The main variables were volume (onstage and in the venue, separately), delay and reverb. We would start with a dry (no effects) sound and get the volume right first. This would already be kinda stressful for her as a dry sound is raw and unkind! We would then bring up the effects to a normal level which would sound way too extreme in soundcheck, as there is no crowd to absorb the sound - so again it wouldn't sound good to her. Then we would bring the effects down in order for it to sound right for an empty room, which felt more comfortable to her. Then for the performance the volume and effects levels would be higher to account for the crowd absorbing the sound.

So for soundcheck it was important for it to sound right for an empty room, whereas we were also somewhat covertly tuning parameters in anticipation of the crowd's effect on the acoustics.

So the key point is that we were building two sounds; one to fit her psychoacoustic perception in an empty room to make for a comfortable soundcheck, and another for the actual acoustic scenario of the performance. The latter was offputting for the performer despite being 'correct'.

Very interesting! Thank you for the detailed reply. I have always wondered about the detail that went into soundcheck. Sound often seems too loud in a bar venue. I don't know if that's where you are more likely to have an amateur sound tech or it is just difficult to get a loud enough sound in a small space. Or maybe I'm just a little baby when it comes to sound.

To get a good sound, or particularly a good loudness, in a small place is a lot to do with the proportionality of the speaker power output relative to the room size. You can have huge speakers sound awful in a small room and vice versa.

Levelling the playing field in 'amateur sound tech' (mostly good craftspeople who are just doing what they can with limited time/resources) is an active area of research in acoustics and DSP. I recommend Dr Josh Reiss of QMUL (where I'm based) for more on this. Check out his work on automatic audio engineering. Machine learning is very appropriate to this area of work:

https://scholar.google.co.uk/citations?hl=en&user=fVlS_EgAAA...