Вы находитесь на странице: 1из 16

YOUR RECORDINGS

34 / COMPUTER MUSIC / November 2013

clean up your recordings / make music now <

Whether youre looking to fix a dodgy recording, restore an old demo track or just do away with irritating mains hum, our curative walkthroughs will patch up all your audio injuries
Picture the scene: after a long day recording in an unfamiliar studio, you get home for a playback session on your trusty monitors, only to find that things arent sounding as great as you thought. The bass is completely lacking in bottom end, the rhythm guitar sounds dreadfully harsh, theres mains hum all over the keyboard parts and the drums sound like a set of suitcases falling down a flight of stairs. Suddenly the whole thing is a disaster. But dont panic s here to make it all better! There are plenty of reasons for recordings going south. A lack of experience on the part of the engineer is a common one, but poor performances, phase issues, time restrictions, dodgy equipment, loose connections, broadband noise interference and numerous other gremlins can also ruin the results of an otherwise successful session. And, of course, these issues arent restricted to those recording bands. Perhaps youre looking to resurrect a noisy old demo from tape, or youre working on a mix or remix using parts that have been recorded and sent to you over the net by other musicians/producers, with highly variable results. Or youve got a sampled drum loop you want to use in your latest electronic track that grooves like a mutha but is ruined by weird resonances Ultimately, what were trying to impress upon you is that every computer musician will find themselves needing to rescue a bad audio clip sooner or later and thats where we come in. You may not realise it, but your computer gives you access to some extraordinarily powerful and sophisticated tools with which to repair audio (and were not just talking about pricey restoration plugins here). While theres no sure-fire, single-click substitute for getting things recorded properly in the first place, the phrase we can fix it in the mix has never rung more true, and theres an abundance of tools and techniques that can rescue your tracks. Over the pages that follow, were going to diagnose and cure a whole host of ailments that you may commonly encounter in your audio engineering and production endeavours, both purely sonic and performance-related. Youll find all the examples and files accompanying these walkthroughs on the DVD and at vault. computermusic.co.uk, so scrub up, grab your audio bone saw and prepare to operate!

DOWNLOAD
Get the samples, video and tutorial files on your PC/Mac at vault.computermusic.co.uk

How to use this tutorial


To demonstrate our techniques, weve created a simple demo recording project that well be using for all the walkthroughs. The multitrack audio for this project can be found in our Tutorial Files folder, which you can download over at vault.computermusic.co.uk or find on the DVD that comes with the print edition. The easiest way to follow along is to make a new project in your DAW, set it to 144BPM, then drag the WAV files from the _STEMS folder onto separate tracks in your DAW. Youre now ready to follow our tutorials!

November 2013 / COMPUTER MUSIC / 35

> make music now / clean up your recordings

Frequency response issues


Thin. Harsh. Dull. Boomy. Nasal. Muddy. Tinny. No one wants their audio to sound like this, and the general cause of such problems is what we call a poor frequency response that is, the relative levels of the frequencies in your signal. There are many ways in which recorded audio can end up like this. Perhaps the instrument being recorded actually sounds that way. Or maybe you used a mic that picks up way too much (or not enough!) of a certain frequency range. The room you record in also greatly affects frequency response. Sometimes the effect can be very specific, affecting a narrow range of frequencies, causing whistling resonances and ringing artifacts. Its hard work mixing sounds with disparate and inappropriate frequency responses, as youll have to apply unusual and unintuitive amounts of EQ, and dynamics effects like compressors may not react in the way youd expect. Its unlikely that youll achieve the best results working this way. We recommend a two-step solution: apply surgical EQ to correct deficiencies, then follow up with a separate EQ for the usual broad tonal shaping and mixing. As you may know, EQ is short for equalisation, so-called because it was invented to even out the tonal character of a sound by cutting and boosting certain frequencies across the spectrum, creating an even, balanced tone with each frequency at an equal level. A good parametric EQ is an absolute must-have in any producers toolbox, and almost all of these feature a graphic display of frequency laid out horizontally across the centre, against gain or volume up the left and right edges. Several frequency bands can then each be centred around a specific frequency and adjusted to make that particular frequency louder (boost) or quieter (cut). A third control, mysteriously called Q, sets the width of the range of frequencies affected when you cut or boost that particular band. Each adjustment you make to a band is reflected by a curve on the graphic display, so its easy to tell visually what effect your changes are having on the sound youre EQing. Filters, meanwhile, are designed to reduce frequencies above or below a set point the rate at which they do this is known as roll-off. A 24dB-per-octave roll-off is a much steeper cliff than a 12dB-peroctave. High-pass filters (so-called because they only allow higher frequencies to pass through them) set at about 60-100Hz are useful for getting rid of rumble, and lowpass filters set between 9-10kHz are good for slicing off top-end hiss.

> Step by step

1. Smoothing out harsh guitars with EQ

FILES

TUTORIAL

Were going to fix a thin, harsh guitar track with too many upper-mid frequencies, although this technique will work just as well to correct signals that are dull, boomy, etc. We can pinpoint the problems using a spectrum analyser plugin, such as Voxengo SPAN (free at www.voxengo.com). Set your project to 144BPM, import the test guitar from the _STEMS folder, and add SPAN to its track.

