"Sampling and Samplers – Theory for Beginners"

Sampling and Samplers: Theory for Beginners

This document on sampling is intended as a primer for the NEWBIE, who is looking to form a deeper understanding of what samplers, samples and sampling is (and are), and what they can be used for. This document is intended to give you the basic knowledge needed to begin to explore the world of sampling, multi-samples, and sample manipulation and composition.

Part I - Introduction

Well, this is indeed a very big topic and we are faced with a very general question, but lets take a preliminary stab at an answer, it should help to get anyone started. We will look at what exactly sampling is and means, through to defining common sampling and sampler terminology, to how we can use the different tools in our samplers to accomplish a wide variety of tasks.

Click Here To Order VST Platinum

This document has been written in a very general format, and is not intended to be used as a guide to any specific sampler technology. For this purpose I have avoided using direct examples of a specific vendors technology I want this document to be useful as a general introduction to people starting to investigate sampling. The choice of a specific technology or approach will depend on one's own unique musical needs and goals, and shouldnt be influenced by this document.

First of all, using samples is just a way to make a sound. Indeed, all you want to do when creating 'digital music' is ultimately make some sounds of some sort.

There are a few different ways to make a sound from a digital device like a computer.

One of the ways is to use 'synthesis', where physical mechanisms (or digitally modeled physical elements) interact to generate a sound where there wasn't any before. These processes usually generate a sound using an oscillator to spit out a big pile of audible frequencies, then a filter to shape and control the actual frequencies that get 'played' and thus heard, and often an 'effect' or two that will add character to the sound. Thus the act of synthesis creates a sound all by itself.

(Please note that the above paragraph was an extremely simplified overview of basic subtractive synthesis, simply for sake of brevity within this document. There are other methods of synthesis that differ in theory from this, such as granular, FM, wavetable, additive, etc. etc., however basic subtractive theory is really the grandfather of all synthesis, and as good a place as any to begin in an understanding of it all.)

In any case, we have not even begun to answer the original question yet. Lets keep going...

So synthesis is purely a way to make a controllable sound.

'Using Samples' is simply a different way of making a sound. Instead of using mechanisms, or digital models of mechanisms, to create a sound 'on their own', we instead forget that approach entirely. We simply record a sound beforehand, and then play it back when we want to hear it. That is the simplest form of using samples.

We still have a way to go before we can say we understand our sampler though!

Part II - The Wonderful World of Samples

Now we've just seen two different concepts and methods of making a sound - synthesis and sample playback. Since we are investigating samples and samplers, we need to look past the gross oversimplifications above, and further into the details of this art.

There are a lot of different ways to use samples. I tend to divide the different ways to use samples into three mental categories:

Let's look at each of these over the next couple of paragraphs, and by then you should have a pretty good idea of how sampling can work for you, and hopefully get a good start on learning more (and hopefully my time spent typing this will have been worthwhile!)

Single use samples (one-shots)

Example lets pretend that you have a .wav file with a drum loop, its a nice break beat, really funky. So you loop it and it plays - maybe in the background of your song, maybe just in part of the song. That part doesn't matter. Perhaps you trim the sample to be a bit shorter, and then loop it good for you! You may also have some synths playing too, nice work.

Now hey! It turns out that you also have this other .wav file; it's a funky bass guitar riff. You can stretch it (so it fits the same tempo) and loop it alongside the break beat! Sweet, you are using some single samples.

Applications like Sony's ACID are all about the 'single sample'. You can get piles of CDs from them full of these .wav file loops. Programs like ACID let you have them all automatically time-stretched to your chosen tempo, and you can layer all the ones you like into 'tracks'. Maybe you have a sound effect you really like, and you have a Boss Dr. Sample machine. You can load that sound effect, and 'fire it off' by hitting a rubber pad. Yay! Youre using single samples. This is one of the many ways to use samples.

Usually programs that work with single samples have tools within them to do things to 'effect' the sample before playback, like filter it, adjust the volume, pan, eq, adjust the loop points, etc. (But not always! Check with your friendly manufacturer or developer.)

Multisamples

Now Im going to talk about the whole concept of multisampling. Please note that this is a generic term that can be applied to a very wide range of different applications, so I'll do what I can to at least help it make sense.

To look into multisampling, let's start with the idea of a 'single use sample' as described above, and then extend that a bit.

