November 30, 2017
November 30, 2017
|By: John Woodward
Executive Business Manager, Industrial Power, Maxim Integrated
If you ask a typical cellphone user what is more important, ability to talk continuously for a little longer on a single charge or having some percentage battery left after days on standby to make a quick call, I suspect both will be deemed important. After all, efficiency is the watchword in modern power design for energy savings, but variation of efficiency at light load or in standby condition affects battery charge availability of portable devices and is equally important to the user.
The DC-DC converters powering the electronics from depleting the battery in portable devices can make all the difference in these two scenarios. A simple design might have a good headline conversion efficiency of 95% or more at full load, but standby losses could vary tremendously, directly affecting your phone’s availability for that emergency call after a few days without charge. In DC-DCs, at light load, the losses tend to be fixed values independent of actual load power, typically coming from ‘housekeeping’ functions or charge and discharge of power MOSFET gates, independent of whether they are passing drain current. Efficiency under these conditions should be as high as possible but the actual figure is not so meaningful; if your standby load is 1mW and your fixed conversion losses are 1mW, you would probably be quite pleased, even though the efficiency is only 50%.
That absolute value of fixed loss is important though; at a system level, if it is halved to 0.5mW, talk time in a cellphone might improve by only a few seconds, but it might double the standby availability time counted in days. Efforts to lower no-load and standby losses are certainly worthwhile.
Synchronous is the Way to Go
For sure, 'buck' converters in portable devices generally need to use the best available technique of ‘synchronous rectification’ to achieve high efficiency. Buck converters consist of a series switch SW1 passing pulses of current to an inductor with a second switch allowing continuous current to flow to the output when the series switch is off. Pulse-width modulation, and the averaging effect of the inductor and a following capacitor, step down and regulate the output voltage.
Figure 1. Buck converter outline
A 'flywheel' diode for the second switch SW2 is a simple solution automatically conducting at just the right time by 'commutation', the action of the stored energy in the inductor causing forward bias of the diode when the series switch is off. The diode has a significant forward voltage drop, though, which is in series with the load for part of the switching cycle. If the load voltage is low, this diode drop becomes a large proportion of the total leading to poor efficiency. The effect is worse at high input voltages when the diode conduction time is longer in the switching cycle. Even Schottky diodes struggle to give decent efficiency levels at higher currents where their voltage drop increases to similar levels to standard fast recovery types.
Replacement of the diode with a controlled switch, dubbed 'synchronous rectification', was always the best solution as long as the switch is close to ideal with little voltage drop. Early circuits used bipolar transistors which beat diodes, but needed complex, power-hungry current drive circuits offsetting the advantages won, particularly at light load.
Enter MOSFETS as switches, which need far less drive power. High Rds(on) figures initially meant that, at higher currents, their voltage drop was comparable to diodes, but as technology has progressed to sub-milliohm devices, they have come into their own (Figure 2).
Figure 2. MOSFET synchronous rectifier
High efficiency at high loads is now easy to achieve but those standing losses still need attention. For example, a modern buck controller might operate at 3MHz but just switching a MOSFET gate at this speed between zero and five volts dissipates 15mW for a device with a total gate charge of just 1nC. The power is not consumed by the gate itself, it’s dissipated in the driver and series gate resistance and is independent of pulse duty cycle. Given that there are at least two switches in the converter, the problem is clear.
PFM or 'Pulse Skipping' Helps Standby Losses
At light load, it’s not necessary for a buck converter to operate at high frequency; the energy required can be supplied by short pulses with low repetition rate. By forcing this, the fixed switching losses can be dramatically reduced - power dissipated is directly proportional to the number of switching events or gate charge/discharge cycles per second. The controller effectively goes into a ‘constant on-time, variable off-time’ mode. The-off times can be so long at lightest loads that it’s actually practical to disable some of the controller internal circuitry for the duration as it has nothing to do, saving yet more power.
Disabling Synchronous Rectification at Light Load
Synchronous rectification has the advantage of bi-directional conduction giving continuous inductor current (CCM) at any load, which helps with your loop compensation design. It can be an advantage, though, to deliberately disable the synchronous MOSFET from conducting reverse current at light loads, forcing the converter into discontinuous conduction mode (DCM). If you can maintain loop stability, the net losses can be lower as the MOSFET is conducting for a smaller part of the switching cycle. If the converter is any way in pulse-skipping mode, the gain is more still. Figure 3 shows the waveforms you would see.
Figure 3. Different buck converter switching modes
The Maxim MAX17501 controller is a good example of a part incorporating these features. It can deliver 500mA of load current with a peak efficiency of better than 90% with up to 60V input and fixed 3.3V or 5V outputs. Pulse-frequency mode occurs at less than about 60mA, and at zero load the switching current reduces to less than 100µA at 20°C. As a comparison, if fixed-frequency PWM is optionally forced at no load, the current draw becomes closer to 5mA. In some applications you might want to suffer this to keep the switching frequency constant and noise levels more predictable.
Of course, if the load current is really zero, downstream circuitry is not operating and you have the option to shut down the controller entirely. In this case the current is less than 1µA. OK, the efficiency is 0%, but I don’t think you’ll mind. Learn more about this topic from Bob Mammano, the father of the first switch-mode power supply, in this video that he hosted on synchronous switching regulators. Mammano teamed up with Maxim to present a video-based Power System Design Seminar – watch it for a deeper dive into power supply design.