DMGAudio Compassion (Launched)

[noidea]

no one?

RTFM...

WOW … very clever answer …


Well, my questions have arisen after RTFM ...

So maybe someone more intelligent?

[DaveGamble]

OK - question time

The RMS setting (under Global) is in ms. I understand a compressor detecting peaks or RMS - but how does that work in ms? I thought it would be more like "either/or".

Does a low setting mean "a little bit RMS"? Does all the way up mean "RMS" as we tend to think of it - ie looking at RMS value of the signal?

There +IS+ no "true" RMS.

Or, if there were, it would be one number which did not change.

What you imagine RMS to mean with a compressor is "root mean squared", yes?
So, we square the signal (straightforward enough; voltage out = voltage in * voltage in)
We take the mean (aaaah.... ok)
And we take the square root (ok, voltage out = sqrt(voltage in))

Ace. Except for that "mean" bit.

In the real world of analogue circuity, "taking the mean" means PRECISELY one thing, which is: time average using a low-pass-filter.

That's what you'll find inside any RMS circuit.

But what low-pass-filter? Well, it's first order, and it has unity gain, so there's only one question remaining... what frequency is the cutoff?

Time is inversely proportional to frequency, so I gave compassion a time selector.

A setting of zero means peak, (because sqrt(x*x) = x).
A low setting means a quick RMS
and a high setting means slow RMS.

Think of it like an extra attack stage.

Dave.

[jbraner]

Well that's certainly a detailed answer! (no surprise there)

A setting of zero means peak, (because sqrt(x*x) = x).
A low setting means a quick RMS
and a high setting means slow RMS.

I think that answers my question.

Thanks!

[DaveGamble]

I sink my teeth into the advanced section and I don't really understand the following parameters:

- AR Coupling

- AR Order (when AR Coupling Off)

- Attack Type

______with AR Coupling On
____________0%: perfect compensated attack circuit
__________100%: uncompensated circuit

______with AR Coupling Off
____________0%: Attack stage behaves like a low-pass Filter
__________100%: perfect attack circuit

- Release Type
________0%: Release to Zero
______100%: Release to Signal

I don't know what is achieved by tweaking these parameters … I don't know when I want to tweak these parameters.

Sorry for my stupid questions.

Please can anybody help me?

Hi,

These are parameters to be tweaked in order to be able to model a specific unit.
So, you take the lid off the unit, draw up the schematics, and take a good look.

Now, if you find one circuit that implements the attack, and then another for the release, that'd be AR Coupling Off. On the other hand, if you find a diode driving a resistor with a cap and resistor in shunt (SSL bus compressor, for instance), that'd be AR coupling On. It determines whether Attack and Release are one circuit (coupled), and hence interact, or are two circuits (uncoupled).

AR Order is for decoupled circuits - which order are the stages in, within the circuit? If the attack is before the release, you'd get a slightly different interaction to if it was after.

Attack type, Coupled: This allows you to choose how well compensated the attack circuitry is overall; you can get a gain drop across a coupled AR stage as a consequence of the resistor values in play. Now, digitally, I can fix that, or I can leave it be.

Attack type, decoupled: Let's be honest, analogue decoupled attack stages are just first-order low-pass filters, but I thought it'd be nice to have a "proper" attack circuit, with instantaneous release, which is what an attack stage /would/ do if it were enormously expensive and well-designed. Your ears are used to the LPF version, but the "perfect" version is... well... "perfect"

Release type: What's the reference signal when the circuit starts releasing? What's the actual discharge path for the release caps? Is there feedback that means that the diode will switch back on as soon as the cap discharges far enough (release to signal), or does the cap just discharge straight to ground, and that diode says "no"? I've seen both, and it's very much a matter of designer's taste. I'd say release to signal is probably more subtle with vocals, release to zero with drums.

These are all incredibly involved parameters which require a decent slice of circuit design knowledge to set deliberately. But the POINT of Compassion is that you can play! Try wiggling them! See what sounds good! The Character Mods are there to give you some starting points (i.e. the older the circuit, the more likely that AR is coupled, British designers love release to signal, etc), but PLAY! What sounds good to you, on this signal?

There are no "right" answers overall, or there'd be a "BEST COMPRESSOR EVAR" that everyone would go out and buy. The fact that dynamics circuits are so various and the results are so subjective is why we have so many different famous compressors! I thought it'd be a lot of fun (and I'm having a lot of fun with it, even if noone else is) to be able to mix and match different circuits, on the fly, at will!

With the right settings, you can get the response of any specific unit (and if you can't tell by ear what unit a Mod is emulating, then it just doesn't matter that much, at least not for this signal), and tweak into others until it's doing what you want.

For problem-solving, you rarely need to adjust any of these things, but for breathing character into things, it can be a magical adventure that I've still not found as satisfying a way of achieving (auditioning a dozen different compressors and starting from scratch with the settings does not appeal to me in the same way as just tweaking things).

Half the fun is creating impossible units (semi-compensated stages!)

Anyway, don't fret too much about the detail; have fun and enjoy!

Dave.

