Ground bounce noise reduction aware combinational multi threshold CMOS circuits for nanoscale CMOS multiplier

Abstract

Multi-threshold complementary metal-oxide-semiconductor (MTCMOS) is often used to reduce the leakage current in idle circuit. Ground bounce noise produced during a transition mode (sleep-to-active) is an important challenge in MTCMOS. In this paper, various noise-aware combinational MTCMOS circuit was used to evaluate the ground bounce noise. An intermediate mode was applied in the sleep-to-active mode transition to reduce the charge stored on virtual lines to real ground. The dependence of ground bounce noise on voltage, transistor size and temperature was investigated with different MTCMOS circuit technique. The peak amplitude of ground bounce noise was reduced up to 78.82%. The leakage current of the circuit was decreased up to 99.73% and the active power of the circuit was reduced up to 62.32%. Simulation of multiplier with different MTCMOS circuit techniques was performed on 45 nm CMOS technology.

Keywords

multi-threshold complementary metal-oxide-semiconductor (MTCMOS) / mode transition / ground bounce noise / sleep transistor

1 Introduction

Compact optical devices are important for dense-wavelength-division-multiplexing optical communications system and optical networks like fiber-to-the-building or fiber-to-the-home in the future [1]. Grating-assisted-couplers, which combine the multi-ports characteristic of directional coupler with the wavelength selectivity of grating, have attracted considerable attention in recent years [2]. Compared with short-period (Bragg)-grating, long-period-grating-assisted-coupler (LPGAC) has several advantages such as easy-fabrication, wide tuning range and much lower insertion loss [3]. By far, most researches of LPGAC have only focused on periodic refractive index modulation. But long-period-grating (LPG) is an optical component that can be used to couple core mode with co-directional propagating cladding mode, whose field distributions are evanescent, which makes it possible for LPG to experience loss or gain induced by external adjust, such as the change of carrier concentration in semiconductor waveguides or pump conditions in rare-earth-doped waveguides [4-6]. In order to comprehensively understand the performance of LPGAC, complex-modulated-grating combined refractive index (real) with loss/gain (imaginary) modulation should be considered. In this paper, by directly deriving from full-scalar wave equations, unified coupled-mode equations and their close-form analytical solutions of the complex LPGAC are obtained. Power coupling characteristics of the complex LPGAC are analyzed and its application in optical signal buffer is discussed in detail.

2 Theoretical model

Figure 1 shows the structure of complex LPGAC, which consists of two parallel but unequal waveguides in close proximity to each other. Because power transferring between asynchronous waveguides is not efficient, a grating is inscribed to facilitate power coupling by reducing the difference of propagation constants between two co-directional modes [7].

Fig.1 Complex long-period-grating-assisted-coupler

Full size|PPT slide

Complex LPG is defined as a periodic longitudinal refractive index modulation (real part) combined with some gain or loss perturbation (imaginary part):

(1)$$\mathrm{\Delta }n=\mathrm{\Delta }{n}_r\cos \left({\displaystyle \frac{\mathrm{2}\mathrm{\pi }}{\mathrm{\Lambda }}}z\right)-\mathrm{i}\mathrm{\Delta }{n}_i\cos \left({\displaystyle \frac{\mathrm{2}\mathrm{\pi }}{\mathrm{\Lambda }}}z+\mathrm{\Delta }\phi \right),$$

where $\mathrm{\Lambda }$ is the grating period. $\mathrm{\Delta }{n}_r$ and $\mathrm{\Delta }{n}_i$ are refractive index and gain ($\mathrm{\Delta }{n}_i>\mathrm{0}$) or loss ($\mathrm{\Delta }{n}_i<\mathrm{0}$) modulation, respectively. $\mathrm{\Delta }\phi$ represents the additional phase difference between the real and the imaginary part.

