Reply to comment

The History and the Pioneers

Under the hood of every virtual instrument is a sound engine designed to bring you the variety of tones promised on the front of your software box. Some methods rely mainly on mathematical models, others make use of audio samples, and there are those that make use of a mixture of the two.

Today we tend to speak about sampling as a normal part of the current music scene without being aware of the long uphill grind it took to get here.

It is now common practice to capture an audio snippet, fashion it to taste, and play it back pitched and at our own convenience.

The current state of technology has brought us amazingly far from the fledgling years of sound reproduction. We have come from the days of Edison's cylinder phonographs and gramophone disks, through the rise of the magnetic tape era; to our now tapeless state of mind. Technological advancements in both tape machines and microphones allowed artists to start experimenting with new ways of using these tools. Musique concrète and early electronic music made use of tape looping and splicing and later, instruments like the Chamberlin and Mellotron placed prerecorded audiotape segments under the control of a two-octave keyboard. But these were not quite "samplers" as we refer to them today.

G-Force M-Tron "Virtual Mellotron"

Today we think of sampling as the ability to capture a segment of audio and reproduce it on command. What differentiates a hardware or software sampler from a simple audio recorder or tape-based instrument is its ability to both record and readily retrigger your sonic snapshot. Some early digital delay lines had the ability to do just this, and they were used extensively to fly-in background vocals or trigger drum sounds to fatten up tracks. And although these samples could be tuned and triggered, they could not really be "played", and were lost when the unit was powered down.

Early Instruments

The first commercially available sampling "instrument" was the polyphonic, green-screened Fairlight CMI. The machine was introduced in 1979 and, like its rivals the New England Digital Synclavier and WaveFrame AudioFrame, it went through several improvement stages before being wholly embraced by those who could grasp what it had to offer. At the time, memory and disk storage were prohibitively expensive, and this added to steep price tags that covered both hardware and research for this new field. Always looking to the future, high-end producers and performing artists like Trevor Horn, Peter Gabriel and Stevie Wonder were eager to get their hands on them straight away. Sampling was a welcome alternative to the instrumentation that was available at the time and it posed some exciting possibilities. With improvements in audio quality, whole prerecorded sections of music could be lifted from the master reel and moved without even touching a tape splicing kit. A short recording of almost anything could be pitched down by several octaves and looped to create some very complex and unexpected soundscapes.

By the middle of the 1980's several lower priced alternatives began to appear. In 1982 PPG would produce the Waveterm, which offered both waveshaping and sampling, and Kurzweil would release the well-designed K250 by 1983 that shipped with some very nicely executed acoustic emulations. In 1981, Emu introduced a lower priced alternative with their Emulator, yet things really took off when other manufacturers helped to make their gear even more affordable while steadily increasing audio quality. Ensoniq, Korg, Roland and Sequential Circuits all had a hand in the game, but it was the introduction of the Akai S series that landed accessible quality hardware in thousands of studios and stages worldwide. More affordable memory and trickle-down technology from higher end models also lead to samples being used onboard countless popular synthesizers like the Roland D-50 and Korg M1. Samples and sampling capability are incorporated in all of the top selling keyboard workstations available today and its inclusion, even as an added option remains a top selling point.

Kurzweil K2600XS Workstation

The Process

With the huge amount of moderately priced memory available today, software-sampling solutions on our own home computers can be both less expensive and more powerful than earlier hardware alternatives. Even with these advances the process remains nearly the same. Hardware samplers work like a sound camera- after you ready the machine to capture audio you can either manually trigger the recording or set the machine to start automatically recording when it detects audio at its input. Signals passing through the analog inputs are then translated into digital information with an A/D converter so that they can be saved and manipulated in the digital domain. If you are looking for the highest fidelity possible to capture acoustic instruments and the like, then it is important to remember that the quality of the unit's inputs and A/D converters can strongly affect your result. The same goes for your D/A or output side of your signal chain.

Analog audio signals are continuous in that they directly reflect the periodic wave cycles associated with them. Digital machines, on the other hand, prefer to speak a more jagged language composed of only two states- on and off or 1 and 0. With clever smoothing techniques analog signals can be represented with a string of 1's and 0's as long as there are enough of them to adequately reproduce the original signal faithfully.

TECH TERM: Nyquist Theorem

The Nyquist theorem states that for satisfactory results, we must sample at twice the rate of the highest frequency we are trying to capture. Our current CD standard of 44.1kHz/16-bit is based upon the assumption that the theoretical upper limit of our hearing range is 20kHz.  By the Nyquist theorem we know that we must at least sample at a rate of 40kHz to capture audio frequencies up to 20kHz. Any frequencies occurring above 20kHz will be reflected back into the sample as a form of audible distortion called aliasing. To keep this from happening, "brick-wall" filters are employed to filter out any frequencies above 20kHz, preventing aliasing. The seemingly uneven figure 44.1kHz is a carryover from the days of converting digital audio to analog videotape as a means of storage and was made standard by the Audio Engineering Society standard in 1985.

