Envelope Detector Input Filtering (Orange Squeezer Content)

[aziltz]

Has anyone considered Variable Hi-Pass at the input to the envelope detector of a simple compressor like the Orange Squeezer.  I think this would tailor the "Feel" quite a bit.  You could have higher compression settings but avoid the clamping that occurs on the bass notes (if you wanted).  I've built a version of the OS that has the detector out of the signal path, so swapping the input cap to the Env Detector affects how much the compressor reacts to the bass notes.  I was looking for a way to get high compression on the treble strings, but avoid over squashing full chords.  I'm thinking about making this an internal trimpot.  Any thoughts?

I think I remember reading that the Dyna Comp does something similar?  It filters, compresses, and then un-filters?

[Mark Hammer]

In fact, much of the ripple that is so problematic in compressors set for slow recovery is actually content in the sub-fundamental (but still audio) range, like beats between notes and such.  So while use of LDRs and full-wave rectifiers and lowpass filtering can be useful for bringing the changes produced by the sidechain closer to what would be the case if the envelope were absolutely pure averaged DC, I would imagine that some highpass filtering of the signal fed to the rectifier (and this is different than filtering after the rectifier), say starting at around 80hz or so, might help to eliminate audible ripple.

[aziltz]

In fact, much of the ripple that is so problematic in compressors set for slow recovery is actually content in the sub-fundamental (but still audio) range, like beats between notes and such.  So while use of LDRs and full-wave rectifiers and lowpass filtering can be useful for bringing the changes produced by the sidechain closer to what would be the case if the envelope were absolutely pure averaged DC, I would imagine that some highpass filtering of the signal fed to the rectifier (and this is different than filtering after the rectifier), say starting at around 80hz or so, might help to eliminate audible ripple.

Interesting.


btw Mark, the compression control you implemented in the Tangerine Peeler works great!

[Rob Strand]

Any thoughts?

I've got more thoughts on this than time.

I don't think pre-filtering the envelope detector will achieve what you want.  You have to think about what is being compressed and what causes the compression separately.

It works on say De-ess'ing because you want the compressor to kick in on the S's.  The S's might seem loud and annoying but the signal energy is small.  Boosting the highs makes the compression kick in earlier at high frequencies.   When it does compress it squashes the S's but there no a lot of other signal at the same time. So the S's are dealt with in isolation with minimal side effects.    In normal speech even with the boost there not enough high frequency energy to trip the compression

The real problem with guitar is you *want* to compress the lows because most of the signal energy is there.  However the compressor acts across the whole band width so when it compresses the low the highs get compressed and it sounds dull.   The key thing here is you *want* the lows to trigger the compression the problem is there's both low frequencies and high frequencies happening at the same time.  If you filter the side envelope detector then to a large degree you are just  raising the compression threshold.

The post tone control tries to boost the highs back but it has the side effect of boosting the highs when the compression isn't acting so you have to found a middle ground.

A better solution is to feed the dry signal through.  This way the highs can be squashed away so much.   Some other options are leak some of the highs only through.  Or split the lows and highs like a multiband but only compress the lows.

For a multi-band compressor when the lows kick in the compression the highs are unaffected, they can be controlled through a separate path.

Another option which might fit in easily into the orange squeezer is the idea of dynamic filtering.    When the compressor cuts in you want the lows to be attenuated but you want the highs to be attenuated less to prevent dulling.  When the compressor is not compressing you want a relatively flat response (unlike the post tone boost method).  What I suggest is replacing the 82k with say a 39k in series with a parallel combination of 39k and 5.6n.   The idea is when the compression cuts in the JFET shunts the signal but the lows will get attenuated more than the highs because of the variable filtering.  You can play with values but try to keep the sum of the resistors around 82k.

[aziltz]

That's some good stuff there.  I really appreciate the input.


