A Novel Multi-Threshold CMOS Based 64-Bit Adder Design in 45nm CMOS Technology for Low Power Application Rumi Rastogi and Sujata Pandey Amity University Uttar Pradesh, Noida, India Email: rumi.ravi@gmail.com, spandey @amity.edu Abstract -- A new 64-bit low power MTCMOS XOR/XNOR based adder has been proposed in this paper. The adder circuits have been designed in 45 nm Multi-threshold CMOS (MTCMOS) Technology. Due to the high-vt sleep transistor added, the leakage power of the circuit is minimized significantly (75%) at room temperature. The leakage power of the circuit is also evaluated at higher temperatures. The proposed adder effectively suppresses the leakage power even at elevated temperatures. Simulation results verify that the circuits operate with high speed due to the low-threshold voltage transistors used in the evaluation block. The proposed adder also shows a significant reduction in Silicon area due to low transistor count. The ground bouncing noise of the MTCMOS adder is also evaluated. It is shown that the ground bouncing noise of the circuit reduces with increase in sleep signal rise delay. Keywords -- MTCMOS, Ground-gating, sleep transistor, Virtual-ground, Circuit Awakening. I. INTRODUCTION CMOS Technology is widely used in all the digital circuits due to low power consumption. Although dynamic power is the major source of power dissipation, but in deepsubmicron technology, leakage power is also quite dominant. Sub-threshold current is the major source of leakage power, especially in deep-submicron technologies [1]-[2]. As the technology is shrinking, the supply voltages are also being scaled down to minimize dynamic power consumption. With reducing supply voltages, the threshold voltage (V T ) of the transistors also needs to be scaled, to minimize the delay of the circuit. Due to reduction in V T, the sub-threshold current increases exponentially, typically when the circuit is in idle mode and the transistors are turned off. This increases the power dissipation and reduces the noise immunity of the circuits. In recent deep-submicron technologies leakage power contributes significantly to the total power of the circuit. Devices like laptops, memories have long idle periods and hence leakage power minimization is a major issue for VLSI designers. Several techniques like input vector control [3], Dual-V T technique [4, 5] have been proposed to minimize leakage power. Dual-V T circuits employ low-v T transistors in the critical path to minimize delay, whereas the high-v T transistors were employed in the non-critical path to minimize subthreshold leakage current. Multi- V T or MTCMOS (Multiple threshold CMOS) is the most effective and widely used technique in deep-submicron digital circuits [6, 7] to minimize sub-threshold leakage current when the circuit is in idle mode. In this technique, a high-threshold voltage (high-v T) transistor is inserted either as header or footer between the logic circuit (low-v T) and the supply or ground rail. The sleep transistor is ON in normal mode and the circuit operates at high speed. The sleep transistor is switched off when the circuit is in idle mode. This tends to minimize leakage power significantly with a minimal delay and area overhead, due to the extra sleep transistor added. However, when the circuit transitions from sleep mode to active mode, large currents flow through the sleep transistor which leads to disturbances (noise) on the power and ground network. This noise needs to be minimized. In this paper low power 64-bit MTCMOS adder circuit is proposed, the basic adder cell being XOR/XNOR based. The proposed adder is power efficient as compared to other adders due to low transistor count and reduced short circuit currents. The adder also exhibits low leakage power due to ground gating. The ground bouncing noise of the MTCMOS adder is also evaluated in thin this paper. The paper is organized as follows. In section 2, the performance characteristics of the XOR/XNOR based full-adder [8] and proposed 64-bit adder in MTCMOS technology are discussed. Section 3 addresses the simulation results and comparisons for the proposed adder. Section 4 concludes the paper. II. SYSTEM DESIGN XOR/XNOR Based Full-Adder Keeping in view the increased demand for low power designs, several adder topologies have been proposed in literature. The power dissipation of a CMOS circuit includes three components given by P total = P dynamic + P short-circuit + P leakage (1) DOI 10.5013/IJSSST.a.17.41.14 14.1 ISSN: 1473-804x online, 1473-8031 print
