Using a high speed op-amp, in this case a HA2525, I built a "rectifier" circuit as an AM detector. Idea is that it should be more linear than an ordinary diode. The HA2525 has a slew rate of 120v/uS, and enough bandwidth to do unity gain to around 10MHz or more. You can find the datasheet here.

This particular op-amp circuit outputs the negative going parts of the RF carrier and the amplitude modulation. Instead of a curved diode knee, this circuit would do a much sharper knee, thus better detection of weak signals and also strong signals (better audio fidelity on music).
Like this (though the polarities are different)
vs for a diode


The RC circuits fed by the op-amp's output filters out the RF part of the rectified waveform to get audio, and AGC voltage.

I took Ian Robertson's advice to make it a peak detector. Added a 3nF cap to the circuit output, and some resistance to the positive supply, here a 100K resistor. Also a 270K resistor to the + input of the op-amp, which also has a 910 ohm resistor to ground, used to minimise the error caused by the input bias current. Found that another resistance to the positive supply, a 270K resistor got the circuit in the sweet spot of detection. I think that these resistors are making up for the fact that my op-amp power supplies are not exactly the same absolute value voltages, and for the AGC line IF amp bias circuit. And it's not at all fussy, varying supply voltages doesn't mess it up. Substantially self biasing. I never much liked having to use trim pots, which usually means that the circuit isn't self bias correcting, and can easily get out of adjustment. I kept the 2nd diode, it acts like a catcher for the op-amp's output pin so it doesn't have to slew back so far when it comes time to rectify again .

Maybe you should try a standard op-amp peak detector circuit. It's self-biassing - no pots needed. 1 op-amp, 1 diode, 1 cap., 1 resistor.

Good stuff, guys ..... interesting ....and challenging ....

Actually, this thread got me wondering......

I think you could now use DSP to demodulate DSB AM. You could reject anything that wasn't identical in both sidebands, thus cutting out noise, interference etc. and getting a full audio bandwidth output Performance could be better than FM.

Remember, you saw it here first!!!