I developed an automatic mixer using DPOTs for gain control, for an analog mixer company that abandoned the project. Automatic mixers involve similar gain manipulation to dynamics processing.
I laid out my prototype to use either DPOTs or VCAs, I wasn't sure DPOTs would sound OK. I populated the DPOTs first and they sounded good enough that I never populated the VCAs.
I forget all the details but IIRC the automatic mixer had 6 mono input channels and at least one stereo line input (maybe two). The stereo channels were treated as linked and both L and R DPOTs got the exact same control signal.
I was worried about zipper noise so put in software hooks for zero crossing coordination. If you only execute the gain changes during zero crossings there are no (less) clicks. It turned out that the DPOTs sounded good enough without zero crossing synchronization, so I never implemented that, on my bread board.
Sadly this product never saw daylight. The analog mixer company I designed this for got crushed by the wave of cheap digital mixers that didn't suck.
For development I was mainly performing AM (automatic mixing) on speech signals grabbed from TV, with additional music inputs to exercise the gain sharing.
I expect if this moved beyond my first cut prototype it would have been exposed to more critical testing.
IIRC I was using 12b DPOTs and a 16b microchip microprocessor. In hindsight updating the gain settings very frequently kept audible perturbations minimal. I also used some trick HF pre/de emphasis. If you pre-emphasize HF content before the DPOT and de-emphasize HF after, you de-emphasize clicks.
It sounded better than I expected, but never got tested in the market place.
JR
PS: Modern CODECs are so cheap this could be done completely in the digital domain for near perfect dynamics crunching.
