Anderton compressor build report and Q

[MartyMart]

The "Flatline" from a few days ago was for someone else, so although I
liked that one a lot, I took a chance to try out Craigs compressor from
his book, "Electronics projects for musicians" ... BTW  I hope that doesn't
mean    "Electronics projects for stupid people"   !!!!       :lol:  - my girl
friend would agree with that one !
Anyway, it took a little while to figure out the "ancient" pin out on Craigs
dual op-amp, for which I used an NE5532ap and I had bought a couple
of CLM6000's from smallbear, so was good to go.
I used two stereo jacks for the bi-polar supply, which I've never done
before - if someone could tell me a way of powering this from an 18vPSU
I would appreciate it !
It worked right away and is actually quite "simple" with a low parts count.
The headroom from the +9 -9 supply sounds wonderful, I would say that
it sounds more like a gentle "studio" compressor  ie: if you push up the
compression until you can "hear it" then you have too much !!
so subtle use seems best, also at higher settings, you get "odd" sounds
as the circuit "pumps" and at one point I had a kind of "tremolo" effect
going on - quite nice if i could "harness" that setting  !!
There's HUGE volume boost available here, enough to "kick shit" out of
my amps, so thats a bonus.

To sum up, a very high quality subtle but very useable compressor, not
for everyone, as the higher settings wont get you that "finger pickin" or
"funk" sound without issues, but a great "leveling" or "limiting" FX which
will give you quite a bit of sustain and minimum noise ....

The "question" is, can I stop the "pumping" somehow ? and also re the
powering options .....

Thanks,

Marty.

[MartyMart]

"BUMP" ---- "anyone" .... ??   :cry:

Marty.

[Mark Hammer]

The EPFM compressor/limiter illustrates why some folks are so lovingly attached to optical gain control.  Unlike lower-cost OTAs and FETs, LDRs have no tipping point when it comes to signal level and distortion, because all they are are resistors.  Clean and noise free.

At the same time, all LDRs have their own intrinsic "speed" characteristics.  In the realm of audio signals, FETs and OTAs can turn on a dime, where LDRs are often mumbling "Hold on, I'm coming along.  Keep your shirt on."  They're not ALL slow as molasses/treacle, but where you never really have to wonder if the response time of a FET will be fast enough, you DO have to select the LDR/optoisolator that is suited to the speed of task.

While you can't make the change rate of an LDR any faster than it already is, you CAN slow it down.  You will note, however, that the components which might normally be present to alter the attack/decay aspects are absent on the EPFM compressor.  The gain stage for the envelope-follower/sidechain is merely that, an op-amp with gain and a 1.5k current limiting resistor going to the LED.  Absent is any cap to ground to "lowpass filter" the amplitude envelope or stretch out the transients.  Given that the CLM6000 is half LED, it accomplishes the half-wave rectification on its own.

Note that turning up the compression in this circuit involves increasing the input signal to a gain stage with a fixed gain of just over 21.  Note as well that, since there is a reciprocal effect of the LDR on the first op-amp, the *actual* gain between input jack and the LED half of the CLM6000 is more like 105 since the first stage has a gain of 5, and 5 * 21 = 105.  That is very quickly reduced as the LDR resistance drops from its maximum of several meg.

Okay, put that on the back burner for a moment.

Like any sidechain-driven device, there are two dissociable aspects of envelope responsiveness: 1) whether and how much it responds, and 2) what it does when it responds.  

Since it is hard to imagine this easily for compressors, we'll use auto-wahs as our model.  Normally, if the sensitivity is cranked up, we would expect each note to produce a filter sweep (wah).  With the sensitivity turned down, we would expect only the hardest-hit notes to produce a sweep, and most others to produce no sweep or a gurgle at best.  Of course, we could also arrange for the unit to respond to every note, but produce only a modest sweep to every note.  That illustrates the conceptual distinction between properties 1 and 2 above.

Okay, back to the compressor.  The LDR has a fixed resistance, which is in parallel with a 1meg resistor.  In combination with the 220k input resistor, those components set the gain of the input stage.  As input signal rises, the LDR value drops, and the parallel resistance of the LDR and 1meg chops the gain.  Of course, that gain is set to a max of around 5.  If the 220k input resistor were smaller (e.g., 110k as in the case of the Henry/Orman "Whisper" compressor), then the potential drop in gain is much larger (since 100k makes it around a max of 9), and the net effect of the *same* amount of resistance drop in the LDR much larger.

The role of the LDR as a parallel resistor is too often neglected.  Like the property-1-vs-2 distinction above, one can use lower values than 1meg for the feedback resistor, and/or fixed resistors in series with the LDR, to yield different sorts of gain reduction curves in response to the same input signal.  If the 1meg resistor is dropped to, say 470k or even 330k, then a greater drop in LDR resistance is needed to produce an appreciable change in their combined parallel resistance, yielding a different feel.  

Of course, that sets the initial gain of the stage much lower.  That can be offset in a few ways.  One is certainly to crank the sensitivity control up, or even increase the gain of the follower stage.  Another is to change the value of the input resistor to stage 1.  

