30.5 A 95.3% 5V-to-32V Wide Range 3-Level Current Mode Boost Converter with Fully State-based Phase Selection Achieving Simultaneous High-Speed $\mathbf{V}_{\text{CF}}$ Balancing and Smooth Transition

Abstract

DC-DC boost converters are widely used for various applications that require a high supply voltage, including solid-state drives (SSDs), LED drivers, etc. A multi-level (ML) boost converter (BST) with a high conversion gain (CG) and low voltage rating for switches (SWs) would be an ideal solution to regulate the high output voltage <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(\mathrm{V}_{0})$</tex> [1–4]. However, in practice, ML converters suffer from a flying capacitor <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(\mathrm{C}_{\mathrm{F}})$</tex> voltage <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(\mathrm{V}_{\text{CF}})$</tex> imbalance due to the parasitic capacitance <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(\mathrm{C}_{\text{par}})$</tex> and gate driving power <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(\mathrm{P}_{\text{GD}})$</tex> , etc., which causes not only large conduction losses <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(\mathrm{P}_{\text{COND}})$</tex> in the on-resistance <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(\mathrm{R}_{\text{ON}})$</tex> of the SWs and the parasitic resistance <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(\mathrm{R}_{\mathrm{L}})$</tex> of the inductor (L), but also a large overlap loss and damage to the SWs. For these reasons, a <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{V}_{\text{CF}}$</tex> calibration <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(\mathrm{V}_{\text{CF}}-\text{cal})$</tex> technique is essential to the ML converter. In particular, unlike a ML buck converter, MLBST requires the faster <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{V}_{\text{CF}}$</tex> -cal because the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{V}_{\text{CF}}$</tex> 's reference voltage <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(\mathrm{V}_{\text{CF},\text{REF}})$</tex> is <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{V}_{0}$</tex> which can vary abruptly. For the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{V}_{\text{CF}}$</tex> balance, prior works [1], [2] adjusted the duration of the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{C}_{\mathrm{F}}$</tex> charging and discharging phase <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(\phi_{\text{CH}},\phi_{\text{DIS}})$</tex> using two mismatched duties <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(\mathrm{D}_{1},\mathrm{D}_{2})$</tex> modulated by the calibration loop. However, that caused sub-harmonic oscillations in the inductor current <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(\mathrm{I}_{\mathrm{L}})$</tex> , which enlarged <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{P}_{\text{COND}}$</tex> in <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{R}_{\mathrm{L}}$</tex> proportional to the frequency. In addition, the bandwidth (BW) limit of the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{V}_{\text{CF}}$</tex> -cal loop by the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{V}_{0}$</tex> control loop and the repetitive <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\phi_{\text{CH}}$</tex> and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\phi_{\text{DIS}}$</tex> every cycle cause difficulty of immediate <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{V}_{\text{CF}}$</tex> ·cal in the start-up (Fig. 30.5.1, top left). Since it can result in breakdown and reliability issues for transistors, it is hard to mass produce the MLBST. Moreover, for the stability of the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{V}_{\text{CF}}$</tex> -cal loop, an additional compensator that uses bulky passive components is necessary. Another structure that does not depend on the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{V}_{\text{CF}}$</tex> -cal loop was proposed, but it needed an additional capacitor and incurred a surge current with redistribution losses due to the repetitive capacitor hard charging [3] (Fig. 30.5.1, middle left). Meanwhile, in power converters, the current mode control (CMC) is preferred thanks to the fast dynamic performance, over current protection, etc. But it is hard to apply it to the MLBST because of the mode transition issue. In the 3-level (3L) BST, there are two operation modes, mode-1 <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(\phi_{\mathrm{M}1})$</tex> and mode-2 <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(\phi_{\mathrm{M}2})$</tex> . Prior CMC 3L converter changed the mode between <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\Phi_{\mathrm{M}1}$</tex> and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\phi_{\mathrm{M}2}$</tex> by the forced mode selection signal intermittently every pre-determined period at the mode transition point <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(2\mathrm{V}_{\text{IN}}\approx \mathrm{V}_{0})$</tex> [2]. This was subject t0 <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{V}_{0}$</tex> fluctuation due to the DC offset of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{I}_{\mathrm{L}}$</tex> between two modes (Fig. 30.5.1, bottom left). To solve the above issues, we propose the following: 1) a fully state-based phase selection (FSPS) technique which is capable of the fast <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{V}_{\text{CF}}$</tex> -cal, stable start-up, and smooth mode transition without the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{I}_{\mathrm{L}}$</tex> sub-harmonics of the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$3\text{LBST};2$</tex> ) a comparator-based low-power <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{C}_{\mathrm{F}}$</tex> -charqinq and discharging phase selector ( <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{C}_{\mathrm{F}}$</tex> -CDPS) that does not require an additional compensator, and simultaneously enables both real-time <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{V}_{\text{CF}}$</tex> sensing and calibration signal generation; 3) an adaptive slope generator (ASG) for optimal slope compensation and smooth mode transition in a wide I/O range of CMC.

Topics & Concepts