Multiband compressor

From Fractal Audio Wiki
Revision as of 00:55, 24 March 2009 by Javajunkie (talk | contribs)
Jump to navigation Jump to search

[Ultra]


Introduction

The Axe-Fx UltraTM contains a three-band compressor that is great for mastering or compression of a whole mix. It also works great after distortion to help thicken the tone. The basic principle of the Multi-Band compressor is that the input is divided into three bands using a crossover. Compression is applied to the individual bands and then the outputs recombined. This allows isolating bands of the input material and applying different amounts of compression. Multi-Band compression is the de-facto mastering tool and can really improve the final mix.


Parameters

  • FREQ1 - Sets the crossover frequency between band 1 and band 2.
  • FREQ2 - Sets the crossover frequency between band 2 and band 3. Each compressor section has it’s own menu page with the following paramters:
  • THRSH - Threshold control. The threshold at which output compression starts to occur.
  • RATIO - Sets the gain reduction ratio. This is the ratio of input power to output power beyond the threshold. For example, if the threshold is -40 dB, the input signal is -30 dB and the ratio is 2.0 the output will be -35 dB.
  • ATT - Attack rate. Sets how fast the compressor reduces its gain once the threshold is exceeded. Higher values equal faster response.
  • REL - Release rate. Sets how fast the compressor restores the gain once the signal has fallen below the threshold.
  • LEVEL - Sets the output level of the band.
  • DET - Selects the type of detector to use, RMS or Peak.
  • MUTE - Mutes the output of the band. You can use the mute controls to audition one or more bands.

Multiband Compressor Tutorial

Tutorial

You need to have a purpose in mind before you use MBC and it should be something that a broadband compressor won’t accomplish. It’s really not that hard to learn if you go about it in an orderly manner. The method that follows is just one of many ways to dial it in. At it’s finest; MBC is an art form - in the sense that there is no ‘correct’. Depending how you use it, or abuse it - it can change everything. Be prepared to get it all wrong because that’s how you get it right. When it all clicks together in your head you will understand why specific parameter settings that achieved a desired result on program material A, will never translate properly to program material B, as it is simply too fine-tuned to specific sources to do this. This applies to all the various rigs out there as well.


What to use it on

A good, simple example of why MBC would be more appropriate than broadband compression is where everything is just so, except for this chug thing that sounds great but sticks out too much. Let’s say an attempt to pinpoint the chug and surgically remove local energy away via parametric EQ just didn’t work. Perhaps the tone became brittle - whatever. Using MBC, you can apply a dynamic solution aimed right at the frequency region where the chug resides - something that broadband compression can’t do. Another case for MBC would be where your highs are too dynamic to fit into a mix, and while broadband compression glued the highs in, the mids disappeared in the process. This is also a simple fix - just compress the highs. A savvy user could also use it to do something like dig deep into a sweet tonal artifact that is masked by much more dynamic surroundings. If you were to compress these surrounding areas leaving the sweet spot alone, and bring the sweet thing forward by raising the output level of the band it resides in - presto. MBC can alter the dynamics in both directions too. Anyone can reduce dynamic range, but expert use allows an increase of perceived dynamic range that was not there to begin with. Beginners will first have to contend with locating the frequency range(s) that need altering. One method is to sweep parametric EQ while listening for the subject area to either poke out or disappear; it’s a whatever-works approach. Another trick that works (if a mix is not to dense) is to let the tuner tell you where in the frequency spectrum you are, and then refer to a note vs. frequency chart. If you’re trying to do this playing live (instead of Memorex) you can still do something like the chug thing or wail on that piercing note and refer to the tuner for guidance.

One thing for sure – MBC is a re-ampers dream come true. Getting sophisticated with MBC while playing guitar will hurt your head, but fixing something like the chug problem shouldn’t take too long to do after a little practice.


