Different types of compression, SSL vs 1176 vs 1a2a vs DBX160 ect...

[DaveGamble]

Would the approach taken with the Gates Sta Level still be a valid approach today? Or more accurately what I'd like to know, is what might be some of the reasons to approach compression the way the device did?

Ok, so the approach taken with the Gates Sta Level is this:
3x valve gain stages; lift some output, rectify, and pull that out of the input.
So it's one large feedback circuit, with some slow-to-react elements.

It's less a compressor, more a self-defeating mic pre.
It's not the kind of thing you could have attack and release controls over (independently), nor is it something that gives you any real control over HOW it compresses (threshold and ratio kinda doo).
From a guess (without actually graphing it), I'd guess that it doesn't really HAVE a threshold; there's an infinitely soft knee that tapers slowly from 1:1 to X:1 at max, then flattens out after 40dB of attenuation.

So, it's not a compressor, per-se, in that it allows you to leave some audio unaffected, but it's an interesting and characterful dynamic effect, and why not?

Now, I get that you're not the designer of the original and I'm not expecting that. But more I'd like to understand a bit more of the possible conceptual idea behind the circuit.

Ah, ok then. It's a preamp, with a lot of gain, which takes some of the signal and subtracts it back off, in feedback... but slowly. Another way to describe it is kinda like a level-controlled distortion. Audio goes through, and the distortion gives a certain amount of gain, but the output is used to determine the gain from the distortion... but those changes take time to take effect.

It's a slowed-down distortion. There you go, that's what it is
It never really calculates any actual data about the signal, it just wiggles

It's not exactly vari-mu though. I'm not REALLY sure we can call it that. Because that's not really what we're doing.
Vari-mu is more strictly when you play with the pentode grids... but that's not what's happening here.

Even if it is just a best guess. An example that does not deal necessarily with a strict compressor, but De-essers (HF limiters more or less). There are a multitude of ways to approach that particular problem. Ranging from a simple HPF feeding into a sidechain to more complex phase cancellation designs. But they all strive to do the same thing, remove excess sibilance. I'm just looking for a best guess on why one might approach compression the way that the Gates Sta Level did.

1. Options. Valves were the only thing around, so it was going to be a valve design.
2. The need to gain reduce. Your choices were vari-mu or... something else.
Most likely the designer wanted a really fast response, and wasn't satisfied with what he could get with a vari-mu design, so he played around a bit. The circuit is a 3-stage (tri,tri,pent) class-AB amp, with the feedback mechanism that adjusts the grid voltages for the first 1,2,3 stages.

If you break it down, it's really the same as any other kind of feedback compressor. Just a very simple, nifty design!

Dave.

[DaveGamble]

Re Portico it has the Feed Forward also but yes we could look at the dbx also...just wondering did you listen to the portico and hear any of what I was talking about?

Phase cancel the two and you're left with roughly the original signal 20dB down, so, whilst it's almost certainly calibration, the hardware and the plugin are not generating the same GR signal. It could be nothing more than threshold calibration, but they're not matched up.

That said, they certainly seem to have the same approximate character overall (I only listened to FF, not FB).
You're in a tricky position, because if you compared two of the hardware with fractionally different settings, you'd be exactly where we are now with the hardware and the plugin.

If you have an audio precision, or prism dsound, or the like, i can tell you how to set it up to do a compressor comparison?
Or maybe it's time to build some test tools?

Dave.

[SWAN808]

Re Portico it has the Feed Forward also but yes we could look at the dbx also...just wondering did you listen to the portico and hear any of what I was talking about?

Phase cancel the two and you're left with roughly the original signal 20dB down, so, whilst it's almost certainly calibration, the hardware and the plugin are not generating the same GR signal. It could be nothing more than threshold calibration, but they're not matched up.

That said, they certainly seem to have the same approximate character overall (I only listened to FF, not FB).
You're in a tricky position, because if you compared two of the hardware with fractionally different settings, you'd be exactly where we are now with the hardware and the plugin.

If you have an audio precision, or prism dsound, or the like, i can tell you how to set it up to do a compressor comparison?
Or maybe it's time to build some test tools?

Dave.

it could be that in this instance - but it wouldnt explain why I hear the same characteristic with most hardware/software comparisons...

Recently Slate has been talking about the amount of saturation present in hardware compressors and how that adds to the sound - and how no one else has got it right yet (as no one does before he launches a new plugin)...

http://www.gearslutz.com/board/8807610-post305.html

[DaveGamble]

it could be that in this instance - but it wouldnt explain why I hear the same characteristic with most hardware/software comparisons...

Lets get a few together and do a test!