Based on weak coupling approximation, electric field $E(x,y,z,t)$ can be described as a linear combination of mode fields in isolation [8]:

(2)$$E(x,y,z,t)=[c_{\mathrm{1}}(z)\exp (-\mathrm{i}{\beta }_{\mathrm{1}}z)]·{\psi }_{\mathrm{1}}(x,y)\exp (-\mathrm{i}\omega t)+[c_{\mathrm{2}}(z)\exp (-\mathrm{i}{\beta }_{\mathrm{2}}z)]·{\psi }_{\mathrm{2}}(x,y)\exp (-\mathrm{i}\omega t),$$

where $c_{\mathrm{1}}(z)$, $c_{\mathrm{2}}(z)$ are slowly varying mode amplitudes, ${\psi }_{\mathrm{1}}(x,y)$, ${\psi }_{\mathrm{2}}(x,y)$ are transverse field distributions and ${\beta }_{\mathrm{1}}$, ${\beta }_{\mathrm{2}}$ are propagation constants.

Inserting the complex perturbation (1) and electric field (2) into full-scalar wave equations:

(3)$$[{\nabla }_t^{\mathrm{2}}+k^{\mathrm{2}}{n}^{\mathrm{2}}(x,y)]E(x,y,z,t)=\mu {\displaystyle \frac{{\partial }^{\mathrm{2}}}{\partial t^{\mathrm{2}}}}[P(x,y,z,t)],$$

(4)$$P(x,y,z,t)={ϵ}_{\mathrm{0}}\mathrm{\Delta }{n}^{\mathrm{2}}E(x,y,z,t)=\mathrm{2}{ϵ}_{\mathrm{0}}n\left[\mathrm{\Delta }{n}_r\cos \left({\displaystyle \frac{\mathrm{2}\mathrm{\pi }}{\mathrm{\Lambda }}}z\right)-\mathrm{i}\mathrm{\Delta }{n}_i\cos \left({\displaystyle \frac{\mathrm{2}\mathrm{\pi }}{\mathrm{\Lambda }}}z+\mathrm{\Delta }\phi \right)\right]E(x,y,z,t).$$

Then, multiply Eq. (3) by ${\psi }_{\mathrm{1}}$and integrate over infinite cross section, the evolution of the mode amplitudes can be expressed as the following unified coupled-mode equations:

(5)$${\displaystyle \frac{\mathrm{d}{c}_{\mathrm{1}}}{\mathrm{d}z}}=-\mathrm{i}\left[K_{\mathrm{21}}^r\cos \left({\displaystyle \frac{\mathrm{2}\mathrm{\pi }}{\mathrm{\Lambda }}}z\right)-\mathrm{i}{K}_{\mathrm{21}}^i\cos \left({\displaystyle \frac{\mathrm{2}\mathrm{\pi }}{\mathrm{\Lambda }}}z+\mathrm{\Delta }\phi \right)\right]c_{\mathrm{2}}\exp [-\mathrm{i}({\beta }_{\mathrm{2}}-{\beta }_{\mathrm{1}})z],$$

(6)$${\displaystyle \frac{\mathrm{d}{c}_{\mathrm{2}}}{\mathrm{d}z}}=-\mathrm{i}\left[K_{\mathrm{12}}^r\cos \left({\displaystyle \frac{\mathrm{2}\mathrm{\pi }}{\mathrm{\Lambda }}}z\right)-\mathrm{i}{K}_{\mathrm{12}}^i\cos \left({\displaystyle \frac{\mathrm{2}\mathrm{\pi }}{\mathrm{\Lambda }}}z+\mathrm{\Delta }\phi \right)\right]c_{\mathrm{1}}\exp [-\mathrm{i}({\beta }_{\mathrm{1}}-{\beta }_{\mathrm{2}})z],$$

where $K_{vu}^r$ and $K_{vu}^i$ ($v,u=\mathrm{1},\mathrm{2}$) are defined as

(7)$$K_{vu}^r={\displaystyle \frac{\omega }{\mathrm{4}}}{\displaystyle \underset{\infty }{∯}\left[\mathrm{2}{ϵ}_{\mathrm{0}}n\mathrm{\Delta }{n}_r\cos \left({\displaystyle \frac{\mathrm{2}\mathrm{\pi }}{\mathrm{\Lambda }}}z\right)\right]}·E_v{E}_u^{*}\mathrm{d}S=\mathrm{2}{k}_n\cos \left({\displaystyle \frac{\mathrm{2}\mathrm{\pi }}{\mathrm{\Lambda }}}z\right),$$

