Does every key still register?
Keyboard tester
Click the map, then press keys — every key you press lights up and stays lit. The browser is told to ignore the keystroke, so nothing is typed anywhere. Press Esc twice to hand the keyboard back.
Held at the same time
0keys
Hold several keys down at once. Under 6 means the controller merges signals (that is ghosting), 6 is the plain USB limit, more than 6 means full n-key rollover.
- Keys checked—
- Last key—
- Repeat delay—
- Double fires—
One gear check a month
MonthlyOne thing a month worth measuring before it fails properly — a mic that reset itself, a stick that drifts, a switch sending two clicks. Two sentences and the test.
Try also
How do you test a keyboard in a browser?
The keyboard tester lights up each key as the browser receives it and counts how many you can hold down at once — that count is your keyboard’s rollover. Press keys one at a time first. Anything that stays dark never reached the page, which points at a dead switch, a tired ribbon cable, or a combination the operating system took for itself.
Then hold several down together. A cheap membrane board usually stops at two ordinary keys plus modifiers, the standard USB keyboard report carries six, and a board sold as NKRO keeps counting past ten. Modifiers travel in their own byte, so Ctrl, Shift and Alt never eat one of the six.
Two things no browser test can show you. The Fn key is resolved inside the keyboard’s own firmware and produces no key code at all, and the operating system claims combinations such as the Windows key, Alt+Tab and print screen before any page gets a turn.
Rollover is the number people argue about without ever measuring it. The table below is what each tier does when you actually hold the keys down, and the only column that matters is the one your board lands in.
| Keys held at once | 2KRO membrane | 6KRO standard USB report | NKRO |
|---|---|---|---|
| Two, plus Ctrl, Shift and Alt | all register | all register | all register |
| Three in the same matrix block | third is dropped | all register | all register |
| Four to six ordinary keys | dropped | all register | all register |
| Seven ordinary keys | dropped | seventh is dropped | all register |
| Ten fingers down at once | dropped | six survive | all register |
The modifier row is the one that surprises people. Ctrl, Shift, Alt and the Windows key ride in a separate byte of the report and never occupy one of the six slots, so Ctrl+Shift+W+A+S+D is six keys to a hand and four to the protocol.
Keys are half of an input desk. The typing test measures what those keys are worth in words per minute, the mouse tester does the same job for buttons and the wheel, and the double click test hunts the switch fault that looks like a software bug.
Frequently asked questions
Why does the fourth key stop working when I hold three down?
That is the switch matrix. Keys sit on a grid of rows and columns, and without a diode behind each switch the controller cannot tell a genuine fourth key from a phantom one, so it refuses the whole set. It shows up first around WASD because that is where gaming hands sit. Boards that solve it are sold either as anti-ghosting, meaning a few protected clusters, or as NKRO, meaning every key every time.
What is the difference between ghosting and blocking?
Ghosting means a key you never touched appears; blocking means a key you did touch is dropped. Nearly every keyboard built in the last fifteen years blocks rather than ghosts, because refusing an ambiguous read is safer than inventing a keystroke. The test shows both cases: a lit key you are not holding is a ghost, a dark key you are holding is blocked.
A key types two characters from one press. Is the switch failing?
Usually yes. The contacts bounce as they close, the firmware ignores that bounce for a few milliseconds, and a worn switch bounces for longer than the window allows. Rule out auto-repeat first: a key held down starts repeating after a quarter to a full second depending on your settings, and that is normal. Two characters from one deliberate tap is not, and the cure is the switch.
Nothing happens when I press Fn or the Windows key.
Correct on both, for different reasons. Fn is handled inside the keyboard itself and never sends anything to the computer, so no software of any kind can see it. The Windows and Command keys do send a code, but the operating system intercepts the shortcuts built on them before the browser is offered a turn. Neither result means the keyboard is faulty.
The tester says KeyQ when I press A. Why?
Because the browser reports two separate things and a hardware test uses the physical one. The code value names a position on a US layout no matter what your own layout is, so on AZERTY the key marked A sits where Q would be. The character your layout produces is reported separately. Position never changes, the letter does, and only one of those identifies a switch.
Can I test a laptop keyboard the same way?
Yes, and the pattern of failure tells you what you are dealing with. One dead key is a switch or debris under the cap. A whole row or column dead at once is a broken trace or an unseated ribbon connector, which is a repair rather than a clean. Keys that only misbehave once the laptop has warmed up point at that connector too.
How the test works
Worked example
With keys held down together 6, this page works out 6 keys at once. Six at once is exactly the plain USB keyboard limit, so this is normal, not a fault: the standard boot protocol carries six keys plus modifiers and no more.
- Reported at once6 keys
- Plain USB limit6 keys
- Verdictnormal, not a fault
Sources:
What a dead key usually turns out to be
Order matters when you are diagnosing. Test the key here, then in a plain text field, then on another computer. Each step removes a suspect: this page proves the browser saw it, the text field proves nothing is rewriting it, another machine proves the keyboard itself.
Software steals more keys than hardware breaks. Remapping tools, gaming overlays, accessibility filters such as Sticky Keys and Filter Keys, and virtual machines all sit above the browser and can swallow a keystroke without saying so.
A cluster dying at once is almost never coincidence. Adjacent keys share matrix lines, so five dead keys in a line is one broken trace far more often than it is five broken switches.