where P dynamic = αcv 2 f P short-circuit = I sc. V P leakage = I leakage. V and α is the activity coefficient, C is the total capacitance associated with the circuit, f is the frequency of operation, I sc is the short-circuit current, I leak is the leakage current associated with the circuit and V is the supply voltage. The conventional static CMOS full adder [9] has been widely reported in literature due to its robust design. But it consumes significant power and has high transistor count. Different power efficient adders designed with body biasing techniques have also been reported in [10]. XOR/XNOR based adders [11, 12 and13] with reduced transistor count, low power dissipation and faster operation are of great interest in VLSI and can be used for cascading and designing bigger blocks. Full adder circuits can be implemented with different combinations of XOR/XNOR modules and multiplexers [13]. The full adder circuit proposed in [8] has been implemented with two XNOR gates and one multiplexer. Here the sum is generated with two XNOR gates and Carry is generated with a XNOR gate and a multiplexer as shown in figure 1. The adder uses only 8 transistors and hence achieves great reduction in Silicon area, power and delay. The full adder has been implemented using the XOR/XNOR logic and the equations governing the circuit are given by: Sum= H XOR Cin = H*Cin' + H'*Cin (2) Carry_out = A*H'+ Cin*. H' (3) where H = (A XOR B ) and H' is denoted as complement of H. The circuit consumes less active power than the static adder(28t) [9], Transmission gate adder(tga) [14], Static Energy Recovery adder(serf) [15] and complementary and level restoring carry logic (CLRCL) adder [16] due to less transistor count and hence lesser parasitic capacitances. The short circuit currents in the circuit are also minimized as the direct path between supply and ground is eliminated. Although the circuit is power efficient, but the issues of leakage/static power have not been taken up. XOR/XNOR Adder In MTCMOS Technology The 8T full-adder proposed in [8] is power efficient due to low transistor count and lesser parasitics. Also the short circuit currents are minimized as no direct path between supply and ground exists. But the static power dissipation of the circuit is not discussed in [8]. In deep-submicron technologies, leakage power minimization is crucial issue. The leakage power is mainly caused due to sub-threshold leakage currents which flow in the circuit when the transistors are in OFF state. The sub-threshold current of a MOS transistor is given by [17]: (4) (a) where (5) I ds is the drain current, µ is the mobility, W and L are the channel width and length V GS is the gate to source voltage, V TH is the threshold voltage, V T is the thermal voltage, C d is the depletion layer capacitance, C ox is the oxide capacitance. The sub-threshold current varies exponentially with gate to source voltage. The sub-threshold slope (S) is given by: S = (d(logi DS )/dv GS ) (6) (b) Figure 1(a) Adder Block diagram using two XNOR gates and a multiplexer (b) Circuit diagram of the 8T adder The sub-threshold slope shows to what extent the transistor can be switched off below the threshold voltage V T. As the technology is shrinking, it is difficult to completely switch-off the transistors below V T. This leads to increase in sub-threshold leakage. In this paper, Multi-threshold voltage ground gating technique (MTCMOS technique) has been applied to the 8T adder to minimize the sub-threshold leakage current and hence the static power dissipation of the circuit is also DOI 10.5013/IJSSST.a.17.41.14 14.2 ISSN: 1473-804x online, 1473-8031 print
