used for the detailed analogy. The 6T SRAM is made
input pin. Then the cell must be selected using its
up of six transistors, whereby two of the transistors are PMOS
corresponding row and column coordinates. This is achieved
and four are NMOS. The configuration is such that the PMOS
by pulling the specific address column low for that cell while
and NMOS form a cross-coupled inverter while two NMOS
enabling the write line (write enable) and also inserting the
transistors are connected one each to the bit lines (see Figure
word line; when the data is established in the cell then the
8). Thus, these NMOS bit lines connected transistors are
word line is turned off to preserve the power. In addition, the
referred to as the “access transistors” which are controlled by
write driver transistors are designed to be stronger than the
the word line. The 6T configuration expels most of the
relatively weak transistors of the cell so that data can easily
limitations of the 4T and 5T to a certain degree especially
override the previous state data of the memory. Thus, careful
with the advantage of better noise immunity.
sizing is necessary to ensure correct operation of the entire
However, external noise during read operation, power
design. consumption, and stability are still issues to be tackled. Static
power dissipation in 6T SRAM is relatively small. The cell
only draws current from the power supply during switching.
But, during the idle state the cell’s leakage current in the

8
 International Journal of Computer Applications (0975 – 8887)
Volume 178 – No. 11, May 2019

deep-sub-micron technology becomes an issue because of the as shown in figure 9b, MN3 and MN4, are switched ON
leakage current and data retention at low operating voltages. thereby connecting them to the bit lines. See figure 9(b) for
The 6T SRAM consists of two PMOS (MP1 and MP2) known simplified schematic during read 0. For instance, if data Q = 0
as the load transistors, two NMOS (MN1 and MN2) known as through the bit line (BL) then the gate of MN2 is turned off
the drivers’ transistors and also two NMOS (MN3 and MN4) and MP1 initially held high due to the pre-charging will then
known as the access transistors. There are three major go “off” - PMOS are turned ON when the gate input is low
operations of SRAM: retention/standby, read and write while NMOS when the gate input is high. Note that Q =0,
operations. These operations are explained below: then QB=1 which is then fed as feedback to MP1 and MN1
hence switching MN1 ON while MP1 OFF. Thereafter the
current (Icell) now moves from the bit line (BL) through MN3
to the storage node Q thereby charging node Q while
discharging the bit line (BL). Since MN1 is ON then the
current from the node Q is further discharged to gnd; this is
possible by making the width (cell ratio) of MN1 wider than
MN3. The bit line voltage, (VBL), having been discharged to
(Vdd - Vth) the sense amplifier detects this voltage difference
in the bit lines and is then triggered and speedily amplifies the
small differential voltage between the bit lines to full swing
close to (Vdd) by identifying the bit line with the higher
voltage raising it Vdd while the lesser voltage is discharged
slowly through gnd. Then the data is kept at a stable state by a
sense amplifier. Conversely, if the data to be stored is “1” (see
figure 9c) the potential at node Q and the bit line potential will
be equal so no discharge will take place; however, at node QB
= 0, the bit line-bar (BLB) potential is higher so the discharge
current will move through transistor MN4 to MN2, thus
discharging BLB to Vdd -Vth. The sense amplifier will pull
BLB to gnd while BL remains at Vdd.
Figure 8: 6T SRAM Architecture [1]
WL = Vdd

3.1.1 Standby (Retention) Operation Vdd

This is the state when the SRAM cell is idle (data is held in Icell
latch) and the bit line and bit line bar (data path) are kept at BL= Vdd
MP2 BLB=Vdd
gnd when the access transistors are disconnected because the
word line is not inserted. Thus, the PMOS transistors will MN3
Q=0 MN4
QB=1
continue to re-enforce each other as long as they are Vdd
connected to the power supply in order to keep the data stored MN1
in the latch as shown in Figure 9a. During this idle/retention
mode, when “1” is stored in the cell, MP1 and MN1 are ON
thus there exists a positive feedback between Q and QB nodes
making Q to be pulled to Vdd. Similarly, when “0” is stored
in the cell, MP1 and MN1 are OFF while QB is pulled to Vdd . Figure 9(b): Read Operation for Data=0 [1]

Figure 9(c): Read Operation for Data=1 [1]

Figure 9(a): Retention Mode [1] 3.1.2.1 Read Constraint

The increase in voltage at node Q should not be too large to
3.1.2 Read Operation switch ON transistor MN2. The 6T SRAM design shown in
This is the state when data is requested from the memory cell. figure 8 is symmetrical, therefore all dimension for left
Therefore, to read data, both bit line (BL) and bit line-bar symmetry will be used for the right symmetry. Therefore, the
(BLB) are initially pre-charged to a logic state 1 (Vdd), when condition below must be satisfied, Cell Ratio (CR): The ratio
the word line (WL=0) is low. After the pre-charge cycle the between sizes of driver transistor to the load transistor during
word line (WL) is enabled (WL=1) thus the access transistors

