Before you start
Choose one question. A missing position, an unmatched release, repeated characters and a dropped combination need different samples. Record the affected key, application and action first. A clear starting observation keeps a successful unrelated control from hiding the symptom you actually need to reproduce.
Choose the layout that represents the board, then compare the received code with the key you intentionally press. KeyboardEvent.code is separate from the character selected by the active layout. Remapping or virtual input can still change delivery, so the map does not directly identify a physical switch from its printed label.
Is a search box, menu or form focused? The test excludes those contexts to protect normal page interaction. Move focus back to the instrument before collecting a sample. Note software that may capture keys globally, but do not assume closing a window necessarily removes every configured mapping or background process.
Run the check
Test isolated keys. Press one ordinary letter, release it fully and watch the position clear. Continue through the relevant block, then modifiers and navigation keys as separate observations. A position that never appears should be compared with another application before any hardware service decision.
Compare modifiers deliberately. Hold one modifier and add an ordinary key that avoids a known system shortcut, then release both. Record the event order rather than only the final highlight. Reserved actions may still occur outside the page even when its shortcut-blocking control is enabled.
For combinations, add ordinary keys one at a time and record the largest held set and the missing event. Repeat another order or region in a fresh run. The keyboard ghosting guide explains how to keep that sample separate from an isolated-key or release complaint.
Read your result
Read proven positions, currently held keys and unmatched releases as different observations. A key can register in one isolated press yet still behave differently during a combination or an intermittent later session. The map cannot turn that sample into a complete certificate that the device is healthy.
Does a key type twice after a full release? Record fresh presses separately from a held key producing repeat=true events. For keyboard double typing, compare the same isolated action in the affected application. A regular interval or a short pair alone cannot prove the operating-system setting or a physical switch is the cause.
Use the keyboard repeat rate test only for automatic held-key repetition. It measures intervals between delivered repeat events, not USB polling. A different repeating interval after one setting change is useful evidence about that setup, but not an independent hardware report-rate measurement.
Measurement limits
The browser cannot inspect the matrix, switch contacts, scan firmware or complete USB input path. It receives events after earlier processing. That is why a missing key or unexpected label needs a delivered-event description rather than a confident physical explanation unsupported by the page.
Key repeat and rollover are separate observations. N-key rollover and 6-key rollover labels need the exact model's documented conditions; neither label is verified by one held group. System shortcuts can be handled before the test receives them. Keep their absence labelled as an incomplete browser observation and compare the required shortcut in its real application.
Timing needs a separate method. The keyboard latency test uses an A-key event timestamp and animation callbacks; it does not observe physical actuation or panel light. Do not mix its milliseconds with an automatic repeat frequency or use either result to verify a device's electrical specification.
What to do next
Repeat before service. Keep the same layout, connection, application and intended action, then change one condition after recording the baseline. A problem that follows the keyboard across applications or computers narrows the investigation but does not remove every mapping or hardware explanation automatically.
If only one application fails, compare its bindings, focus and input settings with the browser sample. Leave the keyboard condition unchanged during that check. A clean browser map is useful evidence, but it cannot inspect the application's private processing or prove that the entire fault belongs to software.
If the symptom persists, record the device, mode, keys, press order and release behaviour for support. Consult model-specific instructions before opening the board or replacing a switch. The guide cannot determine warranty coverage or select a repair from a missing event, a repeat interval or a held-key count alone.
Symptom table
| Symptom | Likely layer | Next action |
|---|---|---|
| An isolated key never appears | The delivered position is missing | Compare the same press in the keyboard test and another application |
| A held combination loses a key | The input route or device constraints need comparison | Record another order and physical group |
| A released key stays highlighted | An unmatched release needs a focused repeat | Clear the run and repeat without changing windows |
Common questions
How can I test a keyboard without typing into a document?
The browser map reads key events delivered to the test area, so a document is not needed. Focus the instrument, press and release an ordinary key and inspect the map and recent log. System-reserved shortcuts may still act outside the page, so do not treat their absence as proof of hardware failure.
Why does my keyboard drop keys when I hold three at once?
A missing simultaneous key is a delivered-input observation that may involve mode, mapping, application or device constraints. Record the exact group and press order, then compare another order and region. The browser cannot identify a matrix or protocol cause solely from the count, or guarantee that an advertised mode resolves every group.
One key types two characters. Is the keyboard broken?
Not necessarily. Separate automatic held-key repetition from a new action after a complete release, then compare the same press in another application. A short pair or a regular interval cannot identify switch bounce or a system setting by itself. Record the sequence before deciding whether a model-specific service check is needed.