Home | Technical | Audio | Sampling Tips and Techniques
Sampling Tips and Techniques
Overview of Sampling
Samples are clips of audio data that can be looped and manipulated to form sections of music. Traditionally the sampler was a separate module (originally using analogue magnetic tape, but now digital RAM). Modern PC's often deal with samples internally, either processing them directly from the hard disk or by storing them in system or soundcard RAM.
The pitch is altered by changing the speed of playback or the frequency of the sample. Samples can be used as looped sections or as the basis for individual notes. Wavetable synthesis and soundfonts use small samples that can be combined and manipulated to form a large variety of instrument patches.
Finding and Creating Samples
Although many samples are readily available both commercially and from the public domain creating your own will generally result in a more original sound. Libraries of samples ranging from loops to individual instrument hits can be bought on CD and the relatively high cost of these is accounted for by the fact that you are also buying the right to use them without fear of copyright infringement. You should always take care in obtaining samples from other sources that you will not end up being sued.
When recording your own samples it is important to remember that the amount of RAM available to you for storing them will almost certainly be limited.
The smaller you can make the sample the better.
Clipping off unnecessary silence and reducing the sample rate are two ways of reducing the size of samples.
Some sounds respond better to having the sample rate reduced than others (bass sounds will lose very little at lower sample rates) and so you may need to experiment to find the best compromise.
Below are the absolute minimum sample rates that you can get away with for various sounds:
If you wish to use looped samples (for a drumbeat, for example) it is essential that the looping is accurate. Small inaccuracies may not be correctable at a later stage and can result in poor timing and clicks. This can be even more important if you are looping a part of the sample in order to create a sustaining sound such as a violin.
When the key is first struck the attack portion of the sample plays. The looped section then continues to play whilst the key is held giving unlimited sustain. On releasing the key the decay portion sounds, ending the note.
Sometimes, however hard you try, an audible click or change in tone remains as the sample loops. One way to overcome this is to crossfade the loop. Most good samplers and sampling software will have a crossfade option and this overlaps the beginning and end of the sample to create a smoother loop. Crossfading should not be used as a remedy for clumsy looping.
Another technique for smoothing loops that some sampling software provides is a draw tool. Zoom in on the start and end of the loop and simply draw the data in as a smooth curve to zero (end) or from zero (start). This method effectively creates a zero crossing point at the exact place where the sample loops. Some software will let you search and automatically select zero crossing points but there is no guarantee that they will fall where you want them to.
If you need to fit a sample to loop at a given BPM you can use a timestretch or pitch change function. Timestretch will have the option to keep the sample's original pitch whereas a pitchshift will obviously change the pitch as well as the length of the sample.
Sometimes a sample that has been looped can end up sounding sharp or flat compared to the original even if it has not been deliberately shifted. Often a loop tuner function is available to correct this.
Looping the Unloopable
String ensembles and sustained pad sounds are notoriously difficult to loop accurately and smoothly but a few techniques will help improve your chances. Start by getting the main body of the loop as close as possible. When you are reasonably happy copy this section and then paste it at the end of the loop. While it is still selected, reverse it. This gives you a ping-pong style loop and will often create a seamless, pop-free loop. Beware of phasing on stereo samples when using this technique. To avoid this do the whole thing in mono and then copy the mono sample to both sides of a new stereo sample, adding a slight delay to one side, or altering the volume envelopes for each side to give the sound a subtle L/R movement. Alternatively you could equalise each side differently, perhaps a +3 boost at 4k on one side and a +3 boost at 500hz on the other.
There are many ways of adding effects to samples. Good sampling/wave editing software will usually include a plethora of customisable effects and many of the latest soundcards and samplers have numerous effects built in. Heavy processor use is common to most (software) digital effects so do not expect to have much joy with lower spec PC's. A separate effects unit is not particularly dear and may be well worth the investment.
One important distinction that should be made is between destructive and nondestructive effects. Destructive effects change the original audio data and are eventually irreversible (undo functions provide some relief from this). Nondestructive effects are more often provided by hardware and are applied over the top of the original sample so that they can be varied without actually changing the saved data. With the increasing speed of PC's nondestructive effects are becoming a more frequent in software packages. It is always far better to use nondestructive methods to produce several mixes of the original so that you do not risk changing the original material into something that you later decide you are not happy with.
Some of the more common effects are detailed below:
Delays can be extremely effective in filling-out and layering sample sections of a track. A drum loop for example can be greatly enhanced by providing delayed variations that add depth to the mix. To work effectively on loops delays should be timed exactly to correspond to the tempo (BPM) of the music. A variety of different ratios can be used to achieve different styles (dub and trance styles often make use of delayed drumbeats). A chart of BPM delay times can be found in the audio section's index and there are also a number of different bits of software that will calculate these for you.
