Mix Rescue: St.Vitus

We get to grips with a mix in which big, up-front vocals are crucial to the impact of the song, and dispense plenty of tips along the way...

When Phil Hall-Patch sent his track 'Word Gets Around' to Mix Rescue, his main issue was that he couldn't get the kind of vocal sound he was after, although there was also a problem with the mix as a whole sounding a bit constrained, despite some adventurous panning and doubling effects. Loading Phil's multitrack files into my Cockos Reaper-based mix system, what really struck me was the vocal performance, which had so many great little nuances that it almost seemed a shame to obscure them by mixing in anything else! (Now that would have been a quick Mix Rescue...) On this basis, I decided to build the mix around the vocal in order to maximise its emotional impact, so I started my mixing with the voice on its own.

Vocal & Acoustic Guitar Tones

Although I tend to use a lot of freeware for Mix Rescue (so that you can more easily try things out for yourself at home), on this occasion I decided to call up my URS Console Strip Pro plug-in so that I could quickly surf a selection of desirable analogue emulations for something that would really suit this particular singer's timbre. As I would often do with singer-songwriter-style material like this, I first tried an LA2A emulation (famed for its near-undetectable gain-reduction and touch of valve sound), but that didn't give a fast enough gain action to exaggerate the vocal details in the way I was hoping for here, and it made the sound a bit stodgy. A Dbx 160 emulation was cleaner and a bit more aggressive with the gain reduction, but gave things a slightly hard edge that wasn't as appealing. Third time lucky, though, because the next emulation I tried (of a Urei 1176LN) hit the nail on the head, giving lots of nice breathy FET distortion components while controlling the dynamics fairly assertively.

The degree of proximity bass boost on the vocal recording was causing the compressor to respond erratically at points, so I preceded the Console Strip Pro with a simple high-pass filter, although I set this to roll off lower than I often do because I still wanted to retain a good deal of warmth in the tone. The raw track was also woolly-sounding at the high end, and I therefore ladled on a generous 7dB high shelving boost at 14kHz to bring all the little breathy elements of the sound to the fore. The combination of this with the low-end warmth really helped to make the lyrics seem like they were right in your ear.

With the basic vocal tone in the can, I started work on the acoustic guitar, which appeared to me to constitute the harmonic and rhythmic backbone of the track. This had been recorded with an SE Electronics SE2A mic in an SE Electronics Instrument Reflexion Filter, with the mic positioned quite close to the fingerboard around the 12th fret so that it picked up quite a lot of finger noise. A DI signal had been captured at the same time, and this had good sustain but suffered from the usual 'dead' sound, as well as some big low-end peaks in the more driving sections whenever Phil's right hand hit the body of the guitar.

The LF thuds could be dealt with simply with a high-pass filter on the DI channel at 65Hz, and I also put in this filtering to deal with the same (albeit less pronounced) problem on the mic channel. My next thought was to experiment with the phase relationship between the mic and DI sounds — because sound takes a small amount of time to travel through the air to the mic, the DI signal always leads the mic signal in this kind of recording setup. One approach is simply to look at the audio waveforms and realign the waveshapes by eye, but that's not as easy with an acoustic guitar part as it might be with an amped electric bass, because the waveshapes don't often match that closely. Besides, adjusting the phase relationship is as much about taste as it is about technical accuracy, because changing the phase relationship and balance between the two signals adjusts the nature of the comb-filtering between them and thereby gives access to a whole set of creative tonal options. You can easily line up the waveforms perfectly and then not like the sound as much as when they're out of alignment!

Different compressors can do different jobs and Mike used two to process Phil's lead vocals. The first one, a Urei 1176LN emulation from the URS Console Strip Pro plug-in (far right), was chosen not only for its assertive dynamics control, but also for the subtle enhancement imparted by its distortion harmonics. To pull up the performance details further, Jeroen Brebaart's PC2 (right), set to its psychoacoustically modelled 'Complex' mode, was used.With this in mind, I decided to use a phase-rotation plug-in (Betabugs' funky little Phasebug) so I could experiment more easily. Phasebug doesn't just delay the whole signal, but rather delays each frequency differently according to its wavelength, thereby delivering a constant phase-change across the audio spectrum. (For some more explanation of how phase works, check out 'Phase Demystified', back in SOS April 2008.) Despite the apparent complexity of this task, the processor has only one control (the amount of phase change) and all you do is insert an instance into one of the two tracks (the DI in this case) and then twiddle the control until you get the sound you most like. A setting of 150 degrees rotation worked best for me here, solidifying the combined sound in a way that would have been impossible to achieve with EQ.

