The digital gear of the 1980s was limited in exact, measurable ways: short word lengths, low sample rates, no dither, steep filters. Those same limits are what the ear reads as musical.
Ask why a track cut on a Fairlight, an SP-1200 or an early digital reverb still turns heads, and you'll rarely hear "because it was accurate." You'll hear weight, grit, glue, air. That's worth taking seriously and it's measurable.
The converters: short words, honest noise
Every digital instrument lives or dies at two points: the A/D converter on the way in, and the D/A on the way out. In the 1980s both were young. Word lengths were short — 8, 12, sometimes 16 bits — and the converters were successive-approximation or R-2R ladder designs whose linearity drifted with heat and age.
The maths sets the floor. Each bit buys about 6 dB of dynamic range (6.02·N + 1.76 dB, exactly). Eight bits gives roughly 50 dB. Twelve about 74 dB. The 16-bit CD standard about 98 dB. A modern 24-bit converter reaches ~144 dB on paper and ~120 dB in practice, with distortion below −100 dB. On the numbers alone, today's gear is cleaner by a factor of thousands.
But "cleaner" is the point. An 8-bit sampler like the Ensoniq Mirage or the Fairlight CMI, or the 12-bit E-mu SP-1200 at 26.04 kHz, doesn't hide its noise floor — it wears it. And because the signal was rounded, not dithered, the error correlates with the music instead of sitting under it as neutral hiss.
Correlated error isn't neutral noise. It's low-order harmonic distortion that moves with the music. That is the "grit."
The bandwidth: a ceiling you can hear
Low sample rates put Nyquist down where you can hear it. The SP-1200's 26 kHz gives a usable top barely past 13 kHz. To stop the images folding back, designers used steep analog reconstruction filters right at the edge of the band — brickwalls that rang, rippled and smeared phase across the whole top octave. Modern converters oversample and filter gently. The 1980s couldn't. The price was a coloured, slightly soft, phase-shifted top — the exact thing people call "vintage."
Samplers added a second trick by accident. Play a sample back at a new pitch with no per-voice anti-aliasing and you get alias tones — but they track the note, locked to the source and harmonically related, not random. The ear hears them as timbre, not as fault. It's why a chord stab pitched down on an early sampler has that metallic sheen.
The chips: fixed-point, no dither, companding
The DSP between the converters was fixed-point and short on headroom. Multiplies were truncated back to the word length, usually without dither, so each stage added its own small, signal-dependent distortion. Some machines companded — the Emulator II, for one — spending more bits where the ear is sensitive and fewer where it isn't. None of it was transparent. All of it was consistent, low-order and coherent.
Why the limits were musical
The pattern is clear. The artefacts were low-order (2nd/3rd harmonic, not harsh), coherent (locked to the signal), and gentle where it matters — a soft, band-limited top and a steady floor. Modern converters remove all of it and hand you the recording as it was. That's the right default for mastering. It just isn't the only thing a musician wants.
What we do with it
At MEY this isn't nostalgia — it's a spec. The Living Digital Circuit models a virtual converter and signal path on purpose: a defined bandwidth, a chosen quantisation behaviour, a noise floor you can point to. And because it's modelled, it's measurable. We run every programme through the MEY Test Bench and read the numbers back — the roll-off, the harmonic profile, the decay — so the character is a design decision with a graph behind it, not an accident we can't reproduce.
The 1980s made musical machines out of hard limits. The honest way to honour that is to understand exactly why — and then be able to prove it.