Imagine that you have a handful of sound FX samples in .wav files, and you'd like to be able to control firing them all off at different times. Heres where a 'sampler' program will be useful. You could be using any one of many different methods to trigger the playing of one of your samples, it could be a rubber pad on an MPC type device, a key on your computer's QWERTY keyboard, a 'note' or 'step' in the piano-roll/timeline of a MIDI sequencer, or a key from a MIDI-capable 'piano-like' controller keyboard.

So that the discussion can be as easy as possible moving forward, let's focus on the idea that we are going to trigger our sample to play by pressing a key on our MIDI-capable piano keyboard (which we will refer to hereafter as a 'controller keyboard') - but bear in mind this is just sending a MIDI message, and that message could just as easily have been created by a rubber pad, a drum machine, a sequencer etc., however as stated we are going to focus on the piano concept for the time being....

Now what we will do is assign each one of our handful of FX samples to a different key on our controller keyboard. We can now fire these off at will by pressing the assigned key, and even play more than one of them at the same time by pressing keys simultaneously. We can record the MIDI data generated by our key-playing, and edit it if we want, and use that recording to trigger the same 'performance' from the samples at a later time.

This is the most basic way of applying the concept of multisampling. Perhaps the samples we are using are of a drum set - one bass drum sample, one snare sample, one crash cymbal sample, etc. We get one mapped to each key, and we can 'play the drums' by pressing the keys on our controller keyboard (or even by playing a MIDI-capable 'eDrum kit').

Anyway, the sampler will have a way to save this 'mapping information', relating a bunch of samples on your hard-drive to a bunch of keys on a MIDI keyboard, so that you can recall it at a later time - this is the 'programming information' for your sample set, and will be in a different format depending on the manufacturer of your sampler.

If you have a 'program' for a sampler device, it will consist of two things: the actual sample data (basically a collection of .wav files), and the programming information that associates the mappings of these sound files to MIDI input notes, or keys on a controller keyboard. Some of the samplers wrap all these files up in a single 'monolithic' file, somewhat like a .zip file contains multiple files inside it. TASCAM's GigaSampler? works like this - you could have a single 2.0gb .gig file, that inside contains thousands of samples of audio data, and lots and lots of programming information.

Other sampler systems keep all the audio files in a directory, and keep a data file to hold all the mapping information. There are many different formats of this type of sample data, and they are not all compatible with each other (although there are good converter programs available for most types). Most sampler programs can open up a number of different formats however. Additionally, almost all sampler programs allow you to drag n' drop in your own audio data files, and create all the mappings and programming information yourself to create your own 'programs'.

Now that we have a basic understanding of multisampling, we can now go to the next step here is where multisampling starts to get pretty serious and impressive.

You now know enough to see how you can start to use a sampler to create something like a 'playable copy' of a different 'real instrument', like the playable drum set we discussed above. This 'playable copy' is indeed the holy grail of sample/sampler development and technology.

The quest for a better and better 'copy' of the original instrument has fueled an amazing growth over the years in the abilities and technologies that are used by samplers. As computers began to take over the real hard-core sampling jobs, and as these same computers became faster and faster, the 'instruments' that we can play with a sampler get better and better, and more and more realistic (not to mention 'larger and larger' as more and more samples are used to create the 'program'). Lets examine how all of this works and we will again go to the old standby example item, the piano.

Imagine that we want to make a playable copy of a piano using our sampler program. The first thing we will do is get eighty-eight .wav files, one of each key of a piano being hit. Then we will load each one of these in our sampler, and 'map' it to the corresponding MIDI note, or key on our MIDI controller keyboard.

There you have it - your very own virtual, sampled piano! Indistinguishable from the real thing! Pat yourself on the back. Well, ok, maybe not. It seems that our wonderful piano has a lot of problems compared to the real thing. It actually kind of sucks to be honest! We need to work on this idea some more - let's look at some of the things that are wrong.

First of all, a real piano responds differently depending on how hard you press the keys. REALLY, VERY differently from the softest press to the hardest press! That creates a challenge for our static set of samples. Additionally, resonances build up inside the piano body that affect the sound, and what these resonances sound and act like depend on what has been played, and how hard, and how recently it's been played. That creates a BIG challenge! We can also easily see three pedals on the piano, each of which affects the sound. Weve got our work cut out for us if we want to create a perfect sampled piano! This is starting to be a very complex interaction of sounds and responses when we sit down at that Steinway Grand and start banging on the old ivory keys.

Fortunately, the modern sampler has given us tools to tackle all of these problems. Its a constant argument about how realistic the 'best' sampled emulations of real instruments are, but it's fair to say that some instruments can be done very, very well, and some still cannot tackle all of the nuances of the original yet. Brass instruments are notoriously hard to do, with the same control of expression that a real player has. In any case, those arguments are a different topic for a longer article, and right now I'd like to at least let you know how samplers address some of these problems before you need to worry about the intricacies of these debates.