Click the Edit button and set the meter type to Cumulative Average (AVG), which gives an overall picture of the frequency levels over time, rather than a constantly updated dancing graph. For greater accuracy, set a high Block Size, and turn up the Overlap to prevent the display blinking. Then set Smoothing to 1/6 Oct this makes the display easier to make sense of.

Play a long section through the analyser the longer, the better, as this gives more of an overview of the general guitar tone rather than just showing the harmonics in one phrase or chord. Theres a big peak around the 3kHz region, which gives the guitar that honky harshness. There are also troughs around 250Hz, 550Hz and 1.7kHz. Note that the high-end roll-off seen here is normal for guitars.

Now we insert a parametric EQ before the analyser in the plugin chain, so that the analyser displays the EQd sound. Were using DMG Audios EQuality, but any fairly flexible EQ with a good number of bands would work just as well. Using the EQ, we boost the areas where the analyser shows gaps, and make cuts where there are peaks, almost creating an inverse of the curve displayed in the analyser.

The resulting graph is much flatter, and the guitar now sounds much fuller and more mix-ready. Troublesome resonances at specific frequencies (whistling) can be isolated by disabling Smoothing and zooming in the analyser display to see if anything pokes out. Were looking for tall, sharp spikes at precise frequencies here we can see that theres one at around 1.05kHz.

To deal with this, we turn up the Q parameter to make a very narrow EQ band at 10kHz, then cut it by 10-12dB. Once youre happy with your flattening EQ, you can add another EQ plugin for broad tonal shaping (for example, a boost around 2kHz often works well on guitars). The signal will also be much more receptive to treatment by other processors such as compressors and virtual amps/cabs.

36 / COMPUTER MUSIC / November 2013

clean up your recordings / make music now <


> Step by step
2. Using notch EQ to eliminate mains hum

Mains hum can occur when recording external electronic kit such as synths, electric guitars and turntables. In the UK, mains hum occurs at 50Hz, while in the US its 60Hz, both with harmonics. Solo the Synth track (from the STEMS_ folder) and youll hear a loud low-end hum. Add a gate plugin and set its sidechain filter to focus on the 500Hz range. Now, the synth will open the gate but the hum wont.

Set the gates Threshold to -50dB and Release to 170ms. Now, when the synth isnt playing, the signal will be silenced so you dont hear the hum on its own. However, we can still hear the hum when the synth is playing, which will really cloud the bottom end of the mix. We can use EQ to notch it out, though. Add DDMF IIEQ Pro CM after the gate. (Audio: 2. Gated synth.wav)

Set all bands to notch mode, and set them to 50Hz and its harmonics: 100Hz, 150Hz, 200Hz, etc. Maximise the Q on all bands, then adjust the Q of band 1 to taste. This takes out most of the hum. Itll also adversely affect the low end of the synth part, but you can use a low shelf boost to help offset this if needed. A 24dB high-pass filter at 86Hz kills any remaining rumble. (Audio: 3. EQed synth.wav)

Perfect match
Weve shown you how to use EQ to flatten a guitar, but you dont have to aim for a flat curve indeed, the guitar (or other instrument) tones on your favourite records may be far from flat. You can use spectral analysis on these golden tones to see their frequency response, then apply EQ to get your tracks closer to theirs. Simply find and loop a part in your reference track where the guitar plays on its own, insert an analyser plugin (with the same settings as used on your guitar), then EQ your guitar until the curves look similar. Mid/side processing on the reference track can help for example, on many rock tracks, turning down the mid signal (Voxengos free MSED is perfect for this) can give you an almost perfectly isolated guitar part. This matching technique is also useful with multiple guitar parts that you want to sound similar. Alternatively, you can use a matching EQ such as iZotopes Ozone 5 (being aware that its Source and Target labels are the wrong way round!) or Logics Match EQ to do it for you. These enable you to capture the tonal characteristics of a signal and automatically equalise your own audio to give it a similar sound, at far greater resolution than is possible doing it by hand. Finally, understand that even the most accurate EQ matching will hardly ever make one signal sound exactly like another but it will get you a bit closer!

> Step by step

3. Fattening a thin bass sound with a synth

In our _STEMS folder, youll find a badly recorded bass the highs arent crisp and theres hardly any low end! Rather than correctively processing the signal, were going to use a synth bass to supply the missing frequencies. Set your tempo to 144BPM, bring up DuneCM on a MIDI track in your DAW, and load the patch 1. DuneCM Bass.fxp from Tutorial Files.

This is a simple mono bass sound comprising sine waves an octave apart and a tight amp envelope. Playing it back in the octave of C-1 alongside our recorded bass restores its beefy low end. You could use your DAWs audio-to-MIDI function to get the parts notes and timing, but for a simple line like this, it might be quicker to just sequence it by hand.

This give us a solid low end, but the highs still need work. Duplicate the bass track, and add IIEQ Pro CM to the new version. Put the first band in HPF24 mode, at 2kHz or so, to isolate the very highest frequencies of the signal, which can be distorted to generate and/or emphasise the lacking upper harmonics.

