Several semiconductor suppliers offer digital power ICs as improved solutions over traditional analog approaches. To replace an existing analog power design, a digital power solution must perform equal to or better than the existing solution and offer key design advantages. Digital solutions match or improve the efficiency, size and cost of analog implementations, while providing many more benefits.
A digital power solution differs from an analog solution in that the pulse width modulation (PWM), loop control and feedback are implemented digitally. Analog signals are converted to digital using analog-to-digital converters. Once the signals are digital, microcontrollers, digital-signal processors or simple state machines control the PWM and the feedback loop.
First and foremost is power conversion. Digital power becomes interesting when the efficiency and cost is equal to or better than a comparable analog power-conversion solution. However, power conversion is only part of the overall system solution. Having the digital controller in a standard CMOS-silicon process allows for the integration of the power management with the power conversion. The integration of the thermal and power management features begins to make the digital power solution compelling.
Integration Reduces Component Count
Key aspects of power management include voltage and current monitoring, voltage sequencing, voltage tracking, fault detection and fault management. Thermal management includes the ability to monitor the temperature throughout the system and respond to overtemperature conditions by controlling fans or shutting down parts of the system. Integrating the power and thermal management with the power conversion removes the need for additional power and thermal management ICs in the system. For example, an analog power system with four power rails may require more than 150 components, whereas the same system using a digital solution would require fewer than 50 components (Fig. 1).
Digital solutions reduce component count not just by integration, but also by improving upon existing features. In analog solutions, RC time constants are used to set delay and ramp times. These times are now part of the configuration of the digital power IC and do not require an external RC to set the time. Similarly, external resistors and capacitors are not required when configuring the loop compensation and output preset on a digital power IC.
Configuring features in the digital domain and the integration of the thermal and power management with the power conversion reduces the overall component count, which lowers system cost and improves system reliability.
The power engineer's job is to design a reliable power distribution system of many voltage rails. These voltage rails are required to sequence or track other rails to properly bias ASICs, FPGAs, microprocessors or other digital logic that is present in the system. As the product design moves through the various design phases, changes in the design may occur. These changes could be an addition of a power rail or the need for more current on any given rail, or a transient response requirement may tighten. This would require the power distribution to be redesigned.
A digital power solution provides the flexibility to easily adapt to the requirement changes. A new voltage rail is easily added to the power-management system through the use of the industry-standard System Management Bus (SMBus). The digital power ICs communicate with each other over the SMBus using the Power Management Bus (PMBus) protocol. The addition of a new rail is easily integrated into the monitoring, sequencing, margining and fault detection schemes that have already been designed. The digital power IC for the new rail is given its own SMBus address and is added to the system. There is no need to reprogram or add more stand-alone power-management ICs just because of the additional voltage rail.
Digital power ICs have the ability to configure the load-line impedance, or output-voltage droop. This flexibility helps balance out errors over load when current sharing or to compensate out inherent droop in the load. Adding compensation flattens out the load line. Sometimes droop can be helpful. Intentionally adding droop can improve transient response. A tighter transient requirement generally means adding more capacitors. However, for relatively small improvements, adding more droop in a digital converter will improve the transient response. Fig. 2 shows a transient response with and without output-voltage droop. The load step causes the output to droop down by 30 mV, which allows the output to reach the target voltage faster and thus a faster transient response.
This technique can be used to achieve a smaller ΔV with the same number of components or a similar ΔV with fewer components.
Other requirement changes are also easily adapted, generally without any hardware changes. Delay time, ramp-up time and ramp-down time can be reconfigured during development through external resistor settings or simple PMBus commands. There is no longer the need to recalculate and change RC time constants all over the board in order to adapt the delay and ramp times as the design progresses. Likewise, there may be times where the required voltage for a given rail will change. Again, this change is accommodated simply by using a PMBus command to the power IC without the need for changing external components. Many ASICs or FPGAs require their voltage rails to ramp up and ramp down in a particular order to prevent latch-up in the device. Digital power ICs communicate with each other and can sequence their voltages for the application. If the sequence order changes, the digital power ICs are reconfigured to the new sequence order. This flexibility makes it easy to adapt a design to the new system requirements without having to change the board layout. This greatly reduces the development time of the product.