Capturing and Shaping Your Samples

Much like opening the shutter of a camera to let in light, a hardware sampler captures audio directly through its inputs. With a software sampler, on the other hand, audio must first be captured in a recording program and then imported into your sampling software. This may not be as immediate but it does allow you to manipulate and edit your sample before it goes into the sampler itself. Once your audio has been digitized it is ready for any number of refinements. This is where you can take care of any tasks that will allow you to shape your sample to taste:

Trimming - Trimming allows you to choose your exact start and end points so that the sample triggers and ends precisely.

Looping - The ability to repeat a segment of your sample that has been defined by two locate points. Because of the memory restrictions imposed on early sample-based keyboard instruments, manufacturers made every effort to get the most out of what little they had to work with. A common practice would be to start with a short sampled attack portion of a sound, a piano for example. In order to stretch valuable memory, the sustain portion would consist of a looped segment that would continue until a key was released. Today's "loopless" presampled libraries command higher prices because great care has been taken to sample each note from beginning to end requiring larger amounts of memory.

Filtering - Filtering can be one of the most powerful tools at your disposal when shaping the tonality of your sample. Some samplers, like the Emu hardware series for example, came equipped with filters that were capable of taming the brightest raw material or twisting it into oblivion.

Keymapping - This will allow you to assign a midi note number to your sample so that it can be triggered via keyboard, drum pad, or other MIDI controller. Some detailed libraries dedicate a set of samples to every note for the most realism. Another more economical method would be to assign a note number to your sample, Middle C for example, and spread its range amongst the surrounding notes. This way you are pitching the surrounding notes by speeding up or slowing down the original recording. Playing back a sample at twice its speed will yield a pitch shift of one octave up and at half-speed an octave down. The problem with using this method on acoustic instruments is that as you do this the formants or fixed resonances that give it its character are altered leading to a very unnatural sound. On the other hand this can be used to great advantage when exploring the unconventional.

Three Key Map Examples: Sample 1 maps a single sample across an octave,  Sample 2 maps two samples,  and Sample 3 maps 8 samples across an octave

Velocity/Keyswitching - In order to add expressiveness to your playing, you could apply velocity sensitivity settings to allow your sample to play louder or brighter with harder keystrokes. In fact, many other parameters are usually available for this function such as effect or modulation depth and keyswitching. By assigning different velocity values to different sample sets on the same key it is possible to trigger for example a soft piano sample at low velocity and louder piano samples at higher velocities for more realism.

Envelopes - As in most of our modern day synthesizers, envelope generators give you the ability to shape the amplitude and filtering of your sample over time. In more advanced systems envelopes may be applied to a host of other parameters for more complex sound design.

Tuning/Time Stretching - As mentioned above, altering the pitch of a sample by a few semitones could make a very noticeable difference in the character of the original recording. Dropping the original sample by an octave will produce highly unexpected results every time. This is a great technique to employ if you are active in sound design or want to try something new. Time stretching will allow you to alter the playback time of your sample without altering its pitch. This can be very useful when tailoring an imported drum loop and is a valuable tool for broadcast engineers. If you are allotted 15 seconds for a short radio spot but your source material lasts 20 seconds you can easily squeeze it to fit without the "munchkin effect" or changing the pitch of the announcer's voice.

Processing - Although later hardware sampler's boasted onboard effects the new class of software and sample based instruments have taken things up a notch with a much greater selection of sonic manipulators, filters and enhancers. Adding effects to what might seem like a dull or uninteresting sample can bring it new life. Since the early days performers have relied on samplers to add spice to their performance or to help them reproduce complex studio parts on stage. By adding the right amount of processing to the sample there was no need to bring along an external processor. Today reverb delay, and the like are considered basic tools and software options like bit-crushing and morphing add to your choices for processing.

Outfitting your computer - Check out our Virtual World Part 3 article for some useful information about setting up your computer for better audio performance.

Hopefully these "hot points" will give you some basic guidelines or refresh your memory on the roots of sampling and some of its more important principles. Hopefully some of this information will help you as you browse through the ever -growing selection of soft-samplers available today. Next month we will take a more detailed look at the specific models but here is a short list of some of the top virtual samplers available today.

Native Instruments Kontakt 3 MOTU Mach 5

Digidesign Structure IK Multimedia Sampletank

E-MU Emulator X2 Apple EXS24 Sampler

Steinberg Halion Tascam Gigastudio 4