Today, non-isolated point-of-load (niPOL) dc-dc converter modules have advanced to a point where they are a common component on almost every circuit board in the proximity of the load. To meet the needs of constantly shrinking board space and increasing demand for higher efficiency, these devices have evolved to encapsulate magnetics, power semiconductor ICs, MOSFETs, and passive components inside the same miniature housing.
Suppliers are calling them power-supplies-in-a-package, or PSiPs. With an integrated inductor in a PSiP or without an inductor as a power-supply-on-a-chip (PwrSOC), these high-density integrated devices have enabled distributed power architecture (DPA) designs and dramatically reduced the overall power board space requirements.
Key Challenges
Traditionally, a DPA system converts an incoming ac to a dc bus, which in turn powers independent isolated dc-dc converters used to drive the output loads. Although this method generates higher overall efficiency due to single-stage conversion for each output rail, it results in higher costs and takes up significantly more board space.
The problem gets worse when the number of IC loads increases on a board (CPUs, DSPs, FPGAs, logic, memory), a common scenario in modern systems. By converting the DPA’s high dc bus voltage (48 V or higher) down to an intermediate level (8 to 12 V) that is adequate to drive a number of narrow-range niPOLs, the intermediate bus architecture (IBA) overcomes the limitations of traditional DPA schemes.
In an IBA scheme, each niPOL then operates from the intermediate bus voltage and produces the desired regulated output voltages for each IC load on the circuit board. While the IBA reduces board area and cost compared to a traditional DPA system, the duplication of conversions reduces overall efficiency. As a result, it is unable to fully exploit the size and efficiency advantages of niPOLs.
A more ideal power solution would be to blend the niPOL’s compact size with the built-in isolation of an isolated dc-dc converter. For example, in a typical communications line card, the galvanic isolation required for the dc-dc converter is usually offered by the intermediate bus converter (IBC) stepping 48 V down to 12 V, followed by a number of niPOLs.
Now, replacing niPOLs with isolated POLs or iPOLs in similar PSiPs could eliminate the IBC altogether. The end result would be the overall efficiency of a true DPA with the space and cost advantages of the IBA. In other words, utilizing iPOLs in distributed design can bring the best of both DPA and IBA schemes in a single, more efficient architecture.
In essence, the use of iPOLs offers the opportunity to design a truer DPA system that doesn’t rely on a centralized power source like an IBC. In other words, an iPOL enables a more “ideal” point-of-load solution.
Take, for example, an automated test equipment (ATE) system where isolation is a key requirement. In this application, an iPOL could concurrently provide isolation, conversion, and regulation directly at the test head (or load). Aside from realizing higher efficiency and smaller board space due to a reduction in power conversion stages, the iPOL also enables routing of higher voltage on the circuit board. Busing a higher voltage lowers IR losses, which ultimately translates into higher efficiency.
However, the design and construction of an iPOL package similar in form factor to a niPOL PSiP comes with many challenges and added complexities. Development of this type of iPOL design requires attention to the controller technology, isolated transformer (integrated within the package), MOSFETs, and packaging.
In fact, accomplishing such a task requires system-level understanding with the ability to achieve an optimal balance between size, power density, packaging, thermal management, and the integrated power-supply topology that works optimally with the semiconductor switching controller. This expertise typically requires a blend of traditional brick design and advanced IC design technologies.
Traditional brick designs have made strides in reducing open-frame or brick-type supplies, but most have not been able to achieve the density and IC-like packaging that an engineer would expect in an iPOL. Hence, the packaging advances on the isolated front have lagged that of niPOLs.
From half-brick to smaller quarter- and eighth-brick sizes, designers have cut the size to a sixteenth-brick form factor. But, by comparison, the sixteenth brick is still two to three times larger than a high-performance semiconductor IC-like package used for niPOL devices. From an IC approach, we have seen limited evidence of an integrated, isolation transformer of any significant power level.
What’s Next?
Needless to say, iPOLs do exist and continue to improve to deliver increasing performance and/or power density. Just as suppliers have continued to advance niPOLs to deliver them in semiconductor packages, the efficiency, density, and cost performance of iPOL dc-dc converters will also continue to improve, further increasing their use in power designs. Ultimately, I can envision iPOLs in niPOL-like dense packages to completely alter the traditional way of doing distributed power designs.
Consequently, when it comes to cost, board space, and efficiency, we know that the use of niPOLs alone in IBA architecture cannot guarantee the most efficient solution. Consideration should be given to blending the attributes of DPA and IBA and deploying IBCs, niPOLs, and iPOLs in a combination. This design approach allows for greater flexibility, power density, and efficiency not possible with traditional DPA schemes of IBCs and niPOLs alone.