Thanks Bitflipper for the infos.
So I downloaded the demo versions of the Analyzer, Modercompressor and MspectralDynamics, testing them on a whole mix and to confirm the feeling of Dstrenz, the MSdynamics seems to sound very good : The quick general impression is that it sustains and "expands" the signal smoothly.Comparatively, I tried DSM, and cannot get this smoothness and still stay with this agressivity ! I also tried the Slate compressor (FGx and Bus compressor) and, no offense to the Slate stuff, but it sounded very tense and rather disgusting !
By cons (concerning the MspectralDynamics) it is true that despite the explanatory menus, it is not easy to choose a particular criterion and the many different options could be more confusing than anything for the novice user I am.

Hi Peter, I've been using DSM for quite some time and it is possible to get it to sound smoother by putting a good parametric EQ in the effect chain before DSM and adjust it so that the source material more closely matches the frequency curve. That way, DSM doesn't need to do as much work. After that's done, play with the ratio and the wet/dry controls. This works with MSpectralDynamics too.
 
For anyone who wants a quick basic example how these can be used, load a wav file and insert MSD. Find the part of the song that sounds best, and press and hold the capture button while that part plays. Now play the song. MSD shows the frequency curve of the part you captured and the real-time frequency graph of the wav file. While it's playing, play with the Threshold fader. You'll see the Threshold control moves the captured frequency curve up and down. Wherever a part of the frequency curve is below the real-time source, those source frequencies are compressed to match the frequency curve. Wherever a part of the frequency curve is above the real-time source, the source frequencies are untouched. If the wav file happens to be a vocal track, this procedure nicely de-essed it.
 
I also use DSM to help bring some life back to old recordings that works better than anything else I've tried. If interested, just ask and I'll show how I do that. The process should work equally well with MSpectralDynamics.

Bit - I have MDynamics in their Mixing bundle.
 
How does MSpectralDynamics differ from that one?

Thank you very much dstrenz for all those precisions ! I was certainly a little confused by the controls of DSM and their interactions, But in fact, I just used the presets and not captured a reference track. And yes, if you could show how you're doing, it could be very informative.

peter434
Thank you very much dstrenz for all those precisions ! I was certainly a little confused by the controls of DSM and their interactions, But in fact, I just used the presets and not captured a reference track. And yes, if you could show how you're doing, it could be very informative.

Sorry, I missed your reply until now. Here's a basic rundown of what I do.
 
Say you've got a digital copy of an old cassette of a garage band and your (possibly unrealistic) goal is to make it sound like Allman Brothers Live at the Filmore. Note that if there is something drastically wrong with the recording, like a super loud guitar, there isn't much you can do to improve it.
 
Capture a curve from the Allman Brothers album and save it. Load up the old recording onto a track and copy that track and mute it. It is all too easy to make changes that make it sound worse than the original, so the unmodified copy is for reference and should be compared with the modified track often.
 
Let's see if the curve we're using will improve the sound. Insert MSD and load the Allman frequency curve. As you play the song, you'll probably see the  source material does not come close to matching the curve. Adjust the threshold so that the entire curve is below the source material, which will cause MSD to compress practically every band all the time to match the curve. The result will probably sound unnatural at 100% wet. Set the wet/dry fader to dry and slowly raise it to mix in some of the compressed signal until it sounds best. You're probably somewhere around 10-20% wet. If there's no improvement, use a different curve until it does.
 
Now put a good parametric eq in front of MSD and use it to make it so the source material more closely matches the curve. Set the MSD threshold so that the curve is, on average, slightly above the source material. Now use the wet/dry control's to smooth things out. It will probably end up somewhere between 50-100% wet now.