Reference designs serve an important purpose, but can also incite misgivings
In the beginning, there was the circuit "cookbook" such as the classic IC Timer Cookbook for the legendary 555 timer IC. The typical cookbook had dozens of basic schematics, each solving a specific, simple problem or objective, such as "Electronic lock has variable timeout".
Most cookbooks included a circuit schematic, bill of materials, and a short description of the circuit function and operation, but had no layout or performance analysis. Further, many of these cookbook designs had not actually been built (I'm sure you are shocked!), but were based on what we'll call politely "speculative design."
Despite their brevity, these cookbook circuits were—and still are—a good source of ideas and starting points for other circuits. Their simplicity, though, is also their weakness: if you wanted or needed to modify the circuit to do something different, or to use the parts you had available, that could be difficult: you often didn't have enough explanation of the operational subtleties to understand what was critical, what was not, what could be changed, and what had to remain as shown. The relationship between the engineer or circuit experimenter was often (pick one or more): "I need it, I like it, it's driving me crazy, I don’t understand why it doesn't work, I need it to do more."
But engineers today demand much more. Now, a vendor can't announce a new part, even a basic building block such as an op amp, without a full-on reference design. These designs go well beyond the cookbook by proving lots of information, including PCB layout, performance curves, photos to verify that the design was actually built, Spice and other models, a description of the operation and the role of key components, driver software, and validation software.
These reference designs serve several purposes: they speed the project's time to market, reduce the need to re-invent the design wheel, and—important for the vendor--shows that the part actually can meet its specifications with the right layout. They are a pre-emptive answer to the user saying "your part doesn’t meet its spec."
So far, so good. But in the mind of many engineers, the problem with reference designs is that they are perhaps too good. After all, who needs a skilled engineer to design and lay out a circuit, when that fine-tuned reference design is waiting there? And if the project team and its competitors all have access to that same reference design, what's the competitive edge going to be? Is the engineers' role reduced to system integration, software, and packaging?
At the same time, reference designs themselves have changed and improved. Some, like the Webench family from National Semiconductor Corp. (click here), allow the engineer to use interactive online tools to adjust circuit parameters and tradeoffs, to find a circuit configuration which hits the "sweet spot" among conflicting goals of performance, power and efficiency, and pricing for the ICs and their associated passive components.
The latest reference design tool goes in a somewhat different direction. Their just-announced Circuits from the Lab from Analog Devices (click here and here for a pair of overviews) starts with a reference design, then adds detailed circuit tutorial, FAQs, PCB details, and more, all rolled together. It addresses both the "how?" and "what if?" of the topology and circuit details.
Reference designs are mostly good news for the hard-working, deadline-constrained designer, when they are done with flexibility and insight. If they are from a reputable source, they let you start with a known working circuit, layout, and BOM, and then make the changes and choices you need to better meet your project priorities. But one from a less-than-qualified source can lead you astray or constrain you unnecessarily, or trap you in blissful ignorance that backfires.
What's been your experience with reference designs? Where do you stand in terms of loving, hating, embracing, or shunning them? And have you ever been misled or even "burned" by one? ♦