We tackle some thorny noise problems, and add a couple of extra parts to breathe new life into this month's track.
Daniel Thompson had been writing and recording songs for years without getting deeply engaged in mixing, for which he'd relied on the efforts of bandmates or professional engineers. The final product had never quite delivered what he sought, though, so he decided it was time to learn to do it himself and emailed SOS to ask for some pointers. Eventually, we hit on the idea of a Mix Rescue: I'd find out what he wanted a specific track to sound like, and would do my best to create a mix that delivered what he wanted.
The track in question is a largely acoustic and somewhat wistful number called 'Every Dream'. Daniel had played and recorded this "with decent mics in one guy's basement", along with his Connecticut-based band MARMO, who've been together just over a year. It comprised lead and backing vocals, acoustic guitar, lap-steel guitar and harmonium. He felt the original mix sounded "like it's just got a lot of busy stuff going on, which makes it feel like you're not sure what to listen to". And when I asked what he wanted the mix to deliver, Daniel responded that "a good record should have a sense of space. And this particular one might give a sense of listening to us in a small club or bar or honky-tonk. Guitar player in front singing his heart out, harmonium to one side, lap–steel on the other. Big backing vocal 'oohs' on the chorus." Importantly, he also said that, "you can do anything you want with these tracks including muting, adding tracks, changing things, whatever you think will make it sound good. But I'd like to get an understanding of your approach and reasoning, because fundamentally I want to get good at this over time."
With that in mind, I auditioned the original mix. A few things struck me immediately. There was no dedicated bass or percussion; the acoustic guitar combined with the harmonium to provide the main rhythmic interest, and the same guitar provided the nearest thing the song had to a bass line. The mid-range was rather congested and, presumably because of this, the vocals and guitar levels were high in the mix. There was also precious little stereo width generally, and it all added up to the whole sound stage just feeling constrained. I liked the song, though, and the performances themselves seemed decent enough.
Having brought the multitracks into Reaper (I chose this DAW since Daniel uses it), I created a couple of group busses to manage the pairs of guitar and backing vocal tracks, and set out to establish a rough 'faders–up' mix, listening all the while for any noises that might need removing. Unfortunately, I could hear several problems, including hiss and rustling on the vocal and guitar recordings and, most notably, a loud hum/buzz that came in part way through the lap-steel part, and decided to attend to this before moving on.
I started with edits, chopping away obvious knocks and bangs and paying close attention to the sound of the fades. I also made extensive use of Reaper's ReaFIR plug-in as a de–noising tool — to use this for noise-removal, you select its Subtract mode and would normally play it some exposed noise to 'learn' and remove. But the files must have been rendered edits, as I had no exposed noise available! Persevering, I drew some noise profiles in manually and, for the vocals and acoustic guitar, I achieved a worthwhile improvement. A few audible edits when fading the exposed count-in were disguised with some more controlled noise sources, courtesy of iZotope's free Vinyl plug-in and a noise generator, both mixed in very low.
The lap–steel's broadband buzz seemed an intractable problem. Not only was it really audible, but it varied in level, becoming particularly bad towards the song's end. I spent about an hour frustratedly tweaking various noise-removal tools before I hit upon my get-out-of-jail trick: I dialled in fairly aggressive static settings for ReaFIR (set to Subtract) and a low-pass filter, and together, these addressed the worst noise. I then automated the filter frequency and ReaFIR's wet/dry mix (Reaper has a mix control per insert slot) to back the processing off where the noise wasn't problematic or the processing dulled the sound too much. Finally, I fed the part into an amp simulator plug-in, which made the changes in tonality less obvious. This painstaking approach allowed me to fashion a sound that was, while in no sense perfect, at least workable. You can still hear buzzing, particularly from about 2:45, but I felt the part was critical to the emotion of the song, making this the lesser of two evils.
With the noises addressed, I returned to building my 'faders–up' mix. I'll typically chop the main tracks into the different song sections and use clip gains, mutes and/or pre-insert gain automation to create a ballpark balance for each section — usually clip gain, as I prefer to see my changes on the arrange page. From there, I'll set up a master-bus EQ, a strategy I've adopted over the last couple of years. Some people dislike 'top-down' EQ, but I like it, the main benefit being how rapidly you can get the overall mix tonality in the right ballpark. These broad-brush EQ moves won't suit every source, but the more we hear something, the more our brains will accept it as being 'normal', and working this way keeps my ears fresh during these critical early stages of a mix. If level and processing judgements for each source are made in the context of a balance that's already broadly right, you're in a far better position to make the right calls!
