Bill Schweber, my friend and mentor in my early career as a technical editor at EDN, writes a weekly newsletter to support the Planet Analog web site he edits at TechInsights. Nothing new about writing a newsletter, most other sites do it as a way to increase visitors count, but Bill almost always shares an insight you will not find on the Planet Analog page. This week he shares his opinion on the use of the term "SoC".

Since you might not have received the newsletter, here is the relevant part:

What's annoying me lately: I'm annoyed by all this attention that the SoC (system on a chip) is getting, as if it is new development, or the end-game of IC design. Hey, folks, the SoC is nothing new, it's all a matter of what you define as your "system". One person's system is another person's subsystem! It all depends on where you are in the design chain. For example, I was at a major audio event a few years ago, they kept talking about audio "components", and what they meant by "component" was a rack-mounted chassis, not an IC, and not even a PCB (ed. Printed Circuit Board).

Sorry to be cynical, but to me, a lot of this SoC talk is just an attempt to create marketing buzz. About a million years ago, as microprocessors were starting to rule the land, I was involved in a 12-bit ADC IC product release. The virtue of this new ADC was that it was double-buffered, so the digital outputs were isolated from the conversion process, and you would not get data transitions when you read out the ADC. That was a big plus for users, so we said the ADC was a "complete system" on a chip. Then the next year, we added an internal voltage reference to the ADC, so we joking called it "the really, truly complete ADC SoC".

And it's not just analog, either: we've seen the same SoC and "complete" hype as processors absorbed and integrated more and more of the associated non-CPU functions, all the way to external I/O, memory, and more. So please, enough of this SoC talk as if it's something new, or we are approaching the single-IC total solution. To me, for a real SoC, you would just hook up the power rail, the input, the output, but no passives, and the product is done and ready to ship. I don't think so.

I think that most engineers will readily agree with Bill, but to make money companies need to commercialize the technology, and EDA vendors have a very hard job convincing their customers to purchase new tools. So they employ marketing professionals whose job it is to find and develop new market segments. Thus SoC was created as a term to define a new market segment. But is it really new, as Bill asks?

Of course not. An SoC is just a really small PCB. An SoC contains cores, often developed by third parties, firmware, most often some analog circuitry, a bunch of memory, some internal busses, and buffers and I/O drivers to interface with the outside.
This description fits both a PCB and a SoC. So why not call it PCBoC (a PCB on a Chip). Well it does not sound as good, and it elevates PCB designers to the same level of chip designers, something the latter group finds unfair. After all, real men build real small transistors and put lots of them together, or so I am told. Bill is correct: a system turns out to be a subsystem of a bigger system. A core is a system, an SoC is a system, the PCB the SoC is on is a system, the device that uses the PCB is a system, and in fact the person and the cell phone together are a system.

So remember that the same people that gave you the SoC also gave you DFM (Design For Manufacturing) which one can only compare to DFF (Design For Failure).