We want the distortion to have a consistent tone, so we put CM-Comp 87 after the EQ, setting its Threshold to -56dB, Ratio to 2.0:1, and Makeup to 21dB. Then we add a distortion effect to grunge up the signal. For a more dynamic, lively sound, leave out the compression. (Audio: 4. Full frequency bass.wav).

September November 2010 2013 / COMPUTER MUSIC / 37

> make music now / clean up your recordings


> Step by step
4. Creating double-tracked parts from one take

Huge, 3D-sounding parts can be made by recording the same thing multiple times, known as double-tracking (or tripletracking, etc). If youve only recorded a single part, though, dont despair, as creating a double-track from an existing part is possible, provided theres enough source material in the original track. Well start with the guitar part from the first tutorial, with our corrective EQ in place.

Remember, our track has been recorded at 144BPM, so set your projects tempo to this. Duplicate the guitar track, then delete the audio thats on the new track. Find a section where a guitar phrase/riff is played twice in a row and cut it into two halves between bars 10 and 18 works for us, with the cut at bar 14. Make each half a different colour so that you can tell them apart.

Copy the two sections onto the second track this is where well assemble our fake doubled track, using the first track as the master. Pan the first track hard left and the second hard right. If we hit play, the guitar will sound louder, still centrally panned and totally mono, as you may have predicted. Thats because the exact same part of the recording is playing back on both channels.

To get a proper double-tracked effect, we simply swap the two clips over on the second track so that the same two sections of the recording never happen at once they now swap over between channels after four bars. Now we have got the illusion of two guitarists playing at once one in the left speaker, one in the right, giving that big, wide, double-tracked rock sound.

Work through the song using the same technique wherever possible. There may be areas where you need to get creative to make the second track. For instance, when the guitar comes in at the start of the track, it seems theres no corresponding part to copy to the second track; but because the lick is repeated later in the tune, we can just copy and paste one of those instances into place.

Occasionally youll come across an unwanted variation, where the aligned parts dont play the same thing. For example, in this segment, there are different phrases at the end of bar 25, and a mistake in bars 18-20 of the right hand channel. Lets first tackle the mistake. We zoom in and snip out the offending portion, then copy and paste a well-played phrase from bars 22-24 in its place.

Joins and edits can be noticeable, so you may need to tweak region boundaries to get patched regions to sit tightly. You can also use your DAWs crossfade tools to smooth out transitions and kill clicks and pops. If youve patched up a section with mistakes, dont forget to copy and paste the corrected version back into the place it came from on the original track to correct the original mistake, etc.

To fix the variation at the end of bar 25, we use the same technique as before, pinching a bar containing a stop from bar 37 and pasting it in onto the right channel at bar 25, so that the two sides are now playing the same guitar part. Matching up the parts where they occur the other way round at bar 33 is trickier, as the left-channel part is only played once in the song. Well have to get creative

To recreate the necessary part, first set snap to 1/8, then copy and paste the chord from 33.4.1.0 to 33.3.1.0, and the chord from 33.2.3.0 to 33.3.3.0. Use the technique described in step 7 if you need to tidy it up further. Not only do our guitars now sound bigger and more impressive, but theyve been pushed to the sides, making space for the drums, bass and vocals to sit centrally.

38 / COMPUTER MUSIC / November 2013

> make music now / clean up your recordings

Multiband processing
Multiband processors split the signal into multiple frequency bands at user-defined crossover points. For example, setting the crossover of a 2-band processor to 1kHz will give us one band covering everything below 1kHz and another for all frequencies above 1kHz. So in a multiband compressor the most ubiquitous of multiband effects wed be able to apply, say, heavy compression to the lower band, and only a little up top. Most DAWs have multiband dynamics processors built in, and there are plenty of feature-packed third-party plugins available too. MeldaProduction deserve special mention here, as they offer by far the most extensive array of multiband plugins on the market. Some engineers avoid multiband processing for mix work, saying that if you have a good mix, theyre just not necessary. Thats all well and good if you only ever get perfectly recorded material to work with, but for the rest of us, multiband processing can be hugely useful for controlling unpredictable resonances, getting a grip on signals with constantly shifting tonal balance, adding targeted processing to mixed sound sources, and much more. Take, for example, a vocal thats been tracked in a living room. The shape of the room and its acoustic properties will mean that it reflects some frequencies more than others, resulting in unwanted peaks at certain points in the performance. While this might not perhaps be all that obvious to the untrained ear were used to hearing sounds in a natural environment after all itll be clearly visible in a spectral analyser, and will make the vocal hard to mix. The obvious solution is to use an EQ plugin to notch out the offending peaks, but while this works fine on some sources, its not really viable in this case, as the behaviour of the stand-out harmonics is just too erratic and varies greatly from note to note. A more elegant way to deal with unpredictable resonances is to use multiband compression. This enables us to reduce the dynamic range of a specific range within the frequency spectrum, reducing volume spikes while maintaining the overall timbre of the sound. Closely related to the multiband compressor, the dynamic equaliser (a great example being Voxengos GlissEQ, the demo of which can be downloaded from www.voxengo.com) responds to incoming volume level, increasing or reducing the amount of cut or boost applied to each band depending on whether the signal level hits a specified threshold.

