TW #70–73 — the realism debate (#trax log, Behemoth, GBlues, Alan Robinson, Raphael, Sturm, brn, It-Alien, Purple Haze, Necros)

A #trax argument about a tracked saxophone, printed as TraxWeekly's second survey question ("is 'realism' something viable for tracked music?", replies to Psibelius), followed by three issues of answers; Behemoth's later "Realism in MODs" (#76, s-tw076-realism-in-mods) belongs to the same thread.

Key claims#

• #70, the IRC log. in_tense: a sax in a tracked song is futile — "tracker's are not meant to replicate music that already exists"; an intricate sax line will never sound real unless you use one direct sample and write no new melody with it. necros: "you can do a sax right if you know how to multisample and have a smidgen of talent" — listen to greynote.s3m, and the sax in "Orchard Street" on his coming musicdisk is better; he uses "10+ samples of bends, hits, overdrives, diff octaves, and keyclick noises". psibelius: the Grey Note lead "is almost a huge wav file chopped up into ten pieces", which is why it works. catspaw: measuring a tracked song by how "real" it sounds means "you're on the wrong channel, with the wrong attitude" — "it's music". Human: nothing short of "50 megs of samples" gives the variation in tone, attack, dynamics and vibrato of a Mozart violin concerto. Nexus: his synth gives a perfect sax sample. DennisC found "the obscure amiga tune whence necros ripped the point of departure drumloop".
• #71. Behemoth: tracked music has moved "from completely computer-sounding to almost realistic"; exact reproduction would take gigabytes and his sound card has 8 MB; his aim is not a Chopin CD but to sound "as little 'digital' as possible"; most S3Ms are still 8-bit 22 kHz samples, and (his argument) a sampled instrument is chopped into 22,000 pieces a second while a real one is a continuous wave. GBlues: realism "is intrinsically impossible" (every squeak of a split reed) but it can be faked, as greynote.s3m shows — "Done correctly it will work just fine. Done incorrectly, it will ruin your tune." Alan Robinson: it depends on the style — classical needs real-sounding instruments (he has never heard a good classical module), techno "doesn't matter one bit" and often wants what cannot be played live; "A really fake instrument just trashes everything (except in the case of techno)". Mark / Raphael: techno is made with synthesizers anyway, so a module with synth samples is as "real" as the record; refuses to leave computer music to the MIDI people; dreams of a modules-only radio station.
• #72, Sturm (Explizit, "not afraid to call myself Alex Cavaye"): "real" depends on your definition — a guitar through wah, distortion and fuzz is still a real sound; "A mod may sound the same every time you play it, but SO DOES A CD!"; modules are not live music. On Music Contest 4: his ambient trip-hop entry M4V-LITA.IT was judged "too boring" on purpose; he refuses to force out a "demo tune" ("invent a cheesy melody around the same notes as a Purple Motion tune, throw H81 commands in all over the place, and reuse some old bassline from a Skaven tune"); write your own style, judges should not mark tunes down for length, and compos should have no time limit. Handles are scene tradition inherited from graffiti taggers via crackers and coders.
• #72, replies. brn (a coder) corrects Behemoth: a 44.1 kHz sample's Nyquist frequency is 22,050 Hz, above the ~20 kHz the ear can follow, so "the DIGITAL NATURE of samples does NOT make it lose any information" — "A 44100 hz sample, 16 bit, is crystal clear"; gigabytes are only needed for real-time replay, and real-time digital filtering on sound cards would change everything. It-Alien: he tracks "exactly what I 'hear' in my mind" — "Who cares if what I imagine is a Syn Brass and not a real Brass?!" — and asks listeners to judge modules by feeling, not closeness to real sound. Purple Haze (a BBS sysop, not a tracker): the real limit is file size — modules should play on a stock SB16 / SB32 / PAS / 1 MB GUS, or the BBS users blame the tracker; his rate table: telephone 8 kHz, radio 16–22 kHz, the average module sample 22 kHz, CD 44.1 kHz, DAT 48 kHz; until a format compresses 44.1 kHz audio to a usable size, samples "attempt to be real, but are yet so far away"; "battle technology" rather than each other.
• #73, Necros ("andy, necros / five musicians"): can you hear the finger squeaks on the acoustic guitar, the 22 kHz intricacies of the choir, the hall reverb on the kick, in a module? "doubtful. mods aren't meant for this" — otherwise everything would be MP3 needing 300 MHz machines, or 40 MB WAVs. "tracked music is inherently synthetic in a very important sense"; getting close is possible but outside the format's scope except as novelty; "the best tracked music revels in it's contrivedness, and uses the digital domain to come up with a sound that ISN'T the acoustic stuff you hear all day on the radio".

Practical takeaways#

• A convincing solo acoustic instrument takes a set: 10+ samples of articulations (bends, hits, overdriven notes, several octaves, key clicks) mapped in one instrument — see multisampling — or whole sampled phrases jumped between with sample-offset (Grey Note).
• Let the style decide: classical and jazz leads need realistic samples; techno and chip styles do not, and a half-real instrument is worse than an honest synthetic one.
• Sample rates in one table: 8 kHz telephone, 16–22 kHz radio, 22 kHz the typical 1996 module sample, 44.1 kHz CD, 48 kHz DAT; 44.1 kHz / 16-bit already exceeds hearing (Nyquist) — the loss is in the number of samples you can afford, not in "digitalness".
• Budget for the listener's hardware: a 1 MB GUS or a stock Sound Blaster in 1996; see recording-samples and chip-samples.
• Judge (and write) tracked music as music, not as an imitation.

Notable quotes#

• "if you're measuring the worth of a tracked song by how 'real' it sounds, I might suggest that you're on the wrong channel" (Catspaw)
• "tracked music is inherently synthetic in a very important sense" (Necros)
• "A mod may sound the same every time you play it, but SO DOES A CD!" (Sturm)
• "A 44100 hz sample, 16 bit, is crystal clear." (brn)

Contradictions#

Behemoth's "continuous wave" argument (#71) is contradicted by brn's Nyquist argument (#72); the wiki follows brn — sampling theory — and records Behemoth's as the common intuition of the time. Necros's "mods aren't meant for this" (#73) sits beside his own demonstration that a sax can be done (#70): his position is that it is possible but beside the point.

Relevance#

Defines realism-in-tracked-music, the opposite pole of chiptune; the concrete multisampling and phrase-sampling advice lands on multisampling and sample-offset; the rate table on recording-samples. Confirms Necros's Five Musicians membership (his #73 signature).

Pages touched#

realism-in-tracked-music, multisampling, sample-offset, recording-samples, chiptune, chip-samples, instrument-design, necros, psibelius, catspaw, traxweekly.