Corviva · gate G6 · heartbeat sound acceptance

The Lub-Dub Test

Before any of the measurement pipeline gets built, one question: does a synthesised heartbeat actually read as a heartbeat — or as a click track with good intentions? An evening's work, no camera, no pulse. If the answer is no, that is worth knowing now rather than after Stage 4.

Instrument

Four sounds, one fixed rate

60 bpm throughout. The rate never varies — this gate tests the sound, not the measurement.

Headphones or a real speaker — not a phone speaker. A heart sound puts almost all of its energy between 30 and 150 Hz. Measured through a simulated handset speaker, this synthesis comes out at roughly 3% of full scale: you would be judging a faint tick, and rejecting the sound for a reason that has nothing to do with the sound. Making it survive a phone speaker means adding treble until it becomes the click track we are testing against.
rate 60 bpm beat 0

Each variant keeps playing until you stop it. Switching between them is seamless — the beat clock keeps running underneath, so an A/B never realigns the rhythm.

AC-6.1 · the three-minute question

Is it a heartbeat, or a click track?

Play Lub-dub, living and sit with it for three minutes — the length of a real session. Not analytically. The way you'd hear it if the app had just started.

0:00 / 3:00

Judge independently. The criterion asks both of you to describe it without conferring first — one person's "well, sort of" becomes the other's answer the moment it's said out loud.

AC-6.2 · blind comparison

Does a living amplitude beat a steady one?

Two versions below. One holds every beat at the same volume; the other lets each beat's loudness follow pulse strength, which drifts with the breath. Which is which is randomised, and hidden until you choose.

Switch between them as often as you like. Then pick the one you'd rather hear for three minutes.

AC-6.3 · a decision, not a preference

One thump per beat, or two?

A heart makes two sounds per cycle — valves closing, then closing again about a third of a second later. Single impulse gives one; Lub-dub gives both. Two costs slightly more to build and is harder to keep clean at high rates.

Closing the gate

What to send back

Three answers and a sentence each. AC-6.1 needs both verdicts to be "heartbeat" — if either of you says click track, that isn't a failed build, it's a finding: the sonification wants a different aesthetic direction, musical rather than literal, and far better learned tonight than after the pipeline exists.

Your answers stay in this browser only — nothing is sent anywhere. Copy them into the thread when you're done.

Synthesis, as built

These are the numbers the sonification engine inherits if the sound is accepted — recorded here so the build doesn't start from a fresh guess.

ElementBodyOnsetDecayLevel
S1 "lub"52 Hz falling to 38, bandpass Q 1.1, + noise0 ms150 ms1.00
S2 "dub"78 Hz falling to 58, bandpass Q 1.3, + noise320 ms110 ms0.48
Metronome1.6 kHz highpassed click0 ms28 ms0.55
Living amplitude±18% following a 0.25 Hz respiratory drift, plus ±4% beat jitter

S2's level is set by ear-equivalent measurement, not by the number alone: a gain of 0.48 renders at 0.65 of S1's peak, because the higher bandpass passes more of the noise. At equal peaks the pair stops sounding like lub-dub and starts sounding like tick-tock, which is exactly the failure AC-6.1 is listening for.

Amplitude drifts; timing does not. Real pulse intervals also vary with the breath, but this gate holds the rate fixed on purpose, so that what you're judging is the sound rather than the liveliness of a simulation.