Compressor parameters measurements?

[dirty1_1garry]

Hi!

Want to know:

Are there any standard methods (may be industrial) to measure operation parameters of audio compressors?
How it's possible to make a correct measure of attack, release and of course ration parameters?

If audio manufacturers make a marks on faces of their compressor so they should know that these parameters correct, right?) So they should have uniform standard... may be...

 

[JohnRoberts]

I never was aware of definitive standards for measuring consumer gear, while there may be some telephone system standards.
=====
There are usually identifiable time constants in the side chain circuitry, but the control law for the sundry gain elements means these defined time constants for control voltage do not always map out to similar gain change time constants.

An objective way to measure the result of dynamics processing is to use a gated tone burst, preferably one with a gated path and variable dry path so you could dial up bursts or level steps on top of steady tones. This way you can see the the time it takes for a compressor or limiter to reduce gain (attack) then release again, after the level step is removed.

It is very difficult to determine when you are fully attacked and fully released so for convenience perhaps use something arbitrary like reaching 90% of target state. 

JR

PS: I rolled my own burst box, and refined it so that it only turned on and off at zero crossings so you didn't get extra clicks why trying to listen to how well a dynamics circuit was handling the bursts.

 

[dirty1_1garry]

I never was aware of definitive standards for measuring consumer gear, while there may be some telephone system standards.
=====
There are usually identifiable time constants in the side chain circuitry, but the control law for the sundry gain elements means these defined time constants for control voltage do not always map out to similar gain change time constants.

An objective way to measure the result of dynamics processing is to use a gated tone burst, preferably one with a gated path and variable dry path so you could dial up bursts or level steps on top of steady tones. This way you can see the the time it takes for a compressor or limiter to reduce gain (attack) then release again, after the level step is removed.

It is very difficult to determine when you are fully attacked and fully released so for convenience perhaps use something arbitrary like reaching 90% of target state. 

JR

PS: I rolled my own burst box, and refined it so that it only turned on and off at zero crossings so you didn't get extra clicks why trying to listen to how well a dynamics circuit was handling the bursts.

Ok, no direct dependence from electronic parameter for time constant. But as you sad there are objective way of how to make a measure the result of dynamics processing.. Could you describe it more open?))

 

[CJ]

Fairchild 670 was done with human ears of good quality,

time constants were picked  to give the best results for the various instruments used in a typical orchestra,

 

[JohnRoberts]

Ok, no direct dependence from electronic parameter for time constant. But as you sad there are objective way of how to make a measure the result of dynamics processing.. Could you describe it more open?))

Like I said you begin with a tone burst generator capable of adequately long on time and off time, with a dry bypass path so you can mix in un-gated signal.

Here is a crude schematic of a gate I built for my own use back in the '70s, drawn from memory, not actual schemo so caveats apply.

I used a simple JFET shunt to gate the signal off, and a 4013 CMOS flip flop to synchronize the gating to sine wave zero crossings, The other opamp establishes the on time and off time,

Using a clean sine wave source, feed the gated sinewave to the dynamics processor then look at the results with an oscilloscope.(there is a trigger output from the FF so you can sync the scope to the burst). 

Attack time will be how long it takes for the tone burst to get to 90% (or whatever) of its final state. The release time, will be how long it takes for the gain to come back after the burst level drops, To see this release time there needs to be a constant signal playing so you can see the level creep up. 

You need to adjust the on time/off time long enough to not step on these time constants. If you put in a 1kHz sine wave you can count the number of cycles for mSec of attack time.

When I designed dynamics processors I was never very particular about precision in labeling attack/release. This is something you really should adjust by listening for what sounds best. I can appreciate a desire for precision and repeatability, I guess.

Note: In addition to att and release, I have also seen a hold time used in some dynamics processors so a comp would delay releasing immediately, this reduces distortion on low frequency sine waves. In my judgement this was mostly a bench trick to measure better, and didn't sound that different on complex waveforms. More effective IMO was to shift to a slower att/release time-constant when close to threshold, then revert to normal faster time constants when not close. In my experience this sounded more natural in use.

JR

 

[q2audio]

I've done this with Audacity or similar software. Make a 1khz sine wave maybe ten seconds long at some nominal level. Then about three seconds into it select an area about one second long (try to make sure it snaps to zero crossings) then increase the volume of that section by 10dB (you had to leave room for this added level when you first created the wave). Input this tone to your unit under test. This gives you a bit of tone to set up your comps threshold, you want it to be just over the first lower level that you created. Then after selecting whatever attack, release and ratio, etc., you want to test play back the file while recording the output of the comp back to another channel. This makes it easy to see attack, release, etc.

I hope that made sense...

 

[JohnRoberts]

I've done this with Audacity or similar software. Make a 1khz sine wave maybe ten seconds long at some nominal level. Then about three seconds into it select an area about one second long (try to make sure it snaps to zero crossings) then increase the volume of that section by 10dB (you had to leave room for this added level when you first created the wave). Input this tone to your unit under test. This gives you a bit of tone to set up your comps threshold, you want it to be just over the first lower level that you created. Then after selecting whatever attack, release and ratio, etc., you want to test play back the file while recording the output of the comp back to another channel. This makes it easy to see attack, release, etc.

I hope that made sense...

Yup, I should have mentioned this (not available back in my old bench days). While my hardware box was useful for my extensive bench work tweaking processing to sound best, for small steps or large. For simple attack-release measurements, make a sine wave sample that alternates on one second intervals between say -10dBu and +10dBu. Making several seconds of this will make measurements easier.

One nice thing about my box for design work, was that I could feed actual audio signals through my burst circuitry and punch up the dynamic range of existing pre-recorded audio to stress a dynamics path more.

JR