I promised I’d post up September’s track a month ago now and I didn’t – sorry about that, but here it as at last. I also said that the next post would be of a more technical nature partly because right this minute I don’t have any great spiritual (or none spiritual for that matter) rantings running around my head and partly because doing this track I learned a couple of really useful tricks on mixing vocals which I think are worth sharing. So here goes:
So for those of you who want the track I suppose its only fair to give you first priority as many of you might not want to sit through pages of technical mumbo-jumbo. As ever you can listen online:
Now, on with the interesting technical geeky stuff – the reason this track presented me with some challenges is that Nathan who is singing has an immensely dynamic voice – by that I mean his loud bits are really loud and his quiet bits are really quiet.
From a listener’s point of view this is nice – its gives a sense of emotion and expression rather than some bland old pop song, from the engineers point of view this is a nightmare because once you add all the other instruments the quiet bits disappear and the loud bits jump out and poke you in the eye and cause your track to peak really badly.
To give you some sense of what I mean by dynamic take a look at this waveform and small audio snapshot:
Snapshot of Nath’s vocal waveform before compression
Se how big the loud bits are vs how small (almost not visible in some places) the quiet bits are. Now the obvious fix to sorting this kind of things is simply to whack the thing through a compressor but in this case that isn’t gonna work, or more to the point if you make it work it’ll sound terrible and obvious… so how do we solve this?
5 steps to tame a wildly dynamic vocal track:
- Do what I forgot to do and teach the singer about good mic use and dynamics. Dynamics are good but need management; for example the singer should move away from the mic at the loud parts and up close at the quiet parts – essentially doing their own natural compression. I failed miserably to do this so was left with trying to sort it with technology
- Automate! Trust me if you can do this its worth it for a track like this. You basically need to go through each syllable that has been sung and automate the volume on the track before the signal is sent to any kind of compressor. Compressors are not always quick enough to catch the first set of peaks so you can find nasty little peaks still creeping through – If you can’t do this never mind you can always put another compressor in.
- EQ first – If you put your EQ & de-esser before your compressor you can get rid nasty peaks that shouldn’t be there anyway such as big bass pops from letters like “B” and “P” or over sibilant ess’. If you don’t do this you’ll end up triggering your compressor with irrelevant noises and not the actual vocal track. It is worth noting that all this can remove some of the crispness of a vocal but you should put that back in at the end.
- Multiple compression – Until learning about this I was trying to just tidy stuff with a single compressor – the problem that is either it doesn’t really work or you can hear it. The holy grail of compression is that you can’t hear it (unless you’re better at this than me…in which case you probably can). To do this you need to use multiple compressors together in series to slowly increase the gain of the track and remove the peaks.
- Use a limiter – Even after all this you’ll still get some nasty peaks, particularly on high frequencies that can cause clipping (or digital distortion) if you pop a limiter on the last compressor you can usually remove these.
So then here’s the process I used for this track:
First the automation:
Nath’s vocal channel after automation
You’ll see I’ve been pretty vicious here – with some 13db of change over just this little bit of audio but doing this just means I can be quite specific about what is lifted or what is turned down. On a more stable vocal track this might not be so important and you could perhaps skip this step but at this level of dynamics its worth doing. To put this in some context – one thing you might often do as an engineer is to put a gate on a vocal track to remove all the breaths between words. On this track when I tried to do this I found that despite the fact that track was normalised (some bits are as loud as they can be) when I set the gate correctly to cut out the breath it didn’t actually open for some of the words – that’s pretty dynamic…..well that or there’s some rather heavy breathing going on!
Setting your EQ
There’s not a lot to say here really, you’ll need to use your ears because this part massively depends on the mic you’ve used, what bass roll-off settings that mic might have and the specifics of the vocalist. However as a rule of thumb you can assume on a vocal you don’t really want anything below 100Hz and perhaps even as high as 150Hz. It is worth pointing out that this part of processing is ONLY clean up the vocal and remove any funny pops or sibilance rather than make it all smooth and creamy – what you’re wanting to do is make it simple for the compressor. Interestingly if this were a kick drum and I was working live I would always EQ after the compressor….but that’s another story.
Series chained compressors
Now this is cool bit – what you’re gonna do is gradually remove the lumps and bumps in the audio, normally you can do this in 2 compressors and if that’s not good enough stick a third in however with this track I actually used 4! – However for the purposes of this tutorial we’ll just look at 2.
Oh by the way if you’re using logic you can’t put this compressor as an insert on the channel with the vocal on otherwise all the automation is pointless cos that is after the automation – I route the channel to a bus and apply the compression there.
The first compressor wants to be pretty evil and harsh but pretty quick:
1st stage of dynamic vocal compression
In logic the best compressor to use for this seems to be the “Opto” compressor – if you’re using outbound “real” compressors then the general advice seems to be that an LA2A will give you the best results.
You’ll see here that I have a fast attack, a fast release and a pretty high ratio of nearly 7:1. Obviously you’ll need to set your threshold to suit your particular track but essentially what you’re after is a total reduction of somewhere between 3 & 5db at your highest peaks.
You set your release quick so that you’re only chopping the top of the peaks off and as you watch the track your db reduction meter should be bouncing around quite a bit and simply not moving at all in the quiet bits.
You could also apply a but of gain increase (2 – 4 db) just to give the next compressor something more to work with, however in this particular instance I’ve not done because the track was so dynamic anyway that was unhelpful.
The second compressor needs to be much more polite:
2nd stage compressor
In this compressor we need a slightly slower attack because although we want to remove some of the dynamics we still want to leave in the attack or punch of the start of words – if you make this too quick you’ll start to kill the words a bit.
We give the compressor a longer release too. STOP – a word of warning here – what you’re after is a release long enough to avoid a pumping noise as the compressor cuts in and out. Your aim should be that generally the compressor has let go (or released) by the time the next word or significant transient in the waveform come along because otherwise you’ll get a bit of a weird effect as the compressor comes in and out at odd places half way through a sentence and you’ll also start to affect the EQ of the track with the compressor which you want to avoid.
You’ll notice that the ratio is much smaller on this compressor, I’d say you want it somewhere between 1.5 and 2.5:1 depending on if you’re adding another compressor after this (smaller if you’re not).
With the threshold you’ll want to adjust this so that the compressor is basically in all the time but never exceeds a 5 or perhaps 6db gain reduction. If you find you can’t to that then that’s when you ‘ll need to start thinking about using another compressor in the chain.
And finally you want to add a bit of gain increase on the output now otherwise you’ll lose a load of level.
The finished product
Once you’ve done all this hopefully you’ll have a vocal track is which much, much easier to fit into your mix and hear, but should still sound reasonably natural.
Here is new waveform and audio snapshot of the same piece of audio I showed you at the beginning of this tutorial but after I’ve done all this processing:
Nath’s vocal after applying the dynamic vocal correction
You can apply the same techniques to spoken word as well however you’ll probably find that with spoken word the first compressor will being doing more as you tend to have less dynamics.
If you’re going to use a knee on your compressor to smooth it out I would put it on the second one not the first one, and remember with compression your general aim is to have it as invisible as possible and normally the biggest give away is the in the release, you can hear a pumping when it’s too quick or if it’s too long and not releasing before the next word you can hear warm word before the compressor cuts in and then it gets thinner as the compressor cuts in but some words are randomly warm again because the compressor has only just managed to release so watch your release times.
If you’ve found this useful you might want to have a look at these websites after all it’s where I stole most of my ideas from so its only fair to quote them: