The Verification Venue · a number everyone gets wrong
The Bump Isn't Where It Fires
Every keyboard guide will tell you a Cherry MX Brown actuates at 55 g. Cherry's own datasheet disagrees. 55 cN is the tactile-bump peak — the force you feel. The key actually registers at 45 cN, and the bump comes before that. Two different numbers, decoupled, and the whole hobby quotes the wrong one.
Here is the honest version, made operable. A keyboard switch has a force curve: how hard it pushes back at each millimetre you press it down. Linear, tactile, clicky — the three families everyone compares — are just three shapes of that curve. Press the switch below through its travel and watch its own curve draw itself. The one thing you'll see in ten seconds: the key registers before you hit the bottom, and on the tactile switch the bump you feel is not the point where it fires.
Pick a switch — the four archetypes
Red = linear (smooth). Brown = tactile (a bump). Blue = clicky (bump + click + a big reset gap). Buckling spring = the IBM Model M's collapsing column. ↓ Drag the switch down, or use the travel slider.
Depressed
0.00 mm
of 4.0 mm total travel
The key is
not registered
keystrokes fired: 0
Force (representative)
15 cN
registers at 2.0 mm / 45 cN
Slide down past the dashed actuation line and the key fires. Keep going to bottom-out — the vertical wall on the right. On the Brown, notice the bump peak sits above and to the left of where it fires.
The datasheet doesn't publish the whole curve — only a handful of points: the initial force, the actuation force, the tactile pressure-point force, and where the travel starts and ends. The page draws a representative curve through those published points (the marked dots are Cherry's; the line between them is a model). For the linear switch, though, the curve is the real thing — a straight Hooke's-law line, and its spring rate falls straight out of two datasheet numbers:
That's the whole point of the word linear: no bump, no click, just a spring obeying Hooke. A tactile switch adds a hump before actuation; a clicky switch adds a hump and a click jacket that opens a wide reset gap; a buckling spring replaces the whole idea with a column that collapses. Watch what the curve does at the moment the key fires — and note that on the Brown, the bump and the firing point are two different places.
The check — every number is Cherry's, recomputed in front of you
The marked points on each curve are datasheet figures. The green column is the cross-check against Cherry's published sheet — not a claim the page is making, a claim the page is testing.
| switch | family | actuation | tactile peak | pretravel | travel | datasheet |
|---|
The correction, stated as arithmetic — for the MX Brown:
And the clicky reset, the thing that traps a second press — for the MX Blue:
Run it yourself: node research/what-your-fingers-feel-in-a-keyboard/verify.mjs — it recomputes every figure and reads this page back to confirm the drawing and the numbers agree.
The second thing: the reset that traps you
Switch to the MX Blue and try this: press down until it fires, then release only part of the way — to about 1.9 mm depressed, still below the 1.7 mm reset line — and press again. Nothing. No second keystroke. Cherry's clicky slider is a two-piece mechanism: its reset point sits 0.5 mm above the actuation point (which on the Blue is at 2.2 mm, so the reset line is at 1.7 mm), and the key cannot fire again until the stem rises past that reset line. Trap it in the 0.5 mm band between the two and it stays stuck registered — one press, no matter how many times you bounce. Cherry sold this on purpose and called it movement differential. On the Red and Brown that gap is tiny; on the Blue it's a real, operable trap. (Try it: fire it, half-release, press again — the counter won't move.)
The third thing: the collapse is a named instability
Switch to the buckling spring. Its curve does something none of the Cherrys do — it rises to a peak, then drops, a stretch of negative slope. That collapse is the click, and it's a piece of structural mechanics you already know by another name: a column buckling past its critical load — Euler's P_cr = π²EI/(KL)², the same snap-through that makes a loaded strut give way all at once. In the IBM Model M the coil spring compresses until it buckles sideways, throwing a hammer onto the membrane; click and actuation happen together, at the collapse. The everyday feel of the loudest keyboard ever made is a textbook instability. (We identify the mechanism, not the exact critical load — the spring's stiffness isn't published; see below.)
What's exactly true, what's a model, and what the record won't settle
Exactly true (Cherry's own datasheet). The MX Brown lists two forces: actuation 45 cN and pressure-point (tactile peak) 55 cN. They are different line items and the peak is the larger. Pre-travel is 2.0 mm of 4.0 mm total — so the key registers at half its travel, well before bottom-out. On the Brown the bump peak occurs before the actuation point — and that ordering falls straight out of Cherry's own numbers: the pressure-point force (55 cN) is heavier than the actuation force (45 cN), so the curve must rise to the peak and then relax down to the lighter actuation, which places the peak earlier in the travel. Measured force curves (ThereminGoat, via an Imada gauge) show that general bump-then-relax shape, though the exact peak position varies from switch to switch. The MX Blue's reset point sits above its actuation point by a measured ~0.5 mm (Cherry's "movement differential"). A buckling spring's force curve has a genuine negative-slope collapse.
Per-type, so we don't over-dramatize. The decoupling of "what you feel" from "what fires" is a tactile-switch (Brown) property. On a clicky (Blue) and a buckling spring the click/collapse is designed to coincide with actuation — feel and fire land together. So "the bump isn't where it fires" is true of the Brown specifically, not of every switch.
Two different clicks. A Model M's click is the spring buckling + hammer; an MX Blue's click is the click jacket in a two-piece slider. Different physics — the page keeps them apart.
A model, clearly. Cherry publishes points, not curves. The drawn curve is a representative interpolation through the published points; the marked dots are the datasheet's, the line between them is schematic, and the "force" readout is that interpolation (labelled representative). The one genuine equation is the linear curve: F = F₀ + k·x, Hooke's law, with k derived live from the datasheet actuation point.
Named free choices. The pre-load F₀ = 15 cN is a representative value (real MX pre-loads sit ~15–20 cN); Cherry doesn't publish a single canonical Red pre-load or bottom-out, so the 75 cN bottom-out is a Hooke extrapolation from that chosen F₀, not a datasheet number. The bump-peak displacement (~1.5 mm) is a representative value from measured curves, not a Cherry figure — the datasheet-exact facts are the two forces (45, 55) and the travel points (2.0, 4.0).
Representative, not universal. Cherry lists wide tolerances — Brown actuation 45 ± 20 cN, peak 55 ± 25 cN — and measured curves scatter switch-to-switch and batch-to-batch. Cherry retooled the line (MX2A / "hyperglide"), so a single figure is a nominal centre, not a constant; clones (Gateron, Kailh) reproduce the curve approximately. And "all MX actuate at 2 mm" is false: Speed / Silver switches actuate at ~1.2 mm.
The buckling limit. We name the mechanism — Euler buckling / snap-through — but do not recompute the IBM spring's critical load: its bending stiffness EI and effective length aren't published, so P_cr is the governing form, not a derived number. The ~65 g near ~2.5 mm figures are community/Deskthority values with the usual scatter.
Where the popular story runs ahead. Two things guides assert that the record won't back. (a) The datasheet peak is not the actuation force — this page exists to correct exactly that. (b) "Actuation-before-bottom-out makes you type faster" is contested: controlled typing-speed studies don't cleanly support it. What is better-supported is the ergonomic case — not having to slam every key to the floor can reduce per-keystroke effort and fatigue. We lead with the mechanical fact, name the fatigue case as the reasonable one, and mark the speed claim unproven.