The transition from one year to the next is filled with sages reviewing the past years’ events, and pundits predicting the new years’ developments. I’ll do neither.
But I will make one plea to designers everywhere: Whenever possible, take some time to look at unconventional approaches to the problems you are working on. Or, as the folks at Apple said in their recent ad campaign, “think different.” (I know, it should have been “think differently”, but I suppose that not following the rules of grammar shows they are taking their own advice.)
For example, look at the Roomba floor vacuum from iRobot, the first home robot to enjoy widespread commercial success, with over a million units sold in just a few years. What are two of the ingredients to the Roomba’s success, in addition to a good overall design?
First, although it is a “robotic” unit, it doesn’t try to do many functions. It is tightly focused in its objective. It does one thing, it does it well, and it does nothing else. This focus translates into ease of use and a streamlined design.
Second, the designers turned the cleaning path algorithm strategy around. Rather than have the Roomba somehow determine an optimum work path in order to minimize vacuum time, and thus power usage—a task which would require somehow mapping the room with sensors and a complex algorithm–they used a different approach. As long as the Roomba battery could support the system for several hours per charge, they opted to have it to the dumb but effective random walk through the room, over and over. This requires no calculations, and no room sensing and mapping. While the overall coverage path trace has many overlaps and redundancies—so what? The tradeoff is well worth it.
I can give you a purely electronic example as well. A student emailed me several months back, asking for some advice on component selection for a fairly accurate peak detector for 10 MHz signals. The design he planned out required a fast, high-resolution A/D converter streaming its output data to a processor running a tight software loop, which would compare new samples to older ones, so as to decide when the signal had peaked and at what value. His design’s error budget didn’t have much slack, the power consumption was high, and the parts cost and PCB footprint were large.
I suggested he abandon the A/D and processor approach, and instead use an analog peak detector. For a few dollars, he could use a high-performance op amp, voltage hold capacitor, and diode—the core of the circuit—to achieve his goal, and with very low power consumption, small footprint, and near-zero processor load.
But what shocked me, although it probably shouldn’t have, is that the student simply could not understand what I was trying to explain to him. It’s not that he didn’t know analog circuitry and how to use an op amp to build a peak detector; I could accept that, since it is not a common circuit. It’s that he simply could not grasp the basic concept that the signal could be assessed and processed using all-analog circuitry. In his world, if the signal was not in digital form, it was useless. Its analog existence was a very temporary state, until he could apply the A/D muscle to it and transfer it to the digital world.
This is why designers need to step back and ask themselves if there is another approach to the design or to achieving the design’s ultimate objective. And it is the same reason that smart designers still use discrete transistors such as the 2N2222 in this era of huge digital ICs–they offer a nice, cheap, basic solution to many circuit needs.