Sram Report
Sram Report
Project Report
On
“ DESIGN AND IMPLEMENTATION OF HIGH SPEED
SRAM CELL ”
Submitted in partial fulfilment of the requirements for the reward of the degree of
Bachelor of Engineering
in
Electronics & Communication
Submitted by
CERTIFICATE
External Viva
Name of the examiner Signature with date
1.
2.
ACKNOWLEDGEMENT
We take this opportunity to express our sincere gratitude and respect to the Bangalore
Institute of Technology, Bangalore for providing us an opportunity to carry out final
project.
We would like to thank Dr. MUKTHI S.L, Prof. GAHAN A V, Assistant Professor and
Project Coordinators, Department of Electronics & Communication Engineering, BIT,
Bangalore.
We express our sincere regards and thanks to Dr. HEMANTH KUMAR A.R, Professor
and HOD, Electronics & communication Engineering, BIT for his valuable suggestions.
We also extend our thanks to the entire faculty of the Department of ECE, BIT, Bangalore,
who have encouraged us throughout the course of bachelor degree.
Static Random Access Memory (SRAM) has been an important memory device in VLSI
design. This proposed work presents design and implementation of 1kb SRAM cell using
single bit six transistor SRAM memory using CMOS technologies by EDA tool. The 6T
SRAM cell is a popular type of SRAM cell, which is widely used in various electronic
devices. In this proposed work SRAM cell is operated with voltage of 1V. Proposed high
speed 6T SRAM cell decreases read and write delay by 29.4ps & 15ps respectively and
improved overall read and write stability by 70 mV & 595 mV respectively compared to
previously published ([7] [11]). In the proposed work static power is 36.6uW which is
very less and dynamic power is 82.2uW.
LIST OF CONTENTS
I
LIST OF FIGURES
II
3.9.2 Simulation Result of 1-bit Write Driver 28
3.9.3 Schematic of 8-bit Write Driver 28
3.9.4 Symbol for 8-bit Write Driver 29
3.10 Design of 32*32 Cell Array 30
3.10.1 Enlarged View of Cells in 32*32 Cell Array 31
3.11 Schematic of 1kb SRAM Memory 32
3.11.1 Simulation Result of 1kb SRAM memory 33
4.1 Simulation of Single bit SRAM Cell 34
4.2 Butterfly Curve Analysis for Read Operation 36
4.2.1 Butterfly Curve Analysis for Write Operation 36
4.3 Power graph for SRAM memory 37
LIST OF TABLES
III
DESIGN AND IMPLEMENTATION OF HIGH SPEED SRAM CELL 2023-24
CHAPTER 1
INTRODUCTION
1.1 OVERVIEW
Semiconductor memories have a significant role has most fundamental units in VLSI layout. It is
the most important unit in SOC and contain memories inclusive of flash memory, read only memory
(ROM), static random access memory (SRAM) and dynamic random access memory (DRAM).
SRAM which is kind of memory used in computer that keeps its information stored as long as
power is supplied. It is described as “static” because unlike dynamic RAM which require frequent
refreshing to maintain its information. SRAM is commonly utilized as a memory for input/output
operations in computer systems and as high-speed cache memory in CPU. Compared to DRAM, it
is more costly and faster and also more dependable, uses less power and has a higher access time.
This makes SRAM cells a fundamental building block of modern computing systems and a key
technology for achieving high-speed data processing.
SRAM is made of flip flops which are electrical circuits storing a bit of information. Four to six
transistors make up each flip-flop, allowing them to maintain their value until power is turned off.
SRAM is used in variety of computer systems and electronics including routers, phone switches
and other networking equipment, as a result of its many benefits. SRAM is typically found in cache
memory in microprocessors and other types of memory in devices. SRAM can be designed using
BJT’s however, power loss in switching is larger. SRAM is now created utilizing MOSFET’s which
reduces power loss in switching. In this nanoscale era, the static or leaky power usage of CMOS
devices predominate over the dynamic power dissipation. Whenever power is consumed in device
it generates heat that heat has to get out of the device and that can create problems for small form
factor device.
Transistor switching characteristics are to blame for dynamic power loss. For the extension of
battery life manufacturers should regulate both static and dynamic power. SRAM is used in the
high-speed, low-capacity levels of the memory hierarchy due to its fast access times. As we move
down the hierarchy, capacity increases significantly (DRAM), but speed decreases. Other memory
technologies like DRAM offer better cost-effectiveness for larger capacity needs, while SRAM
excels in speed-critical applications. SRAM is used as Cache memory which is very fast
and used to speed up the task of processor and memory interface. In this proposed work 1Kb SRAM
is designed using 90nm CMOS technology. All peripherals like pre-charge, Row Decoder, Word
line driver, Sense amplifier, Column Decoder/MUX and write driver are designed. In this proposed
work we have concentrated on the design of high speed 6T SRAM which can be used as high speed.
