These production tips and tricks are designed to help you make the most of your music. Covering everything from mic technique to post-production, we hope you find them useful!
© Tenth Egg Productions 2009
Having covered the basic nuts and bolts of audio editing in part 1, lets now take a look at some of the more advanced stuff, the first of which is Audio Restoration.
Audio Restoration:
Ideally we want to be working with nice clean recordings free from any quality issues, but this isn't always possible. Removing background noise, digital clicks and pops, DC-offsets, distortion and other unwanted audio artifacts are all forms of what is termed 'audio restoration'. The sophistication of the tools will vary widely from one audio editor to another, but most of the same basic principals apply. Ideally restoration would be first in your editing workflow, ahead of any chopping up or plug-in processing. This is because tools such as noise-removers will work far more effectively on the original audio where the unwanted noise will be more consistent and therefore easier to isolate - imagine how much harder it would be to remove a click or pop after the audio has been through a reverb plug-in for example. You may also find that all further processing down the chain will help to mask any negative side-effects of the restoration work carried out.
Clicks and Pops are usually a result of bad edits that don't follow the zero-crossing rule, but they can equally occur during recording from digital clock problems or mains power interference. Rather than drop a click/pop remover across an entire track as a plug-in or process the whole file - both of which are likely to lead to false-positives (i.e. stuff that isn't a click or pop being identified as such) - it’s much better to identify individual clicks and pops by ear and then fix though particular areas in isolation. So in the example below using Soundtrack Pro we have identified a digital click by ear and selected the area around. Soundtrack Pro finds the problem area (highlighted in red) and we then just click 'Fix' to remove it. If the problem doesn't go away first time round, don't be afraid to analyze and process the same area again as this will often do the trick.
Tip 21 : Digital Audio Editing - Part 2
music mastering
Tips and tricks from our engineers to help you with your music making
Industry news, useful links and what’s going on at Tenth Egg
Share us with your friends
A different approach to music production.
Share us with your friends
DC Offset describes a situation where the audio waveform is shifted slightly toward the positive or negative scale rather than oscillating symmetrically around the centre line. It's a common issue in cheap soundcards and although it doesn't cause any audible problems on its own it can make further editing a bit of a nightmare and limit the effectiveness of any dynamics processing.
Again it’s pretty straightforward to remove and in this case, because any offset will tend to run the length of an audio file, it is best to analyze and process the file as a whole.
Noise Removal, namely hiss and hum is perhaps less needed in these days of low noise floor mics, pre-amps and convertors, but it's good to know that when you need it the tools are pretty effective. The Noise-Print removal technique can be particularly effective in cases where the unwanted noise is consistent throughout the audio. It works by analysing a sample of noise and using this to identify frequencies to reduce or remove throughout the track. Needless to say you stand a much better chance of being able to transparently remove the offending noise if you are working on an unprocessed original as any post-processing may effect the consistency of the background noise. You also need to have a portion of audio that you can use to extract a decent noise print from, which ideally contains the noise alone.
Effective though it is, you should accept that you probably won't be able to remove 100% of the noise, that slider ranging from 0-100% noise reduction shouldn't really be pushed much past the 50% mark. A useful option when trying to judge the optimum level of reduction is to listen to the noise actually being removed. Most plug-ins will provide this monitoring option (seen on the right as the ‘Noise Only’ checkbox) and what you want to listen out for is any stray traces of the actual instrument being caught. If you don't have this option then as you adjust the reduction and threshold controls you should listen out for audio artifacts being introduced at the top end. In particular a warbling sound, not dissimilar in sound to heavy mp3 compression, which engineers will affectionately refer to as 'birdies'.
Clipped or Distorted audio is about the hardest thing to repair and while most tools will yield some improvement you shouldn't expect miracle results, so it should only really be relied on when re-recording isn't an option. Repair tools will look out for consecutive samples at the same level and then attempt a combination of peak reconstruction and/or tonal softening to help smooth the worst side effects of distortion.
In all the restoration tasks above I would use a standalone dedicated audio editor such as Soundtrack Pro to perform any restoration before moving back into Logic and continuing with more creative editing.
