Adjusting Effect Levels for Mix/Bus Compression

I spent a few hours yesterday doing final bus compression for the track I’m currently working on. Approaches to and techniques for bus compression were one of the things I learnt most about during 2016, and yesterday I had a kind-of ‘lightbulb’ moment, which will hopefully lead to better results in this area going forward.

I’m a ‘reluctant participant’ in the whole competitive levels/loudness wars thing. Fundamentally I like the groove, emotion, impact, etc which a decent dynamic range can impart on a track. But at the same time I understand the need to achieve an overall loudness level that’s similar to other tracks in the same genre (especially because not doing so simply makes your music difficult for DJs to mix).

In the past, I’d always equated greater amounts of bus compression to a loss in clarity. To some extent this is true, as compression will narrow the dynamic range of the sound and hence simply reduce the ‘depth’ of volume variation available. However I’d always found that compressing the entire mix necessitated a compromise of getting closer to competitive levels while sacrificing some detail and clarity.

About halfway through last year I had a mini breakthrough of sorts, when I realised certain settings on bus compressor plugins can have a big effect on the quality of the resulting audio. Specifically I usually use Cytomic’s ‘The Glue’ as the first stage in the bus compression chain, and I found that simply setting the oversampling rate to the recommended or higher levels (4x or more when auditioning) gave far clearer audio quality than the default lower settings.

For my current track I had spent a bit longer than usual honing the reverb plugin settings, and fine tuning the reverb send levels. After this I was really happy with the result… it had a nice balance of having a good depth/space with sounding too ‘washed out’, and seemed to translate well to several different sets of speakers and headphones. But yesterday it was a bit disappointing to have some of this clarity and balance lost when I started pushing the final mix through bus compression. When I listened closely it wasn’t so much a by-product of compression, but more that the levels of the reverbs and delay effects were stronger. When I thought about it, the reasoning was obvious… I’d squashed down the top 3-6 dB of the volume range, so obviously sounds down at -15 to -20dB (like the reverb layer) had been effectively pushed up by a similar amount.

I usually do final bus compression in a separate Reaper project to the mixing, using just the final stereo mixdown as a source track (my aging PC can’t handle multiple reverb plugins and CPU hungry bus compression at the same time). So I went back to the mix project and rendered another version of the stereo mix with reverbs and main delays turned down around 1.5dB. Running this new version through the same compression chain resulted in a much clearer mix… it sounded a lot more like the former original stereo mixdown… just louder (which is exactly what I was trying to achieve).

Anyway, in hindsight I’m a bit surprised it’s taken me this long to figure out this technique (the basic point of compression after all is to reduce dynamic range), but I’m going to experiment a bit more, and hopefully end up with a lot cleaner, clearer final mix than for past tracks.

Another way to potentially prevent the issue could be to ‘mix into’ a compressor or limiter during writing/sequencing/mixing. This is a bit unorthodox technique historically, but seems to have gained popularity in the last few years (I seem to have read a lot of articles recently where people discuss working this way). The idea is to put a limiter/compressor on the master bus right from the early stages of writing (using generic/default settings close to what you’d usually use for final bus compression). This way you’re always evaluating level balance with compression already ‘baked in’. I don’t usually use this technique because for some reason I like to keep a clear separation between the mixing and final ‘mastering’ stages… but based on yesterday’s experience I can definitely see the merits, so may try it in a future track.

When the Problem is Staring You in the Face

I had an interesting experience over the last couple of weeks, with a mixing problem that should have have been obvious and easy to fix, but because I was too focused on details, I missed the bigger picture and let the problem persist for way longer than it should have.

I’m still in the finishing off stage of a track which has ended up becoming the most drawn out and time consuming piece I’ve worked on so far. I just looked back to previous posts and realised I said I was on the ‘home straight’ with it more than 2 months ago.

Part of the reason this track took longer than others was that it was the first where I’d used an acoustic instrument for one of the main themes… an acoustic piano riff (from NI’s ‘New York Grand’). As with acoustic percussion samples I’ve discussed in a previous post, any recorded acoustic instrument is inherently going to have a much greater dynamic range than synthetic sound. And to fit this into the generally very narrow dynamic of club music, considerable but careful application of compression is required.

The piano riff I came up with, I thought, had a nice dynamic… getting thicker in texture and a bit louder/stronger towards the end of the riff… I felt this gave it a bit greater feeling of tension. Although a fair amount of compression would be required to make the riff fit well in the mix, I was keen to try and preserve as much of that dynamic as possible. Hence when mixing I was too focused on trying to preserve dynamic of the riff that I’d liked in the soloed part. This unfortunately led me to being too cautious in applying compression, and ended up pushing the piano part way too high in the mix (in order to get it to stand out properly). Added to this was the mistake of not following my own advice and regularly checking back against reference tracks, so when I finally did do a side-by-side comparison with my usual reference material I’d created a kind of ‘inverted smile’ in terms of frequency spread… with piano and mid-range way too dominant, and not nearly enough bassline nor cymbals.