There was a place for EQ here as well, though, courtesy of Leftover Lasagne's Pushtec 5+1A, a freeware plug-in clearly heavily inspired by the venerable Pultec units. Given the space still available in the mix at this point, I had pretty free rein to sculpt the sound to taste. Both DI and mic channels benefited from a stiff boost (or 'Push') at 100Hz to warm them up a bit, as well as partly compensating for mid-range effects of the preceding high-pass filtering, while the mic part also needed some cut above around 5kHz to take the edge off the pick noise, which had been exaggerated by virtue of the choice of miking position. The DI was still making the combined sound feel a bit hollow, though, so I boosted the mid-range too, with peaks at 300Hz and 2kHz, to fill it out.

Although the guitar tone was now broadening out, there was still a variety of dynamics problems to deal with. The main thing was that the percussive elements of the part were much more pronounced than the instrument's sustain, which meant that the guitar didn't fill out the texture in the way I was hoping it would. Bussing the DI and mic tracks together and then compressing the bus channel was clearly the way forward, but there was a further fly in the ointment: with traditional compression, the loudest section of a processed track receives the stiffest compression, but the loudest part of this particular guitar track was during the pre-chorus, which had the kind of punchy sound that I really didn't want to compromise in the process.

My solution was to this was two-fold. First of all I used a low-threshold 1.2:1 compression with fast attack (0.4ms) and a very slow release (1.3s) to act almost like an automatic volume control, simply pulling down the louder pre-chorus sections to a level that was more consistent with the verses and choruses. Because of the low ratio and slow release, this had precious little effect on the shorter-term dynamics of the part, for which purposes I turned to parallel compression (in other words mixing together compressed and uncompressed versions of the bus channel), as this is a technique well suited to adding sustain without compromising attack transients. The parallel compressor (Buzzroom's Grancomp, in this case) was set at a 2:1 ratio and with a threshold that was low enough to trigger around 10-12dB of gain reduction on peaks. In order to bring up the sustain phase of each note, I went for a fast 28ms release, but left the attack long enough, at 12ms, to allow some attack transient to break through.

This achieved exactly the kind of extra sustain and body I was after, and meant that I could lay the instrument further behind the vocal without the mix emptying out unduly. However, even at a lower level, the higher frequencies of the parallel-compressed guitar part were beginning to conflict with the all-important vocal intelligibility frequencies, so I shelved 6dB of high end from the guitar's compressed signal at 2.7kHz to clear out a bit of space. I have to say that I preferred the soloed guitar sound without the EQ, but there was no getting away from the fact that the vocal was more important, and the less forward-sounding guitar definitely worked better in context. This is why EQ'ing instruments 'within the mix' is so important — settings that work for the mix aren't necessarily what make an instrument sound best when solo.

Bass Synth & Drum Loops

Normally, when I'm approaching a mix I try to base my decisions as to which tracks are most important on my knowledge of the style, but for a song as left-field as this one, genre judgements aren't tremendously useful, so instead I took some cues from Phil's own rough mix, in which the bass occupied quite a prominent position, while the drums took a slightly more low-key role much of the time.

The synth that Phil had chosen for his bass part (a Moog Voyager patch from Spectrasonics Trilogy) was, in his words, the 'meatiest' he could find, and it was probably the patch's dense low midrange (100-500Hz) which informed his choice. However, powerful as they came across in isolation, these frequencies piled up with the rich lower-range fundamentals of Phil's baritone voice, making the mix as a whole sound rather muddy. It also turned out that fundamental frequecies of the bass synth notes were actually quite weak, so I decided that I'd shift some of part's low mid-range and then add in more of the fundamental frequencies (which slotted very nicely underneath the vocal range) to reinstate some of the meatiness.

Phil had recorded his acoustic guitar part via mic and DI simultaneously. Mike adjusted the phase relationship between the two tracks by ear using Betabug's tidy little Phasebug phase-rotation plug-in, in order to achieve the most suitable combination.Although you can approach a task like this with EQ, with melodic parts like Phil's I find that I can get more useful control over the timbre using other processes which better track the changing pitches in the line. So rather than just boosting the low frequencies of the synth part, I high-pass filtered it instead, and then added the fundamental frequencies back in with a separate additional synth. My normal choice for this in Reaper is the simple little built-in ReaSynth instrument, which I set to its fairly dull-sounding triangle wave and then low-pass filter savagely to remove everything but the very lowest harmonics. It didn't take long to re-program Phil's part, although it was important to be careful with placing the note-offs. When you're working with rhythmic music they can have almost as much impact on the groove as the positioning of note-ons.

