When digital power first emerged as a topic of debate in the early 2000s, most hard core power supply designers cast their doubts at the thought of replacing analog PWM controllers with a DSP based digital power supply controller. Power engineers associated digital power with fears of z-transforms, software coding or blowing up power supplies due to “blue screen of death”, as sometimes seen in a particular computer operating system. Furthermore, prior to the introduction of semiconductors that included all the functionality necessary to digitally communicate and control a complex power system, the fundamental concepts were proven using high cost DSPs or performance limited FPGAs. Even though digital power techniques had demonstrated the ability to raise power system performance, the industry could not accept mimicking digitally what was already being done in analog at more than 10 times the silicon cost. Much progress has been made during the past several years and market analysts as well as financial investors appear very optimistic regarding the future growth of digital power.
Today’s digital power market is made up of numerous semiconductor suppliers and original equipment manufacturers (OEMs) each contributing unique solutions. It’s important to first distinguish what is meant by digital power. Power system supervision, monitoring, fault detection and data logging is one aspect of “digital power management” (DPM) that can be implemented using an inexpensive microcontroller, FPGA or PLD, where the time base requirements are relatively slow. However, the more challenging example of digital power refers to “digital control of power” (DCP) where one or more power supply control loops are closed around a high speed analog to digital converter (ADC) and a micro-controller, state machine or DSP-based control algorithm.
Semiconductor market research firm, iSuppli estimates that the total world wide digital power market will grow from its current 2008 revenue of just over $200M to about $900M in 2011. The digital power rate of revenue growth is summarized graphically in figure 1 indicating a progressive growth pattern with DPM far outpacing DCP for the next several years.
volume, non-isolated, DC-DC power conversion, including voltage regulator modules (VRM) and point of load (POL) converters. These types of DC-DC converters typically deliver power to low voltage, high current digital loads such as microprocessors, DSPs and other high speed digital circuits that are very dynamic in nature. It is becoming increasingly difficult to maintain accurate voltage regulation around 1V, while meeting load transient requirements approaching 200A/ns using pure analog control techniques. Some digital controllers can offer certain features, such as non-linear control which may be more difficult or impossible to realize in an equivalent analog IC. In fact, just about all digital POL controllers include some distinguishing control technique aimed at improving transient response. Since most of these proprietary control algorithms were developed by start-up venture companies, the OEM’s demand for preferred suppliers and dual sourcing became a barrier to gaining market acceptance. This in turn has led to licensing agreements between a number of digital power start-ups and the more established analog semiconductor suppliers. Non-isolated, DC-DC POL applications are viewed as a digital power entry point into the power conversion market. However, it is eventually expected that isolated DC-DC and AC-DC power factor correction (PFC), which currently implement digital control using non-dedicated DSPs, will gain even higher acceptance in coming years.
The functional requirements for a digital PFC controller are much different compared to a POL controller. PFC regulators require less load regulation accuracy and operate at a lower switching frequency and lower overall control loop bandwidth making quantization errors, limit cycling and ADC resolution less of a design challenge. Still, there are unique design issues with PFC controllers demanding significant system knowledge and experience. Although load transient requirements and dead-time optimization are not concerns for PFC, non-linear control could be useful for handling line transient conditions which are characteristic of virtually all off-line converters.
AC-DC power supplies generally require an electro-magnetic interference (EMI) filter capable of attenuating noise generated by harmonic distortion and switching regulators. Since EMI filters are designed using networks of inductors and capacitors that appear in series with the AC mains, they have a negative effect on overall efficiency but are necessary for meeting certain equipment specifications. Some digital PFC regulators can incorporate features aimed at reducing EMI filter requirements, resulting in smaller more efficient EMI filters. Analog PFC controllers are distinguishable by various control algorithms and operating modes such as discontinuous conduction mode (DCM), continuous conduction mode (CCM) or boundary conduction mode (BCM). Each PFC operating mode has advantages and disadvantages that become more evident at various power levels. Similar to the multi-mode operation of a digital POL, a digital PFC controller could be assigned to operate in either PFC control mode. Whether for PFC or POL applications, a controller that is configurable across many OEM product platforms provides value through faster development time and less component inventory to maintain.
