Digital power supplies have quickly become popular, as pressures on energy efficiency, size and cost have sharpened system designers’ focus on power conversion. Understanding the principles of digital power can help us compare new products such as digital DC/DC point-of-load converters with their analog predecessors, and benefit from the improved flexibility and simplicity the new devices can offer.
Introduction: The Digital Conquest Continues
Digital technology has ousted legacy analog systems in fields such as audio, broadcasting and cellular communications, and is now taking over in power conversion. In general terms, its success is due to advantages including greater flexibility, lower cost, lower communication bandwidth, and easier high-volume manufacturability. As far as power conversion is concerned, digital power supplies have been shown to deliver greater efficiency over a wider range of load conditions. They can satisfy complex device-power demands such as multiple voltage rails with specific power-up/down sequencing more easily than analog power supplies, and can adapt to support dynamic changes in strategy during operation.
As one of the most important early adopters of digital power, the data-center industry has quickly come to rely on this technology to supply the hundreds of amps required by standard server boards within the tight space constraints imposed by standard rack and shelf dimensions. Digital power also ensures consistently high efficiency to minimize utility costs and thermal management challenges. With the benefit of the experience and long-term performance and reliability data now coming from the data-center industry, other application areas such as industrial power are increasingly willing to move from traditional analog supplies to access the advantages of digital power.
Comparing Conversion Principles
A digital power supply has the same objective as any of its predecessors: to provide a stable, regulated DC output at the desired voltage value despite changes in input voltage or load conditions. In the switched-mode power supplies (SMPS) or switching DC/DC converters that have been widely used since the 1990s, this has been achieved by feeding back a portion of the output voltage to an analog error amplifier that compares the received signal with a reference voltage representing the desired set-point. Figure 1 illustrates this principle.
Closed-loop feedback to regulate the output voltage of a conventional analog converter.
If any change in the load condition causes the output voltage to deviate from the set-point, the error amplifier generates a signal that is proportional to the difference. An analog comparator then compares this signal with a sawtooth waveform to generate the pulse-width modulated (PWM) driver signal that determines the on-time of the switching power transistor. If the fed-back voltage is higher than the set-point, this PWM signal will reduce the transistor’s on-time to ensure the voltage will reduce towards the set-point. Conversely, if the detected voltage is below the set-point, the PWM signal will increase the on-time to increase the output voltage. The behavior of the feedback loop and error amplifier circuit is determined by components whose values are fixed during the design of the power supply to ensure correct response and keep the feedback loop stable under all operating conditions.
On the other hand, when the feedback loop is implemented digitally, as shown in figure 2, below, the output voltage is first converted to a digital equivalent and compared with the digital value of the set-point. This comparison generates an error term that is then fed into an algorithm, which is fundamentally a Proportional-Integral-Derivative (PID) filter. The output of the PID filter is then converted to an analog signal that determines the on-time of the switching transistor. Since the digital control circuitry is implemented in low-voltage logic, external high-current gate-driving circuitry (in the control and power switches block of Figure 2) is required.
The digital feedback loop is defined by firmware, and sensed values could be recorded to enable flexible control and analysis.
In principle, this appears similar to the analog approach already described. However, the PID filter has much greater influence over the output characteristics than the analog comparator. The proportional and derivative terms determine the response speed and overshoot of the output voltage respectively, while the integral term can use past error signals to help stabilize the output at the desired voltage. Moreover, whereas the analog feedback characteristics are fixed and can only be optimal over a limited operating range, the PID filter is firmware based and so can be adjusted without changing hardware. These properties of the digital feedback loop effectively provide the basis for a sophisticated, fully programmable, firmware-controlled converter.