> Step by step

5. Using multiband dynamics to tame resonances in a vocal

Recording audio in environments that are less than acoustically perfect can have the effect of boosting certain frequencies and suppressing others. Our vocal has been recorded in a front room rather than a recording studio, and using BlueCat Audio FreqAnalyst to examine it, we can see unwanted spikes at certain frequencies, particularly around 1kHz, where things seem decidedly erratic.

We could use EQ to lower these peaks, but this process would be difficult to get right and could have a detrimental effect on the timbre of the vocal. A more natural-sounding solution is to use a multiband compressor to control the level of just the problem frequencies. Insert MMultiBandDynamics (demo at www.meldaproduction.com) on the vocal track.

Select the 3-band compressor preset from the list on the left. To access the plugins parameters, click the Edit button at the top of the interface. The graphical display in the top left hand corner shows us the presets three bands. Click Band 2 (the central one) to bring up its parameters.

Drag the left and right edges of the band to set its frequency range. Set the lower crossover to 700Hz and the higher one to 2kHz. This focuses the band on the range thats giving us trouble. Click the Link button at the top of the Band 2 panel so that we can adjust this regions parameters without affecting the others. Click the S button on the band to hear how it sounds soloed.

By default the Threshold is set to -30dB, so we can reduce the areas dynamic range by simply turning up the Ratio. Move FreqAnalyst CM to the insert slot after MMultiBandDynamics to see the effect on the frequency response, and gradually turn up the Ratio. At 2.00:1 or so, the peaks are less jumpy. Turn the bands Output gain up to 2dB. (Audio: Multiband compressed vocal.wav)

As weve unlinked the bands, we can adjust the Ratio of Bands 1 and 3 separately, to apply gentler compression to them. We can now mix the vocal in the usual way. For instance, we add some width with MHarmonizerCM turning up the Width parameter in the Global Pitch panel slightly 0.35 semitones or so gives us a pretty sweet widening effect. (Audio: Stereo vocal.wav)

40 / COMPUTER MUSIC / November 2013

clean up your recordings / make music now <

The problem with drums


The more mics you use to record a drum kit, the higher the likelihood that youll run into phase problems. This is because any number of the signals picked up by all those mics could well be out of phase with each other, leading to a thinsounding drum kit with an overall lack of impact. Getting a good phase relationship between mics is an essential part of obtaining a great drum sound, and is usually achieved by moving the mics around and checking and adjusting the phase before recording. If you discover phase problems after recording, however, the easiest remedy is to try flipping the phase of individual tracks in your DAW. If you know two tracks are out of phase, simply flip one so that they work together to reinforce the sound, rather than cancelling each other out. You can zoom right in on the waveform to view whether the waves are in positive or negative phase at any given point. If youve checked the phase and your drums still sound thin, another technique is to manually align the room or overhead mic tracks with the kick and snare tracks by nudging them backwards in your DAW by a few milliseconds, shortening the natural delay caused by the sound having to travel the extra distance to the mic. This can work wonders in terms of tightening the sound, but the technique does run the risk of messing with the natural phase relationships between the tracks. If you do decide to try it, then, proceed with caution! Another potential problem with multiple drum kit mics is crosstalk, or bleed, where the sounds of neighbouring kit pieces are picked up by each mic so you end up with loud hi-hats in the snare mic, for example. The traditional approach to taming this has been to use noise gates, which can be tricky, but the current generation of dynamic noise reduction plugins can be useful in this area too.

Be aware of potential phase issues when using multi-miced drum parts

> Step by step

6. Fixing resonance and hi-hat bleed on a snare track

Youll find our drum tracks in the STEMS_ folder theyre all at 144BPM. The snare in our drum track is quite ringy and resonant, and its mic has picked up a lot of hi-hat bleed. Lets see if we can alleviate these problems. Solo the snare track and insert an EQ were using EQuality again, with its built-in analyser enabled. Practically any EQ and analyser combination will work for this, though.

To locate the troublesome resonances, we use the tried and tested seek and destroy method. Simply increase the gain of one band and set a high Q, so that it boosts a very narrow frequency range. Play the snare on a loop and sweep the frequency through the spectrum until the main resonant pitch of the snare becomes really loud, ringing as much as possible here, this happens at 482Hz.

Next, use the Gain control to pull the offending frequency down in level until its no longer bothersome. Now we add another band, repeating step 2 to find the next harmonic a little further up the curve. We repeat the process until were satisfied with the results. While were EQing, we can also make the snare punchier by boosting its fundamental frequency in this case, its around 225Hz.

POWER TIP

>Variable GR
Some EQs let you control the overall strength of the EQ curve. In EQuality, for instance, theres the Range slider, which allows you to increase the maximum possible gain reduction of all bands, from the default -18dB up to a huge -36dB, or right down to 0dB, where cuts/boosts have no effect. Once youve located the troublesome resonant frequencies and got your curve in place, you can set the EQ to be as extreme or as subtle as you like useful if your overaggressive notching has robbed the instrument of some of its tone.

