Why is Al James' poppy and well-arranged track not reaching its potential in the mix, despite extensive EQ treatment and other processing? We dissect his original Cubase project to find out.
Al James, the producer of this month's Mix Rescue track, clearly knows how to put together a slice of pure pop. However, while his song 'Schoolboy Fascination' ticked all the boxes in terms of arrangement, and Al had made a creditable stab at a final mixdown, he eventually had to acknowledge that he'd got too close to the song to be objective about the mix. Although he'd already achieved a demo good enough to pitch to artists, he was keen to find out how he might push things further towards a more polished sound: vocals that would cut through more without being too 'out there', drums that would be punchy without being too dominant, and, in general, a greater sense of space in the stereo field.
Given the well‑developed original mix, I suggested to Al that in this instance it might be more instructive for me to continue working from his final Cubase project, rather than just starting from the raw multitrack audio files, as per usual in Mix Rescue. He duly sent this over for me and it loaded up fine on my system, except for a couple of Waves mastering plug‑ins that Al had inserted in his master output — but this was no great loss, given that we both agreed he'd pushed loudness processing a bit too far, even bearing in mind the pop style.
Looking through Al's mixer settings, the first thing that struck me was that there were a lot of tracks (64 audio tracks, by my count, along with a handful of Group Channel Tracks), but they were laid out somewhat haphazardly in the Project window, with tracks of vocals, percussion, piano and guitar intermingled throughout the track list. There was some grouping of similar tracks together, but it quickly became tiresome skipping about in the track list trying to trace the sounds on which I wanted to concentrate. This may seem like a bit of a petty thing to pick up, but it's really important to lay out your tracks sensibly, otherwise your creative flow when mixing will continually be punctuated by unnecessary interruptions. It's worth pointing out that most of the high‑flying professional mix engineers we've interviewed in the magazine employ assistants to map out each new project, so that the same instruments always come up where they're expected, for precisely this reason.
Opening up Cubase's Mixer window, something else came to light: switching to the Show Sends view in the Extended Mixer demonstrated that Al hadn't used any effects loops at all, and what delays and reverbs there were had been inserted on the tracks or groups in question instead. Reverbs are tremendously important to this kind of mainstream pop sound, not least because they help provide some of the stereo width that Al was ostensibly looking for, and the problem in Cubase is that if you insert a stereo delay or reverb into a mono track channel, it sums it to mono by default — which defeats some of the point of using the effect in the first place! Global reverbs and delays are also very handy for getting tracks to blend with each other, and of course the vocal blend was another area with which Al was dissatisfied.
An examination of the EQ settings across the mix revealed another significant concern: there was a lot of heavy processing going on with Cubase's built‑in channel EQ, including some very large high‑frequency peaking and shelving boosts. The kind of processor‑friendly digital EQ algorithms bundled with most sequencers are useful up to a point, but you generally get the best results out of them if you keep the processing to a minimum, using wide filter bandwidths to reduce unnatural artifacts such as phase‑shifts and resonant ringing. In my experience, this is especially relevant at high frequencies, where large boosts can frequently lead to quite abrasive‑sounding results. I'd already noticed a certain sonic brittleness to the original mix, but had initially put this down to over‑enthusiastic mastering‑style post‑processing. The channel EQ settings, however, provided a much more likely alternative explanation, and I could see evidence that Al had actually been trying to reduce this particular tonal problem in his master output channel with an instance of Cubase's GEQ30 graphic equaliser.
It makes a great deal of sense with built‑in sequencer EQ primarily to cut rather than boost, as that concentrates any undesirable processing side‑effects into areas of the frequency range that you want to hear less of, rather than focusing them into those areas that need to be most prominent. This especially applies to peaking filters, so the presence of more than 12dB of peaking boost on some of Al's tracks was not a good sign. For this mix, I figured that it would be best to go back to the drawing board as far as the EQ was concerned, and see whether a lighter touch might deliver a smoother overall timbre.
