Professional Documents
Culture Documents
DICD Final Paper Jugi Edited 4
DICD Final Paper Jugi Edited 4
their performance.
Jugi Rithanya R P1 Sakthivel R3
School of Electronics Pooja K2 School of Electronics
School of Electronics
Engineering Engineering
VIT University Engineering VIT University
VIT University
Vellore, India Vellore, India
rpjugi@gmail.com Vellore, India
Corresponding Author
rsakthivel@vit.ac.in
poojakeyan20@gmail.com
Abstract: A resistance with memory is what is known and Olumodeji, O. A el al., For neuromorphic system, it is used
defined as memristor. The incidence of the two-terminal as adaptable resistance for pulse-based operation. It has its
device was found in the year 1971. Memristor is a non-volatile, own ability to store data [7]. It can and is used in remote
expandable, high speed, high packing density, and low-power sensing scheme and also in microprocessor, unconventional
device. It will hold/retain the data, even when the power goes computing. Memristor can be used as crossbar-latches and
off. It is used in various applications like programmable logic, can act as a replacement for transistors [8] and in future it
remote sensing, neuromorphic system, and low power can be implemented in heterogeneous systems [9].
applications. Random Access Memory (RAM) executes
search operation by means of memory address and returns
the data from the location stored. Content addressable
memory (CAM) return the location of searched data by
associating it with the kept data. CAM is a superior category
of memory, which is used by searching applications, like
caches in microprocessors, and networking. CAM cells makes
parallel examination over all stowed data and finds out the
matched data in one cycle. In this paper, we make a relative
study on the CAM cells and the consequence of
implementation of Memristor in CAM cells and analyze the
performance of the design. We find that 1T-1M gives an
improvement of 99% when compared to other CAM models
discussed.
Keywords- CAM, Binary CAM, Memristor, Write and
Read, NAND, NOR.
1. INTRODUCTION
In 1971, Leon 0. Chua discovered a new two terminal
device called memristor. Memristor is also presented as a
fourth basic circuit element after capacitor, resistor and Figure 1 Relation between basic components [17]
inductor. Memristor is well-defined as a resistor which can Amamath, N el al., 10-T NOR Type CAM trails a 6-T
retain memory. Figure 1 signifies the association between SRAM structure with 4 additional transistors which are used
the basic components and memristor. It is characterized by
the association between charge and magnetic flux [1]. to compare the stored data and the data that is searched for
Subsequently thirty-six years later, HP Labs created a in the cam cell [10]. The charge division problem and short
physical memristor with assistance from the metal titanium circuit current problem in NAND and NOR CAM restrict
di-oxide. The device designed is of nanoscale size. the power operation, pre-charge free CAM eliminates these
Memristor has two regions: doped (TiO2-x) and undoped drawbacks [11]. The foremost goal of pre-charge free CAM
(TiO2) region. is to decrease power consumption and the speed for search
The resistance in undoped region is considered as application. Furthermore, higher number of searches can be
𝑅𝑜𝑓𝑓 and for doped it is taken as 𝑅𝑜𝑛 . When the potential is achieved in lesser time. Software based search operation are
applied, the memristor holds the value. When the potential slower than the hardware based CAM [12]. Conventional
is dropped to zero, no current flows through this physical NAND and NOR based CAM have the disadvantages of
device, but it holds the value, i.e., memristor retains the data poorer speed and high-power consumption. Bahloul el al.,
even when the power goes off. Memristor is analysed with
MCAM is considered because of its advanced speed and
different models namely, Non-Linear ion drift model,
Linear ion drift model, VTEAM model, Simmons tunnel low power utilisation [13]. Telajala venkata mahendra el
barrier model, and TEAM model. [2]. VTEAM model al., The hybrid architecture of three transistors and two
defines about the voltage controlled memristors whereas, memristors are used in 3T2M SRAM cell. These SRAM
TEAM model deliberates on current controlled memristors cells are used to build CAM architecture. [14]. NAND type
[3]. Memristors are used in synaptic plasticity in the biology and NOR type are very useful due to large searching
neuronal system [4] and also in bio-inspired pattern performance. NOR CAM are much faster than NAND
processing [5]. Memristors can be used in stateful logic
function and so it's used in analog filters. Memristors are CAM, but the downfall is that it consumes higher power
widely used in digital memory and logical circuits [6]. when compare to NAND type [15]. The advantage of the
NAND CAM is that it does not require more transistor The device will remain in this state until a negative
compared to NOR CAM cell. Beside 10T NAND CAM voltage is applied. Now, by applying negative voltage, the
provides complete read and write operation [16]. Yatagiri, device resistance is increased which corresponds to the off
state of the device i.e., stored data will be removed from the
N. H el al., The drawback in NAND CAM is that the
memory.
