Compressor Thumping like crazy/explanation of ORANGE SQUEEZER?

[loss1234]

I tried building THIS circuit:


http://thesquarewaveparade.com/images/amber-press.gif

which is a more control laden version of the Orange Squeezer (has thresh, ratio, and attack/decay)

IT THUMPS like crazy at a lot of the usable settings. (btw i put a preamp of 5x gain before it when using with GTR>but it also works with line level audio.

BUT since I do not quite understand how the gain reduction circuit works, I am a bit lost as to how to troubleshoot it.

I know the 2 FET'S are responsible somehow. do they shunt the output to ground when triggered by the DIODE? (rectifier?)

and how important is the OFFSET?

I tried finding an in depth article on the orange squeezer but had no luck.


any help or links highly appreciated.


thanks.

[Mark Hammer]

There are a LOT of OS-related threads here.  Look through a handful and you'll probably find some ideas (plus the origins of many of the added features of the schematic you linked to  ).

Certainly one of the constants underlying the misbehaviour of anythng with a sidechain is the amount of control element change being asked for in a given amount of time.  In this case the control element is the JFET labelled Q2.

Presumably to offset the impact that inclusion of a 10k attack-time pot has (it will significantly reduce sensitivity), the designer has upped the gain of the opamp appreciably, from 23x in the original design to a maximum of 213x.  A little heavy-handed, if you ask me, but as I say, intended to counterbalance another design change.

If the gain is up high, and the attack time set for fast, then the JFET is being asked to go through tremendous resistance change in a brief period of time.  You will undoubtedly perceive this as a thump.

The original has an 82k fixed resistor used in conjunction with Q2 to form a virtual volume pot, where one leg is always 82k and the other (provided by the JFET and 10k trimpot) varies.  The amount of attenuation provided is a function of the ratio of that fixed resistor to the JFET resistance.  Increasing the value of the series resistor - in this case R4 plus whatever VR8 is - will make the JFET have a more pronounced attenuating effect.  So, if VR8 is up above 70k or so, VR10 is set at less than 40k or so, and VR3 is less than a couple of kilohms, you can most certainly expect to hear a thump as the unit behaves more like a gate than a compressor.

Personally, I think the pot values shown in the schematic are rather excessive, however they CAN be set modestly and still get decent performance out of the unit.

[loss1234]

I hate to be a pain but after a half hour of searching through threads (many of which are build related questions or which compressor is best queries) I am a bit lost as to WHAT to search for.

Anyone have any ideas of what to search for on this forum?

"compressor mods" didnt turn up too much.



basically I am interested in understanding more about how to design and tweak compressors.

compressor build theory.

I want to find some threads and or articles that take a schematic and go through it part by part explaining how it works.

The tools I really need are those which will help me analyze the circuits I see.

for example,


So far my understanding of compression is this:

you derive a control voltage from a Level Detector (which if i understand correctly is just a FWR

you shape this CV (using attack/decay networks and transfer blocks.) (which gives you attack and decay and your knee

you then control the AMOUNT of this CV into the VCA for RATIO?

And somewhere in there, i cant remember, you get threshold.

For a VCA you can use an FET, A vactrol, a tube, an OTA, or a fancy chip like an SSM2164 or a NE571,etc

Also...you have to determine if you are going to do average, or RMS.


i know how to build a FWR. I know how to make attack and decay circuits.
I know how to make a vca.

BUT there seem to be a lot of fine points to it. Like how to control threshold, prevent thumping, how to rolloff frequencies, and how to make the VCA
go to unity gain when there is no CV. and where to get germanium diodes


so...any articles or threads appreciated.

thanks guys

[Mark Hammer]

Just look for Orange Squeezer or OS+, OS++ threads.

