It's not often SOS readers get to mix a session tracked at Abbey Road — and when they did, the results were surprising!
Back in December 2013 our Session Notes column profiled a recording session carried out at Abbey Road Studio Two, by engineer and SAE Institute founder Tom Misner. Tom very kindly made available the full multitrack of a song from the session for SOS readers to practise their mixing skills, with the promise that if he liked any reader's mix better than his own, he'd put it on the album.
By the deadline of January 31st, we had received about 40 entries, and it fell to the two of us to listen through to all of them and make a shortlist of three for Tom's consideration. Both of us were struck by the number of basic problems that recurred time and again, so we thought it worth writing this short 'where did it go wrong?' article to try to point out some of these pitfalls, and explain how they might be avoided in future.
The original article simply gave an email address, with no further instructions about how to submit mixes for consideration. Faced with this lack of information, most entrants displayed common sense, but not all. Several were disqualified because they placed their mixes on streaming sites with no option to download them. Others didn't bother to email us to find out the deadline, and missed it. Still others sent four or five different versions of their mixes, and one person even submitted not only mastered and unmastered mixes, but also full stems. Sending multiple versions just seems indecisive, and imposes too much on the person receiving the mixes.
Other things that didn't impress us, and should have been addressed before sending, included mixes with 10 seconds of silence at the start, mixes that were 15dB away from hitting 0dBFS, and conversely, mixes that had been squashed to oblivion in a vain attempt to stand out from the competition. Oh, and let's not forget the mix that had been wrongly sample‑rate converted and thus played back a tone flat.
Once all the ineligible, inaccessible and unplayable files had been weeded out, we were left with 33 entries. And of these, more than half could be ruled out within 15 seconds' listening (though, for our sins, we did listen all the way through). The reason? They screwed up the most important single element of the song.
Our Session Notes article made perfectly clear that, although this was an ensemble recording, singer Lyndsey Ollard is very much the feature artist. It follows that her vocal is the key component around which the mix must revolve, even if this wasn't already apparent from the song and the arrangement. Several mixers seemed unable to take this fairly obvious point on board, and buried Lyndsey's vocal behind the rhythm guitars and drums. Others made the vocal far too loud. Some did both, at different points in the song. One otherwise decent mix inexplicably panned the vocal noticeably to the left.
The session as made available to download included quite a few different vocal takes, and Tom deliberately left it up to the mixer to choose and comp the right performances. This was an aspect that most entrants handled pretty well, but some decided to use spare takes to create doubling or widening effects at various points in the song. Not a bad idea in principle, but in practice, we didn't feel that anyone added anything worthwhile by doing this.
Those who buried the vocal were, at least, in a minority: it was getting the actual vocal sound right that seemed to flummox almost everyone. The initial vocal entry at around 0'13 is soft and very exposed, accompanied as it is only by a couple of guitar parts. In this context, what it really doesn't need is heavy compression, saturation, or high‑frequency boost, but it was astonishing how many entrants applied all three! This vocal entry sounded unacceptably harsh and sibilant on almost all the mixes that were submitted. Several were so compressed that noise on the vocal track was clearly audible between phrases.
Where, as in many of these mixes, a vocal has been presence‑boosted to the point of being harsh and uncomfortable to listen to, this has often been done simply because everything else in the mix has been made a bit brighter than it needs to be, frequently as the result of individual sounds being 'polished' in isolation and therefore out of context. When everything in a track is bright, the only way to make something stand out is to make it too loud or give it way more presence than is good for it. The key point here is that not everything in a mix needs to be in the spotlight and in full focus, and if you are prepared to allow some elements to sit back, then others, such as the perfectly well‑recorded lead vocal in this instance, will find their place at the front of the mix without needing to be over‑EQ'd for presence.
Reverb and related effects do certainly have a role to play in supporting such an exposed vocal, but mixers who used them appropriately were very much in the minority. Some chose to drown the vocal in epic hall 'verbs, while others employed wildly unsuitable short ambiences or slapback delays. Few seemed au fait with the all‑important techniques of using pre‑delay and EQ to separate the reverb from the vocal.
A number of the mixes also suffered from excessive use of artificial reverb in general, not just on the vocal. When you are listening to the same piece of music over and over again during mixing, it is all too easy to cease to hear really obvious anomalies. As with the bass level, a good reality check for reverb is simply to take it all out every now and again, listen to the track without any for a bit, and then gradually reintroduce it. Routing all your artificial reverbs to a single subgroup fader makes this a very easy process, and also maintains the relative levels of any different flavours of reverb you may be running. (Note that this suggestion refers to artificial 'effect' reverb, rather than anything that might be considered a fundamental part of a sound source, like drum ambience mics — pulling those out can change the sound of a mix so fundamentally that it will feel like you are starting again.)