searching delay is high and charge problem that will occur
at pass transistor. The 16 cross 16 Content addressable Content addressable memory (CAM) gives back the
memory arrays by means of pre-charge free reason and location of the data searched by comparing it with the stored
data. CAM is a superior type of memory, which is utilised
better-quality project to sense match data are assembled by in exploratory applications, like caches on microprocessors,
means of the FinFET 18nm tech. In association to and computer networking. CAMs can be binary or ternary.
traditional CAM, the Precharge free CAMs provides better In binary Cam, it can keep either logic-0 or 1. In ternary
performance, decreases power and energy consumptions CAM, it can keep logic-0,1 or a don’t-care. Binary CAM is
significantly[18]. Sadiq el al., the assistances of these used typically in cache operation, to store data and in many
memory cells are used in Branch Target Buffer, onchip search operations. Ternary CAM is mostly used in
caches, and Translation Lookaside Buffer [19]. Rouhi, S el application tasks. CAM makes concurrent search over all
stowed words and figures out the matched data in one clock
al., this paper provides architecture of memristor-based
cycle. Figure 2.2 represents the block diagram of the CAM
CAMs with substantial optimization with respect to the cell operation.
physical layout area, power consumption, and ease of
performance[20].
Figure 2.1 I-V Characteristics of Memristor Figure 2.3 Circuit of 10-T NAND CAM Cell [3]
2.2 10-T NOR CAM Cell
Figure 2.4 represents the circuit diagram for 10-T NOR
CAM cell. 10-T NOR consists of Word-line (𝑊𝐿),Bit
line (𝐵𝐿), Bit-line bar(𝐵𝐿 − 𝑏𝑎𝑟),D,D-bar, Search-line,
Search-line bar, Read-word line bar and Match line (ML)
respectively. Word line and Word line bar are used to
choose the operation. i.e. write and read operation. Bit-line
and Bit-line bar is used to write the data. Search-line and
Search-line bar is used to search for the stored data in the
cam cell. D and D-bar holds the written data and it is very
useful during search operation. Read-word line bar works
exactly like word-line operation. Match line plays a major
role where the output of the match line determines whether
the cell has stored the data the we searched for or not.
the CAM cells. Table 2 represents the comparison of [2] Kvatinsky, S., Ramadan, M., Friedman, E. G., & Kolodny, A. (2015).
VTEAM: A general model for voltage-controlled memristors. IEEE
average processing time taken for each CAM model i.e. 10- Transactions on Circuits and Systems II: Express Briefs, 62(8), 786-790
T NAND, Pre-charge Free CAM,1T-1M, 10-T NOR ,3T- [3] Eshraghian, K., Cho, K. R., Kavehei, O., Kang, S. K., Abbott, D., & Kang, S.
2M and 1T-1M CAM respectively. We also estimate the M. S. (2010). Memristor MOS content addressable memory (MCAM): Hybrid
architecture for future high-performance search engines. IEEE Transactions
average power consumption for the above-mentioned on Very Large-Scale Integration (VLSI) Systems, 19(8), 1407-1417.
CAM models. [4] Mandal, S., & Saha, A. (2016, October). Memristors act as synapses in
neuromorphic architectures. In 2016 International Conference on
Table 2 Comparative Analysis for Different CAM cell Communication and Electronics Systems (ICCES) (pp. 1-5). IEEE.
designs [5] Truong, S. N., Van Pham, K., Yang, W., & Min, K. S. (2016, October).
Memristor circuits and systems for future computing and bio-inspired
CAM Average Power Improve information processing. In 2016 IEEE Biomedical Circuits and Systems
Models Processing Dissipation ment in Conference (BioCAS) (pp. 456-459). IEEE.
Time (n𝑊) Time [6] Mokhtar, S. M. A. B., & Abdullah, W. F. H. (2014, April). Re-model
fabricated memristor behavior in LT-SPICE and applied in logic circuit.
w.r.t 10-
In 2014 IEEE Symposium on Computer Applications and Industrial
T NAND Electronics (ISCAIE) (pp. 106-110). IEEE.
CAM
[7] Olumodeji, O. A., & Gottardi, M. (2017, May). A pulse-based memristor
10-T 50.59 𝑛𝑠 16420 𝑛𝑊 - programming circuit. In 2017 IEEE International Symposium on Circuits and
NAND Systems (ISCAS) (pp. 1-4). IEEE.
[3] [8] Chabi, D., Wang, Z., Bennett, C., Klein, J. O., & Zhao, W. (2015). Ultrahigh
density memristor neural crossbar for on-chip supervised learning. IEEE
Pre- 50.06 𝑛𝑠 104.6 𝑛𝑊 1.24 %
Transactions on Nanotechnology, 14(6), 954-962.
charge
[9] Li, H., Liu, B., Liu, X., Mao, M., Chen, Y., Wu, Q., & Qiu, Q. (2015, May).