So, with all that in mind, here is a simple suggestion to produce different feels.  Instead of a 220k input resistor, use 100k, but leave one end (the end going to the op-amp) free for the moment.  Desolder the end of the 1meg resistor that goes to the inverting input.  Now, get yourself a 250k linear or log pot, and connect the wiper to the inverting input on the first op-amp stage (where the input resistor would normally go to).  One outside lug goes to the free end of the 100k resistor, and the other goes to the free end of the 1meg resistor.

So, what do you have at this point?  You have an input stage whose gain (excluding the influence of the LDR) ranges from a minimum of just under 3 (feedback R = 1meg, input R = 250k+100k) to a maximum of 12.5 (feedback R = 1M+250k, input R = 100k).  Doesn't sound like a lot, but in combination with what the follower stage adds, trust me its a lot.

More importantly, the same amount of resistance change in the LDR has different proportional impact on the gain reduction.

I'm not promising these values as absolutely magical, but in principle this experiment should yield something interesting.  Note that the "feel" and sensitivity/compression controls will be interactive, in that changing the one will likely necessitate changing the other, since what the LDR does depends on the gain of stage 1 and the signal attenuation en route to the follower stage.

Without wishing to make this posting too long, I will add one more thing and that is treble compensation.  Typically, compressors have a way of sounding like the treble got cut.  One way to offset this is to use a tone control like the Whisper does.  Another way is to use some sort of treble compensation in the gain.  Here we take a lesson from other pedals and use two parallel paths to the input of a gain stage.  Specifically, we're going to provide two input paths to the second gain stage in the audio path.

You will note that it uses a 2.2uf cap and 8.2k resistor as the input path.  This provides a low end rolloff around 9hz, and a max full bandwidth gain of around 12.  There IS a feedback cap to trim high end but we can safely consider the high-end limit of 80khz to be full bandwidth.

If we placed a .033uf cap and a 3.3k resistor in series, and put them in parallel with the 2.2uf/8.2k combo, that would provide an "easier" path for content above around 1.5khz, resulting in greater gain at the output stage for high end, at just about all output volume settings.  If that provides too severe a boost, try something like 4.7k and either .022uf (for boost at 1.5khz) or .015uf (for boost at 2.2khz).  Since that boost doesn't particularly respect our need to avoid hiss, I will suggest upping the 20pf feedback cap to 150pf.  This will give a 10khz rolloff at max gain, which should keep things a little more noise-free.

Here is a drawing of the whole thing: http://img.photobucket.com/albums/v474/mhammer/ModdedEPFM-Comp.gif

Thanks to Stellan Lehrberg for the basic drawing that made it easy for me to cut and paste this together.  It's a small (6k) GIF file.

[MartyMart]

Mark, I can't thank you enough for that very informative AND eloquent
reply  
Not only have I been given an almost "perfect" answer to my small
"gripe" about the circuit, but have learned something here - yet again !!

Have you built the version with your mods ? and were you happier with
the response/feel of the compression ?
I'm also wondering if another LED/LDR combo, with different properties
would work "better" .... for a faster attack perhaps ?

Bless you,

Regards,

Marty.  

[Mark Hammer]

Mark, I can't thank you enough for that very informative AND eloquent reply  
Not only have I been given an almost "perfect" answer to my small
"gripe" about the circuit, but have learned something here - yet again !!

Have you built the version with your mods ? and were you happier with
the response/feel of the compression ?
I'm also wondering if another LED/LDR combo, with different properties
would work "better" .... for a faster attack perhaps ?

Well, always happy to help out the "empire". :lol:
Actually, I think I may have cannibalized my compressor for the optoisolator, or maybe the chassis, since I can't seem to find it anywhere about.  So, the answer to your questions would have to be no and no.  I have a Q&D and a couple of SSM2166 chips lying about, plus I snagged myself a nice rackmount Ashley SC-55 stereo limiter/compressor for peanuts last summer, so I'm pretty much set for compression.

Bear in mind that while the suggested mods may yield a broader range of "feels", they don't necessarily change the *rate* of resistance change; that is, how they (LDR's) travel the distance they do to do WHAT they do.  It may well be that changing the feel/response in terms of gain reduction does what you need.  On the other hand it may be that some sort of medium-sized cap to ground may be needed between the envelope gain stage and the LED to smooth things out.  You don't want to limit the current too much, so splitting the 1.5k up into something like a 1k and 470R in series, with a cap to ground from their junction, or even R/C combo to get a faster release, might be what's needed.  

Alternatively, maybe an optoisolator with a different response curve. Anderton went with the CLM6000 partly because it took damn near a half second for it to recover, making fancy rectifiers unnecessary for removing ripple.  Cost effective, yes, but not the speediest things in the world.

If you haven't taken a peek at either the What Compressor, or the terrific write-up explaining how/why the creator did what he did, you should make an effort to do so.

[MartyMart]

Thanks again, I will have a read of the "what compressor" article asap.
I'm almost a bit sorry about giving the "flatline" away , but it was an
emergency for a friend ... !! - that is a very nice circuit BTW ....

For "studio" use , I have a nice TLA stereo valve compressor and a smaller "opto" version too, I think between this - dyna - OS and a Maxon
CP101, I'm all set for "compression" now ... !!
Regards,

Marty.