Browsing by Author "Chen, Yingping"
Now showing 1 - 3 of 3
- Results Per Page
- Sort Options
Item 15.6 A 10MHz i-Collapse Failure Self-Prognostic GaN Power Converter with TJ -Independent In-Situ Condition Monitoring and Proactive Temperature Frequency Scaling(Institute of Electrical and Electronics Engineers Inc.) Chen, Yingping; Ma, Dongsheng (Brian); Chen, Yingping; Ma, Dongsheng (Brian)With superior figure of merits, GaN switchs are highly anticipated to replace MOSFETs in high-performance power circuits [1], [2]. However, GaN technology today still faces formidable reliability challenges [3]. While GaN device aging and failure mechanisms are not as well-studied as silicon counterparts, its unique structure and operation also induce new aging and failure problems. Use a GaN switch M_H in a buck converter of Fig. 15.6.1 as an example. As a high-side switch, it faces large-switching-current and high-input-voltage stress in each charge phase. After repetitive switching actions, a number of electron carriers can be injected into the AlGaN barrier and buffer layers, known as hot-electron injection. In discharge phase, M_H is off, but M_L becomes conductive, which shorts the source of M_H to ground, creating high VDS stress on M_H. This induces charge traps in the insulator and buffer layers, known as charge trapping. As a joint effect of both mechanisms, trapped or injected electrons in the insulator, AlGaN barrier and buffer layers repel free electrons in the channel when M_H is on, weakening the 2-dimensional electron-gas (2DEG) layer and further elevating hot-electron injection. This effect, known as current collapse or i- collapse for short, degrades channel conductivity, increases the on-resistance RDS_ON, and is a major cause of GaN-switch aging and failure [3]. On the other hand, another aging cause is thermal effect. To reduce manufacture costs and improve technology compatibility, it is common to fabricate GaN transistors on a silicon substrate. Accordingly, to reduce lattice mismatch, an AlGaN buffer layer is often inserted (Fig. 15.6.1). However, this increases the junction-to-ambient thermal resistance R θJA, which, together with the increased R_{DS_ON} due to the i- collapse, causes higher power and heat generation, elevating the junction temperature, T J. According to Arrhenius' Law, as T J increases, the mean-time-to-failure (MTTF) drops exponentially [4]. Even worse, the elevated T_J deteriorates the i- collapse effect with even higher R_{DS_ON}, significantly reducing device lifetime. ©2019 IEEE.Item 15.7 An 8.3MHz GaN Power Converter Using Markov Continuous RSSM for 35dBμV Conducted EMI Attenuation and One-Cycle TON Rebalancing for 27.6dB VO Jittering Suppression(Institute of Electrical and Electronics Engineers Inc.) Chen, Yingping; Ma, Dongsheng (Brian); Chen, Yingping; Ma, Dongsheng (Brian)GaN power switches have gained fast-growing popularity in power electronics. With a similar R DS_ON resistance, they boast 2-to-3-order lower gate capacitance than silicon counterparts, making them highly desirable in high-frequency (fsw ), high-performance power converters. However, at high f sw , switching transitions have to be completed in much shorter times, creating much larger di/dt and dv/dt changes in power stage, which directly link to electromagnetic-interference (EMI) emissions [1]. To suppress EMI, spread-spectrum-modulation (SSM) techniques [2-5] have been proposed. As depicted in Fig. 15.7.1, a periodic SSM (PSSM) is straightforward and easy to implement. However, its EMI suppression is not effective [2]. A randomized SSM (RSSM) can outperform the PSSM, with lower peak EMI and near-uniform noise spreading, but its performance highly relies on the random clock design. In [3], an N-bit digital random clock was reported to achieve a discrete RSSM (D-RSSM). However, the bit number N has to be large in order to achieve satisfying EMI attenuation, significantly increasing circuit complexity, chip area, and power consumption. To overcome this, a thermal-noise-based random clock was proposed [4]. Unfortunately, thermal noise is very sensitive to temperature and is hard to predict. To apply this approach to a practical implementation requires additional signal processing with periodic signals to confine its range of randomization, which, in turn, reduces the benefits of the RSSM. To achieve a near ideal RSSM, a continuous RSSM (C-RSSM) with a cost-effective implementation is highly preferable. Meanwhile, another challenge of applying SSM schemes lies in the fact that the schemes deteriorate V O voltage regulation. As shown in Fig. 15.7.1, as an SSM scheme continuously or periodically modulates f sw , a converter switching period fluctuates cycle by cycle, causing random errors on the duty ratio and thus jittering effect on V O. This is difficult to correct by a feedback control loop, as the duty-ratio error changes randomly between switching cycles. Due to a limited loop-gain bandwidth, the loop response usually lags far behind. Although a ramp compensation scheme was reported to resolve this [5], the improvement is very limited, and the scheme only works for voltage-mode converters. © 2019 IEEE.Item GaN-based DC-DC Conversion Achieving High Reliability, Low EMI and Balanced System Performance(2020-04-15) Chen, Yingping; Ma, Dongsheng BrianDC-DC power conversion circuits improve the switching frequency consistently over the past five decades, pursuing better dynamic response and higher power density. To empower such a trend, silicon power transistors have been advancing continuously. However, they are approaching the theoretical limit of performance, slowing down the developing pace of power electronics. With superior switching characteristic, gallium nitride (GaN) high-electron-mobility transistor (HEMT) rapidly emerged, pushing the operating speed of DC-DC power circuits to a record high level. GaN technology is thus recognized as a promising candidate to enable the next-generation switching power conversion. However, it still faces formidable challenges before the industry-wide adoption, including unique reliability issues, considerable electromagnetic interference (EMI) emissions and intensified power design trade-offs. This dissertation delivers key innovations in power stage, gate driver and control scheme, intending to conquer these challenges. To improve the reliability of GaN power stage, an online condition monitoring is developed to prognose the current-collapse (or i-collapse) effect in GaN HEMT, sensing its dynamic onresistance as aging precursor. A gate leakage inspired junction temperature TJ sensor is integrated to determine the TJ of GaN HEMT, facilitating the calibration of temperature effect on the dynamic on-resistance. As the benefit, TJ-independent online condition monitoring is accomplished, significantly improving the monitoring accuracy. Further, to enhance the system longevity, a proactive temperature frequency scaling scheme is designed to modulate the operating speed according to the thermal stress and power conditions, thereby extending the GaN lifetime while minimizing the impact on the system performance of the converter. To reduce the conducted EMI noise, an adaptive strength gate driving scheme is developed for GaN HEMT. By modulating the driving strength at the start point of Miller Plateau during the switching transitions, it achieves an independent control of low di/dt and high dv/dt. Thus, the conducted EMI noise, mainly caused by di/dt, is reduced, while the switching power loss overhead is minimized. By such a means, the classic design trade-off between EMI noise and power efficiency is effectively balanced. To facilitate such an active control, an emulated Miller Plateau tracking scheme is proposed to identify the critical di/dt and dv/dt instants, which are susceptible to load current and power input voltage conditions. These proposed techniques are incorporated in a GaN-based buck converter for verification. Moreover, a continuous random spread-spectrum-modulation (C-RSSM) technique is utilized to scatter the EMI spectra evenly and continuously, attenuating the EMI further. For demonstration, the proposed C-RSSM scheme is applied to a GaN-based buck converter with peak current mode control. In the meantime, a one-cycle on-time rebalancing scheme is designed to overcome the crossover frequency limit existing in the conventional PWM control, thereby stabilizing the duty ratio under frequency modulation. Beneficially, the output jittering effect induced by RSSM is removed, balancing the trade-off between EMI and output regulation.