Hi-hat bleed is harder to deal with, but one solution is to use a denoiser plugin capable of learning the noise profile of a sound for removal were using iZotopes RX2 DeNoiser. We isolate a couple of single hi-hats in the snare track from bars 4 and 5, trim off any adjacent cymbal or snare hits, then copy and repeat them to create a four-bar strip of hi-hats that we can use to feed the denoiser.

We insert RX2 DeNoiser on the snare track, enable Learn mode and feed it a few seconds of hi-hat. Once the plugin has learned the noise profile, it reduces those frequencies on subsequent playback. This kind of processing is always a compromise, so tweak the settings to find something suitable for your mix. If the snare sounds too dull, try inserting a second EQ and boosting the 5-6kHz range.

November 2013 / COMPUTER MUSIC / 41

> make music now / clean up your recordings


> Step by step
7. Fixing the timing of multitrack drums with Logic Pros Flex Time

Logic Pro 9 and X feature powerful timing correction in the form of Flex Time. Lets use it to correct the timing of a dodgy drum track. (If the rest of your parts have been played to fit the timing of the drums, youll probably need to adjust the timing of those too, otherwise theyll end up out of sync with the corrected drums.) The project tempo has to be set correctly before we start, so set it to 144BPM now.

Drag the drum parts into the arrangement, telling Logic to create new tracks for each part. Delete any unused audio tracks. For proper multitrack editing of these drum tracks, we need to group them. To do this, press X to bring up the mixer, then hold the Alt key and click each of the drum tracks Group slots. This will set them all to Group 1.

Bring up the Group settings by rightclicking a Group slot and selecting Open Group Settings. Click the Editing (Selection) checkbox to tell Logic that we want to edit all the parts in this Group as a single entity. Keep the Phase-Locked Audio checkbox selected this ensures that audio across the tracks stays in phase, which is crucial for preserving punch and coherence in multitracked drums.

Close the Group Settings menu, and click the Flex Time button at the top of the arrange window to activate Flex Time. By default each tracks Flex Time modes is set to Monophonic. This isnt ideal for drums so change it to Slicing, which better preserves transients. As weve told Logic we want to edit all the tracks together, you only have to do this for one track and the others will follow.

Double-click the kick drum track to open its audio editor, then click the File button to view the waveform. From the Audio File menu, select Detect Transients. A series of orange transient detection markers will appear. Check the results by scrolling along and making sure all the kicks have been detected it seems to have worked!

Do the same for the snare track. Youll notice that not all of the snares are detected this time, and thus dont have transient markers. You can use the + and buttons to adjust the sensitivity of the transient detection, but a more accurate way is to hold Alt and click the undetected hits to add markers manually. Use the Eraser Tool to delete unwanted markers.

Before using Flex Time quantisation, we need to tell Logic which tracks we want to use as references for the timing of the other tracks. Generally, the kick and snare are the core of any drum beat, so we deselect the green Q buttons in the Track List on every track apart from those two.

If you turn on the metronome, youll hear that the tempo of the drum track drifts slightly. We can remedy this by selecting the kick track and setting its Quantize to 1/16th Note in the Inspector. This improves the timing, particularly in bar 7. (Audio: 8. Bar 7 quantised.wav)

The timing of some of the fills might suffer from the quantisation, but this can be fixed by zooming in and dragging the hits into the correct positions. For example, at the start of bar 10 the kick with the crash on it has been moved just before the beat by the quantisation. Simply drag the kick back onto the start of bar 10, and the rest of the drums will move with it. (Audio: 9. Fill timing fixed.wav)

32 / COMPUTER MUSIC / November 2013

> make music now / clean up your recordings


> Step by step
8. Replacing drums with high-quality triggered samples

You can augment or replace recorded drums with high-quality samples. This is called drum triggering, and many DAWs offer it, making it easy to trigger drum samples from audio tracks. Were using Cubase. With our drum tracks loaded and the project at 144BPM, load the instrument you want to provide the new sound (BFD Eco in this example) on a new MIDI instrument track.

Double-click the kick drum audio region to open it in the sample editor. On the left, click the Hitpoints tab, then turn up the Threshold setting to generate vertical lines wherever Cubase detects a kick drum hit. Use the horizontal lines as a guide youre aiming to get the peaks of all the required hits to cross the horizontal threshold bar.

Adjust the threshold until every kick drum hit has a vertical line just to the left of it. If necessary, you can manually add or remove hit points by clicking Edit Hitpoints and Alt+clicking to do the deed. Now click the Create MIDI Notes option at the bottom of the Hitpoints panel. This brings up a set of options with which to fine-tune the process.

POWER TIP

>Overhead xing
With our parts being triggered by MIDI regions, its possible to fix dodgy timing by quantising them. If you do this, however, remember that even if you mute the kick and snare mics, those drums may still be heard on the overheads, potentially creating flamming artifacts. This can be alleviated by editing the overhead tracks: group them together, generate transient hitpoints, then use your DAWs audio quantise feature, set to the same grid setting as your MIDI part, to line them up. Alternatively, quantise only the offending hits.