Recently Slate has been talking

I'm sure he has. He tends to. I think it's adorable.

Dave.

[FeelingMachine]

[SWAN808]

it could be that in this instance - but it wouldnt explain why I hear the same characteristic with most hardware/software comparisons...

Lets get a few together and do a test!

Recently Slate has been talking

I'm sure he has. He tends to. I think it's adorable.

Dave.

ok Im up for it!...Mmm...the ones I have at my disposal dont really have emulations...Ive got currently...

Drawmer 1969
Overstayer FET
Alesis 3630
Focusrite Compounder...

Meaning to pick up a dbx 160xt at some point...

[DaveGamble]

it could be that in this instance - but it wouldnt explain why I hear the same characteristic with most hardware/software comparisons...

Lets get a few together and do a test!

Recently Slate has been talking

I'm sure he has. He tends to. I think it's adorable.

Dave.

ok Im up for it!...Mmm...the ones I have at my disposal dont really have emulations...Ive got currently...

Drawmer 1969
Overstayer FET
Alesis 3630
Focusrite Compounder...

Meaning to pick up a dbx 160xt at some point...

Why not run some loops through those, and we'll see what we can do to match up against them!
Sounds like a decent enough set to test the "lower mid" theory!

My preference would be to start with a loop (let's use the drums from the Portico test), and then for each unit, do one test with medium settings, and another with extreme settings (so you can really hear it doing something, quite audibly).
Also, and this might seem like an odd request, can you bounce the loop through each unit with it bypassed? (and do one of the loop through your DAC and back in again, so we can compensate for delay/phase shift through your convertors with ease?)

Then we can test a whole bunch of fun things. I'll happily knock up (and save) a mod in Compassion which I'll limit myself to a few minutes to do (someone will outdo me, of course, but that's half the fun of this).

What I'll do is:
1. Compare the non-processed version with the "just through the DA/AD" version, and see what happens with a phase cancel.
2. Load the non-processed version onto a channel with a plug on and flip phase, and mix that with the processed signal, calibrate overall gain, and then match up the compression characteristics.

This will tell us quite a few things. I'm sure I'll be able to get you a few dB of cancellation, and someone really working on it (to EXACTLY replicate your knob settings, which are never exact with hardware, and even switched is a little hopeful, will take very careful fine-tuning, because pots are a bit random).

To make the challenge interesting, rather than impossible, please stereo-link the units (so there's only one set of controls), or just do it in mono, because without switched controls, it's basically impossible to get both channels knobs set identically.

Dave.

[APZX]

Gamble,

Well that bit pretty much describes the way the in which I notice the compressor to behave. Thank you for explaining it! What I think is most interesting is that it doesn't necessarily behave much like a compressor but at the same time it does behave like a compressor. I guess I had never thought to consider it as a Self-defeating Mic pre. But given the context of when it was designed and such that actually makes a huge amount of sense.

I do think you are right on the design not having a threshold hold per se. I notice with it that it can be driven very hard and that harder it is being driven the more it actually compresses. Getting into more limiting territory. But I think the most interesting characteristic of the design (besides the running stuff through it sounds better aspect) is just how it responds to different material. Example, if you run a bass through it and really ramp it up the bass doesn't actually seem to be compressed as much as it just becomes as solid as a brick. No other way to put it. But it only really does this if the GR is being pushed up towards 15 - 20dB. Another example is acoustic guitars. It just has a way of controlling them even with some pretty crazy amounts of GR going. If you push it too hard then it pumps, but if you back it off a couple steps all of sudden the guitars just come to life. But honestly, where I find it really excels is actually on drum buss and master buss duties and for almost the same reason. The key here is to not drive the compressor hardly at all. Just the occasional move of the needle is all that is needed. What you get out of it is just some wonderful breaking up of transients and a slight presence (6-8KHz) coloration. It is a really fun piece of kit to work with.

[djscorb]

Yo!

Hi Dave, I was going to send you an email but this thread seems to have your attention and the info might be appreciated by others too!

Sooooo....

How does one go about setting up compassion to emulate a classic DBX 160 compressor?

I read that it has program dependant attack as well as release? Is this achieved by the advanced attack settings (obviously there are settings for auto release, but not for attack)?

These are the key specs form the DBX site for the 160A:

Compressor Threshold Characteristic OverEasy® or hard knee
Compressor Ratio Variable 1:1 to :1 thru to -1:1; >60dB maximum compression
Compressor Attack Time Program dependent; typically 15ms for 10dB, 5ms for 20dB, 3ms for 30dB
Compressor Release Time Program Dependent; Typically 8ms for 1dB, 80ms for 10dB, 400ms for 50dB; 125dB/sec rate
Dynamic Range >113dB
Gain Adjustment Range Variable from -20dB to +20dB
Stereo Coupling True RMS Power summing

The infinity ratio of the DBX 160 is easy to get with Compassion, what else do we need?