Higher power AC-DC applications are better candidates for digital power because they require one or more PFC regulators as well as multiple, isolated and non-isolated power rails and complex cooling profiles. Server power systems are one example where digital control and power management can offer performance and efficiency benefits in a distributed power architecture (DPA). As more digital power content makes its way into the server and computing power segments, other high power DPA systems such as those used in telecom and datacom are beginning to embrace digital power solutions. The depth of present and forecasted market penetration for all digital power semiconductors is clearly being led by these types of high power applications, as shown graphically in figure 2. Interestingly, recent government and regulatory requirements for PFC above certain power levels combined with the falling costs of silicon for digital controllers may enable the feasibility for digital power development in some consumer ac-dc power supplies such as game systems, set-top boxes, digital televisions, AC adaptors and battery chargers.
Regardless of analog or digital, power supplies developed for consumer market applications are especially cost sensitive. The cost pressure compared to existing analog solutions has made digital power a tough sell for consumer applications but some industry analysts predict that within 2008 the silicon cost of existing digital controllers may approach price parity compared to similar analog solutions. The value proposition of digital power will never be compelling enough if compared to a simple analog power supply solution. One power supply OEM in particular has designed a platform of digital power solutions that has been shown to beat their existing analog solutions in terms of performance and overall cost. Issues such as system complexity and proprietary intellectual property demanded that the proof come from within their own company. The cost comparison between digital and analog power solutions has to be looked at from the entire system point of view including: the cost of design, process development, test, qualification, manufacturing, inventory control and component savings.
Although wide market adoption of digital power is still early in its infancy, there are several key factors for acceptance that can not be ignored. Tougher global efficiency requirements beg the question: what can digital power contribute towards increasing efficiency? When implemented correctly, digital control can potentially offer some efficiency improvements through duty cycle optimization, adaptive shoot-thru control, meeting transient requirements while enabling lower switching frequency, real time adaptation of critical timing such as synchronous rectification or resonant dead time, phase shedding and varying operating modes based upon load current. Digital control can yield better overall efficiency results because the system power is managed more efficiently according to the demands of the load. Any efficiency improvement is always welcome but in order to set realistic expectations, it is important to emphasize that efficiency is primarily determined by the power stage design.
A reliable, easy to use graphical user interface (GUI) is an essential requirement for the success of any digital power controller. Many power designers are quick to nod in agreement at the thought of wanting all the flexibility a digital controller can offer. Enabling flexibility through hundreds of programmable commands requires careful consideration in terms of software layers, partitioning, protection features and command grouping. The GUI needs to have an intuitive, natural feel but most importantly it needs to work one hundred percent of the time. Power supply design is difficult enough without adding doubt as to whether the controller is behaving the way it was programmed.
Digital controllers also need to be self-configurable, meaning that no calibration should be required by the end user. Some power supplies are manufactured in an environment where the controller module and power stage might be arriving from different suppliers and are not integrated until final assembly. In this case, there is no opportunity to calibrate the controller since the power stage is unavailable at the time of delivery.
Every digitally controlled power converter inevitably requires a certain amount of analog circuitry. Whether external or integrated into the same package, gate drivers are just one example of an interface between the analog and digital domains. ADCs, voltage references, regulators and comparators are also required analog functions that must perform seamlessly in or around a digital controller. Because there are numerous ways to partition digital and analog functions, obtaining the optimal balance between silicon performance, efficiency, intelligence, cost, process capability and die area continually poses a difficult challenge for system engineers. In retrospect, the concept of digital power is really nothing new or revolutionary. DSPs and microcontrollers have been widely used in industrial motor control applications for many years. The power supply industry has been inching toward digital power adoption but the necessary standards and dedicated semiconductors were not available until recently.
Digital content such as combinational logic, clocks, counters and timers have always existed inside analog controllers. As mixed signal design processes continue to improve, expect the relationship between digital and analog components to be optimized to a much higher degree. Depending on the complexity of a targeted application, some functions would benefit best from a digital process while others should remain in the analog domain. At least in the near term, the most cost-effective, optimal solutions appear to be mixed signal approaches no matter how “digital” a digital power system becomes. The fundamental groundwork necessary for shaping the landscape of digital power is in place but there are still questions remaining as to how digital power will complement or compete with traditional analog solutions. Most accomplished power supply designers who recognize the value of digital power will update their design skills accordingly but regardless of different viewpoints on adoption rates, one thing is certain: digital power is inevitable.
Captions
Figure 1. Global Digital Power Semiconductor Revenue Forecast (Millions of U.S. Dollars)
Figure 2. Digital Power Semiconductor Market Penetration Rates
Click here for the illustrations: Figure 1, Figure 2