Once I figured out my mistake, it was pretty easily corrected with a simple application of Waves’ Renaissance Axx compressor (after having spent at least a week going in the wrong direction)… sure I had to sacrifice some of the nice dynamic I had originally wanted to highlight, but looking back, I think that original desire was misguided. The track I’m writing is in a minimal-techno style… where narrow dynamic and very loud overall track levels are commonplace… the expectation to keep a main acoustic instrument part fairly dynamic, and achieve a competitive level in the overall track was a bit unrealistic.

So 3 important lessons I learned for going forward…

  1. Audition parts in the context of a mix. Things that sound good on a soloed part may no longer sound so good, or even be completely lost in the context of a whole mix. I was too swayed by trying to work towards a soloed piano sound which I thought sounded good… it would have been better to have always auditioned it in the context of the mix right from the start.
  2. Be realistic about how much dynamic range you can achieve in styles which are innately highly compressed.
  3. Listen to and compare to your reference tracks regularly!

Spending Time Appropriately

I’ve blogged before several times about sample manipulation and clean up (i.e. EQ, compression, gating, etc…).  I use a lot of live-sampled sounds in my tracks, often as main/key elements in the overall composition.  So in these cases, properly cleaning up and compressing samples is really important, if I want these elements to stand out well and sound clean in the mix.

For the track I’m working on at the moment, I’m in the middle of doing all the incidentals, and part of that is incidental percussion.  I usually first sequence all of these sounds in the track’s Reaper project, and afterwards go through and automate EQ and compression to get them sitting in the mix properly (usually with a single EQ and compressor instance and using automation to adjust the parameters for each incidental).  That process can be pretty tedious, so I decided this time I’d do all the sample EQ and compression up front (i.e. before sequencing) in audio editing software.  This also was pretty daunting at first, because I had about 20 samples to treat, and I usually spend a good 5-10 minutes per sample finding the right EQ and compression settings, and auditioning on monitors and close listening through headphones.  But as I started working through them, I realised I didn’t need to spend so long on each one…

For samples which are used as key elements in the track, overly careful EQ and compression is really important… your key elements will either comprise the main ‘themes’ of a track, or at a minimum occur many times during the track’s progression… hence you need to spend time making fine and careful adjustments to get them sounding as good as possible.  On the other hand, the incidental samples I was working on might play in total a couple of times at most during the entire progression… plus they often occur at sonically ‘busy’ parts of the track (builds, peak points, etc…), where slight quality issues (like a tiny bit of distortion, or slightly wrong EQ) will likely be masked by all the other sounds occurring at the same time.

It made me realise I could afford to be (what I would usually consider) a bit ‘sloppy’ with my approach to this sample editing… only audition on monitors, and sometimes running mild effects like a slight limiting or compression with minimal or no auditioning (I use Waves L1 and C1 a lot for this purpose, and have used them so much I can usually apply mild adjustments without needing to audibly auditioning).  And more generally it made me think about using your time appropriately.  Time is a precious commodity for a producer… particularly if it’s not your profession and you have limited time to start with.  So you need to really think about the areas where it’s necessary to spend time, and the areas where you can afford to take a more ‘quick fix’ approach.

Mix Issues – Prevention Rather than Cure

Thanks to the rapid development of DAWs and plug-ins over the last 5-10 years, as producers we have close to unlimited flexibility in terms of audio processing.  Even my very old (7+ years) music PC is capable of running 10’s to 100’s of simultaneous plugins in a track’s project.  Added to this, the internal digital routing in a DAW, and the ever-increasing quality of plugins, means chains of 10’s of plugins are not only a reality but often the norm in putting together a track.

But with this flexibility can come a complacence to ‘fix problems later’ with plugins, rather than dealing with them at the source.  I’ve read numerous interviews with pro producers who emphasise the importance of getting sound right in the first instance… particularly with things like tracking… finding good sound through good mic selection and placement, rather than fixing it with EQ in the mix.  Yet, it can be easy to forget or ignore this advice given how simple it is to throw extra plugins in an effect chain.

While writing ‘Dystopia‘, I came into this kind of situation… a problem which could have been fixed by additional tweaking, or extra layers of compression… but which actually had a simple, and probably ultimately better sounding solution at the source.

