Best Clicky Switches You've Tried

[Muirium]

In what sense do you find them very, very tactile? Just curious.

Oh, it'd be so fun for me to post, "+1 what zrrion said."

Anyhow, TL;dr of the below:
* Bjerrk does have a point.
* I rather like Space Invader clicky switches, too.

---

Bjerrk, you do have a point. It would be great if I could break down why I choose switch X over switch Y and zirron's post shows how difficult that can be. I'm not sure that just a force graph is sufficient to answer the question. I think I'd have to sit down and have the tactile and clicky keyboards I have spread out in front of me to be able to tell you what I like about them.

The reason I don't like the Model F XT much is because it feels like hitting too many keys is going to make the keyboard either explode or implode. Brown Alps are about the same for me, but they do feel more under control.

I mentioned, above, that I have a 45g Topre Realforce and an 45g capacitive dome Abko. Why do I prefer the Abko? I think the Realforce is more like bubble wrap and the Abko is feather pillow-y. It just took me a long while to decide that the Abko was better, even though both are outstanding. The reason why I mentioned that I'd like to try an HHKB, which is another Topre keyboard, is because a few people (Mu) have been telling me that the plastic mounting plate in the HHKB further softens the "ride characteristics" of the Topre switch.

---

I wanted to mention that there has been a couple mentions of the "Space Invader" Hi-Tek 725 clicky switches. They do have a delightfully light click and a pleasant weighting. The major drawback of either collecting keyboards or what I do, which is restore keyboards, is that you'll constantly say something like, "Yes, switch A is great, but I like switch B more." Enjoy searching for that white whale.

It’s more the landing characteristics, to borrow your terminology, where HHKB differs from Realforce. The plastic plate damps the landing quite nicely. Realforce feels sharper and harder on bottoming out. Which is why you shouldn’t do it, of course. You will be punished!

Mind, you’re going to bottom out a lot with Topre. The collapsing force curve invites over-travel. It’s really more about how hard you’re still pressing. If you’re heavy handed, a Realforce is going to whack back!

[Muirium]

Zrrion’s chart is magnificent, by the way.




I’d never thought of the integral (the area under) the force curve before. And I’d never really thought of deconstructing the curve into two components: base linear and tactile overlay. That actually makes a world of sense to me. Bet MX brown’s is a whole different story to brown SKCM!

This technique should get all interestingly pear shaped when applied to Topre’s S-shaped force curve. Collapsing domes may well have positive and negative phases of their tactile force component. I suspect that’s also where Niz domes really differ from Topre’s.

[Bjerrk]

If you draw a line from the initial weight to the final weight on a force curve you can get a curve that represents the linear version of the switch. the area under this linear curve (marked in blue) represents the weight of this hypothetical switch.

A tactile switch will have a significant amount of it's own curve above/below this line. In the case of a heavily tactile switch the area of the curve above/below the linear curve (represented in red) is going to be significant when compared to the area under the linear curve since the tactility is a large amount of the total force of the tactile switch. For brown alps you can see that of the total area (red and blue combined) the blue area only makes up 2/5 while the red area is 3/5, which is significantly more! So while the tactility isn't shape, it is clearly highly significant.

Substantial, yes, but perhaps not significant. It's a good idea, but the problem with that metric is that it is not very robust. You placed the initial point at approx. 20gf, but the initial rise from 0 to 45gf all happens within 0.1mm. It would be equally reasonable go places the initial point at 45g, yielding a drastically different ratio of integrals.

But I do agree that there are different dimensions of tactility - magnitude and sharpness for instance, as mentioned earlier. And I do see that SKCM Browns have a decent tactility magnitude, but I don't generally find switches with such low sharpness of tactility very tactile. Perhaps my internal definition of tactility doesn't align with others'. I think mine is something like magnitude*sharpness (note the multiplication: the product is zero if either factor is).

but that sounds like a lot of work for a metric that won't help people who can't properly read the graph anyway so presenting that info in a hard to understand way isn't productive. Would still be fin though

Not sure who that would be, though. A force displacement curve isn't exactly rocket science (although it certainly is a part of rocket science )

[Bjerrk]

I’d never thought of the integral (the area under) the force curve before.

