1 Abstract— With the current trend in designing circuits that have low power operation characteristics, the focus is slowly shifting from size reduction to Low power consumption. From the semiconductor industry perspective, we have reached a point where shrinking the device size or packing a few more million transistors in a given space has reached a moot point. Only if a way to minimize the alarming increase in power consumption is satisfied, can we think of the next step in semiconductor process evolution. One such way is supply voltage scaling but having said that, we have a slew of problems that come with this innovation. This paper deals with the concept of sub-threshold operation and how this can be effectively utilized in memory circuits to improve the read/write stability even after operating them below the threshold voltage. The first part of this paper shows how an adaptive body biasing technique can be used to vary the threshold voltage. Rather than generally going ahead with the body biasing, this process is dependent on the operating frequency. The second part of this paper discusses on utilizing various subthreshold SRAM circuits and finding the optimum design in terms of reliability. This is done with the help of N stability curves. The whole project has been implemented in 45nm technology and the ensuing results are tabulated and analysed. Index Terms—Sub-threshold conduction, Adaptive Body Biasing, Ring Oscillator, SRAM, Schmitt trigger, N Stability curves, SVNM, SINM, Monte Carlo Simulation. I. INTRODUCTION Subthreshold circuits have gained immense usage in wireless sensors and biomedical applications mainly due to their low Power consumption characteristics. Unfortunately, Disadvantages such as increase in leakage current, performance issues and high sensitivity to noise have confined the usage of this domain to low performance spheres. Before moving forward with the project work, some of the earlier works in this field have to be mentioned. Authors in [1] have explained in the detail the need for operation in subthreshold region, its possible advantages and disadvantages. Moreover, some ways to negate the shortcoming though not altogether have also been given. The next paper listed as [2] gives a good idea about how the concept of adaptive body biasing can be employed to gradually scale down the threshold voltage, thus aiding in the operation in subthreshold region. The paper [3] was highly useful in explaining the concept of varying the body biasing as a function of operating frequency. Though the main concept of using Ring Oscillator as a part of adaptive body biasing is found in much detail in [4], the authors of [3] were very accurate in using this concept with the correct peripheral circuits to achieve the biasing as a function of the operating frequency. The paper [5] details on the possible SRAM circuitry and the reasons why most of them are not suitable for subthreshold operation. The same authors have proceeded to discuss in detail about Schmitt trigger based circuitry in [6] and [7] and how these can be used effectively as a subthreshold SRAM design technique. The proposed design to overcome the disadvantages of the Schmitt trigger based designs in the above mentioned papers has been used from the paper mentioned in [8]. To model the stability for all these circuits, the basics STABILITY ANAYSIS OF CMOS BASED SUBTHRESHOLD SRAM CIRCUITS Narayan Aiyer Venkatesan
17
Embed
STABILITY ANAYSIS OF CMOS BASED SUBTHRESHOLD SRAM …rsridhar/cse691/Present15/... · The whole project has been ... and how these can be used effectively as a subthreshold SRAM design
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
1
Abstract— With the current trend in designing circuits that have low power operation
characteristics, the focus is slowly shifting from size reduction to Low power consumption. From the
semiconductor industry perspective, we have reached a point where shrinking the device size or
packing a few more million transistors in a given space has reached a moot point. Only if a way to
minimize the alarming increase in power consumption is satisfied, can we think of the next step in
semiconductor process evolution. One such way is supply voltage scaling but having said that, we have
a slew of problems that come with this innovation. This paper deals with the concept of sub-threshold
operation and how this can be effectively utilized in memory circuits to improve the read/write stability
even after operating them below the threshold voltage. The first part of this paper shows how an
adaptive body biasing technique can be used to vary the threshold voltage. Rather than generally going
ahead with the body biasing, this process is dependent on the operating frequency. The second part of
this paper discusses on utilizing various subthreshold SRAM circuits and finding the optimum design
in terms of reliability. This is done with the help of N stability curves. The whole project has been
implemented in 45nm technology and the ensuing results are tabulated and analysed.
Index Terms—Sub-threshold conduction, Adaptive Body Biasing, Ring Oscillator, SRAM, Schmitt
trigger, N Stability curves, SVNM, SINM, Monte Carlo Simulation.
I. INTRODUCTION
Subthreshold circuits have gained immense usage in wireless sensors and biomedical applications mainly
due to their low Power consumption characteristics. Unfortunately, Disadvantages such as increase in leakage
current, performance issues and high sensitivity to noise have confined the usage of this domain to low
performance spheres. Before moving forward with the project work, some of the earlier works in this field
have to be mentioned. Authors in [1] have explained in the detail the need for operation in subthreshold
region, its possible advantages and disadvantages. Moreover, some ways to negate the shortcoming though
not altogether have also been given. The next paper listed as [2] gives a good idea about how the concept of
adaptive body biasing can be employed to gradually scale down the threshold voltage, thus aiding in the
operation in subthreshold region. The paper [3] was highly useful in explaining the concept of varying the
body biasing as a function of operating frequency. Though the main concept of using Ring Oscillator as a
part of adaptive body biasing is found in much detail in [4], the authors of [3] were very accurate in using
this concept with the correct peripheral circuits to achieve the biasing as a function of the operating frequency.
The paper [5] details on the possible SRAM circuitry and the reasons why most of them are not suitable for
subthreshold operation. The same authors have proceeded to discuss in detail about Schmitt trigger based
circuitry in [6] and [7] and how these can be used effectively as a subthreshold SRAM design technique. The
proposed design to overcome the disadvantages of the Schmitt trigger based designs in the above mentioned
papers has been used from the paper mentioned in [8]. To model the stability for all these circuits, the basics
STABILITY ANAYSIS OF CMOS BASED SUBTHRESHOLD
SRAM CIRCUITS
Narayan Aiyer Venkatesan
2
of SRAM stability was refereed to from paper [9]. A few more papers listed as [10], [11] and [12] deals with
the same topic but provides a more clear understanding of how to perform stability analysis with a SRAM
circuit. The Monte Carlo simulation for this project implemented using Cadence virtuoso was performed as
per the Iowa state Wiki entry in the link [13]. Low power designs need proper proof of reliability to show
their validity in comparison to other existing circuits. Thus this paper starts with discussing the problems and
the usefulness of subthreshold operation and finally end in proposing circuits to prove the reliability claim.
Section 3 discusses the adaptive body biasing technique and the components involved in it in depth. Section
4 discusses the SRAM design techniques and analyses the advantages and disadvantages of each of the
designs mentioned. Section 5 discusses the results by comparison with the existing SRAM techniques and
shows quantitative proof of increase in reliability. Section 6 discusses future works that can be done with
these design techniques and the scope for improving reliability using even further changes to the process and
thereby concluding the paper.
II. SUBTHRESHOLD CHALLENGES
Some of the main problems associated with subthreshold operations is their susceptibility to external noise.
This is due to the fact that in the subthreshold region, the ration of ON current to the ratio of OFF current that
usually is about 108 reduces to near thousands. The physical Implications of this being that the ON current
in subthreshold ranges in the domain of noise. And the OFF current also is in the same range. This causes
the circuit to be most likely be turned on by rampant noise signals. Another main problem in this mode is the
fact that as the supply voltage reduces, the leakage current increases. This causes the entire power
consumption of the circuit to be based on the leakage power. The equation to calculate the total power for a