Capacitance Scaling Based Energy Efficient Internet of Things (IoTs) Enable CAM Design on FPGA

Capacitance Scaling Based Energy Efficient Internet of Things (IoTs) Enable CAM Design on FPGA

Tarandeep Kaur, Sunny Singh, Bishwajeet Pandey

Dept. Of Computer Science Engineering Chitkara University, Punjab Campus Chandigarh, India

AbstractThis paper deals with Capacitance scaling based

Energy Efficient Content Addressable Memory. We have analyzed IO power reduction by scaling down the capacitance from 4000pF to 50pF with intermediate value of 2000pF, 1000pF, 500pF at different frequencies i.e. 1 GHz, 2.4 GHz, 3.6 GHz, 4.9 GHz, 5 GHz, 5.9 GHz, 60 GHz bands of WLAN channels. We have tested the compatibility of our device with wireless network by operating it on different frequency ranges of WLAN Channel. There is 79.57% reduction in IO power, when we scale down capacitance from 4000pF to 50 pF at 1 GHz and 93.33% reduction in IO power, when we scale down capacitance from 4000pF to 50 pF at 60 GHz. In this work, we are using 28nm technology based Kintex-7 FPGA and Verilog Hardware Description Language. Our CAM is Internet of Things (IoTs) enabled in which we are using 128 bit IP address of Internet Protocol Version 6 (IPv6). With the help of this IP address, connectivity to the different objects from this CAM can be provided.

KeywordsCapacitance, Energy Efficient, Internet of Things, Content Addressable Memory.

1. INTRODUCTION

   Core dynamic power is not dependent on output load capacitance [1]. IO power and static power is dependent on output load capacitance [1]. There is 99.72% reduction in IOs power consumption of Universal Asynchronous Receiver Transmitter (UART) when output load is scaled from 10,000pf to 5pF in IOB setting of FPGA [1].

   Figure 1: Shunt Capacitance for Reduction in Impedance [1]

   Fig.1 shows the shunt capacitance which reduces the

   impedance that ultimately reduces overall power consumption of our device [2]. Memory mainly refers to the physical device used to store the sequence of instructions (program), data permanently or temporarily. Memory devices usually store and retrieve data with the use of address at specific memory location.[3-15] Different types of memories used are Random Access Memory (ROM), Content Addressable Memory (CAM) etc. The only difference in these type of memories is the way in which the data is fetched. In case of RAM the user supplies the memory address and in return data word [8] stored at that memory address is given to the user. On the other hand CAM is designed to make it easy for the user to search data by simply searching the data through the content [12]. In CAM user supplies the data word that is related to the data to be searched and CAM tries to search its entire memory to find the presence of data anywhere in its memory. On the result of this search , list of storage addresses are given where the data word was present. As compared to the other memory devices CAM is fast in all the searching application.[4] due to which CAM is used for High Speed search operation [13]. CAM involves a large amount of circuitry due to which it is more power consumptive [12]. So we use CAM only at those places where the slow speed is not tolerable. Two types of CAM mainly used are Binary CAM, ternary CAM [4]. Binary CAM uses data search words containing 0s and 1s, on the other hand Ternary CAM uses a third state i.e 'Don't Care' or 'X' [9] which making our searching process more effective. For e.g. Ternary CAM [10] can have a stored word "X010X" which will match any of these words "10101", "10100", "00101", "00100". It increases our option for search by making it more flexible but also increases the cost of it .CAM that we are going to design is Capacitance based which means the ability of an object to store the electric charge [3]. Any object (body) that can be electrically charged exhibits capacitance. The output load is the sum of capacitance of pin and capacitance of device. The unit of capacitance is Farad

   but we have used pico Farad (pF) which is one trillionth (10-

   12) of a farad [4]. Power that is directly proportional to capacitance reduces when capacitance is reduced i.e when we reduce capacitance, and then there is reduction in IO power, leakage power. In the cases below there is no significant difference in the leakage power but in IO power and the total. Energy is being saved through this

   device which varies from 79.57% to 93.33% by using different frequencies bands (1 GHz, 2.4 GHz, 3.6 GHz, 4.9

   GHz, 5 GHz, 5.9 GHz, and 60 GHz).

