Orange Squeezer - freq dep comp mod?

[lion]

I've finally got my OS working - and I like it very much - very soft an musical.
I build it a year or more ago - but couldn't get it to function properly. The sounds was OK - but the compression was very very subtle, if any at all. (Couldn't believe that this was what a lot of people was raving about). Tried a lot of transistors and different diodes - and followed the ongoing talk and advise  on this board - with not much luck and change in function. Last week I ordered some new 2N5457's and a 1N5818 schottky to give at a last try . Fitting the new components, I discovered a wiring error. I had the feedback circuit (1.5K > diode > 4.7mF >.....) conneted at the wrongs side of the output cap. After correcting the error and changing the transistors everything was OK - and very little noise also. Don't know whether is was the new transistors or the corrected error that made the difference ().
Never mind - I'm very impressed with the circuit now. Found the adjustable attack and recovery - and MH's brightness control - essential for perfect adjustment.

Having used the OS for at couple of days now, I've found that in some instances, I would like to have a little less compression on the lower notes (say the wound strings) in relation to higher notes - a sort of (simpel) frequency dependent compression.
In theory it should be possible to stick some EQ/filter in the feedback loop to the JFETS. I imagine some kind of highpassfilter with a soft lowfreq rollof would do the trick. Or maybe another kind of adjustable 'tonecontrol' - balancing the amount of bass/treble to the JFETS.

Maybe I'm totally wrong - so before giving it a try, I would like some advise. Anyone?

Erik

[Mark Hammer]

Not an inappropriate question and in fact part of what separates high end compressors from stompboxes.

Ignoring the output level pot, the signal essentially passes through the 4u7 cap and 1k5 resistor before hitting the diode.  The cap/resistor combo result in a highpass filtering action that rolls off the low end around 22hz.  That provides full bandwidth compression, or rather, compression that is influenced by the full frequency range.

Normally, it would just be a mater of dropping the value of that cap to raise the lowend rolloff frequency.  For instance, a 220n cap would rolloff around 480hz, and a 470n would roll off around 220hz.  Because the output signal is tapped from the junction of the 4u7 cap and 1k5 resistor, though, the 4u7 cap serves double duty by permitting the full frequency range to not only reach the audio output but the envelope detector as well.

Fortunately, though, it should simply be a matter of taking two taps from the output of that op-amp.  One goes to the standard 4u7 cap then on to the output pot.  The other goes to a 220nf cap (or whatever value suits your needs; smaller rolls off higher up) and then to the 1k5 resistor.  That will provide a different passband for the envelope detector than for the audio output and less influence of bass notes over the compression.

Just bear in mind that bass notes have greater amplitude, so roling them off in the envelope detector means there will be less overall envelope signal, hence less compression.  You can offset this by adjusting op-amp gain and judicious use of the trimpot, but just be ready for it.

This gets me thinking.  Isn't it about time we put our heads together and produced a simple multiband compressor using stock circuit fragments?  

Imagine the following....

Common input buffer goes to what are essentially two separate parallel OS units, each tuned for either highs/mids or lows, both in terms of pre-compression drive (which sets envelope amplitude).  The outputs of those then go to a common mixer which has adjustable high/low level-balance, a bit of gain (naturally), and a common mono output level control.

One *could* use complicated filters, but it doesn't have to be all that complicated.  A simple large value cap in the feedback path of the low-end half would restrict bandwidth.  For the high/mid section, you culd use a smaller cap to ground from the inverting leg of the op-amp and, as described above, a smaller value cap going to the envelope follower.  That would likely provide enough rolloff to differentiate what each half deals with.

Once you have two separate sections, you can then play with the attack and decay constants optimized for each range.  By having a high/low balcne control at the output mixer, you could also adjust the tone far more than is often the case.

So, to recap, one dual op-amp plus 4 FETs for the compressor sections, one dual op-amp for the input and output sections.  One pot for low-end drive, another for high-end drive, a high/low balance pot, and an output level pot.  Since the time constants are optimized for each range, there is no need to futz with a variable attack control for when you use different types of input signals.

Should sound even MORE transparent, and tonally flexible, without that muchmore cost, complexity, or size.

You see, this is why its good to ask dreamers a question on Monday morning.

