Don't delete those unsatisfactory samples until you've tried resampling them. Tom McLaughlin explains how you can turn a bad sample into a good one.
Some of your samples may be poor - made when you were inexperienced, misinformed or simply made in a hurry. What do you do with a sample that isn't what a sample ought to be?
WHEN THE SAMPLING bug bites, it bites hard. If you're anything like me, not long after bringing your sampler home from the shop and pulling it out of its box you will have sampled just about everything under the sun... probably before you've completely absorbed the instruction manual.
Chances are though, that you've some samples in your collection that aren't quite what you think they should be. Maybe you sampled at too low a sampling rate, and there's an ugly aliasing "undertone" present, or on second thought, you might like to equalise your favourite percussion sounds differently. Fortunately there's an answer... you can re-sample your sounds.
Re-sampling can be executed in three ways: by feeding the output of your sampler back into its input, by feeding the output of one sampler into the input of another (preferably of the same type), by recording a sample onto tape and then back into your sampler. It can also be executed in software, but we'll leave that for another day.
To find out if your sampler will allow you to re-sample directly, put it into Record mode and hit a key - if you hear the sample currently in memory you're away. Some samplers won't allow this (Korg, Mirage) and others will only playback in one area of the keyboard and record on another.
If your sampler won't re-sample directly you'll either have to borrow (or hire) an identical sampler, or "bounce" samples using a tape deck.
MAKE SURE THAT the VCF and VCA on your sampler are wide open. This will ensure that all the top end and level that are present get re-sampled. Check that there's enough level going to the input of your sampler to open its gate... otherwise valuable attack transients may get lost. Make sure effects are properly earthed to prevent mains hum from getting sampled. Pump as much level into effects as possible to achieve the best signal-to-noise ratio. Effects, especially semi-pro and effects pedals, are notorious for adding unwanted noise to a signal. Keep lead lengths to a minimum as more top end is likely to be lost with longer leads. Switch loops off - there's no reason to sample loops and eat up valuable memory space. Monitoring the new sound as you experiment will save repeating samples to get things sounding right.
Granted, you'll be adding a small amount of digital distortion and will probably lose a bit of fidelity and top end as you re-sample sounds. In many cases this won't be discernible and in others resampling may be the only option open to you. The bottom line is to try re-sampling a sound, make an A/B comparison between the original and the re-sampled sample and let your ears be the ultimate judge. Executed with proper care, a resampled sound should lose no more fidelity than when mastering a multitrack tape - without tape hiss to worry about.
Here's several applications for resampling that I've found useful:
1. Corrective equalisation
2. Creative equalisation
3. Adding ambience or effects
4. Pitch correction
5. Saving memory space
6. Adjusting sample length
7. Sampling chords/rhythms
8. Simulated stereo samples
SEVERAL OF MY favourite samples were executed before I had a firm grasp of sampling procedures and suffered badly from aliasing. Many of these could never be recreated and no matter how I tweaked my sampler's filter and resonance controls, I couldn't get rid of the unwanted low frequencies. I'd have to either use the samples as they stood or toss them on the scrap heap. Or so I thought.
I found that, if I took the output of my sampler and fed it through the highpass filter from a frequency-conscious noise gate (a gate with a filter in its side-chain), I could get rid of most of the offending frequencies that were lower than the given note's fundamental. You'll know you've started cutting out the fundamental when the tone starts sounding thin.
Once the aliasing frequencies were minimised I took the output from the filter and fed it into the sampler and resampled it. After hearing my favourite samples for so long with low frequencies that were totally unrelated to the pitch of the sample, it was quite a buzz to be able to play the samples without them - even if a little bit of fidelity was lost in the process. In some cases, getting rid of alias frequencies may be impossible, but corrective equalisation can help minimise its effect.
I prefer using simple filters for corrective equalisation as the signal passes through fewer circuits. In the case of removing unwanted top or bottom end this is all you really need. My favourite "tool" for this is a Frequency-Conscious Noise Gate with its single highpass and lowpass filters - you can't get much simpler than that. I almost always have one of these "in line" when sampling anyway, making use of the lowpass filter to ensure that no frequencies higher than twice my sampling rate get into the sampler I'm using. Although there are supposedly internal lowpass filters on the way into many samplers, I've had occasions where aliasing occurred despite these, and feel safer setting my own top cutoff frequency.
Those of you with modular analogue synths can make use of their highpass filters for removing bottom end, or you could hunt through old electronics magazines for highpass filter circuit designs if you prefer to make your own. It shouldn't cost you more than a tenner to put one together.
