This month, we embellish some repetitive drums by adding percussion, and get busy with the send effects.
Singer‑songwriter Joe Murphy has been crafting tunes for a few years, but had hit a wall when it came to fine‑tuning the arrangement and mix of 'Casting Shadows'. He'd picked the sort of sounds he knew he wanted, but was having difficulty making them work together as they should, so, having heard the original mix, I asked Joe to send me the audio files and the Cubase project.
On a first listen, I could hear plenty to commend the song: there was a strong, characterful vocal performance of a good, hooky melody, and several interesting guitar parts interplayed nicely during the chorus. Joe had evidently spent time picking the right sort of timbres, and the basic feel that he was going for came across quite effectively.
Sonically, though, it felt rather lacklustre and the arrangement was too repetitive, particularly when it came to the drums. It sounded as if a drum library had been used and, indeed, Joe confirmed he'd used Toontrack's EZ Drummer, which is a decent‑sounding library, but the parts themselves needed a bit more variety. The choruses and outro really needed an injection of life, too, and though they needed to be prominent, the vocals were too high in the mix. I also thought the track could benefit from some new elements: a guitar solo, perhaps, or a string arrangement.
I fed back some of these thoughts to Joe, and was pleased to find that he was in broad agreement. Usefully, he also gave me a bit more background on the sort of sound and feel he was trying to achieve. Eventually, we managed to boil down the references he gave me to some specific artists and tracks, and he told me that the song had been inspired by his relationship with his late father — so there were some strong and complex feelings that needed to be conveyed sensitively. This gave me something to aim at, and even if we ended up taking the mix in a slightly different direction, we had an agreed starting point.
My first step was to break the monotony of the drum part, to make it feel more natural and to give more contrast to the different sections of the song. I quite liked the sounds Joe had used, but didn't have the same EZ Drummer kit, so rather than reprogramming, I decided to try layering some tambourine and shaker parts over the drums. Laziness led me to my sample library: I imported a few suitable loops into new tracks in Cubase. They were 5-6bpm slower than the project tempo, so I used Cubase's time‑stretch tool to make them fit, and found the few artifacts acceptable.
I left these new parts out of the first stages of the song, allowing some room to build the rhythm as the song progressed. The busy shaker part served to speed things during the second and third choruses, and the outro, whereas I used the tambourine a little more sparsely, doubling the snare drum. This seemed to work really well, particularly when I later sent the tambourine to the snare reverb.
I also got busy with the scissor and mute tools on the drums, to create a breakdown after the second verse — and I'd later mute some of the other elements of the mix to make the break more effective.
I wanted to create a catchy, almost bouncy groove, so I'd need to tighten up the timing of the bass, and really beef it up sonically. I'd hoped I could use the existing part with a bit of editing and processing, but as I worked my way through an arsenal of processors, I realised the song would be better served by re‑tracking the bass. Fortunately, Joe has some capable bassist friends, so he was able to send through a new part, which worked much better: the player seemed to be digging more assertively into the strings, and there were a few nice variations. The sound itself was fuller, too, and all I needed to do was copy some loops to replace a couple of short, slightly weaker sections, remove some of the more adventurous fills, and tweak the timing of the odd note here and there.
It's a common mistake to start a mix with the kick, working to get the 'ultimate' kick sound, before tackling the bass, then the snare, and so on. That approach may work for composition, but you need some context to make EQ and dynamics decisions. The lead vocal, not the drums, clearly needed to star in this show, so I started by soloing the drums, bass and lead vocal, and soon added to that the rhythm guitar in the verses. (Only when I was confident that these elements were working together did I start introducing other elements.) Most of the kit parts didn't require much processing, but the kick and snare sounds both needed a little work to sit with the new bass. On the kick, I used the UAD Neve 1073 EQ, simply dialling up a 'deep and snappy kick' preset as a starting point, and tweaking to taste. I then ran it into a limiter, just to nail down any headroom‑hogging transients.
