Proceedings of the 2nd International Conference on Current Trends in Engineering and Management ICCTEM -2014

INTERNATIONAL JOURNAL OF ELECTRONICS AND

17 – 19, July 2014, Mysore, Karnataka, India
COMMUNICATION ENGINEERING & TECHNOLOGY (IJECET)
ISSN 0976 – 6464(Print)
ISSN 0976 – 6472(Online) IJECET
Volume 5, Issue 8, August (2014), pp. 107-116
© IAEME: http://www.iaeme.com/IJECET.asp ©IAEME
Journal Impact Factor (2014): 7.2836 (Calculated by GISI)
www.jifactor.com

AN IMPROVISED DESIGN IMPLEMENTATION OF SRAM

Poornima.H.S1, Deepu M2, Jyothi V3, Ajay.M.N4

1
Assistant Professor, Dept of ECE VVCE, Mysore
2, 3, 4
Student, Dept of VVCE, Mysore

ABSTRACT

Memory arrays are an essential building block in any digital system. Static random-access
memory (SRAM or static RAM) is a type of semiconductor memory that uses bistable latching
circuitry to store each bit. The term static differentiates it from dynamic RAM (DRAM) which must
be periodically refreshed. SRAM exhibits data remanence, but it is still volatile in the conventional
sense that data is eventually lost when the memory is not powered. The aspects of designing an
SRAM are very vital to designing other digital circuits as well. The majority of space taken in an
integrated circuit is the memory. SRAM design consists of key considerations, such as increased
speed and reduced layout area. This paper is aimed at creating an efficient SRAM design using
Cadence. The focus was on developing simplified design by reducing the transistor count and
replacing some of the conventional circuit designs.

Keywords: Bistable Data Remanence, Volatile, Cadence.

I. INTRODUCTION

Static random-access memory (SRAM) is a type of semiconductor memory that uses

bistable latching circuitry to store each bit. The term static differentiates it from dynamic RAM
(DRAM) which must be periodically refreshed. SRAM exhibits data remanence. SRAM is designed
to provide an interface with CPU and to replace DRAMs in systems that require very low power
consumption. A SRAM cell must meet the requirements for the operation in submicron/nano ranges.
The scaling of CMOS technology has significant impacts on SRAM cell – random fluctuation of
electrical characteristics and substantial leakage current.

II. SRAM ARRAY

The basic architecture of a SRAM consists of an array of memory cells with support circuitry
to decode addresses and implement the read and write operations. SRAM arrays are arranged in rows
and columns of memory cells called wordlines and bitlines, respectively. Typically, the wordlines are
107
 Proceedings of the 2nd International Conference on Current Trends in Engineering and Management ICCTEM -2014
17 – 19, July 2014, Mysore, Karnataka, India

made from poly silicon while the bitlines are metal. Each memory cell has a unique location or
address defined by the intersection of a row and a column.

2.1 6T SRAM cell

A typical SRAM cell comprises of two cross-coupled inverters forming a latch and two
accesses. Essentially, the data is latched at the cross-coupled inverters. The bit-lines are
complementary and are input to the I/O of the inverters. Thus, the value is latched during a write and
maintained as long as power is available.

2.2 Transistor sizing

An SRAM Cell has to provide non-destructive read and a quick write-the two opposing
requirements impose constraints on the cell transistor sizing. The 6-transistor (6T) SRAM core
shown in Fig 2.1 stores one bit of data. The size ATION of the transistors used is the primary factor
that determines the performance of the SRAM cell. Since power dissipation is a constraint, we
minimize the sizing as much as possible without compromising performance significantly.

Fig. 2.1: Sram cell

Fig. 2.2: Schematic of Sram cell

There are some issues to be considered when sizing the transistors. The latch inverters (M1,
M2, M3, and M4) form a positive feedback loop, so that the stored value is maintained as long as
power is available. Since the bit lines are pre-charged to VDD-Vtn, the cell NFETs (M1 and M3)
cannot be smaller than the pass NFETs (M5 and M6) to overcome the current value on the bit line
when pulling it to a low value. Note that though a transmission gate may be used for the pass
transistors, only NFETs are used so that the area for a single SRAM cell may be small. It will be
shown later that special circuitry (bit-line conditioning and sense amplifiers) is needed to recover
from the performance losses due to using just NFETs .In an array of RAM cells, a single word line is
connected to an entire row of RAM cells, forming a long word-line row. Since the word line uses
108
 Proceedings of the 2nd International Conference on Current Trends in Engineering and Management ICCTEM -2014
17 – 19, July 2014, Mysore, Karnataka, India