   Fig. 2: Top Level Schematic of CAM

   In this research we have used 9 register, 9 Flip-Flops, 507 multiplexers out of which 254 are 1-bit 2-to-1 multiplexer and 253 8-bit 2-to-1 multiplexer. Other components are

   599 basic elements (BELS) which contains 1 inverter (INV), 3 LUT2 (Look Up Table), 69 LUT3, 65 LUT4, 224 LUT5, 237 LUT6, 9 Flip-flops/Latches (D Flip Flop with reset), 1 clock buffer (BUFGP), 266 IO Buffers out of which there are 357 Input Buffer (IBUF) and 9 output buffer (OBUF). Our CAM is Internet of Things (IoTs) [5] enabled in which we are using 128 bit IP address of Internet Protocol Version 6 (IPv6) [3]. With the help of this, connectivity to the different objects from this CAM can be provided. Earlier, Capacitance scaling is used to design energy efficient UART [1], power optimized Register [2], and ALU [15]. Here, we are extending that approach to energy efficient CAM.

2. RESULT

   Unit of capacitance is Farad (F). The output load capacitance is in range of 50pF-4000pF. We are doing power analysis in form of clock power, logic power, signal power, IO power, and leakage power. Unit of power is Watt (W). Power is directly proportional to capacitance and frequency. Therefore, power dissipation will increase with increase in either capacitance or frequency or both.

   1. When CAM is Operating at 1 GHz Frequency.

      C Power	50pF	500pF	1000pF	2000pF	4000pF
      Clock	0.008	0.008	0.008	0.008	0.008
      Logic	0.005	0.005	0.005	0.005	0.005
      Signal	0.024	0.024	0.024	0.024	0.024
      IO	0.029	0.042	0.056	0.085	0.142
      Leakage	0.147	0.147	0.147	0.147	0.148
      Total	0.213	0.226	0.240	0.269	0.326

      C Power	50pF	500pF	1000pF	2000pF	4000pF
      Clock	0.008	0.008	0.008	0.008	0.008
      Logic	0.005	0.005	0.005	0.005	0.005
      Signal	0.024	0.024	0.024	0.024	0.024
      IO	0.029	0.042	0.056	0.085	0.142
      Leakage	0.147	0.147 /td>	0.147	0.147	0.148
      Total	0.213	0.226	0.240	0.269	0.326

      TABLE 1: POWER DISSIPATION AT DIFFERENT CAPACITANCE

      There is 79.57% reduction in IO power, when we scale down output load capacitance from 4000pF to 50pF at 1 GHz as shown in Figure 3 and Table 1.

      Fig. 3: Power Dissipation at 1GHz Frequency and Different Capacitance

   2. When CAM is Operating at 2.4 GHz Frequency.

      TABLE 2: POWER DISSIPATION AT DIFFERENT CAPACITANCE.

      C Power	50pF	500pF	1000pF	2000pF	4000pF
      Clock	0.019	0.019	0.019	0.019	0.019
      Logic	0.011	0.011	0.011	0.011	0.011
      Signal	0.058	0.058	0.058	0.058	0.058
      IO	0.076	0.150	0.232	0.396	0.724
      Leakag e	0.148	0.148	0.149	0.151	0.155
      Total	0.312	0.387	0.470	0.636	0.967

      There is 89.50%, reduction in IO power, when we scale down capacitance from 4000pF to 50pF at 2.4 GHz device operating frequency as shown in Figure 4 and Table 2. The frequency of WLAN Channel 802.11b/g/n is 2.4 GHz.

      Fig. 4: Power Dissipation at 2.4GHz Frequency and Different Capacitance

   3. When CAM is Operating at 3.6 GHz Frequency.

      TABLE 3: POWER DISSIPATION AT DIFFERENT CAPACITANCE.

      C Power	50pF	500pF	1000pF	2000pF	4000pF
      Clock	0.029	0.029	0.029	0.029	0.029
      Logic	0.017	0.017	0.017	0.017	0.017
      Signal	0.087	0.087	0.087	0.087	0.087
      IO	0.118	0.253	0.404	0.706	1.310
      Leakage	0.149	0.150	0.152	0.155	0.162
      Total	0.399	0.536	0.689	0.994	1.605

      There is 90.99%, reduction in IO power, when we scale down capacitance from 4000pF, 2000pF, 1000pF, 500pF, 50pF at 3.6 GHz as shown in Figure 5 and Table 3. The frequency of WLAN Channel 802.11y is 3.6 GHz.

