Key Drivers for Power Module Density

Power vs. Volumetric Density

Power solutions are notorious for being the long pole in the tent—for driving overall system size, volumetric efficiency, system bill-of-materials costs, and power density. In general, these are broken down into the common figures of merit (FOM) for a system, such as its size, weight, and power (a.k.a. – SWaP factors) characteristics. When combined with a cost metric, this can also be referred to as SWaP-C factors [1]. Power density is typically a function of total available power versus overall solution volume, which is why component size tends to have an inverse relationship with power density. The power density metric is taken a step further when combined with overall solution mass (typically translated to weight on Earth), which can be a critical FOM in nontethered applications, as is reviewed from many perspectives in the content that follows.

It is also good to differentiate power density from volumetric density, where the power density can be characterized specifically in the context of the power solution, which is a subset of the overall system volume. In general, power density ALWAYS goes up, where volumetric density can decrease as major system loads shrink in size (and perhaps in power requirements) and/or increase their functionality to perform more work in the same volume from generation to generation, lending themselves to a different trend than is seen directly for power solutions. The industry has tried to normalize these trend discrepancies with oversimplified, terrible metrics, such as dollar per watt ($/W), which makes little to no sense unless it makes highly similar power supply comparisons.

As with just any aspect of evaluating power solutions and assessing their technical impacts and financial contributions, it is crucial to look beyond first-order analysis. Power consumption and energy efficiency can often be a game of “whack-a-mole,” in which optimization in one subsystem may lead to a lesser performance in another area; thus, the effective system-level impact is the same or even worse when taking this approach. Some classic examples are when the enhanced power density of a wide-bandgap power switch, such as gallium nitride or silicon carbide, enables a physically smaller (even with increased power-handling capability) power train by taking advantage of increased switching frequencies that enable the reduction of some power components. However, it may also consequently require a larger (and perhaps costlier) thermal mitigation solution to handle the denser power dissipation in smaller geometries or even push a system into requiring liquid cooling. It can often be the “little” features that are “nice-to-haves” but can have a disproportional impact on solution size and/or cost. For instance, connectors (especially blind-mate type) and fans can be very significant contributors to all the FOMs in a SWaP-C analysis since they can be large, and electromechanical components are also the bottlenecks for maximizing system quality and reliability.

Power solutions do not scale at the same rates we see for items on the load side, such as those driven by Moore’s Law and microelectromechanical system devices. This means that system roadmaps cannot plan for an exponential reduction of power solution size (or exponential increase in power density, conversely) due to nearly year-over-year process node improvements. That being said, a power solutions can help keep the pace of enhanced load size/performance by meeting the increasing demands of loads in its own way [2].

Feature Set Plays a Huge Role in Power Density

Look at the picture of this power supply (above) with its chassis cover removed. Does it look like most of the volume in this box is consumed with the actual power supply components, or does it look like the connectors, wiring, fans, and even the heat sinks and enclosures consume the majority of the volume (accompanied by empty space)? Sometimes, it is quite surprising to learn what little impact the actual power train has on the overall power solution volume and, therefore, the overall maximum power density achievable. Again, this is why a metric such as $/W is inappropriate for just about any power supply design evaluation that is not made to be extremely apples to apples (e.g., features nearly all identical, with the only major difference being the power rating of the power train components).

Requirements driven by safety certification and support for needs such as high-voltage inputs (e.g., increased 2D and 3D spacing requirements) and harsher operating environments can greatly impact solution density. If more stringent electromagnetic compatibility (EMC) and/or shock-and-vibe levels must be met, such as for applications requiring network equipment-building system certification, bulk, and volume are consumed by larger filter components along with enhanced mechanical support used to secure larger-mass components. This may require some adhesive/sealant (liquid silicone rubber that solidifies at room temperature; hence, “room temperature vulcanizing” (RTV) compound is common for this use), strapping, and even full potting (i.e., totally immersing the solution in an epoxy/polymer material to facilitate thermal transfer and to insulate both electrically and from exposure to the external environment). All these larger components and retrofitting of supporting materials for safety/certification and/or thermal/environmental support contribute to the overall weight of the solution and, therefore, density metrics.

Given the comprehensive quality and accelerated life-testing power solutions endure, in addition to functional electrical bench qualification testing, these test setups and pass requirements should be taken into consideration in the design phase and how a test plan may be impacted. If a long-term expensive (in terms of $ and schedule time) qualification test is being run, assuring a pass the first time is a goal, but this can be hampered by more frequent failures of larger and/or more complex designs, as well as the ability to access assemblies and components upon failure to perform diligence in the failure analysis, which drives the appropriate corrective action.

Opportunities for Density-Driven Power Solution SWaP Improvement

The biggest contributors to driving SWaP metrics are also the biggest opportunities for improving the associated FOMs. Namely, the primary contributors are the filter components, electromechanical components, and anything required to support the mass of these larger and/or looser components. Identifying these factors and isolating the components and their individual contributions to system design enables designers to focus their overall optimization on a series of subtasks and dedicated testing for validation.