You can choose to generate notes at a fixed velocity, but if you want to preserve the dynamics of the original part, leave Dynamic Velocity engaged for now, as we have here. Leave the Pitch at C1 (the standard note for triggering a MIDI kick drum), and set the MIDI note Length parameter to a short value weve chosen 1/32. Finally, choose the First Selected Track option, then click OK.

The resulting MIDI region appears on our BFD Eco track. To keep things simple, well use a separate instance of BFD Eco for each piece of the kit (kick, snare, etc), so on this track we can mute the other drums giving us just the sound of the kick drum, without any bleed from the other pieces. If you can unload the other drums altogether, then even better, as youll conserve RAM and CPU cycles.

Because we chose Dynamic Velocity mode in step 4, the volume of each beat has been converted to note velocity; however, all the hits are too quiet. Open the MIDI part in the piano roll editor and adjust the velocities as you see fit we select all the notes then drag their velocities up to the top of the velocity graph, so that the hits are loud but still have some variation.

The beauty of triggering is that youre free to change the drum sound by simply loading a new sample into your virtual drum kit instrument. Put a section into loop play mode and flick through the available samples until you find one that suits the track. You can then adjust the levels of the miced and triggered kick to get the perfect balance, or simply mute the mic altogether, as is quite common.

Here we have a triggered snare drum part, created in the same way as our kick. You can blend in the miced snare if you like, to help maintain dynamic feel and realism. If you hear phasing artefacts between the miced and triggered sound, try nudging the MIDI track forwards or backwards in time by a few milliseconds until it goes away. Our audio examples are in the Tutorial Files folder.

44 / COMPUTER MUSIC / November 2013

clean up your recordings / make music now <

Restoration software
The term audio restoration brings to mind tedious and unsexy processes such as repairing wax cylinder recordings of Thomas Edisons dog barking. Despite the negative connotation of its name, audio restoration software is wholly relevant to todays musician, with the ability to salvage dodgy recordings. Say you have a take thats perfectly performed, but has been recorded too hot and as a result is clipped in places. Once upon a time, youd have to simply accept this and move on but not any more! Thanks to plugins like iZotope RX (version 3 of which was released as we were going to press), you can simply run the audio file through a declipper process to automatically restore the missing waveform tops, resulting in a perfect signal. You can download the demo of RX3 at www.izotope.com.

Feel the noise


An issue that often comes up with recordings taken from tape is hiss, which in the past could only be remedied with EQ. Even in todays tapeless studio, background noise can creep in from various sources, and the problem with broadband noise is that it covers a relatively wide frequency range, so the necessary EQ will usually have a detrimental effect on the signal as well. Thankfully, many audio restoration plugins and audio editors now have facilities to deal with constant background noise, including freeware favourite Audacity (get it at audacity. sourceforge.net). In Audacity (the procedure is similar in most alternatives), highlight a section of the recording that features just the noise and none of the signal, select EffectsNoise Reduction and click Get Noise Profile. Audacity analyses the noise, and you can then use the various sensitivity and frequency parameters to tailor the level of noise reduction for the best possible result. Once youre done, click the OK button to process the waveform, leaving you with a hopefully hiss-free version.
iZotope RX3 offers powerful spectral editing and restoration effects that make light of hum, noise, clicks and more

The great thing about these learning noise reduction systems is that you can use them to remove any undesirable signal elements, particularly those that are constant in nature, such as mains hum or fan noise. It can even work on musical sounds, as we demonstrated on p41 with our hi-hat bleed.

Record breakers
If youre taking samples from vinyl records, then youre almost certainly going to end up with pops and clicks in your recording. While its possible to remove these unwanted glitches manually by zooming in on the waveform in an audio editor and using a pen tool to redraw it, this can be an extremely long-winded process if youre dealing with anything more than a very short sample. Those in a hurry can

take advantage of the pop and click removal functions offered by software such as iZotope RX and Adobe Audition. You can download a demo of the latter from www.adobe.com. Another problem that you might encounter when dealing with tracks recorded on malfunctioning or poor-quality kit is DC offset, where the horizontal centre of the waveform is offset from the zero line, resulting in a variety of issues, including causing compressors and distortion plugins to respond strangely. If you have a recording that appears to have a DC offset problem (itll be visible in your audio editor), simply apply your editor or DAWs DC offset correction process to automatically realign the waveform. Alternatively, applying a high-pass filter set to its lowest possible frequency should also correct the issue.

> Step by step

9. Patching up dropouts and other unusable sections of audio

Load up the vocal track from STEMS_ in a project at 144BPM. Check out the vocal track and take a listen to bar 50. Theres a potentially disastrous problem: a technical issue has resulted in a silent gap in the vocal! We need to find something to fill it, so lets scour the vocal for a section with another sustained note at the same pitch. (Audio: 1. Vocal gap.wav)

Luckily theres something suitable at the end of bar 52. Cut out the missing section at bar 50 and delete it, then slice out a section of the sustained note at the end of bar 52 and copy it to a new track, at the point where the vocal cuts out. Make sure to route this new track to the same destination as the main vocal track, eg, to a vocal processing bus.

