Recording samples (sources, cabling, sound cards, rates)

What it is#

Getting a sound from a synth, CD, tape or microphone into a sample cleanly: the part of sampling that happens before sample-looping and multisampling. Cowboy's 1996 article is the wiki's source on the hardware chain (source: s-tw066-sound-recording); the realism debate supplies the sample-rate facts (source: s-tw070-realism-debate).

Why it matters#

An analog-to-digital conversion "can never do better than what you started with" — level, cabling and noise decide the sample before any editor sees it (source: s-tw066-sound-recording). Making a sample that has "No clicks, loop[s] easily, correct pitch, and volume" is hard enough that the scene traded them for years (source: s-tw024-sample-ripping-debate; see sample-ripping).

How to do it#

Sources and cabling#

• Set the source's output level optimally first; then check the connectors — clean (Cowboy wipes them with a rag and a little WD-40), correctly plugged, shielded 3-conductor cable, kept under about 3 feet so impedance balancing is a non-issue. "Oxygen free copper is a myth."
• Record CDs digitally (fasttracker-2 "or some other direct CD recorder"; OS/2 had it built in) — "This is the way to go."
• Keep analog runs out of the PC case: "The inside of your computer case has more RF crap than any other source in your home"; a scope test found up to −40 dB of pickup on a CD-audio loop cable routed through the case. If you must, wrap the cable in foil and insulate it, or run a lead from the drive's outside plug to the card's line-in outside the case.
• If you insert an EQ, synth or mixer, use one with active, not passive, filtering (a bass/treble/bass-boost box was a sub-$25 build in 1996).

(all: s-tw066-sound-recording)

Sound cards (1996)#

On Cowboy's Fluke scope the GUS had the least distortion in playback and recording — S/N 72 dB at 50 Hz to 81 dB at 8.5 kHz — with the AWE32 "a CLOSE second" (70 dB at 45 Hz to 79 dB at 7 kHz); a foil-and-cardboard shield around the card, one layer grounded to the case, might help but was untested (source: s-tw066-sound-recording). The listener's card sets the size budget, not the quality: a 1996 module was expected to play on a stock SB16 / SB32 / PAS or a 1 MB GUS (source: s-tw070-realism-debate).

Formats and rates#

• Rates, from the realism debate: telephone 8 kHz, radio 16–22 kHz, the typical module sample 22 kHz (most S3Ms then were 8-bit 22 kHz), CD 44.1 kHz, DAT 48 kHz. A 44.1 kHz sample reproduces everything up to its Nyquist frequency of 22,050 Hz — beyond hearing — so "A 44100 hz sample, 16 bit, is crystal clear"; the limit is memory, not "digitalness" (source: s-tw070-realism-debate). Record high and convert down (see chip-samples).
• Cowboy's file-format notes: WAV up to 44.1 kHz stereo, but keep samples under about 2 MB so playback does not skip; avoid AU and SND except for effects; he states IFF as 8 kHz mono only (source: s-tw066-sound-recording; see To verify).

Recording the mixdown#

For a tape of a finished module: ordinary hi-bias cassettes. In a lab comparison an expensive metal tape had more noise than a $1.50 one (but helped below 200 Hz) and CrO2 was only ±2 dB better across 200 Hz–10 kHz; no cheap cassette holds up above 16 kHz or below 200 Hz — reel-to-reel if it matters; DAT needs 48 kHz and digital out, which few cards had (source: s-tw066-sound-recording).

Rigs in the interviews, 1995–98#

• 1995. leinad sampled on an Amiga 500 with AudioMaster III because it beat his GUS for quality (source: s-tw002-interview-leinad); blaze-runner sampled the AWE32's own synth and mastered through the card's digital out to DAT at 44.1 kHz/16-bit (source: s-tw008-interview-blaze-runner); stalker sampled a loop from a Sega CD game disc played in an audio CD player (source: s-tw004-interview-stalker). vicious's GUS ACE "can't sample, it has no recording abilities", so sample CDs were his source — "I prefer 16 bit 44khz samples. They're much cleaner" (source: s-tw031-interview-vicious); mick-rippon delayed his debut musicdisk until he owned a keyboard to make his own samples (source: s-tw041-interview-mick-rippon).
• 1996 budget rigs. acidfrog's mixer, turntable, tape deck and CD player feeding a GUS MAX in a 386DX40, hi-fi from garage sales (source: s-tw044-interview-acidfrog); vadimvs on a borrowed P75/SB16 with no money for sampling hardware (source: s-tw050-interview-vadimvs); scirocco's Sound Blaster Pro of 1993 (source: s-tw054-interview-scirocco); clef waiting for a GUS PnP to sample at all (source: s-tw055-interview-clef). Shawn makes his own samples, rips some and credits them (source: s-tw041-interview-shawnm); ganja-man samples one sound he likes from a CD and builds from it (source: s-tw052-interview-ganja-man). u4ia's gear: a 486-33 with an SB16 (after an Audio Trix Pro, "as nothing supports it") plus a partner's studio — Akai S-950, E-mu ESI-32, Cheetah MS6, Roland JD-990, Alesis Quadraverb, a Studiomaster 16-channel mixer, Cubase on an Atari STe (source: s-tw067-interview-u4ia).
• 1997–98. Cosmic/RR's chain: CD, vinyl and keyboards, edited in the fasttracker-2 sampler, GoldWave and Cool Edit, sometimes a Kurzweil K2000; a GUS 2.4 with the 16-bit daughterboard (source: s-tw090-interview-cosmic). mental-floss's next purchases: "A sampler with filters", then an effects box — samples give "a bigger sound stock … and more control" than synths (source: s-tw119-interview-mental-floss).

