Convert Orange Squeezer to 1176 Style compressor

[POTL]

Hello.
I thought it would be nice to add the 1176 functionality to the simple and compact Orange Squeezer circuitry in order to get a functional compressor in a pedal format.
For obvious reasons, we are not talking about copying sound, only about functionality.
Here are the diagrams:



How I see this diagram
1) Standard buffer on an operational amplifier, it gives high input impedance, then standard volume control, which works like Gain.
2) Q2 + Q3 + Q4 in 1176 is a discrete operational amplifier, we can replace it with a standard operational amplifier like in orange squeezer circuit.
In both circuits, the FET works as part of the voltage divider, in front of the op-amp. We will leave the diagram as it is.
3) Q12 + Q13 + Q14 + Q15 in 1176 is a full wave rectifier.
In the orange squeezer, we have a primitive chain R10 + D1 + C7 + R11.
The idea is to remove D1 and put an op-amp between C6 and R10.
R10 is replaced with a 5kΩ variable resistor and acts as an attack control.
R11 is replaced by a variable resistor of 500kOhm-1MOhm and works as a decay control.
Again, the standard solution.
4) Between C6 and the full-wave rectifier we set the standard volume control, this will work as a ratio.


I have no way to draw a diagram at the moment.
What do you think of a similar solution?

I googled and found something similar to my idea.

[Mark Hammer]

That's one of Rod Elliot's ideas.  Just thought I'd mention that.

Take a gander at Dave Thomas's "What" compressor:  http://dt.prohosting.com/hacks/what.html  Does it compare to an 1176?  Couldn't tell you.  But it's optical and has a more sophisticated envelope-follower than many.  Why mention it here, in the midst of a thread about the Orange Squeezer?  Because, like the OS, it operates on the principle of fixed gain and variable attenuation, rather than envelope-dictated gain.

[POTL]

Thanks for the answer.
It seems to me that the idea of ​​vixed amplification is more practical since we avoid problems with sound distortion.
In addition, classic studio-grade circuits such as the 1176 and LA2A work on this principle.

I adhere to the principle that the fewer components, the better.
I like the Diamond Compressor topology, where you can replace the optics with fet.
I would also replace the detector with something more modern, an op-amp.

[caspercody]

POTL

Did you get this idea to work, and if so, how does it sound?

[POTL]

Hello. No, I don't have time for DIY lately. but the question of sound is rhetorical. The circuit will continue to sound like an Orange Squeezer, just get a full set of controls. I came to the conclusion that I do not like the work of Jfet compressors, because they strongly color the sound and, with strong compression and no mixing, do not sound very nice, like OTA compressors. if I collect compressors, it will be either a transparent VCA or an opto compressor that colorizes the sound.

[Rob Strand]

Hello. No, I don't have time for DIY lately. but the question of sound is rhetorical. The circuit will continue to sound like an Orange Squeezer, just get a full set of controls. I came to the conclusion that I do not like the work of Jfet compressors, because they strongly color the sound and, with strong compression and no mixing, do not sound very nice, like OTA compressors. if I collect compressors, it will be either a transparent VCA or an opto compressor that colorizes the sound.

The feedback compressors are inherently 2:1.   They rely on the detector creating the threshold and sharpening-up the compression curve to get > 2:1 compression.   For the Orange Squeezer, if you bias the JFET just off then it's essentially 2:1 with no threshold.   Things like the Dynacomp have a 2x transistor rectifier which creates a sharper knee on the detector.

There's a good discussion on JFET gain control in National Semiconductors Application note AN129.
See PDF page 252 (pages 6-39 to 6-46)
https://www.rsp-italy.it/Electronics/Databooks/National%20Semiconductor/_contents/National%20Semiconductor%20FET%20Databook%201977.pdf

Some conclusions:
- lower distortion with high VP JFETs
- linearization networks cause control delays (and some feedthrough)
- suggest mods to linearization networks to reduce side effects (by removing the d to g resistor).

FYI, the extra 2k7 resistor on the LA2A circuit can do some weird things to the compression curve if it isn't adjusted to suit the gain element.

[PRR]

The feedback compressors are inherently 2:1.

