Hello huntest,
It's the same as with an Orange Squeezer. Q1 is used as a variable resistor for the compressor. The bias totally depends on the JFET you use. For it to work properly as a variable resistor, it needs to be put into the linear region, meaning it needs to be barely "on" when there's no audio signal. The more audio signal goes into it, the more "on" it will go.
If it's biased so it's off when it's idle, then it's no longer in the linear region. So that's the wrong bias then.
George Giblet provided a way to bias the Orange Squeezer, see
this link. It's easy to follow.
Alternatively, if you know Vgs,off from your JFET, you could just measure the voltage between gate and source of the JFET and make sure it's barely on when idle (no audio signal). So for example, if your Vgs,off is -1,50V, then use the trim pot to adjust the voltage between gate and source to -1,49V while no audio signal goes to the JFET. Then it's just barely "on" and the Median is biased properly.
