Prompted by my recent build of a DC reference, and another freak's mention of a video calibrator I've been wondering about AC calibration, in particular in the context of audio levels. A quick simulation...
...shows exactly what you'd expect. The precision elements would be the voltage reference, V1, and the two resistors R1 and R2. U1 just needs to make a decent virtual earth mix point and plays no real part in the precision of the unit. By using an AVR to drive the switch, V5, you can make the switching frequency have a non-integer relationship to the input signal which means that the output signal, when free-running on a 'scope, will not appear 'locked'. All you need to do is observe where the positive peaks and negative peaks meet and adjust the input level until they touch.
It should be trivial to reach 0.5% basic accuracy which is a 0.04dB error.
Can anyone see any problems?