4612-Article Text-8827-1-10-20201230
4612-Article Text-8827-1-10-20201230
4612-Article Text-8827-1-10-20201230
Design A Low Power And High Throughput Error Detection And Data
Correction Architecture By Razor II Method
ABSTRACT
The proposed error detection and correction circuit designed due to the existing
circuits accommodate the worst-case delay. To prevent Error in the system, detect and
determine violation to maintain correctness to help on the fly mechanisms. The proposed
circuit is to present speculative error detection technique along with an error recovery
mechanism. Circuits are wanted to oblige the delay and to get to be deficient in their
execution. To enhance the execution, they oblige fly system to forestall, identify and correct
errors. In this paper, low power speculative error detection and error recovery architecture are
4393
PJAEE, 17 (9) (2020)
to be developed. The main aim of the circuit is to reduce delay, power and area. This paper
demonstrates their ability to operate under worst-case accommodation. The proposed error
correction and detection circuit give 226nW, propagation delay 1ps, throughput 792MHz..
1. Introduction
The enforcement of I.C. designers laid a great exertion in diminishing the
energy utilization of VLSI systems and suspension gave the inclusive
demand for higher speed and more efficient electronics. Circuit and
architectural techniques are using the difference in the consumption of VLSI
system [1]. Even though the standard method of pipelining is valid and
reliable, but the clock frequency is not flexible, and it always stays at the
critical path of the framework. In few others, they improve their adaptability
to bypass the critical path in the system and more to design. In contrast,
other circuit technique focuses more on decreasing the margin to the
frequency of the clock because of voltage, thermal vibration, and process.
The sources of the changing variant have many different characteristics
which can be vigorous to realize. It is essential to understand that diverse
variation sources have their attributes which cause Error to be recognized,
and recovery methods are utilized [1]. There are two main variations, such
as random variation and systematic variation, respectively. Random
variation can be a label under static dynamic. Mostly for line end roughness
effect and the non-predictable voltage fluctuations could show in
unpredictable variation performance. Static random variation is a random
dopant fluctuation that gives effect to the transistor threshold voltage (Vth).
Moreover, the dynamic random variations can be voltage vacillations,
usually not sure which way the voltage will change, and even though we
know how it will change after some time. It is difficult to say how it will
precisely influence a specific circuit. The second major of variation is
Systematic variations [2]. Systematic variations can adequately describe and
displayed as they are indeed known and unsurprising. Sources of variation
in CMOS circuits are static variation and dynamic variation. Static variation
usually is fabricated dies, which results in more significant power
consumption and a maximum frequency of degradation and usually has
many problems in sub-100nm technologies. The local and global procedure
variation impacts in feature size, the rate of the transistors can shift
drastically from dying to die or gadget to gadget. A few planners use
procedures to lessen the timing edges and enhance speed. However, others
utilize the same methods to spare dynamic energy by diminishing VDD and
working at perfect or lower rates. The examination had figure out that
minimum energy accomplishes when VDD enters the subthreshold, where
can figure out how to attain ten reductions in energy per computation [2]. It
4394
PJAEE, 17 (9) (2020)
2. Design methods
Design is the passage through which creativity realized. It is necessary to
understand the different aspects of fundamental laws to practical
considerations that conduct it. There are some fundamental principles of
low power design [8, 9] that are by using the smallest geometry of the
highest frequency devices or the lowest possible supply voltage.
2.1 Error Detection Circuit
A Razor I flip-flop circuit is used metastability detection circuits [10]. The
configurations of the metastability indicator are expanding process variation
because it expected to make a reaction to the flip-flop outputs. Other than
that, it additionally obliges the utilization of more significant devices which
near to harmful effects of the area and power dissipation of the Razor I flip-
flop. So, the additional or included danger of metastability at
the restored signal that will spread to pipeline control logic, conceivably
major to flip-flop disappointment. The micro-architectural domain is to
4395
PJAEE, 17 (9) (2020)
successfully place the timing issue and design in Razor I, which shown in
Figure 1.