SRAM is used as Cache memory which is very fast and used to speed up the task of processor and
memory interface. With the recent improvements in VLSI technology, processor speeds have
increased intensely. To take the advantage of high speed processors, i.e. high clock rates, need to
provide instructions and data to the processor with little or no delay. Therefore, the very high
frequency is required for instructions and data.
With improvements in VLSI technology, the speed of the logic gates has increased significantly,
but memory speed is not improved equivalently because memory densities have increased
simultaneously. Therefore, for high speed computers, SRAM memories are very important to
improve speed and can be used with processors to do so. In this work we have concentrated on the
design of high speed SRAM which can be used as high speed memory for high speed computers.
Thus, for high speed PCs SRAM memories are critical to enhance operating speed and DRAM are
utilized as a part of Main Memory where Density has more significance than Speed.
CHAPTER-2
LITERATURE SURVEY
In order to get a foothold and basic understanding of the idea of our proposed project, we need to
review and analyze previously published technical papers in the design and implantation of SRAM
for high speed. The list below presents the details about the major such papers.
Six transistor SRAM memory design of 16-bit storage, in different nodes is presented in [1]. The
performance evaluation of all 3 designs was carried out. The performance criteria were total power
dissipation & average delay. It was observed that reduction in technology results in reduced power
dissipation & average delay. The comparison of the results is done at 180nm, 90nm, 45nm technology
nodes. It shows that power dissipation and average delay is improved as the channel length is reduced.
Inference
There is a lack of specific details regarding the methodology employed in the design and
implementation process.
V. Panduranga Vemula, S. Priyanka, A. Sairam [2] proposed many processors and SoC devices now
employ SRAM and CMOS technology, which necessitates new SRAM design innovations. SRAM
bit cells are formed of lowest geometry devices in order to achieve high density and stay up with
CMOS technology scaling as a consequence. To deal with the nano-regime issues, numerous SRAM
bit cell topologies and array layouts have recently been suggested. Different SRAM bit cell topologies
and their benefits and drawbacks are also discussed in which we have considered the design
architecture which shows speed improvements along with scaling of technology and delay time also
decrease.
Inference
Fail to overcome primary issues like poor stability, process variation tolerance in SRAM designs.
The read, write, operations were performed for the 6T SRAM. Also, DC response for temperature
and power is simulated in [3]. Another major factor that is the Static Noise Margin has been done and
the noise responses obtained for both read and write operations. Here we have designed the Read and
Write Operations for 6T SRAM Cells. The analysis of 6T Static Random-Access Memory is done in
[3]. In the future, we can upgrade the Static Random-Access Memory cell and make it dependable
and robust to the noise present in the environment.
The SRAM cell design and its operational aspects, there's limited discussion on design trade-offs or
optimizations made to achieve specific performance goals. So, to overcome this we provide
Elaborating on design decisions and their impact on performance would enrich the circuit analysis.
Designing of 32*32 Memory Array SRAM is done in [4]. 1Kb SRAM array in bit orientation with a
6T SRAM cell in CMOS technology was designed and analyzed. It features a 1024-bit capacity for
storage. The power consumption for read and write operations was 48.22 µW and 385 µW,
respectively, and used for the write driver circuit for low-power applications. The performance
attributes of the CMOS based SRAM arrays, such as the power dissipation for the read and write
operations, were contrasted with those of previous works. This paper was validated using the Cadence
Virtuoso tool in CMOS 22 nm technology.
The paper does not discuss security aspects such as susceptibility to side-channel attacks, which can
be crucial in memory designs, especially for applications dealing with sensitive data. So, to overcome
this addressing security concerns is essential in modern memory design which is incorporated in our
design.
16- bit SRAM cell using different configurations is designed and analyzed in180nm CMOS
technology in [5]. Based on the simulated results, Low power SRAM cells have the lowest power
consumption as compared to 6T and 8T SRAM and read and write operation is performed. Low power
SRAM shows the reduction in power consumption for the read cycle and write cycle. In comparison
with 6T SRAM, the low-power SRAM is 20% faster. The only drawback is area overhead due to an
increase in the number of transistors.
The paper mentions improved charging and discharging times for higher operating frequencies but
does not provide specific details about the read and write access times. So, to overcome this optimized
access time is a critical parameter in memory design which is incorporated in our design.
Govind Prasad, Bipin Chandra Mandi, Megha Jain [6] proposed the basic 6T Static random access
memory (SRAM) cell experience relatively high static and total power loss problem, to solve this 8T,
NC cells were designed. But this all consume more area as well as delay. Hence in this paper, 5T
SRAM cell proposed and checked by Monte Carlo Simulation. In this we have Considered the speed
of read and write operation.
