One question: could a microphone-triggered drum app ever feel playable on
this device? Measures the real round-trip latency, then lets you play against it
and find the point where it stops feeling connected to your hand. No classification and
no calibration — onset detection only. If you just want to hit a desk and hear a
drum with none of this apparatus, use desk drum.
Before you start
Round-trip test needs the speaker on — phone flat on a desk,
nothing covering it, quiet room.
Play-along tests want headphones — wired if you have them.
Bluetooth adds 100–300 ms and will read as unplayable.
Turn the volume up to about 3/4.
Hit the desk with a pen 5–10 cm from the phone.
Input
-70-50-30-150 dBFS
Onset detected
Round-trip latency
The page fires a sharp impulse out of the speaker and times how long
until the microphone hears it. That total — output buffer + speaker + air + mic +
input buffer — is the delay you will feel between hitting the desk and hearing the drum.
Phone flat on a desk, speaker unobstructed, room quiet.
not measured yet
Samples accepted
0
Median
—
Min / max
—
Spread (σ)
—
Reported by browser
—
Unreported input path
—
Play — onset to click
Every detected onset fires a click. Try to play a steady beat.
You are hearing the real acoustic tap immediately and the click one round trip later,
so what you are judging is the flam between them. That double-hit feel is the whole
verdict on this idea — no amount of classification work removes it.
Onsets
0
Last peak
—
Last interval
—
Implied tempo
—
Clipped blocks
0
Suspected echo retriggers
0
If hits vanish, lower the threshold. If one tap
fires twice, raise the hold-off. Clipping means the mic is saturating — move the phone
further from the strike; saturated hits flatten velocity and wreck any future classifier.
Metronome — can you stay in time?
Play along with the click. Each hit is measured against the nearest beat.
Spread (σ) is the number that matters — a constant offset your brain
learns to compensate for; inconsistency it cannot. Run the round-trip test first and the
offset below is corrected for it.
Headphones are not optional for this test. On the speaker the
mic hears the metronome itself and logs it as a perfect hit, so the numbers
describe the phone playing along with itself rather than you.
Hits counted
0
Mean offset (raw)
—
Mean offset (corrected)
—
Spread (σ)
—
Feel it
A number does not tell you whether something is playable. This does.
Two trigger paths fire the same drum, and a delay you can dial in on top of both.
Sound
Tap hereon-screen trigger
Hit the desk → sound
run test 1 first
Tap the pad → sound
—
How to use this. Start with
the pad, delay at 0 — that is the fastest this device can respond to anything. Tap a
steady beat, then raise the delay until it stops feeling connected to your finger.
Whatever number that is, it is your threshold, and it is worth more than any
figure I could quote you. Then hit the desk and see which side of it the microphone
lands on.
Two honest caveats. The pad is not a zero-latency
reference — a touchscreen has its own input lag, and the tap has to cross the main
thread, neither of which is measured here. And with headphones on you lose the
acoustic tap of pen against desk, which is half of what makes the delay obvious;
the flam is starkest with earbuds or one cup off.
Device
Microphone track settings actually granted
Report
—
Deliberately no service worker and no caching — a stale probe would lie to you.