'Catching Up' is not a song with a static arrangement: it's a builder, which starts quietly and contains two major dynamic steps up. The first of these, at around 1'03, sees the vocal and guitars joined by drum kit played with brushes; then, at around 2'17, the brushed kit gives way to a kit played with sticks. (It's possible that many problems with the intro vocal sound can be traced back to the need for the vocal to cut through the mix towards the end of the song. If so, more differentiation between the vocal sound in different sections was needed.)
One of the big challenges in mixing a song like this is to preserve or even enhance the overall dynamic 'arc' of the music. The individual sections need to work in their own right, but they also need to fit together in a satisfying fashion. Many of the mixes submitted achieved respectable balances at least until the second drum entry, but this was definitely the point that separated the men from the boys. Some threw everything at the song too early, or overdid the mix‑bus compression, meaning that the second drum entry was robbed of its impact. Others mixed the sticked drums far too loud, so that they overpowered the song.
Most entrants managed to achieve a reasonable mix tonality in the quieter sections, but many failed to compensate for the additional mid‑range frequency content introduced by the sticked drums and other loudly played instruments, allowing the mix to become tinny or brittle from 2'17 onwards. As described in the article, Tom and his team employed multiple room mics to capture the sound of the drum kit in Studio Two's celebrated live room. Several mixers got carried away with these, and at the other end of the spectrum, some relied too heavily on the close mics, ending up with a disjointed and unbalanced kit sound.
As well as the core rhythm instruments, the multitracks made available by Tom included a number of overdubs: backing‑vocal 'oohs' and several tracks of electric guitar licks and ornaments. Quite a few mixers felt they had to use all of these, and ran into problems as a result, especially as not all the overdubs were perfectly in time with each other. The most successful, in our view, were those who pruned reasonably aggressively, but made sure to mix the ornaments they did use loud enough.
If it sounds like we are being somewhat prescriptive over what constitutes a good mix, that's really not the case here at all. There are always several significantly different‑sounding options for any mix, all of which could be said to be equally viable as a finished balance. Indeed, there's also a number of different, but equally valid, methods of arriving at the same result. Successful commercial‑release mixes can and do sound significantly different from one another in terms of how dark or bright they are, and how dry or ambient. Provided that nothing is too far out of whack, our ears happily adjust to the relative balance of frequencies we are presented with. And as a similar overall tonality tends to be maintained throughout any related collection of tracks, there is nothing of contrast to reveal the brightness or dullness of what we are listening to.
Put your iPod or iTunes collection on shuffle, then, and some wide variation between tracks will be revealed — yet the professionally mixed tracks will all still sound like commercial mixes, even if some are intentionally rough and others smooth and expensive‑sounding. None of them will be wallowing in uncontrolled deep bass and the top end won't be drilling a hole in your head, and if your mix is doing either (or both) of those, it will be immediately obvious with a simple level‑matched A/B test against some commercial CDs — not comparing loudness or density, but just basic tonality. Even if you only do it once right at the end of your mix, it can be enough to prevent you from getting something really basic, really wrong.
Mixing is, above all, a learned art, however, and every one of the participants in this process will have benefited from the experience — benefited to the extent that every one of them would doubtless produce a better mix if they were to do it again now, with fresh ears and a few weeks' distance from the process. Home studio owners are notorious for not finishing their tracks and never producing final mixes, but you only get good at mixing by doing a lot of mixing. Maybe there is a lesson there.
Getting the bottom end of a mix right is always a challenge, particularly for those working on headphones or in compromised monitoring environments, and so it proved here. Several otherwise decent mixes had the bass guitar far too loud or completely inaudible. Many others failed to control its very low frequencies, resulting in an unwanted surplus of sub‑bass or, worse still, a wildly uneven low end. Still others achieved a reasonable balance of frequencies at the expense of making the bass guitar itself sound muddy, boomy or indistinct.
Even highly experienced mix engineers agree that getting the bottom end right is one of the more challenging aspects of trying to achieve a satisfying, commercial‑sounding mix. An excess of any single frequency area will always make it difficult to appreciate the other elements in the mix, but too much deep bass is especially destructive. Even if the oppressive low end doesn't have the listener reaching for the volume control to turn the whole track down, it will still be masking a big chunk of the lower mid‑range, obscuring a lot of the fundamental tonality of the other instrumentation.