Correcting timing:
Improving the musical timing of a performance is a very common editing task and modern sequencers have developed various tools to help you do just that. One thing that you can do at the recording stage to make life a lot easier during editing is record to a click, not just because you can use your sequencer's beat/bar grid for a reference for manual edits, but also because you can then use some of the automatic timing correctors such as the Quantize Engine in Logic or Beat Detective in Pro Tools. If you haven't used a click then all is not lost! If the tempo is reasonably stable then it should be possible to work in sections to tighten up the timing using an average tempo.
But let's start with the manual approach first as often this is what you'll have to fall back on. As we saw last month in Audio Editing Part One most sequencers offer two different approaches to editing namely what we called 'Overview' editing and 'Sample view'. Timing correction can be carried out in either but I would always start in 'Overview' because it's much easiest to play around with your edit and undo what you've done if it doesn't work out.
So having identified a timing problem the first thing to do is to splice around that particular note or phrase: Zoom right in and try and chop your audio as precisely as possible and then slide the chopped-out region back or forward to move it into time. It's usually pretty easy to identify the start of a note and then line that up with the nearest beat. - One thing not to get caught out with when working with guitar recordings is that often part of a chord strum can naturally fall before the beat and moving the very start of that strum onto the bar will actually make the timing sound worse - In moving our note or phrase we will inevitably be left with a gap or overlap at either end of the edit. To smooth the transition between overlapping audio it's relatively straight-forward to drop in a short cross-fade which 9 times out of 10 will do the job and also get rid of any clicks or pops if our original edit wasn't quite on a zero crossing point.
To fill out gaps the quickest solution when moving an individual note is to use a time-stretch to fill any resulting gaps. In Logic Pro this is as simple as moving your mouse to the bottom right of the particular audio region and alt-dragging out the end. Logic will then stretch the audio region out to the desired length without changing the pitch and as long as you don't go too far the audio degradation is pretty negligible. When working on a phrase you'll just want to stretch out the last note to fill the gap, so first you must splice it off using the scissor tool.
The other option is to copy a portion of the last note and paste it into the gap using cross-fades at the beginning and end. This can sometimes work better on larger gaps than the time-stretch, but also don't be afraid to use a combination of both!
When identifying timing problems it's often best to look at your audio from a bird's eye view first and try to identify any overall patterns. If in general a performance is late but there are occasional notes within that that are early you'll save a lot of time by shifting the region as a whole to correct the overall lateness and then zooming in to focus on just the early notes rather than working through on a note-by-note basis.
Another important point to note is that when editing any sound-source which has been captured with multiple mics (e.g. drum kit) we need to always make timing adjustments to the tracks as a group to retain phase coherence. Moving the position of part of one track in isolation (e.g. the snare) would push the audio out of line with the spill of that part in other tracks (e.g. the overheads).
Ok, so having learnt the manual way, lets see how the computer can do some of the hard work for us! With a bit of a helping hand from us something like the Audio Quantize Engine in Logic can do all that splicing, shifting and stretching in a couple of seconds. Much like MIDI Quantizing, Audio Quantizing is all about moving note events to the nearest bar/beat/division except that it has to guess where the start of each note is and this is where it needs help from us. By adjusting various parameters (see below) and manually removing false-positives we can help Logic identify each note in an audio region and after selecting a suitable quantize value we just hit 'Process' and let Logic do the rest. Typically the more percussive the instrument the better this will work and in these cases automatic timing correction can be a real time-saver.
Adjusting tempo
Thankfully this is pretty much a one-click process, but as with all time-stetching / compressing there are limits to what can be achieved before the negative audio artifacts become a problem. As as a good rule of thumb when working on an entire mix you should be able to get away with 5bpm slower or 10bpm faster without too much compromise, depending of course on the material (electronic music will be more forgiving than acoustic). There are mixed opinions about whether it's better to adjust the tempo of each mix element individually rather than the mix as a whole. Potentially mix processing and summing of individually stretched elements could help mask some of the negative artifacts and give more natural results, but it is also a lot more time consuming, so it's probably worth trying a tempo adjustment on the entire mixdown first.
One piece of good practice is also to leave some headroom available in the audio file as stretching can have an affect on level, creating higher peaks. Therefore it's always a good idea to carry out stretching ahead of mastering.