Recording a sound chip#

Recording hardware you are playing yourself is a different job from sampling a record, and little-scale's chip sample packs (2008–12) are the wiki's worked method for it (source: s-little-scale-chip-sample-packs):

• Build the interface, don't use the console's output. Everything was "recorded from hardware using custom designed interfaces", taken off the chip and before the machine's analogue stage — the SP0256 pack omits "the two pole filter that normally follows the audio output from the chip", the SAA-1099's "audio output stage of the sound chip has not been filtered". His reason is the one that makes a recording reusable: "you can always add a digital filter via software if you so desire." You cannot subtract one afterwards.
• Generate the material yourself, and say so. "I have created all of the sounds myself on the hardware (as in, these are not sounds or samples recorded from games or demo programs)" — repeated across every pack. That sentence is what separates a chip recording from sample-ripping, and it is worth writing into your own release notes.
• Capture high, process later. 16-bit 44.1 kHz mono WAV (with an MP3 mirror for browsing), "normalised and trimmed, but not compressed nor equalised in any fashion", with fades added only to stop clicks. Down-sampling and filtering belong at the point of use — the kit-preparation numbers are on chip-samples — not in the archive copy.
• State the coverage. Each pack says exactly what is in it: the C64 pack is "every C and G note from C-2 to G 6" of triangle, saw, pulse (duty ≈ 50 %) and noise, from an 8580 — so it documents that revision, with no filter, ring, sync or envelope variation (sid, multisampling).

Recording for a target that is smaller than you are#

If the recording is destined for an lsdj kit or any 8-bit sampler, decide the target before you press record, because two of the steps are not reversible. Know the playback rate and its Nyquist limit, keep the take clean of low-frequency rumble and DC offset so that normalisation has something to work with, and expect to normalise and then push 3–6 dB further into clipping on purpose, because sample playback is quiet against the chip channels: "yes, it looks terrible but it doesn't sound too bad". The full preparation chain — resample, crop in three places, lowpass at the Nyquist limit, export mono 8-bit PCM — is on chip-samples, with an Audacity recipe (sources: s-little-scale-lsdj-kits, s-chipmusic-lsdj-kit-threads; levels-and-headroom).

Tips & pitfalls#

• Level first, cable second, card third.
• Check the result on speakers as well as headphones before trusting it (source: s-tw119-reverb-and-audio-gear; see reverb-and-compression).
• Editing after recording — normalising, filtering, reverb and compression — is on chip-samples and reverb-and-compression. The one rule worth repeating here: when compressing a sample to bring it forward, set the attack just longer than the sound's own attack phase (a 14 ms attack → 15 ms) so the transient is never crushed, and normalise after compressing, not before (source: s-tw118-reverb-and-compression).

To verify#

• Cowboy's "IFF only supports 8kHz mono" is a statement about his tools, not about the format. Two wiki sources say so indirectly: openmpt lists what it loads as "Amiga IFF (8SVX / 16SVX / MAUD)", so the family includes a 16-bit member (source: s-openmpt-manual-samples), and protracker both saves samples "with IFF-loops" and offers "Load Loop: Will load loops from an IFF file", so the container carries metadata beyond raw 8-bit data (source: s-protracker-23-readme-and-docs). Whether an 8SVX chunk states its own sample rate is still not stated by any source here (unverified) — the format specification itself would settle it.

Recording a Game Boy: ProSound#

The game-boy's headphone output runs through an amplifier that adds hiss. The ProSound mod taps the signal after the volume potentiometer — so the knob still works — and sends it out of the headphone jack; when it is done internally the speaker comes out, since it is fed from the same amplifier. Invented by Timothy Lamb (Trash80), and little-scale believes Nanogrrl was first to fit it to an original grey DMG, "but I could be mistaken" (source: s-little-scale-overclocked-game-boy).

Related#

sample-looping · multisampling · chip-samples · sample-ripping · reverb-and-compression · realism-in-tracked-music · fasttracker-2 · gravis-ultrasound · levels-and-headroom

Sources#

s-tw066-sound-recording · s-tw070-realism-debate · s-tw024-sample-ripping-debate · s-tw119-reverb-and-audio-gear · s-tw002-interview-leinad · s-tw008-interview-blaze-runner · s-tw004-interview-stalker · s-tw031-interview-vicious · s-tw041-interview-mick-rippon · s-tw044-interview-acidfrog · s-tw050-interview-vadimvs · s-tw054-interview-scirocco · s-tw055-interview-clef · s-tw041-interview-shawnm · s-tw052-interview-ganja-man · s-tw090-interview-cosmic · s-tw067-interview-u4ia · s-tw119-interview-mental-floss · s-little-scale-overclocked-game-boy · s-openmpt-manual-samples · s-protracker-23-readme-and-docs · s-little-scale-chip-sample-packs · s-little-scale-lsdj-kits · s-chipmusic-lsdj-kit-threads · s-tw118-reverb-and-compression