Maybe some improvement for Aphex Exciter material

[John Dyson]

(Not in DSP or other sections -- this doesn't seem to fit)

 

One thing that I have always thought to have an ugly sound -- the Aphex Exciter.  I have noticed either the 'Exciter' or other kinds of specialized EQ being used in various material from at least the 1970s.  This example seems to work well for Linda Ronstadts recordings, and a modified version seems to help control the overly scrambled 'wall of sound' effect on ABBA.

It might not correct all of the 'Exciter' sound, but seems to help with the 'chalk on the chalkboard' graininess.  The script below seems to help, but an additional amount of treble boost might also be helpful -- additional treble boost doesn't seem to re-instate the 'grainy'/'makes me itch' kind of sound.  It seems to make the 's' sound move to the correct place in time.  (mitigates some of the reverse lisp on Linda's recordings.)

 

The additional treble boost -- I have tried +3 dB at 3kHz/Q=0.500, and it seems to help bring the highs back without adding the grain back into the audio.  This is just a first iteration of the anti-Exciter, and really does help get rid of the 'grain'.  How effective -- on an ad-hoc scale, maybe 2/3?  I wrote a more complicated anti-Exciter, but it was done with FIR filters, which was stupid, since the original 'Exciter' used analog style filters (IIR near-equivalents.)

 

Basically, this is an interleved set of filters at logarithmically (almost) even spaces, +3 then -3dB at the same frequency with different Q values.

 

Here is a first cut script: (I verified this exactly -- I included a version that can be tweaked in case someone wants to play with it.)

 

HAVE FUN!!!

 

Q0=0.500
G0=3.0
Q1=0.840
G1=-3.0

DECVARIABLE="treble ${G0} 1.5k ${Q0}q treble ${G1} 1.5k ${Q1}q treble ${G0} 2k ${Q0}q treble ${G1} 2k ${Q1}q treble ${G0} 3k ${Q0}q treble ${G1} 3k ${Q1}q treble ${G0} 4k ${Q0}q treble ${G1} 4k ${Q1}q treble ${G0} 5.2k ${Q0}q treble ${G1} 5.2k ${Q1}q treble ${G0} 6.84k ${Q0}q treble -3 6.84k ${Q1}q treble ${G0} 9k ${Q0}q treble ${G1} 9k ${Q1}q treble ${G0} 10.4k ${Q0}q treble ${G1} 10.4k ${Q1}q treble ${G0} 12k ${Q0}q treble ${G1} 12k ${Q1}q treble ${G0} 15k ${Q0}q treble ${G1} 15k ${Q1}q"

 

DEC="treble 3.0 1.5k 0.500q treble -3.0 1.5k 0.840q treble 3.0 2k 0.500q treble -3.0 2k 0.840q treble 3.0 3k 0.500q treble -3.0 3k 0.840q treble 3.0 4k 0.500q treble -3.0 4k 0.840q treble 3.0 5.2k 0.500q treble -3.0 5.2k 0.840q treble 3.0 6.84k 0.500q treble -3 6.84k 0.840q treble 3.0 9k 0.500q treble -3.0 9k 0.840q treble 3.0 10.4k 0.500q treble -3.0 10.4k 0.840q treble 3.0 12k 0.500q treble -3.0 12k 0.840q treble 3.0 15k 0.500q treble -3.0 15k 0.840q"

DECEQ="treble 3 3.0k 0.500q"

Use a command line this in 'bash', but something similar in cmd.com or elsewhere will also do the job:

 

> play -q infile.wav gain -3 ${DEC} ${DECEQ}

'DEC' does the 'grain' removal, and DECEQ undoes the HF cut (at least partially.)  This does work, and if there is interest, I can upload some examples.

It is NOT perfected -- but MIGHT be a good first step.

 

John