UPDATE: most of what I've written in this post is confused. Check out my later post down in the thread... I think I've figured this out better now....
I think Veet's diagram is correct, except possibly the knee knob direction.
What's really weird is the attack and decay settings (127 is fast) are the opposite of what was quoted by the Elektron email that Tarekith just posted.
But after listening to what the comp is doing, and (to check my head) running test signals through the compressor and back into my DAW, I agree that 127 seems like 'fast' action.
EG: To get a hard pumping sound, play a pattern with a hard 4/4 kick and plenty of long-decay stuff playing during the non-downbeats. Set THRSH low-ish and RATIO above 30 (which I agree sounds like about the max ratio strangely). If you set attack to 127 and tune the release somewhere between 0 and 64, you get a really strong "pump" as the compressor releases slowly. If you want more snap from the kick, dial attack back from 127 to a slower attack mode.
What can get confusing about this, especially when trying to get analytical using pure sinewaves to test, looking at the waveform etc, is that a) this compressor seems to be very much a peak signal detector, so any transient spikes will cause it to clamp down. b) the GND-sin machines don't phase-init every time you trig, and you often get very loud spikes during the first cycle of a note. That's enough to signal the compressor to clamp down very hard -- but depending on your threshold that might only last through that initial transient spike. So a very high ATK setting can allow the compressor to *completely* gain reduce during that initial transient, and then release during the rest of the signal. If you have ATCK high and RELEASE low, then your note will have an attenuated start and then swell up as the compressor releases -- which can make the ATTACK knob seem like a release since you have to have a really high attack to get the gain reduction during the initial spike to make the slow release sound audible. Hope that makes sense to someone.
I'll post some images when I get some more time.
But doing listening tests on program material, it really does sound like high attack = fast and high release = fast.
Ironically, I took a closer look at the Dynamix on the MNM, and guess what -- it's completely different. Attack of 0 is a fast attack, release of 0 is a fast release, the ratio knob actually does something meaningful above 30, and there's a peak-rms dial for the detector function. That said, I couldn't get the release to do anything useful on the MNM compressor yet!
-j
PS: just mailed Elektron about this, and I'll post if I hear anything back.