Press one key repeatedly at a steady pace. We measure interval variance between presses to estimate switch bounce.
Switch bounce lasts about 5 milliseconds. That is the window in which a mechanical keyswitch, after you press it, rattles its metal contacts together and fires a burst of duplicate signals before settling. Keyboard chatter is what happens when that burst grows instead of settling. The keyboard chatter test catches it before a game punishes you for it.
Chatter is also the most misdiagnosed keyboard fault going. Owners blame software, blame the game, blame a spilled drink from 2022. The switch sits there quietly double firing the whole time, and ten seconds of measurement ends the argument.
Inside every mechanical switch, a metal leaf contacts a plate to close the circuit. On a healthy switch the contact lands once and the firmware's debounce routine, about 5ms in most boards running stock QMK firmware, masks the natural bounce. On a worn or contaminated switch the contact arcs or grinds, and the noise window stretches past the debounce window. Two presses arrive where you threw one.
Dust and switch wear are the usual culprits. Cherry rates its MX switches for 50 million keystrokes, and Kailh sells switches rated at 80 million. Ratings describe the mechanism in clean conditions, and real keyboards collect skin oils, desk dust and humidity. A switch at a fraction of its rated life can start misbehaving. The offending key is often the most used one, which for gamers means W, and for everyone else means space.
Pick a key and press it repeatedly, one clean press at a time. The engine records the interval between consecutive presses on the browser's performance.now() clock and watches the spread. A healthy finger on a healthy switch produces intervals that cluster tightly, since repeatable motion lands around 150 to 250ms per press.
Three numbers come back. The mean interval shows your tempo. Standard deviation shows how far your intervals wander. The consistency score turns that ratio into a percentage, and it is the diagnostic that matters. Consistency above 90 percent at a normal press rate means the switch reports what your finger did. Intervals spiking under 30ms in bursts mean the switch fired two or three times for a single press, and that signature belongs to the hardware. Fingers do not move 30 milliseconds apart on purpose.
Judge the switch, not yourself. If one key scores 60 percent consistency while its neighbours score 94, that key is the problem and the data just paid for the test. If every key scores low, slow down, because deliberate presses about a second apart give the cleanest diagnosis. Fast hammering inflates variance on healthy switches and muddies the reading.
Chatter scores matter most in games that punish double inputs. In Counter-Strike a chattering W key lurches your movement. In osu! it counts extra taps you never made, and in an MMO it can fire a cooldown ability twice and waste the rotation. Single player typing mostly forgives it, since autocorrect quietly eats the evidence. That forgiveness is exactly why the fault survives for months unnoticed.
Firmware can silence chatter by extending the debounce window, ignoring any repeat signal inside say 8 to 10ms instead of 5. Some boards ship with a user adjustable setting for this. The repair is real, and the trade is real too. Every millisecond of debounce is a millisecond added to how long a genuine fast double press waits before counting. Masking chatter with debounce is painkiller, not surgery, because the switch is still dying underneath.
The honest fix is the switch itself. On a hotswap board you pull the culprit and seat a replacement, which costs a few dollars and takes two minutes with the right puller. On a soldered board the job needs an iron, so at that point weigh the repair against the warranty. Chatter inside the warranty window is a defect, and the manufacturer knows it.
Run the keyboard chatter test on your most used keys, one at a time, pressing about once per second. A consistency score far below your other keys confirms chatter, as do interval spikes under 30ms. In games the symptom is a duplicate input, a double jump or a stutter step you never commanded.
Partly. Raising the debounce window in firmware masks duplicate signals at the cost of a few milliseconds of delay, and it works while the switch degrades slowly. The switch itself only gets worse, so treat debounce as a stopgap.
Yes, because double firing inputs are a hardware defect. File the claim rather than voiding the warranty with a soldering iron.
Lube changes feel more than electrical behaviour. Contamination inside the switch is the electrical problem, and a flush with contact cleaner sometimes restores clean contacts. It is a 10 minute fix worth trying before a warranty claim. Never spray anything into a board that is plugged in.
No. The fault stays inside the switch, and the worst symptom is wrong inputs, though a chattering key in a ranked match can do real damage to your standing.
If inputs vanish instead of doubling, that points at rollover, and the NKRO test maps the ceiling. For raw press speed on a single key, run the keyboard CPS test next and compare your tempo against the benchmarks.