Much the same process applies when adjusting the tempo of individual loops or sections within a track. Many shop-bought loops won't need time-stretching at all because they come in a special format that is tempo aware and will automatically adjust as you change the tempo of your project. We'll be looking at how to use Apple's Loop Utility and Propellerhead's Recycle to make your own intelligent loops in a future production tip, but essentially they perform a similar analysis process to the Quantize Engine - trying to identify individual beats within a loop - to create tiny audio splices which can be squashed together or moved apart individually to adjust tempo.
Correcting Pitch
Correcting pitch is the last editing process we're going to look at and probably the most controversial. Really pitch correction should be seen in the same light as restoration, a tool that can be used in extreme cases to salvage a performance when re-recording is just not an option. I would avoid letting poor tuning slip through at the recording stage because you know you can 'fix it in the mix'. Just as with time-stretching there is a limit to what can be achieved without compromising audio quality and ironically pitch correction works better on good performers than bad for that very reason.
There are two basic types of pitch adjustment, 'shifts' and 'correction'. Shifts move the entire selection up or down by the desired amount and are good for correcting entire tracks which are consistently sharp or flat, or manually adjusting individual notes in otherwise acceptable recordings. Although you can use a plug-in to shift the pitch on the fly, better results are usually obtained by using the tools built into your sequencer's sample editor. In Logic this is called the 'Time and Pitch Machine' and having selected the portion of audio to process the controls are pretty basic. The principal one you are interested in is the 'transposition' measured in Cents, where +/- 100 Cents represents a semi-tone shift in pitch. Although processing time will be extended it's worth switching in the Harmonic Correction to help maintain the character of the original sound when working with monophonic material as this can produce more natural results. Even using harmonic correction you'll probably find that going much beyond a couple of semi-tones up or down starts to sound unnatural and compromises audio quality.
Correction is a more dynamic process and involves the plug-in or processor tracking the pitch of the audio as it passes through and automatically adjusting the amount of pitch shift to push or pull the performance in tune. This is how the infamous Auto-Tune works and is the best tool when there are a whole host of tuning issues in a performance. Again the more information you can give the tool the better the job it will do so if you know the key/scale used enter these. In general you should also avoid placing the plug-in across the length of the song because of the potential for false-positives. It's best to splice out sections with pitching problems onto a separate track and place the pitch corrector only on that channel. And remember auto pitch correction will only work on monophonic material so anything with chords of simultaneous notes isn't going to be fixable this way.
The aim of this two part series has been to introduce you to the main audio editing techniques and provide some basic guidelines. There's plenty more detail to cover on each individual topic and we'll be looking at each again in more depth in future production tips, meanwhile - happy editing!
Click and Pop removal in Soundtrack Pro: Having identified a click by ear, selected that area of the waveform and run the click and pop analysis Soundtrack identifies the problem highlighted in red. It then just a case of selecting ‘fix all’ on the left and Soundtrack removes it with ease.
Manual correction: In this example we can clearly see that the third note is late. So we chop around it and shift it back so that it lines up with the bar/beat grid in Logic. In this case because there are nice big gaps between each note in the phrase, cross-fades / stretching aren’t needed to smooth the edit.
Stretch to fill: Here Logic’s stretch tool is used to end the first region in an edit where the two notes need to flow into each other. A simple cross-fade helps smooth the transition.
Usually we work with a current tempo and destination tempo. By default the current tempo will be set to that of your project, so if you didn't record to a click and haven't set this it's worth quickly working out a rough tempo. In Logic Pro you can also adjust by length in samples, length in SMPTE and length in bars and as you adjust one all the others will automatically change so you can see how much longer or shorter your tempo adjustment are making the audio region. Another function that is found on many time-stretchers is the option to select what is called the 'Algorithm';. The same stretching method won't work as well for all material so having some control over this by selecting something like 'Beats Only', 'Pad' or 'Monophonic' will help get the most transparent results.
Pictured above: The Tempo section of Logic Pro’s Time and Pitch Machine. We can see that as well just entering a new destination tempo we can dial in our change as a percentage, new length in sample, SMPTE time or even as bars.