The track has the following background percussion pattern in various sections…

Within an 8 beat phrase, the first percussion ‘hit’ occurs on the third 16th beat, and has a quieter and lower pitched ‘grace note’ a 16th before that.  The below screen shot shows the MIDI sequence for the pattern, with the grace notes highlighted…

mix-issues-prevention-rather-than-cure-1

At the point of final mixdown and applying bus compression, I noticed that there were occasional waveform spikes at the points of these grace notes… the highlighted peaks on the below waveform show an example…

mix-issues-prevention-rather-than-cure-2

These spikes were not only quite strong (almost hitting 0dB), but occurred on a rhythmically odd (syncopated) beat of the bar… i.e. the second 8th beat of the bar… at the same point as the offbeat hi-hat sound.  When I was trying to apply compression, the strength and syncopation of these spikes were causing the same type of uneven, pumping compression I mentioned in my second bus compression article.  The problem could have been cured at the final mix stage by potentially applying a limiter or a fast acting compressor at the start of the effect chain.  But instead, I went back to the MIDI sequencing and took at look at the part itself.  Considering the note at the second 8th beat was just a grace note, and that it occurred on the same beat as a rhythmically far more important part (i.e. the offbeat hi-hat), the MIDI velocity of that note seemed quite high (at around 81).  Hence, I tried simply reducing the velocity of the grace note to about 70 as per the below screen shot…

mix-issues-prevention-rather-than-cure-3

…and this simple change benefited the mix in 3 ways…

  • It left more room for the offbeat hi-hat, and hence made the hi-hat clearer.
  • It wasn’t in any way detrimental to the in-context sound of the percussion part (actually, I think it sounded better after the change).
  • It had the effect of removing those waveform peaks, and hence let the compressor work more smoothly and musically (see the ‘after’ waveform below)…

mix-issues-prevention-rather-than-cure-4

Ultimately, a simple changing of MIDI velocity fixed the problem, and was far easier to implement than extra layers of limiting and compression would have been (and also avoided the additional side-effects that limiting and compression could have introduced).

Clips of the ‘before’ and ‘after’ full mix are below…

Before

After

The interesting take-home from this experience, was to always think a bit ‘out of the box’ with regard to mix problems… to consider whether there’s a simple preventative measure that could avoid or correct the problem in the first instance.  In 99% of cases, as the pro producers advise, such a prevention is probably going to be easier and more effective than the equivalent cure.

Sidechain Compression on Basslines

The use of sidechain compression on various elements of club tracks is these days fairly well prolific.  Given that fact, you might wonder why I’m posting about sidechain compression of basslines… a fundamental technique, that’s been used in club music for years.  Whilst it’s a well known technique, it’s also not always easy to execute, nor to judge whether you’ve achieved the correct effect. Hence I wanted to document a few notes about my typical approach to sidechaining basslines, and a real example from a recent track.

The first thing to note is that it’s very hard to give ‘ballpark’ settings for bassline sidechain compression. Appropriate settings depend a lot on the tonality of the kick drum, and moreso the pattern of the bassline. If I’m trying  to come up with initial settings for the compressor, I tend to think about it analytically, along the lines of the following. ..

  • Fundamentally, you want to avoid having the bassline and bass drum overlap by playing the same frequency range at the same time.  I aim to set the compressor to compress the bassline where any such overlapping exists.
  • As bass drums typically start with a higher frequency attack (e.g. click) and then quickly descend in pitch to a low drone as they decay, the decay portion is typically where you need to be careful of overlapping the bassline.
  • The attack and release parameters are important as they’ll typically contribute a lot to the feel or groove of the bass. For example if your bassline notes are sounding at the same time as the decay of the kick, you’ll typically want to use a fast attack setting, so the compressor is engaged quickly when the drum and bassline notes overlap. Conversely, if the bass drum and bassline notes are more separated in terms of timing (i.e. little overlap), a slower attack setting might provide a more natural sound.
  • Depending on how quickly a bass drum hit sounds after a preceding bassline note will similarly contribute to the release setting.  As with the attack setting, generally the greater the overlap of notes in this regard, the quicker release setting you might want to use.
  • EQ filtering of the sidechain input from the bass drum can sometimes be useful to prevent certain frequency ranges from triggering the compressor.  E.g. if the pitch of your bassline is relatively high compared to the drum,  you may want to apply some high pass filter to the sidechain input,  so that the low frequency parts of the bass drum don’t trigger the compressor (if the bassline is relatively higher pitched, there should be less overlap at the lower frequencies)

That said, I usually use the above thinking as a guide only, and ultimately the sound and feel/groove should determine the final settings.  This is a case where it’s highly preferable to use monitors with a sufficient low frequency response, so you can really hear (and/or feel) the effect of the sidechained compression.