Click Here To Order VST Platinum

We are going to first take a look at how samplers address the problem of 'velocity', which is the problem of 'how hard the piano is played'.

Remember that when we 'tell' our sampler to play back a sample, we do so with a MIDI message. We are assuming in our discussion that it is a MIDI controller keyboard that is generating this message (but we also recall that it could be many other things too!). Within the MIDI specification, which is basically the language that these 'MIDI messages' are written in, there are a few different bits of information.

Of course one of the pieces of information is 'which MIDI note is being sent' - which corresponds to the key you press on your controller keyboard. However, whenever a message like that is sent, it's always accompanied by a 'velocity' value, which corresponds to how HARD you pressed the key on your controller keyboard (sequencer programs will let you adjust the velocity of any recorded MIDI notes as well).

The MIDI spec allows for 127 levels of velocity, so when you press a key, the MIDI message may say something like "C1, 122" meaning you pressed the C1 key, at almost full velocity. A gentle press of the same key may generate the message "C1, 46".

So how does this help us make our sampled piano more realistic? You will find out very soon, but first we are going to have to get a lot more than 88 samples of our original piano!

We originally obtained one sample for each key on the real piano. Now we want to collect a whole bunch of samples from each key, some hard, some soft. It might be good to build a machine to make sure that the 'hard' hit is the exact same strength on each of the 88 keys when we hit them! This is going to be a big challenge.

Technically speaking, we could record up to 127 samples, of 127 different volume levels of each key on the piano. This would yield a HUGE amount of samples - which may or may not be what we want. Some commercial sample sets provide an 'XXL' and a 'Lite' version of the same program - the XXL will use a lot more samples for deeper detail, but it has a much larger performance cost, memory cost, storage cost, etc.

The number of samples that we will record of our piano will depend on the final size we have targeted for the sample set that we are building. And since this is a completely theoretical sample set, let's not even bother figuring out the 'target size' and how many velocity levels we will take. That would be a pain in the ass! Let's just say we take a 'bunch' of samples per key, maybe 8 or 12 or 24. Cool. Lets go to our sampler and do some programming!

Remember how we had the 88 samples lined up over the 88 keys on our controller keyboard? In a sampler's interface, that is usually a horizontal mapping - you do it from side to side. Keys move from left to right, and the Samples that are assigned to the keys are layered vertically over each key. You usually map velocity 'layers' vertically, over the key you are mapping them to. You map the higher (louder) levels to the higher velocity 'zone' and the quieter sampler to the lower velocity zone - all 'per key'. So a single key from a MIDI keyboard may end up having 8 or 12 or 24 samples stacked on top of it, and the sampler itself has the job of selecting and playing the proper sample depending on how hard (velocity, remember?) the key was pressed.

I hope you are following because we've gotten into the tricky territory now. If it's a little cloudy, maybe have another read through the last two or three paragraphs! We want to be sure we know what we are doing before moving on..

Now the good news is that once we have done the programming, our sampler will automatically take care of playing the louder notes when we hit the keys on our controller keyboard harder. Our virtual sampled piano is starting to sound a LOT more realistic. Cool!

Now the trick we just learned about mapping different samples to different velocity layers is called 'velocity cross fading' - you are using the sampler's programming to control your ability to play a different sample from the same key, depending on how hard you hit it.

It doesn't even have to be different volume samples of the same instrument mapped onto one key. Theres no limit to creativity here. You could map piano samples to the low velocities, and trumpet blasts to the high velocities! Then you actually have the ability to control two instruments at the same time from your controller keyboard - playing softly gets you piano, playing hard gets you trumpet. Cool stuff!

Now Ive just introduced a new sampling concept to you - sample cross fading. If you understand the above paragraph, you'll see that we are controlling the cross fading of different samples, using velocity messages. There are other different ways we can control sample cross fading too! Heres where we can tackle more of the problems of our sampled piano project.

Lets look at the sustain pedal on our real piano - a very important part of a piano player's expression. OH NO! I THINK WE ARE GONNA NEED TO RECORD SOME MORE SAMPLES!

Lets assume that we originally did 12 velocity layers of each key of our keyboard. We are going to have to record 12 more samples per key, this time with the sustain pedal held down on the piano. When that's done, we can take them into our sampler program and do some more programming!