Stability is decreased which needed to be increased for the high speed processer. This consumes more
Vibhash Choudhary, Dharmendra Singh Yadav [7] proposed 6T and 8T SRAM cells have been
compared on 180nm technology using an industry-standard Cadence Virtuoso Tool. It’s challenging
to make an SRAM cell with low power consumption and stay in a small space. The consumption of
power on both the SRAM cells are compared. The parameters of SRAM Static Noise Margin (SNM),
Write Delay, Read Delay and average power consumption are examined and discussed thoroughly.
Results illustrate that Read delay and Write delay will decrease of 8T as differentiate to 6T SRAM.
In this we have considered the variation of average power consumption and delay has been observed
as an operation of the supply voltage.
Results can vary depending on the chosen CMOS process technology (e.g., 45nm vs. 130nm).
Debasis Mukherjee, Hemanta Kr.Mondal [8] proposed the different types of analysis such as noise,
voltage, read margin and write margin of SRAM cell for high-speed application. The design is based
upon the 0.18 µm CMOS process technology. Static Noise Margin (SNM) is the most important
parameter for memory design. For demand of the high-speed application of the SRAM cell operation,
supply voltage scaling is often used that is why we have done Data Retention Voltage. We took
different types of curve by which straight forwardly we could analyses the size of the transistor of the
SRAM cell for high-speed application, In this we have considered the butterfly method to calculate
static noise margin, write noise margin and read noise margin of a 6T SRAM cell.
Primary issues like poor stability, process variation tolerance in SRAM design.
T. Wada, S. Rajan, and S. Przybylski [9] considered configuration, organization, and process
parameters for formulation of the access time of on chip cache.
It doesn’t consider interconnect delays and tag paths, for each delay stage it assumes a step input
waveform, and overgeneralized delay models and circuit models.
Hansraj, Ajay Rana [10] proposed the power dissipation of the projected new SRAM cell is sort of
stable for top speed operations. Since the number of transistors and therefore the space has
exaggerated compared to the standard SRAM 6T cell, this disadvantage will simply be overcome by
low power dissipation even at terribly high frequencies. This new style approaches the voltage mode
methodology to scale back voltage swing around writing switch activity. Within the standard SRAM
cell thanks to method variation, the hold on knowledge is also destroyed throughout the read
operation. As a result, the outpouring current thanks to bigger power dissipation is reduced. Dynamic
power is calculated for various frequencies and compared to the standard SRAM 6T cell.
Fails to provide specific details about the read and write access times and area overhead due to an
increase in the number of transistors.
Designing a basic 6T SRAM cell in which READ and WRITE operations are observed one after the
other. Each operation is done using the Tanner tool in the SEDIT in [11]. There is no need to refresh
the circuit each time because the power supply is given as the bits are already stored in the memory.
Also, the circuit operates at low power that is READ operation power is about 1.028720e-007 watts
and WRITE operation power is about 2.483478e-004 watts. In ideal conditions, it consumes very less
amount or negligible amounts of power in the circuit.
The paper doesn't provide information about the fabrication process or technology node used for the
design.
2.2 OVERVIEW OF LITERATURE SURVEY
From the above referred research paper there are few observations associated with Read and
Write Stability, Read and Write Speeds, Design Complexity, Power Consumption, and Cost. In
the research paper we have found that read, write, hold, operations were performed for the 6T &
8T SRAM and the performance criteria were total stability & average delay. The various design
techniques are used to overcome factors like power dissipation, read and write delay, and
stability. After analyzing the required research papers, we found that the fast access time and low
latency need to be upgraded so our main aim of this proposed work is to enhance the throughput
with read and write speeds. We have about to minimize access times to improve memory
bandwidth and system performance by multi-threshold voltage (Vt) design techniques using
advanced EDA tool. Since in current day world speed seems to be a prominent while designing
the integrated circuits with memories, if this not met slower SRAM can lead to reduced
responsiveness in computing systems
This proposed work focuses on designing and implementation of 1kb SRAM memory with
operating voltage of 1V using CMOS technology. By enhancing the speed and efficiency of
SRAM cells through CMOS technology, the project seeks to overcome existing limitations in
access times and power dissipation, thereby enhancing overall system performance in integrated
circuits.
2.4 OBJECTIVES
Design and Implementation of peripherals like sense amplifier, pre-charge circuit, 32X32
cell array, row decoder and column decoder.
CHAPTER – 3
METHODOLOGY
The Write operation updates the contents of an SRAM cell, whereas the Read operation fetches
the contents of SRAM cell.
Fig. 3.1 shows a schematic diagram of a 6T SRAM cell. 6 MOSFETs make up a typical SRAM
cell. Each bit in an SRAM is stored on two cross-coupled inverters made up of four transistors
(PM1, PM0, NM1 and NM0).