While the pitch is constant, the joins between sections of vocal are obvious because of the clicks where the audio stops and starts. We can fix this with the judicious application of fades at the start and end of the repaired section. With the fades set up to keep a consistent volume level as shown, the join should be pretty much impossible to detect. (Audio: 3. Fixed vocal.wav)

November 2013 / COMPUTER MUSIC / 45

> make music now / clean up your recordings

The nal mixdown


Throughout our tutorials, weve worked our way through practically every element of our demo mix, knocking audio gremlins on the head left, right and centre. With that done, its time to move on to the mixing stage; however, when dealing with lowgrade source material like this, a few special considerations have to be made. through, so you may prefer to back off the corrective EQ. Whats vital is that you can separate these processes, to make the distinction between corrective and creative processing. For instance, if you prefer the sound of compression before EQ when mixing, dont make the mistake of putting the compressor before your corrective EQ. If you do, it will then react to the highly unbalanced rogue elements in the original, uncorrected signal. shape means it will not dull neighbouring frequencies. Try it yourself! We then applied de-essing, our multiband compression, some gentle shaping EQ, the MHarmonizerCM widening effect, u-hes Satin, and finally a slapback pingpong delay. Drums were simple enough: we replaced the kick and snare with EZdrummer equivalents, applied our snare-resonance busting techniques to the toms, then sent all the drum channels to their own bus to be phattened up by Slate Digitals Virtual Tape Machines and Universal Audios FATSO for UAD-2. In the end, we applied quantising only to the drum fills, to preserve the groove during the other parts. We also used cutand-paste editing to add extra cymbals to liven up the repetitive performance! On to the guitars. We added gating with FabFilter Pro-G to these to tighten up the stops, then our corrective EQ followed by UAs API 560 for tasty tonal shaping, cutting a little at 1kHz to make the upper-mid presence frequencies stand out. A touch of CM-COMP 87 compression and Slate VTM tape saturation rounded it off. To make our bass tone more cohesive, we routed all three tracks (the original audio, the low end synth, and the excited top end) to their own bus, where we ran the lot through Studio Devils Virtual Bass Amp, then standard EQ, compression and tape simulation. This made it sound less like three individual elements and more like one solid bass sound. To add a unified sense of space, we used a reverb effect on an aux send channel, sending the synth, vocal, snare drum and guitars to it, with a little EQ on the reverb to roll off its lows and highs. On the master bus, CM-EQUA 87 gives a gentle 2dB dip around 1kHz, with stronger shaping from Universal Audios Manley Massive Passive EQ plugin. For dynamic control, we opted for Slate Digitals VTM and Virtual Buss Compressors, and finally, iZotopes Ozone Maximizer for limiting. While our track might still not be perfect, and the mix weve done here is a quick one purely for demonstration purposes, were sure youll agree that its a massive improvement over the raw tracks, now sounding quite rich and smooth. Job done!

Keep em separated
As we stressed earlier in this article, it makes sense to tackle problem audio in two stages: first, apply plugins and editing processes that will correct for obvious deficiencies and problems with the sound; second, go about your normal mixing routine of using sweetening EQ, phattening compression, sumptuous reverb, and so on, adding any such plugins after the corrective ones. Naturally, there is going to be some back and forth between the two stages for instance, applying mix compression to the snare drum may make any annoying resonances even more audible, so you may want to go back to your notching EQ and pull down some of the cuts a little more. Or conversely, you might find that in the context of a mix, the twangy ringing of the snare adds character and helps it to cut

In the mix
In our Tutorial Files folder, youll find our final mixdown of our demo project, so lets quickly run through our mixing procedure for this tune. The synth track needed no more work we simply treated it to a little tape saturation from u-hes Satin processor. When working on our vocal, however, further pokey resonances became apparent, this time in a narrow band centred around 2150Hz. To attenuate these harsh frequencies, we used one of our secret weapons: DDMF IIEQ Pro CMs Butterworth Band Shelf with the Order setting turned up. This unusual flat-topped filter shape is extremely useful for busting clusters of harsh frequencies, as its square

The Butterworth Band Shelf filter in DDMFs IIEQ Pro CM is a fantastic harshness-busting tool, and the built-in analyser of the commercial version shown here makes fixing up your frequencies even easier

Mixed signals
One thing that we havent touched upon in this article is how to deal with situations where youve wound up with audio that is actually a recording of more than one instrument. For instance, guitar and vocals, or piano and percussion, etc. Aside from applying the corrective techniques described elsewhere in this article, you may need to find creative ways of applying targeted processing to the elements of that recording. The most common problem is that the instruments are not balanced in level, so first try some fairly quick compression on the entire signal to see if you can pull down the louder instruments peaks to make it seem quieter. If that doesnt work, instead turn to multiband dynamics, to control overbearing elements in certain frequency ranges and bring up low levels in others. For instance, if you have a combined guitar and shaker part, and the shaker is too loud, try quick, highratio compression on the upper (shaker) frequencies to sharply pull its level down on each hit, and on the lower frequencies, gentler compression to boost the guitars musical tones. Another possible route is to duplicate your source audio track and use filtering and gating tricks to isolate elements of the signal on each of the tracks. In our example, you could use filters to focus on the shaker, then set up a gate (or edit the audio by hand) so that it only opens on each shake. Now you can do things like send the isolated shaker to a reverb to give it its own sense of space.

