🚀 go-pugleaf

In article <f6j94q$9dh$1@panix2.panix.com>, kludge@panix.com says...
>
>
> <muzician21@yahoo.com> wrote:
>>What's the typical way to use de-essing - do you go through the file
>>and painstakingly cherry-pick the problem spots or should you be able
>>to treat the entire file at once?
>
>In the analogue world, we normally use it on the whole track.
>
>>Is a de-esser more than just an EQ tweak? I discovered that Nero 6
>>wave editor has a de-esser but it seems to take the sparkle off the
>>overall file if I just apply it globally. I assume that there are
>>better de-essers.
>
>So, don't use so much.  The de-esser is a compressor, with an equalizer
>in the sidechain.  Only a narrow range of frequencies go to the 
>compressor control, and only those frequencies cause the compressor to
>squash.  So it squashes only when that narrow band is found.
>
>If it's taking some "sparkle" off, change the tuning.  Make the filter
>Q narrower and move it up and down.  Set the compressor threshold higher.
>Listen to the music and switch the thing in and out of bypass so you
>can hear what it's doing.
>
>>Any such thing as a de-esser plugin for Cakewalk Home Studio?
>
>I don't know, but you should be able to make one with a compressor 
>and a parametric equalizer, just like we do in the analogue world.

It actually takes *two* compressors and a highly selective bandpass or 
highpass filter, which a parametric EQ may or may not have. A good de-
esser looks at the ratio between the broadband level and high frequency 
level and adjusts the ratio as necessary to prevent the high frequency 
level from sticking out from the broadband level. The problem with the 
parametric EQ/single-compressor technique is that the de-essing action is 
highly level-dependent.

The Orban 516EC and its successors use the "ratio" method, as do the dbx 
de-essers, which were introduced later. The Orban de-essers use a 
broadband level detector to modulate the threshold of a feedback 
compressor with a frequency-dependent sidechain, while the dbx de-essers 
use two decilinear RMS detectors, one of which is fed by a bandpass 
filter, to compute the ratio, which is simply a subtraction in the log 
domain.

The esses can be reduced by at least two methods. The Orban units use very 
quick wideband compression, which I find to be essentially artifact-free 
if the designer has chosen appropriate attack and release times, which 
tend to be *much* faster than commonly used for normal vocal compression. 
(The release time of the 536 is about 30 ms.) By "artifact-free," I mean 
that there is no audible gain pumping or other dynamic instability.

The dbx units allow you to use either broadband GR or lowpass filtering. 
For voice-only tracks, I strongly prefer the wideband technique, which 
does not color the esses. Lowpass filtering can be useful if one is trying 
to rescue a mix in which the vocal has not been correctly de-essed, but I 
don't know of any "set it and forget it" technique for de-essing a mix.

I use an Orban 536 de-esser in my standard vocal recording chain after all 
EQ and compression. I find that I *never* have to adjust it. There is a 
certain "natural" ratio between voiced sounds and esses that everyone 
knows because everyone has heard natural speech all of his/her life. A 
good de-esser should maintain that balance, automatically adapting to 
microphones, EQ, compression, and different singers The only exceptions I 
know of are pathological cases where the sibilance is in the form of a 
narrowband whistle instead of the much more common broadband highpass 
filtered noise. These cases should ideally be fixed in tracking, even if 
it requires a little temporary dental work, as another poster suggested.

Bob Orban