I can add two little additional tales, one.. why does my amplfier oscillate at random, answer when the wife is using the (wireless) phone in the next room, and two, why does my servo controlled DC offset change when the sun comes out answer- photoelectric charge in the bare film integrator capacitors. Obvious when you know, much head scratching till you do! Always expect the unexpected.
AD does mention that the 797 needs a little resistance at the input to avoid oscillation at low gains---an unfortunate thing, as it inevitably adds thermal noise. But they point out that most of the time users will be needing gain, so in most applications it's not that big a deal.
As far as keeping the bandwidth as low as possible--that's a half-truth. Given the right high-frequency modulation, you can get local rectification and spurious ~d.c. offsets, right down in your seismic signal passband. It's been noted that JFET input amplifiers are perhaps 100x less susceptible to this rectification than are bipolars. Sadly, most of them also have higher intrinsic voltage noise and higher temperature drift. You can roll your own from selected discretes, but that's anathema for volume production.
I believe it, having run into similar problems with wideband op amps being used as audio amplifiers. I was working on a home project with a low noise, high bandwidth op amp, an AD797, I think it was, and I started hearing the noise level rise and fall, along with some faint whistling sounds in my headphones, particularly when the gain was turned DOWN. When I looked at the datasheet, I realized that all specs were tested at gains of +5, -4, or higher (in absolute value). The datasheet did not say that the op amp was not unity gain stable, but evidently it wasn't! When I turned the gain down to near unity, it started oscillating at about 200 MHz rail-to-rail, so all bets were off as far as audio specs went! Moral of the story is the same as in James Hahn's story: more bandwidth is not necessarily better. Also, beware of specs that aren't in the datsheet.
That's one way of looking at it. I usually treat this as an EMI problem. Hardening a product against electromagnetic interference requires considering a lot more effects than just radio signals. Of course, this is an entire engineering sub-specialty, but I recommend reading the EN 61000 series of standards as a start, to get a comprehensive look at what effects and levels to protect against in a typical industrial environment.
Great article James. I remember as a young engineer working on sensitive RF circuits, I was shocked the first time I ever saw local TV stations on a broadband spectrum analyzer. The signals were so big, how could I keep them from mixing with my UHF local oscillator? I quickly learned that cardinal rule you mentioned -- keep the system bandwidth as small as possible.
Blog Doing Math in FPGAs Tom Burke 16 comments For a recent project, I explored doing "real" (that is, non-integer) math on a Spartan 3 FPGA. FPGAs, by their nature, do integer math. That is, there's no floating-point ...