Measuring Compression Release Time

I've been thinking about how one would go about measuring the release time of a compressor and haven't come up with anything that makes too much sense or seems very easy. The only thing I can think of is to do an amplitude sweep up and down, and then look at the difference between the two traces around the known threshold point. The time it takes for the trace to join back up after the deviation is the release time.

How is it really done?

When I was designing comps I used a tone burst generator that stepped between two amplitudes to measure attack and release times.

JR

..but note, that there are no common standards for what the time number actually refers to..

> only thing I can think of is to do an amplitude sweep up and down

How fast?

Think real-life. Loudmouth Bob interviews soft-spoken Lois. For listener (or recorder) convenience, you want them both the same loudness. Set the gain so Lois is good and loud, set the threshold so Bob gets smacked-down to Lois' level. How long does it take after a Bob-shout for Lois to become fully loud again?

As JR says, it is usually done with a tone-burst. The burst drives the compressor into gain reduction. You usually don't un-burst to zero because you want to watch the gain, which means watching a signal. And since most compressors have some knee-softness, you don't want to work too close to threshold. Conceptually you have a burst to +10dB over threshold and an off-state of say -10dB below threshold.

We usta use a slow-sweep o'scope. Today it may be convenient to get a digi-file and open in a wave editor which has time and amplitude scales.



As gyraf says: there is a lot of interpretation and arbitrariness. It takes infinite time for an exponential to arrive at the final value, but it gets "close enough" long before infinite time. How close is close enough? So that Lois' speech rises to within 1dB? 3dB? And real boxes are rarely perfect exponentials, will have different timing at 5dB overdrive and 25dB overdrive; in fact that's a plus for pleasant audio.

PRR said:

As JR says, it is usually done with a tone-burst.

Conceptually you have a burst to +10dB over threshold and an off-state of say -10dB below threshold.

Click to expand...

What real world hardware box signal source does this. I've been scouring the HP manuals of various oldish models and the only way I can see to do this is to AM modulate a sweep. I don't see a way to switch between two tone bursts in memory. Most of the models need an external modulation source which is impractical for any kind of repeatable timing. I must be missing something. I want a box that doesn't have Intel inside or AP on the outside.

The most common spec for release time I see is dB/sec. I realize where the second starts and ends is somewhat arbitrary. I will label the switch 1-6 because it's easy to write down. And I will tune by ear but I'm curious none the less.

I rolled my own.. I think I may have even redrawn a schematic and posted it here at least once...

Basically I made a simple on/off fet shunt (like a noise gate). I carried it one step further and detected zero crossings, so I could always start and stop on a zero crossing.

I then added a pot to feed signal around the gated off part for the baseline or off amplitude.

Then just add a signal from a sine wave generator or whatever signal source you want to use.

JR

PS: The zero crossing switching detection is useful when designing dynamics circuitry by ear, otherwise the clicks from the tone burst would mask any clicky  artifacts from the processing. I also grabbed the gate drive signal to trigger my scope.

PPS: A rule of thumb for measuring rise time time constants is 90% of final value... this is also close to within 1dB, so I suspect either criteria is adequate for quantifying.  Some comp circuits have signal dependent or adaptive att/rel so may not give uniform time constants in all cases. I also used a hold time before release in some designs so that needs to be looked at separately from the release time.

I used a tone burst generator and a scope (slow sweep) when I was developing my tube limiter. There was much futzing around with controls to get a readable waveform, but it worked.

Thanks gentlemen. So it seems there isn't an off the shelf signal source that switches easily between tone bursts of different amplitudes and durations. I would have thought a fancy HP model would do this. Are us black sheep audio people the only ones who want to do this? I imagined dynamic range manipulation would be used for all types of signals, and the ability to test would be needed.

I made a load of burst tones with a software signal gen... play them through the comp and record back, not quite real time but good enough.

Edited at sample level to ensure I crossed at zero from small to large amplitudes. It's well worth doing it at a few frequencies.

<100Hz is great at showing the attack/release speed mangling of LF waves.

I'd suggest an overall burst length of 6 seconds or so, with the last low amplitude section set to about 4seconds, so long-ish release times can be seen.

-T

My tone burst source was just a regular benchtop function generator.
The main cause of the futzing I mentioned earlier was adjusting the burst duration and the triggering on the scope to get the most readable waveform. But once it was dialed in, all was good.

Here's a snapshot. Obviously, I was mostly concerned with attack time during these tests:




I'm sure you could generate a suitable test waveform in a DAW and use that.

NewYorkDave said:

I'm sure you could generate a suitable test waveform in a DAW and use that.

Click to expand...

Yeah, but I don't want to. The DAW is in the studio. I don't want a computer in the shop. I've decided to collect test gear to do tests. I want to be able to look at phase, bandwidth and amplitude for EQ's so I bought an HP spectrum analyzer. I want to be able to look at compressors too. I'd like to find a function generator that will allow me to do that. Since I'm an aesthete I like the late 80's early 90's HP look. I'll probably get an HP3314A or 3325A. I'm leaning towards the 3314A because it has better bursts.

I vaguely recall some old Wavetek signal generators that did bursts, but they were expensive and sine waves weren't very clean.

These days I suspect some computer generated tone bursts written to a simple MP3 player, perhaps with a click track on the other channel to sync scope from..

JR

What is frustrating is the little handheld Lindos Mini MS10 - or MS-20 will do tone bursts for PPM testing but I don't think there is a way to alter the timing of the bursts or the amplitude.

I might have to pop them an email and see if there is a way to alter the software that drives the thing to allow you to create user test sequences easily. There might well be and in my lack of trying skipped that part in the manual... it would have to provide sync on the right channel output. Hmmm

I'm pretty sure an audio precision can do this.

If they could do this test they would be a much better unit IMO and worth the cash.

-T

Mind you thinking about it, they would need a PC to control them, at which point you may as well go the DAW route... which seems not at all to your liking!

-T

Find the Collins 26U-1 manual PDF online.  I believe it characterizes their method.

A clean, fast-acting attenuator or switch with variable duty cycle and a trigger output would be a great add-on for your typical benchtop signal generator.