Wait, what? The integral of a force displacement curve is the work (energy) performed. That's an important metric - Physics 101

And I’d never really thought of deconstructing the curve into two components: base linear and tactile overlay. That actually makes a world of sense to me.

Yes! Except for the lack of robustness - you can get a whole range of very different ratios depending on very slight changes in the initial state chosen.

[Muirium]

Hey, it’s morning. Still having coffee here.

Aye. I see Zrrion’s chart as illustrative of a compelling idea, not mathematically objective. The linear line must be rigorously defined to have numerical meaning. But for making a visual point, arbitrary ends will do!

Are these force curves averaged over a large sample size of key presses, by the way? Trying to understand their obvious noise. Is this scratchiness or just sample error?

Edit: the legend seems to suggest it’s just 4 presses. I’d be happier with 100, averaged, with error bars.

[zrrion]

You placed the initial point at approx. 20gf, but the initial rise from 0 to 45gf all happens within 0.1mm.

That's an artifact of the machine doing the testing, not really an aspect of the travel. It's the same thing as when the graph goes super high at the end.
Going from 0 to 20 grams so quickly is because the switch doesn't provide any resistance at all if you aren't touching it and the machine generating the graph doesn't start on the switch. The rise at the end represents the bottom out where increases of force don't get any movement because there's nowhere for the slider to go at that point.

This is part of what I mean where I say using this info to make some sort of "tactility unit" wouldn't be helpful. Tactility units would just be a summary of the graph's important info re: tactility but that doesn't do any good when simple stuff on the graphs isn't necessarily straightforward if you don't know how they're measured. EDIT: if someone does make a tactility unit please let me name it

for this specific switch you may notice that the end of the tactile curve is in line with the linear curve I drew on, this is because the tactile elements of an alps switch don't effect the curve after they are cleared. They can't meaningfully resist the downward motion of the slider once the slider has pushed them out of the way since they're no longer under the slider. This means that for alps switches there is a portion of their travel where they behave just like they would as a linear. You can extrapolate from that portion of the measured curve what the curve would look like if you had opened the switch, removed the tactile leaf, and measured that curve. Ideally we would have measurements for a linearized brown alps switch to compare but this works for basic demonstration. If we were using an MX switch where the tactile mechanism is much different you would need to replace the slider and measure again to get a linear curve to compare against. for something like buckling spring or philips torsion spring you have to just pick a starting value that seems reasonable (as in not part of the steep incline at the start where the machine isn't on the switch). Not scientific but mechanisms like that aren't really capable of being linearized so you can't really measure an equivalent linear curve.

[Bjerrk]

for this specific switch you may notice that the end of the tactile curve is in line with the linear curve I drew on, this is because the tactile elements of an alps switch don't effect the curve after they are cleared. They can't meaningfully resist the downward motion of the slider once the slider has pushed them out of the way since they're no longer under the slider. This means that for alps switches there is a portion of their travel where they behave just like they would as a linear. You can extrapolate from that portion of the measured curve what the curve would look like if you had opened the switch, removed the tactile leaf, and measured that curve.

Now that is a solid (and physically meaningful) point!
However I would argue that even the leaf spring has a Hookean component to it which isn't tactility per se, but that's another detail.

[Bjerrk]

My general impression here is that we should have a "Force curves - insights and limitations" workshop thread I'll bring the croissants and coffee.

[zrrion]

yeah, there's a few sorts of switches which are of a mysterious 4th sort. not linear, not tactile, and not clicky (although that's just tactile again in terms of force curves) which would be super hard to pin down. It's kinda weird that there aren't more of that sort of switch around but I guess that may be because the MX design isn't well suited to it. Clickbar switches could potentially be modified so that the torsion spring is the only source of return force (and doesn't click) and that would get you an MX style switch of this category. Stuff like philips torsion spring have a change in resistance over time that's certanly nonlinear, but none of the action of the switch could be said to represent any sort of clearing/change in state that you get in tactiles. It doesn't clear a bump, doesn't buckle, doesn't doesn't engage a 2nd spring or disengage a magnet or anything weird, it's just bending a torsion spring. I have no idea how you would go about attempting to make a linear comparison to these that isn't just picking an arbitrary line and hoping the number of tactile units feels right.

this sort of switch also needs a name but all of the ones I've heard suggested kinda suck.

