The Great Switch Debate

[Hypersphere]

@PollandAkuma: I've tried every Cherry mx switch there is, and I have not liked any of them. Cherry mx switches are inherently linear, and if I were to go with Cherry, I would be inclined to choose one of their purely linear varieties (red or black) unfettered with a grafted bump on the stem or a bump plus an add-on clicker. Unfortunately, the contemporary Cherry mx red or black switches feel scratchy to me. Moreover, reds are too light and blacks are too heavy.

I had heard good things about Gateron, and some time ago I bought a custom 60% with a glass fiber case, HHKB layout, and Gateron yellow switches. I like the keyboard and the switches. Gateron yellows are very smooth and they are an ideal weight in between that of Cherry red and black.

I also like Matias Click switches, but white (pine) or blue Alps are much better -- more refined and elegant, and certainly quieter.

[Mattr567]

While I am a Alps person myself, I find these (vintage 1997?) MX Browns from my Compaq MX G80-11800 to be rather nice. Smooth and pretty refined with a little tactility thrown in. They don't feel like the bumpy polarizing switches they normally are. Also I would find Brown too light but the thick Cherry profile caps make up for it. My old 11800 with the thin caps didn't feel nearly as nice.

I wouldn't buy any new MX board though.

For anyone who has had a bad experience with dirty Alps, they can usually be restored with a dusting and if not an ultrasonic of the switch top will do the trick. Only in extreme cases do they not come back all the way, like the SKCM Blue that is my profile pic Good 'ol Taobao...

[Daniel Beardsmore]

Cherry mx switches are inherently linear …

People say this, but what do they mean?

What precisely is and is not allowed in a tactile switch?

There's a lot of variation in force curves. For example, Marquardt designed a switch where the force drops then hits a wall instead of rising and falling:

https://plot.ly/~haata/237

Or it can start out being forceless, before the wall, as with Tokai:

https://plot.ly/~haata/tokai

I would say that the issue with Cherry—if there even is one—is more that the magnitude of the tactile peak is comparatively low. Nobody calls Alps switches "inherently linear", but they are: Alps switches are linear, with a piece of metal stuck in to obstruct the switch. Not many switches are truly tactile, and those that are, are often less tactile than inherently linear ones. Mitsumi tactile switches for example, typically use a design with a horizontal spring that serves as both the return spring and the source of tactility, but these are widely panned for not being tactile! (Yet, if you press a loose switch, it's extremely tactile!) Membrane buckling spring feels pretty linear to me, and it is: most of the force curve is a straight line, with a drop near the end.

It's a much more complex than the Good Guys versus Bad Guys approach that people like to take against Cherry, which is why I've been contemplating the idea of an ASDR-like analysis of force curves to classify switches based on their fundamental principles.

Cherry switches are just more gentle, and honestly that's not a bad thing, as increased tactility is like your fingers engaging in a constant argument against keys that are trying their hardest not to be pressed. MX Blue offers just enough feedback, although it does seem that they take some time to break in and soften — I found them very harsh and jagged feeling for a long time. MX Brown is more targeted to the subconscious — I find that no matter how tactile the switch, nothing registers in my head from it (that is, I still don't know if a key registered from touch alone).

[fohat]

Unfortunately, Alps-based designs age terribly.

This keeps getting repeated. Age is not the enemy of Alps, dust is.

I have an AT101W that I got new with a Dell system in 1997-98 that still works very well. I used it for several years, enough to get the keys quite shiny, but it has been stored in a clean closet since.

On the other hand, I bought a near-identical keyboard "almost new still in box" just a couple of years ago that is not nearly as good as my well-worn one.

[Chyros]

Cherry mx switches are inherently linear …

People say this, but what do they mean?

The way I myself have always regarded it is that the linearity of the switch is maintained even if a tactile bump is introduced on Cherry MX. If you look at the force curve of MX blues - which have one of the more defined tactility types in the MX series - you can see that the slope before the bump matches up almost perfectly with that of the slope after the bump - they're basically exactly in-line. The tactile bump provides a small hiccup, but the travel is not quintessentially changed. Contrast this to buckling springs, and it's obvious that that with that, the slopes are not in line; the keyfeel undergoes a genuine change. With Alps it's even more pronounced, when the slopes have nowhere near the same gradient for tactile switches, and even the linear switches have more offset than tactile Cherry MX do.







Someone recently actually drew this out with lines on the force curve, but I forgot where it was.

[DeChief]

Now this is more like it. I was expecting this thread to complete go under, but I'm glad to see some healthy discussion going on.

The superiority of the MX switch design doesn't mean they offer the most refines typing experience in all of their different variants.

But I have yet to see a modular buckling spring switch ... or a PCB-mountable Alps switch ...