PM0 and PM1: Pull-up Transistors
NM0 and NM1: Pull-down Transistors
NM2 and NM3: Access Transistors
WL: Word Line
BL: Bit Line
BLB: Bit Line Bar
The digits 0 and 1 reflect the two stable states of this secondary cell. Two more access
transistors control the cell's entrance during read and write operations. The word line WL governs
whether the cell is linked to the bit lines BL and BLB, providing access to the cell. The bit lines
BL and BLB were utilised to transfer data in both read and write operations. Inverters in the
SRAM cell move the bit lines high and low when the cell is being read. Simulation result for 1-
bit SRAM cell is shown in fig 3.1.1.
A. Write Mode
The term "write mode" refers to the process of changing data. To write anything in SRAM, either
BLB or BL are discharged to ground. When logic 1 must be written, charge BL to VDD and
discharge BLB to ground. When writing logic 0, BLB is connected to VDD and BL to ground.
If data is to be written into the cell, WL is turned on. Assume a 0 is stored at NM1-PM1(node A)
and 1 must be written; 1 logic is applied to the BL line, which begins charging node A through
the NM2 transistor. As a result, PM0 is activated, and the output of inverter NM0-PM0 (node B)
begins to discharge, causing MM1 to activate. As a result, the value 1 is now written on node A.
B. Read Mode
Read Mode can be defined as fetching data from memory. The two access transistors NM2 and
NM3, which are coupled to the bit lines, are enabled when WL is set to read mode. The data from
nodes A and B is now transmitted to the bit lines. Assuming 0 is stored at node A, BL will
discharge through NM2 and PM1 transistors, and the BL BAR will be brought up to VDD
through NM0. In this mode of operation, the NM1 and PM0 transistors are turned off, but the
PM1 and NM0 transistors function in linear mode. To Read data from memory we need a sense
amplifier circuit.
The One-Bit 6T SRAM schematic with peripheral components. The signals used in the
simulation results: ‘pre-charge’ from the pre-charge circuit, ‘word-line’ from 6T SRAM circuit,
‘write_en’ corresponds to write enable signal from the write driver circuits, ‘bit-line’ and ‘bit-
line bar’ acts as input or output based on write/read operation, ‘q’ and ‘qb’ corresponds to the
storage nodes Q and QB of 6T SRAM during write operation.‘read_en’ corresponds to the read
enable signal given to the sense amplifier circuits and ‘out’ signal gives the output of reading
operation from sense amplifier circuit.
Fig. 3.3 shows 1-kb SRAM block structure which includes Cell Array, decoder, sense amplifier,
write driver, column mux, precharge circuitry, address latch and read and write control as the
peripheral circuit to the SRAM cell array. Since the memory core trade performance and reliability
in reduction area, memory design relies exceedingly on the peripheral circuitry to recover both speed
and electrical integrity. While the design of the core is dominated by technological consideration and
is largely beyond the scope of the circuit designer, it is in the design of the periphery where a good
designer can make an important difference. Generally, for smaller memory designs monolithic
architectures are preferred, but in the Design of bigger memories monolithic architecture will not
give efficient performance. The frequency of operation of the circuit is reduced by a factor of two as
the number of rows doubles. Similarly, the frequency of memory, reduced by a factor of four as the
number of 42 columns doubles, hence in bigger memory designs memory portioning technique is
used which is known as memory banking.
Write Circuit: This block is responsible for storing data in the memory cells. It involves
generating and controlling signals to write information into the selected memory cell.
Precharge Circuit: The precharge circuit prepares the bitlines in the memory array for read and
write operations. It sets the initial conditions for proper sensing and writing.
Row and Column Decoder: Row and column decoders are crucial for selecting the specific
memory cell in the array. The row decoder activates the desired row, while the column decoder
selects the appropriate column.
Sense Amplifier: The sense amplifier is used during read operations to detect and amplify the
small signals from the memory cells. It helps in accurate reading of the stored data.
In this work, we have designed precharge circuit for 32-bits which is required to operate in 1kb
SRAM memory. Test circuit for 32-bit precharge circuit is shown in fig 3.4.2 and the symbol is
shown in fig 3.4.2. Simulation result for 1-bit precharge circuit is shown in fig 3.4.4.
Sense Amplifiers plays a crucial role in the design of memories to achieve performance, reliability
and functionality of memory circuits. Figure 3.5 shows the design of sense amplifier circuit.
Normally sense amplifiers perform various operations like voltage amplification, reduction in
delay, power reduction and restoration of original signal. Generally, sense amplifiers are used in
the memories to speed up the read operation. Sense amplifier takes the small signal difference bit
line voltage as input and gives full swing single ended output. Access time and power consumption
of memory is affected by the sense amplifier hence the performance of memory is improved by
reducing both sensing delay and power dissipation.
Bit line and bit line bar is the input to the sense amplifier, these are highly capacitive bit lines and
are driven by SRAM cell. NM5 and NM7 are the differential input devices and inputs are given to
them, whereas transistors PM4 and PM5 act as an active current mirror load. The sense enable SE
signal is used to turn on and off the sense amplifier.
Test Circuit of the Sense Amplifier is shown in Fig 3.5.1.