Mixing in the clean does have its advantages, and I really enjoy that when I use my BYOC 5 Knob, as an always on thickener.  But for the current project I'm working on (the Orange Squeezer Derivative) I had some design constraints I want to work in.  Mainly, I wanted to restrict the design to 1 dual-opamp and only add FET buffers where needed, staying as close to the OS-Audio Path as possible.

I'm not really going for a "transparent" compressor here.  I noticed at high compression settings, there was this great snap to the high notes.  Not quiet the Country Click (that's there at max sustain), but it was really great for clean chords stabs, perhaps in some funkier rhythms, or maybe Maroon 5 for an example. HOWEVER, when I hit a full chord at that setting, I got too much compression, simply because of the energy in the bass notes.  It felt, like you described, that the highs were also "disappearing".  To keep things more or less "Orange Squeezer", I wanted a passive solution to keep the compressed highs, and lessen the clamping on the lows.


So I filtered the bass out of the envelope detector.  Simple, and it gave me the effect I was looking for.  I still have the same compressed highs, but they don't disappear when bass notes are played.  The compression sounds more "even" to my ears, as in, the highs and the lows are compressed by more or less the same amount, and the lows don't have too much more weight.    I still may need to get more highs at the output, it seems a tad dark overall.

The allure of the OS is that its so simple.

You're last idea sounds the most interesting to me though.  Boosting some of the highs going into the attenuation.  I will give that a shot.  I can always use a brighter signal!

[liquids]

I'm a little late to the party and commenting on earlier posts...

But from my experiences, filtering a low frequency roll off for any 'envelope detector' circuitry is a good thing in fair dose.   I'm thinking mostly in terms of envelope filters, but this may well apply to the OS too...

That being said, note that a lot of wisely built envelope detectors are fed by a buffered signal.  Oftentimes though, a theoretically small value cap is fed from said buffer into envelope circuitry.  This only work in theory since the impedance of the signal is often that even extremely tiny caps seem to pass all signal, unless combined with a resistor to ground to set a low frequency roll off point such as here: http://www.muzique.com/schem/filter.htm

If you choose your values careful and control this resistor in combination with a chose cap value (within reason), and maybe even get creative with some simple low pass filtering (series resistor with cap to ground) too, if you so desire, you can really tweak and manually control the frequency range of signal the envelope detector is responding to for compression or envelope filtering, and also help smooth ripple indirectly in the process....I've not tried it in depth so it may be overkill, but it works in my head, and it may be quite useful...  

[aziltz]

thanks for the input liquids.  I was tinkering around a bit more tonight.  I need to come up with a nice way to keep the compressed dynamics of the highs, but boost them back a little when the sustain is turned up.

[aziltz]

here's something I tried that I liked.  I added some feedback, from the output back to pre-attenuator, this brightened the signal a bit.  I also swapped the op-amp out for a 2262.

Question, is the TLC2262 a "slew rate limited" op-amp or a "rail-to-rail" op-amp or both?  I'm starting to get those two things confused.

[Mark Hammer]

Rob and I might well be on the same page.  I think the missing ingredient is what we are respectively thinking about when we say "filter out low end".  There's low and there's low.  My position was not so much to eliminate the content which forms part of the desired guitar signal (the part to be compressed).  Rather, it was to completely eliminate audible ripple.  And for ripple to BE audible AS ripple it actually needs to be low enough in the spectrum that it is well below the lowest fundamental on a guitar.  Certainly you can have ripple at 1khz, but you wouldn't really hear it as a trill.  Moreover, ripple at that frequency would be unlikely past the initial attack portion of the note (for most notes), and the sort of ripple we're collectively trying to eliminate is that heard during the decay phase when slow recovery times are used.  IOf there was ripple around 1khz, more than likely, you'd hear it simply as a kind of harmonic component that imparted a sort of tonal flavour.  So, a 2-pole 80hz highpass on the input to the rectifier would eliminate most of the energy coming from beats, but have no impact on the extent to which one could elicit squish by picking notes on the low E.