I, too, have yet to see a modular buckling spring switch or mountable alps. Is there a reason for either of those not existing? Surely they would both blow Cherry MX out of the water, this thread wouldn't be necessary at all.

I can't speak for the longevity of Cherry MX switches due to not owning a "modern" mechanical keyboard (I've only got an IBM 5576-001, 5576-002, and a TeraDrive Model M, all seem to have been very lightly used), but I can say that when I tried a friend's Cherry MX blues they didn't feel very good at all. The sound they made was quite unpleasant too, rather dull-ish, and as far as I know he's only had his keyboard for 3 - 4 years.

It would help to know your target market. Is it gamers? Typists? Collectors? Vintage fetishists? You won't be able to catch them all with a single net (i.e., switch type/brand) so you'll have to make some sacrifices.

Definitely vintage collectors, and to some extent, fetishists too; the project revolves around a brand/line of vintage computers after all. The type of switch I use ultimately doesn't matter a whole lot, just as long as it's better than these horrendous stock keyboards (which isn't difficult, any rubber dome keyboard from the past 25 years or so would feel miles better). It's just a matter of choosing quality sounding and feeling switches that will justify the product's existence.

I checked out some videos on Matias switches and they seem to be a much better alternative to Cherry MX. The only problem is that I can't find any Matias clicky switches... Their website says that they'll be out of stock until May, and I'm far too impatient to wait that long. Any ideas?

[Wodan]

Unfortunately, Alps-based designs age terribly.

This keeps getting repeated. Age is not the enemy of Alps, dust is.

Indeed, that's what I've been meaning to say. With old Alps keyboards, there is so much more fuzz about the condition than with old MX keyboards. Some people even like their MX switches well "broken in"

So when I say they "age terribly" I mean they "age terribly with use"

And the clickyness of well used Alps switches is also appaling. Most of the Alps boards I find have seen some good use and I can usually not even tell if they are a clicky version or not. Funny enough, they often gain some clickyness back after resting in my basement for a few weeks, probably loosing some humidity.

[DeChief]

Since the Matias clicky switches are out of stock on their website and I wasn't able to find them elsewhere, I've gone with some Cherry MX blues for now. I'm pretty sure they're clones, but I'll have to wait and see.

[emdude]

I, too, have yet to see a modular buckling spring switch or mountable alps. Is there a reason for either of those not existing? Surely they would both blow Cherry MX out of the water, this thread wouldn't be necessary at all.

There are PCB-mount Alps switches, specifically the SKFL and SKFR/SKFS (may be others) series of switches. However, these were typically found only in portable computer keyboards. I haven't tried them, but they apparently feel different from typical SKCM/SKCL.

There was a fellow who made a "prototype" discrete buckling spring switch a couple of months ago, but I haven't heard anything about it since its original post and the following (failed) kickstarter for further development.

[PollandAkuma]

This is not helping, I thought I settled my inner clicky debate to only get either blue alps or model F, now I'm tempted to give mx blues a chance ;-;

[Lynx_Carpathica]

This is not helping, I thought I settled my inner clicky debate to only get either blue alps or model F, now I'm tempted to give mx blues a chance ;-;

If evrything is true, I'll get a board with MX Blues tomorrow
I'll tell you my thoughts about it.

And yes: Well broken-in MX Blacks are smoother (or as smooth as vintage mx blacks).

[Sigmoid]

I, too, have yet to see a modular buckling spring switch or mountable alps. Is there a reason for either of those not existing? Surely they would both blow Cherry MX out of the water, this thread wouldn't be necessary at all.

As for pcb mountable alps, it's just that nobody bothered to make one. (edit: and anyway, most enthusiast keyboards are plate mounted, so this doesn't disqualify them) The mechanical keyswitch market is pretty much legacy-driven, most of the available swithces are decades-old designs, probably made on decades-old molds...

Discrete buckling spring, well it would take some impressive mechanical engineering to make it practical and feasible - you'd have to shrink the spring, probably redesign the barrel... The reason Model Fs are not discrete is that this was the most elegant solution - capsense switches need a very simple pcb base, it's totally different from a leaf switch that actually has contacts and all...

Definitely vintage collectors, and to some extent, fetishists too; the project revolves around a brand/line of vintage computers after all. The type of switch I use ultimately doesn't matter a whole lot, just as long as it's better than these horrendous stock keyboards (which isn't difficult, any rubber dome keyboard from the past 25 years or so would feel miles better). It's just a matter of choosing quality sounding and feeling switches that will justify the product's existence.

Would you share some details? Sounds like a cool project.

Since the Matias clicky switches are out of stock on their website and I wasn't able to find them elsewhere, I've gone with some Cherry MX blues for now. I'm pretty sure they're clones, but I'll have to wait and see.

