Mix Rescue: Eastern European Folk Music

Paul White's studio, where he undertakes the remixing of your track.

Can't get your mix sounding right? Let SOS sort it out!

Everyone who's tried their hand at mixing will know the feeling. You work on a mixdown for hours, carefully balancing faders, tweaking compressors and EQs, and automating every last aux send, but you still can't get the sound you can hear in your head. You end up doubting your monitoring, your ears, and even the quality of the song itself.

You're not alone. In fact, it's because this kind of creative burn-out is so common that so many commercial productions end up requiring the services of a specialist mix engineer. The idea behind this is that someone with a fresh pair of ears and a wide-ranging experience of different productions can bring valuable objectivity to the final product, focusing the mix around those elements which really make the production stand out. But such luxuries are not really an option for most home studio owners, so in reality most just go it alone and hope for the best.

We hear the results of this every month, either via CDs sent direct to the SOS office or via MP3s emailed to us, and in a significant number of cases the mix is the main weakness of the final result. So we've decided to tackle this malaise head-on by taking problem mixes from real readers and remixing them from scratch ourselves to try to get closer to the sound they had in mind. This, of course, means that we're not going to entirely reconstruct tracks and add significant new parts in the way that a dance remixer would — instead, we'll try to work wherever possible with the existing recorded material.

Introducing Ouja

The first candidate for our remixing challenge was an instrumental track recorded by violinist Georgina Clarke as part of her music technology college course. Outside college, Georgina is one half of a duo that goes under the name of Ouja, playing material, much of it original, based on the style of eastern European folk music. Her partner Paul plays Mandola and, for this particular recording, Georgina had brought in a couple of friends to add a bass guitar line and a djembe drum part. A solitary MIDI track was used for a sampled accordion counterpoint line in the middle section, and this was originally added at the college using IK Multimedia Sample Tank.

The Bass Amp plug-in settings Paul used for processing his replacement bass-guitar line.

Before starting out, I asked Georgina how the track had been recorded. She explained that both the mandola and the violin had been recorded using two mics, one close to the sound hole and one further away, but the choice of mics was probably not optimal for these instruments, as she'd used a combination of AKG C1000 back-electret mics and Shure SM58 dynamics. These are fine mics for their intended live applications, but a more sensitive capacitor mic with a better high-end extension would have been preferable. Using two mics on a single instrument also brings up the potential problem of phase cancellation affecting the tone in a negative way, while placing one mic very close to the sound hole runs the risk of over-emphasising the body resonance of the instrument. As the tracks I was provided with comprised the mix from both mics, I wasn't able to separate these.

The bass guitar was simply plugged directly into the line input of the 02R used for recording, while the djembe was recorded by pointing a mic into the open bottom end of the drum. The final recording was to an Alesis HD24 hard disk recorder, which made it easy for Georgina to bring me the separate tracks as WAV files, which I could then import into my own Apple Logic Pro system for mixing and processing. The MIDI part for the accordion line wasn't on the CD-ROM she brought along, so she played that part in again. She also brought along her original mix, which to my ears was over-EQ'd, making it sound rather dull, and she'd also used a lot of reverb with the aim of creating a big and spacious sound. The balance Georgina had set up was pretty good, but because of the way the track was treated, it sounded almost like listening to the band from a distance, or from outside the room in which they were playing.

Preliminary Edits & Bass Fixes

Fortunately for me, Georgina had decided to record by overdubbing one instrument at a time to a click track rather than playing everything live, so separation wasn't a problem, and, because the tempo was fixed to the click, I was able to do some basic editing. After loading the WAV files into a Logic Song, the first thing I did was to try to establish the tempo at which they had been recorded, because Georgina didn't have a record of this. Logic 's tempo counter actually made a pretty good estimate working from the djembe track, but I found I could also get the same result by adjusting the tempo manually while listening to Logic 's own click. Fine tuning could then be done by looking at the waveform display for the last bar of drums (which of course would drift the furthest out of time in the case of any error) and then making very fine tempo adjustments until these sat exactly on the bar and beat lines. I did this check anyway, just to ensure that there was no drift, and ended up with a very odd tempo value to three decimal places, so apparently their click didn't run at exactly the same rate as the one on my system.