I wanted to give more weight and snap to the snare, but the frequencies just didn't seem to be there to work with. I decided that the quickest way to get the sound where I wanted would be to blend in another snare, so I inserted an instance of WavemachineLabs Drumagog 5, set the trigger threshold, and selected a nice beefy rock snare. Drumagog's pitch‑shift control allowed me to match the pitch to the original snare, and I ended up blending around 11 percent of the new sound in to reinforce the original snare, without being too obtrusive. I moulded the attack and sustain of the resulting sound using SPL's Transient Designer plug‑in, before adding a bit more bite using UA's Harrison EQ. I also boosted quite a bit in the lower mids — but that followed later, because the snare became a bit submerged as the mix evolved. I also set up a dedicated send reverb for the snare, using TC Electronic's Megaverb.
I routed all the drums and percussion to a Group channel and whacked a tape emulation plug‑in in the first insert, to round everything off a bit. Two of my favourite plug‑ins for this are Bootsie's Ferric and Jeroen Breebaart's Ferox: Ferric is arguably a better tape emulation, but Ferox offers a much warmer, rounded sound at the bottom end, which can sound fabulous in its own right on drums and bass. Unusually, I ended up using both: I placed Ferric first in the chain, in a bus-compressor-ish role, but used Ferox to give the warm bottom end I wanted, and to smooth out the highs a little. The only other processing I added, after receiving feedback on a draft mix, was via DDMF's almost‑freeware (you pay what you want!) LP10 linear‑phase EQ, boosting at the bottom end to pick out the kick a bit more, and slightly boosting around 2.2kHz, adding a little air back in at the top end.
First in the chain for the bass was Bootsie's Density, to tame and thicken the part: again I 'cheated', using the 'bass pin‑down' preset as a shortcut, and tweaking the threshold and drive settings, the result being a slightly richer sound and about 6dB of gain reduction.
I wanted the bass to sound quite deep and large, but low frequencies have a lot of energy and can really eat into your mix headroom, so rather than EQ'ing and boosting the level, I used Waves Renaissance Bass (beware of overdoing it with this!) to add in some harmonics. Finally, I used Cubase's compressor plug‑in in a post‑fader insert slot, and sent to its side‑chain input from the kick drum, so that the bass ducked when the kick drum sounded. You want this effect to be barely audible, so short attack and release times work well, and you're only looking for a few dB of gain reduction. Also, remember to switch off any make‑up gain, or you'll make the ducked part sound louder!
I like to work with a compressor on the master bus, just to glue things together, and add a bit of character to the mix. With the drum and bass now working well, and enough other parts playing to suggest a rough mix balance, I tried some different compressors, including some modern tools, as well as emulations of classic hardware, before settling for the UA Fairchild, more for the pleasing sound of the modelled circuitry than for the compression itself. There's no point in leaving bus‑compressor decisions until the end of the mix, but I find it equally distracting if you put it on when working on the first few elements.
I quite liked the lead vocal sound Joe had achieved and wanted something similar. But although there didn't appear to be much to do, I actually had to work quite hard. There was some 'essing' that needed taming, particularly when I started to boost the higher frequencies to bring out some air — so much so that Joe and others suggested I take the de‑essing further after the first draft mix. There was also a section of the recording where Joe had evidently stood back from the mic when singing louder, the result being an undesirable change in the room sound.
The first job was to get busy with the mouse, separating the odd phrase and syllable here and there, just to tighten the timing in places, and to raise or lower the level of certain few words and syllables for clairty. It's better to do this on the events themselves, rather than with level automation, as you need to get the part right before any processing.
With these edits done, I started on the corrective processing. I fired up Cubase's in‑built SPL De‑esser, which has one of the simplest interfaces of any de‑esser, but still seems to work rather well: I set it to 'Male' and played with the S‑reduction control until things sounded right. A little 'essing' remained in places, and there were a few occasional resonances that needed tackling, so I used TC Electronic's Dynamic EQ to solve both problems. Rather like a compressor or gate, this EQ is only triggered when the level exceeds a threshold, but as you have control over the frequency, 'Q' setting, number of bands and degree of boost/cut, it's a very precise tool. The final bit of corrective work, was to remove a little hiss using the gate on Waves' Renaissance Vox processor, before using the same plug-in's compressor to give 6-10dB of gain reduction, which seemed to pin the vocal pretty much in the place I wanted it.