Rather than showing both corrections in flip flop and also in error detection,
Razor II shows the only detection in flip flop. Still, the correction in it is an
architectural replay [11]. Although the price of the IPC penalty gets higher
during recovery, these allowed the complication and the size reduces
Architectural replay is a conventional method that consists of better
performing microprocessors to help in the operation, for example, out-of-
request execution and branch prediction. Besides, that is easy to overwhelm
the remaining framework, which helps recap in the timing errors event.
These methods need for pipeline restore signal, whereas expressively
unwinding the timing imperatives on the error-recovery way [12]. This
Razor II characteristic is agreeable to utilize in better-performance
processors.
The procedure of Razor II flip-flop comprises a positive level latch contrast
with the master-slave flip-flop [13]. It describes that any changes on the
input information of the decisive clock stage with timing mistake. Rejection
of master latch expressively brings down the clock-pin limit by reducing its
energy furthermore overhead area. Other than that, it additionally offers the
opportunity to the Razor II flip-flop to regularly find the Single Event
Upsets (SEU) sensibly and registers without the overhead. So, it gives lower
area value [14].
4396
PJAEE, 17 (9) (2020)
4397
PJAEE, 17 (9) (2020)
The area of the Razor circuit is 0.644nm. The numbers of transistors are 17
NMOS and 31 PMOS.
2.3 Proposed Error Correction Circuit
When the Error is detected, a particular form of procedure needed to
approach for the Error noticed to be handled systematically and in a proper
manner. This section describes the design and implementation of the
proposed error correction circuit. A one-bit comparator circuit used in this
architecture. The comparator circuit compares the two bits and gives high
output if A bit greater than B bit, which is known as F= AB'. The proposed
Error Correction Circuit show in Figure 5. The design obtained using the
DSCH2 method, as stated earlier. The circuit uses three NOT gate, six AND
gate, four OR gate and two 2-to-1 Multiplexor circuit to give the output
logic. The MUX takes in 2 input to produce the cumulative output data with
the help of the selection input, either logic "1" or logic "0". The MUX
designed in order of selection input and its complement. The output of the
MUX circuit is one of its inputs [18]. A 2-to-1 multiplexer Boolean
equation shows in Equation 3.1. Where A and B stands for the two inputs,
selector input is S, and the output gives as Z:
𝑍 = (𝐴. 𝑆̅) + (𝐵. 𝑆) (1)
As shown in Figure 4 proposed error correction circuit, the error correction
inputs are given in the input sector of a0, a1, a2 and a3, respectively. It can
be feed either logic low or logic high. These four inputs are data bits or error
bits stream for the circuit. Then, the data bits pass into one bit 2-to-1
comparator circuit. The error bits stream is compared by two bits
consequently. Using the given data bits, the A greater than B one-bit
comparators compare the first value with the second value, which is in
sequential order. The circuit is because the error bits maybe one or zero in
the logical terms. If A is greater than B in logical wise, the comparator
circuit gives logic "1" otherwise logic "0". According to the parity check,
the comparison has done from a higher value to lower value. After that, the
comparing values sent to the two-input OR gate as one of the inputs.
4399
PJAEE, 17 (9) (2020)
4401
PJAEE, 17 (9) (2020)
180nm feature size, respectively. Similarly, the propagation delay, chip size,
EPI, latency, throughput and PDP than other feature sizes. The error
correction circuit achieved than other existing circuits. The proposed circuit
has the lowest power dissipation, lower total chip area and lower power
delay product. It achieves the aim of the paper, which is to achieve circuit
with low power, small area and low delay.
Figure 7 is shown the LVS graph of voltage vs time obtained from the
simulation of the proposed Razor II, which measure between sum and input
gives a more excellent performance. For the feature size 65 nm, the output
voltage from Figure 7 is 0.689V. Based on the NMOS technology, the
output needs to be VDD /2 for linear operation [19]. This proposed error
detection gives higher output voltage that is more than a linear region
voltage due to the actual arrangement of NMOS transistors and reduced
critical path in the circuit. The critical path increases the parasitic
capacitance is reduced, therefore decreases the delay of output (spike).
