There has always been some debate over the various types of springs that are available for Cherry MX switches and how the spring rates relate to switch actuation force. Adding to confusion are the aftermarket Korean springs that are sold based on the spring force at the bottom of the switch movement.
I decided to try to build a rig that could consistently measure the force of these springs for the purposes of comparing them. If we can measure the spring force at the top and bottom of the switch stem movement and assume that the springs are linear (which is mostly the case), we can then plot the spring force and compare different spring designs. A few Cherry switches were sliced open to accurately measure these two positions and to observe the spring behavior inside the switch.
I do not have access to expensive spring force measurement instruments. Instead, I used a heavily modified arbor press as a jig to consistently replicate spring compression at the top and bottom of a Cherry MX switch movement. I then use a digital jeweler's scale (supposedly accurate to 0.05g) to measure the spring force at these two positions. The large spring at the top of the jig is used to take the backlash out of the arbor press's gear mechanism. The block on the front of the press shaft provides the top spring position while a pin at the top provides the bottom position. A removable shim allows the press to be completely lifted off of the spring so that the scale can be calibrated with the spring's own weight before each measurement. A dial indicator is used to re-verify each measurement position.
The biggest problem was measuring the spring force at the bottom of the movement on the lighter springs. It turns out that because they have so many coils, the lighter springs are almost fully compressed when the switch is bottomed-out. Since the coil spacing is not perfectly even, some coils bind-up before others. This has the effect of increasing the spring rate near the bottom. Particularly bad samples were skipped, but this effect is still apparent in the variance of those measurements. To help cope with this, the range of motion was reduced by about 0.2mm for a total of 3.8mm.
The press proved to generate reproducible results, once I got the hang of its little quirks. For a given spring, it reproduced measurements well within 1%. For each spring type, at least 20 samples were measured and the results averaged. Where possible, I mixed springs from different batches.
So far, I have measured 6 spring types:
- Genuine Cherry MX "light" (blue, brown, red)
- Genuine Cherry MX "heavy" (black, green, white)
- Genuine Cherry MX "clear"
- Originative "45g" Korean
- Originative "55g" Korean
- Originative "62g" Korean
Others to follow:
- Genuine Cherry MX "vintage black"
- Genuine Cherry MX "vintage brown"
The weighted averages were used to generate this plot, which compares the various types. The center, vertical line represents approximately where switch actuation would occur.
It's important to remember that these plots show the
bare spring force, not the switch actuation force. The leaf spring that makes up the switch contact interacts with the coil spring and other frictional forces to make up the total switch actuation force. Even linear switches exhibit this interaction; the leaf spring actually works against the coil spring, slightly reducing the return force. The absolute force values from these measurements are mainly useful for comparing each type to one another.
| Cherry MX Light | Cherry MX Heavy | Cherry MX Clear | Originative 45g | Originative 55g | Originative 62g |
Position: | Up | Down | Up | Down | Up | Down | Up | Down | Up | Down | Up | Down |
Average: | 35.2 | 63.4 | 45.7 | 87.8 | 40.6 | 89.9 | 25.8 | 56.3 | 32.2 | 59.2 | 35.5 | 66.7 |
Variance: | 0.99 | 4.59 | 1.03 | 2.46 | 2.22 | 4.14 | 0.49 | 4.08 | 0.37 | 1.12 | 1.05 | 1.12 |
Error: | 5.6% | 6.0% | 3.4% | 3.7% | 5.3% | 5.7% | 6.9% | 6.0% | 4.5% | 5.0% | 6.9% | 3.7% |
k(g/mm): | 7.43 | 11.08 | 12.96 | 8.03 | 7.08 | 8.21 |