A while ago, one of my colleagues was trying to map a simple, common analog circuit onto one of the programmable system-on-chip devices that my company makes.
It's an interface to a passive infra-red (PIR) sensor, and variants of it have been appearing in data sheets and app notes for decades, not only from the sensor manufacturers but also from opamp vendors who rightly see it as a good example of how to do something very non-critical using some very cheap amplifiers and some very pedestrian passives. And indeed he was able to fit it into this new, low-end product in our range. Two opamps, check. Two comparators, check.
A powerful CPU to do some simple calculations and drive some new-fangled digital interface — that's a bonus! The resulting circuit took almost as many components, almost as much board area and product volume and — don't tell anyone I said this — cost more. The figure shows the circuit (or one very much like it), extracted from the Murata PIR data book (which is freely available online, no secrets there).
A smart marketer once wrote that you shouldn't confuse your customer's compelling reason to buy with your compelling reason to sell. This looked like substitution for substitution's sake, to get our part number onto the very busy schematic. It seemed to me that it completely neglected the real reason why you might think of using such a cool new technology as ours, which is to implement a cool new way of solving the problem at hand!
Now, you might argue that for such a humdrum application, there's no economic return in changing from a design that already uses components with prices so low that you need double precision arithmetic to represent them. Well, perhaps not quite. But, by coincidence, I had also started looking at this very circuit with a view to addressing its major problem: It's too darn big!
The design contains seven capacitors (not including the rail bypass caps) of value up to 22 uF and about 15 resistors and two diodes. Even if we put the opamps and comparators into a bare die version of the target PSoC device, the board would still be way larger than the sensor — and I wanted it to be smaller than the sensor, for obvious reasons we don't need to go into.
Well — skip to the end, this is only a blog, after all — I figured out that with so much digital processing capability available, we ought not to need anything like that amount of auto-pilot analog passivity. By combining a lightly filtered low gain front end with a low power ADC and a digital servo loop to filter off baseline drift, I managed to chop the passive component count right down (to five, in the simplest case), and eliminate any capacitors higher in value than a few nF. And make the system's transfer function highly programmable for sensitivity and time response into the bargain.
One of my bright, young colleagues implemented and optimized it on a PSoC 4100, and, hey presto, a PIR interface that will fit on the rear surface of a PIR sensor, if you use the small bare die. We're going to write a more detailed article on that some time, but I only use this to make a point: Sometimes the old ways aren't the best ways. I'm an analog guy through and through by training and experience. But I'd be letting down my company and my customers if I wasn't open to the possibility that a new way of working might be better in some respects.
So, don't let people browbeat you into thinking that retro is the only respectable approach in analog. Paradigms shift discontinuously, and some analog engineers fear discontinuity. When you encounter a design problem, read between the lines. Or in this case, between the huge, old-fangled passive components.
There's a less-than-amusing coda to the story. I wanted to get some demo boards made up to show the solution off to some excited customers. We instructed an external group, which included several people with years of analog electronic experience, to do the job. But for some reason they never received the project describing the new design. They used their initiative and built… the old application circuit from the sensor data sheet. Hey ho.