Free [11] The applications of memristor devices in next-generation cortical processor
10-T 49.96 𝑛𝑠 42180 n𝑊 1.04 % designs. In 2015 IEEE International Symposium on Circuits and Systems
NOR [3] (ISCAS) (pp. 17-20). IEEE.
1-T 1-M 𝟐. 𝟎𝟗𝒆 − 𝟓𝒏𝒔 𝟖. 𝟖𝟕𝟖𝐞 − 𝟑 𝒏𝑾 𝟗𝟗. 𝟗 % [10] Amamath, N., Vijay, H. M., Satyanarayana, S. V. V., Ramakrishnan, N. V., &
Sriadibhatla, S. (2018, March). SEU sensitivity analysis of low power,
1-T 19.59 𝑛𝑠 1.93 𝑛𝑊 61.276 % precharge-free modified CAM cell. In 2018 4th International Conference on
2-M [13] Devices, Circuits and Systems (ICDCS) (pp. 101-105). IEEE.
3-T 39.12 𝑛𝑠 6.7 𝑛𝑊 22.67 % [11] Zackriya, M., &Kittur, H. M. (2016). Precharge-free, low-power content-
2-M [12] addressable memory. IEEE Transactions on Very Large Scale Integration
(VLSI) Systems, 24(8), 2614-2621.
4. CONCLUSION [12] Sarwar, S. S., Saqueb, S. A. N., Quaiyum, F., & Rashid, A. H. U. (2013).
Memristor-based nonvolatile random access memory: Hybrid architecture for
We have compared the various circuits of 10T NAND Cam, low power compact memory design. IEEE Access, 1, 29-34.
10T NOR cam, pre-charge free cam cell, 1T-1M cam cell, [13] Bahloul, M. A., Naous, R., & Masmoudi, M. (2017, March). Hardware
emulation of memristor based ternary content addressable memory. In 2017
1T-2M cam cell and 3T-2M cam cell using cadence 14th International Multi-Conference on Systems, Signals & Devices
virtuoso. And from the 2nd Table which explains the (SSD) (pp. 446-449). IEEE.
comparative analysis on different CAM Cell designs. From [14] Telajala venkata mahendra, sandeep misha,” Self-Controlled High-
the table, we can observe that 10-T NAND consumes less Performance Precharge-Free Content-Addressable Memory”, IEEE
Transactions on Very Large-Scale Integration (VLSI) Systems Volume:
power when compared to 10-T NOR, but it is bit slow when 25, Issue:8, Aug. 2017
compared to 10-T NOR. Meanwhile, Pre-charge free CAM [15] Chi-chum hung, jun-han wu, “A self –disable sense technique with differential
cell consumes less power when compared to 10-T NAND NAND cell for CAM”.2008 15TH International Conference on Devices,
circuits and system.
and 10-T NOR respectively. We can clearly say that
[16] Yang, Y., Mathew, J., Chakraborty, R. S., Ottavi, M., & Pradhan, D. K.
memristor based CAM cells provides better performance (2016). Low cost memristor associative memory design for full and partial
when compared to other non-memristor CAM models and matching applications. IEEE Transactions on Nanotechnology, 15(3), 527-
538.
1T-1M CAM cell stands out with very less power
[17] Kavehei, O., Iqbal, A., Kim, Y. S., Eshraghian, K., Al-Sarawi, S. F., & Abbott,
consumption and faster processing operation. When we D. (2010). The fourth element: characteristics, modelling and electromagnetic
consider memristor based CAM Models we can observe theory of the memristor. Proceedings of the Royal Society A: Mathematical,
that as number of transistors increases, power consumption Physical and Engineering Sciences, 466(2120), 2175-2202.
[18] Yatagiri, N. H., & Sowmya, K. B. (2019, June). Low Power Self-Controlled
also increases and speed decreases. This proves that
Pre-Charge Free Content Addressable Memory. In 2019 3rd International
number of transistors plays a major role in power conference on Electronics, Communication and Aerospace Technology
consumption and performance of the design. From the (ICECA) (pp. 1225-1229). IEEE.
analysis, we can conclude that 1T-1M CAM model [19] Sadiq, Z., & Hasan, S. (2021). MemCAM: A Hybrid Memristor-CMOS CAM
Cell for On-Chip Caches. IEEE Access, 9, 21296-21305.
provides an improvement of 99.99% in the average time
[20] Rouhi, S., Mirzakuchaki, S., & Zeyrani, M. (2020, September). Optimized cell
taken for one complete operation and consumes very less structures for memristor-based content addressable memories. In 2020
power when compared to other proposed models. International Conference on Electrical Engineering (ICEE) (pp. 1-6). IEEE.