Keyboard Chatter Test

What this test checks

Key chatter is a switch fault where a single press is registered two or more times, so you type "heello" when you pressed E once. Our chatter test timestamps every keydown event in your browser and measures the gap between consecutive presses of the same physical key. Human fingers cannot repeat a key in under roughly 30 milliseconds, so any pair closer than the threshold is flagged as suspected chatter.

The threshold is adjustable. 30 ms is a sensible default; 20 ms is stricter and only catches very fast bounces, while 50 ms is more sensitive and may occasionally flag a deliberate, extremely fast double-tap. The per-key table shows how often each key was flagged and its shortest interval, which is the number that matters most when deciding whether a switch needs cleaning or replacing.

How this test works

Inside every keyboard, pressing a key closes a small electrical contact. A healthy contact closes once and stays closed until you release it. A worn, dirty or oxidised contact can flicker open and closed several times in the first few milliseconds of a press, and if the keyboard's controller does not smooth that flicker away, the computer receives two or three separate key presses from one physical action. That flicker is what people call key chatter, switch bounce or double typing, and it is one of the most common faults on heavily used mechanical keyboards.

Our chatter test listens for the keydown events your browser receives and reads two pieces of information from each one: the physical key code, which identifies the key position independently of language layout, and a high-resolution timestamp recorded when the event was generated. For every key we remember the time of the previous press, and when the same key arrives again we compute the interval between the two. Events that the operating system marks as auto-repeat, which happen when you hold a key down, are discarded before they reach the comparison, so a long press can never be mistaken for chatter.

The decision rule is simple: if two consecutive presses of the same key are closer together than the selected threshold, that pair is counted as a suspected chatter event. The test does not need to know what a normal typing rhythm looks like, because the threshold sits far below anything a human can do. Even very fast typists need well over 50 milliseconds to lift a finger and press the same key again, so a 10 or 15 millisecond gap can only come from the hardware.

Because the raw intervals are kept rather than only the verdict, you can switch the threshold between 20, 30 and 50 milliseconds at any time and the table is recalculated instantly from the same session. That makes it easy to see whether a borderline key is genuinely bouncing or just sitting near the edge of a sensitive setting.

Reading the results

The three counters at the top summarise the session: how many presses were recorded in total, how many pairs fell under the threshold, and how many distinct keys produced at least one such pair. The verdict line beneath them turns into a warning as soon as any key is flagged. A clean session after a couple of hundred presses across the keys you use most is a good sign that the switches are behaving.

The flagged keys table is where you make the real judgement. For each key it lists the total number of presses, how many of those were flagged, and the shortest interval ever seen on that key. A key with dozens of presses and a single flag at 28 milliseconds is probably fine; a key with a handful of presses, several flags and a minimum interval of 2 or 3 milliseconds is almost certainly bouncing. Real chatter tends to show intervals in the low single digits because the contact flicker is electrical, not mechanical.

The recent presses list shows the last ten events with the interval since that same key was last pressed, with flagged pairs highlighted. Use it when you want to watch a specific key in real time: press it once, pause, press again, and see whether an unexpectedly short interval appears beside it. A first press of any key shows no interval because there is nothing to compare it with yet.

When a key is flagged, repeat the test rather than trusting one result. Bouncing contacts are intermittent, and a key that doubles only occasionally may need several dozen presses before it misbehaves. Conversely, if a key flagged once never flags again across fifty more presses, the original event was most likely a hiccup in the operating system or browser rather than the switch.

When to use this test

The most common reason to run a chatter test is that doubled letters have started appearing in your typing: "thee", "annd", or a space bar that inserts two spaces. Rather than guessing which key is responsible, type a few sentences here and let the per-key table point to it. The fault is usually confined to one or two keys, typically the ones that get the most use, such as the vowels, the space bar, Backspace and Enter.

It is also a useful check when buying or selling a used mechanical keyboard. Chatter develops gradually with wear and does not show up in a quick visual inspection, so a two-minute pass over every key gives a much better picture of the board's condition than its appearance does. Run the test once before cleaning or lubing switches and once after, and you will be able to tell whether the work actually helped.

Gamers should run the test after changing debounce settings in their keyboard's software. Lowering debounce reduces input latency slightly but increases the chance of chatter on ageing switches; if you have pushed the setting down, this test will tell you whether the board can cope with it. The same applies after a firmware update, since some updates reset debounce to a default value.

Finally, the test is a quick way to confirm that chatter is a hardware problem before you go hunting for software causes. If a key doubles here, in a plain browser page with no editor plugins or macros involved, the switch is responsible and software changes will not fix it.

Limits and things to know

The test measures the events that reach your browser, not the electrical signal inside the keyboard. If the keyboard's own firmware is already filtering bounces, the test will report a clean result even though the contact is physically worn, which is fine from a practical point of view: what matters is what the computer sees. Likewise, a software debounce tool running on your computer will hide chatter from this page as well as from your editor.

Timestamps come from the browser's event clock, which is precise to a fraction of a millisecond in most environments, but the operating system can occasionally delay delivery of events under heavy load. A single flag with an interval close to the threshold during a moment of stutter is therefore not conclusive. Very short intervals, in the range of one to ten milliseconds, are far more trustworthy evidence than intervals just under the limit.

Keys handled inside the keyboard never generate events the browser can see. The Fn key, media controls and vendor macro keys fall into this group and will not appear in the counts. The test also prevents the browser's default reaction to most keys so that Tab, Space and the function keys do not move focus or scroll the page; refresh shortcuts and the developer tools key are left alone so that you can always leave the page.

Accessibility settings can interfere with the readings. Filter Keys and Bounce Keys on Windows are designed to suppress repeated presses and will hide genuine chatter, while Slow Keys can make presses disappear entirely. Turn those features off before testing if you want an accurate picture of the hardware, and turn them back on afterwards if you rely on them.