Now for some tonal tweakage. I used Focusrite Liquid Mix's 'Viking 1' compressor, again as much for the sound as for the compression, and its 'Tranny 1' EQ to apply a gentle, broad boost centred around 200Hz, to reinforce the body of the voice. More boosts centred around 5kHz and 7kHz brought out the breath and sparkle (it was safe to do that with the de-essing in place!). It's worth mentioning that I didn't arrive at these settings straight away: although I got close to these early in the mix, I performed several tweaks as the mix progressed.
The rhythm part in the verse was important to the groove, but felt a bit sluggish and needed more bite, which wasn't helped by the fact that it had been recorded 'wet', with reverb. I used Waves' Renaissance EQ to roll off the low frequencies, with a high‑pass filter turning over at just under 400Hz (there was plenty else filling out those frequencies in the mix). An SPL Transient Designer plug‑in allowed me to bring out the guitar's attack, while pulling back some of the reverb. I then used Waves Renaissance Axx to reduce the dynamic range, with a hefty 12dB of gain reduction, but a lazy 4.2ms attack allowing the transients through.
That gave me the basic 'performance', but I wanted to make it sound more 'vintage', so I ran it through Softube's excellent Vintage Amp Room plug‑in, which allows you to move the mic around and tweak the amp settings to achieve the desired tone, and this gave a bit more attitude without swamping things. To emphasise the vintage sound still further, I then used another instance of Ferox, mainly to lend a bit more thickness to the sound, using a 20 percent hysteresis setting. I later used the plug‑in's high‑cut filter set right down at 2.3kHz to push the sound back in the mix, making space for the vocals and slide guitar.
Most of the sound of the slide and the harmonica was achieved by blending and automating various send reverbs and delays, with the delay tails, in particular, helping to add much more width and depth to the part. On the slide, after deliberately dulling it down a little (using Ferox yet again), I sent it to three separate reverbs, as well as to the PSP84 delay that I'd set up earlier for the harmonica. The only other processing on this part was a side‑chain compressor fed by the vocal (to keep the slide level up without clashing with the vocal), and a little EQ, courtesy of Cubase's built-in channel EQ, just to roll off the bottom end from about 300Hz, and to pull the brightness down around 2.5kHz. Finally, a little level automation brought out some details at the end of each slide riff.
In the second verse, a staccato rhythm part came in. It worked well, but I didn't want it to draw too much attention to itself, so again I used send effects to blend the part into the 'mix space'. I sent it to the three main reverbs, the levels of which were static, but automated the send to the PSP84, so that the sound was reasonably dry while vocals were present, but the delay kicked in at the end of the section.
Two guitar parts in the chorus and outro worked nicely together, with a chiming, picked melody weaving itself around slow‑strummed chords. With the new bass part and the more confident snare/tambourine sound, I was able to allow the guitars to 'float' a bit more than they had originally. I panned the melody hard left and the rhythm hard right, just to give a bit of width, but wanted the two parts to interplay across the stereo field, so I sent a bit of each to the PSP84, opposition panning the sends so that the part on the left hit the delay on the right, and vice versa: success!
I wasn't as keen on the sounds as the parts themselves, and ran both chorus guitars through amp simulators, which tends to be more effective than over-massaging with EQ. For the rhythm part, I applied some gentle compression with long attack and release times and a ratio of about 3:1, before running it through IK Multimedia's Amplitube Jimi Hendrix plug‑in. I find the presets on that plug-in really usable, and dialing up the 'Axis: Bold As Love Rhythm 1' preset got me 90 percent of the way there — with tweaks to the gain and output level, and a bracket high‑ and low‑pass filter at 150Hz and 4kHz respectively completing the job.