The MIDI spec again helps us - most MIDI controller keyboards have an input for a sustain pedal. When you press the MIDI sustain pedal, a MIDI message is sent. When you release it, another message is sent. If you don't have a MIDI sustain pedal, you may be able to assign the function to something else, like a modwheel.

Anyway, you can probably see where this is going now - we will program these new 'sustain' samples so that they play when the MIDI sustain pedal is held down, and the sampler plays the original 'dry' samples when the pedal is not pressed down. That old sampler is really starting to do a lot of work for us! And our sampled piano is REALLY starting to get expressive too, this is cool stuff.

In the above concept, again we are controlling cross fading of samples, only this time we are using a MIDI sustain pedal message to control them, instead of velocity. Lots of things can control cross fading between different 'layers' or 'sets' of samples - sustain pedal, expression pedal, modwheel, velocity, aftertouch, etc. And you can combine any or all of these to create very expressive, controllable sample sets. And remember - you can cross fade between totally different sounds, not just samples from the same instrument.

Earlier we said that we could switch from piano to trumpet with different velocity layers - you could do it with the modwheel instead, so that you can preserve the ability to have lots of velocities of the original instrument, but still have the ability to crossfade between instruments. Roll the modwheel up, and playing your controller gives you velocity-sensitive piano notes. Roll the modwheel down, and your same playing gives you velocity-sensitive trumpet sounds. Good times!

Another term often given to controlling the switching between 'layers' or 'sets' within a sampler program is the word 'keyswitching' - this originates from the ability (found in many software samplers) to assign a NOTE (or key on the controller keyboard) to switch or toggle between different sampled layers - in lieu of using a modwheel or expression pedal etc. to do it.

You've now been able to learn a lot about multisampling - thanks to these concepts, we have 'playable copies' of instruments we can load up in our samplers - if you look at the free and commercially available sample sets, you'll see every instrument imaginable has been multisampled and is available in different formats - and as the technology keeps getting better, newer more detailed sets are arriving all the time - it's the power of the sampler that makes this all possible. For the most intense, cutting-edge example of what high-end sampling is doing these days, take a look at the Vienna Symphonic Library project, found at http://www.vsl.co.at

FSU Sampling

The final type of sampling application I would like to discuss is what I call FSU Techniques. Indeed, FSU stands for Fu*k Sh*t Up, and, as you probably imagine, is an approach used to create something that sounds very different from the original. Many effect processors and plug-ins specialize in FSU effects, but we can also harness the power of sampling to create our own FSU results! I actually consider this use of sampling to be closer in nature to synthesis as it is to traditional sampling, however it does need to be addressed in this document so that the reader will have a full idea of what can be done with sampling.

The basic idea behind FSU-based approaches to using samples is probably very obvious: Use the sample (or sampler) as a starting point to create something very different from what is in the original samples. A simple FSU strategy could be to assign a random sample to a variety of keys and velocity levels, thus resulting in unpredictable results when you play your mixed up and random multisample set.

A more complex and powerful approach, has to do with raw waveforms. A sample that is loaded into a sampler typically consists of around a half-second (or more) of recorded audio material. A second of recorded audio material at 44.1khz actually consists of 44,100 cycles of a waveform. It is the rapid repetition of this waveform that creates the timbre of a recognizable instrument.

Instead of working in human ear timelines such as seconds or minutes, we can use the sampler (or wave editor) to look at much smaller pieces of time we can go right to the single cycle of the waveform. By combining the waveform shapes from various sounds, or by taking a single cycle of a recognizable sound and then changing its shape, we can create something that sounds quite different when the waveform is sped up and played back to us at audible rates. In fact, this part of the main idea behind wavetable synthesis, however it is also something that can be accomplished through sampling. You can take a sample of a bell being hit, zoom in so that you can view a single cycle of it (or perhaps a few cycles), and the change the shape.

You may change it by drawing a different line using a pen sort of tool in your audio editor. Perhaps you will process it through an effect, or combine it with shapes from other waveforms. Regardless, the end result is that you have started with a recorded sound (the original sample) and then have manipulated it to create a very different sound. This in essence is Fu*k Sh*t Up approaches to sampling. The fun thing here is that the sky is the limit as you bend and twist your samples to create new sonic worlds.

Hopefully this document has been useful in helping you come to terms with all of the different elements conjured up by the word sampling. In specific, the understanding of how a sampler works, and what it can do in particular within the concept of multisampling is essential knowledge for someone about to delve into the world of sampling, sample playback, and sample manipulation. I hope that your hands, eyes, ears and mind are now open and excited about the many possibilities before you.

Now go forth and create!