      Fig. 5: Power Dissipation at 3.6GHz Frequency and Different Capacitance

   4. When CAM is Operating at 4.9 GHz Frequency

      TABLE 4: POWER DISSIPATION AT DIFFERENT CAPACITANCE

      C Power	50pF	500pF	1000pF	2000pF	4000pF
      Clock	0.039	0.039	0.039	0.039	0.039
      Logic	0.023	0.023	0.023	0.023	0.023
      Signal	0.119	0.119	0.119	0.119	0.119
      IO	0.162	0.362	0.585	1.029	1.918
      Leakage	0.150	0.152	0.154	0.159	0.170
      Total	0.493	0.695	0.920	1.369	2.269

      There is 91.55%, reduction in leakage power, when we scale down capacitance from 4000pF, 2000pF, 1000pF, 500pF, 50pF at 4.9 GHz as shown in Figure 6 and Table 4.

   5. When CAM is Operating at 5 GHz Frequency

      TABLE 5: POWER DISSIPATION AT DIFFERENT APACITANCE

      C Power	50pF	500pF	1000pF	2000pF	4000pF
      Clock	0.040	0.040	0.040	0.040	0.040
      Logic	0.023	0.023	0.023	0.023	0.023
      Signal	0.121	0.121	0.121	0.121	0.121
      IO	0.166	0.371	0.599	1.055	1.968
      Leakag e	0.150	0.152	0.154	0.159	0.170
      Total	0.500	0.708	0.938	1.400	2.323

      There is 91.56%, reduction in IO power, when we scale down capacitance from 4000pF, 2000pF, 1000pF, 500pF, 50pF at 5 GHz as shown in Figure 7 and Table 5.

      Fig. 7: Power Dissipation at 5GHz Frequency and Different Capacitance

   6. When CAM is Operating at 5.9 GHz Frequency

      TABLE 6: POWER DISSIPATION AT DIFFERENT CAPACITANCE

      C Power	50pF	500pF	1000pF	2000pF	4000pF
      Clock	0.047	0.047	0.047	0.047	0.047
      Logic	0.028	0.028	0.028	0.028	0.028
      Signal	0.143	0.143	0.143	0.143	0.143
      IO	0.197	0.452	0.735	1.302	2.425
      Leakage	0.150	0.153	0.156	0.163	0.176
      Total	0.565	0.823	1.110	1.682	2.829

      There is 91.87%, reduction in leakage power, when we scale down capacitance from 4000pF, 2000pF, 1000pF, 500pF, 50pF at 5.9 GHz as shown in Figure 7 and Table 6.

      Fig. 6: Power Dissipation at 4.9GHz Frequency and Different Capacitance Fig. 8: Power Dissipation at 5.9GHz Frequency and Different capacitance

   7. When CAM is Operating at 60 GHz Frequency

   C Power	50pF	500pF	1000pF	2000pF	4000pF
   Clock	0.482	0.482	0.482	0.482	0.482
   Logic	0.111	0.111	0.111	0.111	0.111
   Signal	1.325	1.325	1.325	1.325	1.325
   IO	2.114	5.486	9.233	16.727	31.715
   Leakage	0.195	<>0.248	0.323	0.539	0.993
   Total	4.227	7.653	11.475	19.184	34.626

   C Power	50pF	500pF	1000pF	2000pF	4000pF
   Clock	0.482	0.482	0.482	0.482	0.482
   Logic	0.111	0.111	0.111	0.111	0.111
   Signal	1.325	1.325	1.325	1.325	1.325
   IO	2.114	5.486	9.233	16.727	31.715
   Leakage	0.195	0.248	0.323	0.539	0.993
   Total	4.227	7.653	11.475	19.184	34.626

   TABLE 7: POWER DISSIPATION AT DIFFERENT CAPACITANCE

   There is 93.33%, reduction in IO power, when we scale down capacitance from 4000pF, 2000pF, 1000pF, 500pF, 50pF at 60 GHz as shown in Figure 8 and Table 7.