When would you use the Peak as opposed to the Sum detectors in Compassion? What hardware units would you be emulating when you use Peak as opposed to SUM?

It really is time more light was shed on how to emulate different compressors using Compassion. So many people ask about it and I'm sure many more copies would be sold if people like myself who are obsessed with compressors but haven't your advanced knowledge could go about replacing their other plug in compressors with it

Cheers

Scorb

p.s. I really want to capture as many of the API2500's characteristics in Compassion mods too. I can go about getting the attack and release shapes right, but feedback mode is affected by make up gain (especially in Auto makeup) which I really like. Is this beyond the scope of Compassion's dynamics engine?

[DaveGamble]

Gamble,

You are more than welcome to call me Dave! Everyone else does

Well that bit pretty much describes the way the in which I notice the compressor to behave. Thank you for explaining it! What I think is most interesting is that it doesn't necessarily behave much like a compressor but at the same time it does behave like a compressor. I guess I had never thought to consider it as a Self-defeating Mic pre. But given the context of when it was designed and such that actually makes a huge amount of sense.

Great! I wasn't completely sure how much help I was being

I do think you are right on the design not having a threshold hold per se. I notice with it that it can be driven very hard and that harder it is being driven the more it actually compresses. Getting into more limiting territory. But I think the most interesting characteristic of the design (besides the running stuff through it sounds better aspect) is just how it responds to different material. Example, if you run a bass through it and really ramp it up the bass doesn't actually seem to be compressed as much as it just becomes as solid as a brick. No other way to put it. But it only really does this if the GR is being pushed up towards 15 - 20dB. Another example is acoustic guitars. It just has a way of controlling them even with some pretty crazy amounts of GR going. If you push it too hard then it pumps, but if you back it off a couple steps all of sudden the guitars just come to life. But honestly, where I find it really excels is actually on drum buss and master buss duties and for almost the same reason. The key here is to not drive the compressor hardly at all. Just the occasional move of the needle is all that is needed. What you get out of it is just some wonderful breaking up of transients and a slight presence (6-8KHz) coloration. It is a really fun piece of kit to work with.

Yeah, it's a nice strategy. To Compassion it, you would:
1. Kill Attack and Release (to zero) and use VCA smoothing (which is really more like GR smoothing)
2. Set a midrange threshold (just to get things going), and set ratio to limit
3. Use "analogue knee" at a ridiculously huge value, so you get it progressively tapering in
4. Dial in "bleed", to get it to give you that overeasy behaviour
5. Keep the ceiling engaged at about 40dB, with a max ceiling curve setting
6. Set curve law to linear pls.
Adjust to taste.

Dave.

[Funkybot's Evil Twin]

Dave, you mentioned developing some testing tools a few times already, I just want to chime in and say that would be fantastic. What specifically were you thinking of building?

I'd love something that could setup a blind A/B test from two signals, that would also make level matching signals easy. It would be great if it could clearly show the before/after waveform.

[DaveGamble]

Yo!

Hi Dave, I was going to send you an email but this thread seems to have your attention and the info might be appreciated by others too!

Fair enough!

Sooooo....

How does one go about setting up compassion to emulate a classic DBX 160 compressor?

Is the 160 a classic already? I can't shake the idea of it as a really standard piece of outboard.
When I'm mixing FOH, there's a set of 160s around maybe 50% of the time.

I read that it has program dependant attack as well as release? Is this achieved by the advanced attack settings (obviously there are settings for auto release, but not for attack)?

Oh, no. The dbx "program dependent attack and release" is an RMS detector set to "slooow".

These are the key specs form the DBX site for the 160A:

Compressor Threshold Characteristic OverEasy® or hard knee

I love OverEasy, because it's just sticking a diode in the GR path.
Use analogue knee.

Compressor Ratio Variable 1:1 to :1 thru to -1:1; >60dB maximum compression
Compressor Attack Time Program dependent; typically 15ms for 10dB, 5ms for 20dB, 3ms for 30dB
Compressor Release Time Program Dependent; Typically 8ms for 1dB, 80ms for 10dB, 400ms for 50dB; 125dB/sec rate
Dynamic Range >113dB
Gain Adjustment Range Variable from -20dB to +20dB
Stereo Coupling True RMS Power summing

That means use "sum" rather than peak, for stereo linking.

