overall 'best' mastering compressor?

[JumpingJackFlash]

Well, we've reached 100 votes!

PSP MasterComp and Voxengo Soniformer are clearly emerging as the favourites, with the likes of Voxengo Maquis and Kjaerhus GCO-1 following just behind.

Will this remain the case? - Please keep your votes coming in!

[justjazz]

...

[Aleksey Vaneev]

justjazz, I would say that Marquis Compressor should be suitable for both mixing and mastering - it has enough mode switches to be usable in both scenarios.

[JumpingJackFlash]

justjazz; it should be noted however, that Aleksey Vaneev is biassed, because he makes the Voxengo products!
I don't want to put you off; while a non-multiband compressor, if used well, may be perfectly adequate to mastering, it is generally believed that a multiband compressor, if used well, may be better suited for mastering most styles of music.

[Kingston]

generally believed

indeed. *believed*

multiband compression is, how should I put it, so nineties. The unfortunate Finalizer legacy still shows.

(but if absolutely necessary, which is rarely the case, soniformer is pretty much the ultimate multiband compressor.)

[bmanic]

Hmm.. While I do agree, partly, with Kingston I still think that for punchy, totally controlled bass and that modern pop sound there is little choise but to use some kind of multiband processing unless the mix is very good to begin with. I think that bass can be gotten under control with certain high end outboard compressors but so far I have found no software, single-band compressor that can do it adequately without pumping the whole mix in a non-musical fashion. I wish I'd get some more experience with the weiss compressor as it seems to be one of the few tools that can control the levels of bass without any crossovers at all. It simply is so "transparent" that it can be used in RMS detection mode where bass usually becomes the highest average energy level and hence hits the compressor first. At least this is how I think it could work. Kilroy, any ideas?

I usually end up using Buzzroom multiband compressor to separate frequencies below 100hz and compress them separately. This also has the benefit of putting the crossover right at the rather nasty 80-90Hz area where bad mixes tend to have too much boomyness.

However, multiband compression is bad simply because the crossover filters always colour the signal in some way, usually in a bad way. Voxengo Soniformer is rare in that it does colour the signal but very little.

So, what I'd like developers to focus on when creating a multiband device, targeted at mastering, is to very carefully code the crossover filters. Of course the compression action itself should be good but this can be more "utilitarian" than "wannabe vintage".

Cheers!
bManic

[Aleksey Vaneev]

bmanic, for your information, if multi-band compressor uses 6 dB/oct crossovers - it almost adds no audible coloration (its group delay function is exponential, without any fluctuations on it). 12 dB/oct crossover is probably a possible maximum (which Soniformer uses as well). Above 12 dB/oct you may hear sound of crossover (it makes the group delay plot look like a hair comb). Linear-phase crossovers solve this problem, but they can potentially create a problem of pre-ringing in some bands (expecially if gain reduction achieved is high) - then again, this problem can be partially resolved by using massive number of bands (like 32 in case of Soniformer).

[Kingston]

thanks Aleksey, that is useful information.

I've always been adverse to multiband compressors due to the all too common steep (more than 12db/octave according to you) crossover slopes creating audible phase distortion/smear.

[bmanic]

Thanks for the information Aleksey. Btw, a useful plugin would be a multi-output, cross-over filters only, plugin that would provide the user with very high quality cross-over filters to be used as desired in plugin/host hybrids like energy XT. This would allow us to create our own multiband-what-ever plugin chains. I've been planning to build such a plugin myself once Synth Maker matures a bit more and I get my math skills up to spec.

For instance, multiband-phaser/flanger/chorus can be a really interesting thing, not to mention multiband delays and reverbs!

Cheers!
bManic

[bleebsen]

Can I vote for UAD-1 Fairchild?

I tend to use multiband-dynamics only to fix broken mixes in mastering. A proper mix can do without IMHO.

[Aleksey Vaneev]

bmanic, unfortunately, I can't add such plug-in into my workflow: its potential userbase is too small as far as I'm concerned. Why not ask energyXT's authors to build such module into energyXT itself? Such module can be created using stock plug-ins as well.

[Kingston]

also, it's dead easy to build multichannel crossover templates in cubase/nuendo as well.