I don't believe that is true, if you have gain in the loop.

Do you have Barry Blesser's papers? Example:
Audio Dynamic Range Compression For Minimum Perceived Distortion
BARRY A. BLESSER, Student Member, IEEE
IEEE TRANSACTIONS ON AUDIO AND ELECTROACOUSTICS VOL. AU-17, NO. 1 MARCH 1969

[Rob Strand]

I don't believe that is true, if you have gain in the loop.

Do you have Barry Blesser's papers? Example:
Audio Dynamic Range Compression For Minimum Perceived Distortion
BARRY A. BLESSER, Student Member, IEEE
IEEE TRANSACTIONS ON AUDIO AND ELECTROACOUSTICS VOL. AU-17, NO. 1 MARCH 1969

I'm pretty sure it is true because the math for it is fairly straight forward.  The NE57x compander datasheets and app-notes contain the textbook equations.   IIRC the gain shifts the unity gain point but not the 2:1 compression characteristic.

If you add thresholds or non-linearity to the detector you can get different compression ratios.

The common case where the gain of the feedback compressor is adjusted normally shifts the input threshold but in order to have a threshold the detector must have non-linearity.   For the dynacomp there's a 0.7V *output* threshold set by the transistor rectifier and the gain changes the input threshold.   The thing here is the rectifier is very non-linear, once you hit 0.7V it's almost a hard limiter.   The orange squeezer only gets a threshold when the gain control JFET is biased off with no signal.

It's also possible to make an adjustable compressor with a NE57x compander.   It creates a pseudo knee when the gain goes from linear (no compression) at low levels to 2:1 (incremental) at higher levels.  Here the compression curve is less than 2:1.  To go higher than 2:1 you need to incorporate a Feedforward side-chain somewhere (or non-linearity, like the extreme case of a hard limiter example in the NE57x Philips docs.)

I'm pretty sure I don't have that paper, it looks interesting.   It's possible the test set-up has some non-linearity, say at the knee.   The threshold is the most likely point where non-linearity enters the equation.

[PRR]

> The NE57x compander datasheets

Telephony likes 2:1 and 1:2 ratios. Nice round numbers and good performance.

The NE57x were specifically aimed at telephony. (If it had been a little more successful (cheaper) they would sell Billions.)

Feedback control can not reach infinite ratio but can come as close as you want to fuss. 10:1 is easy, 20:1 common on pro gear.

[Rob Strand]

> The NE57x compander datasheets

Telephony likes 2:1 and 1:2 ratios. Nice round numbers and good performance.

The NE57x were specifically aimed at telephony. (If it had been a little more successful (cheaper) they would sell Billions.)

Feedback control can not reach infinite ratio but can come as close as you want to fuss. 10:1 is easy, 20:1 common on pro gear.

They are all feedback.   The point is the NE57x are what happens naturally with a feedback compressor 2:1.  Audio compressors have non-linearity in the detector,  that's *how* they get more than 2:1 compression.

[Rob Strand]

FWIW, here's a JFET feedback compressor with gain in the forward path yet it  produces a 2:1 compression ratio.

The key points:
- The rectifier/detector output is linear.  The control voltage proportional the Vout level (here an averaging rectifier).
- The JFET is biased to be just off when the control voltage is zero.
  This closely approximates the gain being inversely proportional to the control voltage.

By nature of the linear rectifier there is no threshold.  The CR is a constant 2:1 for all inputs.  There is a unity gain "reference" level.

Here's the circuit, simulation and results,


(There's some very fine points I've not elaborated on.)

[ElectricDruid]

The other element that could be modified to become non-linear would be the gain control itself, presumably?

LM13700 or other OTA-based compressors are using a linear gain element. I don't suppose a JFET-as-voltage-divider is very linear, volts-at-gate/output-level, and probably vactrol-based stuff isn't either. So there's some complications going on here beyond just the rectifier response.

But you could also use V/dB VCAs like the THAT Corp parts, or the 2164 VCA, which all have a proper exponential response.

[Rob Strand]

The other element that could be modified to become non-linear would be the gain control itself, presumably?

Definitely, it's the overall non-linearity that counts.  You can think of the transistors of the dynacomp as modifying the linearity but it's not so obvious as the rectifier and non-linearity are combined.   You could split the two processes if you wanted.