Set it up

Without a doubt, the best way to learn MBC would be to unplug the guitar and route a decent fidelity CD (or something) though the Axe as if you were going to do some mastering. Just feed the Axe’s rear input(s) the source and feed the outputs to your stereo system/whatever. It might be a good idea to grab the manual, and glance over of the MBC page first.


The Parameters

FREQ1 and FREQ2 draw lines in the aural sandbox. These two lines define where the single, broadband frequency spectrum is divided up, creating three smaller bands. On the Axe, Compressor 1 affects the lowest frequency band. Compressor 2 affects the mid-band, and Compressor 3 affects the high-band.

Compressor 1 (C1) always operates between a fixed position of 20 Hz, to whatever frequency you set FREQ1. Compressor 3 (C3) always operates from a fixed position of 20,000 Hz; to whatever frequency you set FREQ2. Compressor 2 (C2) operates within the resultant frequency range having been determined by FREQ1 and FREQ2 settings.


Operation

Start with C1, as it is a good frequency region to learn with. To prepare, set the FREQ1 crossover to around 500 Hz and FREQ2 crossover to around 3000 Hz. Set the Threshold, Attack, and Release fully clockwise and set Ratio fully counter-clockwise. Then set Level to 0 db, Detector to Peak and Mute to off. Do C2 and C3 in the same manner. Now feed the Axe some music, a clip, or something with decent dynamics (that might be hard to find these days) and set the input levels normally. All should be well.


Threshold and Ratio

Reduce the ratio into the 2.0 region, and then slowly reduce the threshold level until you see the meter bar drop about 20% or so. The threshold simply tells the compressor where to start doing its thing relative to input signal level. You are now altering the dynamics of a frequency range between 20 Hz and 500 Hz and the meter is indicating the resultant output gain changes. Twist the threshold down to –80 and see/hear the whole bottom-end go away. Bring the threshold back to where the meter is around 20% and then play around with the ratio. Higher ratios will ‘clamp’ on more than lower ratios, and threshold determines the point that the clamping starts. Play around with various combinations of the threshold and ratio, and you will find that these two controls have some serious interaction going on. Now find a relationship between the two that creates some aggressive gain reduction. Something like -40 on the threshold, and 2.0 on the ratio should be close if the Axe’s input level is driving the input meter properly.


Attack and Release

Next thing is to decrease the attack time by adjusting counter-clockwise towards 0. You will hear some life returning because the leading edges of dynamic transients are sneaking past the compressor. This occurred because you just told it to hold off for a split second after the input signal crossed above the threshold. Leave the attack time somewhere you like and now vary the release time - which adjusts the ‘let-go’ part of the equation. It releases in a slower manner as you adjust counter-clockwise. Since the compressor has a hold of the gain, you need a way to adjust just how quickly the gain is returned to normal when the input signal drops back down below the threshold. Go back and forth and see/hear how they interact. With the right material, you can really get some dynamic pumping going on, which can be good or bad. Find a relationship you like and leave it there. Since all the parameters are interactive - and you just changed two of them, go back and fine-tune the threshold vs. ratio to see if further progress can be made. It is not uncommon to do this whole procedure many times over, and it’s likely that smaller and smaller adjustments will be made as you go. Don’t forget that you can change the lines in the sand too. Effort equals results here for sure.


Detector

The best DET type to use is program dependent. Basically, you are telling the compressor what to respond to - sensing transient peaks (Peak) or more of an averaged signal (RMS). Compared to the other variables, it has little overall affect; so don’t sweat details now.


Level

Perhaps the best way to think of this is to realize you can independently vary the volume of each dynamically doctored band relative to the others. A stereo mix could have six variables to rebalance. If you are hitting all of them you will definitely have some work to do. Balancing these bands containing doctored innards is pure artwork, and there is a lot of bad art out there. Crank them up and down relative to one another. Dynamics aside, you will be surprised with newfound artistic freedom.