[loss1234]

thanks

[gsmith850]

If you are still struggling with this, I can help you and take you through the concepts of this circuit.
There are a number of questionable issue with this modification of the classic OS circuit and I'm not surprised its thumping !
The first thing you need to do is get the schematic for the original OS and set all the pots and associated resistors (very important) back to the values of the original, especially the input chain R4,VR8 and R5/R6, labelled "Ratio", which it isn't !!!
The OS is basically an Auto Gain Control (AGC) device and as such attempts to hold the output level constant, so the compression "Ratio" is always "infinity" ?!? Isn't it ?
Just to get you started, Q2 is the control element and its resistance is controlled by driving the gate to turn it ON and act as a level pot in conjunction with R4+VR8. Remember also that R5 in parallel with R6 are also in parallel with the FETs resistance and limit the effect (likely cause of this cct thumping).
To control an n-channel FET, the gate needs to be negative of the source pin. In the OS, Dan did this by lifting the source slightly positive of ground (by Q1, R11, VR5 and C1).  Q1 and R11 are a constant current generator driving into VR5, which converts it to a voltage.  You set the "turn off" point of the FET by VR5 (1 - 1.5V usually).  C1 effectively "decouples" (connects) the ac signal directly to ground, so we have a dc voltage, but the signal thinks its at ground.  Also, shouldn't C1 be 10uF not 10nF ?
C11, R12 and R13 help "linearise" the FETs response & reduce distortion (a standard cct and values).
This is the main operation of the circuit, the rest is just a simple (one diode) half-wave rectifier, into C7 (the control voltage). The attack ad decay resistors are obvious.  As the output goes up, the control voltage goes up, turning Q2 ON, which reduces the input to the Opamp.
The "sensitivity" will be controlled by the gain of the Opamp, using VR10. Dans original was set about x22.  So this pot should really be called............ ? Answers on a postcard to anybody but me....

Cheers GS

ps, there are better circuits than this. It has a low input impedance (can be corrected by a high impedance FET buffer at the input (yahoo! my treble has "magically" returned)) and lack of a FWR causes "breathing" effects (cure by using other half of Opamp as an inverter and adding a second diode for full-wave). Was this what you wanted ?

[loss1234]

thanks

i love explanations!!

in the meantime, I built myself a nice TRIDENT fet compressor on the breadboard.

it blows the amber press out of the water!

and i have also been trying my hand at desining some simple compressors

i had thought it odd that there was no FWR myself

[Mark Hammer]

Actually, the OS is not, strictly speaking, an AGC circuit.  The amount of gain applied to the signal is held constant at 23x, and what is varied is the level of the signal fed to the gain circuit.  So, more precisely, it is an Automatic LEVEL Control circuit....as is the Amber Press.

You won't see a FWR in the schematic because it doesn't really need one.  For a long time I was of the opinion that full-wave rectification was implicitly better than half-wave in just about any application, because it minimized the amount and audibility of envelope ripple.  In this application, however, that turns out to be of little concern.  Why?  Well, because the period during which the envelope voltage is "doing something" in the OS/AP circuit is quite brief.  Were the envelope applied to something like a filter, where one expected and wanted the sweep to extend over a period of, say, a couple of seconds, the envelope signal would have entered that phase of its lifespan where it gets pretty choppy.  Part of that choppiness is because the diode itself sets a floor on what amplitude the signal must be to pass on to the control element.  Even if the diode is a germanium, with a forward voltage of 200mv, there is going to come a point during the decay phase where the signal dipsy-doodles above and below 200mv, simply because of what strings naturally do.  There are other factors too, such as what finger vibrato does to level, and so on.

When the interval that the signal peak is "held" by the averaging/storage capacitor (C7 in the Amber Press schematic) is brief, the signal never really reaches that point in its lifespan where those fluctuations will occur.  I've found on the autowahs I've made for myself or modded that shortening the decay time makes a lot of the concern about envelope ripple from a half-wave circuit all but disappear.  Of course, extend that decay/release time a bit, and you start to hear the difference a FWR makes.