Similarly, we have designed schematic and test circuit for 8-bit sense amplifier, required to operate
in 1kb SRAM memory. Test circuit for 8-bit sense amplifier is show in Fig 3.5.2. Symbol of 8-bit
sense amplifier is shown in Fig 3.5.3.
When SE is logic low then both bit line voltages are charges to supply voltage, when SE is logic
high then sense amplifier is getting ON and one of the bit line voltage discharges to ground via
pull down transistor. It takes BL and BLB voltages as an input and generates single ended output.
When BL voltage is greater than the BLB voltage then current through is increases and
simultaneously current through decreases to maintain as a constant, then the drop across is
decreases hence output voltage increases, which interprets output as the logic 1. Similarly, when
BL voltage is less than BLB voltage then it indicates output as the logic 0. In this way, Sense
amplifier plays a crucial role in the memory read operation. Simulation result is shown in fig 3.5.4.
Whenever a memory allows for random address based access, address decoders must be present.
The design of these decoders has a substantial impact on the speed and power consumption of the
memory. Since only one transition determines the decoder speed, it is interesting to evaluate other
circuit implementation. Dynamic logic offers a better alternative. A first solution is presented in
Fig. 3.6, where the transistor diagram of decoder is depicted. Notice that this structure is identical
to the NOR ROM array, differing only in the data patterns. The same structure can be used to build
5:32 decoder for 32x32 memory array. Test circuit for 2:4 decoder is shown in fig 3.6.1 and the
simulation results in fig 3.6.2.
Below given fig 3.6.2 represents the simulation result of 2:4 decoder. A 2:4 decoder typically has
five output lines: two inputs (enable, A, and B) and four outputs (Y0, Y1, Y2, Y3). The enable line
controls the overall activation of the decoder. When enabled, the specific combination of A and B
determines which output line (Y0 to Y3) goes high (typically logic 1) and the others go low (logic
0).
The waveform for a 2:4 decoder would show the voltage levels on the enable, input, and output
lines over time. During transitions, there might be short spikes or glitches as the outputs switch
states.
Row decoder, whose task is to empower one memory row out of 2^M, where M is the width of
particular fields in address word. When the number of inputs is more than or equal to four then the
speed of operation of the decoder is affected so pre-decoders are to be used which reduces the large
fan-in. The structure of 2:4 decoder is used to build 5:32 decoder. Fig 3.7 shows the schematic, Fig
3.7.1 shows the test circuit of the row decoder and the simulation result is shown in the fig 3.7.2.
While considering these decoders, it is imperative to keep the complete memory floor plan in
context. When the number of inputs is more than or equal to four then the speed of operation of the
decoder is effected so pre-decoders are to be used which reduces the large fan-in such that the speed
of the decoder is improved. The principal level is the pre decoder where two groups of address
inputs and their complements are first decoded to initiate one of the pre-decoder yield wires
separately to obtain the partially decoded outputs. The pre-decoder yields are consolidated at the
following level to enable the word line. The decoder delay comprises of word line wire delay,
interconnect delay of pre-decoder and gate delays in the critical path. As the wire RC delay develops
as the square of the wire length, the wire delays inside the decoder structure, particularly of the
word line, gets to be critical in extensive SRAMs. From delay analysis, it was observed that the
NOR based decoder is quicker than the NAND based decoder.
For 1-kb 8-bit SRAM block, Array is of size 32x32 and we have to select 8 out of 32 bit lines for
that 4:1 column mux is used. 2:4 decoder is needed to select one of the select lines of column mux.
It is better to design 4:1 mux using transmission gate logic. NMOS pass transistor logic mux works
very well during write cycle as one of the bit line need to discharge to ground and NMOS is good
to transfer zero logic. Fig 3.8 shows the schematic of column mux for write operation and the
waveform shown in fig 3.8.1.
During read cycle sense amplifier is connected after mux and due to bad logic 1 transfer of NMOS
sense amplifier can’t detect the difference in BT and BB until it goes above threshold voltage and
thus increases the delay in operation so we have used transmission gate logic mux to improve the
performance. Fig 3.8.2 shows the schematic of column mux for read operation and the waveform is
shown in fig 3.8.3.
A multiplexer (MUX) has been designed using transmission gates because they require fewer
transistors, resulting in a smaller chip area. Additionally, the transmission gate MUX provides a full
swing in the output, reducing static leakage. Both PMOS and NMOS transistors used in the design
are of 90nm technology.
In this proposed work, we have designed 8-bit column multiplexer using 8 single bit column
multiplexer which is shown in the Fig 3.8.2. Consider the design of 1Kb array, there is totally 32
numbers of SRAM cells present in a cell row and these 32 cells are divided into 4 portions such that
the output of SRAM memory is having a size of 8 bits. So, 4 to1 multiplexer is used to select one of
the portions out of 4 portions.