poly silicon (which has high resistivity), it is necessary to keep the two pass transistors (M5 and M6)
small. This improves signal integrity on the word lines and reduces power dissipation. Therefore, the
size is kept small.
During the read operation, it was concluded that transistor Q1 had to be stronger than
transistor Q5 to prevent accidental writing. Now in the write case, this feature actually prevents a
wanted write operation. Even when transistor Q5 is turned on and current is flowing from BL to the
storage node, the state of the node will not change. As soon as the node is raised transistor Q1 will
sink current to ground, and the node is prevented from reaching even close to the switching point. So
instead of writing a ’1’ to the node, a ’0’ will be written to the inverse node. Looking at the right side
of the cell we have the constellation Q4-Q6. In this case BLB is held at gnd. When the wordline is
raised Q6 is turned on and current is drawn from the inverse storage node to BLB.
At the same time, however, Q4 is turned on and, as soon as the potential at the inverse
storage node starts to decrease, current will flow from VDD to the node. In this case Q6 has to be
stronger than Q4 for the inverse node to change its state. The transistor Q4 is a PMOS Transistor and
inherently weaker than the NMOS transistor Q6 (the mobility is lower in PMOS than in NMOS).
The Sram array is functionally divided into four blocks

• I/O section
• Decoder Section
• Control section

III. I/o SECTION

This section of SRAM is responsible for pre-charging bitlines, sensing differential analog
voltage during read operation and driving bitlines during write operation. It has the following
functional blocks.

• Pre-charge
• Sense Amplifier
• Write Amplifier

3.1 Pre-charge circuit

The pre-charge circuit ensures charging of bitlines before read operation. When two RAM
cells containing opposite value in the same columns are accessed subsequently, the output has to
switch first to an equalized state and then to the opposite logic state. Since the capacitance on the bit
lines is quite large, the time required for switching the differential from one state to the other
becomes a significant portion of the overall access time. Equalization of the bit-lines between the
accesses can reduce the access time.
The size of the charge transistors must be as small as possible, so that they do not override
the value in the latch during read and write operations. The schematic of pre-charge circuit is as
shown in Fig 3.1.

Fig. 3.1: Schematic of precharge circuit

109
 Proceedings of the 2nd International Conference on Current Trends in Engineering and Management ICCTEM -2014
17 – 19, July 2014, Mysore, Karnataka, India

3.2 Sense Amplifier

Sense amplifiers are an important component in memory design. The primary function of a
sense amplifier in SRAMs is to amplify a small analog differential voltage developed on the bit lines
by a read-accessed cell to the full swing digital output signal thus greatly reducing the time required
for a read operation. Since SRAMs do not feature data refresh after sensing, the sensing operation
must be non-destructive, as opposed to the destructive sensing of a DRAM cell. A sense amplifier
allows the storage cells to be small, since each individual cell need not fully discharge the bit
line.The design using NAND latch overcomes time criticality of SEN signal by implementing a
NAND-latch. The output of which is ANDED with WEN (Write enable) which is active low that is
for read operation WEN=1. The circuit is as shown in Fig 3.2.

b’ Pch’

wen

dout
Fig. 3.2: Sense amplifier using NAND latch

3.3. Write amplifier

This circuit helps in driving bitlines high or low during write operation. The design shown in
Fig 3.3 is a circuit consisting of 4x and 8x inverters used as drivers (as the bitlines have to be driven
through 64 rows to reach the topmost cell, they require drivers). The pass transistors as shown in
figure isolates Din (data in) during read operation, they are enabled by the write select signal which
arrives from the write select generator of the control block. To drive the bitlines we have used 8x
inverter for complementary bitline and two 4x inverters for bitline.

din

Fig. 3.3: Write amplifier

110
 Proceedings of the 2nd International Conference on Current Trends in Engineering and Management ICCTEM -2014
17 – 19, July 2014, Mysore, Karnataka, India

3.4 Decoder section

The decoder decodes the input address to access the particular row. The multiple-stage
decoding uses several hierarchically-linked blocks. Normally, the most significant bits are decoded
(pre−decoded) in the first decoder stage, effectively selecting the array that is to be accessed by
providing enable signals for the subsequent decoder stage(s) that enable a particular word line. The
number of outputs of the last decoding stage corresponds to the number of rows (word lines) to be
decoded
A pre-decoder is used in the design to reduce the fan-in and the size of the NAND gate
needed to create the decoder. This results in a faster decoder. For a 64x8 memory array that is
designed, two stage decoding is chosen.
The optimized decoder leaf cell has two stages with 3 input NAND gates at the first stage and
a 3 input NOR gate with inverted clock as one of the inputs at the second stage as shown in Fig 6.
The transistor count reduced to 24 decreasing the area as compared to the conventional D flip flop
using NAND gates.

