These days, nearly every released component is supported by vendor tools such as Spice or other models, evaluation software, critical code segments, and reference designs.
The term “reference design” covers a lot of range, of course. It can be a basic “how to apply” schematic; a detailed schematic with bill of materials (BOM); or, more likely, a schematic, BOM, PC-board layout, operating and test code, and more. Even a basic building block such as an op amp usually comes with some suggested circuits for likely applications, if not full reference designs.
You can't blame a design engineer for wanting to have reference designs. After all, who wants to re-invent the wheel? Who, in theory, should know more about a new component than its designer? There are occasional lapses in good quality by some vendors.
But that's the exception. Reference designs make a lot of sense, especially for a constrained application which is fully defined by an industry standard, such as one in the IEEE 802.xx family, where it is easier to evaluate a successful outcome.
A reference design, especially for analog/RF/power-related circuits, can be a good start but may also lull the designer into underestimating the actual project, for many reasons:
“…perhaps just one board was built…”
Just because you see the design in an application note doesn't mean that it was actually built, nor that it was adequately tested (don't be shocked). Or perhaps just one board was built and run through its paces, but the design was not evaluated for component tolerances, temperature drift, and other real-world factors which afflict designs;
- A successful analog design requires lots of passive components, and many these of the have critical first- and second-tier parameters. For example, it is not enough to know that a 4.7mH inductor was used and thus is called out on the schematic and BOM; you usually also need to know its DC resistance, frequency-related specs, and more. So the reference design really needs to identify the vendor and model number of the part, so you can check out those other parameters;
- While there is a lot of commonality among analog applications, there are also critical differences which may make the reference design an “almost OK” fit rather than a really good match — and the gap may be enough to require careful consideration. For example, a reference design for a 300-watt motor drive/control may not be a good fit for a 350-watt motor situation, although it may be a good starting point. Many things may need to be changed when going from 300 watts to 350 watts, depending on the margins and headroom of the lower-power design;
- Finally, layout, layout, layout. Even if the reference design has been vetted and includes a PC board layout, most applications will also require other circuitry which is unique to the end-application in addition to the reference design. That other circuitry brings with it subtleties in power, grounding, system noise, IR drop, EMI, thermal drift… the list goes on. So even a very good, properly qualified reference design may need some moderate-to-major tweaking based on the ripple-effect of the rest of the design.
There's nothing wrong with using a properly designed and evaluated reference design. But it's a naïve engineer, or one under serious schedule pressure, or one with a manager who believes in miracles, who thinks such a design can just be dropped in and that's that, end of story. In some ways, a less-complete reference design is more realistic, because it doesn't lull its user into overconfidence with the veneer of completeness, and the engineer knows that a lot more work will be needed.
What's been your experience with using reference designs? Have you ever been able to use one “as-is”? Have you ever been badly mislead by one? Have they given you a good head start towards project completion?
- Can We Have Too Many Op-Amp Performance Curves?
- Do We Really Need Yet Another Op-Amp?
- Can a Car Be ‘Over-Sensored’?
- What’s Your Breadboard Look Like?
- What Analog’s ‘Imperfections’ Taught Me
- Projects vs. Products: What That First Real Engineering Job Actually Teaches
- When Engineers Are a Little Too Smart for Themselves