In write cycle, initially both bit line BT and bit line bar BB are charged to VDD. After that,
according to the data on data lines which is to be written in memory, either BT or BB is selected
to discharge to ground. If logic 1 need to write in a cell the BT is charged to VDD while BB gets
discharged ground and if logic 0 has to write then vice versa. Then the word line goes active and
data gets written in the cell. Now, after recharging BT and BB to VDD to discharge it to ground it
takes time, to minimize this we have designed a write driver for 1-bit as in Fig. 3.9, the test circuit
is shown in fig 3.9.1 and the waveform is shown in 3.9.2, which makes bit lines down to the ground
depends on input data. Size of transistors is more as it has to handle large current.
Before write operation both bit line voltages are charging to supply voltage and the write
operation is performed by enabling W_EN signal. Suppose if we want to write logic 0 in to the
memory cell, then the BB line voltage charges to supply voltage VDD and BT line voltage is
discharges to lower potential i.e. ground. The data stored in bit line, BT and bit line bar, BB is
accessed by enabling word line.
Similarly, we have designed 8-bit Write Driver to operate in 1Kb SRAM memory. Fig 3.9.3
shows the schematics of 8-bit Write Driver and the symbol for 8-bit Write Driver is shown in
Fig 3.9.4.
SRAM cell array size and its orientation are most important to consider before the start of circuit
design. For the design of bigger memories with particular operating frequency, we need to design
the small blocks of memory which satisfy the frequency requirement and multiple use of such
small blocks will give the bigger memories. Array size that is the number of rows and column of
SRAM array will work for some maximum frequency, if we want to increase the speed, array size
will go on reducing so that delay will be less. Not only has the speed that decides the number of
row and column in the SRAM array, but aspect ratio also needed to be considered. 1:1 is the perfect
aspect ratio of no. of rows to column. Suppose, for 1-Mb 8-bit SRAM memory, 1-kb is block we
are using so that operating frequency will be very high. Therefore, for 1kb SRAM the memory
array should have an aspect ratio as 128x8. It is not practically good to go with such aspect ratio
and also bit line capacitance increases with increase in bit line length. So, we have divided the
128 rows into 4 portions and used a 4:1 Mux to select one out of these. Now the memory array is
perfectly square that is aspect ratio of 32x32 and is shown in fig 3.10. Enlarged view of cells in
32*32 cell array is shown in the fig 3.10.1.
1-kb SRAM memory contain blocks like precharge circuit, row decoder, cell array, column mux,
sense amplifier, write driver. Complete schematic of 1-kb SRAM is shown in fig. 3.11. For 1-kb
SRAM the memory array should have an aspect ratio as 128x8. It is not practically good to go with
such aspect ratio and also bit line capacitance increases with increase in bit line length. So, we have
divided the 128 rows into 4 portions and used a 4:1 Mux to select one out of these. Now the memory
array is perfectly square that is aspect ratio of 32x32.
The SRAM can perform read and write operations at the granularity of a single bit. The row and
column decoders enable efficient access to any memory location within the 1kb SRAM. Sense
amplifiers are essential for reliable data reading due to the small voltage differences within the
SRAM cells. Precharging bit lines helps maintain a stable starting point for read operations.
The sense amplifier plays an important role in the functionality, performance and reliability of the
memory circuits. Precharge circuit is used to charge both bit lines to VDD. It is to be performed
before every read and write operation. As bit lines have high capacitance, precharge circuit needs
to provide large current to bit lines to get charged quickly. In write cycle, initially both bit line BT
and bit line bar BB are charged to VDD. After that, according to the data on data lines which is to
be written in memory, either BT or BB is selected to discharge to ground. If logic 1 need to write
in a cell the BT is charged to VDD while BB gets discharged ground and if logic 0 has to write then
vice versa. Then the word line goes active and data gets written in the cell. Now, after precharging
BT and BB to VDD to discharge it to ground it takes time, to minimize this we have designed a
write driver which makes bit lines down to the ground depends on input data. For 1-kb 8-bit SRAM
block, Array is of size 32x32 and we have to select 8 out of 32-bit lines for that 4:1 column mux is
used. The simulation result for 1kb SRAM memory is shown in fig 3.11.1.
CHAPTER-4
RESULTS AND DISCUSSION
The above figure 4.1 shows the simulation result of single bit SRAM cell which is used for read
delay and write delay calculation. Read delay is calculated by knowing how much time DIN will
take to give input for Q and write delay is calculated by knowing how much time SE will take to
give output to SOUT. Delay is calculated using above Fig 4.1 and values od delay is compared
with already existing recent research paper which is shown in below comparison table.
The stability of SRAM cell can be decided by mostly three parameters i.e. Static Noise Margin
(SNM), Read Margin (RM) and Write Margin (WM).cell ratio (CR) and pull-up ratio(PR) are two
parameters which effects the above three parameters.
