We get weeding this month, to thin out a busy mix without sacrificing the big sound.
The song 'Perfect Day', recorded entirely by Hamlet Sweeney, came to us as a mammoth session of almost 60 tracks. However, a bit of judicious pruning and the use of a mono submix of the four handclap tracks (created from around 20 handclap overdubs, I'm told) helped whittle the project down to a rather more manageable size. The song sounded, in many ways, quite conventional, but I could also see the potential for it to be a very contemporary-sounding pop track.
There were some problems evident in the vocal recordings (both the main lines and the multitracked harmonies), which sounded very coloured and lacking in presence. I thought this was possibly due to an under-treated recording environment, and Hamlet later explained that he'd recorded in an untreated room, but improvised a screen behind the microphone using a duvet. If you only have one duvet available, better results can be achieved by putting the duvet behind the singer, as that's where most of the problematic reflections come from when you're using a cardioid mic, but if you also put another duvet or something like a Reflexion Filter behind the microphone, this will tighten things up even more. Another cause of dullness is singing too close to a cardioid mic, which can allow the bass boost that is caused by the proximity effect to dominate. It really is best to get this right at the recording stage, as if room coloration does get into the mic, compressing the vocals later only exacerbates the problem.
The song includes several multitracked, picked guitar parts, and on the original mix the picked arpeggios sounded a little wooden and stilted (which is a common problem when one person layers all the parts). Nevertheless, with the exception of the slightly chesty-sounding vocals, the tracks sounded pretty well recorded when heard in isolation. That included the acoustic drums, which came from the EZ Drummer sample-based plug-in, the audio files from which had been separated into close mics, overheads and room mics. The room ambience track seemed to be adding nothing useful, so I discarded it, along with a second acoustic guitar part that seemed just to be doubling the original. However, I did use the last chord from the otherwise unused acoustic guitar part, as there was a little noise during the decay of the original, possibly due to moving the instrument too soon at the end of the song.
In addition to the usual guitar, bass and drums line-up, there was also an organ part that came in towards the end of the song, and a handful of layered string parts. Additional percussion was also present in the form of claps, tambourine, shaker and triangle, and multiple electric guitar parts and layers were deployed throughout the arrangement. In isolation the organ sounded rather bland, as the same rotary effect was applied all the way through. Had this been more prominent, I would have suggested using the rotary speed control during the performance, but as it was, I used Logic's Guitar Amp Pro modelling plug-in to dirty it up a bit, then applied some very fierce low cut EQ to thin it right out, which kept it from muddying the bass and drums. I added some distortion and echo to the handclaps to sharpen them up a little, then mixed them so they sat nicely alongside the snare drum.
The only way to get a handle on a mix with so many separate parts is to create subgroups, which enables you to control groups of sounds via a single fader and, where necessary, to process them via a single chain of plug-ins. The obvious subgroup to set up first is one for drums, which I did, but I also created groups for all the strings, all the harmony vocals, the various tracks holding different sections of the lead vocal, the picked guitars and the percussion. This enabled me to control most of the final mix using around 12 faders.
Rescued This Month
Here's what composer and musician Hamlet Sweeney said about the recording of his track:
"I recorded on my Dell PC running Cubase SX in the spare room of my home, which is without any acoustic treatment. It's pretty boxy-sounding and, to be honest, it sounds pretty damn bad. I erected an impromptu booth of sorts, with a thick duvet hanging over a few old mic stands, into which I faced when recording vocals and acoustic guitar, using my AKG C3000B (vocals) and Neumann KM140 (acoustic guitar). Getting the vocals done was the trickiest of all, as I find it tough to get into the right head space I need to be in to sing well and also be keeping an eye (or ear) on things like levels and clipping. Unfortunately, the technical end usually suffers a bit, and I often find the vocal might have a touch of clipping. I also don't get to record too late, as the neighbours have threatened to call the police if I do... it's the eternal struggle of all us home-recordists, I guess!