[jbraner]

Dave,

You should put this stuff in the manual

It explains what things are - and it would probably stop most of us from asking so many questions.

[noidea]

Hi Dave,

now I understand!

Thank you very much!

Another Question relating to the Main EQ Parameter LP Split EQ:

I don't understand the difference between "PL" (parallel) and "Split".

Could you help me again?

[DaveGamble]

Hi Dave,

now I understand!

Thank you very much!

Another Question relating to the Main EQ Parameter LP Split EQ:

I don't understand the difference between "PL" (parallel) and "Split".

Could you help me again?

Oh. Favour for Bob Katz.

Split linear-phase splits the signal into "whats between the filters" on that graph, and "everything else", and pushes the filtered part through the compressor, then remixes with "everything else".

PL was something Bob wanted; it pushes the filtered part through the compressor, then remixes with the input signal, so you can do parallel subband processing.

Dave.

[noidea]

Thanks again!

[DaveGamble]

SERIOUSLY, once you start sub-band processing, you will NEVER go back to multiband.

Just ask anyone who bought the algorithmix lp-splitcomp, or the weiss DS!

Dave.

[djscorb]

Hey Dave

Was cool to read another of your explanations about AR coupling etc as they were the parts of the compassion manual I had to re-read several times and keep notes of on a piece of paper for quick reference!

While I agree that it is great to just tweak and see whether the advanced parameters have a positive or negative impact on the material at hand, I, like many others enjoy being able to actually understand what's going on and how the advanced controls relate to various hardware units and/or circuits etc.

Interestingly I just read the compressor thread that you've been contributing loads of great info to this weekend and was also poring over the SSL schematic last night. Me and a buddy are currently building a heavily modified SSL 4000 G series clone at the moment. It has the API2500 thrust filters and extra side chain HPFs, stereo side chain (turbo mod) and I am also aiming to soften all the ratios and change the release times to make it more suitable for mastering.

I am very keen to model the compressor in Compassion so that I can approximate what releases I will likely want to end up with on the hardware for when we come to calibrate things. I have already researched into how we'll go about modifying those things and we've mounted sockets or VRs where values need tweakage. I would like to ask you a question about the way SSL calculate their release times. I know its the resistance multiplied by the capacitor values in farads but I believe it is then multiplied by another constant, perhaps square root of two or something from reading posts on groupdiy?). Can you shed any light on that for us? At the very least, I want to establish the timing values for the fast and short release in the auto section and experiment with changing them slightly:

91k resistor with 6U8 capacitor for short time = ? ms
750K resistor with U47 cap for longer release = ? ms
(or maybe I got these backwards?)

I also plan to get rid of the longest manual release time of 1.2secs and put in a faster one instead and shunt the attack times along a slot

Any insight into how to calculate the labelled release times from the resistor and cap values would be great. I realise that these timings aren't usually accurate or measured consistently from manufacturer to manufactureer, but I would like to understand how SSL arrive at their labelled times

We spoke at length once about some software you had knocked up for measuring compressor knee response and curves. Any chance you've developed that further because I am still well up for modelling other units for generating some more useful Compassion mods?

Cheers mate and sorry for the long post hehe!

Scorb

[DaveGamble]

Hey Dave

Hello

Was cool to read another of your explanations about AR coupling etc as they were the parts of the compassion manual I had to re-read several times and keep notes of on a piece of paper for quick reference!

While I agree that it is great to just tweak and see whether the advanced parameters have a positive or negative impact on the material at hand, I, like many others enjoy being able to actually understand what's going on and how the advanced controls relate to various hardware units and/or circuits etc.

I agree, but I have to be judicious about it. I think it's fascinating, and I love to talk about it, but it's kinda tangential to actual audio work. You can get a lot (all?) of your dynamics work done with compassion without knowing how to configure the "deep" parameters, and I have to avoid scaring people off. Eh, maybe I'm wrong about it all, but anyway, here we are talking about it, so that's good

Interestingly I just read the compressor thread that you've been contributing loads of great info to this weekend and was also poring over the SSL schematic last night. Me and a buddy are currently building a heavily modified SSL 4000 G series clone at the moment. It has the API2500 thrust filters and extra side chain HPFs, stereo side chain (turbo mod) and I am also aiming to soften all the ratios and change the release times to make it more suitable for mastering.

I'm going to assume you mean the bus comp.
So thrust and HPFs are going before the THAT chip that you assume is a logging amp, but turns out to be a VCA, and you're confused.
Stereo sidechain... I can think of 4 points where you could implement that...
- Pre THAT chip
- Run a wire between the FW rectifier outputs
- Tie the AR stage diodes
- At the thresholding stage
Each will give you a different vibe with how the channels link.
My strong advice is pre the THAT chip, so you don't find yourself trying to compensate for the change inside what is a very sophisticated feedback loop That's if you want one set of controls.
If you want different controls for L+R then do it at the thresholding stage.

