Press a key slowly and watch the screen. The letter appears while the key is still moving — with perhaps two more millimetres of travel underneath it, and the firm stop at the bottom happening well after the machine already knows what you typed.
That gap is not a defect and it is not slack. It is the design, and it is the single most useful thing to understand about a mechanical keyboard, because almost everything sold as a performance feature is really a claim about where in that stroke the signal happens.
Where in the stroke the signal happens
A conventional mechanical switch has about 4 mm of total travel and closes its contact at around 2 mm. Half the distance your finger moves is after the keyboard has already reported the press.
There is a second point that gets less attention: the switch does not release at the same depth it actuates. It stays closed until the stem rises slightly above the actuation point — a deliberate gap called hysteresis, which exists so that a finger resting near the threshold does not produce a stream of presses and releases as it trembles.
Those two numbers, actuation depth and the reset gap, are the switch's whole electrical personality. Everything else — the click, the bump, the sound, the weight — is about how the stroke feels and has nothing to do with when the letter appears.
Bottoming out is a habit, not a requirement
Since the signal happens at the halfway mark, the second half of every keystroke is optional. Most people bottom out anyway, and it is worth being honest about what that costs and what it does not.
It does not cost accuracy or speed in any way that shows up in a typing test. What it costs is force and impact. A typical switch actuates at 45 to 55 grams of force; driving it to the bottom takes more than that, and then delivers the remainder as a small collision between the stem and the housing. Multiply by several thousand keystrokes in a working day and the difference between a typist who floats at the actuation point and one who drives every key into the floor is a real difference in accumulated impact — which is why dampened switches and rubber landing pads exist, and why they change comfort far more than they change performance.
Learning not to bottom out is possible and it takes weeks, not minutes. A tactile switch helps, because the bump gives your finger a place to stop that it can feel. It is a legitimate reason to prefer tactile over linear for long typing sessions, and it has nothing to do with the marketing case for either.
The millisecond arithmetic that the shopping ignores
Here is where the subject gets interesting, because the feature people pay most for is the smallest term in the sum.
Break down what happens between your finger moving and the character arriving:
- Travel to actuation — about 2 to 4 ms on a standard 2 mm switch; 1 to 2 ms on a short-travel switch actuating at 1.2 mm.
- Debounce — the controller waits to be sure the contact has settled, because a metal contact bounces for 1 to 5 ms as it closes. Conventional firmware waits 5 to 10 ms.
- Matrix scan and processing — roughly 0.5 to 1 ms.
- Polling — 1 ms at 1000 Hz, less at higher rates.
Total: something like 8 to 16 ms for an ordinary board, and 3 to 6 ms for one built deliberately for low latency.
Now put the shopping decision against that. Moving the actuation point from 2 mm to 1.2 mm buys you between 1 and 2 milliseconds. The debounce wait, which appears on no product page, is 5 to 10. You are being sold the smallest term in the chain while the largest one goes unmentioned.
What magnetic switches actually change
Which sets up the thing worth knowing about Hall-effect and optical switches, because it is not what the advertising leads with.
The headline is the adjustable actuation point: set it to 0.2 mm, react sooner. From the arithmetic above, that is worth a couple of milliseconds.
The real change is underneath. A magnetic switch has no contacts to bounce, so there is nothing to debounce. Instead of waiting 5 to 10 ms for a mechanical contact to settle, the controller reads a magnetic field through an analogue-to-digital conversion that completes in well under a tenth of a millisecond. That is the multi-millisecond saving — and it comes from the sensing method, not from where you set the threshold.
So the honest summary is: magnetic switches are meaningfully faster, and the reason is the term nobody advertises rather than the one on the box.
Rapid trigger is a separate feature and often the one you want
The other capability that arrives with analogue sensing is frequently confused with actuation depth, and it answers a different problem entirely.
On a conventional switch, once a key is pressed it stays registered until the stem rises above the reset point. Press, release halfway, press again, and the second press does not register — the key never got high enough to reset.
Rapid trigger changes the rule from a fixed depth to a change of direction: the key releases as soon as it starts moving up by a set amount, and re-actuates as soon as it starts moving down again. For anything involving repeated taps of the same key — most obviously the sideways movement keys in a competitive shooter — this is a genuine change in what the hardware can express, and it is not measured in milliseconds of latency at all.
It is also entirely wrong for typing, where it will produce doubled letters from any finger that wobbles.
What this means if you mostly type
None of the above is the reason to choose a keyboard for writing, and it would be a poor article that pretended otherwise.
For typing, the numbers that matter are the ones about feel: the force curve, whether there is a tactile bump and where it sits relative to actuation, the sound, the stability of the larger keys, and the height and angle of the whole board relative to your wrists. A 2 ms difference in actuation is imperceptible in prose. A switch that is 20 grams heavier than your hands want is something you will feel by the afternoon.
And a very shallow actuation point is actively bad for typing. Set the threshold at 0.5 mm and fingers resting on the home row will trigger keys. The adjustability that helps a gamer is a liability for a writer, which is why boards that offer it also offer per-key profiles.
What to do
- Find out your board's debounce setting before shopping for switches. On boards with open firmware it is adjustable, and dropping it from 10 ms to 3 or 5 is free and worth more than any switch swap.
- Choose the switch on feel, not on actuation depth — unless you are playing competitively, where the analogue sensing and rapid trigger are real.
- If you type all day, try to stop bottoming out, and pick a tactile switch whose bump sits at the actuation point to help you learn.
- Match the actuation setting to the task if your board allows it. Shallow for the movement keys, conventional for everything else.
Two things we would not buy
A keyboard chosen for a 1 mm actuation point as a latency feature. That is one to two milliseconds bought at a premium, sitting inside a chain that has a five-to-ten millisecond wait in it that the same product page does not mention.
And rapid trigger for a machine that is mostly used for writing. It is an excellent feature aimed precisely at a problem you do not have, and its side effect — doubled characters from an unsteady finger — is one you will meet on your first long email.
How this was put together
The travel and actuation figures come from the switch manufacturers' own datasheets, which specify actuation force, actuation depth and total travel with tolerances, and from the analogue switch specifications that state their adjustable ranges. The latency breakdown — contact bounce duration, conventional debounce windows, scan and polling contributions, and the totals they add to — comes from the measurement work published on keyboard input latency, which instruments the board rather than timing it in software, and from the firmware projects whose debounce implementations are open to read.
The derived point is ours: that the 1-to-2 millisecond saving sold as a shorter actuation point sits underneath a 5-to-10 millisecond debounce wait that is never quoted — so the feature on the box is the small term, and the real advantage of magnetic switches is that they delete the large one.