Vocoding is an effect achieved by using one channel of a sample to manipulate the other. It was particularly popular in the early days of electronic music but is still heard frequently today. More often than not the effect is applied to a vocal or spoken sample to create a metallic or robotic sound. A number of standalone and plugin programs can be obtained to create vocoded samples and best results are achieved using a long pad sample to manipulate the spoken one.
Reverb, Chorus and Flange
All three of these effects use delays in one way or the other to change the sound of a sample. Reverb simulates the echoes associated with different acoustic environments such as rooms, stadiums, tunnels etc. Good quality reverb effects are very processor intensive if run as software and so some form hardware unit is recommended. Chorus is similar to reverb but also changes the pitch of the delay to create rich, choral effects. Do not be tempted to overdo the reverb or chorus as your track will end up sounding very muddy. Be selective in which parts you choose to affect - clean sounding vocals and instruments often benefit the most from a little extra depth. Flanger effects modulate the delay over time to give a sweeping effect - the rate at which it does this is perhaps the most significant factor in determining the quality of the sound and common presets will usually include 'fast' and 'slow' flanges. Matching the flange time to the rhythm of your track will often work well - try variations of the same times you use for BPM delays.
Compression works by narrowing the volume range of a sample. Frequencies with lower volumes are boosted or frequencies at higher volumes are reduced to create a tighter, punchier sound. Compression can be very effective when used on sampled drumloops and can be the difference between pathetic sounding toy town drums and tight, professional-sounding loops. Apply 2:1 compression and listen to your drum parts acquire new life. Add a little reverb before compression for deep tom sounds. Chop off the tail of the sound (or use a noise gate) to create gated drum sounds or cut the sound off abruptly to get a dance kick feel. Many great drum loops are created in the effects stage rather than with outstanding kits and compression is generally the starting point.
Wah-wah is an effect achieved by changing the volume of sound dramatically as it plays. Commonly associated with the guitar it was originally achieved manually by means of a volume pedal but can be effectively simulated digitally. The effect is often described in terms of crying.
Filters and EQ
Filters and EQ effects work by selecting bands of frequncies to manipulate. A hiss reduction filter, for example, works by analysing a sample for high frequencies that are associated with background hiss and removing or damping them. Combinations of frequency band filtering can produce some interesting results.
Graphical EQ slider panels provide the most intuitive interface and particular styles of music will often tend to favour certain EQ settings. Classical music, real instruments and unaffected vocals benefit from a fairly balanced EQ setting with bass, mid-range and treble frequencies being equally emphasised. Complicated natural sounds carry numerous harmonics and overtones and so require all parts of the frequency spectrum to be well represented. This is sometimes referred to as a 'warmer' sound. Dance and pop music, however, often relies on a more primal approach to music with heavy bass lines and thumping kick drums. In this case the mid-range frequencies are often cut, with the emphasis being placed on the treble and bass ranges. This results in less presence to the music but a generally cleaner and more penetrating sound.
Using EQ on individual instruments is another way to add clarity and separation to the final mix. Often an instrument part will have bass frequencies that can be removed as they add little to the sound and simply muddy the final mix.
Aural Exciters/Harmonic Enhancers
Aural Exciters or harmonic enhancers are used to give a cleaner, crisper sound to an instrument or final mix. Our hearing is particularly sensitive to the attack portion of a sound and so exciting this (first 20 or so samples) can make the instrument stand out. Too much excitement is invariably a bad thing resulting in thin, overbright mixes.
Mono vs. Stereo
Stereo samples are great but unless you ensure your stereo pans are hard/Left and Right, you may be inviting problems at other mix settings. The most obvious is the phase-cancellation problem where the sound entirely disappears from the mix when both sides are summed to mono (or panned center). Most of the time, your sound won't entirely disappear, but its volume will suddenly drop out and become lifeless or stressed. This never happens with mono samples, which can be panned anywhere with no signal loss. Mono samples are always safe. However, there are a number of tricks one can do with very small offsets in a stereo sample.
You can, for example, make the sound seem to come from the ceiling, from behind you, from to the extreme far left or right, farther back than a mono sample panned 100%. You can also make the sound 'jump out' of the speakers and slap you in the face with stereo. All of the above stereo effects are achieved by hard panning a "similar" sound left and right, with small deviations in pitch, slow LFOs corded to volume and filters, and tiny delays added. Go ahead and take the same sample, hard pan L/R, then play with the attack offset, and detune a tiny bit. You will hear the stereo field jump out. Now do some micro-edits and listen to the field change. What was a simple static pad suddenly becomes an aural landscape with richness and texture. When you lock into something great, save the patch. Just remember to hard pan in the mix. When you are in a mix, the patch will either work or it won't. if it doesn't, there is little chance to make it work. With mono samples, making it fit in the mix is infinitely easier. For backing or supporting instruments, this may be the way to go. For the "signature" instrument in the mix, however, you may need stereo to give it more fullness and depth.