9
 International Journal of Computer Applications (0975 – 8887)
Volume 178 – No. 11, May 2019

read operation. node Q then MP2 pulls down Q and cause the SRAM to
switch values
…………….(1)
…………..(3)
Since = = 45nm, then > by a factor
of at least 1.2 in order to ensure adequate noise margin and In equation 3, the NMOS has mobility higher than the PMOS,
no-destructive read. And as CR increases, the speed of the therefore for correct write operation the NMOS can be sized
SRAM cell increases [9]. to be equal to or greater than PMOS at minimum size.
Similarly, WMN3 = WMN4 and WMN1 = WMN2 Write Operation Summary:
The differential voltage developed on the bit lines depends on 1. Drive one of the bit lines high and the other low;
the cell current, Icell, bit line capacitance, Cbit, and the length
of time, t, the word line is activated. The current should be 2. Load data into write driver input pin;
large enough in order to discharge the bit line capacitance. 3. Enable the “write line” and the “word line” simultaneously;

…………………………………… (2) 4. Data is overwritten due to the weak SRAM cell transistors
compared to the write driver.
Read Operation Summary: 3.2 Transistor Scaling and Challenges
1. Pre-charge the bit lines to Vdd while the word line and sense Moore’s law states that the shrinking in density of transistor
size will double every 1.5 years. This predictive law for
amplifiers are disabled; CMOS has always proven to be true; the technology-node
scaling is driven by the need for high integration density and
2. Lower the column decoder of the given memory cell to be performance required in cache designs and microprocessors.
read; This has led to increase in statistical variation in the process
parameters which can cause increase in the total leakage
3. Enable the word line and sense amplifier after the pre- current. Hence, the reduction in threshold voltage, channel
charge cycle; length, drain/source junction depth, gate oxide thickness and
Vdd has become a major contributor to increase in leakage
4. Sense amplifier reads data from bit lines; current. The sub threshold leakage is the drain-source current
of the transistor when the gate-source voltage is less than the
5. Read outputs using the bit lines, a drop in bitline (BL)
threshold voltage; this is large for short channel devices. The
indicates data = 0 else data =1.
gate current leakage is due to low oxide thickness and the
3.1.3 Write Operation high electric field resulting to current flow through gate of the
This is the state when data is been written/updated in the cell transistor even during the off state because the classical
(see Figure 9d). To write data into a cell, the sense amplifier infinite impedance assumption of MOS is destroyed by the
and pre-charge circuits are deactivated while write enable and energy field. Due to this increase in leakage current, the static
the word line are first activated then the input data is driven power consumption thus exceeds switching component of the
through the write driver input pin then the bit line is pulled to power consumption.
the value of the given data while the bitline bar (BLB) takes
the complementary value. For instance, if data=0 then BL =0 3.3 Effect of Process Variation in SRAM
while BLB = 1 (Vdd); conversely, if data=1 then BL =1(Vdd) Process variations are the critical design parameters – die to
while BLB = 0 (gnd). Hence, given that transistors MP1 and die and intra-die variation – from equipment processing in the
MN3 are correctly sized then cell will flip and the data is semiconductor design technology due to inability to precisely
effectively written. control the fabrication process at small feature technologies at
the nano-scale which in turn results in large variation in the
Vdd
operation and functionality of the design. This is very severe
in the case of memory components as minimum sized
Icell transistors are used in their design [11]. These variations
BL=Vdd
MP2
include the film thickness, lateral dimensions, doping
concentration and threshold voltage variation. All these
BLB=0
Q=0 MN4
contribute to the circuit optimization for performance and
MN3 QB=1 power consumption. Doping concentration affects the
Vdd
threshold voltage, the Vth increases steadily as a result of more
MN1 random dopant fluctuations in channel, source and drain due
to increase delay distribution and delay spread. Consequently,
these random and systematic fluctuations affect the stability of
the SRAM [1]. Therefore, in the 6T SRAM design, the read
Figure 9(d): Simplified Schematic During WriteOperation stability of the cell is determined by the ratio of the current
(switching data 0 1) produced by the access transistors MN3 and MN4.
Furthermore, the impact of variation increases as the supply
Thus, consider when data = 1 to be written to a cell node voltage, Vdd, scales down to Vth because the sensitivity of the
initially storing a “0”, then the transistors MP2 and MN4 will circuit delay amplifies. Temperature and voltage variation are
function as pseudo-NMOS (MN3 is ON) inverter then current environmental variations which are primarily a function of
flows through the storage node to bit line-bar (BLB) and also intra-die (within die) variations, and contribute to failure rate
through MP2 to the storage node (QB) as soon as the potential (write ability and read stability) in SRAM cells.
at this node starts decreasing. This results in a voltage drop at