Although the mandola part had been well-played, there was quite a bit of room sound on the recording. However, this was soon sorted out with some low-frequency EQ cut and a dash of ambience reverb from the TC Powercore Classicverb plug-in (above). A second reverb, created using Apple Logic's Space Designer (top) was applied to flatter just the violin part.The next check was to listen to the individual tracks to see if there were any problems, and it turned out that the bass-guitar part was going to be the hardest to deal with. This had been DI'd and the tone was rather dull, but with a lot of high-frequency fret noise and rattle that stifled any attempt to use EQ to better focus the sound. The timing was also a bit on the rough side, and to my mind the player had picked the notes too gently, so there was no real definition to the sound. After a few minutes playing with EQ, compression, and Logic 's Enhance Timing plug-in, I decided that it was never going to sound entirely satisfactory, so I replaced it with a new bass-guitar part played with a pick and following the notes of the original as closely as I could make them out. This was DI'd via the high-impedance instrument input of my MOTU 828MkII interface and then shaped using the Garage Band Bass Amp plug-in, which now comes as part of both Logic Express and Logic Pro. Users of other platforms could achieve much the same result using any amp-modelling software or by DI'ing via a modelling guitar or bass preamp. The end result was a more middly bass sound with more definition, and I felt it would sit better in the mix as a result. Compression was added after the amp plug-in to even out the sound, but otherwise it seemed fine for this track, where it played a basic supporting role.

Mandola & Djembe

Checking Paul's mandola track showed it to be well played, with no timing or tuning problems, though the low end seemed excessively boomy, presumably because of the location of the close microphone. There was also a certain amount of room tone in the sound, because of the use of a second more distant mic, but it wasn't too serious. Where artificial reverb is going to be used at the mixing stage, it is invariably best to eliminate as much room tone as possible at source by using acoustic absorbers (foam or duvets, for example), as room reverb tends to dictate the tonal character of the sound no matter what artificial reverb you apply later. My strategy in this case was simply to use some low-frequency EQ cut to clean up the low end, adding a hint of high-end boost to put some gloss back into the sound. I also applied compression to even up the level using the Waves Renaissance Compressor plug-in. A short ambience reverb was then set up on one of the sends to add life to the mandola and djembe drum without clouding the mix.

We made the artistic decision to drop out the mandola for one short section in the middle of the song, in order to add variety and to improve the general dynamic of the piece. This was achieved simply by cutting the file at the appropriate places, then fine-tuning the cuts so that they fell exactly before new 'strums'. As the rhythm part was fairly staccato anyway, the lack of sustain at the drop-out point wasn't really noticeable, and the ambience reverb helped conceal it further. All that remained to be done then was to clean up the starts and finishes of the track to ensure there was no unnecessary noise.

Here you can see the complete processing chain for the Djembe drum part. After some quantisation from Apple Logic's Enhance Timing plug-in, the Noveltech Character plug-in was used to bring out the attack of the sound dynamically. The Universal Audio UAD1's LA2A compressor then reigned in the overall levels a little, while the Waves L1 Ultramaximizer kept tight control over signal peaks.Turning to the djembe track, the timing turned out to be slightly loose in places — it's not easy playing to a click track when you're not used to it, as the conscripted bass player no doubt also discovered! The tone was also lacking in attack and brightness, but with a little cheating I figured that it was probably going to be usable. To establish a more solid, confident-sounding rhythm, I located a bar of straightforward four-to-the-bar playing, copied it to an adjacent track, and then looped it throughout the song so that it underpinned the generally more busy sound of the djembe part. I also used Logic 's Enhance Timing plug-in (a type of audio quantisation) to tighten up the main djembe part. This worked pretty well, except for some sections where very fast fills were being played.

I could have used EQ or a harmonic enhancer to add more attack to the djembe sound, but settled on a rather clever plug-in called Character, which is part of the TC Electronic Powercore suite of plug-ins. This emphasises certain parts of the spectrum after the manner of an enhancer, but it does so dynamically and in a way that isn't simply confined to bringing out the high end. As the name implies, it seems to lock in on whatever gives a sound its character, then emphasises it. I tuned this by ear to bring out more of the sound of the hand hitting the drum, then compressed the result to even up the levels — I used the Universal Audio UAD1's LA2A compressor plug-in for this, but there are lots of other compressors that would have done a perfectly adequate job here. The same treatment was applied to the drum loop, and both parts were then treated with a short ambience reverb to give them a sense of space. In the mix I panned the two drum parts slightly to each side to give an impression of width, but without separating them too much.

