Stereo Tool

I'm releasing a preset I'm calling "DewBX9"; geared toward subtle, but effective, enhancement of music.

This is actually an offshoot of a previous project I was working called; called ToxicMagic. ToxicMagic was designed to be an enhancement done in a DAW to provide some dynamic expansion in combination with StereoTool's declipper as well as additional EQ effects. The original system was modeled after a dbx 1BX; using RMS-triggered expansions with a 10ms attack and a 250 ms release; which was as close as I could get to the dbx release since it was velocity based rather than time based. ToxicMagic later became "BX4"; applying some additional teaks to the expansion module, and moving to a multitrack environment. A stereo track was copied to 4 stereo tracks; each track was band-filtered and run through the slightly modified ToxicMagic process. It turned out to work well; as this screenshot of Californication shows. (Top is original; middle is declipper only; bottom is the full BX4 treatment.)

In fact, I was so focused on playing with Declipper and working on ToxicMagic/4BX I completely missed a bunch of upgrades that added NaturalDynamics; which rendered my BX4 work obsolete. So, I decided to focus on tweaking it instead.

DewBX9 aims to be a subtle yet effective enhancement. The goal was not to run the effects to high that you notice them; but just enough that you notice not having them. DeClipper was tweaked to work on a large majority of the clipping you stumble upon in most music; forcing restored peaks to be only 103% over original. NaturalDynamics was programmed to use RMS detection on the lower and upper frequencies; with Peak elsewhere. Azimuth correction was also enabled; but set to a slow correction speed as to prevent over-correction and degradation. Finally, StereoBoost is used in conjunction with the phase amplifier to provide a subtle boost in the stereo image. The -6db post-amplifier is to protect the output from clipping due to declipper and/or NaturalDynamics. I have opted to go this route rather than use a limiter; which will destroy much of the enhancement in the dynamic range.

I am continually throwing tracks through this from a large archive of over 20k tracks. Revisions will be continually updated; allowing you to track the progress of development.

Comments, suggestions, and other feedback are greatly welcome and appreciated.

All updates will be edited in to top-post; replies will indicate when update is made.


*****Revision History******

Revision 5 - Final Candidate 1 12-Oct-2013

Added funky multiband configuration to fight essing once and for all.
Additional tweaks all around; specific revision notes lost during dev.

This is a candidate for a final release. I feel I'm pushing the audio far enough while still staying "subtle" Please note: If you try to play with the tresholds; and possibly other settings in Multiband; it seems to "break" it. You'll notice this if the band VU's seem to max out. You have to reload the preset to restore it; sliding back to original settings has no effect. It's probably due to the bizarre way I designed the multiband (still figuring out; I'm sure beer was involved when I made this one).

Revision 4 - "Because bleeding ears are no good" 3-Oct-2013 Widened ND band #8; narrowed #9
Reduced follow speeds of bands #8, #9
Lowered "punch" settings for band 8
More adjustment for stereo image


Revision 3 - "The voices; too shrill!" 1-Oct-2013 Tightened bands in ND
Increased look-aheads of bands 8 and 9
Decreased follow speed of bands 1 and 2
Slight tweaks to individual band strengths
Changed all ND bands to "Peak" mode in response to new settings.
Adjusted Stereo Image and Phase Amplifier

Notes: Some tracks are mixed in a way that Stereo Image and Phase Amplifier have a negative effect; it may be necessary to turn them off in this case.

Case/recommended tracks: Donna Lewis - I Love You Always Forever; Dido - Stoned; Radiohead - Karma Police


Revision 2 - "Its roots are in dbx!" 30-Sept-2013
*Temporarily Removed due to message limit*

First public release
changed mid-frequency bands to RMS detection mode
activated stereo boost and phase imager
deactivated equalizer
backed off azimuth speed

Revision 1 - "All my work for nothing." 06-Sept-2013

No public release - overwritten by V2
Initial core: defaults
Declipper changed to ToxicMagic settings; dropped force peak to 103%
added equalizer
added aggressive azimuth correction

*****Reference System for DewBX9 (as of V3)*****


Winamp using ASIO output with SSRC resampling
M-Audio Transit USB DAC - locked to 96khz
Yaqin MS-12B Tube amplifier using 5963 "computer-tubes" for preamplification stage
Sansui SE-8 Graphic EQ
Denon PMA-770 Integrated amplifier (also use Motorola HS-619 monoblock 6V6 PP amps)
Bose Triport "Over The Ear" Headphones. Original version - Model TP-1A
DCM TimeFrame TF500
*The dbx 1BX was for testing and not used for Rev. 3


****Reference System for development of V2****
(This setup died 2 weeks before public release of V2)


Winamp using ASIO output with SSRC resampling
M-Audio Transit USB DAC - locked to 96khz
Onkyo TX-SR606 Receiver with optical S/PDIF to Transit
*The dbx 1BX was used during initial design of ToxicMagic; not used for Rev. 2

Revision 3 of DewBX9 has been released; you can find it in the top post of the thread.