"The organ and strings are from Reason. Everything on this song was recorded, written, sung and played by me, though I'm really looking to start working with others now. I find it extremely rewarding to play everything, but it has its limits, and I miss that chemistry and sense of surprise that comes from recording with others.
"The guitars were all recorded through my Pod XT, and I worked hard to find two or three settings that I then used on everything. The bass was DI'd through my Groove Tubes Brick preamp, while the drums were programmed using EZ Drummer. The project wasn't easy, but I'm happy enough with how it sounds for a home recording. I write a lot of catchy, melodic, 3.5-minute songs and have always fantasised about recording a pop/rock song at home, alone on my basic gear, that would sound good enough to be played alongside the big boys on the radio."
While it is futile spending too much time trying to optimise tracks in isolation (as they will always sound different in the context of the mix), I did spend some time trying to get the sounds into the kind of ball-park I'd normally expect them to be in, especially in the case of the drums. Although EZ Drummer is based on an accurately-recorded set of samples, the kick drum used in this particular kit had what sounded like a hint of an audible pedal bounce following each hit, so I used a gate to kill it and then opened up a parametric EQ, to add more punch to the sound. I could have used any number of different equalisers to do this but I settled on the Logic channel EQ, as I find the clear, graphical picture of the EQ curve useful. I beefed up the usual 80Hz region, but balanced this with some 270Hz cut to stop the lower-mid range from sounding boxy. A 3kHz presence hump brought out the beater-hit definition nicely. I didn't do anything to the hi-hats, other than add minimal amounts of emulated plate reverb, which I set up on a send. Even the snare drum responded nicely to gating (both the top and bottom mics) with a bit of 3.25kHz presence boost added to the top mic to crisp up the sound, plus small amounts of the plate reverb. The overhead mics were EQ'd to roll off the low end below 200Hz and to boost the 'air' region up at 12kHz. Rolling out the low end keeps the necessary definition, while making the overall kit sound less muddy, and when you're working with a real kit this can be an effective strategy if the kit was recorded in a small (or otherwise non-flattering) room. I also inserted the UAD LA2A compressor over the drum submix to add to the punch, but used it sparingly. EZ drummer generally sounds good right out of the box, so the small adjustments I made are probably no more than you might make to any other studio recording of a good kit. It was really down to making the drum sound sit in with the rest of the track.
Do you have a mix that you feel isn't quite hitting the mark? Well, don't be shy: get in touch and you could have your mix revamped by Sound On Sound!
To apply, send an email with an MP3 of your mix to firstname.lastname@example.org, explaining the sort of sound you're trying to achieve, and a bit about how you wrote and recorded the track.
If you're asked to send in the multitrack files, so that they can be used on any system we ask that you provide them as WAV or AIFF audio files at 16-bit or 24-bit, 44.1kHz, and starting at bar one, beat one. Ideally, there should be no more than about 24-30 tracks to the mix, and they should be unprocessed mono files where possible, or interleaved stereo files where not (in the case of drum loops, for example).
I adopted a similar thinning-out strategy for the miked acoustic guitar part, as in a busy track like this all you really need to hear is the high end, and if you can EQ out some of the low end it prevents it from contributing to the all-too-common lower-mid congestion. As shown in the screenshot above, I rolled off the sound gently below 500Hz, with a generous 'air' presence lift of four or five dBs at 16kHz, but using a wide bandwidth setting, so the EQ spills down into the 5-10kHz region too. In the context of the mix, this gives the acoustic guitar a nice zing, without letting it take up too much spectral real-estate. I didn't add any reverb to the acoustic guitar, or to the majority of the electric guitar parts, though I did use delay on the picked arpeggios and on the guitar solo section.
As the electric guitar parts (which were all recorded using a Line 6 Pod XT), sounded pretty good in isolation, the only thing I needed to do was balance them as I saw fit, and, to take away some of the rigid feel on the layered picked arpeggios, I inserted a tape echo plug-in over the group, and set up an echo that fell roughly in time with the picking speed. This actually worked better than I had hoped, making the performance seem less rigid and also more contemporary-sounding. Initially, I also muted the picked parts during the first verse and during the first half of the second verse, to allow the song to build in a more obvious way, but later put the parts back for the entire second verse, as both Hamlet and I thought it sounded more natural that way.