(8)$$K_{vu}^i={\displaystyle \frac{\omega }{\mathrm{4}}}{\displaystyle \underset{\infty }{∯}\left[\mathrm{2}{ϵ}_{\mathrm{0}}n\mathrm{\Delta }{n}_i\cos \left({\displaystyle \frac{\mathrm{2}\mathrm{\pi }}{\mathrm{\Lambda }}}z+\mathrm{\Delta }\phi \right)\right]}·E_v{E}_u^{*}\mathrm{d}S=\mathrm{2}{k}_{\alpha }\cos \left({\displaystyle \frac{\mathrm{2}\mathrm{\pi }}{\mathrm{\Lambda }}}z+\mathrm{\Delta }\phi \right),$$

where $k_n$ and $k_{\alpha }$ are the real and imaginary coupling coefficients corresponding to the refractive index modulation or gain/loss perturbation, respectively.

Taking the second differential $c_{\mathrm{1}}^{″}$ of Eq. (5) and eliminating $c_{\mathrm{2}}$, $c_{\mathrm{2}}^{\prime }$ according to Eq. (6):

(9)$${\displaystyle \frac{{\mathrm{d}}^{\mathrm{2}}{c}_{\mathrm{1}}}{\mathrm{d}{z}^{\mathrm{2}}}}+\mathrm{i}\mathrm{2}\delta {\displaystyle \frac{\mathrm{d}{c}_{\mathrm{1}}}{\mathrm{d}z}}+\left\{k_n-k_{\alpha }\exp \left[\mathrm{i}\left({\displaystyle \frac{\mathrm{\pi }}{\mathrm{2}}}-\mathrm{\Delta }\phi \right)\right]\right\}·\left\{k_n+k_{\alpha }\exp \left[\mathrm{i}\left(\mathrm{\Delta }\phi -{\displaystyle \frac{\mathrm{\pi }}{\mathrm{2}}}\right)\right]\right\}=\mathrm{0.}$$

Thus, close-form analytical solution of the coupled-mode equations can be obtained in the matrix form as following:

(10)$$\left[\begin{array}{c}{c}_{\mathrm{1}}\left(z\right)\\ c_{\mathrm{2}}\left(z\right)\end{array}\right]=\left[\begin{array}{cc}\left(\cos \left(\gamma z\right)+\mathrm{i}{\displaystyle \frac{\delta }{\gamma }}\sin \left(\gamma z\right)\right)\exp \left(-\mathrm{i}\delta z\right)& -\mathrm{i}\left({\displaystyle \frac{k_n-k_{\alpha }\exp \left[\mathrm{i}\left(\mathrm{\pi }/\mathrm{2}-\mathrm{\Delta }\varphi \right)\right]}{\gamma }}\sin \left(\gamma z\right)\right)\exp \left(-\mathrm{i}\delta z\right)\\ -\mathrm{i}\left({\displaystyle \frac{k_n+k_{\alpha }\exp \left[\mathrm{i}\left(\mathrm{\Delta }\varphi -\mathrm{\pi }/\mathrm{2}\right)\right]}{\gamma }}\sin \left(\gamma z\right)\right)\exp \left(\mathrm{i}\delta z\right)& \left(\cos \left(\gamma z\right)-\mathrm{i}{\displaystyle \frac{\delta }{\gamma }}\sin \left(\gamma z\right)\right)\exp \left(\mathrm{i}\delta z\right)\end{array}\right]\left[\begin{array}{c}{c}_{\mathrm{1}}\left(0\right)\\ c_{\mathrm{2}}\left(0\right)\end{array}\right].$$

Here, $\gamma =\sqrt{{\delta }^{\mathrm{2}}+k_n^{\mathrm{2}}-k_{\alpha }^{\mathrm{2}}}$ with $\delta =\mathrm{\pi }/\mathrm{\Lambda }-({\beta }_{\mathrm{1}}-{\beta }_{\mathrm{2}})/\mathrm{2}$.

