IEICE TRANS. INF. & SYST., VOL.E101–D, NO.4 APRIL 2018 1116 PAPER Analysis of Body Bias Control Using Overhead Conditions for Real Time Systems: A Practical Approach∗ Carlos Cesar CORTES TORRES†a), Nonmember, Hayate OKUHARA†, Student Member, Nobuyuki YAMASAKI†, Member, and Hideharu AMANO†, Fellow SUMMARY In the past decade, real-time systems (RTSs), which must in RTSs. These techniques can improve energy efficiency; maintain time constraints to avoid catastrophic consequences, have been however, they often require a large amount of power since widely introduced into various embedded systems and Internet of Things they must control the supply voltages of the systems. (IoTs). The RTSs are required to be energy efficient as they are used in embedded devices in which battery life is important. In this study, we in- Body bias (BB) control is another solution that can im- vestigated the RTS energy efficiency by analyzing the ability of body bias prove RTS energy efficiency as it can manage the tradeoff (BB) in providing a satisfying tradeoff between performance and energy. between power leakage and performance without affecting We propose a practical and realistic model that includes the BB energy and the power supply [4], [5].Itseffect is further endorsed when timing overhead in addition to idle region analysis. This study was con- ducted using accurate parameters extracted from a real chip using silicon systems are enabled with silicon on thin box (SOTB) tech- on thin box (SOTB) technology. By using the BB control based on the nology [6], which is a novel and advanced fully depleted sili- proposed model, about 34% energy reduction was achieved. con on insulator (FD-SOI) technology. Thus, combining the key words: silicon-on-insulator, SOTB, body bias, low power design, time- benefits of SOTB and adaptive BB can drastically suppress overhead, energy-overhead the leakage current. However, when controlling the BB on RTSs, the BB must be controlled dynamically so as not to 1. Introduction miss the deadline. Although the energy for statically main- taining the BB voltage is quite small, the dynamic control of Real-time systems (RTSs) are part of our daily lives; they are the BB requires considerable energy. Although several stud- used in different domains, such as home appliances, medi- ies on dynamic BB control have been conducted [4], [7], [8], cal systems, robotics, security, aeronautics, and many oth- they were not based on accurate real-chip measurements that ers. One class of these systems is used for highly time- include the BB switching-voltage overhead. critical tasks that should be executed in a predefined dead- Based on the above, we investigated RTS energy ef- line. When failing to meet this deadline, the executed task’s ficiency by analyzing the dynamic BB control on perfor- results can be corrupted, or even the entire system might fail, mance and energy, including the physical energy and tim- possibly leading to catastrophic consequences. ing overheads when executing the voltage transitions. To At the same time, and with the increasing popularity this aim, we propose a practical timing and a power mathe- of Internet of Things (IoT), the need to design RTSs that matical models capable of determining the energy consump- can be embedded in small devices has become a necessity. tion based on the task execution while taking into account a Also, these types of embedded RTSs require a long lasting given deadline constraint. Thus, the contributions of this battery life and should operate on a limited power budget. paper can be outlined as follows: As technology continues to scale, the leakage current will • keep increasing, and a strict control is needed to find an op- The timing and energy overheads when switching the timal operational region. Hence, the energy consumption BB voltages are measured with a real microcontroller should be kept minimum while making sure that the timing implemented with SOTB technology. • constraints are met. By using these measurements, a practical power model The RTS energy efficiency has been extensively stud- for scaling the BB according to the switching behav- ied, and some have focused on very-large-scale integration ior, operational frequency, clock cycles per instruction (VLSI) designs. Various techniques, including power gating (CPI), and time for a deadline is proposed. The pro- (PG) [1] for dynamic power management (DPM) [2], and posed model can calculate the energy consumption for dynamic voltage scaling (DVS) [3] have been introduced each task of a given RTS application. • The energy saved with the proposed model is analyzed. Manuscript received August 11, 2017. Manuscript revised November 16, 2017. This paper is organized as follows. Section 2 provides Manuscript publicized January 12, 2018. a background of studies related to energy optimization in † The authors are with the Department of Information and Com- RTSs. Sections 3 and 4 are dedicated to explaining our tim- puter Science, Keio University, Yokohama-shi, 223–8522 Japan. ing and power models. The evaluation setup and results are ∗This study was supported by Keio Kenkyu no susume schol- arship. presented, analyzed, and discussed in Sects. 