For the melody part, I really wanted to emphasise the bell‑like chime, by bouncing it across the stereo field, so I introduced another delay send effect (Cubase's Stereo Delay), ending up with a delay time in the left channel of about 42ms, and 150 in the right. As the mix filled out, this melodic part became increasingly suffocated: the bright 'chime' shone through, but it began to sound rather thin. I tried EQ'ing, and running the part through different amp‑emulations, and while that rewarded me with a ballsier sound, it was at the expense of losing the 'bells'.
I decided, therefore, to leave the original part in place, and duplicate it on another track to reinforce it that way, as I'd done with the snare. I panned the new track a little more centrally, and ran it through a compressor, for more solidity. Vintage Amp Room then gave me the solution I'd been looking for: I dialled up the 'Brown Jazz' preset, and tweaked the mic position until the sound fell into place. A high‑pass filter and a generous boost just above 2kHz got it sitting nicely in the mix, then I just balanced the two parts with fader settings and a little level automation, before bussing them together so that their combined level could be automated in the mix.
The acoustic guitar recordings were a little noisy, but both the body and the transients had been well enough captured. There was plenty else going on in the intro, though, so I opted to use only the higher frequencies of the acoustic guitar. I used TC Electronics's Denoise plug‑in to remove the unwanted noise, and panned the two parts to different sides, though not quite so hard panned as the electrics. Other than a touch of very gentle compression on the rhythm part, the processing was identical for the two acoustics: a low‑cut filter rolled off the frequencies below 150-300Hz, and a notch EQ removed an unwanted resonance or two. I then added a bit of sparkle using UA's Precision Enhancer, to generate some ear-candy harmonics.
The last plug‑in I used in both chains was (yet again) the SPL Transient Designer: I reduced the level of the attack on both parts, as the enhancer had really emphasised that area, and extended the sustain portion of the rhythm part, helping it fit better with the overall groove. Again, to push these parts away from centre‑stage (where the harmonica needed to be) I sent healthy levels of both to the room and hall send reverbs, using a similar send‑panning strategy as with the chorus electric guitars.
I felt that the backing vocals should lift the track in places, as would the string part that I planned to program. I had a lot of difficulty getting these vocals to work, and Joe even tracked some new parts, but I continued to believe that the originals could be made to work, even if it meant some heavy editing and effects to disguise the sound.
I went through each of the layered parts, manually editing each phrase, both on the grid and in Cubase's Variaudio, to lengthen some of the notes, and to make sure first syllables, consonants and endings were aligned. That removed any jarring phrases, and stopped the combined part shifting around the stereo field. I then pushed things back by sending the backing-vocal group channel in varying amounts to to the five different reverb sends I'd set up during the course of the mix (see the 'Return To Sender' box for more on this).
I programmed a violin part in an early version of Garritan Personal Orchestra, which I happened to have on my machine. I kept the strings for the outro, leaving space for the backing vocals to build during the second and third choruses. The melody worked well, but the sound needed shaping a bit, and I though it could benefit from a lusher arrangement featuring cello, viola and bass. That's not a strength of mine, so I asked SOS contributor John Walden if he'd do the honours, and he kindly sent a few parts back. They worked well enough on their own, but unfortunately made the outro feel rather muddy, so I ended up using the original melody but with some of John's viola and cello brought in at a low level to reinforce it: even then, I had to use an aggressive low‑shelf EQ to avoid the mud.
The strings needed to cut through the mix without pushing anything else back, and I turned to UA's Harrison EQ for this job, cutting a couple of dB around 300Hz, and boosting by 4‑5dB around 5kHz. That gave the strings enough presence that I could afford to pull back the high‑end a bit, to make things sound less brittle... before drenching everything once again with the send reverbs!
