Publication | Closed Access
4.2 A 6A 40MHz four-phase ZDS hysteretic DC-DC converter with 118mV droop and 230ns response time for a 5A/5ns load transient
44
Citations
4
References
2014
Year
Unknown Venue
Low-power ElectronicsElectrical EngineeringLoad TransientEngineeringPwm ControlPeak Processor CurrentsEnergy EfficiencyPower Electronics ConverterComputer ArchitectureComputer EngineeringPower DissipationElectric Power ConversionPower Electronic SystemsPower ElectronicsPower-efficient ComputingPower-aware DesignResponse TimePower Management
In recent years, the clock frequency, the number of cores, and the power dissipation of application processors (APs) for portable electronics have dramatically increased. As a result, peak processor currents have reached several amperes with slew rates on the order of 1A/ns. These fast large steps incur large output voltage (V <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">OUT</sub> ) droops, which induce failed paths and cause processor black-outs. Present voltage regulators (VRs) combat these challenges by using bulky output capacitor (C <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">OUT</sub> ) arrays that could add up to over 100μF. However, this practice is untenable for next-generation APs, which have severely limited PCB area and require fast dynamic voltage scaling. These challenges have led to great demand for ultra-fast VRs. PWM control requires a bandwidth 5 to 10× less than the switching frequency, f <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">SW</sub> , which results in a slow response [1]. Hysteretic control has been proposed to achieve faster response [2, 3], however, it still suffers from an inherent delay (t <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">delay</sub> ) up to the discharge period, (1-D)T, due to realistic hysteretic window size and inductor current (I <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">L</sub> ) slew limit. Consider, for example, a fast hysteretic VR achieving 10% voltage droop for an instant load step of 5A. For an inductor (L) chosen to have I <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">L</sub> ripple <; 200mA, and C <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">OUT</sub> under a few μF, t <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">delay</sub> cannot exceed a few ns. This requires a f <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">SW</sub> from 0.5 to 1GHz, which in turn causes a large switching loss and restricts the feasible power level of the converter. This is against the power demand trend of APs. An interleaved multiphase topology can be the most effective way to improve both the system response and the equivalent I <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">L</sub> slew rate by changing the number of phases; however, clock and phase synchronization and current sharing for conventional hysteretic control are challenging.
| Year | Citations | |
|---|---|---|
Page 1
Page 1