10
 International Journal of Computer Applications (0975 – 8887)
Volume 178 – No. 11, May 2019

4. SRAM CHALLENGES MITIGATION Nano-CMOS Technologies”, M.Sc. Theses submitted to

SRAM is designed to reduce power loss, mitigate scaling University of Bristol, United Kingdom.
challenges, process variation and single and multiple event [2] Uddalak Bhattachara et al.,2008 “45nm SRAM
upset and other challenges. Many architectures have been Technology Development and Technology Lead
proposed to evade these problems, some of which include: Vehicle” Intel Technology Journal, Volume 12, Issue 2
(i) The use of Statistical DOE-ILP Power- [3] Milad Zamani, Sina Hassanzadeh, Khosrow Hajsadeghi
Performance-Process (P3) optimisation of Nano- and Roghayeh Saeidi, 2013“A 32kb 90nm 9T -cell Sub-
CMOS SRAM technique [10] is very efficient threshold SRAM with Improved Read and Write SNM“
because this approach can reduce 61% power 8th International Conference on Design and Technology
consumption and 13% SNM increase when 6T and of Integrated Systems in Nanoscale Era(DTlS).
8T were subjected to this method at 45nm
technology node. [4] Arvind Chakrapani 2018, “ Survey on the design
methods of low power SRAM cell” in International
(ii) Block permutation scheme [3] is another proposed Journal of Pure and Applied Mathematics
scheme to minimize process variation by permuting
cache blocks to maximize the distance between [5] Singh Jawar, Mathew Jimson, Pradhan Dhiraj K.,
blocks with consecutive addresses, thus if the area Mohanty Saraju P. 2008 “Failure analysis for ultra low
increases then power density is minimized as a power nano-CMOS SRAM under process variations”.
result of increase in the working sets. Soc Conference, IEEE international, IEEE conference
publications, P251 -254.
(iii) The N-curve based power metrics (SPNM and
WTP) [5] which takes into account both voltage and [6] Mohammad, M. O, Saint-Laurent, P. Bassett,
current for the power metric analysis to measure the andAbraham J., 2008 “Cache design for low power and
stability of an SRAM cell, thus, in addition high yield,” in Proc. 9th International Symposium on
statistical model for estimating the static power Quality Electronic Design ISQED 2008, 17–19 March
noise margin (SPNM) and the write trip power 2008, pp. 103–107.
(WTP) are given for process variation in threshold
voltage, Vth. [7] Hoang Anh Du Nguyen, Lei Xie, Mottaqiallah Taouil,
Razvan Nane, Said Hamdioui, Koen Bertels, 2015
(iv) Double ended read decoupled 9T SRAM designed “Computation-In-Memory Based Parallel Adder”
[4]. Laboratory of Computer Engineering, Faculty of EE,
Mathematics and CS Delft University of Technology,
(v) Read-Write Assist techniques, Vdd lowering [6], Mekelweg 4, 2628 CD Delft, The Netherlands.
Vss raising, boosted word-line gate voltage and
negative bit-line. [8] Luigi Dilillo, Patrick Girard, Serge Prevossoudovitch,
Arnaud Virazel, “ Resistive-Open Defect Influence in
(vi) The use of dynamic sleep design to lower SRAM SRAM Pre-Charge Circuits: Analyshis and
power supply to effectively reduce static power Characterization” Proceeding of European Test
consumption by reducing leakage [12]. Symposium(ETS’05) Copyright 2005, IEEE.
5. CONCLUSION [9] Shalinin, Anand Kumar, 2013 “Design of High Speed
In this paper, the design principles of SRAM including the and Low Power Sense Amplifier for SRAM
peripheral circuitry, operations, challenges, mitigation Applications”, International Journal of Scientific &
techniques, how to improve the stability were clearly Engineering Research Volume 4, Issue 7, ISSN 2229-
explained in a very simplified manner including the design 5518, pp 402 -406.
schematic to show the memory cell and logic gates of the
SRAM block. Meanwhile, designers and researchers are faced [10] Thakral Garima, Mohantu Saraju P., Ghai Dhru,
with issues with process variation which include: process Pradhan Dhiraj K. 2010 “P3 (Power- Performance-
technology, voltage, temperature, stability and leakage power Process) Optimisation of Nano-CMOS SRAM using
as transistors sizes are scaled down, and effects under single statistical DOE-ILP ”. Quality Electronic Design
and multiple event upset. However, SRAM is still currently (ISQED), 11th International Symposium on, p176- 183.
the most widely used memory in embedded systems; its use is [11] Mutyam M , Narayanan V. 2007 “Working with Process
now extended in designing non-volatile memory using Variation Aware Cache”. Design, Automation & Test in
memristors [1]. Consequently, researchers are currently Europe Conference & Exhibition, p1-6.
exploring its use for SRAM-based physical unclonable
function architecture, SRAM-based Computation-in-Memory [12] K.-S. Min, K. Kanda, and T. Sakurai, 2003 “Row-by-row
architecture, and field programmable gate arrays. It is worth dynamic source-line voltage control (RRDSV)
to note that SRAM will still continue to play major role in schemefor two orders of magnitude leakage current
embedded system and computing devices even in decades to reductionof sub-1-V-VDD SRAM’s.” In Proceedings
come as more applications of its use are been explored. IEEEInternational Symposium Low Power Electronics
andDesign (ISLPED), pp. 66–71.
6. REFERENCES
[1] Ezeogu, Apollos. 2013 “Process Variation Aware Non-
Volatile (Memristive) 9T SRAM Memory Design in

IJCATM : www.ijcaonline.org 11

View publication stats