I left refinements to the song's outro until I had the strings programmed. I'd held back on the levels of the other parts to give me a bit of room to 'big this up', but with so many elements competing for space, it was a challenge to get everything heard. To be honest, you don't need to hear every part all of the time: the trick is to bring things forward for just enough time to catch the ear, before something else takes up the baton. Side‑chain compression can be useful for this: for example, I slapped another compressor on the strings and fed its side-chain input from the backing-vocal group track, so that the strings were up in the mix, but allowing the vocals through. The other key tool is automation, and lots of it: it can be a painstaking process, but paying sufficient attention to level and send‑level automation is still the only way to get the best results.
I sent a near‑final mix to Colm O'Rourke, a mastering engineer friend of mine, who tweaked the final mix for me. Colm: "Some control of dynamics was needed in the mid‑frequency area. I also felt the track needed a more contemporary 'in your face' sound, particularly in relation to the vocals. I thought the bass could use a bit more definition, and that the track would benefit from a little width and loudness enhancement. I'd recently mastered some albums in a similar genre, so I called on those for referencing.
"I started by checking the track with a number of different‑sounding limiters, and decided to use Waves' L2 Ultramaximizer: I lowered the threshold until it started to limit by about 3‑4dB at the peaks, though I knew I'd need to fine‑tune this later.
"Next, I tackled the mid‑range with Waves' C4 Multiband Compressor. I switched in a single band at a crossover setting between 200Hz and 1.930KHz, set the range to ‑4.1, the threshold to ‑18.6, the attack to 7.29, and the release to 25, which seemed to bring the mids nicely back under control. I then turned to the Massenburg DesignWorks EQ for a little boosting in the lows and highs, because this EQ has a character that I felt would work well here. I used a low shelf at 60.1Hz with a boost of 0.6dB, a low peak of 63.7Hz, a boost of 6.6dB and Q of 4.5, and a 1dB high boost at 3.45kHz with a Q of 1.5. Waves' S1 Stereo Imager increased the width a bit (to a setting of 1.04), but the stereo guitars were getting a little lost as a result, so I used the Brainworx BX Digital EQ to rescue them, with a slight boost of 0.6dB at 3.8kHz. I also increased the stereo output of this EQ by 0.8dB.
"As a final touch before dithering, I thought the track would benefit from a little psychoacoustic enhancement, and used the Aphex Aural Exciter Type III for this, with a 'Tune' setting of 4021Hz, tweaking the other controls until I found the sound I wanted, but only mixed in 11 percent of this.”
Send effects were pivotal in getting this mix to hang together, and I ended up with seven different effects channels on this mix: five reverbs and two delays. I didn't consciously set out to use all of these effects, though. I left in the Alitverb patch that Joe had created, and I deliberately set up both room and hall reverbs, using Lexicon's native reverb plug‑in, and a vocal plate using the UAD Plate 140, but the rest were almost happy accidents. I set up the TC Electronic Megaverb as a dedicated snare reverb, the PSP84 for the harmonica and slide, and the Cubase Stereo Delay specifically for some of the guitar parts — but I ended up sending differing amounts of many tracks in the mix to these as well.
The one exception was the lead vocal: in the initial mix I sent for feedback from both Joe and Colm (see the 'Mastering' box), the lead vocal was much wetter. After having problems with sibilance and intelligibility, I decided to try muting the sends on that so that it cut through better, and quite frankly forgot to un‑mute everything... but I rather liked the drier sound, which felt a bit more contemporary. The change was initially a bit of shock for Joe, but after we both sought impressions from some trusted ears, he seemed happy with it — and I think it was probably the right decision.
You can find audio files, including before and after versions of the full mix, and some of the individual elements, on the SOS web site: /sos/sep10/articles/mixrescueaudio.htm.
Joe Murphy: "The early changes in the mix and arrangement helped me decide in which direction I did and didn't want to take the song. Thanks to my friends Rob and Al for sorting out the bass, and of course to Matt and Colm for all their time, effort and patience in all aspects of shaping the song. The Mix Rescue experience has given me a better understanding of how to create and improve a song, and turned an idea into a finished song. I've really enjoyed and appreciated this opportunity.”