deskthority

Just put this draft up: http://deskthority.net/wiki/Force

Am I right in thinking that peak force is the force required to fully depress the key?

Does anyone have a good definition for preload? This is one for bhtooefr I think.

Obviously the page needs a lot of work, diagrams etc.

In the end it may be possible to split up http://deskthority.net/wiki/Switch_terminology completely, as that page just seems to be a motley mixture of some terms that's a bit hard to link to. Certainly what I've written about force couldn't possible fit in there, and there's more on that page that could be expanded out into separate pages.

The dithering is just some bullshit/troll ripster made up when confronted with the inherent problems of his method for which he had no serious answer. (The lack of significant figures with this kind of precision means you could measure three forces, 49, 51 and 59 as 45, 55 and 55.)

There's actually several forces involved.

Peak force may actually be a bit of a misnomer, as the true force peak is the force immediately before the keycap, switch, PCB, or casing breaks, when continuing to increase force after bottoming out. Except on the cheapest, crappiest keyboards, if it's a direct press, this is probably in the hundreds or even (in the case of buckling spring, especially Model Fs) thousands of newtons.

However, I use "peak force" as the peak force encountered in the normal travel (that is, right up until bottoming out) of the switch, for linear switches. Tactile switches, it gets into further weirdness (which makes me really think I should switch to "tactile force" - and note that this can be different from actuation force, especially on switches such as Cherry MX), but essentially, the peak force encountered up to the tactile event.

As far as preload, this is easier - the initial force that the spring is under. For instance, a keyboard with 20 cN preload requires 20 cN of force to begin moving the key.

Edit: Added tactile force, and some explanations.

Ah, so I did completely misunderstand. I took high load/preload to mean the characteristic of Alps and domes where you have to increase from 0 to actuation force within perceptually 0 mm. (That's one reason for having this wiki -- a go-to place for terms that are hard to derive from context, especially if people (including me) then abuse them by using them for incorrect meanings :)

Glad I left that as "..." and generally left the page as a draft :-P

I might leave the rest of this page to you as I don't really know what I'm talking about. I only created it to have somewhere to document gf and cN, and then realised that the whole topic of force wasn't actually covered properly.

Man, by next year's wiki award there have got to be more than three candidates!!!

For what it's worth, the MINIMUM preload of a switch is that of the weight of everything supported by the springs (so, the stem and keycap, usually). That's right, if a switch has no inherent preload on the spring (which is usually accomplished by compressing the spring slightly on installation), adding a keycap will actually lower the position that the switch returns to ever so slightly.

On a force graph, this is usually the force at zero travel. (Some switches will show zero travel force with no or extremely low preload, a quick jump as it encounters a spring, and then a linear increase. Usually in that case, preload is at the quick jump.)

Also, Cherry prefers the terms "pressure point" and "operating point" to refer to the positions at which peak tactile force and actuation (or make) occur respectively. (Also, "reset point" for the break point.)

Yeah definitely leaving this one to you :P

I am only the #32767 Keyboard Expert on the Planet.

webwit wrote:The dithering is just some bullshit/troll ripster made up when confronted with the inherent problems of his method for which he had no serious answer. (The lack of significant figures with this kind of precision means you could measure three forces, 49, 51 and 59 as 45, 55 and 55.)

It's not complete bullshit - it's just the Ripster effect where everything said in Ripster's spam becomes unbelievable.

I'm actually the one who proposed him to use "dithering" in the mechanical sense - adding very small vibrations to your test setup to overcome static friction inherent when weight-stacking and increase accuracy (you face it even when doing it webwit style, no coins). The greatest downside of this technique lies in the reproducibility: you need vibrations which are large enough to overcome friction and yet small enough that they only add random noise to the measurement (which can then be filtered out given enough measurements). But then it's best to do it using some motorized device, as you need to have the same magnitude of vibrations for each and every test.

In the larger scheme of things, I do know that a force gauge pushed by a press is a far better method seeing how you don't have any static friction in the first place. But that doesn't mean that weight-stacking is completely invalid, neither that "dithering" doesn't help. Weight-stacking can give approximate results, one only needs to do enough measurements and use weights which are granular enough. Dithering is then only an attempt to shave those few extra percent off the error margin, especially for buckling springs.

Dithering is a valid technique to increase measurement accuracy but it's tricky to get it right. The basic idea if I am not mistaken is to add random noise to a measurement (add a random value of known distribution to be precise) and determine the actual value you want to measure from the distribution of the results.

Still claim it's bullshit unless proven otherwise. No prove this is actual a valid technique which can be applied for this purpose. Show me the math and logic and not just the buzzword. Are you dithering, or simply adding extra force, and you're dithering an inaccuracy? And then if it would logically work out, it would only make sense when applied in a professional measurement setting. When applied to something like an amateur setting involving stacking coins and tapping, it only introduces an extra uncertainty factor. May make more accurate, may make less accurate. Less significant figures.

