Device Bench

Device test · Input

Keyboard Tester

Press every key once: each one lights up when it registers. Nothing you type leaves this device.

Key— Code— Keys working0/0 Stuck0 Chatter0 Held0 Max rollover0 Non-mod peak0 Tested0 Combos0/8 Blocked0 Capsoff

Tab, Enter and Space are being tested instead of operating page controls. Press Escape to use Tab for navigation.

Verdict

NOT RUN

—

Type a few sentences on the diagram above to check it. Chatter is the same key registering twice under 30 ms apart.

Published thresholds: 0 chatter repeats on a key is PASS. 1–2 is CHECK SETTINGS — a single bounce can happen on a clean switch. 3 or more repeats on the same key is LIKELY FAULT. A key reported held for 5 s or more with no release is CHECK SETTINGS (lift your fingers: if it stays lit, it is stuck); 15 s or more is LIKELY FAULT. The worse of the two results is the verdict. Rollover and the keys never seen by this page (see below) are reported but do not change this verdict: they describe what the browser can observe, not a pass or fail. Full reasoning in the methodology.

If it's flagged

    Checking how many keys register at once for gaming? The key rollover and ghosting test runs preset combos such as WASD + Space + Shift with its own verdict. The two panels below are the quick version.

    Key rollover / NKRO test

    Hold several keys together, then add another. Max rollover is the most keys reported held at the same moment; Non-mod peak is the same count without Shift, Ctrl, Alt and the OS keys. Try the combinations you use in games or shortcuts as well as a wide spread.

    Hold several non-modifier keys, such as letters, to start the rollover check.

    Most keys reported at once: —

    Modifiers in that set: none

    Most non-modifier keys reported at once: —

    The page sees only the keys the keyboard sends, never your fingers, so every number here is a floor: the board can do at least that much, and its limit may be higher.

    Keyboard ghosting test

    Hold each three-key combination below together for about a second and watch "Reported now". When all three register at once, the row records "all 3 reported". If you held the whole combination but one key never appeared, release the keys and press the "Held all 3" button on that row. On a keyboard wired as a plain grid, firmware usually handles a combination it cannot resolve by dropping one of the keys, so what you notice is a key that goes dead in one particular combination.

    CombinationReported nowResultIf a key never arrives

    No combination run yet.

    A blocked key sends nothing, so the page cannot see it. That is why "blocked" is recorded on your report and shown next to the largest set the page did see together.

    Keys the page cannot see

    Some keys and combinations never reach a web page. From inside a browser, an intercepted key and a dead key look the same, because nothing arrives, so the diagram leaves them untested and never marks them as failed. Some shortcuts can also switch tabs or take you away from the test.

    How this test works

    The tester lights a key on its keydown event and releases it on the matching release event. Two properties matter: code names the physical key position whatever your language layout, and key is the character or action your layout produced. The diagram maps by code and shows US labels, so a key lights in its physical position whatever your layout prints on it (same switch, different legend), while the Key readout shows the character you typed. Codes outside this reference layout appear in the extra-keys list.

    The classic USB boot-keyboard report has six ordinary key slots and separate bits for modifiers, which is why the page keeps a non-modifier peak as well as the all-key peak. Each peak stores the keys that were down when it was reached, since the two records can come from different moments. The browser does not reveal which report format your keyboard uses, and a combination that registers in full tells you about that combination only.

    The guided combinations are groups of nearby keys and common game controls (WASD and the letters around it, plus Shift, Space and the arrow keys), because keys that sit close together in a typical wiring grid are where a keyboard without per-key diodes runs out of combinations it can tell apart. They leave out Ctrl, Alt and the OS keys to avoid common system shortcuts. If you report a combination as blocked, the row keeps your report beside the largest subset recorded so far, and a later attempt that registers all three replaces it with "all 3 reported".

    Frequently asked questions

    How do I test key rollover or NKRO here?
    Hold as many keys down at once as you can and watch Max rollover and Non-mod peak. Non-mod peak matters most: it leaves out Shift, Ctrl, Alt and the OS keys, which use separate bits in the classic six-slot USB boot-keyboard report. A non-modifier peak of seven or more shows the keyboard is sending more than that report's six key slots. NKRO means every key can register at once, so ten keys here is a good sign, but other combinations can still behave differently: test the ones you use.
    What does the ghosting matrix show?
    Which three-key combinations your keyboard reports in full: a row passes once the page sees all three of its keys held at the same time. If you held all three and one never appeared, press that row's "Held all 3" button to record it as blocked. Other non-modifier keys that show up during a combination and are not part of any listed combination are listed below the matrix without being called ghosts, because the page cannot tell a stray press from an unintended event.
    Which keys might not reach this test?
    Fn, which the keyboard often handles itself without sending an event. Alt+Tab, Command+Tab and similar system combinations can be claimed by the operating system first, and this page lets Ctrl and Command shortcuts through to the browser. On Windows, Print Screen reports only when released, so it lights when you let go. An F row in media mode sends media codes until you press F-Lock or hold Fn. Anything that never arrives is simply left untested.
    Does this work with AZERTY, QWERTZ, or other layouts?
    Yes. The tester follows the physical position of each key (the KeyboardEvent code), so every key lights its own position on any layout. The diagram shows US legends: on AZERTY, the key printed A lights the Q position, and the Key readout shows the A your layout typed. Anything the US diagram lacks is listed with the extra keys.
    My key double-types. Is it broken?
    Not necessarily; it may just need cleaning. Press the key slowly several times and watch Chatter, which counts presses of the same key less than 30 ms apart, then compare a nearby key. One press registering twice is often a worn or dirty switch, though firmware, the connection or remapping software can also send duplicate events. Try cleaning under the keycap with compressed air first, following your keyboard's service instructions. On a mechanical keyboard, a switch that keeps chattering usually needs replacing; on a laptop it usually means a keyboard service.

    Related tools