EDIT:

My general impression here is that we should have a "Force curves - insights and limitations" workshop thread

That's not a bad idea at all. hopefully someone who actually has a machine for measuring force curves would be able to set up such a thing. I can read them well enough but I'm sure there's all sorts of nuance that I am missing since I don't have first hand experience with them. There are likely a lot of little things about the setup that could impact the measured results that I would never think of.

[Bjerrk]

My general impression here is that we should have a "Force curves - insights and limitations" workshop thread

That's not a bad idea at all. hopefully someone who actually has a machine for measuring force curves would be able to set up such a thing. I can read them well enough but I'm sure there's all sorts of nuance that I am missing since I don't have first hand experience with them. There are likely a lot of little things about the setup that could impact the measured results that I would never think of.

The primary limitation, if you ask me, is the lack of velocity-dependence. A force-velocity diagram would give all sorts of very valuable information.
If you measure force-displacement through a series of static or quasi-static measurements, you miss those all-important dissipative effects -- i.e. friction. I'd wager that force-velocity curves would reveal an awful lot about why some switches feel much worse than others, even though their force-displacement diagrams look inconspicuous.


I'll just quote myself from another thread - classy, right?

I'm not sure if there's an ISO standard that applies since they're all pay for access. I did find a number of research papers proposing measurement methods though. There's some pretty good considerations and test setups but none that are particularly easy to do at home. I didn't find anything that specifically looks for off-center presses, etc.

https://deepblue.lib.umich.edu/bitstrea ... sAllowed=y

https://citeseerx.ist.psu.edu/viewdoc/d ... 1&type=pdf

https://www.researchgate.net/publicatio ... s/download

A force measurement setup that can vary stem torque would be ideal.

Cheers! Those are very worthwhile reads - especially the 1999 paper. I'll read through them and perhaps start a "journal club" type topic.

I can already say that I appreciate these paragraphs, though:

Although prior research has investigated key layout (Kroemer, 1972) and keyboard shape
(Marklin et al, 1999; Simoneau, et al, 1999), studies of key tactile feel have been limited to static
properties only. The tactile feel is elusive because our kinesthetic sense is a composite of several mechanical factors,
both static and dynamic, that are not easily isolated nor characterized, and the kinesthetic sense
is a gestalt sensation. Our inability to isolate mechanical properties of keys may explain
why we prefer some keyboards but cannot articulate our reasons well.

This is essentially the point I was making about static force-displacement graphs being inadequate since the kinematic aspects are such an important part of keyfeel:

In part, this sensation or feeling can be quantified by determining the key's mechanical impedance,
i.e., (i) the stiffness, which relates the contact force to the displacement of the key,
(ii) the damping, which relates the contact force to the rate of displacing the key,
and (iii) the mass (inertia), which relates the contact force to the acceleration
of the key. These three properties can be found from graphs of force-displacement,
force-velocity, and force-acceleration characteristics, respectively.

[Muirium]

if someone does make a tactility unit please let me name it

I suggest Tacton.

[Muirium]

The primary limitation, if you ask me, is the lack of velocity-dependence. A force-velocity diagram would give all sorts of very valuable information.
If you measure force-displacement through a series of static or quasi-static measurements, you miss those all-important dissipative effects -- i.e. friction. I'd wager that force-velocity curves would reveal an awful lot about why some switches feel much worse than others, even though their force-displacement diagrams look inconspicuous.

Yes! I find pressing keys sssslllllooooowwwwlllllyyyy truly brings out the very worst of the worst among them. MX reds: don't you dare! You can't unfeel it!

