Keyboard Chatter Test

What keyboard chatter actually is

Chatter, also called double-typing, is when one deliberate press of a key produces two or more separate key events. You type the and get thhe. On a mechanical switch the usual physical cause is contact bounce: the metal contacts do not close cleanly but rattle for a fraction of a millisecond as they meet, and if the keyboard's debounce logic no longer filters that rattle, the firmware reports it as two presses.

Debounce timing is a firmware setting, typically in the region of 5 to 10 milliseconds, though it varies by design and can often be raised on boards running custom firmware. Chatter tends to appear on one or two specific keys rather than the whole board, and usually gets worse as a switch ages or picks up contamination.

How this test decides an event is suspicious

The page listens for keydown and keyup on the window. For each key code it remembers when that key was last released. When the same code produces a fresh keydown, it records the gap between the release and the new press. If that gap is at or below your threshold — 15 ms by default, adjustable from 5 to 50 ms — the event is listed as a suspected duplicate.

One rule matters more than all the others: every event where the browser sets event.repeat to true is discarded before detection runs. Those are operating-system auto-repeat events generated while you hold a key down, and they arrive far faster than any chatter threshold. A test that counts them reports hundreds of imaginary chatter events the moment you lean on a key. This page counts them separately and shows the total so you can see they were handled, and the same is true of untrusted synthetic events, events that arrive while the window has lost focus, and events with an unusable timestamp.

Why a flagged event might not be chatter at all

The honest reading of a flagged event is “two presses of this key reached the browser unusually close together”. Several ordinary things produce exactly that:

• Fast intentional double taps. Gaming and shortcut habits produce genuine double presses, though almost never inside 15 ms.
• Rollover during fast typing. When keys overlap heavily, releases and presses can land very close together in the event stream.
• Remapping, macro and layer software. Anything that synthesises key events — AutoHotkey, driver macro engines, keyboard layer tools, text expanders — can emit bursts the browser reports as ordinary presses.
• Virtual machines and remote desktops. Input is re-injected on the far side, so event timing is reconstructed rather than original.

Real chatter has a shape: it clusters on one or two keys, the gaps are consistently tiny, and the press that precedes the duplicate is often held for only a few milliseconds. A single isolated flag across hundreds of presses is much more likely to be one of the causes above.

Filter Keys, Bounce Keys and Slow Keys

Operating-system accessibility features sit between the keyboard and the application, and they can both hide and create odd repeat behaviour. On Windows, Filter Keys covers Bounce Keys (ignore a repeated press of the same key that arrives too quickly) and Repeat Keys / Slow Keys (change how long a key must be held before it registers). macOS has an equivalent Slow Keys setting alongside its key-repeat controls.

If Bounce Keys is active, real chatter may be filtered out before the browser ever sees it, and this test will report nothing while you still see double letters in some other application that is fed differently. If Slow Keys is active, presses are delayed and reshaped, so the gaps on this page no longer describe what the keyboard sent. Check these settings before you trust either a clean result or a noisy one.

Reading the gap histogram

The histogram plots every same-key re-press gap, not just the flagged ones. Comfortable typing usually puts the same key down again somewhere around 100 to 400 ms later, so a healthy board produces one broad hump in that region and nothing else. Chatter shows up as a second, separate cluster jammed against the left edge, in the 0–15 ms range, with a clear empty gap between the two groups. That separation is the useful signal; a smear that runs continuously from 5 ms up to 200 ms is much more likely to be typing behaviour than a switch.

Timer resolution sets a floor here. If your browser reports a 1 ms clock, every gap is quantised to whole milliseconds and the fast cluster collapses into a single bar. The measurement environment panel above shows the resolution measured in your browser.

If the pattern looks real

Reproduce it first. Try a different USB port and cable, quit remapping and macro tools, switch off Filter Keys or Slow Keys, and if you can, try the keyboard on a second computer — that separates the board from the software on any one machine.

If the same key still produces tight clusters everywhere, the common remedies are raising the firmware debounce time where the keyboard supports it (QMK, VIA and several vendor configurators expose this), cleaning or reseating the switch, or replacing it on a hot-swap board. The CSV export from this page is a concrete thing to attach to a warranty ticket.

Read the full guide to keyboard chatter, or measure overall input timing with the keyboard latency test.