automatic timing in compressor confusion

[Petrol Vendor]

Hi everyone,
This is my first post, i am very happy to become part of this community.

I have been checking out some ways to add Automatic release to my current compressor design (diode ring using quad transistor arrays wired as diodes).
Nothing seems to work...
I am confused, maybe someone has a thought or can point me to some reading.

1st try
The circuit in the ssl clone at gyraf has two sets of resistors and caps in series when in auto mode. http://www.gyraf.dk/gy_pd/ssl/ssl_sch.gif
This seems to behave like a single cap/resistor.
I was looking at the cv in my scope while hitting the sidechain with short, long, loud and soft bursts of noise. The release timing doesn´t seem to be affected.

2nd try
Then i built the non linear capacitor as in THATs Design note, but driven by a simple peak detector, a 10k resistor and a diode.
http://www.thatcorp.com/datashts/dn114.pdf
I used the exact components, but an old tl084 as the opamp.
Acts as a normal 1uf cap, no matter what. One of the leds is giving a short flash with every note.

3rd trie
Another thing i tried was a normal AR generator, and another one with fixed slow values in parallel. They were summed with diodes.
This generates something like a dual slope characteristic. The longer the signals, the more the cap from the slow network will be charged, the sooner it´s slow release will be of higher level than that of the short network. THis sounds kind of smooth, but i´m not sure if it´s whats normaly considered auto release.

I´m especially confused why the ssl and that circuits won´t work for me.

Cheers,
Jens

[mediatechnology]

In your second try using the NLC you mention a 10K resistor. Where was that in the circuit?

Thank you for joining us!

[Petrol Vendor]

Its not in the circuit and i forgot to talk about it.
Without it the whole thing was acting like a single cap to ground. So there was no filtering, just the rectified audio.
But now as i think about it, with the resistor the thing still acted like a single cap...

[JR.]

I'm not sure what you mean by "automatic" .

The non-linear cap THAT application, and some circuits I've posted here and about, use an opamp configured with a cap to act like a capacitance multiplier. This makes a small cap look like a larger cap only while the opamp is active. By clamping the opamp output for a finite voltage or current swing, large currents charging or discharging the active cap circuit will exceed the opamp's ability to make it look like a big cap, and it will respond like a small cap, until the charging current falls back to within the correction range of the opamp.

This can also be though of as a ripple filter. I have used this in dynamic processing side chains before and it works well to reduce side chain ripple for small changes while responding quickly for large changes.

JR

[Petrol Vendor]

Thank you for your replie.
So could it be that my Sidechain level was simply too high? I have used some gain, giving me a rectified Wave with about 8V above ground for loudest signals.
I have always been doing this to reduce the voltage drop effect of the diodes before the AR generator.

So maybe even soft signals were too much for the capazitance to be multiplied?

If i get this right, there is a destinct threshold point where the capacitance seen will switch from 1.01uf to 11uf with the given values? Or is there a soft trancision?

[JR.]

without seeing a specific schematic I can't guess what is happening..

referring to the THAT circuit you linked to, if you look at the opamp output pin you can see what is going on for small signal changes you will see the opamp swinging up and down . For a large signal the opamp output is clamped by the LEDs either positive or negative. When clamped by large changes the c looks like the 1 uF in parallel with .1uF, when active it looks like many times more than 1uF.

JR

[Petrol Vendor]

Well, it was exactly that circuit i was using, driven by a peak detector , diode and 10k resistor. 2m to ground to discharge the cap.
One thing i noticed: One of the LEDs was flashing with every note thet hit the sidechain. Just a short flash, no matter if the signal was long or short, loud or soft.
When i removed the LED the responce from this circuit was many times slower.
I start to figure that the signal level i was feeding to the circuit was way too high.
But you are saying that its large changes that will clamp the opamps outputs?
So it´s not signal level, but signal envelope that determins if a small or large cap is seen?

[JR.]

Well, it was exactly that circuit i was using, driven by a peak detector , diode and 10k resistor. 2m to ground to discharge the cap.
One thing i noticed: One of the LEDs was flashing with every note thet hit the sidechain. Just a short flash, no matter if the signal was long or short, loud or soft.

When the leds are visible, the opamp is clamped and the circuit is fast, when the leds is dark the circuit is slow.

The concept is for steady state signals to be changing little enough to enjoy slow time constant, but be fast for large changes.

It appears the circuit is actually working.

When i removed the LED the responce from this circuit was many times slower.

Yup... if you don't clamp the opamp it works until it runs out of power supply. I did one design where I used the opamp clipping on the voltage rail as a limit for a really fast attack, but that is a more complicated concept.

I start to figure that the signal level i was feeding to the circuit was way too high.
But you are saying that its large changes that will clamp the opamps outputs?
So it´s not signal level, but signal envelope that determins if a small or large cap is seen?

It's a little complicated to describe, the opamp has multiple feedback paths, the 8 Meg R is actually in parallel with the 100nF cap until the LEDs clamp the signal swing, then the LEDs are in the path, and the LEDs also charge the bottom of the 1Uf cap. So its a current and rate of change thing. If you had a smaller discharge R than 2M you might see the other LED light during fast decay. It might already be clamping on decay, but the 2M release R is only pulling microamps through the decay LED so it won't look very bright.

An oscilloscope is useful for circuit design. The circuit works, but you need to dial it in for specifically what you want...

JR

[Petrol Vendor]

Hi everyone!

First of all, thanks for all the input!
The circuit was working. Only i was expecting different behavior so i didn´t notice everything was fine. When i compared the circuit on my scope to an 1uf cap charged by the same signal the difference was very clear.
Now in the past weeks i spend some 60+hours on the breadboard and with the soldering iron, building up several compressors and toying with the values.

Some observations:
The circuit can be used with variable Attack and Release pots. The release Pot shoud be connected to ground, not V- as the 2M2 resistor in the That document. Otherwise you get an negative offset variing with the release time.

It seems like by using two leds in series the circuit can handle louder and faster signals without clamping, although i´m not sure about that.

I replaced the 220r resistor on the opamp output with a trimmer. I can see no change in behavior for values between 0 and 500r

By connecting the inverting input of the opamp to ground you get the responce you would expect from just a 1uf cap. Makes a completly different charakter.
With some settings and signals either one of these sounds better, so i includet a switch in my designs.

Also a 10uf cap to ground in parallel can be nice for a slower responce. Even in that case it can sound very different if the opamp is active or not (mod mentioned above)

I don´t like having too many switches, but these things can be really interesting soundwise.