Zero Attack with Variable release...Ideas?

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

Swedish Chef

Well-known member
Joined
Jun 7, 2004
Messages
351
Location
London
So I'm trying to make a peak detector which responds instantly to catch a...well...a peak, so I want it not to chop a 10kHz waveform off in it's prime, or even a 20kHz come to that, otherwise you get nasty artifacts, but I then need the release to be variable so that the thing doesn't just end up sounding distorted because it's envelope following.
A bit like a PPM meeeeteeeeer....Hang on!
I'll be back!!!!! :green: :thumb:

chef
 
..find a peak-and-hold circuit in your opamp cookbook..

If you want to use such insanely fast attack times in a compressor, you may want to use a zero-crossing detector and'ed to the voltage, so you only adjust gain when zreo crossings occurs.

Yep, this will set you back half a cycle timing-wize, but it will lower distortion significantly.
 
I did think about the sample and hold thing but what do I use to decide when it samples? Use a threshold so it only samples above that threshold, then holds after it drops below??? :shock:

chef
 
Yeah if i could articulate my thoughts that's what I'd have written...Idiot :oops:
That's what i was thinking when i was talking about the thresh...
 
You probably do not want an infinite attack. There are many peaks too short to worry about. Dealing with them means unnecessarily reducing gain.

But the concept is simple. A perfect rectifier and a capacitor, with zero series resistance in the rectifier and large shunt resistance on the cap.

For a 0.0000125 second (0.01 millisecond, 10 microsecond) time constant (1/4 cycle at 20,000Hz), and a 1μFd cap, this suggests a 12Ω rectifier series resistance.

Also if your control voltage goes to 12 volts, a 1 AMP peak charging current.

We see why the Fairchild 660, which only tried for 50 μSec attacks, uses a 10 watt amplifier to drive the rectifier.
 
Ok if I don't want infinite attack but do want the one you spec, how practical is that?
Is it possible or impractical? Is a 12R series resistance likely?
Yours flounderingly...etc etc.
 
If you want to play around with some values, you can stick a small resistor in series with the main charging cap (like 47ohms, re: a 0.1 - 0.5 uF cap). this will give you instantaneous peak voltage near where you want to be, and utilizing a peak-hold release shouldn't produce enough distortion to hear readily.

Some broadcast peak-limiters used this (Urei BL-40 for example).
 
[quote author="PRR"]Also if your control voltage goes to 12 volts, a 1 AMP peak charging current.
We see why the Fairchild 660, which only tried for 50 μSec attacks, uses a 10 watt amplifier to drive the rectifier.[/quote] Methinks PRR is being very factual, but a bit tongue-in-cheek, if you know that phrase or cheeky if you know that word instead!

These requirements are possible, but quite a tall order.

HTH!
Peace!
Charlie
 
> how practical is that?

What is "practical"? A Fairchild 660 is quite practical in a record cutting plant, quite impractical in field recording.

And it matters a LOT how much voltage you need. The Fairchild had to swing 8 grids 60V without a DC amplifier. The grid leakage set an upper limit on the release resistor value. The longish decay time sets the capacitor value. And the very short attack time and high control voltage requires a large peak drive power. Many folks would think the 660 is way too large. With modern parts we could do something like it much smaller, but still not a teeny job.

And if this is a limiter (sounds like it; but not certain), are you sure you really want a hyper-fast attack time? Clipping less than 1 millSecond (1,000 uSec) is pretty darn inaudible. The exception is something like record-cutting, where a peak may throw the cutter into the next groove or even right across the room. That's why the 660 uses 50uSec, and in practice there aren't any tracks that have faster peaks. It isn't easy to generate very rapid risetimes with mechanical objects (musical instruments or their speakers), and would be painful if they did.

And without a multi-rate time filter, catching every little transient is likely to keep the limiter in GR and low output when the peaks themselves could be clipped inoffensively.

Also, if you want the same box to be able to give the more common and musical 1mS-50mS attack times, it is hard to get a pot to vary 1,000:1 in resistance.

But, just for discussion, you want 10uSec in 1uFd, you can probably use a fast op-amp driving a power transistor and rectifier, with feedback to negate the rectifier drop. 1A peaks would be trivial for TL071 and 2N3055.