Part of the "transparency" of the basic OS circuit is that release time is VERY quick.  The unit really behaves more like a peak limiter than a compressor, by virtue of the fact that it quickly attenuates then "unattenuates", without any of the breathing and pumping that normally occurs with longer release/recovery/decay times.

In the universe of compressors, LDR-based units have a lot going for them noise-wise and distortion-wise.  Unfortunately, a great many LDRs have a more sluggish response than FETs and OTAs, such that release time is slower.  Envelope ripple is not necessarily of more concern in those instances, because the sluggish response also means it does not react to ripple the way that a FET might.  And that is why you might see a FWR circuit for a FET-based or OTA-based compressor, but not necessarily for an LDR-based one.

[gsmith850]

Ok, gain control, level control, it still functions the same way and although it doesn't adjust the "Opamp's" gain (like say a DOD LDR design), the gain of the circuit as a whole, input to output is still being manipulated to achieve a balance of constant output level, i.e., a small input is boosted up and a large input reduced (or at least not boosted as much ?) by means of the FETs conduction. The mechanism is just different.
"Actually", I don't care.
Loss1234, I'm glad you got sorted out, though I don't know what a TRIDENT is, but it has to be better than the Frankenstein design you were working with. Good luck with your own designs. Try a few, FET (n and p channel), LDR and VCAs. FET p-channel is good, as you don't need to raise the source pin above ground, just drive the gate voltage up to turn the FET OFF and towards the source to turn it ON. Much simpler. Just make sure your rectifier cct works the right way round !
One tip if you do get around to trying an LDR/LED combo is to never actually allow the cell to fully turn OFF, by trickling a small amount of current through the LED. The LDR has an extremely high resistance in the dark and people try to limit this by putting a resistor (say 1M - 3M3) in parallel with it. This is Ok, but as Mr H quite rightly pointed out LDR's are sluggish and it takes the device a while to get back down into the control region once the leading note hits the cct. Far better IMHO to tweak the LED current to get the cell to a Meg or so instead, ready to go and near to the region you want to control it in. It seems to respond a lot faster (DOD cct builders take note) and removes some of the initial thump as the cct over-corrects on the leading edge. No doubt someone will disagree with me....... guess I must of just dreamed all those ccts working correctly.

[Mark Hammer]

Sorry if my reply came off a little stiff - trying to stuff in a speedy reply between work things.

The distinction between adjustment of gain and adjustment of level is a moot one in some ways, but a pertinent one in others, depending on context.  You are correct in asserting that they CAN sound exactly the same.  If the gain is fixed, then whatever noise sources feed the gain stage/s remain in effect all the time.  If the gain is varied, then the noise sources are most applicable when the gain is restored.  Now, of course, if the gain stage being varied has a bunch of other crap that comes after it (e.g., EQ, exciter stages, etc), then clearly the variable gain stage's contribution to overall noise is of minimal concern. 

In the case of the OS, though, we have one op-amp that remains at a fixed gain.  What gets fed to it will be turned down in response to signal amplitude.  So, whatever noise contribution the input signal might make is attenuated, but whatever noise contribution the op-amp stage makes on its own will remain in place no matter what.

The other manner in which variable-gain/variable-level distinctions can make a difference is in the ease of determining the feel or response of the unit.  The virtual pot in the OS, formed by the 82k input resistor and JFET+trimpot, observes all the other rules of pots, including taper.  This makes it delightfully easy to modify the attenuation response, merely by monkeying with the appropriate resistance values.  This is precisely what has been done with the Amber Press, by means of the R4/VR8 combination.  These two provide different possible values for the "input leg" of our virtual pot, to determine how it will respond to transient peaks.  Accomplishing those different sorts of responses would be somewhat harder to do with a variable gain arrangement.

So, in this case, it is not so much a matter of whether they do or don't sound different in their compression action, but rather what each type of design permits in the way of easily achieving a given compression action.