I now had controllable low-end 'oomph', but still needed to shift the emphasis in Phil's synth part onto the higher frequencies, in order to clear out that low-mid range. Again, I chose not to do this with EQ, but used distortion instead: a combination of Silverspike's Rubytube and Voxengo's Tube Amp, both of which model tube circuitry. When I'm after extra harmonics in a sound it's not uncommon for me to use more than one distortion plug-in, as they all have different characteristics. For example, here Rubytube's Limiter was useful for controlling some of the filter-sweep-induced level peaks, while Tube Amp provided more of the actual distortion products. By controlling the bass sound fairly strictly in this way, I was able then to balance it at a lower level behind the vocal without losing its character, and the extra sub-bass information helped make the spectrum of the mix as a whole wider and more hi-fi.

In the frequency domain, the drums presented few problems that a bit of high-pass-filtering with a couple of narrow notches couldn't sort out, although I did use a little distortion on one of them, as more of a special effect to differentiate it from the others and therefore help the arrangement contrast. However, the dynamics in a number of places exhibited the opposite problem to the one I'd had with the guitars: they had lots of sustain but not really enough attack. The first thing I tried was Stillwell Audio's Transient Monster, a transient shaper plug-in offering a similar effect to that of the well-known SPL Transient Designer. Boosting a few decibels with the Attack gain control certainly helped pull the transients out more, which meant I could reduce the overall fader level to clear some space for the more sustained parts in the mix. However, I still hankered for an extra helping of punch at the low end, so I supplemented the Transient Monster processing with an added kick-drum sample, which I copied and pasted to support some of the kick drums in the loops.

All that then remained in terms of parts were a couple of string sections and an echo-y granular-style synth pad, which I mixed in at a suitable level before cutting away any frequency ranges that were conflicting with the other parts, particularly in that tricky 100-500Hz region. I also felt that a little bit more background complexity and atmosphere was needed during some of the verses, so took advantage of Heavyocity's excellent Evolve Kontakt Instrument (left), which I've recently had in for review, sprinkling in some extra subtle rhythmic effects and swells at strategic points.

Expanding The Stereo Picture

Adding the Evolve parts also had a beneficial side-effect, in that they provided more in the way of stereo width and interest. Because all the main parts (vocal, guitar, bass and drum loops) all seemed to work best at the centre of the stereo image, and none of them really provided much in the way of stereo width (even the drum loops were mostly in mono), I had to do a fair bit of work to create a satisfying stereo spread to the production overall. The most useful tools for this were two ping-pong delay effects — one an eighth-note patch for the drum loops, the other a three-sixteenths patch for the vocals and acoustic guitar. Using Reaper's ReaDelay plug-in, I set them up with four delay taps apiece, each tap moving more towards the stereo extremes.

Mike added some subtle atmostpheric elements from Heavyocity's Evolve Kontakt Instrument to give the mix a bit more background complexity and stereo interest. Here you can see one of the patches he used: the evocative sound of bowed cymbals.Clearly the most obvious effect to give a sense of space and width is reverb, so naturally I turned to that on a number of occasions too. The acoustic guitars had some smooth synthetic reverb from Stillwell Audio's Verbiage, while Silverspike's Room Machine 844 did its usual efficient job of adding a room signature to the main verse drum loop. A long and very dark plate patch from an impulse response running in Christian Knufinke's SIR2 provided the pillow-like special-effect wash behind the lead vocal, but there were also two other less obvious effects spreading the vocal across the stereo picture: my usual short pitch-shifter delay patch and a short 0.35s reverb patch that I'd widened further by pushing the Stereo IR control in SIR2 up to 138 percent.

A handful of modulation effects played their part, too. A fairly simple phaser effect from Reaper's bundled Jesusonic plug-ins swirled one of the string lines around, while a chorused tempo delay did something similar for the acoustic guitar parts. A second, more distorted bass layer that Phil had added for the second pre-chorus section provided an opportunity to widen the impression of the combined bass sound with an instance of Schwa's Oligarc. The modulation possibilities of this analogue-style multi-effect plug-in are pretty mind-boggling, but now that I've managed to get the hang of how it works I find myself turning to it surprisingly often. For this application I used the Drive and Chorus sections, and modulated them both from different sources in order to achieve a really wide effect.