You could use any EQ, but I used FabFilter's Pro-Q 3 because I find it particularly easy to work with. I like that you can set any EQ node to operate on the full stereo signal or exclusively on the left, right, Mid or Sides channels, and that it has a tilt EQ facility. I applied a nearly 5dB high-shelf boost to the stereo signal just above 11kHz, with a broad enough bandwidth that it started to lift things from about 2.5kHz (I would add even more high end later!), and a 17kHz 12dB/octave low-pass filter countered the shelf's effects at the very high end. A 55Hz broad Mid-only bell boost helped to bring out some fullness in the acoustic guitar, and a steep–ish (36dB/octave) 40Hz high-pass filter prevented that doing anything harmful lower down.
I can't recall quite why, but I then found myself experimenting with another EQ, Acustica Audio's Purple P1. This Pultec EQP-1A emulation adds subtle harmonic distortion as well as EQ, so I was probably lazily searching for some 'analogue colour', but inevitably I also ended up tweaking the EQ controls. The EQP-1A's controls encourage you to boost and cut at the same frequency but with different bandwidths, and I added just a touch more boost than cut at 60Hz and 3kHz. I'd add a little more bus processing later (see box), but for now I was content.
My early panning decisions were largely of the LCR variety, with each source panned hard left, hard right or dead centre. I'd make finer adjustments later, but for now my aim was to carve out space for the vocal, and establish a nice, wide and balanced stereo image. The lap steel went hard left and the harmonium hard right. The two backing vocal tracks were opposite-panned hard left and right — I'd later use the fader and stereo width control of the backing vocals' group bus to tuck them into place, and nudge the lap steel in a touch.
The acoustic guitar posed a tricky question. Should it be panned opposite the lap-steel, or the harmonium? Or did it want to remain in the centre, since it provided all the bottom end? Should I perhaps try adding a bass part, to give me more freedom in placing the guitar? Maybe... but I decided I'd at least see what could be done with what I had. I started with this guitar a tiny bit to the right of the vocal, but this didn't sound quite right. Inspiration came in the form of Melda's freeware MSpectralPan, with which you can draw a curve to map the signal's frequencies across the stereo field. I left the low end in the centre, but panned the guitar from about 200Hz up increasingly to the right. It worked nicely... but I'll return to that additional bass idea later!
But first, I wanted to refine that stereo balance using some send effects, and to work on the lead vocal, since I planned to build the rest of the mix around that. I almost always use multiple reverbs/delays, each with a different character and decay, and usually bracketed with high- and low-pass filters. The high-pass filtering prevents the reverb turning the low mids to mud, and the low-pass allows brighter dry sources to remain fairly up-front while still benefiting from a sense of space. A short (13.8ms) delay with a hint of feedback stood in for early reflections, and a medium-room instance of Reaper's ReaVerbate reverb was bracket filtered at 345Hz and 5.8kHz, and slightly narrowed in the panorama to exploit pockets of space between the LCR extremes. Two FabFilter Pro-R reverb instances were bandwidth-restricted fairly heavily, each providing a different character: one was slightly darker (yet fairly clean-sounding in terms of modulation and distortion) with a long decay and the other brighter and shorter, and with a touch of chorus.
There's little point listing my send-effects decisions for each source, since these vary for each mix and, in any case, seem largely to be borne of experimentation! But I do tend to adhere to a few principles. One is to make sure the whole mix is playing when I set up, send to and tweak my reverbs. They often sound dreadful when a single source is soloed, but all that matters is how they help you shape, spread and place elements in the mix. Another is to set them up fairly early on, so that they're there available when needed, and setting them up doesn't become a distraction.