Starting the remix, then, consisted in the first instance of simply colouring the tracks in groups and organising them using Cubase's Folder Tracks so that it was easier to find my way around. By virtue of that process, I now had an idea of what each track contained, and could then mute everything and begin building the balance again from scratch. There was little point in working from Al's balance, given that I'd already decided to substantially rework his EQ settings.
As befits a pop track, I started my mixing by concentrating on the vocal parts, and Al had already got the right kind of idea here with his original processing, where he'd looked to emphasise the breathiness and intelligibility by tilting the balance of frequencies towards the high end. However, by using 14dB of high‑shelving boost at 6.5kHz and a further 3.5dB peaking boost at 3.4kHz, he'd created a sound with lots of extreme high end, and (by virtue of the default high‑shelf EQ curve he'd selected) a big dip at around 4.5kHz. The result was a curiously crispy sound, which nonetheless seemed veiled in the mix, owing to the fact that one of the regions that can really allow a vocal to cut through had been recessed. Sibilance was also a problem, and Al had clearly toyed with using Cubase's SPL de‑esser to deal with it, but had switched it off upon discovering (as I usually do) that it wasn't actually much help!
Given the severity of the EQ on this track, I was surprised to find that Al's raw recording was actually rather good. Indeed, bypassing the EQ immediately improved the sound no end; there's a lesson in there somewhere! My main impression was that there was a little too much low end for a pop mix (a common problem because of the proximity effect when close‑miking lead vocals with a cardioid microphone), but a simple high‑pass filter at 128Hz fixed that. Like Al, though, I fancied a slightly airier and more enhanced tone at the high end. Experience has taught me that this is the kind of situation where bundled sequencer EQ simply isn't the tool for the job — subjective tonal changes derive from many other factors other than just frequency‑balance changes. Instead, I pulled up a few higher‑quality (and more processor intensive) modelled analogue EQ plug‑ins to see what they could offer, including my usual standby URS Console Strip Pro and the Harrison, Helios, Neve and Pultec plug‑ins from Universal Audio's UAD2. In the end the Harrison seemed to suit Al's vocal best, bringing out extra intimacy and breathiness in the vocal with only minimal boosts: a fraction of a decibel at around 2kHz and a further 2dB at 17kHz. Plug‑ins like these (and, indeed, the analogue hardware on which they are modelled) change the sound even before you do anything with the EQ, so if you find one that suits the sound you're processing, half the battle is won already.
Given the quality of the vocal recording, why the heavy‑handed EQ in the original mix? My guess is that Al's compression settings were the main culprit, because the vocal was only just tickling the gain‑reduction meter. While I'm all for leaving as much musical dynamics in a part as possible, where appropriate, a pop‑style lead vocal like this needs to have its dynamic range firmly controlled if it's to remain audible over a busy backing track. My guess is that it was the search for audibility in the absence of adequate dynamic range control that led Al to overcook the high‑frequency vocal EQ. It also suggests a reason why Al might have used vocal double‑tracking so extensively, as this tends to produce a more consistent combined level for the double‑tracked parts.
The performance itself was fairly consistent, so I aimed for compression settings that would reduce the dynamic range overall, reducing Al's ratio setting for the built‑in Cubase Compressor plug‑in to 1.7:1 and setting the threshold low, so that it would grab the signal pretty much all the time. With a ratio that low, I was able to let the compressor take off 12dB or more, even with fairly fast attack and release times (roughly 2ms and 60ms respectively) to pull up the small‑scale details, yet the subjective dynamics still seemed fine in context.