SNM (Static noise margin) is a parameter which measures the stability of the SRAM cell to hold
its data compared to the noise. At the storing nodes, the minimum amount of noise voltage present,
which needs to be there to change the state of SRAM cell, is called SNM.
There exist two methods to measure the static noise margin of the SRAM cell.
The first method is the graphical approach in which SNM can be obtained by mirroring the
CMOS inverter characteristics and then obtaining the maximum possible square between
them.
The second approach involves the use of noise source voltages at the nodes.
SNM Dependences includes cell ratio, pull up ratio and supply voltage. The SNM of SRAM cell
can be calculated by using Butter fly method which is shown in the below figure 4.2 & 4.3. Cell
ratio, pull up ratio and power supply are the three important parameters which affects the noise
margin of the memory cell. Around seventy per cent of noise margin value is responsible by the
driver transistors.
During the reading of the data, ratio of driver to load transistors is Cell ratio, whereas pull up ratio
is defined as the ratio between the sizes of the load transistor to the access transistor while writing
of the data into the SRAM cell. In this proposed work cell ratio is 3.16 and pull up ratio is 2. Here
Static Noise Margin is found out to be 70.578 mV and 595 mV for read and write operation
respectively.
Dynamic Power: This is the dominant power consumption during read and write operations. It occurs
when transistors in the SRAM cell switch states. In this proposed work 82.23uW dynamic power is
consumed.
Static Power: This is the leakage current that flows even when the SRAM is not actively reading
or writing data. It's a continuous power drain. In this proposed work 36.617nW dynamic power is
consumed.
Power analysis in SRAM cells typically involves analysing the current flowing through the cell during
read and write operations to understand the cell's power consumption. This analysis can be done
through Simulation and Measurement. From the Fig 4.3, Simulating the SRAM cell circuit to estimate
the dynamic and static power consumption under various operating conditions. Measurement can be
done by using power analyzers to measure the actual power consumption during read/write operations.
CHAPTER-5
5.1 Conclusion
The proposed work involves the performance analysis of a 1kb Static Random-Access Memory
(SRAM) using 1-bit 6T SRAM cells. The simulations were conducted using Spectre, a popular tool
for circuit simulation, and the memory is designed using a 90nm CMOS technology node. The
primary focus of the project is on achieving high speed, and thus, attention is given to analyzing the
delay to ensure that the memory array operates within specified speed requirements. In this proposed
work successfully achieved its objectives, focusing on the development of a high-speed 1kb SRAM
using single bit 6T SRAM cells in a 90nm CMOS technology node.
The proposed work placed significant emphasis on achieving high-speed operation. Various design
techniques, including transistor sizing, topology optimization, and circuit-level innovations, were
employed to minimize access time and enhance overall speed. In the stability analysis of a single 6T
SRAM cell, the design achieves a Read Noise Margin of 70.578 mV and a Write Noise Margin of
595 mV for a supply voltage of 1V. These parameters are crucial indicators of the stability and
reliability of the memory cell. The Read Noise Margin ensures reliable read operations by providing
a sufficient voltage difference between the stable states, while the Write Noise Margin ensures robust
write operations by allowing for sufficient voltage margins to write data reliably into the memory
cell.
Overall, the analysis demonstrates that the designed 6T SRAM cell meets the stability requirements
for reliable memory operations. By achieving adequate noise margins and ensuring stability, the
memory array can operate effectively within the specified speed requirements, thereby enhancing its
performance and reliability in practical applications.
The future works that can be conducted for manufacturing of 2kb ,1MB chip using designed 1-kb
block. For that 1MB memory need to design using 1-kb block so that operating frequency will be
near to 1kb design. Whereas the design of bigger memories beyond the range of Mb such as Gb can
be designed by using scripting languages like Perl and SKILL because the layout design of such a
bigger memory could not be drawn with the hands of a man. Hence in industries the bigger memory
designs can be automated by using scripting languages to achieve better performance . By carefully
considering these strategies, challenges, and future trends, which create a roadmap for designing
larger memories using smaller building blocks while striving to maintain performance characteristics
comparable to the original 1kb design. The optimal approach will depend on specific application
requirements, technology limitations, and cost constraints.
REFERENCES
[1] “Design and Implementation of 16Bit 6T SRAM Using 45nm, 90nm,180nm CMOS
Technology” proposed by P Vinay Kumar [1], Ashish Kumar, Bhoomika S S,Ravika G , Sushma
K Sattigeri in the year 2021.
[2] “Design and analysis of 6T SRAM in 45 nm Technology”, V. Panduranga Vemula, S. Priyanka,
A. Sairam, IJERT, 2022.
[3] “Performance Evaluation of 6T SRAM cell using 90 nm Technology”proposed by Varanasi
Koundinya, Jithendra Pulivarthi, Madanu Karun Chand, D.Jayanth Sai Kumar in the year 2021.