An important consideration is where I want to place each source's reverb in the stereo image. For example, I ended up with the dry lap-steel track most of the way (76 percent) to the left. It needed to be there to balance what was happening on the other side. But I also wanted to give it some distance and space, which I felt would be in keeping with the emotion of the song. To do this, I sent it (via Reaper's routing window) only to channel 2 of the reverb tracks, so the dry sound was on the left, with the reverb (which included a little pre-delay) opposite. It's a trick that spreads a sound across the stereo field, but your ears still locate the source as being where the dry signal is panned. Similarly, I sent the lead vocal to some (not all) of its effects only in the left channel. The idea wasn't to spread the vocal to the left, but to give it a sense of space — and there was room on that side of the mix to accommodate that, whereas the right side of the mix was a little busier.
The lead vocal had a character that befitted the song, and the performance was nice. Other than tackling the aforementioned noise issues, processing was mostly about taming the essing and some boxy resonances, and then bringing up lower-level details (using compression) to improve lyrical intelligibility without having to mix the vocal too high. I addressed the sibilance and resonances with a conventional de-esser (I spent some time refining the trigger frequency and threshold) and a combination of conventional and dynamic EQ. I used FabFilter Pro-Q 3 again, since it allowed me to do several jobs in a single plug-in, but there are more affordable dynamic EQs if you need one (for instance, the free version of Tokyo Dawn's Nova).
A tilt EQ node centred on 1.5kHz pushed up the highs a touch and brought down the low end, and high-pass (100Hz) and low-pass (11kHz) filters cleaned out some rumbles, knocks and residual hiss. Three or four narrow dips nixed some ugly resonances, but as some of these caused collateral damage I made those nodes dynamic — the dynamic EQ nodes knock as much as 10dB off at times, but only when the problem frequencies are, um... a problem.
To exert some control over the vocal's dynamic range, and maybe add a little analogue je ne sais quoi, I looked to Acustica Audio's Ceil. This plug-in emulates a Focusrite channel strip, and I'd recently acquired it for free (a time-limited introductory offer) so was keen to try it out. I dialled in a medium attack and a medium-slow release on Ceil's compressor, to clamp down on wayward sustained vowel levels but reset before the next syllable. I started with a 4:1 ratio and set the threshold so that gain reduction occurred some but not all of the time, and tweaked by ear from there, ending up with about 7-8 dB of gain reduction at most, but rather less most of the time. Rather than apply broadband gain to compensate for the loss in level, I used Ceil's high and low shelves, and settled on small boosts at 160Hz and 10kHz. Another compressor (Reaper's ReaComp) followed, set to attack and release a little faster and with a slightly more assertive ratio. This attacked some stray peaks with, again, up to about 7dB of gain reduction... and suspecting I'd overcooked this slightly, I simply blended some of the dry signal back in using the plug-in's wet/dry faders. Note that these two compressors' settings meant that they were acting at different times, rather than heaping on the gain reduction!
That whole question of the acoustic guitar and the mooted bass part was bugging me, so I turned my attention to the guitar. I wanted to make the levels a little more consistent, but also beef up its rhythmic and low-end contributions. I reached again for Ceil — to put my new toy through its paces more than for any other reason — and a low-shelf boost (270Hz) brought out the guitar's body a bit more. The high-pass filter (60Hz) removed the few knocks and bumps that this boost had highlighted, and a 10kHz low-pass filter reined in some finger squeaks and fret rattle the master-bus EQ had similarly magnified. I set a 4:1 ratio on the (pre-EQ) compressor, and a fast-ish attack and release meant it acted only on the louder high-frequency parts.
From there it was a case of painting in layers. An instance of ReaEQ provided broad, gentle dips at around 200Hz and 4.5kHz, where the guitar seemed to be fighting other things. Then another Pro-Q 3 instance was used for lots of dynamic EQ cuts, to get some booming notes and that fret/string rattle under firmer control — there were seven different dynamic EQ notches by the time I'd finished. It still rattles and squeaks a bit, but that's really an instrument/performance thing, and I decided that it all just served to make the part feel human.