Work on Revision 3 quickly began after the public release of Revision 2. The goal was to fix a few problems that did not creep up during the 2 weeks of listening to Revsions 1 and 2. Female voices; as well as most S sounds; had a habit of sounding extremely shrill and piercing, which didn't creep up while I was listening to Beatles and other classic rock. More modern pop tracks were sampled and revision 2 quickly morphed in to revision 3. Also of note; Revisions 1 and 2 were generated using a different audio setup than 3; which the setup for 3 being my "ol' standard". Originally; the 7khz band was switched to "Traditional Peak" to compensate for the shrillness; however after adjusting the frequency band and slope; this was unnecessary and switched back to "Peak". Also of note is this revision completely abandons all methods based off dbx technology; switching the newly adjusted lower and upper frequency bands giving better performance with Peak mode. Various adjustments were made to the look ahead and rise speeds to help prevent essing.

Revision 4 of DewBX9 is now available.

This revision continues to tackle and fix essing problems.

Revision 5 of DewBX9 is ready for release.

Among the major updates is the inclusion of a funky multi-band configuration to force-fix essing without, hopefully, completely destroying the benefits of ND in the upper bands.

This version has undergone extensive testing over the last few days; and it may be the final revision unless I can figure out a way to squeeze any more enhancement out. However, I feel I'm at the extreme end of what I feel is "safe".

See, I kept working on it even though no one said anything positive or negative about it.

You have wrong settings in De-clipper.. making it to actually do almost nothing.
Anyway.. here is "completed" settings with fine AGC leveling (12dB max), plus compressor as protection!
Final Output from MB almost never hit clipper so all peeks are there, it's also as a protection. (uses latest beta!)

btw i see nice amplifier.. but what are speakers?
since you killing ~8kHz too much, in my oppinion, are your speakers "flat" ?

The De-Clipper was done that way on purpose. I didn't want it to go attacking *everything*; because quite honestly most of the clipped audio in higher frequencies doesn't repair that well. This one was more focused on fixing the massive distortion they seem to mix in to the bass. I was trying to completely avoid AGC; as that somewhat negates the idea of doing dynamic range enhancement only to just squash it later. However I've not really played much with the AGC in stereo tool; so it may work similar to the auto-gain from Goldwave which is able to maintain dynamic range...but even then; some pieces have parts that are supposed to be very quiet where AGC would negate that; and I'm mostly thinking of classical music where AGC causes a problem.

As far as 8khz; it's a slightly narrow band that was done solely to prevent essing I was getting from enhancement. Mostly on vocals; Donna Lewis almost made my ears bleed. The *should* be RMS triggered; so they wouldn't attack peaks; although I *may* have accidently left it in Peak.

The speakers aren't pictured because I had the stack outside to take photos in better light. My main pair are DCM Timeframes TF500s. They have a decent response and Vifa tweeters; so if anything they tend to be a little piercing, at least to me. I also have a pair of TimeFrame TF400s and a really old pair of JBL's with the D130 15" driver and 075 bullet tweeter. I did notice when I demo'd this through cans the other day the upper range seemed to have a tad bit of attenuation.

I appreciate your feedback; and I'll take a look and listen to your modifications in a little bit; see if I can't continue to make some tweaks. Like I said, it sounded perfect to me; but my ears are entirely different than most people. Of the few CRT's I encounter; I can still hear the 15khz HV whine.

Point is that your de-clipper settings do 'almost' nothing. So 2 ways to sound better is to set all settings at default and rise "step size reduction" to maximum as you can. That will make de-clipping high-end better. Or to just turn it off to save cpu.

AGC - does not have nothing with dynamic range. Filters after AGC, like compressors and limiters makes that. So proper adjusting AGC's output makes how much signal will go into compressor or limiter or clipper. I left enough room so peeks barely reach clipper. i can agree that AGC in my settings might be fast for your goal (release!) - you can always make that slower and also make less range in AGC by lowering pre-amplifier (depend on your input).

Declipping on modern music is real subtle anyway. It's worked fantastic for analog clipped audio pulled off tape; but mine was done using some *very* bass-heavy/bass-clipped stuff; that seems to be more audible than one or two sample hitting clip. Any decent DAC is going to anti-alias that to a degree.

I still don't buy the AGC having nothing to do with dynamic range; it may not affect it in the same way as a limiter/compressor; but it *still* modifies the volume by boosting it in ways I didn't want for this. Relative dynamics; but there are some portions of a track that are *supposed* to be down that low; and AGC tries to boost everything to have an average RMS.

On my end; your modifications sounded louder than I was going for. I know louder sometimes equates to enhancement; but I'm really trying to get away from the loudness war sound; which is what your preset had; at least on my equipment. Any average person is going to call it great; but I'm not the average person.

I also don't understand the +10db preamp. That seems to go against everything I'm doing.

The results sounded louder and slightly more limited than Rev 5; that's not what I was going for.