3 Analyses and discussion

First, we assume the phase difference $\mathrm{\Delta }\phi$ is $\mathrm{\pi }/\mathrm{2}$ and the signal is initially injected into core 1 ($c_{\mathrm{1}}(\mathrm{0})=\mathrm{1}$, $c_{\mathrm{2}}(\mathrm{0})=\mathrm{0}$) (as shown in Fig. 1). In order to investigate the influences of gain/loss perturbation on the power coupling characteristic, variations of normalized power ${\left|c_{\mathrm{1}}(z)\right|}^{\mathrm{2}}$ and ${\left|c_{\mathrm{2}}(z)\right|}^{\mathrm{2}}$ with the propagation distance for different relative modulation ratio $k_{\alpha }/k_n$ are calculated and results are depicted in Fig. 2. When both $k_{\alpha }=\mathrm{0}$ and $k_n=\mathrm{0}$, which corresponding to mismatched coupler itself, the power coupling efficiency will be too low to be useful, as shown in Fig. 2(a). This is because the propagation constants between the two waveguides are asynchronous and mismatched, which restrain the coupling process. By introducing the periodic refractive index grating, conventional LPGAC ($k_{\alpha }=\mathrm{0}$ but $k_n\ne \mathrm{0}$) will exhibit complete reciprocal and symmetric power exchange, as can be seen from Fig. 2(b). The initial power injected into one waveguide can be transferred totally to another waveguide by choosing suitable coupling length.

If both refractive index modulation and gain/loss perturbation ($k_n\ne \mathrm{0}$ and $k_{\alpha }\ne \mathrm{0}$) are included, power coupling characteristic of the complex LPGAC is different from the conventional grating-assisted-coupler. As shown in Figs. 2(c)-2(f), although the two waveguides still be coupled alternately and periodically, the symmetry of power transferring can be broken by controlling gain/loss perturbation. Depending on how the perturbation (loss or gain) is induced, the maximum power coupled to cross waveguide is different and coupling length is kept increasing with relative modulation ratio $\left|k_{\alpha }/k_n\right|$. Especially, in the case of $k_{\alpha }/k_n=\pm \mathrm{1}$, the coupling length is extended to infinity, which means the coupling process is nonperiodical and nonreciprocal. When $k_{\alpha }/k_n=-\mathrm{1}$ (as shown in Fig. 2(h)), both waveguides will remain their initial injected power due to the effect of refractive index modulation eliminated totally by the loss perturbation. However, when gain perturbation is introduced and matched with refractive index modulation ($k_{\alpha }/k_n=\mathrm{1}$), the coupled power will grow quadratically with the propagation distance. Simultaneously, the signal in the injected waveguide is unaffected (Fig. 2(g)).

Fig.2 Influence of gain/loss perturbation on power coupling characteristic (solid line: ${\left|c_{\mathrm{1}}(z)\right|}^{\mathrm{2}}$, dotted line: ${\left|c_{\mathrm{2}}(z)\right|}^{\mathrm{2}}$). (a) $k_n=\mathrm{0}$, $k_{\mathrm{a}}=\mathrm{0}$; (b) $k_n\mathrm{\ne }\mathrm{0}$, $k_{\mathrm{a}}=\mathrm{0}$; (c) $k_{\mathrm{a}}/k_n=\mathrm{0.5}$; (d) $k_{\alpha }/k_n=-\mathrm{0.5}$; (e) $k_a/k_n=\mathrm{0.7}$; (f) $k_{\mathrm{a}}/k_n=-\mathrm{0.7}$; (g) $k_{\mathrm{a}}/k_n=\mathrm{1}$; (h) $k_{\mathrm{a}}/k_n=-\mathrm{1}$

Full size|PPT slide

However, if the signal is initially injected from another core 2, the power coupling behavior will be different from above. When loss perturbation is introduced ($k_{\alpha }/k_n=-\mathrm{1}$), the signal can be transferred from core 2 to core 1. But, in the case of gain perturbation ($k_{\alpha }/k_n=\mathrm{1}$), power of the two waveguides are no longer intercoupled, as can be found from Eq. (10). Therefore, the complex LPGAC also demonstrates different coupling characteristic according to different input conditions.