This was better still, but still not good enough, so I finally decided to try adding that bass part. Having recently moved to a new computer and not yet installed all my plug-ins, I was curious to see what freeware bass instruments might be available, and discovered AmpleSound's Ample Bass P Lite II, a VST Instrument that plays samples of a Fender Precision bass. And I really like it! Wanting this to sound unobstrusive, almost like an extra bottom string of the guitar, I lined up the MIDI notes for a simple melody first by eye (using Reaper's spectrogram view of the guitar as a reference), and then by ear, as I nudged individual notes into place. I kept the notes fairly short, except for a couple that definitely needed to ring out, and gently rolled off the bass part's top end at around 1.25kHz to dull the note attacks and leave it sounding softer and warmer. An instance of Melda MSaturator thickened the sound a bit, and it it all worked a treat!
In extending the 'height' of the virtual sound stage, everything else just felt a bit more anchored in place. It made the guitar sound fuller too, and, emboldened by this success, I tried another stealth trick to beef it up further... Early in the song, the performance included some almost drum–like hits on the acoustic guitar's body. I rendered a clip of one hit, complete with processing and effects, and brought it into Reaper's ReaSamplOmatic 5000 sampler. I then triggered this from MIDI to subtly reinforce the rhythm throughout the song. Again, I was really pleased with how well this worked!
Hopefully I've been able to give a sense not only of how I tackle problems, but also how I typically approach a mix: painting with broad strokes early on, to create a meaningful context for the more detailed decisions I'll make further down the line; and not being scared to make big decisions like layering new parts in if I feel it will improve the end result. I've obviously not had space to cover every aspect of this mix. In particular I've not focused much on the backing vocals, despite their important role in this song. But other than the gentle use of a tuning plug-in on each part, and a touch of multiband compression on the backing vocal bus, just to gel the voices together, I did little to them that I've not decribed for other sources.
Reading back over what I've written, it strikes me that you could form the impression that mixing requires considerable time sorting out problems in your recordings, but that's really not true — if the recordings require attention and you don't have the option of re-recording, then yes, you need to do this if you want a good result, and it helps to get that stuff out of the way at the outset, but really, your recordings shouldn't demand that of you in the first place! Listen out for such issues when recording and take action then, and the mixing will give you far fewer headaches and you'll enjoy much more satisfaction — whether you're doing the mixing or farming it out to someone else.
Anyway, it seems my approach worked in this case — I'll leave you with Daniel's reaction to hearing my remix: "Man, this sounds fantastic. Great changes. Thank you so much for the help and inspiration!"
I mentioned in the main text how and why I used master-bus EQ early in this mix, but that wasn't the only master-bus processing I used. I mixed into a bus compressor from quite an early stage, opting for Acustica Audio's Aquamarine 3, which is an emulation of the analogue Shadow Hills Mastering Compressor. This was as much about adding subtle analogue-style thickening distortion as about compression. Having experimented with the different coloration options I settled on the plug-in's 'Bronze' setting. This plug-in actually includes two compressors, an optical one and a VCA one. I used them to deliver tiny amounts of gain reduction — only half to one dB from each, and where I found the compressor was working too hard, I identified the source that was responsible and tweaked its EQ, level or dynamics processing accordingly.
I used yet another EQ on the bus — my third in total. That's atypical, but adding in those bass and percussive sounds skewed the overall balance a tad, and EQing the bus offered an immediate fix. Rather than risk unpicking what I'd done earlier, I chose to use a fresh instance of Pro‑Q 3. The most notable moves were a broad, Sides-only bell boost at around 4.7kHz, which was mostly about balancing the acoustic and lap-steel guitars against the new bottom end, a 14kHz low-pass filter to counter some of the extreme HF lift that boost entailed, and a broad LF bell boost of a couple of dB which seemed to help the bass and low end of the guitar work together.
Finally, right at the end, I deployed Tokyo Dawn's Limiter 6, which features a number of different dynamics processors, as a sort of pseudo-mastering process. I'm certainly not a mastering engineer, but I did want to be able to deliver a 'finished' product, and I used the compressor and limiter to apply a light-touch squeeze and raise the levels. It wasn't an obscene amount: the limiter applied up to about 1.5dB of gain reduction on the loudest peaks, and left me with a 'master' with an integrated loudness somewhere between -13 and -14 LUFS, with a short-term loudness no higher than about -9.5LUFS, and True Peaks reading below -1dBFS.
You can find a number of audio examples, including both the original mix and Matt's remix, on this web page:
Alternatively, you can download a ZIP file of hi–res WAV audio examples in the righthand Media sidebar or use the link below.