A real world example

My buddy just joined a jazz rock band that never had a guitar player, and he wants to add his stuff to a demo CD the band uses for promo. After dubbing the guitar over 10 songs or so, I start mixing. Fortunately, the CD was not mastered so it had plenty of dynamics remaining to squeeze the guitar tracks into. Problems exist in the CD however. Overall, the drums are too loud – especially the kick, and the bass guitar is somewhat masked by kick boom. It’s a really good recording otherwise so it’s worth spending some time on. I set up a stereo layout in the Axe like this: PEQ – PEQ – MBC – GEQ. Starting with a tune done in G, it was roughed it in using GEQ. Using the dual-stereo PEQ chain, I located the relevant frequencies of the kick (65 and 130 Hz) and bass guitar (49 and 98 Hz) in each channel. After some poking around, boosting and carving away limited success was found. Very little energy could be cut from the kick before it went into tonal hell. The bass was more tolerant and hung in tonally so it was treated with a healthy boost, raising it in level without adversely affecting the kick. With the GEQ finally removed from the problem area, and observing how the dual-chained PEQ was administered to best affect, I switched from peaking to shelving and freed up two PEQ blocks - by functionally duplicating (via shelving) what had taken two chained PEQ blocks to define. In the end, a small cut was applied to the kick frequencies, while the boosting nearby bass. The definition between the two was certainly better, but the kick and bass combination was still too loud.


MBC to the rescue

After about an hour of work the overall volume balance is much better. Another hour and it will be there. Right now, the left channel FREQ1 crossover point is 165.4 Hz, and FREQ2 crossover is at 3220 Hz. C1 threshold is -80, with a 1.10 ratio. Both attack and release is at 10.00. Level is .9 db and the detector is in peak mode. I haven’t started on C2 and C3 yet - for either channel. While the level of C1 is up .9 db, the perceived volume of is much lower than before - primarily because of the 1.10 ratio that gets applied at a very low threshold.

The right channel FREQ1 crossover point is at 50 HZ and FREQ2 crossover is 2995 Hz. Like the left channel, threshold is also –80, but the ratio is 1.91, and that’s some squash. Both attack and release are at 10.00, just like the left channel. Level is full on at 20.00, which is what it takes to compensate for all the compression applied. If you read into the numbers, you will see why this is working. The left channel has moderate compression applied to a region that both the kick and bass occupy, and definition that is commonly lost after compression was retained because of the shelving PEQ applied earlier. The right channel is hitting the kick hard - but only up to 50 Hz, so only the fundamental boom is reduced, and the bass guitar sneaks right on by. A cool spatial relationship between the bass and kick is happening too. The original mix had the kick and bass almost on top of one another - panned center and somewhat boring. Now the bass is moving around in space a bit and the kick is tight, with body leaning left and snap on the right. The whole mix opened up significantly and there is much more room available to stuff in guitar tracks. Note that I’m still messing only with C1 parameters. Even at this stage of the game there is no way that a setup like this would ever translate to anything else.


Sanity retention

Resist the temptation to use it for more than a fix until you are fluent at it. Stop for a head break before you lose track of what you’re doing because one wrong move is all it takes to bring on the clean sheet. Also be careful of using EQ or any gain altering device prior to MBC because any change to input level will alter your perfectly set output dynamics, upsetting the whole shebang. MBC is usually last in the chain, and because it does not play nice, frequent adjustments to at least the threshold will be required when those input levels change. Also watch for ear fatigue that quickly takes its toll on proper tonal judgments. By the time you have finished three bands of this stuff your tired ears will certainly mislead you.


Axechange Presets having MBC

Dialing in an axechange preset for your personal use can be difficult to start with, and doing this to one that incorporates MCB can really be tough because it is such a user, rig- dependent thing. Hopefully you now understand why wholesale bypassing of MBC will probably not work well either. Try to gradually lift MBC from a source by just raising the threshold, rebalancing output levels and such to suit your rig. If you subjected yourself to the earlier exercise you will know how to pull it off because you will be able to look at the existing parameters, figure out what is happening and make changes to suit.