Sorry to jump in so late! Am I still welcome? I've done a fair bit of research on dynamics...
"Are we close to getting plugin compressors that sound like real ones?"
Yes and no.
First the no:
Analogue compressors don't sound like analogue compressors. When we were working on the Liquid stuff, we discovered very fast that no two analogue units are calibrated exactly the same. Grab two unmatched 1176s off a shelf, pick the same settings, and you'll get a slightly different sound. This said, you /can/ calibrate them to match, although it's fiddly, and not everyone has the time to do it. Also, a big part of the sound of the units is their noise profile - a lot of (though certainly not all) software compressors abstain from adding in the noise. As it stands, for a lot of the classic units, the noise that they add (which comes right up in volume when you add gain, and swamps over the signal when there's enough GR) is a massive part of their character. Also, distortion. It's not strictly compression, and it tends to come after the fact. In designing Compassion it became quickly apparent that it wasn't enough to just add noise to the output, but adding noise (at an independent level) to the sidechain circuit was important too.
Now the yes:
There's only a finite number of topologies that will allow you to implement a dynamics circuit. You're attempting to achieve gain reduction which is some function of the input level, and it needs to be smoothed. Again, in designing Compassion, I wanted to cover all bases, so I built a generalised model of all the possible topologies, which was perhaps a little overkill. Does Compassion sound identical to unit X? It's a function of how much time you're prepared to spend tweaking it. It's not for nothing that there are >100 parameters on it, as fiddly as that can be, but all the factors I've ever seen in a decade of modelling compressor circuits are in there, and can be calibrated. In a sense, it's the same game as trying to calibrate two mono analogue compressors to match up - in the factory, someone tuned some potentiometers to approximately the right place. Roughly. Since then, a few of the capacitors have gone past their best, and it's all a little wonky now. But the framework is in place.
Reading KVR or GS, you'll see people gushing in droves about plugin X or Y, which for them has replaced some piece of hardware. And it comes down to this - they're making a statement that YES, these plugin compressors are doing the right thing, and have replaced the hardware. So the question is whether these people form a majority.
Our understanding of dynamics circuitry is far from poor - the existence of something like Compassion is (I hope) testament to the depth of our knowledge of dynamics circuits.
Sorry for the wall of text, but I'm happy to join in the conversation here if there's anything I can elaborate on!
Dave.
Ninja edit: And in answer to the question of whether there are software compressors that pass ABX tests against their hardware equivalents, yes. Ten years ago, when I was working on the Forte suite, the compressor was indistinguishable from the ISA130 hardware compressor, for trained listeners in a good monitoring environment. And things have only improved since.
Ninja edit 2: I /believe/ they haven't modified my DSP from Forte for the Midnight suite, so you can grab those, and ABX against the 130. Please note: there are calibration differences between the ISA130 and the Red3, and topology differences between the 130 and the later ISA modules (e.g. the ISA430, which moves the attack circuitry to make the circuit sound smoother).
