I've been using circuit A as a peak detector with a bleeding resistor (R38/R40) to discharge the storage cap (C18/C20) over time and an extra (R39/R41, C19, C21) low pass filter stage to eliminate any ripple, but am wondering if there's any reason NOT to switch to design B.
In A a base-emitter junction of an NPN transistor acts as a diode for a peak detector, because this BE drop is often more precisely tuned at manufacture and more predictably temperature variable than the drop of an actual diode. The collector is connected to the base simply to prevent it floating and maybe doing something weird.
B would have the advantage of a higher input impedance, whereas design A is such that at the very peak of an incoming waveform C18 presents almost zero resistance so a relatively large current very briefly flows. If the source supplying the signal had a series resistance within it then B would lose less voltage and have the capacitor charged up right the way to ( V_peak - V_B_E_drop ) , with no source resistance losses affecting V_peak, and as far as I can understand would otherwise be exactly the same in behaviour. Any noise or change in voltage on the power rail would seem to have no effect, as the power rail would just be supplying current for the emitter to output at a voltage of ( V_peak - V_B_E_drop ).
Can anyone suggest circumstances where B would be inferior, or problems it can have that A is immune to?
Thank you