You better bypass the heck out of the supply rails, and be very sure where the capacitor return current is dumping to. That's one advantage of the 660's transformer coupled rectifier: cap current is local to the time filter and never gets back into the supply rails to click and thump.
 
Some of my beach reading on holiday for the last 2 weeks was the Dual Noise Gate/Fader/Panner from Home & Studio Recording Sept 1988. The attack specification was 10usec to 20sec, and very different circuitry to anything else I have seen. Now I should think this could be adapted into a compressor side chain, could it not? I'll get it posted when my jet lag has worn off!
Stephen
 
[quote author="PRR"]> it matters a LOT how much voltage you need. The Fairchild had to swing 8 grids 60V without a DC amplifier. The grid leakage set an upper limit on the release resistor value. The longish decay time sets the capacitor value. And the very short attack time and high control voltage requires a large peak drive power. Many folks would think the 660 is way too large. With modern parts we could do something like it much smaller, but still not a teeny job.

And if this is a limiter (sounds like it; but not certain), are you sure you really want a hyper-fast attack time? Clipping less than 1 millSecond (1,000 uSec) is pretty darn inaudible. [/quote]

Well now. this slipped between the cushions didn't it?!! :grin:
Yep it certainly is a limiter I'm thinking of.
Basically I have a VCA comp and I want to graft a limiter onto it so that and big spikes get hacked off at the knees. I have no idea what it sounds like to have an attack time in the order of 10s of milliseconds so I am tempted to try and build it just to see what it's like, good or bad. I shouldn't think i'll ever hear a 660 so although it's a totally different beast I'm still intrigued.

check out http://www.symetrixaudio.com/tech_support/schematics/501_1F05.pdf
and the limiter section that is on page 2 i think. someone beat me to it but the diode gate thing looks simple and I almost understand the i/p circuitry :?

[quote author="PRR"]> But, just for discussion, you want 10uSec in 1uFd, you can probably use a fast op-amp driving a power transistor and rectifier, with feedback to negate the rectifier drop. 1A peaks would be trivial for TL071 and 2N3055. [/quote]
Ok it's worth a punt. Will get on it next week.

[quote author="PRR"]>You better bypass the heck out of the supply rails, and be very sure where the capacitor return current is dumping to. That's one advantage of the 660's transformer coupled rectifier: cap current is local to the time filter and never gets back into the supply rails to click and thump. [/quote]

Ok heavy bypassing but if this thing isn't isolated like the gr amp on the 660 how to get around those pesky clicks and thumps?

Sorry just thinking out loud so don't worry about me to much :green:
Will get my head back own soon...

Chuf[/quote]
 
Swedish Chef - have a look at this:
http://img.photobucket.com/albums/v317/StephenGiles/timecomp.gif

The specs state attack times of 50usec to 50msec.
The circuit description states "Q1 buffers the output of IC4C, increasing the amount of current which can be sourced. C11 is a reservoir capacitor which provides a generous supply of current with which to charge C12. R36 serves to isolate the supply rails. In this way fast attack times of the order of 50usec can be achieved"
Stephen
 
Swedish Chef - Here is the 3G Dual Panner/Gate sidechain circuitry which provides attack times as short as 10usec. Let me know if you would like me to post the circuit description. I don't know if this is of any help in your quest for a zero attack compressor, but I should think this circuit could be adapted fora compressor - valve police allowing!!!!!!

http://img.photobucket.com/albums/v317/StephenGiles/gate1.jpg
Stephen
 
For very fast attack and 1uF cap you pay very high price-
insignificant at first glance "crackhtz" induced on power supply.
Of course, it can be filtered, but I did fast detector for my
dual detector compressor this way: buffered "video" opamp-fast high conductance low leakage diode-cap-47nF-variable release circuit
( voltage controlled current source).
If you don't want to mess with current source, put something like 2 meg.
to gnd for release.
Also, cap must be poly, great results-silver mica
(if you can find silver mica 47nF :)
And allways have in mind, detector's power supply must be filtered
(decoupled) best way possible from audio.
Have a fun!
 
ijr - do you have a schematic for your "dual detector compressor " that we could look at? I suppose you are still having warm weather in your part of the world - it's getting colder each day here in England!
Stephen
 
Back
Top