Power-supply sequencing is controlled through the SMBus as shown in Fig. 3. Once ENABLE goes high, output 1 waits a predetermined delay period and then ramps to its final voltage. The first device sends a power-good signal over the SMBus to the second device. Output 2 waits its delay period and then ramps to its final value.
A flexible design reduces the number of board revisions during the development phase. Reducing or eliminating board revisions reduces product development time as well as product development cost.
In a typical power-distribution system today, the voltage rails are monitored for voltage faults only. The added components required to monitor the currents effectively, and thus the power, for each rail adds much cost to the system. Having the power management integrated with the power conversion allows for a more reliable system where all of the rails are monitored for voltage and current. Monitoring both the voltage and current of each rail is easily realized and inexpensive in a digital power system. Temperature monitoring provides additional benefits. Each digital power IC has an internal temperature diode and the ability to monitor external temperature diodes. With the known temperature data, current measurements at each power rail are easily compensated for temperature changes. The external temperature data can provide a thermal map of the entire system, which can be used to reduce hot spots in the system.
Having all this monitored data can be useful. The efficiency of each power rail can be calculated from the VIN, VOUT, IOUT and duty cycle. The efficiency can be tracked over time to predict pending failures in the system. Early warnings of a system failure are easier to deal with than catastrophic failures. All voltage rails can be monitored and fault detection can be acted upon accordingly to improve the operation and reliability of the system.
A digital power solution provides multiple methods of reacting to a fault. Overcurrent and undercurrent, overvoltage and undervoltage, and overtemperature faults are common faults to monitor in a power system. Fault and warning thresholds can be configured and adjusted throughout the life cycle of the product. When a fault does occur, the supply can shut down, wait for a period of time to see if the fault clears, and then either shut down or act as a warning and do nothing. Equally important, the fault is communicated to the other voltage supplies in the system. This allows for the other supplies to react to the fault as well.
Faults occur because a voltage, current or temperature went beyond certain limits. It is important for the safety of the system that the power supplies react according to the fault when it occurs. It is also beneficial to the design engineer to understand what the fault was that caused the system to shut down. This information is stored in the digital power IC to allow the engineer to debug and improve the reliability of the system.
Digital solutions allow for all the power rails to be monitored for voltage, current and temperature faults to provide a more reliable system.
Lower Cost of Ownership
Many functional elements in a digital power system contribute to the lower cost of ownership. The lower component count makes a more reliable and longer-lived system. The highly configurable digital power IC allows the designer to use the same device for each voltage rail and to make changes to the operation of the device without having to make hardware changes. The power engineer does not need to learn how to use multiple devices to support high power rails versus low power rails. Design tools allow the engineer to build and model the entire power system prior to building hardware. This reduces the development cost of the system by having a faster development time and fewer board spins.
The monitoring and fault management with reporting allow for a system design that will have a longer life, lower service costs and lower warranty costs. The fault reporting can help the engineer troubleshoot system issues in order to improve the design of the system and future systems.
Boards are getting denser and are consuming more power. Inefficient power supplies cause boards to run extremely hot. Improving the power-supply efficiency can help reduce the generated heat in the system. Efficiency is measured as the output power divided by the input power. Power lost in the power conversion leads to a lower efficiency. Contributors to lost power in a synchronous buck converter are the on-resistance of the switching FETs, the energy required to turn on and turn off the FETs, the power required to bias the digital controller circuit and the resistance of the output inductor.
A digital solution can help improve the efficiency and thus reduce the heat. In the synchronous buck converter case, power is lost when both switching FETs are on (cross conduction) and when both FETs are off (body diode conduction). This is known as dead time since no power is being delivered to the load. Ideally, one FET should be on while the other FET is off and there would be zero dead time. Digital controllers can optimize the switching of the FETs and minimize the dead time to a point that the dead time nears zero. This optimization can improve the efficiency by 1% to 2%, depending on the application.
The figure below shows scope plots before and after optimization. The trace on the left is prior to optimization. Note there is a period where both the high-side and low-side switching FETs ore off. This is the dead time. The trace on the right shows the waveforms after optimization. Here both switching FETs switch at just about the same time, but with opposite polarity, eliminating just about all of the dead time.