OK – I thought I had it kind of working before, so I spent some time with it today.
First – as Nick mentioned – you need to crank up the input feeds for both instrument and voice. I had them almost pegged to test it.
Next – unplug the left output from the H8000FW and only use the right output (I’ll explain why in a minute). Use something like a large loud string section for the instrument feed.
Now play and talk – you should hear some vocoding going on. If you can’t hear it by talking, try plosives like making a clicking or popping noise with your mouth while holding down a chord.
What seems to be happening is that the vocoder presets are passing through the original signals and this overwhelms the vocoding effect that is mixed in with it. By unplugging the instrument output, you have a shot at hearing the vocoding effect.
Basically, I think the presets have a bug and/or the wrong ones are in the machine (the manual implies there are 96k versions of the vocoder presets, but they are not available in my machines). A standard vocoder should not produce any output if only 1 input is used, yet these presets allow you to hear voice and instrument separately. In addition, you really only need a mono output unless you want to provide some type of mega-mixer that allows a user to blend varying amounts of voice, instrument, and vocoded output and pan each of them.