You're missing my point and changing yours.
Strategies for compression is a different story than trying to recreate a hw leveler behavior with a digital tool that lacks the features for that specific task.
If you just had said that you wanted a soft knee and a slow attack in your op rather than specifying you wanted to recreate hw compressors behaviors in MCompressor, then I would have answered differently.
My point was simply to underline that hw compressors, levelers, limiters are more in their caracteristics than just a respond time, and recreating their behaviors requires to take in all caracteristics that defines them and translate them in the digital world with the appropriate tool(s).
Compassion will get you closer in terms of behavior, but there would still be some aspects missing anyway. Those can be addressed but it also requires work, and other tools. So the quest is achievable, as long as the right path is chosen. .

 

Please man... read carefully my original post!!! I'm asking for graphs/curves to replicate and study! Not the hw compressor!
Please, read carefully!!!
Oh! And also the title of this thread. It's only about curves! (No i don't like my women curvy :p )

 

Last edited: Nov 11, 2018

so you have a manley compressor

 

Give McDSP manuals a read, the classic oldies (Analog Channel, CompressorBank and FilterBank) did the things you're interested in back in 1998 (Pro Tools was at version 4.3 which required a proprietary PCI card and didn't have MIDI/piano roll support yet, and VST was introduced by Steinberg 2 years earlier but didn't support MIDI so no instruments yet). The plugins are all about morphable emulations with control that's only possible in digital - Analog Channel does tape saturation with separate control over low end bump and high rolloff, FilterBank did Pultec-style curves with separate controls for peak, slope and dip, and CompressorBank went with the knee control which morphs between all kinds of compressors, all described in the manual. Later SPC2000 and MC2000 extended the idea with serial/parallel and multiband with separate "emulation" per band

 

The easiest to emulate is the GUI. And people trust more their eyes than their ears. That's why there's so many vintage emulations, and the faithfulness of these are highly questionable.

I tend to use digital compressors for mild compressing, and analogue for heavy compressing. Wonder why? Because digital compressors, no matter the price, just [still] don't sound as good.

 

Well, at least the GUI of these emulations (usually) doesn't give me eye cancer, unlike the McDsp plugins.

 

Short answer. Well kind of.
SPICE is a software to design digital, analog and mixed electronic circuits. A simulator.
Long answer.
I am really no expert in Spice Modeling, but as far as I know, Spice doesn't match amplitude or the phase response near Nyquist and most important SPICE is not designed to run in real time. So it is not as easy as transferring a schematic into spice and you get your vst.
Some developers, however, use methods that spice uses (i.e Modified Nodal Analysis with a root finding Algorithm) in their plugins.
Cytomic does this for a long time and I think DMG too. And there are probably a lot more that took a close look at the SPICE code . But the open source software SPICE is still a tool to simulate circuits with the goal in mind to actually build them. And it is not designed to primarily run the simulations in real time.

 

Last edited by a moderator: Nov 12, 2018

From KVRaudio by Cytomic dev. Andy

Spice is not designed to run in realtime, and it doesn't match amplitude and/or phase responses near nyquist either. Developers that say the have optimised Spice I doubt very much have gone and edited the Spice source code to make it run faster, otherwise they would be selling their own version of Spice and everyone would buy it since speed matters for everyone, including people designing circuits, and they would pay decent money for something that runs x10,000 faster! What those developers probably mean is they are using the same methods that Spice does, ie Modified Nodal Analysis (MNA) with a root finding algorithm - this is what Cytomic has done since launch around 9 years ago, so welcome to the party!

I use phrases like "uses the same algorithms that circuit simulators use but optimised to run much faster", and don't bother going into too much more detail as the details are quite technical and not suitable for general marketing. I do have a netlist based semi-automated solver, and in my latest plugin The Scream https://cytomic.com/scream you can import spice diode statements via a diodes.lib file, as well as edit the value of every parameter of every component in the circuit. The circuit cannot be changed, only the values of the components, but I have added extra components so you can change the pedal into any of a number of Tube Scream type pedals. This I think is one of the most detailed analog modelling plugin on the planet, and solves 14 stereo pn-junctions in realtime, including op-amp slew rate limiting and output voltage limiting, diode normal and recombination saturation and non-linear capacitances and more. There is a schematic of the circuit that is solved here, but this doesn't include the full boyle op-amp internals but you can edit all their parameters in the plugin: https://cytomic.com/files/TheScream-Schematic.pdf

Getting a fully automated, stable, low cpu generic circuit simulator running as a plugin requires a lot of work. The trouble is that not even Spice is stable all the time, it is very easy to get it screw up and not give a reasonable solution, and it doesn't have any fixed sample rate or low cpu constraints!

 

The thing is, when doing it properly (from schematics, SPICE, derivation and then simplification for real-time performance), the sound depends only on the mathematical models of parts you used, which are *really* good these days. The biggest tradeoff is faithfulness vs CPU usage, really. Also keep in mind different units have different performance, so your idea of a 1176 or LA-2A might be completely skewed because of that one version you're used to. Remember, this stuff is 50 years old. It's often barely functioning and rarely serviced, so huge variations do exist.

It does the job, I guess. It's hard to present a multi-emulation with a GUI that suggest a particular one, when the biggest feature is morphing between them all. I'd rather have the developer focus on the sound than the looks.

Straight up copying schematics will give you models that take 20 minutes to process a second of audio. Also lots of parts are now obsolete, not documented and you don't really have any measurements of how they behave unless you get an actual unit or come up with some equation that looks reasonably close. Also, SPICE is a command-line program so you could only have a drum loop sample, run the simulation, listen to the processed version, tweak some stuff and run it all again. That's why it needs simplification.

I'm actually learning analog emulation and have to say, Cytomics are one of the cool ones. Native Instruments people also contributed a lot by adding precious resources and books on how to do stuff. iZotope does this as well. Really, the whole field is still largely people with scientific backgrounds who talk about their research on audio conferences. At least the people who are in charge of the sound, the senior DSP developers, not the GUI programmers.