minimized which is not taken up by the existing adders utilizing XOR/XNOR technique. The proposed MTCMOS adder cell dissipates very low power in normal mode due to low short circuit currents. When the circuit is idle or in sleep mode, the static or leakage power dissipation is also minimized. A 64 bit adder has been designed in 45 nm MTCMOS technology shown in Figure 2. Figure 2. Proposed 64-Bit MTCMOS Adder Block. A High VT sleep transistor is added for grounded gate A high-v T sleep transistor (NMOS) is added as a footer to the circuit. The circuit functions in two modes: Normal mode and Sleep or stand-by mode. In normal mode, the sleep transistor is ON and the circuit functions normally with low VT transistors used in the evaluation block to minimize delay. In the sleep mode i.e. idle mode the sleep switch is OFF, and the path between supply and ground rail is cut off which minimizes the sub-threshold current and hence the leakage power. The sum and the carry evaluation blocks have been designed with low-vt transistors so as to enhance the speed of the circuit. A high-vt sleep transistor is connected as footer between the logic circuit and ground rail. During normal mode, the signal connected to the sleep transistor is high and the sleep switch is on. At this time the circuit operates normally with less delay due to the low-vt logic block. During the sleep or idle mode, the sleep transistor is off and the low-vt logic block is cut-off from ground rail. Hence the sub-threshold leakage current is minimized. MTCMOS proves to be an efficient technique for leakage power minimization but a major issue associated with this technique is the noise generated [18], when the sleep transistor is switched on. The noise is produced due to high currents that flow through the sleep transistor. There are several techniques to minimize this noise. The sleep transistor can be sized to minimize the on-current. However sizing the sleep transistor increases the delay of the circuit in active mode. The noise can also be reduced by slightly increasing the rise delay of the signal applied to the sleep transistor. The noise associated with the proposed MTCMOS adder has been evaluated in the paper for different sleep signal rise delays. It is shown that as the sleep signal rise delay increases, the noise is reduced. This is because if the sleep signal rises slowly, the sleep transistor turns on gradually and the current through the sleep transistor flows in a broader time frame. III. RESULTS AND DISCUSSION 45nm Multi-threshold CMOS technology (High_V T_NMOS=395mV,Low_V T_NMOS= 219mV, High_V T_PMOS = -318mV, Low_V T_PMOS= -283mV) has been used in this paper for evaluating the adder circuits. A 64 bit MTCMOS adder is designed, the basic cell being the XOR/XNOR based 8T adder[8]. The simulations have been carried out using Cadence Virtuoso with supply voltage of 1V. For XNOR, Width of NMOS transistors N1 and N2 has been taken as 500nm and 120nm respectively. Width for PMOS transistor P1 is taken as 200nm. For designing the multiplexer section, typical values of width for NMOS and PMOS are taken as 120nm & 200nm respectively. The sleep transistor width is chosen as 500nm.The delay is computed at 50% threshold, i.e. when the signals reach 50% value of the supply voltage. The 8T XOR/XNOR based adder cell has been designed and simulated in 45 nm CMOS Technology. The simulated results have been shown in Figure 3. Figure 3. Waveforms for the 8T adder designed and simulated in 45 nm technology. A. Performance of the 8T Full-Adder The adder circuit is also compared with other adders and shows a significant reduction in average as well as leakage power. The adder also has a low transistor count and reduced parasitic capacitance. The conventional static CMOS full adder cell (28 transistors) [9] is most widely reported in literature due to its robust design. However, the circuit shows significant power and delay due to large number of transistors. The present work is also compared with various low power DOI 10.5013/IJSSST.a.17.41.14 14.3 ISSN: 1473-804x online, 1473-8031 print
adders like the Transmission gate adder (20T TGA) [14], Static energy recovery Full Adder (SERF 10T)[15], CLRCL (complementary and level restoring carry logic) adder[16]. The SERF adder does not contain a direct path to the ground and hence the short circuit currents are minimized. The load charge is reapplied to the control gate (energy recovery) which further minimizes power. Another low power adder denoted CLRCL, again a 10 transistor adder has, higher computing speed and lower energy consumption. All the adders reported in [9, 14, 15 and 16] have been designed and simulated in 45nm technology. The 8T adder considered in this paper shows significant reduction in power and delay as shown in Table I and II. CMOS adders. The leakage power is reduced significantly in the standby mode as compared to the single low-v T CMOS circuits. The effect of temperature on the leakage power of the circuit