4 Application

Because gain or loss can be interconverted and controlled accurately by external adjusting, such as applying additional stress in the coupler, or injected carrier concentration or optical pumping conditions in rare-earth-doped cores [7], many new optical devices can be exploited based on the nonreciprocal unidirectional coupling characteristic of the complex LPGAC.

By connecting the output port with the input port of waveguide 2, a novel optical signal buffer based on the complex LPGAC and a self-closed loop are proposed. Schematic diagram of the optical buffer is shown in Fig. 3. The grating period is chosen to make the wavelength of 1.55 µm satisfy the phase-matching condition. And the grating strength is $k_n{L}_g=\mathrm{\pi }/\mathrm{2}$ so as to provide complete signals coupling when the pump is turned off. Length of the self-closed loop is chosen to make the group delay of one roundtrip be much longer than the input signals.

Fig.3 Schematic diagram of optical signal buffer based on complex LPGAC and self-closed loop

Full size|PPT slide

Working principle of the optical buffer is described as follows. We assume three Gaussian pulse signals with spectral width much smaller than spectral bandwidth of the complex LPGAC to be injected into the input port of waveguide 1, as shown in Fig. 4(a). By turning on the optical pump to make the gain perturbation matched with the refractive index modulation ($k_{\alpha }/k_n=\mathrm{1}$), the complex LPGAC not only transfers the input signals to the output port of waveguide 2, but also provides an amplification to the signals’ power (as shown in Fig. 2(g)). Then, the signals will be transmitted along the self-closed loop and arrive in the input port of waveguide 2 with some attenuation. Because coupling behavior are checked when signals are injected from waveguide 2 in the case of $k_{\alpha }/k_n=\mathrm{1}$, the signals cannot be transferred back to waveguide 1. Thus, the signals will be circulated and stored in the self-closed loop. Figure 4(b) shows the signals after circulating two roundtrips. When the stored signals need be picked up and exported from the loop, we can turn off the optical pump to make the complex LPGAC become the conventional LPGAC ($k_{\alpha }=\mathrm{0}$ but $k_n\ne \mathrm{0}$). The reciprocal and symmetric power coupling action (as shown in Fig. 2(b)) will transfer the stored signals from input port of waveguide 2 to the output port of waveguide 1, as depicted in Fig. 4(c). Therefore, signals can be buffered or picked up discretionarily by controlling the optical pump.

Fig.4 (a) Pulse signals in input port of waveguide 1; (b) pulse signals buffered in self-closed loop; (c) pulse signals picked up at output port of waveguide 1

Full size|PPT slide

5 Conclusions

In this paper, based on the unified coupled-mode equations, influences of gain/loss perturbation on the performance of complex LPGAC are investigated in detail. Results demonstrate that a unique unidirectional nonreciprocal power coupling can be realized by matching gain/loss with the refractive index modulation. And a novel optical signal buffer combined the complex LPGAC with a self-closed loop is proposed. This compact low-cost device will have many attractive applications in the future all-optical network.