Removing Unwanted Background Noise

Hearing Georgina's violin part in isolation exposed some bow noise, but, given her fairly assertive 'gypsy' style of playing, this wasn't surprising, and for me it was part of the character of the sound. Again, I would have liked to re-record her using a single good mic in an optimal position, but the sound was at least natural enough to work with. Although the sound was a little bright, I felt that it would be better to work with it rather than aggressively shaping it with EQ.

The only technical problem with the violin track was a low-level ticking noise that could be heard during the decay of the final note of the piece. I don't know where this came from, but it was a similar sound to what you hear when interference from a computer disk drive is picked up on the output of a soundcard. However, whatever the reason for this, I was more concerned with hiding the problem than discovering the true cause. To achieve this, I used a high-cut filter set to around 10kHz for the main body of the song, taking just a little edge off the violin sound, and then drew in a control envelope on the final note which rapidly closed the filter to 1kHz as the sound decayed. This turned out to be enough to hide the ticking sound without changing the subjective tone of the violin, even when I added compression to smooth out the level peaks in Georgina's playing.

I added a little of the short ambience reverb to the violin to help it integrate with the other parts of the mix, but then added a second concert-hall reverb, as this really flattered the sound and came closer to what Georgina had envisaged. Logic 's Space Designer was used for both reverbs, but any number of hardware- or software-based reverbs would have done the job.

With the tweaks we'd made, it was now possible to have the djembe drums higher in the mix to drive the track along, as the timing was now more accurate and the articulation of the hand playing was more clearly audible. Having the straight rhythm loop underneath also helped keep the beat going, whereas originally it faltered a little when the player was adding complicated flourishes.

The final touch was to set up a sampled accordion to play the counterpoint Georgina had recorded onto the MIDI track, and I also added a low-level sampled tambourine part played with no quantisation to give more of a human feel. This helped reinforce the rhythm part and also added a bit of interest higher up the spectrum, which is something Georgina would have liked to have done originally given the studio time. We ended up copying the accordion part over three octaves to thicken the sound, making the lower line legato to produce a more believable drone. Once this was mixed to just support the violin melody, it appeared perfectly natural, even though it sounded less so when soloed!

Remix Reactions

Georgina: "Recording this track was a bit of a last-minute job, because we didn't have much time in the studio and we'd been having problems with the equipment. I also originally used a lot of EQ on the violin because I wanted a warmer sound, but I wasn't filtering out the really low frequencies, so the sound just became boomy. In only used a little compression on the violin and mandola parts, just to keep the levels consistent throughout, but I also managed to drown everything in reverb! As for the djembe line, Paul and I were never happy with the way it came out at all.

"Now that Paul's worked on the mix, everything's sounding much brighter, and you can hear from comparing the two mixes how much Paul has bought our tune to life. There's also better separation in the mix, because there's less reverb now — the individual instruments are much more distinct. I particularly like the way Paul has made the most of the djembe line by selecting the best bits and then processing them to dramatically improve the sound of the whole tune. I originally wanted to use more percussion to give the feeling of lots of musicians playing, but we ran out of time before we could do this. Now that the djembe part has been changed it has really beefed up the track without us having to add much else.

"It was very helpful to go through Paul's mix in the studio, because he demonstrated to us how he had used simple techniques to get big results. One particular point that stuck with me was how to EQ the bass to allow the other instruments space in the mix — I had simply turned it up and swamped the overall sound."

Paul: "This remix session was an intriguing insight into how much you need to know about sound and technology to do a decent recording, but also into how much is on offer in terms of sound quality and capabilities once the technology has been mastered."

The Final Result

While not perfect, the final mix was more open and 'present', and also closer to what Georgina wanted. But it also showed the limitations of the 'fix it in the mix' approach, so I can only reiterate that the way to get a great mix is to get your sounds right at source when recording them.