5 and 6. Finally, a) E-mail: [email protected] Sect. 7 summarizes the findings and outlook of this paper. DOI: 10.1587/transinf.2017EDP7258 Copyright c 2018 The Institute of Electronics, Information and Communication Engineers CORTES TORRES et al.: ANALYSIS OF BODY BIAS CONTROL USING OVERHEAD CONDITIONS FOR REAL TIME SYSTEMS 1117 a power supply is cut-off for idle states, volatile data are dis- carded. When data need to be preserved, a certain level of voltage has to be supplied as a power supply. Hence, power- leakage reduction is restricted to such conditions. The DVS is a technique that decreases the power supply voltage while keeping application deadlines [3]. It can drastically reduce the dynamic power due to the quadratic power-supply de- pendency. However, the range for power-supply scaling is highly restricted when the power supply voltage is near the Fig. 1 Cross-sectional view of SOTB MOSFET: (a) pMOS and (b) threshold region [11]. Such limitations can drastically im- nMOS pact the efficiency of energy saving. Some studies have analyzed the benefits of combin- ing DVS and adaptive BB for energy reduction [4], [7]. Yan et. al. proposed a task-scheduling algorithm and en- 2. Body Bias Control ergy models for RTSs usage [4]. These models can calcu- late an optimal power supply and BB voltage for each op- 2.1 Silicon on Thin Box erational frequency. By using the obtained voltages, the al- gorithm schedules a task so as not to violate the deadline. An SOTB is a FD-SOI device [6]. It has benefits of latch-up Namely, the authors assume ideal voltage regulators that immunity, superior high temperature, performance, radia- can output any voltage obtained from the models. How- tion hardness, and high BB sensitivity. These characteris- ever, the actual voltage drivers have a certain limitation in tics are possible due to its insulating “buried oxide” layer terms of output-voltage resolution. Akgul et. al. proposed a widely used in SOI devices [9], [10]. These body-driven power-management method considering these voltage con- characteristics enable high caliber energy reduction using straints [7]. The authors assumed discrete power-supply the BB. Unlike other conventional FD-SOI devices, a SOTB voltages and succeeded to reduce the energy even under the device is formed on an ultra-thin box layer (about 10 nm), as restrictions mentioned earlier. However, these studies did shown in Fig. 1, enabling a wide range of BB control. Con- not treat actual overheads of BB control, which also has en- sequently, SOTB ensures more efficient reduction in leak- ergy consumption and switching delay. age current using BB control than other conventional metal- Several works have been proposed to improve energy oxide-semiconductor field-effect transistors (MOSFETs). efficiency. When considering overhead conditions or an- We denote the BB voltage of nMOS as VBN, that of alyzing idle regions, however, all these approachesare at pMOS as VBP, and supply voltage as VDD. As other circuit level [12]–[14]. In our previous study [5], [8],we FD-SOI technologies, the default state of a given MOSFET developed a power model using BB control. The model (VBN = 0 and VBP = VDD) in SOTB is called zero body is based on real-chip measurements in terms of leakage bias (ZBB). If a lower voltage is applied to the nMOS body current, switching current, and maximum operational fre- (VBN < 0) and higher voltage is applied to the pMOS body quency. However, ideal BB switching is also assumed. (VBP > VDD), the depletion width increases; hence, the Hence, to the best of our knowledge, none of the studies threshold voltage increases. This condition is known as re- presented above incorporated these timing and energy over- verse body bias (RBB). In contrast, if a higher voltage is head conditions in their energy-saving approaches targeted applied to the nMOS body (VBN > 0) and a lower voltage for RTSs. is applied to the pMOS body (VBP < VDD), the depletion width decreases; thus, the threshold voltage decreases. This 3. Proposed Model condition is known as forward body bias (FBB). The FBB can achieve high operating speeds, increas- Without power-saving control, a task is executed in time ing the performance at the cost of leakage current, while the top and finishes at the given deadline. The frequency and RBB reduces performance and leakage current at the price voltage are constant all through the deadline. Hence, the of gate delay. Although the RBB for RTSs is useful in re- power leakage consumed in the idle region is wasted, as ducing the leakage current in the sleep mode, the timing and shown in Fig.