You can save multiple bus/channel settings at once (select the channels at once and right click 'save as' channel/bus setting), and recall them at later time in any other cubase/nuendo project.



also gives you the oppportunity to use your choise of filters as well. (PLparEQ would be pretty optimal as a mastering crossover)

[bmanic]

So true! I completely forgot about PLpar4 EQ. Will have to try that one and see if I can come up with good cross-over points.

Cheers!
bManic

[kilroy]

Hmm.. While I do agree, partly, with Kingston I still think that for punchy, totally controlled bass and that modern pop sound there is little choise but to use some kind of multiband processing unless the mix is very good to begin with. I think that bass can be gotten under control with certain high end outboard compressors but so far I have found no software, single-band compressor that can do it adequately without pumping the whole mix in a non-musical fashion. I wish I'd get some more experience with the weiss compressor as it seems to be one of the few tools that can control the levels of bass without any crossovers at all. It simply is so "transparent" that it can be used in RMS detection mode where bass usually becomes the highest average energy level and hence hits the compressor first. At least this is how I think it could work. Kilroy, any ideas?

First, concerning the Weiss...a very, very good box as you have said. You know, I would much rather use/own the DS1 than the Manley Vari-mu if I had to choose...there is just so much more you can do with the DS1 to shape a mix. By the by, the DS1 can indeed be configured for precise band select, or high/low split band operation. It utilizes phase linear crossover topologies, so if you are in band select mode you have two active PL crossovers dividing the total frequency network. Bmanic, if you can do so, try to get some more time on this box. Even if you have to pay something for that privilege it will be very much worth it.

However, multiband compression is bad simply because the crossover filters always colour the signal in some way, usually in a bad way.

Concerning crossover networks in general...I think a gentle 6db per octave crossover is the most practical to model digitally, providing the number of bands are few. In the case of Aleksey's unique multiband processor, the use of steeper 12db per octave crossovers makes sense, because the number of bands in close proximity to each other would result in too much dynamics processing information being "shared" between them if the more gradual slope were used. Personally, I think gentle slopes sound the best, but as the number of discreet bands increases the probabilty that there may be insufficient isolation between them in an active multiband dynamics environment can create conditions that effectively cancel this advantage. This will be most obvious if adjacent bands are operating at widely disimilar gain reduction behaviors, resulting in excess gain reduction "chafing" at the crossover points. A gentle 6db per octave slope would allow a greater amount of gain reduction behavior intended for one particular band to be "crossed over" into any other adjacent bands. The implimentation of steeper crossover slopes in order to improve frequency band isolation would seem a logical enough counter but you then must weigh the fact that you now lose some transparency of the crossover itself. My question, I suppose, is whether better frequency band isolation is a worthwhile tradeoff at the cost of greater phase error. Also, there will still be some potential for significant gain reduction "chafing" even with steeper crossover slopes, only now you are more or less dynamically "modulating" any phase errors that the steeper slope introduces of itself. It's kind off a mad go round, and the real loser in the end is the material being processed.

The "best" approach to the multiband "problem" would seem to be using the smallest number of bands necessary, evenly spaced as far as possible from each other across the full frequency band, using gentle 6db per octave slopes for the crossovers. An alternative would be the choice Aleksey has made...a greater number of bands using slightly steeper slopes. As I see it, the primary merit of this particular topology lies in the ability to widely distribute any gain reduction duties so as to avoid any locally significant crossover chafing...a kind of "many bands makes light work", wherein a greater number of smaller reductions are used to accomplish the same end as what a smaller number of bands would using more extreme reductions. Given these two approaches, the designer must weigh the sonic advantages/disadvantages and decide where he/she wishes to place their efforts. When you introduce crossovers of any design into a variable dynamics environment you are going to get some colouring. One just has to simply decide for oneself whether that colouring has come at the right price...or not.

Sorry chaps, I know...long bleeding post.

[Kingston]

If you would've been there with that post long a go when there was one particular FinalMix related thread that escalated to fullscale war, I recon you would've saved many lives (well, faces ).



"Multiband compressor crossover phase error tradeoffs." I can't imagine anyone comprehensively explaining that in a shorter post.