Visualising a 3D plot now of multiple force curves, projected in sequence to form a surface, from high velocity mashing to low velocity teasing…

Spoiler:

And yes, we really need to crop those meaningless and downright misleading ends off these force curves. (Preferably by re-measuring them.) What does it matter when the key's not even being pressed yet?

[Bjerrk]

if someone does make a tactility unit please let me name it

I suggest Tacton.

In the scientific tradition of naming after pioneers in the field (think Kelvin, Newton, Professor Kilogramme etc), we could go for the Harris.

[Muirium]

There was a while, when a boy, that I did presume the brilliant Prof. Doc. Wizard. "Gram" had invented weighing scales *and* telegraphy…

Besides, Harris is already the unit for tweed. Obvs.

[zrrion]

If they aren't derived from Harris I think calling them Dick units would be kinda funny. (Tacton is actually pretty good though)

[Bjerrk]

But a tacton would be a fundamental particle of tactility. See proton, neutron, fracton, phonon, parton, anyon, ...

[Muirium]

"The tactile potential gradient induces tacton emission from the tactile leaf, duly absorbed by the slider and conveyed to the finger with the physiological sensation of umami."

[Bjerrk]

I think that settles it. A unit of tactility is called a Dick and the associated particle is the Tacton.

[Findecanor]

This does not tell me much about how Dicks are supposed to be measured, though...

[Muirium]

They are a derived unit. To be inferred through the process: "that's what she says."

While I don't like "dick" as the tactile unit, I do like the idea of "dicking a switch" to construct its force/velocity chart. You can just picture the increasingly enthusiastic wee fella ratta-tatting at his work!

[Bjerrk]

While I don't like "dick" as the tactile unit

You're just afraid that lil' Topre won't rise to the challenge

[Muirium]

[mmm]

for this specific switch you may notice that the end of the tactile curve is in line with the linear curve I drew on, this is because the tactile elements of an alps switch don't effect the curve after they are cleared. They can't meaningfully resist the downward motion of the slider once the slider has pushed them out of the way since they're no longer under the slider. This means that for alps switches there is a portion of their travel where they behave just like they would as a linear. You can extrapolate from that portion of the measured curve what the curve would look like if you had opened the switch, removed the tactile leaf, and measured that curve.

For the ideal measurements you'd also have to remove the actuation leaf in this case, but regardless I think the figure you made is pretty accurate. However, measuring both with and without the tactile leaf / actuation leaf could make a pretty interesting 3-part figure, showcasing each elements impact on the tactility/force curve.

My general impression here is that we should have a "Force curves - insights and limitations" workshop thread I'll bring the croissants and coffee.

Please do! I would love to see a theoretical thread from you on this.

The primary limitation, if you ask me, is the lack of velocity-dependence. A force-velocity diagram would give all sorts of very valuable information.
If you measure force-displacement through a series of static or quasi-static measurements, you miss those all-important dissipative effects -- i.e. friction. I'd wager that force-velocity curves would reveal an awful lot about why some switches feel much worse than others, even though their force-displacement diagrams look inconspicuous.

Would it make more sense to make a derived friction-displacement-force curve to make it more readable? (If it is even possible to meaningfully derive an approximate friction coefficient for any given displacement given the above data?) But then of course there are buckling springs that would possibly result in negative friction, but that negative friction is time dependent, possibly, which makes me pretty confident that I am not quite sure what I am saying.

[timw4mail]

By switch type (that I've tried):

Buckling spring: IBM capacative (Model F/ New Model F)
ALPS: Matias Clicky
MX-like: Outemu Blue

[Muirium]

Matias? Are you okay? Should we airdrop real Alps to your famine-stricken location?

[thefarside]

Matias? Are you okay? Should we airdrop real Alps to your famine-stricken location?

Yes please . I’ve tried the Matias Clicky on the Tactile Pro and didn’t think they were bad, but didn’t like the board and ultimately returned it. That was before I tried real Alps too.

I want to do a proper comparison so I purchased Matias Clicky and Quiet Click switches and will replace the orange Alps in two M0116s with them to compare the sound and feel to my other Alps boards.

