I recently received an email from a student who needed to design a circuit to detect peaks in a 10-MHz analog signal. His plan was to use an A/D converter sampling at well beyond the Nyquist frequency of 20 Msps, and then pass the successive converter outputs to a fast processor running a tight, high-priority software loop while disallowing interrupts. In short, his approach was expensive, power hungry, and had a large footprint.
While I hesitate to offer free design advice (such free advice is often worth what the recipient pays for it), I did suggest an alternative approach: the classic analog peak detector. Using a fast, precision op amp, a few low-leakage diodes, and a low-leakage capacitor, you can have the core of a very nice peak-detector circuit for about $5. It will be low power and have minimum real-estate needs; it's also fast, cheap, reliable, and easy to test. What's not to like?
Unfortunately, the student–a potential practicing EE–simply could not grasp the concept of an analog-based peak detector. It's not that he wasn't familiar with this particular circuit (I can understand that) but all he could think about was getting the signal into digital format so it could be properly assessed. The idea of computing and analyzing in the analog domain was absolutely foreign to him; as they say, for him it “does not compute.” This is too bad, because for many signal-processing needs, analog is the better, and often only, way to go. If you need a logarithmic (log) amplifier, an rms-to-dc converter for line power or RF signals, think analog. You may have no choice.
The peak-detector topology was a popular analog circuit back in the days before processors ruled our design world. For example, the data sheet for the OPA111 op amp from Burr-Brown (now part of Texas Instruments) showed a possible circuit (http://focus.ti.com/lit/ds/symlink/opa111.pdf), and Analog Devices offered the BNK01 IC (http://www.analog.com/en/prod/0,,773_925_PKD01,00.html) which I believe (I'm not sure here) was a monolithic version of an older, multichip Burr-Brown device. There some irony here, since today's op amps are so much better and cheaper than those of the past, that they can be used to implement really high-accuracy, high-speed, inexpensive peak detectors.
-x x x-