There is no transition delay between the output data (spike).
4402
PJAEE, 17 (9) (2020)
4403
PJAEE, 17 (9) (2020)
A3 A A A O1 O2 ER ER ER ER
2 1 0 1 2 3 4
0 0 0 0 0 0 0 0 0 0
0 0 0 1 1 0 0 0 0 0
0 0 1 0 0 0 0 1 0 0
0 0 1 1 1 0 0 0 0 0
0 1 0 0 0 1 0 0 1 0
0 1 0 1 1 1 0 0 0 0
0 1 1 0 0 1 0 1 0 0
0 1 1 1 1 1 0 0 0 0
1 0 0 0 0 0 1 0 0 1
1 0 0 1 1 0 0 0 0 1
1 0 1 0 1 0 1 1 0 1
1 0 1 1 1 0 0 0 0 1
1 1 0 0 0 1 1 0 1 0
1 1 0 1 1 1 0 0 1 0
1 1 1 0 1 1 1 1 0 0
1 1 1 1 1 1 0 0 0 0
4404
PJAEE, 17 (9) (2020)
The values for the designed Error Correction circuit are taken from the
layout structure. The analyses are made for the in-depth sub-micron process.
The result obtained for the 4 CMOS technology fixture size is summarized
in Table 3. The Ultra Deep Sub Micron (UDSM) feature size (180 nm, 120
nm, 90 nm and 65 nm) simulated results are shown in Table 3.
For the feature size of the 65 nm, the output voltage is 0.696V. It is
calculated from Voltage vs Current graph. The corresponding output voltage
for 90 nm, 120 nm and 180 nm is 1.001V, 1.197V and 1.988, respectively.
Simulation layout results check the Average Current and Output Power. The
65 nm feature size gives the lowest average current while 180nm gives the
highest average current due to transistor channel complete. The delivered
power is calculated using the standard equation. For 65 nm, the output
power dissipation 0.226μW is calculated from subtracts the value of total
power of the circuit and power delivered in the component (output node) of
the circuit. Therefore, the output power dissipation obtained for the
corresponding technologies is 0.298 µW, 0.126 µW, and 1.972 µW,
respectively. Feature size 120nm gives the lowest value of power
dissipation. The circuit signal propagation delay is measured from the
output signal of the designed error detection circuit, correspondingly the rise
time and fall time of the output signal. The propagation delay acquired is
1ps for all the four technologies.
Table 3 also illustrates the total area of the designed gate. The total chip
area has been calculated in terms of transistor area, interconnect area, wire
area, input and output pad area. For the 65 nm CMOS technology feature
size, PDP is 0.226 fWs. The results obtained for the corresponding CMOS
technology feature sizes are 0.298 fWS, 0.126 fWS and 1.972 fWS,
respectively. The overall best feature size for the proposed error correction
circuit is CMOS technology 65nm. It achieves the aim of the paper, which
is to achieve circuit with low power, small area and low delay.
Table 3: The simulation results and performance analysis results of
Proposed Error Detection Architecture.
Output variable Proposed Error Detection Architecture (Feature size)
65nm 90nm 120nm 180nm
Vo(V) 0.696 1.001 1.197 1.988
IDDAvg(mA) 0.002 0.008 0.013 0.081
Po(µW) 1.618 8.306 15.687 163.0
PD (µW) 0.226 0.298 0.326 1.972
td(ps) 0.896 0.916 1.562 8.963
Area(µm2) 330 385 738 2349
EPI (fJ) 19.29 40.473 62.993 2349
Latency, T(ns) 5.051 5.012 4.999 5.008
Throughput, R 0.792 0.798 0.800 0.799
PDP (fWs) 0.226 0.298 0.126 1.972
4406
PJAEE, 17 (9) (2020)
4. Conclusions
The main aim of this paper is to design low power error detection and data
correction architecture. The error correction and detection circuits are Razor
II circuit proposed and simulated. The proposed error detection and error
recovery architecture show a significant improvement in terms of power
dissipation, area and Energy Per Instruction (EPI) value. Apart from the
reduction in low power, low area and low delay are also achieved.
