I've often said that analog and digital circuits have their respective roles, but power circuits are inherently analog and always will be so. Yet while power itself–the amps and volts–are indeed analog, power management itself is certainly feeling the heat of digital control.
This shift makes sense. Digital power management offers clear advantages. It lets you tailor the same physical supply to different performance attributes and needs. Want a supply with fast or nonlinear transient response? Prefer to trade absolute accuracy for another dynamic parameter? Need a special sequencing of multiple output rails, with adjustable turn-on and turn-off delays? Want a tunable PID control loop? No problem, in principle, using digital control.
The benefits don't stop with the supply itself, however. Digital management lets you collect performance data and closely watch internal parameters. You can have real-time diagnostics and monitoring, watch supply dynamics, get early insight into impending failures, and much more. Assuming it is as reliable as conventional analog management, and cost-competitive (it will be soon), it sounds like a designer's dream come true with its inherent flexibility, insight, foresight, and on-the-fly reconfigurability.
Yet I still have that little voice in my head saying “careful what you wish for, you might get it.” Whenever you allow firmware to manage a function, you reduce your ability to get detailed insight into the design and its foibles. With an analog circuit, an engineer with the schematic, component data sheets, and an oscilloscope can really delve into what is working well or not so well. You are not at the mercy of the embedded code. Instead, the design details and what you need to work on its problems are clearly there, right in front of you.
Even more important, an analog circuit starts and runs due to its components and topology, not a software-sourced kick. If the digitally managed supply doesn't start up, where and how to you begin the debug? You can't see initialization parameters or strategy from the schematic. Did the person who wrote the software understand the idiosyncrasies of magnetics?
There's one more thing that worries me. Software's flexibility is a virtue, but also a cloak. It allows designers to patch around a problem with clever code. Often, though, this is done without understanding of the true problem: “hey, just get it working and out the door; there's no need to worry” that a passive component is drifting excessively or has nonlinear response when it saturates. Code over or around it, and all will be well, for now.
Eventually, digital power will work its way into almost all supplies, except perhaps those at the very low end. But before you jump in, think about its impact on your design, development, and debug process.