[petemoore]

I've built three of them.
 They are tricky in that the effect is subtle, and there are [it seems] numerous ways to miswire, and have signal pass, and gain be 'added', but still not work right.
 A dual OS, as described by Mark Hammer [one for mid/highs, one for lows] sounds like a very peachy compression build technique.
 I've found that adjusting the pickup height can become more important
[or more blatantly obvious that imbalance exists] to balance highs/lows content when using booster and especially compression effects.
 It seems to change somewhat depending on what content of the guitars output frequencies are accentuated by X circuit [whether highs or lows seem louder after being gained up], but I should have thumbscrews on the pickup height adjusters to keep them 'right' under varying conditions of the signals amplification [effect gain/attenuation].
 Most pickups seem to have more output under the thinner strings [compensating for the smaller signal/smaller string], there must be some compensation in most pickups, otherwise the bass would totally dominate with a 'flat' pickup positioning.
 Better to 'set and forget' the pickup height levelling [between the bass side and the treble side of the six strings], and doctor the circuits to provide fairly even response across the Freq range...[or not].
 I hear very uneven response on records at times, Page being the standout in this area, Harrison too...certain famous recordings sound like one strings pickup pole is wayy high [and as a result louder] compared to the rest of them [high E string being a good candidate], so that when THAT string is used, a very different tone is produced. Cool, EZ trick. This combined with a slight string/fret buzz can really make that string standout as very different sounding. This effect IS there [I swear], in cameo appearances, on Zep, Beatle, and other recordings. One strings sound somewhat blatantly stands out from the others....proving Imbalance too can be very cool.

[lion]

Mark - thanks for the advise. I'll give it a try - splitting the output, maybe with a couple or more switchable caps in the detector circuit. Having upped the 1.5k resistor a little for a slower attack - throws your math a little off - but I'll let my ears be the judge anyway. How step is the roll off - I guees this must be a first order filter, so it 6dB/oct - IIRC?

A multiband OS sounds interesting - wouldn't there be a potential noise issue, mixing several parallel OS circuits.

Erik

[Mark Hammer]

Erik,

You are correct on several counts.  One, the math DOES change, but is nothing that the old standby F=1/(2pi*RC) can't cure.  Second, the filtering is single pole, so 6db/oct.  And third, more stages generally does equal a tad more noise.  On the other hand, the OS also currently lacks a decent input stage so maybe inclusion of one would offset any acquired risk for noise.  Besides, one can always stcik a cap in the feedback loop of the high/mid op-amp and another in the feedback loop of the output mixer and roll off objectionable highs that way.

So, what do we call such a 2-headed beast?  The Lemon-Lime?

[lion]

Mark - Lemon-Lime (LOL), suits me fine!

Still sticking with my single OS for a moment. If I was to do a mod for less compression of the higher freqs - I could simply add a small value (100-500pF) after the 1.5 k right before the diode - for some HF rollog, right?

Not that I'm not interested in the Lemon-Lime, just want to explore the possibillities with the circuit I have first - gives you some time to think, and design the LL ;-)

Erik

[Lonestarjohnny]

Mark, You gonna Draw somthin up and post it ?  sounds like a great Idea and for sure I would love to do a build on it.
JD

[Mark Hammer]

JD,

Drawing is unlikely this week.  Though I have some airport time ahead of me this week that might be put to productive use.  My guess is someone will likely beat me to it, anyways.  They usually do, thank goodness.

Lion,

Yes, I suppose some simple treble rolloff at the junction of the 1k5 and diode would work, as would rolling off top end in the op-amp itself.  Of course, trimming high end from the op-amp itself  changes not only the frequency sensitivity of the follower, but the output tone as well, so best to avoid that.

[lion]

I've tried the low freq roll off in the feedback circuit mod.

As Mark suggested I split the output from the opamp in two:
- output to vol. control with a standard 4.7 mF cap
- a seperat line with switchable caps 0.47 - 1.0 - 3.3 mf to the 1.5k/diode/..

But - using a freq generator and my old scope I couldn't detect any LF roll off in the signal to the FETs at all - regardless of the cap value. Splitting the output/feedbackloop seemed to raise the outputlevel by a fair amount.
Wonder if I'm missing something here - Mark?

In order to create a highpass filter shouldn't there be a resistor to ground after the cap?

Erik