Theoretical systems give a better description of circuit delay with
acknowledged timing to enhance throughput or possibly hypothesize timing
impelled errors. Usually, error detection and data recovery can be a dispute,
mainly when working under a sub-threshold region. The essential strategies
that accepted will keep on being utilized. In the futures, a well-designed
4407
PJAEE, 17 (9) (2020)
Razor Circuits gives lower power dissipation less occupying area and the
Multiple Issue using the pipeline technique that matched with higher-level
techniques, for example, error-correcting codes
References
Agarwal, K.; Nassif, S. Characterizing Process Variation in Nanometer
CMOS. In Proceedings of the 44th Annual Design Automation
Conference (DAC '07), San Diego, CA, USA, 4–8 June 2007; pp.
396–399.
Hanson, S.; Zhai, B.; Bernstein, K.; Blaauw, D.; Bryant, A.; Chang, L.; Das,
K.K.; Haensch, W.; Nowak, E.J.; Sylvester, D.M. Ultralow-voltage,
minimum-energy CMOS. IBM J. Res. Dev. 2006, 50, 469–490.
Krimer, E.; Pawlowski, R.; Erez, M.; Chiang, P. Synctium: A near-threshold
stream processor for energy-constrained parallel applications.
Comput. Archit. Lett. 2010, 9, 21–24.
Tschanz, J.; Bowman, K.; Walstra, S.; Agostinelli, M.; Karnik, T.; De, V.
Tunable Replica Circuits and Adaptive Voltage-Frequency
Techniques for Dynamic Voltage, Temperature, and Aging Variation
Tolerance. In Proceedings of the 2009 Symposium on VLSI Circuits,
Kyoto, Japan, 16–18 June 2009; pp. 112–113.
Mintarno, E.; Skaf, J.; Zheng, R.; Velamala, J.; Cao, Y.; Boyd, S.; Dutton,
R.; Mitra, S. Self-Tuning for Maximized Lifetime Energy-Efficiency
in the Presence of Circuit Aging. IEEE Trans. Comput.-Aided Des.
Integer. Circuits Syst. 2011, 30, 760–773.
H.T. Nguyen, A. Chatterjee, "Number-splitting with shift-and-add
decomposition for power and hardware optimization in linear DSP
synthesis", IEEE Trans. Very large Scale Integr. (VLSI) Syst. 8 (4)
(2000) 419-424.
D Kodandapani P Velrajkumar, C Senthilpari, P Ramesh, G
Ramanamurthy Development of Smart Number Writing Robotic
Arm using Stochastic Gradient Decent Algorithm International
Journal of Innovative Technology and Exploring Engineering, Vol 8,
issue 10, 2019, 542 – 547.
D. Soudris, C. Piguet, C.G. Esigning, CMOS Circuits For Low power.
European Low Power Initiative for Electronic System Design,
Kluwer Academic Publishers, Boston/Dordrecht/London.
C Senthilpari, A Low-power and High-performance Radix-4 Multiplier
Design Using a Modified Pass-transistor Logic Technique IETE
Journal of Research 57 (2), 149-155.
Kazuo yano, Toshiaki Yamanaka, takashi Nishida, Masayoshi Satto,
Katsuhiro Shimohigashi, and Akihiro Shimizu "A 3.8-ns CMOS 16
x 16-b Multiplier Using Complementary Pass-Transistor Logic"
IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 25, NO. 2,
APRIL 1990.
4408
PJAEE, 17 (9) (2020)
4409
PJAEE, 17 (9) (2020)
4410