is also evaluated in the paper. The study of sub-threshold leakage characteristics with temperature is important as the VLSI circuits normally operate at higher temperature due to power dissipation. The leakage power of the circuit is evaluated at 27, 70 and 110. With increase in temperature the sub-threshold current in the circuit increases and hence the leakage power consumption of the circuit is also increased as shown in Figure 5. But with the ground gating technique, it is effectively suppressed even at higher temperature. TABLE I: POWER COMPARISON OF CMOS ADDERS Adder Type (at 45 nm) Average power dissipation (nw) Leakage Power (pw) Static adder (28T) [9] 91.81 166 TGA 20T [14] 87.76 156.32 SERF10T[15] 76.23 145.52 CLRCL [16] 73.86 141.39 8T Adder [8] 64.75 101.54 TABLE II : DELAY COMPARISON OF SINGLE-BIT CMOS ADDERS Adder Type (at 45nm) Delay (ns) TGA 20T[14] 0.426 SERF 10T[15] 0.358 CLRCL 10T[16] 0.293 8 T Adder[8] 0.237 Static Adder 28T[9] 0.195 The power delay product of the 8T adder is also lowest as compared to other adders as shown in Figure 4. Figure 5: Variation of leakage power with temperature for the 64 bit adders. The proposed MTCMOS adder shows significant reduction in leakage power. The average power, leakage power and delay of the 64 bit MTCMOS and CMOS adder has been compared in Table III. TABLE III: COMPARISON OF 64-BIT CMOS AND MTCMOS ADDER 64-Bit Adder Leakage Power (nw) Average Power (µw) Delay (ns) CMOS 56.38 78.35 26.11 MTCMOS 14.29 75.28 31.46 Although the MTCMOS adder shows significant reduction in leakage and average power, the delay of the circuit is increased slightly due to the parasitic capacitances associated with the sleep transistor added. The propagation delay of a CMOS circuit without sleep transistor is given by [19] T pd (C L -V DD )/(V DD -V T ) α (7) Figure 4: Comparison of power delay product of 8T adder with other adders B. Power and Delay Analysis of the Proposed 64-bit MTCMOS Adder The proposed MTCMOS adders show a significant reduction in active and leakage power as compared to the where C L is the load capacitance, V DD is the supply voltage, V T is the threshold voltage of the low-v T logic block and α is velocity saturation index. When a sleep transistor is inserted the propagation delay increases slightly and can be expressed as T pd (C L -V DD )/(V DD -V X -V T ) α (8) where V x is the voltage of the virtual ground node. DOI 10.5013/IJSSST.a.17.41.14 14.4 ISSN: 1473-804x online, 1473-8031 print
C. Noise Associated with Circuit Awakening A major issue associated with the MTCMOS technique is the noise produced on the ground network when the circuit awakens from standby mode. The noise of the proposed MTCMOS adder circuit has also been characterized in the paper. The noise produced on the ground network and the virtual ground voltage have been plotted with the sleep signal rise delay in Figure 6. 1.25 Sleep Signal Voltage (V) Real Ground Voltage ( µ V) Virtual Ground Voltage ( m V) 1.0.75.5.25.25 75 50 25 25 50 75 1000 75 50 25 25 5.0 1 Time (ns) 5.0 1 Time (ns) 15.0 2 15.0 2 5.0 1 Time (ns) 15.0 2 Figure 6: Voltage waveforms on the ground and virtual ground lines when the circuit transitions from SLEEP mode to ACTIVE mode. The patterns are plotted for the proposed MTCMOS 64-bit adder. The noise associated circuit with the circuit can be minimized by increasing the rise delay of the sleep signal. When the sleep signal rise time is less, the SLEEP transistor switches on faster and large current flows through it. As the sleep signal rise delay is increased, the current through the sleep transistor flows in a broader time frame. Hence the sleep transistor switches on slowly, which minimizes the disturbances and noise. Figure 7 plots the maximum noise produced with the rise delay of the signal applied to the sleep transistor. Maximum noise produced (µv) 600 500 400 300 200 100 0 0 5 10 15 20 25 30 Sleep signal rise time Delay (ns) Figure 7: Maximum noise produced by the proposed adder versus sleep signal rise delay. IV. CONCLUSION A power efficient 64 bit adder circuit is proposed in 45nm Multi-threshold CMOS technology. The proposed MTCMOS adder shows a significant reduction in leakage power. The leakage power is reduced by 75% in the standby mode as compared to the single low-v T CMOS adder. The ground-gating technique offers significant reduction in leakage even at higher temperatures. The