"Executed with care, a re-sampled sound should lose no more fidelity than when mastering a multitrack tape - without tape hiss."
WHILE I WAS on that track, I experimented with re-sampling "ooh's" and "aah's" through the highpass filter of my frequency-conscious noise gate to bring out their breathiness, trying to emulate the superb airy vocal sounds that come with the Fairlight.
By reducing the fundamental and those frequencies close to it (rather than boosting the top end with a graphic equaliser) a most interesting and useful phenomenon came to light. Not only were the results breathier and less thick than the original sample, the new samples also had far less of that "munchkin" quality to them when moved around the keyboard.
According to the meagre amount of information available on vocal formants in acoustics text books, the lowest (and apparently most prominent) vocal formants lie in the 375-925Hz region of the audio spectrum. Shifting formants around as you change the playback rate of a sample has a lot to do with one of the main drawbacks of sampling.
FORMANTS ARE ACCENTUATED frequency bands that give the brain information about:
1. The size and shape of the "acoustic resonator" being listened to.
2. The register in which a specific note is produced.
3. With vocals, what vowel is being produced.
These formant bands stay fixed (very much like the bands on a graphic equaliser) for a given vowel or instrument, no matter what pitch is produced. For example, a male "ooh" will always have its first two formant bands around the 400Hz and 800Hz marks, an "aah" around 825-925Hz and 1200Hz.
While the ratios between formants remain roughly the same, female vocal formants are generally 17% higher than males, and children 25% higher. This helps explain the problems of transposing samples too far from their original pitch.
When you play a male vocal sample as little as a third higher than its original pitch you're raising its formant bands up into the range associated with children's voices; any higher and you're in "munchkin" territory. By reducing (but not totally eradicating) the lowest formant, the brain seems to have less information to go on and more "mileage" can be gleaned from fewer samples.
With a graphic or parametric equaliser, you can re-equalise samples in ways not possible with the low-pass filters provided in most samplers.
You can cut out the middle frequencies of a sample, making it sound more hollow, an effect impossible with lowpass filters alone. Samples treated in this way might not sound like much on their own, but in a mixing situation they leave more room for other sounds.
With re-sampling, the bottom end of bass guitars and bass drums can be boosted while leaving their mid and top end intact, snare drums and guitars can be EQ'd so that they cut through a mix, and with either EQ, or an Aural Exciter, more sizzle can be added to murky cymbals. The list goes on, but you've got the idea.
"Adding ambience to dry samples is a piece of cake - route your sampler's output through a reverberation device, then re-sample it."
Creative equalisation is a whole new kettledrum of sampled fish, anything goes... Graphic and parametric equalisers would seem to be the first things to try, but consider re-sampling sounds fed through combo amps (or even transistor radios) and DI'ing or miking them up for the unique tonal qualities they impart to sounds.
Miking up amplified sounds in a room not only adds ambience to samples, it also adds the tonal character of the room and can be thought of as natural equalisation.
Static flanging and phasing alter the harmonic structure of sound by cancelling certain frequencies and creating notches in the audio spectrum of a sound - something you might not use a lot, but another tool in your bag of aural tricks.
Aural exciters add top end to a sound in a manner most unlike any form of equalisation, in that they generate upper harmonics, even where none exist, that are actually related to those present in the sound itself. Used in moderation, aural exciters can help lifeless samples stand out a bit more. Go over the top though and you risk having a sound sizzle six feet in front of everything else in a mix. Be careful with this effect, it's easy to overuse it.
REVERBERATION, NATURAL OR synthetic, adds life to samples. Even a miniscule amount can add sparkle and perspective to an otherwise dull sound. Drum and percussion sounds come to life and can be re-sampled so that they all sound as though they're being played in the same room, even if the samples themselves come from a variety of sources. Orchestral samples cry out for a concert hall to play in, they're just not believable without it - but remember, as a sample is played higher up the keyboard, the room will get smaller, and vice versa.
Adding ambience to dry samples is a piece of cake. Simply route your sampler's output through a reverberation device or mic up an amplified version of your sample in the room of your choice, then re-sample it.
You'll probably need to allocate more sample memory space for ambient samples. If the room is not too large you could try using higher playback rates for a given sampling rate to fit things back into the same amount of space you started with.
Looping sounds with reverb sometimes poses a problem. The initial reflections of a room will create a repeating echo if a loop is started too early on in a sample. In some instances, this produces an interesting effect, but usually not. Moving the loop further back will help. If you have the facility, loop crossfading will minimise signs of any loop join.