The components calculated and selected for filters to meet EMC requirements are frequently at the top of the list of items to focus on. Large capacitors and even larger/denser magnetics are commonly the worst offenders, but surprisingly, they tend to get less attention when looking to optimize because many designers are not comfortable with filter design. While filter design can be quite subjective and even an art in more complicated solutions, it is highly recommended that every designer, ever even remotely associated with power design/qualification, familiarize themselves with some basic training on filter design and optimization [3] [4] [5]. The key trade-offs are between the filter component FOMs (better performance tends to be with larger/heavier components) and acceptable compliance levels (typically in terms of emissions levels). NOTE: The best strategy for dealing with undesired energy for a given frequency is mitigation. In other words, try to optimize the design to eliminate or reduce sources of noise before putting the maniacal focus on filtering to deal with and capture such undesired excursions. An example of this is if a power driver/controller supports spread-spectrum clocking to help spread the energy across a broader frequency spectrum, thus reducing the need for heavier-duty filtering [6].

Disaggregating power subsystems can also be a great methodology for improving density. Separating power solutions may seem a little counterintuitive in discussions driven by integration to improve density metrics, but there are times when it can reach a point of diminishing returns when trying to cram too much functionality into a single solution. Particularly given all the knobs and variables in power supply design, it can sometimes make more sense to take more of a “divide and conquer” approach. An example is a system power rail that requires support for a wide range on the input side and isolation and/or tight regulation on the output side, which may be optimally suited by separate solutions, with one optimized for a wide input ratio followed by another optimized for regulation/isolation. Another common example is taking a large single-phase converter and changing it into smaller multi-phase converters that each process less power, thus allowing for smaller components, reduced electrical/thermal stresses, and even an opportunity for pushing switching frequencies to further improve component FOMs.

Whether optimizing filter values, individual components, or the most effective disaggregated solutions, there are certainly all kinds of solutions to assist a designer in meeting these targets while still taking advantage of state-of-the-art (SOTA), particularly with commercial off-the-shelf (COTS) solutions. Major advancements in the area of three-dimensional power packaging (3DPPR), particularly for lower-voltage, DC/DC power converters are a sweet spot. Advanced packaging techniques have facilitated power conversion and power management solutions with the ability to take advantage of many of the SOTA technologies tabled above and integrate them into highly dense integrated components. Filter components, in particular, are heterogeneously integrated into power modules in the form of planar magnetics, packaging over molds, and multichip modules. 3DPPR allows the best of these technologies to contribute to SWaP optimization while still taking advantage of access to COTS solutions.

Conclusion

There is no Moore’s law for power solutions, especially when considering the energy storage devices that tend to dominate SWaP-C metrics, which in turn drive nearly all knobs to determine power density and overall system density. Packaging tends to be a very big driver that enables power solutions (particularly modules and other ready-made products) to help keep up with advancements in density on the load side of things.

Chasing power density just for the sake of an improved metric (such as W/m3) can be a costly endeavor that comes with many project tradeoffs, ranging from increased cost and development time to decreased efficiency and reliability. It is important to consider the true impacts of desired features and to consider whether the impacts on cost, space, and efficiency (and project schedule, of course) are truly justified in the application at hand.

That being said, it has also been reviewed how driving density with advanced packaging and taking advantage of 3DPPR techniques and manufacturing automation can improve SWaP-C metrics, so there are no absolutes here. Increasing design complexity generally leads to design risk in terms of decreased manufacturing yield (or extra rework, thus slowing production and adding cost), but automating otherwise manual assembly processes may enable highly integrated solutions of more tightly controlled process steps that can enhance reliability while concurrently driving power module density. The growing use of planar magnetics is a perfect example of this.

Increased power density tends to drive more challenges in any thermal mitigation strategy. The more heat trapped in a smaller space makes effectively getting that heat spread and transferred to the ambient environment more challenging. When not transferring the heat out as effectively, components see a greater temperature rise, which translates to decreased reliability. This is where it is important to consider the holistic impacts of power design on the overall system to ensure that the pursuit of power module density does not sacrifice other SWaP-C targets for the design. Particularly when it comes to thermals and quality-related product lifetime predictions, density can impact warranty analyses and support costs.

References

[1] “Power Supply Design for maximum Performance,” RECOM Blog, Oct 21, 2022, https://recom-power.com/en/rec-n-power-supply-design-for-maximum-performance-229.html (accessed February 15, 2023).
[2] “Power Modules are Catching up with Moore’s Law,” RECOM Blog, Nov 11, 2022, https://recom-power.com/en/company/newsroom/blog/rec-n-power-modules-are-catching-up-with-moores-law-235.html (accessed March 13, 2023).
[3] “Specifying line inductors for power converter noise filters,” RECOM White Paper, https://recom-power.com/en/support/technical-resources/whitepaper/whitepaper-specifying-line-inductors/whitepaper-specifying-line-inductors.html.
[4] “Very low noise filter for isolated DC/DC converters,” RECOM Blog, Mar 4, 2019, https://recom-power.com/en/company/newsroom/blog/rec-n-very-low-noise-filter-for-isolated-dc!sdc-converters-46.html (accessed March 13, 2023).
[5] S. Roberts, “DC/DC BOOK OF KNOWLEDGE – Practical tips for the User,” Fifth Edition, RECOM Engineering, 2021.
[6] Wikipedia contributors, “Spread spectrum,” Wikipedia, The Free Encyclopedia, https://en.wikipedia.org/w/index.php?title=Spread_spectrum&oldid=1138317993 (accessed March 13, 2023).