For my track ‘Cantana 1‘, I used the following settings…

sidechain-compression-on-basslines-1

The controls which don’t have marked scales were set as follows…
Sidechain HP: 29Hz
Threshold: -12.6dB

The bassline in this track was very much a ‘drone’ type sound, rather than having clearly separated, distinct notes.  Because of this, there was not much (if any) natural gap between the bass drum hits and the bassline notes.  Hence I lent towards fast attack and fast release settings to try and maximize the separation between the two parts (i.e. avoid the aforementioned ‘overlap’ as much as possible).  I started out with a fairly fast release setting also to try and make the initial attack of the bass drum more isolated, and more prominent.  The threshold and ratio were set by ear… I just kept adjusting both until I achieved what I thought gave the best balance between the levels of the bass drum and bassline.  I.e. I wanted to strike a balance between…

  • Not having too much compression so that the bassline level was pushed down too far with each drum hit.
  • Not having too little so that the two sounds were too smeared and indistinct.

The final attack and release settings were also set by just auditioning various combinations and deciding which gave the best rhythm and feel (after starting with the dials on the faster end of the scale, as mentioned).

Before and after audio clips of the 2 parts soloed are below (again you’ll need headphones or monitors with adequate bass response to properly hear the difference)…

Without sidechain compression

With sidechain compression

Although the audible difference is fairly subtle, the bassline’s waveform changes considerably…

sidechain-compression-on-basslines-2
Normal bassline waveform
sidechain-compression-on-basslines-3
Bassline waveform with sidechain compression

Bus Compression 2

my experience in this process was one which helped me discover, and then cement the approach of progressive compression

This is my second instalment on bus compression, and the story of how I discovered what has since become my default effects chain for master bus compression.

I discussed in my last post on this topic the importance of applying compression in a staged and progressive manner through the entire sequencing and mixing process, rather than leaving all the work to a couple of plugins at the point of final ‘mastering’.  My experience in this process was one which helped me discover, and then cement this approach of progressive compression.

It centered around my track ‘Push On‘… when I’d completed mixing of this track, I was really happy with the sound… the bass line was strong and solid, and the snare and synth ‘stab’ sounds both had a nice punchy and aggressive dynamic.  What I basically wanted to achieve with bus compression was to maintain the general sound and dynamic as much as possible, while making the overall level more ‘competitive’.

My ‘goto’ compressor for bus compression is Cytomic’s ‘The Glue’, so first step was to insert an instance of this over the master mix.  Initially I used a slowish attack setting of 0.3ms, but I was finding a lot of transients were slipping through and causing digital clipping when the level was raised.  If I moved the attack back to 0.01ms, it caught all the transients, but also resulted in a fairly un-musical pumping effect.  The clips below contain 1. the track directly after mixing, and 2. the same track plus ‘The Glue’ with similar fast attack settings to what I used originally used…

Original track:

The Glue with fast attack:

The pumping of the compressor is evident from the uneven level of the offbeat open hi-hats.  While originally working on this I also noticed that the virtual ‘needle’ in The Glue jumped around unevenly, not moving with the rhythm of the track.  (From further experience since this,  I’ve come to learn that un-rhythmic movement of the gain reduction needle like this usually always results in uneven, pumping compression).

I bounced the uncompressed track to an audio file so I could get a look at the waveform (see below)…

bus-compression-pt-2-1

Notice the highlighted uneven peaks caused by the synth stab sounds (red) and snare hits (blue).  The synth stab used a fairly heavy chorus effect which caused the individual notes to peak at varying levels in the waveform.  My assumption was that at fast attack settings, the uneven peaks of both instruments were causing the compressor to trigger inconsistent amounts of gain reduction.  Coupled with the fact they were on syncopated beats of the bar (especially the synth stab which was on the 3rd 8th note), this was likely what was causing the un-rhythmic pumping of the compressor.

I used a combination of techniques to ultimately correct this, but the one which had the biggest impact was to put a limiter (Waves L1) in the signal chain before the compressor. This had the effect of evening out the transient peaks, and allowing the compressor to operate much more consistently, aswell as permitting the use of a slower compressor attack setting.  This resulted in a much more transparent compression.  The clip below demonstrates a recreation of the same signal chain…

…notice that the sound generally is much smoother, and the consistent levels of the hi hats in the original version have been maintained.

The end result was a considerable amount of compression was applied by The Glue, but it remained relatively transparent.  The final step was to use another instance of the L1 limiter at the end of the chain to bring the track up to competitive levels.  The final chain and settings were as follows…

bus-compression-pt-2-2

bus-compression-pt-2-3

bus-compression-pt-2-4

I mentioned that ultimately I used several techniques in addition to the above signal chain to achieve the result in the published version, and this included actually going back to the mix project, and applying more compression, and some limiting to the individual snare and synth stab tracks in that project (supporting my earlier comments about a ‘progressive approach’ to compression).  That meant that the first L1 instance in the master bus chain had even less work to do, and again produced an overall more transparent result.

The whole process was a really good learning experience, and the L1 > Glue > L1 effect chain has become my default for master bus compression.  If you struggle to get clean but loud results with your own master bus processing,  I hope sharing my experiences will be useful.