The biggest problem with this much compression is that it tends to emphasise vocal sibilance. As I've already mentioned, Cubase's built‑in de‑esser and I are hardly on speaking terms, and I've prefered to use Digital Fishphones' Spitfish in the past. However, inspired by Bob Clearmountain's reference to de‑essing in our February 2009 'Inside Track' feature, I opted instead to use Cubase's dynamics side‑chaining facilities to implement de‑essing from first principles. First of all, I created a duplicate vocal channel, which I EQ'd especially to isolate sibilants, using a high‑pass filter at 4.5kHz and 24dB of peaking boost at 8kHz. This then fed only the side‑chain of a second Compressor plug‑in on the main lead vocal track, causing it to reduce the vocal gain only for sibilant sections, using fast attack and release times. A bit of juggling of the threshold and ratio controls was required to get the appropriate control over the problem, but it certainly did the trick. (I was also later able to automate the level of the side‑chain feed in a couple of places where I was getting some lisping side‑effects, which was useful.)
The bass was next to be introduced, then piano, then drums, and finally the other parts were slotted around those as space permitted, until the bulk of the arrangement was in place. As I'd found with the vocal part, most of the tracks sounded pretty good without much EQ at all. As such, more than half of all the EQ bands I used were simple high‑pass filters just to keep the low end of the mix clear of clutter, although I wasn't afraid to ratchet these well up into the mid‑range if those frequencies on a specific instrument proved to be redundant within the mix as a whole — there was, after all, a lot going on a lot of the time.
Beyond that, I was mostly just homing in on over‑prominent frequency areas in each new instrument as I added it to the mix, cutting away with the EQ so that I could achieve an appropriate fader level for that instrument and not trample over the other more important parts that I'd already set up before it. All but a couple of these cuts were 6dB or less: one of the exceptions was a narrow, deep notch to surgically target a drum resonance. As for boosting, only about eight of a total of around 90 EQ bands in the final mix were boosts, and (with the exception of my de‑essing side‑chain, a telephonic special effect for the opening TR808 channel, and one of Al's leftover 14dB peaks on a pizzicato sound) none of these were more than 4dB.
My point in lapsing into a bit of quantity surveying is to demonstrate that you don't need to use every band of EQ on every channel to put a mix together if your sounds are already reasonably clean, as is usually the case if you're working with modern virtual instruments. Also, it's a pretty good rule of thumb, in my experience, that if you're finding yourself dialling in more that about 6dB of gain in an EQ band, then either you're probably using the wrong tool for the job (especially if you're boosting), or there's another track in your mix that needs processing to clear some space for the track you're working on.
Reworking the EQ was definitely the bulk of the processing I needed to do on this mix to get to a decent rough balance, but there were also some aspects of the drum programming than needed a bit of attention. The first thing was that the kick-drum part had contained some MIDI duplicate notes that were causing the kick‑drum sample occasionally to trigger twice in very quick succession, creating a phasing sensation and robbing the drum of its weight. I could have sorted this problem out with some patch‑up audio editing work, but I was also less than thrilled with the specific kick‑drum sound: it hadn't been particularly well chosen for the track, and sounded more like a hard‑rock drum, with a very aggressive, slappy beater click and too much thick ambience in the sustain phase. My decision was to replace it completely, and because Al had helpfully left his original MIDI files in the Cubase Project, this was a straightforward matter of just zapping the offending duplicate notes from the trigger part and loading up a replacement instrument.
When I recently reviewed Sonic Reality's Ocean Way Drums Kontakt Instrument, I was most impressed by how effectively you could mould the sound of each drum by rebalancing its multi‑miking setup, and I fired that up, reasoning that it would make it easier to find a sound to suit this particular mix. I tried a couple of the different kick drums on offer before finding one that sounded promising, then adjusted the microphone balance to favour the close mics and achieve a nice tight, punchy beat. It could have taken all day to chase that kind of improvement if I'd looked to audio editing and processing, but as it was, all I spent was 15 minutes finding the new sound and a further two on some very minimal EQ: a 50Hz high‑pass filter to lighten the tone and clear room for the bass, and a 1.4dB peaking boost at 1.4kHz, which I used later to increase the beater definition after I'd already bounced down that instrument track as audio.
