The 808
The 808 is one sine with a pitch envelope. It starts at 147 Hz and lands on 49 in 155 ms, and the tail takes 2.4 s to fall 60 dB. A phone speaker plays 1% of it, and 84% of that is the first 50 ms. Clipping it hard buys the phone under 3 dB. Playing it an octave up buys 10. Play it on the keys, or let it play itself.
What it is
The 808 kick, as an instrument you can play, and the first of a new series.
The Synthesis series took the knobs apart one at a time. Patches puts them back together into the sounds they were for, starting with the low end. Each part is a real instrument: a keyboard across the bottom that answers your fingers, your mouse and the letters A to K, and a demo button that plays a pattern in the genre the sound comes from until you touch a key. The picture is a spectrogram of whatever is actually playing, time running left to right and pitch running up, which is where a patch shows what it is.
The 808 is one sine wave. Its pitch starts about an octave and a half above the note and dives to it in tens of milliseconds, and its level decays over a couple of seconds. The dive is the punch and the tail is the boom, and that’s the whole patch. Here it’s on G1, 49 Hz, with clean, soft clipped and hard clipped versions, a phone speaker to listen through, and an octave switch.
How it works
The pitch is read off the sound from the spacing of its rising zero crossings, which puts a reading on every cycle. It starts at 147 Hz, three times the note, and comes within 2% of 49 Hz at 155 ms. Every reading after the first few cycles matches the patch’s own curve to a tenth, and clipping the sine hard doesn’t move a single reading, because a clip doesn’t move a zero crossing. The tail is an RMS envelope, and it takes 2.39 s to fall 60 dB from a 350 ms time constant.
The phone speaker is a second order Butterworth highpass at 250 Hz, which is generous to the phone. The share of the sound’s power that comes through it is what the phone gets to play, and a second number says how much of that share sits in the first 50 ms.
The sound you hear is the same recipe in Web Audio nodes: a sine whose frequency is set three times up and told to approach the note with a 30 ms time constant, a gain envelope with a 2 ms attack and a 350 ms decay, a WaveShaper holding the clipping curve, and a highpass whose Q is -3.01 dB, because Q is in dB on that filter and 1 over root 2 is what Butterworth means. One hit rendered offline through that graph gives 155 ms, 2.39 s, -20.3 dB and 84%, the same four numbers the sines give, and the same again with the phone and the clip switched in.
What surprised me
A phone speaker plays 1% of an 808, and 84% of that 1% is the first 50 ms. The number that matters isn’t how quiet the phone makes it, it’s where the little that survives sits in time. Everything above 250 Hz in a clean 808 is the click of the 2 ms attack and the first two cycles of the dive, and everything after that’s a 49 Hz sine the speaker can’t move. On a phone, an 808 is fifty milliseconds long, and what a phone plays of a trap record is the punch and none of the boom.
I expected distortion to fix it, because that’s what everyone says and does. It barely does. A hard clip at four times drive puts the third harmonic 14 dB under the fundamental, which sounds like a lot, but the third harmonic of 49 Hz is 147 Hz and the fifth is 245, and both are under the 250 Hz corner. The phone gains 2.7 dB. A soft clip gains 1. The harmonics that would help are the seventh and up, and a tanh puts those 30 to 50 dB down. Clipping an 808 makes it louder and dirtier on a full range system and leaves the phone almost exactly where it was.
What does help is the note. Playing the same patch an octave up, at 98 Hz, gains the phone 9.8 dB, because now the third harmonic is 294 Hz and through. The note is the knob, and it’s why bass that translates to phones is written higher than bass that doesn’t.
One instrument note. With the phone filter in the graph and the fundamental gone, the pitch tracker reads the same clipped 808 at 169 Hz instead of 49. Nothing is wrong with it. There’s no 49 Hz left in that signal, so the crossings follow the harmonics that remain, and the number is a fair description of the pitch a phone is left to imply.
What I would do next
Real phones. The 250 Hz Butterworth is a stand-in, and the day’s numbers are only as fair as it is. Measuring three actual phone speakers with a microphone and fitting each one’s rolloff would replace a guess with a curve, and I think the corner is higher than 250 on most of them, which makes every number here kinder to the phone than the truth.
Then the rest of the low end: the reese, the wobble and an FM bass, each playable, each with a pattern, each measured the same way.