Samplers

A Fairlight CMI sampler, consisting of two six-octave keyboard consoles sitting atop one another, and a boxy black computer keyboard console and CRT monitor sitting on the top keyboard console. Built into the CRT monitor housing is a vertical slot for a five and a quarter inch floppy disk. The lower keyboard console has dials and buttons on panels to the left and right of the keys, and the centermost C key is in its depressed state, possibly broken. The entire assembly is sitting on a wooden desk next to a large cabinet speaker. Underneath the desk is large processing unit, roughly the size of a safe, labeled with the Fairlight logo and encompassing another vertical floppy disk drive. A 1979 Fairlight CMI sampler. Designed by Australian sound engineer Tony Furse, the Fairlight CMI was one of the first popular samplers to be produced, and was used by popular musicians including English singer/songwriter Peter Gabriel.
(Joho345 | Public Domain)

A sampler is a musical instrument that, like a synthesizer, generates sounds electronically. Instead of building sounds by combining basic waveforms, however, a sampler plays individual audio recordings for each note played.

Sampling Fundamentals

Because samplers create sound using audio recordings, they require some form of data storage.

File Storage

While a few early samplers made use of magnetic tape, beginning in the 1970s units made use of disk-based storage — either floppy disks which allowed the user to switch sounds by exchanging disks, or hard disks built into the unit or connected to it for storing larger files.

A diagram showing different disk sizes throughout recent history. Listed are: an 8-inch floppy disk, which held 1.2 megabytes or about 8 seconds of audio, popular in the mid 1970s; a 5.25-inch floppy disk, which also held 1.2 megabytes or about 8 seconds of audio, popular in the early 1980s; a 3.5-inch floppy disk, which held 1.44 megabytes or about 10 seconds of audio, popular in the late 1980s; a CD-ROM, which holds 650 megabytes or about 74 minutes of audio, popular in the late 1990s, and a flash drive, which can hold 1 terabyte, or 83 days of audio. Hard disk sizes are also shown: 40 megabytes, or about 4.5 minutes of audio, popular in the mid 1980s; 500 megabytes, or about 56 minutes of audio, popular in the early 1990s; 4 gigabytes, or about 8 hours of audio, popular in the late 1990s; 500 gigabytes, or about 40 days of audio, popular in the late 2000s; and 2 terabytes, or about 169 days of audio, popular in the late 2010s.
Figure 1: Various common storage media, shown with their common sizes and how much uncompressed stereo audio can be stored on them. Dates show approximate years of popularity.

Digital audio requires large file sizes, and samplers used special techniques to work within the limited storage capacities of early computer disks.

Compression

Because audio files contain a large amount of data — one minute of high-fidelity stereo sound requires about 10 MB of storage space — some samplers use audio compression techniques to reduce file sizes. Some of these techniques include removing the highest and lowest frequencies of sound, or "rounding off" finer details, a process which sacrifices sound quality for smaller files.

Key Mapping

A sample could theoretically use a single recording — for example, the recording of a cello playing the note F3 — and pitch-shift it to match other notes across the frequency spectrum. However, because changing an audio file's pitch also changes its speed, this causes speed distortions for notes at large intervals from the original pitch.

Figure 3: Pitch-shifting a single sample. The original sample is a recording of a cello playing the note F3. When this sample is pitch shifted across the entire range, high notes and low notes have significant speed distortion.
(Cello audio: MTG | CC BY-3.0)

High-quality samples will sometimes include a separate audio recording for every pitch, but these samples have much larger file sizes as a result.

As a compromise, most samples use multiple recordings, but pitch-shift them to cover small groups of nearby pitches. The process of specifying which samples are used for particular pitches is called key mapping.

A keyboard showing notes from C3 to B4. Groups of notes are colored differently: C3 to E3, F3 to A3, B flat 3 to D4, E flat 4 to G4, and A flat 4 to B4. The center note in each group, D3, G3, C4, F4 and B flat 4, are labeled and marked with a dot.
Figure 4: A subset of samples can be pitch-shifted to cover small groups of notes without much distortion. In this case, five recorded samples can cover two octaves of individual notes.

Sample Design

Creating a sample from a recording — or set of recordings — requires a unique process.

Recording

Recording processes for a samples require a careful approach. For a piano sample, for example, a recording engineer will likely select a high-quality instrument like a concert grand piano and ensure that the instrument is tuned and ensure that all its mechanisms are in good repair.

The recording itself is generally done in a controlled acoustic environment like a recording studio. Any background sounds or even ambient sound will be captured in the sample and sound unnaturally repetitive when the sample is used to create a piece of music.

Figure 5: Using poorly-recorded audio — in this case, an out-of-tune piano recorded in a noisy environment with a smartphone — creates an ineffective sample.

In some cases, engineers create samples to capture a particular individual instrument, perhaps one with historical significance. Some instruments, like pipe organs, must be recorded on location, which requires more significant challenges for the engineer.

Stitching

Professional samples will sometimes include multiple variations of a particular sound, such as different articulations of a single pitch. However, it would be impractical — if not impossible — to record every possible length of note for a sample. Instead, sample recordings are designed to have three parts: an attack, a body, and a release.

Figure 6: Audio for a sample is separated into three sections: attack, body, and release.
(Cello audio: MTG | CC BY-3.0)

When a performer plays a note, the attack portion is first played, and then the body of the sound is repeated until the note is released, at which point the release portion is played. The exact dividing points between the sections must be located so there is a smooth transition from attack to body and body to release. Further, the end of the body must segue to the beginning of the body so the body section can be repeated as many times as necessary. This process is called stitching.

The most effective way to stitch sounds together is to ensure that the dividing lines are at zero points on the waveform. This prevents "jumps" in the waveform, which translate into audio as clicks or pops.

Figure 7: When combining components of a sample, cuts must be made at zero points on the waveform to prevent pops or clicks.

It can be challenging to find the right length for the body of a recorded sound. If the body is very short, it can introduce an unnatural pulse to the sound. A long body can provide a more expressive, full-bodied sound, but releases will not be as responsive to the user's control.

Special Features

Some samplers support more advanced features. For example, a sample might be designed to have a steady body that switches to a vibrato when a key's aftertouch is engaged. A complex string sample might include multiple articulations, like regular bowings, pizzicato, and col legno.

Samplers: Summary

  • A sampler is an electronic musical instrument that plays audio recordings for individual notes.
  • Because samplers rely on audio recordings, they must have a method of storing digital audio.
  • Older samplers were limited in sample size and quantity due to the relatively high storage requirements of digital audio.
    • One method for maximizing storage space for audio is the use of audio compression.
    • Another method is key mapping, in which a single audio sample is pitch-shifted to use for a limited range of notes.
  • To use a digital audio recording as a sample, certain procedures must be followed.
    • Recordings must be made in carefully controlled environments to eliminate extraneous noise.
    • So that they may be adapted to notes of varying lengths, audio recordings are split into at least three segments: an attack, a body, and a release.
    • The various segments of a sound must be spliced so that waveforms are not interrupted when they are combined with one another; cutting waveforms at zero points simplifies this.
  • Some samplers support samples with advanced features, such as changing the tone in response to a key's aftertouch.

Exercises

Exercise 1: Constructing a Sample