The snare and ride parts were lacklustre for a different reason: the samples in question simply didn't have enough velocity layers to deliver musical‑sounding variation. Again, because the sounds themselves weren't exactly stellar, I pulled up another couple of instances of Ocean Way Drums to replace them. That library has one or two velocity layers to spare, so things were much improved. I didn't completely replace the snare part, but rather layered the original behind the new sample, to retain a slightly stronger link between the old and new sounds. The other drums were mostly good enough for their purposes in the song, although I did use Cubase's Envelope Shaper plug‑in on each of the toms to bring up the stick attack, as they were all a bit soft‑sounding compared with the snare. Another Envelope Shaper removed the attack from the original snare track, scotching a bit of flamming I'd noticed between the new and old snares. As a finishing touch, I added in a few little accent cymbals from Spectrasonics' Back Beat sample library.
Once I'd got the EQ readjustments in the bag, I was able to get a rough balance pretty quickly, give or take a few instances of Cubase's simple Vintage Compressor. I could then turn my hand to another aspect of the mix that hadn't received much attention in the original version: send effects. I kept in some of Al's original special effects, particularly the modulation processes he'd chosen, as they were clearly part of the sound he had in mind, but I bypassed a number of the delay and reverb inserts to leave room for my own.
In the first instance I just loaded up Christian Knufinke's SIR2 with one of my favourite impulse responses (a hall ambience that isn't too long), and used that to glue the dry sound sources together to some extent, as well as effectively increasing the warmth, depth, and width of the stereo picture into the bargain. In fact, there was rather too much warmth to the reverb sound, as is often the case, so I gradually rolled off the low end of this reverb from around 800Hz to avoid too much mush. The upper ranges of the reverb also felt a bit splashy for this mix, and I shelved off a few decibels of that as well. While I was tweaking the reverb controls, I took the opportunity to add in a bit of pre‑delay, which helps keep the reverb more in the background, and widen the stereo image of the effect too, as a little extra stereo width rarely goes amiss in pop.
This one reverb was pretty successful at binding the mix as a whole, but was less successful at bringing the drum sound together. As they stood, the drums (unsurprisingly) felt a bit dislocated from each other, as if overdubbed one instrument at a time, and needed more of a feeling of a room around them, so I loaded up the Pantheon II reverb that's bundled with some of Lexicon's interface hardware, and looked for something suitable from that. I've always found Lexicon's reverbs to be great for drums, dense enough to feel real and yet devoid of the nasty‑sounding resonances that can afflict other reverbs when they're presented with percussive material. I tried the different preset algorithms on offer and settled on the Chamber, before reducing the room size to its minimum of 10m, leaving around 0.6s of fairly bright reverb tail that I then high‑pass filtered, as I usually do. This was then fed from most of the drum tracks in the mix, particularly the cymbal and tom 'close‑mic' tracks. Reducing the pre-delay to zero seemed to work best in this case, allowing the reverb to pull the drums back into the mix behind the other instruments.
The blend had, again, improved, and feeding a little of some of the other parts to the drum chamber also helped, but there were a couple of instruments — most notably the piano and a middle‑section electric guitar — that just felt as if they wanted a bit more warmth and sustain, beyond what could be achieved with the two reverbs I already had on the go. I chose the UAD Plate 140 plug‑in for this task, because it doesn't impose much additional sense of space. Plate 'C' of the three available sounded most appropriate for the piano, so I went with that, and didn't even bother with EQ in this case, because the instruments I was feeding to it already had a controlled frequency range by dint of their channel processing.
My fourth and final send effect was a simple mono tempo‑sync'ed delay patch, which I used to change the character of the lead vocal parts to suit the different song sections: muting the effect gave the verses a tighter sound; a fairly reserved effect level opened out the choruses; and a more obvious echo worked well for the more expansive pre‑chorus and middle sections. Tweaking the send level from moment to moment was also vital, to prevent stray delay repeats from smudging the lyrics. As I often do with vocal delays, to make them less audible in their own right, I de‑essed the send to the effect using an additional insert in the effects channel's plug‑in chain, but I also sent a little of the delay return to the drum chamber to give the repeats a more background character.