While theoretically you can move the non-linearity between the blocks, something perhaps not so obvious is where you put the filtering.   The control voltage is DC but in practice it has to be constructed, for example by a rectifier and filter.  If you put filtering before a "sharp" non-linearity the filter ripple can create large gain modulations and distortion.   That's where sensible implementations will sound better.

But you could also use V/dB VCAs like the THAT Corp parts, or the 2164 VCA, which all have a proper exponential response.

A good example of "good" non-linearity.

LM13700 or other OTA-based compressors are using a linear gain element. I don't suppose a JFET-as-voltage-divider is very linear, volts-at-gate/output-level, and probably vactrol-based stuff isn't either. So there's some complications going on here beyond just the rectifier response.

You can't beat the OTAs for being close to ideal from the control side.

The JFET gain controls are actually quite linear, see the National Semiconductor App Note AN129 - I posted the link in reply #5.  There's more than one way to use a JFET for gain control.  For a JFET used as a divider (like the orange squeezer and my example) some small control non-linearity creeps in because you have attenuation = rds/(Rin + rds) = 1/(1 + Rin/rds).   IIRC, the 1 in the denominator doesn't get cancellled out fully.   Also the JFETs themselves don't agree 100% with the textbook variable resistor theory.   In practice it's small enough to ignore for compressors.

[Rob Strand]

While theoretically you can move the non-linearity between the blocks, something perhaps not so obvious is where you put the filtering.   The control voltage is DC but in practice it has to be constructed, for example by a rectifier and filter.  If you put filtering before a "sharp" non-linearity the filter ripple can create large gain modulations and distortion.   That's where sensible implementations will sound better.

As an example of this, think about a true rms converter.   The squaring is done before the filtering and the softer square root is done after.  While the overall process is actually linear with level, the filtering is done in a non-linear domain.

[Rob Strand]

To complete the story.   The aim of this post is to show what happens when you add a threshold to the detector.

Adding only 0.5V shift to the control voltage on a JFET with a 4V VP greatly affects the behaviour:
- a threshold is created, as expected.   See schem for prediction of knee voltage.
- compression ratio at high levels increases to 13:1. Adding the threshold changes the CR.
- the offset greatly promotes the JFET distortion

The offset is the V0 parameter,



As a side note look how bad the distortion is without JFET linearization.  For the 2:1 compressor example a few posts back linearization wasn't required.   I've even up'ed the time constant by a factor of 10 without improvement.

[caspercody]

I did not like the JFET compressor.

I did find a V2181 chip, anyone have a compressor schematic for this chip? I know there is the Boss CS-3, but i would be better to buy the Behringer version for like $24.00.

Curious about the Keeley GC-2, or Seymour Duncan Vise Grip. These appear to be VCA compressors

[ElectricDruid]

The V2181 is a Coolaudio clone of one of the THAT Corp chips. They have a load of useful application notes for their stuff, and compressors are amongst those applications. Here's the top-level link:

https://thatcorp.com/design-resources/

This is directly what you were after:

https://thatcorp.com/datashts/dn00A.pdf

But this is very much on-topic here too:

https://thatcorp.com/datashts/dn01A.pdf

[Rob Strand]

The THAT docs are very good but keep in mind some of the derivations assume the control signal sets the gain in dB, which is how their chips work.   When you have basic OTA's or other brands of VCA the control signal sets the linear gain.

[Rob Strand]

I did find a V2181 chip, anyone have a compressor schematic for this chip? I know there is the Boss CS-3, but i would be better to buy the Behringer version for like $24.00.

The Boss LMB-3 uses a THAT2181.    Some of the other stuff like Behringer's BLE-100 clone use uPC1252.  There's a lot of similarity between the 2181 and 1252.

It's true those Behringer pedals can be bought for not much more than the price of the chips.   The limiter pedals have a full "professional" structure.   With some tracing effort they can be modded.

[caspercody]

I found a Behringer CS400 on Amazon, and after my points, I got it for $20.00. Cant beat that.

They say it is a clone of the Boss CS3, so guessing it uses a VCA circuit