Streamlining the System Design Process

There are plenty of motivations for wanting to optimize the design process, especially when it comes to the power supply subsystem that can often be seen more as an inconvenient necessity than directly contributing to high-value system feature sets. Power supply solutions also tend to be some of the most expensive components in the system bill-of-materials (BOM). These reasons, along with the confidence that accompanies direct reuse of a qualified design block or commercially-available power module, are the main motivators for driving a heavy leverage/reuse strategy from one project to the next.

A key takeaway is that even though a power stakeholder shall be beholden to the outputs, they are rarely an integral part of the process that contributed to the inputs. Given a specialty area of focus that requires a multidisciplinary background (often only derived from many years of field experience), it is often puzzling how little power stakeholder perspectives are sought out early in the process for a subsystem(s) that tends to be a primary, gating agent to system size, weight, power, and cost (a.k.a. – the infamous SWaP-C factors) optimization. Since no electronics run without power, add performance and reliability to that list too. Just for icing on the cake, a project timeline that is structured around a perfect, error-free development process (minus 10% to be even faster time-to-market or TTM than the previous product) will also accompany all these idealistic demands.

Now comes the negotiating process. Engineers are trained to be problem solvers so when faced with a list of challenging problems, a kneejerk response is to start digging into the solutions (i.e. – Is there an existing part that can meet this power density and footprint? Should airflow go from front-to-back or back-to-front to meet the system thermal envelope? And so on and so forth…). Even this starting point is the first opportunity to take pause and dig very deeply into the system budget and how it came to be. For instance, how often are all loads (especially the bigger ones) drawing their max currents simultaneously? Surely, there are many subsystems that are designed to be in antiphase with another subsystem (e.g. – the classic examples of compute vs. memory power demands or sleep/wake/transmit operating cycles) so it is pretty rare when the sum of maxima (typically derived from datasheets that may already be starting from a point of an unrealistic max with safety margin) makes sense for an aggregated power budget. Consider each touch point of that power budget from inception until finalized. Each stakeholder will also be sure to add their own margins to cover their own guidance, which really adds up when aggregated. Those extra layers of fat cost a whole lot of money and resources to design to truly unrealistic operating scenarios in even the extremist of corner-case usage modeling.

Another key point in the fight against overinflated system power budgets is to know when to recognize the biggest opportunities for budget optimization. Start with the largest, most demanding loads in the system and talk to the critical stakeholder(s) that best understand what the load really needs in terms of power requirements and try to take real characterization data whenever possible. Doing so shall likely open the door to implementing intelligent power management (IPM) techniques, such as aggregating lower-voltage power rails, load sharing/shedding, and short-term power allocation. IPM is a “combination of hardware and software that optimizes the distribution and use of electrical power in computer systems and data centers” [1]. Though the term was coined for data center applications, the applicability is fairly universal as this is more a frame of mind in design approach than anything else. For instance, changing the approach to power subsystem architecture from an “always on” to an “always available” mentality can bring paradigm shifts in the results of the end solution. This will involve extensive discussions with team members as well as external vendors.

In other words, it tends to be far simpler, faster, and cheaper to put the maniacal work into reducing the system budget into a REALISTIC summary of even true, worst-case, maximum power loading (from each individual power supply’s perspective) as opposed to putting all that sweat equity into trying to bend physics and available components to the whim of unreality. Given that time and cost-down pressures are a constant, following this strategy will enable a far more amenable process to negotiate amongst team stakeholders and find a pragmatic balance between time, cost, and quality. These inevitable tradeoffs are inexorably tied to each other regardless of how much we wish they were not at times, as illustrated by the figure below. For instance, a product can be optimized for either time/cost/quality without optimizing for the other two.

Direct reuse refers to taking an existing design and copying it exactly. Effectively, this is the same thing as buying COTS components though this can sometimes be a little bit of a grey area since some fixed designs can actually be created with flexibility in mind. For instance, reusing power bricks with a wide-input-voltage range or programmable output for different applications. It can also be common to leverage a part family, particularly when referring to power modules designed for common footprints, optimizing specific module features (i.e. – input/output voltage range, power density, current handling, pinout, filtering, etc.) to the application.