Too little deep bass, on the other hand, isn't half as damaging: in fact, you may only notice its absence by A/B'ing your mix against one that has a bit more in it. A mix without much deep bass will often simply sound like it was meant to be like that. Slightly bass‑light mixes — 'slightly' being the operative word — are often perceived as being cleaner, clearer and more detailed, simply because there is less masking going on.
A simple technique for determining the final bass level towards the end of a mix is to take the bass out altogether for at least half a pass through the track to allow your ears to adjust, and then slowly re‑introduce the bass track. If you do this without looking at the fader, and stop at the point where your ears tell you there is enough bass, you'll often find the fader is as much as 3dB lower than where you previously had it.
The bass guitar is inconsistent in level in many of the mixes, in some instances sounding as if both compression and EQ have been used without managing to achieve the objective of a smooth, consistent bottom end. There are two reasons for EQ'ing a bass guitar, and they require very different approaches. The first is to tackle any obvious resonances or dead spots inherent in the instrument that might be causing some notes to be louder or quieter than the others. Over‑loud notes can easily be trimmed back with a high‑Q (narrow bandwidth) EQ, as high‑Q settings tend to be audibly benign when cutting. High‑Q boosts for dead spots, however, are less advisable, as they tend to have a distinct audio signature — slightly coloured or nasal — that is often worse than putting up with the odd weak note. A better solution can often be found with some precise level automation, or even editing the audio file itself, if the offending note only occurs a few times.
The second common objective when EQ'ing bass is to achieve a desired tonality. Here you need a very low‑Q (wide bandwidth) setting, with the aim of moving a large area of the spectrum up or down in level, whilst keeping the relative level of the notes intact. If you use a high‑Q setting to try to adjust bass tonality, you are in danger of affecting the level of just a couple of notes, thus re‑generating the kind of problem that we were trying to solve with the first stage of EQ.
A shelving EQ, which maintains the same attenuation at all points below its corner frequency, allows you to trim back the bottom octave without losing all the weight in the sound. If that doesn't give you enough control of the really low stuff, a very wide‑bandwidth, bell‑curve cut with a centre frequency in the low mid‑range can achieve a slow progressive roll‑off to compensate for the higher energy exhibited by very low notes. A second EQ band can be used to restore the missing mids, if required. In general, corrective EQ should be applied before compression, to prevent the odd hot‑spot from driving the compressor unnecessarily, with 'tonal' EQ applied after the desired dynamics have been achieved.
The other instrument with significant low‑frequency content is the kick drum, and the combination of bass and kick drum is indistinct in many of the mixes. The best way to combine these two rather depends on the role each plays in the piece and to some extent on the genre of music but, in general, it is often preferable to let the bass appear to be the lower of the two — though this appearance can be something of an aural illusion. The aim is to give the kick enough of a presence click to make it easy to follow as part of the overall rhythm pattern, whilst still maintaining enough bottom end to keep it sounding like a kick drum on any hits that don't have a corresponding bass note. If you take some lower mids out of the kick as well as adding a click, it will appear to be 'above' the bass without needing to be thinned out at the bottom too much. A well-controlled bass guitar and kick drum should be able to sit together perfectly well in a mix, with neither having to be excessively 'carved out' to make room for the other.
The uppermost of these three plots shows an averaged frequency spectrum for half a dozen commercial mixes of tracks in a similar genre to 'Catching Up', all featuring female vocalists and similar instrumentation. The response is entirely typical of contemporary pop mixes, with the peak level appearing around 70Hz, roughly 6dB above the peaks in the mid‑range. The high‑frequency energy drops away smoothly from 3kHz down to around 12kHz, and the 500Hz to 2kHz region is roughly level. The middle plot shows an averaged frequency spectrum plot of a random half‑dozen of the competition entries (excluding the three best). The peak is still at 70Hz, but the big difference is what happens below that. In the commercial mixes, the bottom end falls away smoothly, resulting in 20Hz being ‑18dB relative to the peak. In the competition mixes, 20Hz is just ‑8dB and there is still a very significant amount of energy at 30Hz: very audible, but not particularly pleasant or useful in such quantity!
The third screen shows a 'difference' plot between the commercial spectrum and the competition mixes — in other words, the transfer function necessary to make the latter sound more like the former. This makes the uncontrolled bottom end and relative shortage of upper mid‑range in the competition entries very clear.