I am very keen to model the compressor in Compassion so that I can approximate what releases I will likely want to end up with on the hardware for when we come to calibrate things. I have already researched into how we'll go about modifying those things and we've mounted sockets or VRs where values need tweakage. I would like to ask you a question about the way SSL calculate their release times. I know its the resistance multiplied by the capacitor values in farads but I believe it is then multiplied by another constant, perhaps square root of two or something from reading posts on groupdiy?). Can you shed any light on that for us?

Naw! It's just 2. British means "release is two time constants", or 86.4% release. (1-e^-2)

Then round to the nearest "nice" number for the labelling.

So longest time is 1.2M*0.47u*2 = 1.128

Seriously though, just bang a log pot in there! It's not 1990! That was 23 years ago.

At the very least, I want to establish the timing values for the fast and short release in the auto section and experiment with changing them slightly:

91k resistor with 6U8 capacitor for short time = ? ms
750K resistor with U47 cap for longer release = ? ms
(or maybe I got these backwards?)

Oh. Well... that's not exactly how that works.
I mean... the literal reading of that is 1.2376 and 0.705, but that's not exactly what you get.
Those two caps are talking to each other. See the release path for the .47u via the 750k... into the 6.8u?
So, the 6.8u charges as the .47u releases...

You will have success with tweaking the values (compassion AR coupled autorelease is the same "stacked capacitor" strategy), but it's not completely obvious what will happen, because the ratio of the two resistors is significant!!

I also plan to get rid of the longest manual release time of 1.2secs and put in a faster one instead and shunt the attack times along a slot

It'll work. resistor = time/(2*cap)

Any insight into how to calculate the labelled release times from the resistor and cap values would be great.

Just 2* time constant (=2*R*C). And when you silkscreen labels, you pick numbers that are within 20%, and that look nice.

I realise that these timings aren't usually accurate or measured consistently from manufacturer to manufactureer, but I would like to understand how SSL arrive at their labelled times

As above.

We spoke at length once about some software you had knocked up for measuring compressor knee response and curves. Any chance you've developed that further because I am still well up for modelling other units for generating some more useful Compassion mods?

Not further, but I can. Maybe we should flesh out a plan.

Cheers mate and sorry for the long post hehe!

Scorb

It's a pleasure

Dave.

[_dada_]

Ooo K !
To bring this back to a more 'grounded' level :

Are there still plans for the "Quick-Comp" ?
Any ETA ?

And OT:
EQuilibrium still does not show the basic mode ( FIR IIR ) on the 'front page'.
Will this be added, sometime ?

Jan

[DaveGamble]

Ooo K !
To bring this back to a more 'grounded' level :

Are there still plans for the "Quick-Comp" ?
Any ETA ?

I don't know. There's more thinking to be done.

Right now there are a few concurrent ideas:
1. Compassion engine, trimmed to remove non-compression controls, advanced controls mapped into subspaces, so you can adjust "character" in a more creative way, with new information hierarchy to make that work. That might be COMPact.
2. A plugin with specific circuit models, where you can pick a model and play. I'm concerned that this might be more interesting to me than to anyone else. It wouldn't exactly make for a focussed audio tool.
3. A Compassion Mod player, and a Mod exchange, so you can have a standardised set of controls/meta controls, and just load mods to play; then you collect Mods and off you go

I am extremely interested in people's responses to these ideas.

And OT:
EQuilibrium still does not show the basic mode ( FIR IIR ) on the 'front page'.
Will this be added, sometime ?

First request for it. Logged.

Dave.

[Funkybot's Evil Twin]

Dave, your approach sounds good. I'm guessing you're going for something similar to Equilibrium in terms of being able to load up preset models with the appropriate parameters.

My advice would be to focus on making it simple, with a fast workflow. I'd also be very careful about the knob scaling. I surprisingly never got on much with COMPassion. I think there's two reasons: 1) I don't know enough about how compressors work to get the most of it (i.e. too complex), and 2) I didn't get on with the knob ranges when it came to the attack/release settings.

I'd also recommend shipping with a lot of mod presets for existing comps hitting all the usual suspects (1176, LA2A, 160, 33609, Mu, etc.). Finally, I know you're not a strong believer in adding saturation stages (or at least, not with EQs), but it's something I'd implore you to consider.

[DaveGamble]

Dave, your approach sounds good. I'm guessing you're going for something similar to Equilibrium in terms of being able to load up preset models with the appropriate parameters.

Well, nothing is certain yet. Much pondering.

My advice would be to focus on making it simple, with a fast workflow. I'd also be very careful about the knob scaling. I surprisingly never got on much with COMPassion. I think there's two reasons: 1) I don't know enough about how compressors work to get the most of it (i.e. too complex), and 2) I didn't get on with the knob ranges when it came to the attack/release settings.

ABSOLUTELY agree.

I'd also recommend shipping with a lot of mod presets for existing comps hitting all the usual suspects (1176, LA2A, 160, 33609, Mu, etc.). Finally, I know you're not a strong believer in adding saturation stages (or at least, not with EQs), but it's something I'd implore you to consider.

I'm firmly against adding it when it's not really there to add.
But I AM a firm believer in adding saturation when it's actually part of the process.
I'm looking into something that saturates in a VERY interesting way at the moment.

Dave.