Calculating Attack & Release Times..... help, I give up:(

I'm glad you returned to this thread, Ruckus, its a very interesting one!

Heres some food for thought, the schematic for the Trident CB9146 FET compressor, which has a meter and a ratio control:

Urgh.....ok, so I'm at my whits end with this thing.  Have probably spent 2 dozen hours in front of spice trying different things, unfortunately I feel I'm no closer to a solution than I was when I started.  Metering issue is solved though, I'm just deploying an LED meter and measuring the CV, with adjustment possible at each indicator LED, so I can adjust for any nonlinearity of the FET.

Here's where I'm at, and the issues I'm having (all ref designators referencing the images below, not my previous pics).  I have alot of questions, have been trying to write this post for about 2 weeks now to ask the board for help, but every time I start to I didn't even know where to begin, so rather than go into all of the issues I guess for now I'll just ramble about these 2 issues for the moment:



1) Still stuck on making variable attack/release.  Problem is that if I parallel a resistor with the cap (C50), I need to block the discharge back at the opamp (U205A).  I think I can do this just via a diode between D11 and R72, but I think that might mess with the compression knee.  But then there's still another issue, the resistor paralleled with C50 would have to be significantly higher than the attack resistor (R72).  Otherwise the voltage divider it will create lowers the signal to the FET's gate, hence never making any compression possible (or best case very little compression).  This means I'd lose any sort of fast release time, unless the 2 values were linked, or I encorporated a way to keep the resistors from interacting. 

Abbey (or anyone else) - you had mentioned doing this, any place you can think of as an example of what I'd want to do?  There's a method for seperate attack & release used in one of the THAT app notes, don't know if it could be tweaked to work in this circuit or not:

http://www.thatcorp.com/datashts/dn115.pdf

I'm wondering if maybe I should just take the easy path and put a variable pot in series with R72, this would link attack and release speeds, but then relying on the resistor back at the TL072 for release speed would be an issue if I wanted to have variable ratios.

Which brings me to:

2) Ratios - I don't know if it's possible with the present rectifier circuit.  Taking some inspiration from the LA4, I've had the best simulation results with changing the rectification section to this:



Problem I'm seeing with this version is in spice with a lower (say 4:1 or 2:1 ratio), by the time any significant gain reduction is achieved the transients in the feedback path are so high that it's clipping U17.

Also, spice seems to confirm what I've always wondered about the LA4 Peak reduction knob - since that is directly in the feedback path, how is it not affecting the ratios?  I mean, depending how much that peak reduction knob is turned, it will be lowering the amount of input signal fed back to the g/r element, so at a lower setting wouldn't it in effect mean lower ratio?  If so then I would have the same issue in my version as well.

Anyone?........Bueller???

I am too busy to immerse myself completely in this, but simple crude FET shunt limiters will have their limitations.

There should be numerous examples of independent attack release controls.  Depth of attenuation, can be limited with variable resistance in series with shunt FET.

Perhaps you need a combination of more tire kicking of old schematics, and slightly lower expectations for what is possible from crude shunt element. 

or not...  I've used plenty of FET shunts for gates, not so much for limiters. but AFAIK pretty popular for bass guitar compression pedals and the like.

Sorry for the non-answer.

JR

ruckus328 said:

1) Still stuck on making variable attack/release.  Problem is that if I parallel a resistor with the cap (C50), I need to block the discharge back at the opamp (U205A).  I think I can do this just via a diode between D11 and R72, but I think that might mess with the compression knee.  But then there's still another issue, the resistor paralleled with C50 would have to be significantly higher than the attack resistor (R72).  Otherwise the voltage divider it will create lowers the signal to the FET's gate, hence never making any compression possible (or best case very little compression). 

Click to expand...

That's the main problem with the rectifier you chose. First, do you need this rectifier that is accurate even for very low input voltages? I say no, because the compressor acts only at high levels, and you need to build some threshold in it, so you can use a very basic two-phase rectifier (see Fig. 1).

This means I'd lose any sort of fast release time, unless the 2 values were linked, or I encorporated a way to keep the resistors from interacting.

Click to expand...

You need a rectifier with almost infinite discharge path or a buffer with infinite discharge path (see Fig. 2).

There's a method for seperate attack & release used in one of the THAT app notes, don't know if it could be tweaked to work in this circuit or not:

Click to expand...

This circuit works well with log control ports as found on most VCA's, low voltages with perfectly linear dB/V law. It may need some tweaking to adapt if for FET control law.

I'm wondering if maybe I should just take the easy path and put a variable pot in series with R72, this would link attack and release speeds, but then relying on the resistor back at the TL072 for release speed would be an issue if I wanted to have variable ratios.

Click to expand...

That wouldn't be nice at all. It's easy to make it right.

2) Ratios - I don't know if it's possible with the present rectifier circuit.

Click to expand...

I don't think so either.

Taking some inspiration from the LA4,
Problem I'm seeing with this version is in spice with a lower (say 4:1 or 2:1 ratio), by the time any significant gain reduction is achieved the transients in the feedback path are so high that it's clipping U17.

Click to expand...

That is quite normal; that would probably define the operational limits of your design. Remember, with a ratio of 1.1:1, you need 20dB over to get 2dB GR.

Also, spice seems to confirm what I've always wondered about the LA4 Peak reduction knob - since that is directly in the feedback path, how is it not affecting the ratios?  I mean, depending how much that peak reduction knob is turned, it will be lowering the amount of input signal fed back to the g/r element, so at a lower setting wouldn't it in effect mean lower ratio?

Click to expand...

remember the LA4 and the 1176 do not have a threshold control; on the contrary, they use a fixed threshold cell with a quite rounded transfer characteristic, which could also could be called variable-ratio. The actual unit's threshold is controlled by the amount of input signal that is fed to this cell.

If so then I would have the same issue in my version as well.

Click to expand...

I guess you would. That's probably the reason why there are no commercial units having both a threshold control and an input gain.

Maybe you should pop open the Ridge Farm Boiler for some inspiration.  It's basically an LMC with attack, release, threshold controls. Also has LED metering.

http://www.ridgefarmstudio.com/rfinew/bprod.htm