[4] “Design and Performance Analysis of 32 × 32 Memory Array SRAM for Low Power
Applications” proposed by Aruru Sai Kumar, Osamah Ibrahim Khalaf, Rajendra Prasad ,2023.
[5] “Ultra Low Power SRAM Cell for High Speed Application using 90nm CMOS Technology”
proposed by Hansraj, Ajay Rana 2020.
[6] “Design of 16-bit Low power SRAM in 180nm CMOS Technology” proposed by Shubham
Santosh Kadam , Piyush Bhatasana in the year 2022.
[7] “Design and Analysis of Area and Power Optimised SRAM Cell for High-speed Processor”,
Govind Prasad, Bipin Chandra Mandi, Megha Jain, ICPC2T 2022.
[8] “The Stability Performance Analysis of SRAM Cell Topologies in 90nm and 130nm CMOS
technology”, Ajjay Gaadhe, Ujwal Shirode, Rajendra Kanphade, ESCI 2021.
[9] “Static Noise Margin Analysis of SRAM Cell For High Speed Application”, Debasis Mukherjee,
Hemanta Kr.Mondal 2021 IJCSI.
[10] “An analytical access time model for on-chip cache memories,” T. Wada, S. Rajan, and
S. Przybylski, IEEE J. Solid State Circuits, vol. 27, no. 8, pp. 1147–1156, Aug. 2022.
[11] “Ultra Low Power SRAM Cell for High Speed Application using 90nm CMOS
Technology” proposed by Hansraj, Ajay Rana, 2021.
[12] “Design of 16-bit Low power SRAM in 180nm CMOS Technology” proposed by
Shubham Santosh Kadam, Piyush Bhatasana in the year 2022.
[13] “Design and Analysis of 10T SRAM Cell with Stability Characterizations”, Deepika
Sharma, Shilpi Birla, ICAECT, 2023.
[14] “K. Sreevani, Sankara Reddy, “Design and Analysis of 1KB 6T SRAM cell in deep
submicron cmos technologies”, IJETTS, Vol. 5, Issue 3, September 2015.
[15] Preeti S Bellerimath, R. M Banakar, “Implementation of 16x16 SRAM Memory Array
using 180nm Technology”, International Journal of Current Engineering and Technology
(IJCET), September 2013.
Department of Electronics and Communication Engineering, BIT 40
DESIGN AND IMPLEMENTATION OF HIGH SPEED SRAM CELL 2023-24
[16] Nirali K. Khokhara, Bharat H. Nagpapa, “Comparative Analysis of 1 Bit SRAM using
different SRAM Cell in 45nm CMOS Technology”, International Journal of Innovative Research
in Computer and Communication Engineering (IJIRCCE), Vol. 5, Issue 1, January 2017.
[17] Nirali K. Khokhara, Bharat H. Nagpapa, “Design and Performance Analysis of 256 Bit
SRAM using different SRAM Cell in 45nm CMOS Technology”, International Journal of
Modern Trends in Engineering and Research (IJMTER), Vol. 4, Issue 3, March 2017.
[18] Shilpi Dubey, Pankaj Shrivastava, “Design and Analysis of Low Power 8x8 SRAM
Memory Array”, International Journal of Research and Analytical Reviews (IJRAR), Vol. 5,
Issue 4, December 2018.
[19] Pullareddy A., G. Sreenivasulu, R. Veerabadra Chary, “Improvement of SRAM stability
using read and write assist techniques”, International Journal of Engineering and Technology,
Vol. 6, Issue 3, 2017.
[20] Gaurav Hemant Patil, Irene Susan Jacob, Dada Bhagwan Sargar, Sneha Revankar,
“Design and Implementation of SRAM”, Proceedings of 22nd IRF International Conference,
22nd March 2015.
[21] Apoorva Pathak, Divyesh Sachan, Harish Peta, Manish Goswami, “A Modified SRAM
Based Low Power Memory Design”, 29th International Conference on VLSI Design and 15th
International Conference on Embedded Systems, 2016.
[22] Ezeogu Chinonso Apollos, “Design Principles of SRAM Memory in Nano-CMOS
Technologies”, International Journal of Computer Applications, Vol. 178, Issue 11, May 2019.
[23] Rohit Kumar Sah, Inamul Hussain, Manish Kumar, “Performance Analysis of a 6T
SRAM Cell in 180nm CMOS Technology”, IOSR Journal of VLSI and signal processing, Vol.
5, Issue 2, April 2015.
[24] Subhashree Rath, Siba Kumar Panda, “Analysis of 6T SRAM cell in different
technologies”, Circulation in computer science, 2nd National Conference on Mechatronics
Computing and Signal Processing (MCSP-2017), September 2017, pp:7-10.
[25] Abhishek Kumar, “SRAM Cell Design with minimum number of Transistor”
Proceedings of 2014 RAECS UIET Panjab University Chandigarh, 06 - 08 March, 2014.