Cyclic effects such as phasing, flanging, chorusing/ADT and so on, all add magic movement to sounds but can wreak havoc when attempting to loop sustained sounds. You have not only to find a looping point in the sound itself, but also a loop point within the effect's modulation cycle. As with room sizes, an effect's cycle will get shorter/longer as you play an effected sample higher/lower on your keyboard.
Using a harmoniser for static harmonisation shouldn't cause too many problems with looping, since it just adds another interval above or below the original pitch. Modulated harmoniser effects, however, pose the same problems as cyclic effects mentioned above.
"Pitch correction is probably the most important application of resampling - mix samples and you'll soon discover its importance. "
Compression can be added to samples to get as much level as possible into your machine. With a slow attack, more snap can be added to the attack portion of plucked and struck string notes.
Fuzz, overdrive and distortion effects can be added to otherwise timid sounds by employing effects pedals made for this purpose, or you can crank up a combo amp (many have overdrive provisions) until you achieve the sound you're looking for.
"PITCH CORRECTION? I have coarse and fine tuning on my sampler", you might say - and right you are. Unfortunately, these tuning controls only affect the playback rate of samples.
Pitch correction is probably the most important application of re-sampling. When you start mixing samples you'll discover the importance of samples being in tune with one another from the start. Those of you with samplers at the upper end of the market will have provision for some form of digital sample rate conversion, as on the Fairlight III, and shouldn't have any problems, but for us mere mortals the pitch that gets sampled is what gets stored in memory. Hopefully software houses will realise the need in the mid-market for advanced editing facilities such as this and include it in more software packages - it makes mixing samples in software so much less of an ordeal.
With a reliable tuning source, you can tune samples to a common pitch with your sampler's tuning controls, re-sample them at this pitch and then mix them with the greatest of ease. The slight loss of fidelity through re-sampling is a small price to pay for the added convenience.
AS MENTIONED EARLIER, you can fit more information into a given sample memory space by re-sampling at a higher playback rate. This technique really comes into play when you're doing multi-samples and need to squeeze one more sample into your set. Go for re-sampling the lower samples first.
The trade-off is losing top end but for many sounds this isn't too important. The only way to find out if you can get away with the loss is to try it. As a rule of thumb, you should always experiment on a floppy set aside for this sole purpose to avoid losing your original sample. If your experiment is successful, you can then copy it to your working disk.
IF YOU HAVE a sampler with a fixed loop length and have a sample that refuses to loop without glitching, no matter where you place the loop, before you blast the floppy it's on to smithereens, try sampling it at another pitch. Loop positions, like digitally recorded sound material, only have so many notches available, and stretching or squashing the sound just might help.
RATHER THAN USING up six of your sampler's voices for a simulated guitar part, you could re-sample the chords you plan to use as separate samples. Even though this may use up more sample memory space, in a sequence you'll only be using one VCA/VCF at a given time as opposed to six.
Chords can be rolled up or down, either manually or with the aid of sequencing, mimicking up and down strums of a guitar. Harp and piano arpeggios and rolls can also be simulated in this manner.
Likewise, rhythms can be sequenced and re-sampled. An example that comes to mind is a castanet part I was working on. I'd only one "voice" left in my sampler and with the rapid rhythm required of the castanet, the ambience at the end of the sample kept getting cut off by the following hit, sounding quite unnatural. (If I'd another "voice" left I would have alternated the rhythm between the two thus eliminating a problem common to many drum machines.) What I ended up doing was to take my castanet rhythm and divide it up so that each hit was coming out of a separate VCA and then resampled it, each hit having a natural decay. Fortunately the part required little more than the same rhythm repeated throughout the piece.
STEREO SAMPLES CAN be simulated by re-sampling mono samples through a stereo effect. If your sampler has no provision for stereo sampling but can layer one sound upon another and send each sample to its own output, you could sample each side of the stereo effect separately. You'll most likely have to align the starting points to avoid phasing problems with this method.
Effects that come to mind for simulating stereo from a mono soured are: reverb, delay, auto-panning, phase-shifting/flanging, ADT and chorusing. Try experimenting with different effects, settings, modulation rates, or EQ on each side.
So you see, you don't have to possess the most brilliant editing software to carry out corrective or creative re-working of samples and, although it's preferable to get things right in the first place, all is not lost if you happen to change your mind about how a sample sounds once it's been stored onto disk. Re-sampling is as valid a technique with sampling technology as any in the world of audio and sound synthesis.