Jim - you are correct! I didn't think the circuit through enough. An all-analog solution may not be easily possible.
You can definitely use an ADC, microprocessor, and DAC to do this, provided they are fast enough to take enough samples along the waveform to get a good idea of the minimum and maximum value. What this number is, I suspect the customer would have to choose that to be comfortable. If you only take 2 samples per cycle, you're not likely to get the min and max. 100 samples per cycle might be too many. This is something to consider.
Also, how often does this signal change frequency? If the frequency changes from low to high, the number of samples per cycle may not be enough. Speculation - if the goal is to look at the minimum and maximum of the last 500 readings of one 400kHz cycle, you will get a good idea of the shape and peaks of the waveform. With the same sampling rate and a 200MHz you will get only one sample per cycle, which will tell you basically nothing about the min/max of the signal.
Next, how does the signal change in amplitude, and how often would you want to update the DAC? You could do a min/max of the last (some number) of readings, but again you run into the sampling frequency issues and post-processing problems.
A smart algorithm in the microprocessor could look at the rise and fall in real time, and only store the most recent highest reading and lowest reading, and when the waveform changes from rise to fall or fall to rise, then it would know that a max or min has been reached for that cycle. How many cycles of data do you want to average to get the DC value? How often do you want to update the DAC output?
These are things I would certainly consider for this project.
May I ask what the application is, and the customer, and the potential volume, and any cost or design restrictions?
Thank you!
Jason Bridgmon