References

[1] Singh H, Agarwal K, Sylvester D, Nowka K J. Enhanced leakage reduction techniques using intermediate strength power gating. IEEE Transactions on Very Large Scale Integration (VLSI). ystems, 2007, 15(11): 1215–1224

[2] Johnson M, Somasekhar D, Chiou L Y, Roy K. Leakage control with efficient use of transistor stacks in single threshold CMOS. IEEE Transactions on Very Large Scale Integration (VLSI). ystems, 2002, 10(1): 1–5

[3] Calimera A, Benini L, Macii A, Macii E, Poncino M. Design of a flexible reactivation cell for safe power-mode transition in power-gated circuits. IEEE Transactions on Circuits and Systems I, Regular Papers, 2009, 56(9): 1979–1993

[4] Mutoh S, Douseki T, Matsuya Y, Aoki T, Shigematsu S, Yamada J. 1-V power supply high-speed digital circuit technology with multithreshold-voltage CMOS. IEEE Journal of Solid-State Circuits, 1995, 30(8): 847–854

[5] Pakbaznia E, Pedram M. Design of a tri-model multi-threshold CMOS switch with application to data retentive power gating. IEEE Transactions on Very Large Scale Integration (VLSI) Systems, 2012, 20(2): 380–385

[6] Shi K J, Howard D. Challenges in sleep transistor design and implementation in low-power designs. In: Proceedings of 43rd Annual Design Automation Conference. New York, 2006, 113–116

[7] Kudithipudi D, John E. Implementation of low power digital multipliers using 10 transistor adder blocks. Journal of Low Power Electronics, 2005, 1(3): 286–296

[8] Anuar N, Takahashi Y, Sekine T. 4×4-bit array two phase clocked adiabatic static CMOS logic multiplier with new XOR. In: Proceedings of 18th IEEE/IFIP VLSI System on Chip Conference (VLSI-SOC). Madrid, 2010, 364–368

[9] Meher M R, Jong C C, Chang C H. A high bit rate serial-serial multiplier with on-the-fly accumulation by asynchronous counters. IEEE Transactions on Very Large Scale Integration (VLSI) Systems, 2011, 19(10): 1733–1745

[10] Kim S, Kosonocky S V, Knebel D R, Stawiasz K, Papaefthymiou M C. A multi-mode power gating structure for low-volyage deep-submicron CMOS ICs. IEEE Transactions on Circuits and Wystems II, Express Briefs, 2007, 54(7): 586–590

[11] Chowdhury M H, Gjanci J, Khaled P. Controlling ground bounce noise in power gating scheme for system-on-a-chip. In: Proceedings of IEEE Computer Society Symposium on VLSI. Montpellier, 2008, 437-440

[12] Jiao H L, Kursun V. Ground bouncing noise suppression techniques for MTCMOS circuits. In: Proceedings of 1st Asia Symposium on Quality Electronic Design. Kuala Lumpur, 2009, 64–70

[13] Jiao H L, Kursun V. Ground bounce noise suppression techniques for data preserving sequential MTCMOS circuits. IEEE Transactions on Very Large Scale Integration (VLSI) Systems, 2011, 19(5): 763–773

[14] Jiao H L, Kursun V. Noise-aware data preserving sequential MTCMOS circuits with dynamic forward body bias. Journal of Circuits Systems and Computers, 2011, 20(1): 125–145

[15] Jiao H L, Kursun V. Ground-bouncing-noise-aware combinational MTCMOS circuits. IEEE Transactions on Circuits and Systems I, Regular Papers, 2010, 57(8): 2053–2065

[16] Kumar R, Kursun V. Reversed temperature-dependent propagation delay characteristics in nanometer CMOS circuits. IEEE Transactions on Circuits and Wystems II, Express Briefs, 2006, 53(10): 1078–1082

[17] Roy K, Mukhopadhyay S, Mahmoodi-Meimand H. Leakage current mechanisms and leakage reduction techniques in deep-submicron CMOS circuits. Proceedings of the IEEE, 2003, 91(2): 305–327

[18] Neil H E W, Harris D, Banerjee A. CMOS VLSI Design: A Circuit and System Perspective. 3rd ed. New Jersey: Pearson Education, 2011

Acknowledgements

This work was supported by ITM University Gwalior, with collaboration Cadence Design System, Bangalore.

RIGHTS & PERMISSIONS

2014 Higher Education Press and Springer-Verlag Berlin Heidelberg