How to Test a Keyboard Properly

Start by finding out which fault you actually have

A key that "does not work" is doing one of four things, and they look identical from the chair. It might not register at all. It might register on its own but drop out when other keys are held. It might register twice from one press. Or it might register perfectly and simply feel wrong. Those are a hardware failure, a matrix limitation, a worn switch and a preference — and only one of them is worth a replacement keyboard.

The tests below separate them. Run them in order, because each one rules out a class of fault and makes the next result meaningful. Jumping straight to a typing test tells you almost nothing, which is why it is last.

Test 1 — does the key register at all?

Open Keyboard Test and press every key on the board once. Keys light up as they register. Anything that stays dark is not reaching the operating system.

Before assuming a dead switch, rule out the two boring causes. Debris under a keycap is by far the most common, and a can of compressed air fixes more keys than any repair. The second is software: a remapping tool, a game overlay, or a language layout that has quietly moved the key somewhere else. If the key is dark in this test but works in another operating system or on the login screen, the switch is fine and the problem is software.

If a whole row or a whole column is dark rather than a single key, stop. That pattern is a broken trace or a failed connector in the keyboard matrix, not a switch, and no amount of cleaning will help.

Test 2 — rollover and ghosting, which are not the same thing

Now hold several keys at once with the rollover test. This is where most "gaming keyboard" complaints actually live, and the two failure modes get confused constantly.

Blocking is when a key you press simply does not appear because too many others are already held. Ghosting is worse and rarer: a key you did not press appears. Both come from the same cause — the keyboard is a grid of rows and columns, and the controller reads intersections rather than individual switches, so certain combinations are genuinely ambiguous to it.

A membrane keyboard typically manages two or three simultaneous keys before blocking, and the limit depends on which keys, not just how many. That is why W+A+Space fails on one board and works on another. Keyboards advertised as NKRO — n-key rollover — put a diode behind every switch so every combination is distinguishable, and they will report every key you can physically hold down.

If a key fails this test but passed test 1, the switch is fine. You have found a design limit, and the fix is a different keyboard, not a repair.

Test 3 — chatter, the fault that looks like a typing problem

Key chatter is a single press producing two characters. It comes from a worn or dirty switch whose contacts bounce, and the keyboard firmware's debounce window is no longer long enough to hide it.

It is worth naming because of how it presents: you do not notice a hardware fault, you notice that you have started making typing mistakes. People blame themselves for months. If double letters appear in words where you know you pressed once — and always on the same one or two keys — that is chatter, not you.

The key chatter test times consecutive events on the same key and flags a repeat too fast to be human. Cross-check by watching for doubled characters in Typing Test. A switch that chatters is often recoverable with contact cleaner; on a hot-swap board, replacing that one switch costs almost nothing.

Test 4 — latency, and how much of it is not the keyboard

Only now is latency worth measuring, because a key that blocks or chatters will produce nonsense timings anyway.

The honest caveat first: a browser cannot see the keyboard directly. What you can measure is the time from switch actuation to the page receiving the event, and that includes USB polling, the operating system's input stack, the browser and the display. The keyboard is one term in a chain, and usually not the largest.

That is exactly why the useful tool is an input lag budget rather than a single number. It splits click-to-photon across every stage and names the one worth fixing — which for most setups turns out to be the display, not the keyboard. Upgrading from a 125 Hz to a 1000 Hz polling rate saves about 7 ms; moving from a 60 Hz to a 144 Hz monitor saves nearly 10 ms on frame time alone.

Test 5 — only now, the feel

With the electrical faults ruled out, what is left is preference, and preference is worth measuring rather than guessing at.

The typing weakness analyzer times every keystroke and reports the specific letter pairs that slow you down, then builds a drill from them. That is a more useful output than a words-per-minute score, because a wpm number tells you that you are slow and nothing about where. If your slow pairs are all same-finger combinations, the layout is fighting you rather than the switches.

The layout comparison runs finger travel, same-finger pairs and hand alternation over text you paste, which matters because the answer for prose and the answer for code are very different. Before buying an ergonomic board on the strength of a review, run your own writing through it.

What the results actually mean