I played around with the multiband ducker by Carla and found it interesting. But the backgroud pads were leaking through between some of the partials of the solo instrument.

What I needed was like the conventional EQ used to scoop out room in the spectrum for the solo instrument, with a fixed spectral placement of that scoop, just like you normally use in a mix. But I wanted the scooping to occur only if and when the solo instrument were present.

I did the math and found that by subtracting a 2nd order bandpass filtered version of the background from itself, phase cancellations were not present that I had initially feared. In fact it works very well. The difference between that subtraction and a typical EQ notching filter is that for the same depth and breadth of effect, you have to use slightly less amplitude on the subtractive component, and the Q of the BPF has to be made somewhat lower.

This Sound uses my 2nd order BPF filter. I'm not bothering to filter for the presence of the solo instrument. I just take a broadband power estimate use the RMS level plus some boost to determine how much subtraction to apply to the background sound. The !Level control is that extra boost.

So if your solo instrument is riding typically down around -18 dBFS, then by setting !Level to -18, or even lower, that causes near complete cancellation of the background sound at the center of the BPF. Use higher values (toward zero) of !Level for less severe subtraction.

In actual fact, you never get complete and total subtraction of the background, but it can approach 30 dB or more for very high Q settings. But for the most part, I want to scoop out a broad region, so that means low Q settings. For the Oboe, I chose a Q of 0.2 and a center frequency of about 750 Hz. That seems just about right.

The RMS level detector for the solo instrument is hard coded with a time constant of 20 ms. Feel free to change this Sound to suit your own needs.

-- DavidMcClain - 16 Jul 2005