@webwit: feel free to mercilessly redo that section if you disagree with it. I just put up a few bits and bobs on that page to get the ball rolling, as switch forces is not something I know enough about to document properly.

In other news, has the definition of preload changed? The definition added to the page doesn't seem to match the one above, and in terms of Alps vs Cherry, the preload by the page's definition is still ~0, but the force increase within 0.5 mm or 1 mm is very different. It's probably something where we'd need to invent a term and a definition unless the industry has one, and I imagine that it does not.

Also, I was looking at Cherry's force diagrams earlier — didn't realise for some reason that the pressure point and operating point don't correspond!

I was thinking about redoing a few force graphs, copyright-free and consistent in style, SVG. Probably linear (MX black), tactile (ML) and clicky (BS). (Except, ugh, the text would all screw up, as the fonts on the server are still not fixed. Argh. I've just been posting diagrams with sliced off text as I just gave up waiting.)

Surely using manufacturer's force graphs as part of describing the switch is fair use, no?

The terminology needs to be consistent - it's not right to talk about pressing a key when describing a switch, since as you say, the keycap may vary. (All the keycap does is move the whole graph down by a force corresponding to its weight).

Preload is >0 for (nearly?) all keyboard switches. For Cherries, the spring is compressed by roughly 2 to 4 mm depending on the switch type, so the preload is roughly between 1/3 and 1/2 of the activation force.

The other terms are hardly ever used in a consistent way... I'm not sure it's even possible to give firm definitions for them! Cherry's 'operating force' tends to mean 'peak force before activation', and obviously that's useful for comparing linear and tactile switches. (Activation force is fairly irrelevant, or even unmeasurable alone, for tactile switches). So 'Peak force' is ambiguous, since often it's used in an equivalent way to 'operating force'.

Not an easy task!

Keep in mind that a force graph may show the resting state of the switch with no keycap (or, even, in the case of a Selectric, the keylever forced UP), but a keycap increases the effective preload if there's insufficient natural preload in the switch. Zero preload can't ACTUALLY happen unless the keyboard is in a zero gravity environment.

According to Cherry's force graph, blacks and browns have somewhere around 36-38 cN preload, blues have around 30 cN preload.

Also, to get the terminology of points correct:

Cherry terminology:
Pressure point: Point of peak tactility
Operating point: Point at which the switch is "made"
Reset point: Point at which the switch is "broken"

IBM terminology:
Snap over: Point of peak tactility
Make: Point at which the switch is made
Break: Point at which the switch is broken

ISO terminology:
Initial resistance: Also referred to as "starting force" or preload (must be between 25 and 75% of character generation or snap point)
Snap point: Point of peak tactility (should be between 50 and 80 cN, must be between 25 and 150 cN)
Character generation point: Point at which a linear ("ramp action") switch is made (should be between 50 and 80 cN, must be between 25 and 150 cN)
Switch make point: Point at which a tactile ("snap action") switch is made (should be at minimum force point after snap point, must be after the snap point, before key force returns to snap point force)

Whoah... 'points' are in mm, not cN

A keycap decreases the effective preload force, since it acts in an opposite direction to the spring force. Well, I see preload as being the initial spring force, anyway. So the force required to start moving the switch is equal to the preload, but has a different name

Soarer wrote:Surely using manufacturer's force graphs as part of describing the switch is fair use, no?

Probably. Even so, they're not self-consistent in style.

Blah. I have one force graph drawn, but owing to my luck, my new PC has gone one week before Windows has screwed over its own boot configuration data file. I'm sure I'll be able to post the graphs one day.

I'm (slowly), writing a javascript app to draw force curves conveniently (and export as data/images). It should be ready in the next month or so (bug me otherwise).

My force graphs are all drawn and up. I did them on my laptop instead — yay for a 4-button trackpad.

http://deskthority.net/wiki/Force#Force_graphs

Careful with the membrane BS graph.

IBM merely metricated the capacitive BS one, and we know that in the real world, membrane BS has a much shallower slope, with higher preload.

It's just this graph: http://deskthority.net/wiki/File:Bsgraph.jpg from http://deskthority.net/wiki/Switch_terminology

Is that capacitive then? If so I'll move and re-caption it.

That's the capacitive one.

Although, actually, the membrane one is a re-draw of the same graph (and I thought it was fully metricated, it isn't) - they redrew the behavior immediately before the switch is broken. (Figure 9 of US4528431.)

Fair dos, I'll relabel it and move it to the correct name. Not sure the caption will fit even before the font mangleriser.