dbx 902 de-esser

[dmp]

I have a 902 de-esser that has distortion when turned up for gain reduction.

I'm trying to understand the circuit a little bit more. A few schematics attached.
My version is the most common that matches these schematics as far as I can tell, with RMS daughter boards and a dbx202x with 8 parallel THATs.

As I understand it so far, the Audio is split into a low pass (LP) and high pass (HP) components (the HP is created by subtracting the LP from the full bandwidth). The user set frequency control adjusts where the HPF is set. The HP and the full bandwidth (FBW) are sent to separate RMS detectors that each produce an average voltage corresponding to the signal levels, which go to a comparator. The difference of the HP to the FBW with the threshold (set by the range control) becomes the GR control voltage (CV), which goes to the VCA. The more high frequency energy relative to the FBW, the more compression.

There have been some discussions on here years back on operation of the 902, but all the images, etc... have been lost.
I'd be interested to understand what the trimmers are doing before I check the calibration.

 

[JohnRoberts]

Back in the day DBX used to issue AES preprints, effectively white papers describing how their new dynamics processors worked. Maybe search for deesser AES preprint, or AES paper.

JR

 

[dmp]

I'll look for that, but don't see it online after a quick search.

The frequency filter is a 12dB/oct Sallen-Key Butterworth filter adjustable from 818 Hz to 11441 Hz

 

[dmp]

I replaced the two 3.3uF tantalum caps at (under) the RMS filters and it fixed the distortion I was hearing. I had re-capped this years ago but missed those.
I think these caps set the time constant for averaging.

 

[JohnRoberts]

I replaced the two 3.3uF tantalum caps at (under) the RMS filters and it fixed the distortion I was hearing. I had re-capped this years ago but missed those.
I think these caps set the time constant for averaging.

This is a little esoteric but Tantalum dielectric generally exhibits higher DA which can make a subtle difference in time constant side chain applications. For a single ended dynamic processor there is no encode/decode tracking to worry about.

JR