As a final touch, and to help bind the whole track together despite the obvious verse/chorus mix contrasts, I layered a background ambience sample under the entire song (a kind of gentle 'factory floor' ambience), high-pass filtering it to keep it from muddying the low end. This is such a simple thing to do, but it's amazing how it can help glue a track together, give some stereo width, and also add something of that elusive feeling of 'air' around the sound.

Further Vocal Adjustments

At this point, most of the sounds within the mix were now balancing well together, and I'd also used multing to help differentiate the sections of the song structure (see the 'Multiple Multing' box for more details). So as my final task I turned my attention back to the lead vocal, to reassess its processing in the light of the mix as a whole. The first thing I did was to tighten up a few aspects of the performance in Melodyne, because when vocals are so far up-front in the mix, timing and tuning discrepancies can have a destructive impact on the way the whole mix is perceived. That said, I was particularly careful to avoid overdoing the pitch manipulation, because many of the subtle pitch fluctuations contributed a great deal to the expressive nature of the performance.

For maximum de-essing control, Mike used an EQ plug-in under automation control. Although manual de-essing can be a slightly tedious process, it's at least fairly straightforward to do, because the dense, noisy waveforms of the 'S' sounds are usually clearly distinguishable visually, so you can just line up the automation data by eye (as shown in this screenshot).Toning down the sibilance was the next stop, because the EQ boost I'd already applied to get the main vocal tone in shape was over-emphasising the 's' sounds. Where a vocal is very exposed or up-front, like Phil's, I tend to steer clear of automatic de-essers, and instead de-ess manually using an automated EQ band, because it gives me more accurate and consistent results. The setup for Phil's vocal was quite typical in this regard: a high shelf at 8kHz with its gain under automation control. Automating an EQ like this might seem a bit tedious (and it is), but it's comparatively straightforward to do in most sequencers, because the dense and unperiodic waveform shapes of sibilant sounds are quite easy to spot on a zoomed-in waveform display, which means that it's very quick to line up the required automation points by eye.

The main vocal task, however, was further reducing the vocal's dynamic range, so that every tiny facet of the performance could be heard. My existing compression had already done some of the work, but I couldn't really push that any harder without it sounding too obviously processed, so instead I plugged in Jeroen Breebaart's great little PC2 compressor as well. This plug-in has a special complex detection mode that uses the principles of psychoacoustic loudness perception to achieve very transparent dynamic-range reduction, even when you need quite fast attack and release times to really even out the small-scale dynamics of a part. Using this compressor at a 3:1 ratio I managed to dial in 6-9dB of further gain reduction without too many side-effects, even with a release time of only 36ms.

Even then, however, I knew there was only so far I could get with compression. The problem is that it's an automatic process, without the intelligence either to gauge how changes in the song's arrangement affect vocal intelligibility, or to decide which particular syllables are most important to a vocal's lyrical delivery. It's for these reasons that I'm convinced that the only way to squeeze the last ounces of audibility from a vocal in your mix is to tweak its level using the most intelligent processor you have available: yourself. What I mean, of course, is that you need to automate the level of the vocal manually.

Given the importance of the vocal to this mix, I went to town on the level automation for the voice, riding the fader over more than a 12dB range to pull up all the little details that had prompted me to build the mix around the vocal in the first place. These were usually at the beginnings or ends of syllables, as the voice was in that unstable transition between one note and the next, or at those points where the flow of breath was weakening at the beginnings and ends of phrases — in short, anywhere where the sound had a bit more 'grain' to it.

The automation also gave me an opportunity to ride up a few consonants where the words weren't coming through well enough, and I made sure to check the audibility of the lyrics on the various different monitoring systems in my mixing setup (as well as at different playback volumes) to try to ensure that the lyrics would translate as well as possible to real-world listening environments.

Big Vocals, Big Mix

This turned out to be an unusual mix in a lot of ways, but a great deal of what I did can also serve more general-purpose ends. Obviously, up-front vocals are vital to a lot of styles, and rarely is time spent on that area of a production wasted. But if you're feeling that your mixes don't sound expansive enough, then some of the other tricks I used (background ambiences, additional sub-bass, stereo enhancement effects) may also come in handy.