46 / COMPUTER MUSIC / November 2013

> make music now / clean up your recordings

12 tips and tricks


PATCHING UP
If theres a word in your vocal thats just plain wrong (ie, a wrong lyric or mispronounced word) and you dont have any extra takes to comp from, try reconstructing the missing word using syllables taken from other words elsewhere in the song. Is the same phrase sung in the next chorus, for example? Your DAW has the editing power to copy, paste, tune and tweak the smallest fragments of audio, so its always worth a try. Bear in mind that many consonants (such as T, C, P, K, and S) are unpitched, so you dont need to take tuning into account for these they can be taken from anywhere they occur in the song, regardless of the note the word is sung at!

MATCH OF THE DAY


If you dont have a matching EQ (one that superimposes an EQ curve from a goodsounding source onto a different one to make it sound similar), try using two spectrum analysers, one on the track you want to emulate, the other on the track you want to adjust, and using EQ to make the displays match up manually. Even better, some analysers have multiple inputs so that you can see curves from two sources overlaid on one display. Or if your analyser has a stereo mode, hard pan the two signals left and right and route them into the analyser.
Zynaptiqs magical Unfilter is designed to remove the effects of filtering and generally wonky frequency responses

is not playing precisely with the click. So if you quantise only the drums, or just the instruments, you might kill the groove!

plugins, re-rendering if necessary.

TOOLS OF THE TRADE


There are some fantastic dedicated tools out there for restoring, rescuing and repairing dodgy audio. Zynaptiqs Unveil and Unfilter allow you to remove reverb and dodgy frequency responses from signals respectively, while dedicated correction tools like Celemony Melodyne and Antares AutoTune can be used to adjust the pitch/ timing of performances.

MAKE IT CLICK
Our demo project was recorded to a metronome click, which helped our editing processes greatly. If your audio hasnt been played to a click, though, you might want to spend some time building a tempo map before getting stuck into editing. Some DAWs have tools to help with this, or even features to build a map by analysing the audio.

FEEL DENOISE
If youre using a noise reduction plugin such as iZotopes RX2 Denoiser on a sound, be sure to place it before any compressors in the plugin chain. This is because compressors raise the level of the noise floor relative to the required audio, not to mention making the noise floor move dynamically, which would make the job of any noise reduction processors placed afterward more difficult.

MAKE EM BOUNCE
When youve got your corrective processes and plugins in place, consider rendering/ bouncing the track to a fresh audio file to cut down on plugin clutter and general confusion. You can then delete the original (keeping a backup in case you need to go back to it), or just mute it and disable the

KNOW YOUR LIMITS


Theres a limit to how far badly recorded tracks can be made to sound better, so dont drive yourself crazy trying to make yours sound like your favourite commercial records. Indeed, sometimes its best to embrace the weirdness of your audio and make it a feature of the track.

GLOBAL APPLICATIONS
Our demo project throughout this article is a pop/rock track, but dont let that stop you applying our techniques to other styles. For instance, the approach we used to flatten out our guitar sound in the first tutorial can also be used to big-up dubstep synths and the like.

PERFORMANCE IS EVERYTHING
If youve got the opportunity to rerecord your parts with a better sound, then go for it! But dont ever forget that a great performance with a shoddy sound is often preferable to a pristine recording of a mediocre playthrough. A perfect example: the vocal on Christina Aguileras Beautiful was recorded as a rough demo take, complete with headphone spill. The raw emotion of that take could not be bettered by later studio efforts, so the producers went with the demo version.

TRIGGER HAPPY
We looked at drum triggering, but why not trigger other instruments? Most DAWs now have some kind of audio-to-MIDI conversion for pitched instruments, which could be a possible solution when faced with a part whose sound is otherwise beyond salvation. Simply convert the part to MIDI and use a high-quality virtual instrument instead. For converting polyphonic parts, youll have to look into Celemonys Melodyne. Alternatively, get stuck in and reprogram the part by hand time-consuming, but possibly less so than the work needed to make a badly played/recorded part work in a mix!

RECOMMENDED READING
Obviously, the best way to clean up your recordings is simply to get them right in the first place! To help you do this, we put together Computer Music Special 62: Home Recording The Ultimate Guide. Packed with tutorials on everything from capturing drums, guitars and vocals, to turning your house into a recording studio and multitracking a live band on your iPad, its a must-have for anyone even vaguely serious about recording. Its available in shops, at www.myfavouritemagazines.com and on Zinio and Apple Newsstand now.

OFF THE GRID


As we touched upon elsewhere, in a real recording, musicians generally play along to the drummer, even if the drummer themself 48 / COMPUTER MUSIC / November 2013
Make great recordings from the off with our Special

Вам также может понравиться