For the bass, I added compression via the TC Tube Tech compressor plug-in I was reviewing at the time, and I also used this to compress the main and backing vocals. I didn't EQ the bass part, as it seemed to sit well in the track as it was, despite having been DI'd. Once I had all the tracks sounding generally OK, I put up a rough balance and then started tweaking the EQ settings.
"When I first heard Paul's mix, I couldn't believe the difference from what I've been getting, and the hairs on the back of my neck stood up when I listened to it on a stereo. I have never had as clear a lesson about the value of a pro mix.
"I think that Paul and I had slightly different ideas of where this song should go. It's hard to write and record a song entirely alone and then hand the whole thing over to another person. You find yourself moving from being your own little 'audio overlord' to suddenly being faced with an opinion from someone who really knows their stuff — and who sometimes disagrees with you!
That said, there's no doubt that it sounds much, much more posh and professional than anything I've done at home. It was a great experience, and I will be using the mix that Paul has made to send to publishers, producers and managers. But I still wouldn't mind a bit more 'balls' next time round!"
For the strings, I again rolled off some of the low end, but also fed the whole string submix through Universal Audio's Roland Dimension D plug-in, which adds a sense of richness to the ensemble without sounding too obviously chorus-like. Even so, the strings, which seemed to be mainly lowish bowed-cello parts, were pretty inaudible unless turned up so loud that they made the mix seem messy, even when I tried rolling off more low end. Ultimately, I decided to mute everything apart from the nice bowed section over the break halfway through the song, until the very end section, where I brought them all back in to help build the dynamics.
Arranging & Mixing
Before going further with the mix, I decided to snip out all the sections of the audio tracks where nothing was playing, as some of these ostensibly silent sections contained low levels of noise or headphone spill. It was also necessary to fade some last notes, especially on the guitar and bass tracks, as some hum and noise was evident during the note decays.
To me, the track also ended rather unnaturally. Hamlet had originally ended with a single bass note and cymbal, but for some reason I couldn't find this on the files I'd received, so initially I looped a section of the picked guitar part, and then allowed this to die away gradually beneath the final vocal line. My looping kludge wouldn't be my ideal ending either but, given the available options, it seemed the tidiest solution at the time. However, when I revisited the song a day later, I still wasn't entirely happy, so I decided to cut off the original 'exposed vocal' ending to give the song more of a positive finish. This also entailed cutting short the harmony vocal, which was longer than the main vocal line. On re-checking the harmony parts, I found an overhang at the end of each chorus that didn't (for me) work musically, so I cut those off too, and used a short fade to stop them finishing abruptly.
When it was time for panning, I did the usual trick of separating any instruments that had been layered playing the same line, and did the same for the backing vocals. The drum kit and bass were kept dead centre, along with the lead vocal, while the percussion parts were panned in a fairly subtle way. A little level automation was needed to keep the vocals riding above the mix as it built in intensity, and I also lifted the level of the guitar solo just before the final vocals.
Initially, I tried cutting the guitar dead as the vocals came back in, so as to stop the vocals getting swamped, but Hamlet liked the way the song continued to build, so instead I restored the guitar and lifted the vocal a bit more, to keep its head above water during the busy last chorus. After several level adjustments this seemed to be working OK, so I left things there.
Compared with the original mix, I had the rhythm section up a little higher and sounding crisper, with the guitars slightly further back in the mix, but still making sure the vocals sat right up front. There were some vocal timing issues, where the main and backing vocals weren't as tight as they could have been, particularly evident on those sibilant 'S' and 'T' sounds, but that is something better fixed by re-doing the parts than by editing, as it takes much less time and sounds far more natural. A common trick employed when a singer is layering vocals is to soften sibilant sounds, and to miss or fade word endings on the overdubs so that the main vocal dictates the timing. I felt that all I could do here was trim off the ends of some harmony parts to get them to sit more tidily.
