Last month, in response to my post The Other Side of the Integration Coin, reader TheMeasurementBlues asked if we could consider discrete op-amp or data converter devices as opposed to ICs. My first response was to consider the question from the manufacturer’s perspective. From this view, I asserted that “vendors design, verify, fabricate, and package these devices as they do more complex ICs. In that regard, they are subject to the same kinds of challenges and benefit from the same kinds of advancements as do more complex analog and mixed-signal blocks.”
Looking at the same question from the system designer’s perspective, I came to the same conclusion but for a completely different reason: “Integrated devices,” I wrote, “provide barriers of abstraction that relieve designers from a great deal of technical detail compared to discrete designs. This changes the way we design with integrated functional blocks, be they as conceptually simple as an op amp or as complex as a complete AFE.”
Throughout this discussion and, indeed, its precursors, the assumption I had made was that our conversations about integrated circuits were inherently about monolithic devices — that the terms “integrated” and “monolithic” are essentially synonymous. A recent product announcement, however, has made me question that assumption.
One of the many new-product announcements at APEC (the Applied Power-Electronics Conference and Exposition) this year was one from Vicor, which introduced a new packaging technology that the company has dubbed “Converter housed in Package” or ChiP technology. According to Vicor VI chip product line vice president Stephen Oliver, the company is making available devices that exploit this packaging technology in sizes ranging from 1323 (13 x 23mm) to 6123 (61 x 23mm). These are not the smallest set of footprints in the power converter market, but impressive when coupled with their high-power densities (up to 3kW/in3 ), high-area densities (up to 850W/in2 ), and efficiencies (up to 98%).
The packages can house a variety of complex power-management functions including AC-DC conversion with PFC (power-factor correction); isolated and non-isolated DC-DC conversion; buck, boost, and buck-boost regulation; and PoL current multiplication. Devices in the ChiP form factors can include microcontrollers with the power train to implement factory- and user-defined functions, command sets, and telemetry capabilities.
Like large ICs, members of the ChiP family can take advantage of several cooling methods depending on their worst-case operational dissipation and the system’s thermal design: These include direct single- or dual-side cooling plus thermal conduction through the leads, cold-plate cooling, or cooling with standardized finned heatsinks.
Clearly, these are not monolithic devices. At minimum they contain a power-train silicon and off-chip passives — notably magnetics. In their larger forms, they might contain a controller IC, MOS switches, passives, and logic silicon.
But from a system-designer’s perspective, do I care how many individual devices the package contains, or am I more interested in the nature of the package as a unit object? Assuming that commercially available products — be they monolithic or otherwise — provide adequate reliability, more and more I’m thinking that I’m interested in the package as a unit object: It embodies a great deal of engineering value and provides a barrier of abstraction that relieves me from a great deal of technical detail compared to a discrete design. And that, recall, is a key attribute of integrated circuits.
What do you think? What attributes are defining when we refer to integration? Is it monolithicness [sic]? Or does the barrier of abstraction prevail and, in so doing, relieve us from needing to know about the underlying implementation? In an age when modular components and monolithic ICs may overlap in footprint, mounting method, cooling method, design-in procedure, and evaluation process, is there a distinction in the difference?