At this point, the 'sound' of the new mix was in place, but there was still a crucial ingredient missing: automation. In less complex mixes this would perhaps be less of a concern, as a good arrangement will, to some extent, mix itself, but in this kind of pop there's so much going on that you need to help the listener find the most interesting features at any given point in the track, by adjusting the balance moment to moment. In some styles you can just compress everything into oblivion and hope for the best, but that's not really an option in pop, where such heavy dynamics processing can swiftly become OTT.
I've written frequently in the past about detailed vocal automation and, as usual, this was important to keep intelligibility high, but solidifying the vocal levels also allowed me to remove the vocal double‑tracking in some sections, making the vocal more intimate and direct. Many of the other tracks benefited from automation too. The bass and piano, in particular, offered lots of nice little melodic fills that could be brought into the limelight, and there was some benefit in further evening out the levels of these instruments in response to changes in the overall arrangement. Fader levels were used to refine the levels of the drum fills, and while I was doing that I found that editing out the odd tom note here and there helped with the musical flow as well; over‑egged fills can make a track feel like it's stumbling over itself if you're not careful.
The send effects had some automation too, beyond my aforementioned send‑level rides for the vocal delay. The piano's reverb could, likewise, be pulled back for the verses and expanded for the choruses, and I muted all the effects returns for a moment just before each chorus, in response to Al's request for an improved staccato feel at those points.
This month's Mix Rescue illustrates quite a common misconception amongst recording musicians, namely that you have to crank your processing really hard to get results. In reality, it's more likely that a combination of moderate settings for each of the primary mix tools (EQ, dynamics, delay/reverb and automation) will yield more appealing results, particularly where you account for the strengths and weaknesses of each. Al was trying to use EQ to do jobs best dealt with by switching instrument patches, applying more compression or adding global send effects, but in other mixes I've encountered it's the compression or reverb that have been overdone. The moral of the story is to try as many different tools as you can, and then use the ones that actually produce the goods most naturally. That way you'll use less heavy processing overall — and your mixes will almost certainly thank you for it.
Singer, songwriter, and producer Al James is based in Edinburgh and has been in the music industry for many years as a performer and songwriter. He's currently writing songs specifically to target emerging artists and their publishers. Although he mostly writes and records alone, this month's song was co‑written with his friend Marc Mowbray, after which Al put the song together in Cubase 4, primarily using the bundled Halion One soft synth, Steinberg's Virtual Guitarist, and Synthogy's Ivory. All the vocals were sung and recorded by Al.
Al James: "Having invested dozens of hours in putting my original mix together, as usual I found I was way too close to it to be objective, so I wondered if Mike Senior could improve on my mix. And I'm happy to say he has done just that! As they say in Edinburgh, I'm chuffed to bits with the remix Mike has achieved. Directly comparing the old and new mixes on headphones, Mike's mix certainly sounded different. To be honest, it didn't sit well with me at first. However, I knew my ears were too close to my own mix to be subjective, and as I listened many times I grew to love the new direction Mike had taken. Clearly, Mike has an excellent understanding of what I'm wanting to achieve!
"On hearing the first draft of Mike's remix, there were some minor tweaks I wanted doing, particularly in terms of the level of the Virtual Guitarist part — although had it been real guitar this would not have been a problem. Also, I felt that the 'stabs' prior to each chorus needed to be more staccato to create more impact for the subsequent choruses. Mike was only too happy to address these tweaks, and after a couple of emails he had a new mix for me that just 'worked': everything in its own space, vocals up-front, and punchy drums. I'm very, very happy!”
We've placed before and after audio files on the SOS web site at /sos/aug09/articles/mixrescueaudio.htm so that you can listen to the different elements of the mix before and after Mike's reworking of the track.