average power dissipation of the circuits is also minimized. However the delay of the circuits is increased slightly due to the parasitics associated with the sleep transistor added. The proposed adder also requires the minimum number of transistors and hence achieves significant reduction in Silicon area. The noise associated with the proposed MTCMOS adder is also evaluated. It is also shown that the noise of the circuit reduces with increase in sleep signal rise delay. REFERENCES [1] Gielen, G. and Dehaene, W. "Analog and digital circuit design in 65 nm CMOS: end of the road? IEEE Proceedings on Design, Automation and Test in Europe, pp. 37 42, (2005). [2] Kursun and E. G. Friedman, Node Voltage Dependent Subthreshold Leakage Current Characteristics of Dynamic Circuits, Proceedings of the IEEE/ACM International Symposium of Quality Electronic Design, pp. 104-109, (2004). [3] Abdollahi, A. ; Fallah, F. ; Pedram, M. Leakage current reduction in CMOS VLSI circuits by input vector control IEEE Transactions on Very Large Scale Integration (VLSI) Systems,Volume:12, Issue:2 pp. 140-154, (2004). [4] S. Yang ; H. Wang ; Zhi-jia Yang Low leakage dynamic circuits with dual threshold voltages and dual gate oxide thickness seventh international conference on ASIC, ASICON-07, pp.70-73, (2007). [5] V.K. Sharma and M. Pattanaik, Process, Voltage and Temperature Variations Aware Low Leakage Approach for Nanoscale CMOS Circuits, Journal of Low Power Electronics, vol. 10, no. 1,pp. 45-52, (2014),. [6] M.W. Alam, M. H. Anis, and M. I. Elmasry, High-Speed Dynamic Logic Styles for Scaled-Down CMOS and MTCMOS Technologies, Proceedings of third IEEE/ACM International Symposium on Low Power Electronics and Design, pp. 145-160, (2000). [7] M. Anis, S. Areibi and M. Elmasry, Dynamic and leakage power reduction in MTCMOS circuits using an automated efficient gate clustering technique, Proceedings of the 39th Design Automation conference, Page(s): 480-485, (2002). DOI 10.5013/IJSSST.a.17.41.14 14.5 ISSN: 1473-804x online, 1473-8031 print
[8] Manoj Kumar, Sandeep K. Arya and Sujata Pandey, Single bit adder design using 8 transistors with novel 3 transistor XNOR gate, International Journal of VLSI design and Communication Systems (VLSICS), Vol. 2, no.4, pp.75-80, (2011). [9] Y. Leblebici, S.M. Kang, CMOS Digital Integrated Circuits, Singapore: Mc Graw Hill, 2nd edition, (1999). [10] Manoj Kumar, Sandeep K. Arya and Sujata Pandey, Low power CMOS full adder design with body biasing approach, Journal of Integrated Circuits and Systems, Brazil, vol. 6, no.1, pp. 47-59 (2011). [11] M. Shams and M. Bayoumi, A novel high-performance CMOS1-bit full adder cell, IEEE Transaction on Circuits Systems II, Analog Digital Signal Processing, vol. 47, no. 5, pp. 478 481, (2000). [12] Jin-Fa Lin, Y.T. Hwang, M.H.Sheu and C.C.Ho, A Novel High- Speed and Energy Efficient10-Transistor Full Adder Design, IEEE Transactions on Circuits and Systems, vol. 54, no. 5 pp.1050-1059 (2007). [13] H. T. Bui, Y. Wang, and Y. Jiang, Design and analysis of lowpower 10-transistor full adders using XOR-XNOR gates, IEEE Transactions Circuits Systems II (2002), Analog Digital Signal Process, vol. 49, no. 1, pp. 25 30, (2002). [14] M. Aguirre-Hernandez, M. Linares-Aranda, CMOS full-adders for energy-efficient arithmetic applications, IEEE Trans. Very Large Scale Integration(VLSI) Syst, 19 (4), pp. 718 721, (2011). [15] R. Shalem, E. John, and L. K. John, A novel low-power energy recovery full adder cell, Proceedings of Great Lakes Symposium on VLSI, pp. 380 383(1999). [16] Jin-Fa-Lin, Yin-Tsung Hwang, Ming-Hwa Sheu, and Cheng-Che Ho, A Novel High-Speed and Energy Efficient 10-Transistor Full Adder Design, IEEE Transaction on Circuits and Systems I (2007), Vol. 54, No. 5, pp. 1050-1059(2007). [17] J. M. Rabaey, Digital Integrated Circuits : Prentice-Hall, ch. 2, pp. 55 56(1996). [18] Hailong Jiao and V. Kursun. "Ground-Bouncing-Noise-Aware Combinational MTCMOS Circuits", IEEE Transactions on Circuits and Systems- I, Regular Papers. Vol. 57, no. 8, pp. 2053-2065, (2010). [19] C. Hwang, C. Kang and M. Pedram, Gate Sizing and Replication to Minimize the Effects of Virtual Ground Parasitic Resistances in MTCMOS Designs Proceedings of the seventh International Symposium on Quality Electronic Design (ISQED 06), pp. 741-746(2006). DOI 10.5013/IJSSST.a.17.41.14 14.6 ISSN: 1473-804x online, 1473-8031 print