My favorite clicky switches are:

• Model F (original or reproduction)

• Model M (older and broken in)

• Blue or pine white Alps (hard for me to tell the difference)

[hellothere]

Going back, oh, 20 or 30 posts ago, someone was talking about how everyone should know how to read a force graph. I'm not afraid to say that I don't. I thought I understood a little bit, but I guess not. Mind you, I haven't Googled it. I also wouldn't mind an easy-to-understand walk-through.

"The tactile potential gradient induces tacton emission from the tactile leaf, duly absorbed by the slider and conveyed to the finger with the physiological sensation of umami."

Good explanation.

[milneraj]

I'd love to try Buckling Springs or Alps, but they're not easily obtainable where I am. So I'm kind of looking at MX-style switches and would love to have some recommendations For reference, I have Box Pinks which I enjoy as well, but looking for something a little different (emulate a more retro/vintage feel)

There are two that are available where I am at these price points (including shipping):

• Box Navy ($0.4/switch)
  NK_ Sherbet ($0.24/switch)

[JCMax]

I have thought about it and my final verdict is the best overall switch I've tried is definitely the New Model M BS. I am not going to count the switch testers because I know the rest of the board can have a big impact on how they feel. I'll list those out separate (They're basically the ones I want to try in a board).

I still need to learn how to read a tactility chart, but I will sum up the feel in plain English the best I can.

I am typing this on my Hyper X Alloy Origins, and I have to say these switches feel pretty damn good. They are pretty light, the tactility feels good, and they are very reactive with good push back feel. The only thing the BS's don't have on this is the actuation weight. Since they are stiffer and I have to press down a bit harder I don't type as fast and it kind of dampens the reactivity on them.

Of course it's only Cherry MX clones as far as modern switches. Cherry mx blue I rate towards the bottom. They are mushier with scratchy feel and not very reactive, not the best tactility, etc. Clones like Oetemu and Gateron I like. They're OK. I know some people say they can't feel a difference, but I can. It's very slight though, not as much of a difference between cherry and a BS, so I can see why some people would say that.

Best Switches with keyboard:
1) Unicomp New Model M BS
2) HyperX blue
3) Gateron blue
4) Oetemu blue

Key switch testers:
1) Beamspring
2) Model F
3) Model M
4) Kaihl Speed Navy
5) Kaihl Box Navy
6) Kaihl Jade
7) Matais Clicky

The closest I've come to Alps are Matias clicky, but frankly I wasn't that impressed. They're OK switches. But I still want to see what real Alps are like.

[Muirium]

Real Alps vary quite a bit. Much of it is condition (bad ones are so BAAAAAAAD). But there's also a lot going on inside the famous "Alps Vortex" by design. Different colours are truly different beasts, and then there's the distinct generations of design. It's more than a little confusing! But I must say the best thing about great Alps is how… umami they can be. The quiet ones especially, I've quite a thing for them. Well worth seeking out.

Your switch tester list is impeccable. I've never tried a beamspring tester, but as I'm writing on the 3276 just now, I'll never need to! They are just what everyone says about them: egregious, Neanderthal, downright seventies beasts who, despite it all, have the angels' feel! The caps are top notch as well. They're absolutely *different* to Model F, but I rate them much closer together than I do humble Model M. Why? Beamspring's playing a different game: it's got Topre swing. Model F doesn't bother with any of that, but takes its sharp clickiness straight on out to infinity. Plus you don't have to adjust your posture like you're typing with a volume of the Encylopedia Britannica wedged under your keyboard. I quite appreciate that about my compact Kishsaver, which has almost as many keys as this, in a quarter of the volume, albeit the same weight!

The first Topre I tried was a single key tester. Ditto Model M, actually. (I already had a real Model F, mind.) Topre intrigued me, it really did, from just that one disembodied switch. It felt like some out-of-place artifact amongst those Cherry and even Alps switches Mr. Interface sent with it. Fair haunted me so it did! I knew I had to have one. Good luck to you if you find yourself on that beam quest…