For this particular mix, after applying some compression, I initially used a de-esser to remove some of the more obvious sibilant sounds from the harmony submix, but that made the vocal sound a touch 'lispy', so I decided against it. The Channel EQ was used to tame the low end and to add a bit more presence. Then I tried the Antares Avox Choir plug-in, which turned out to be quite effective when set up in four-voice mode, adding depth and interest by making it seem as though there were more backing singers than was really the case. However, it also added a synthetic quality that I eventually found irritating, so at the last minute I swapped it for a tape delay plug-in, with the top rolled off the delays to soften the sound. Had this been a single for release, I would have had the backing vocals sung again to tighten up the performance. Hamlet suggested dropping the offending parts out of the second verse altogether, but I felt this would compromise the way the song builds, so I just tweaked their level slightly. You can mess around with editing and, indeed, some engineers seem to love the challenge, but unless the singer has actually died it's invariably quicker and better-sounding just to sing the part again.
To further sweeten the main vocal, I chose the TC Powercore Classicverb, which is a traditional synthetic reverb with a very simple set of controls. I added around 90 to 100 milliseconds of pre-delay to a vocal plate. I left the early reflections at maximum level, but reduced the level and decay time of the reverb tail, to give a clearer and more intimate sound. This worked well on the voice, after I'd used EQ to compensate as best as I could for the rather coloured recorded sound. Sweeping a parametric boost through the low mid-range revealed the expected boxy honk in the 200-250Hz region, which was then treated using a gentle dip at the same frequency. Mid-range and high boost was added from 3kHz upwards to put back some presence and air, which turned out to be reasonably successful — though you can never completely lose the coloration added by an untreated room, as the resonant frequencies hang on like a reverb that enhances only the room's less endearing attributes. I used Logic's Channel EQ, but any competent parametric EQ would be just as effective. The Tube Tech CL1B compressor plug-in I mentioned earlier worked very nicely to strengthen the vocal without making it sound 'pumpy', so I used that — though, again, any number of compressors would have given perfectly acceptable results here.
Once the track was balanced, and with only minimal use of automation, I inserted the usual limiter in the main signal path to avoid clipping, but for a bit of a change I selected the rather nice Universal Audio UAD1 Precision Limiter. With the barest hint of limiting on the signal peaks, the mix actually sounded pretty solid without further processing, but I weakened and dropped in another instance of the Tube Tech CL1B plug-in, just to see what would happen. Using the minimum compression ratio, and with only a decibel or two of gain reduction showing on the meter, the mix seemed slightly fatter, a hint louder and better integrated, so I left it in — though it wasn't essential. No doubt a modern mastering studio would squeeze more loudness out of it, but as we've commented in the past, pushing for maximum loudness doesn't necessarily do the sound of the music any favours.
In all, Hamlet had made a pretty good job of this recording, other than the difficulties he'd had recording the vocals in a virtually untreated room. If anything, he fell into the trap of using too many parts that occupied the same area of the frequency spectrum, and perhaps also of introducing some of these too early in the song, which limited the extent to which the song could build further without becoming too busy. That's an easy trap to fall into, and a combination of pruning and 'bracketing' EQ can really help.
If you find yourself fighting a congested mix, the mute buttons could be your greatest allies, because sometimes less really is more. Once I'd dropped the electric guitar out of the first verse, it was amazing how much clearer everything sounded, and when the chorus came around next time, the re-introduced guitar part gave it more impact. Of course, there is no one 'right' way to mix or arrange a song and what I came up with is just one possible interpretation, which ended up being an amalgam of what Hamlet wanted the mix to be and what I felt I could bring to it.
Hear The Files For Yourself
We've placed both Hamlet's and Paul's mixes, with a selection of before and after audio files of some of the individual elements of the song, on the SOS web site, at www.soundonsound.com/sos/oct07/articles/mixrescueaudio.htm.