The infinity ratio of the DBX 160 is easy to get with Compassion, what else do we need?

RMS. Lots of RMS time. Try something like 60ms.
And some analogue knee.

When would you use the Peak as opposed to the Sum detectors in Compassion? What hardware units would you be emulating when you use Peak as opposed to SUM?

Oh, that's to do with linkage!
So, peak is "respond to the peak of either channel", whereas sum is "mix the two channels (and allow them to potentially phase cancel)".
Only worry about this is your signal is highly decorrelated (stereo wide), and you're not getting as much compression as you feel like you should. If this happens, hit Peak.

It really is time more light was shed on how to emulate different compressors using Compassion. So many people ask about it and I'm sure many more copies would be sold if people like myself who are obsessed with compressors but haven't your advanced knowledge could go about replacing their other plug in compressors with it

The trouble is, emulating different compressors with compassion still requires being able to understand the circuitry. If you can do that, then you can emulate compressors without code... which is neat... but it's still quite an obstacle. :S
I feel like it's a problem I haven't solved yet. I've been considering a few things.

I had one idea which is to build a plug (or a set of plugs) in which I can have specific circuit models, and you can flip through 'em to play. Might be interesting. Might be lame.
Still pondering. (it should not escape the observant, that having finished EQuilibrium now, I am on the lookout for a new quest...)

p.s. I really want to capture as many of the API2500's characteristics in Compassion mods too. I can go about getting the attack and release shapes right, but feedback mode is affected by make up gain (especially in Auto makeup) which I really like. Is this beyond the scope of Compassion's dynamics engine?

Errrmmm. Yes and no. With the makeup gain INSIDE the feedback loop, it's actually not a makeup gain, it's an input gain... so if you can remember to use the input gain to drive it, instead of makeup, then it'll work out fine.
Also, don't forget bleed and ceiling!!

Dave.

[DaveGamble]

Dave, you mentioned developing some testing tools a few times already, I just want to chime in and say that would be fantastic. What specifically were you thinking of building?

Weeeeeelllllllllllllll since you're aaaaaaaskiiiiing........

1. A plugin Audio Precision; could be great for testing things, and it'd bring everyone back to measuring things in a precise way. Obviously the magic of the hardware is the build quality, but I still would LOVE to have the signal generators and analysers in a handy plugin. (click menu, click thd+n, smile broadly. click menu, click a-weighting, smile again). Has anyone beaten me to this? I mean something where I can click "sweep" and it makes graphs; the whole works.
2. An ABX tester: you insert it on a channel, route in B via the sidechain, pick the number of tests you want to run, and the type of test, calibrate the fadeout/in time, press go, and test things. You press A or B, it fades out, fades in new audio, and you repeat. Then it presents you with results and statistics.
3. Dynamics testing plugins - the very simplest is a stepped-tone generator (how sick would this be as part of (1)??), loud tone, quiet tone, silence, repeat, which measures attack/release curve shapes. That and a compressor knee measurement. And the frequency response plot from (1). Tells you a lot about a dynamics processor.

I'd love something that could setup a blind A/B test from two signals, that would also make level matching signals easy. It would be great if it could clearly show the before/after waveform.

Why can't I build something that level matches? I can.
I can just correlate two signals; you could leave it playing for a while and it would line up level & delay (and frequency response). You'd need to leave it for a while. Or do it offline. Actually that would be MUCH better. That would really work.

There's a mastering console on my drawing board which is sorta related to some of these things.

Hmmm.

Ideas.

Dave.

[Shy]

I'd love something that could setup a blind A/B test from two signals, that would also make level matching signals easy. It would be great if it could clearly show the before/after waveform.

I think the quickest and simplest way is to use foobar2000 (install the ABX Comparator component). Encode two wav/aiff files to flac (right click, convert), apply ReplayGain, which is generally very accurate for level matching, to both (right click, replaygain, scan per file track gain), mark both tracks in the playlist, right click and run the ABX tool. If you want ABC/HR (Hidden Reference) testing, that's something else, check available tools.

[Funkybot's Evil Twin]

I had one idea which is to build a plug (or a set of plugs) in which I can have specific circuit models, and you can flip through 'em to play. Might be interesting. Might be lame.
Still pondering. (it should not escape the observant, that having finished EQuilibrium now, I am on the lookout for a new quest...)

That'd be great! I've been hoping you'd release a COMPact for a while now. I helped a friend mix a record at his appartment, and it was my first time using the most recent version of Logic. There's a compressor in there that does just that and offers a few different circuits/modes. I thought it was a pretty solid compressor and really liked the ergonomics of it. If you were to go that route, do you think you would build in different saturation stages too?