   REFERENCES

   1. P. Singh, O. J. Pandey, B. Pandey, T. Das, T. Kumar, Output Load Capacitance Based Low Power Implementation of UART on FPGA,

      IEEE International Conference on Computer Communication and Informatics (ICCCI) at Coimbatore, Jan 3-5 2014. DOI: 10.1109/ICCCI.2014.6921826

   2. S. Banshal, B. Pandey, S. J. Bendra Capacitance Scaling Aware Power Optimized Register Design And Implementation on 28nm FPGA, IEEE

      International Conference on Computer Communication and Informatics (ICCCI) at Coimbatore, Jan 3-5 2014. DOI: 10.1109/ICCCI.2014.6921838

   3. Capacitance http://en.wikipedia.org/wiki/Capacitance. Accessed on october 2014

   4. Content Addressable memory. http://en.wikipedia.org/wiki/Content- addressable_memory. Accessed on october 2014.

   5. IP_address http://en.wikipedia.org/wiki/IP_address Accessed on

      Fig. 9: Power Dissipation at 60GHz Frequency and Different Capacitance

3. CONCLUSION

   We achieved Capacitance scaling based energy efficient Internet of Things enabled Content addressable memory Design on FPGA. We used capacitance as a main parameter and by operating device on different capacitance and Wireless network channels using IEEE 802.11 protocols at different frequency range from 1 GHz to 60 GHz we have shown how the power is saved. The CAM design we represented is energy efficient as we can see the reduction in IO power i.e 93.33%-79.57% when we scale down capacitance from 4000pF to 2000pF,1000pF, 500pF,50pF. Although there is no significant reduction in Leakage power, signal power, Clock power but there is reduction in IO power and Total power.

4. FUTURE SCOPE

In the near future we will design the Energy efficient Internet of things enabled ALU on FPGA. We will also use Ubiquitous computing in our design which is the part of internet of things. We have used 2D IC in this paper but we will implement the design in future using 3D, 4D IC. We have implemented this CAM using 28 nm FPGA, but there is a large scope of implementing CAM on 14nm, 32nm, 40nm, 90nm and large scale FPGA. This design is energy efficient but we can make high performance CAM that will be internet of things enabled.

october 2014

1. Farad http://en.wikipedia.org/wiki/Farad. Accessed on October 2014

2. Internet of http://en.wikipedia.org/wiki/Internet_of_Things .Accessed on october 2014

3. Data Word,

   http://en.wikipedia.org/wiki/Word_(computer_architecture). Accessed on october 2014

4. A. G. Hanlon, "Content-addressable and associative memory systems", IEEE Trans. Electronic Computers 15.4 (1966): 509-521.

5. K. Pagiamtzis, & A. Sheikholeslami, A low-power content- addressable memory (CAM) using pipelined hierarchical search scheme. Solid-State Circuits, IEEE Journal of, 39(9), 1512-1519, 2004.

6. I. Arsovski, T. Chandler, & A. Sheikholeslami, A. A ternary content- addressable memory (TCAM) based on 4T static storage and including a current-race sensing scheme, IEEE Journal of Solid-State Circuits, 38(1), pp. 155-158, 2003.

7. R. D. Adams, Content Addressable Memories. High Performance

   Memory Testing: Design Principles, Fault Modeling and Self-Test, pp. 67-75.

8. H. Jarollahi, et al. "A low-power content-addressable memory based on clustered-sparse networks." IEEE 24th International Conference on Application-Specific Systems, Architectures and Processors (ASAP), 2013.

9. H. M. Kittur, "Selective Match-Line Energizer Content Addressable Memory (SMLE-CAM)." arXiv preprint arXiv:1406.7662, 2014.

10. Computer Data Storage, Accessed on october 2014. http://en.wikipedia.org/wiki/Computer_data_storage

11. T. Kumar, Sweety, S.M.M. Islam, B. Pandey and T. Das, "Energy Conversion in 64-Bit ALU Design on FPGA Using Mechanics of Capacitance", International Journal of Current Engineering and

Technology (IJCET), ISSN:22774106, 2347-5161(print), Special Issue-3, April 2014.