Fig. 3.4: Decoder leaf cell

3.5 Control section

The control section is responsible for generating the signals to control the i/o and the decoder
section. It includes the flip flops and the write select generator which go to decoder section and i/o
section respectively.

3.6 Flip flop

The flip flops are used to latch the inputs given to the memory that is 6 bit address, data to be
written and the write enable (w_en).The master-slave D flip flops are used to latch the inputs. The
2:1 mux realized using the tri-state inverters is shown in the Fig. 3.5. This can be used as a D flip
flop. The master slave D flip flop is used instead of a single stage D flip flop to avoid race round
condition. The circuit is as shown in the Fig. 3.6 and the schematic is shown in Fig 3.7.

Fig. 3.5: Schematic of 2:1 mux realized using tristate inverters

111
 Proceedings of the 2nd International Conference on Current Trends in Engineering and Management ICCTEM -2014
17 – 19, July 2014, Mysore, Karnataka, India

Fig. 3.6: Master-slave D flip flop using 2:1 mux

Fig. 3.7: Schematic of master-slave D flip flop

3.7 Write select (w_sel) generator

The read and write operations are synchronized with clock. Both the operations take place
when the clock goes high after the precharging phase. The write operation should take place only
when the write enable (w_en) goes low (active low pin) and clock goes high. Therefore the write
amplifier in the I/O section should be activated when both these signals satisfy the above conditions.
The latched write enable signal is ANDed with the delayed clock to generate the write enable
signal as shown in Fig 3.8. The write enable signal from the flip flop is inverted before ANDing as it
is an active low pin. The clock is delayed to compensate for the propagation delay in the flip flop and
the wire delay from flip flop output to the input of nand gate. If the clock is not delayed, the write
select generated will be a small firing pulse insufficient to activate the write amplifier in the I/O
section.

Fig. 3.8: Write select generator

112
 Proceedings of the 2nd International Conference on Current Trends in Engineering and Management ICCTEM -2014
17 – 19, July 2014, Mysore, Karnataka, India

IV. DIGITAL VECTOR FILE AND MIXED MODE STIMULI

Vector files come into picture when long bit patterns are used as stimuli. SPECTRE and
SPECTRE RF input netlists support digital vector files. A VEC file consists of three parts:

• Vector Pattern Definition section

• Waveform Characteristics section
• Tabular Data section.

4.1 Vector Patterns

The Vector Pattern Definition section defines the vectors, their names, sizes, signal direction,
sequence or order for each vector stimulus, and so on. A RADIX line must occur first and the other
lines can appear in any order in this section. All keywords are case-insensitive.
Here is an example Vector Pattern Definition section:

; start of Vector Pattern Definition section

RADIX 1111 1111
VNAME A B C D E F G H
IO IIII IIII
TUNIT ns

These four lines are required and appear in the first lines of a VEC file:

1. ADIX defines eight single-bit vectors.

2. VNAME gives each vector a name.
3. IO determines which vectors are inputs, outputs, or bidirectional signals. In this example, all
eight are input signals.
4. TUNIT indicates that the time unit for the tabular data to follow is in units of nanoseconds.

4.2 Defining Tabular Data

Although the Tabular Data section generally appears last in a VEC file (after the Vector
Pattern and Waveform Characteristics definitions), here we describe it first to introduce the
definitions of a vector. The Tabular Data section defines (in tabular format) the values of the signals
at specified times. Rows in the Tabular Data section must appear in chronological order because row
placement carries sequential timing information. Its general format is:

time1 signal1_value1 signal2_value1 signal3_value1...

time2 signal1_value2 signal2_value2 signal3_value2...
time3 signal1_value3 signal2_value3 signal3_value3...

Where timex is the specified time, and signaln_valuen is the values of specific signals at
specific points in time. The set of values for a particular signal (over all times) is a vector, which
appears as a vertical column in the tabular data and vector table. The set of all signal1_valuen
constitutes one vector.

For example,
11.0 1000 1000
20.0 1100 1100
33.0 1010 1001

113
 Proceedings of the 2nd International Conference on Current Trends in Engineering and Management ICCTEM -2014
17 – 19, July 2014, Mysore, Karnataka, India

This example shows that:

1. At 11.0 time units, the value for the first and fifth vectors is 1.
2. At 20.0 time units, the first, second, fifth, and sixth vectors are 1.
3. At 33.0 time units, the first, third, fifth, and eighth vectors are Specifying a Digital Vector
File and Mixed Mode Stimuli

4.3 Input Stimuli

SPECTRE or SPECTRE RF converts each input signal into a PWL (piecewise linear) voltage
source, and a series resistance. Table 15 shows the legal states for an input signal. Signal values can
have any of these legal states.

4.4 Expected Output

SPECTRE or SPECTRE RF converts each output signal into a .DOUT statement in the
netlist. During simulation, SPECTRE or SPECTRE RF compares the actual results with the expected
output vector(s). If the states are different, an error message appears. The legal states for expected
outputs include the values listed.

V. RESULT

The simulation using vector files and the layout of complete array is shown in fig. 4.1 and fig.
4.2 respectively.

Fig. 4.1: Waveform of SRAM layout simulation

The waveform obtained using Wavescan window for layout of SRAM is shown in Fig 4.1.
The Tcq (Clock to Q) for layout of SRAM was found to be 1500 ps for a maximum clock frequency
of 250MHz.

114
 Proceedings of the 2nd International Conference on Current Trends in Engineering and Management ICCTEM -2014
17 – 19, July 2014, Mysore, Karnataka, India

Fig. 4.2: Layout of SRAM array and peripheral blocks

VI. CONCLUSION

The complete SRAM array includes peripheral components such as memory bit cell, writer
driven circuit, pre-charge circuit, sense amplifier and flip flops. The circuit is designed for storage
capacity of 512 bits. The decoder is a 6:64 decoder implemented in order to select 64 wordlines in
the bit cell array. The sense amplifier and write amplifier are designed in accordance to the
requirement of the circuit. The D flip flops used here, synchronize the data and address lines with the
clock signal of the circuit in order to perform the necessary read and write operation on the bit cell.
The proposed work is operating within an input voltage of 0 to 1.8v. The DRC and LVS are
performed for all the components and correspondingly necessary optimization is carried out for the
same. Eventually, designing and implementing a highly integrated & efficient SRAM circuit design.

VII. FUTURE WORK

We intend on extending the memory size to 1KB and with the help of Cadence SKILL code
designing a memory compiler. An SRAM compiler can be developed for the automatic layout
generation of memory elements in the ASIC environment. The compiler can generate an SRAM
layout based on a given SRAM size, input by the user, with the option of choosing between fast vs.
low-power SRAM.
The language that will be used to perform the layout automation is Cadence’s SKILL.
SKILL, which stands for Silicon Compiler Interface Language, has tool specific functions for several
of Cadence Suites –Virtuoso (Layout Editor) and Composer (Schematic Editor), among others.

115
 Proceedings of the 2nd International Conference on Current Trends in Engineering and Management ICCTEM -2014
17 – 19, July 2014, Mysore, Karnataka, India

VIII. ACKNOWLEDGEMENT

First and foremost we pay our due regards to our renowned institution Vidyavardhaka
College of Engineering, which provided us a platform and an opportunity for carrying out this work
and our guide Sunil Kumar H V, layout team lead manager, Sankalp Semiconductors, Bangalore.

IX. REFERENCES

[1] Bhavya Daya, Shu Jiang, Piotr Nowak, Jaffer Sharief Synchronous 16x8 SRAM Design,
Electrical Engineering Department, University of Florida.
[2] Mehdi Alipour, Mostafa E. Salehi1, Hesamodin shojaei baghini, Design Space Exploration to
Find the Optimum Cache and Register File Size for Embedded Applications Islamic Azad
University.
[3] Meenatchi Jagasivamani Development of a Low-Power SRAM Compiler Virginia Polytechnic
Institute and State University.
[4] Andrei Pavlov and Manoj Sachdev CMOS SRAM Circuit Design and Parametric Test in
Nano-Scaled Technologies, Process-Aware SRAM Design and Test.
[5] Andrei S Pavlov Design and test of embedded SRAMSs University of Waterloo.

116