"Pretty sure they are clones"? Where on Earth are you buying from? Aliexpress?

Fyi: http://www.digikey.com/products/en/swit ... herry%20mx

And yes: Well broken-in MX Blacks are smoother (or as smooth as vintage mx blacks).

Aren't vintage keyboards "well broken-in" by default?

[Daniel Beardsmore]

I, too, have yet to see a modular buckling spring switch or mountable alps. Is there a reason for either of those not existing? Surely they would both blow Cherry MX out of the water, this thread wouldn't be necessary at all.

PCB-mountable Alps (which is what I assume you mean) — why would anyone have wanted that? Oddly there was another way to do it: Datanetics DC-50 series switches are secured by screwing them to the PCB! No need for a mounting plate, and you still get to have discrete switches (unlike Hi-Tek, Stackpole and MEI designs that resolve the problem by forming all the switches from a single moulding).

I can't offer you specific answers, as nobody knows any more, but anyone who really did want a smaller, lighter keyboard tended to use Alps's smaller switches. If anyone demanded Alps make PCB mounted SKCL/SKCM switches, they could have done so, as there are clones that are exactly that (Hua-Jie AKF and Xiang Min KSBF).

Pure modular conductive buckling spring is possible — photos of the prototypes are on Reddit somewhere (basically, instead of a membrane and hammer, there's a conductive strip inside the shell that the spring makes contact with, and the spring forms part of the circuit). The aspect that reduces its desirability is that it can't take Cherry keycaps without being extra tall. IBM keycaps have a gap in the keycap mount to make room for the spring, which allows for reduced height, but you can't do this with a Cherry style mount. With IBM-style keycaps it would be fine.

The way I myself have always regarded it is that the linearity of the switch is maintained even if a tactile bump is introduced on Cherry MX. If you look at the force curve of MX blues - which have one of the more defined tactility types in the MX series - you can see that the slope before the bump matches up almost perfectly with that of the slope after the bump - they're basically exactly in-line. The tactile bump provides a small hiccup, but the travel is not quintessentially changed. Contrast this to buckling springs, and it's obvious that that with that, the slopes are not in line; the keyfeel undergoes a genuine change. With Alps it's even more pronounced, when the slopes have nowhere near the same gradient for tactile switches, and even the linear switches have more offset than tactile Cherry MX do.

…

Someone recently actually drew this out with lines on the force curve, but I forgot where it was.

You're probably thinking of my graph over at Geekhack a few days ago. That's where we established that the published MX Clear graph is completely wrong. That published buckling spring graph is also wrong. Buckling spring is linear path divert:

https://plot.ly/~haata/200

It really is just two linear paths joined by a step. It's a non-linear design with (at least, with membrane buckling spring) nearly linear feel — the tactility in membrane buckling spring is quite minimal. It's the exact same category of switch ("path divert") as Futaba clicky which is also widely held to be of very low tactility.

Someone recently made the point about impulse — that's what matters: the "catapult" effect of your finger being fired downwards when the force drops, and it's not about the shape of the whole curve so much as the magnitude and suddenness of the drop in force.

Alps seems to have a combination of two factors: the tactile peak is longer and taller (more "charging" of your finger, which makes the switch heavier), and it has a period where the force is lower towards the end (it's not a straight drop to the lower path, but the impulse is enough to clear the hurdles in most cases, unlike with SKCM Black and simplified where the hurdles start interfering with the feel). It combines the positive peak of MX Blue and the trough of buckling spring to create a strong drop in force.

At this stage, I don't know that it's fair to say anything for certain until we have a larger assortment of graphs, and we can see what does and what doesn't work. Considering that I already had trouble in the past with MX Blue feeling too sharp, I think it's actually weighted fairly well, but it could do with being a tad stiffer.

[DeChief]

Definitely vintage collectors, and to some extent, fetishists too; the project revolves around a brand/line of vintage computers after all. The type of switch I use ultimately doesn't matter a whole lot, just as long as it's better than these horrendous stock keyboards (which isn't difficult, any rubber dome keyboard from the past 25 years or so would feel miles better). It's just a matter of choosing quality sounding and feeling switches that will justify the product's existence.

Would you share some details? Sounds like a cool project.

Since the Matias clicky switches are out of stock on their website and I wasn't able to find them elsewhere, I've gone with some Cherry MX blues for now. I'm pretty sure they're clones, but I'll have to wait and see.

"Pretty sure they are clones"? Where on Earth are you buying from? Aliexpress?

Fyi: http://www.digikey.com/products/en/swit ... herry%20mx

Shhh, no project details yet, it's all very secretive!

I ordered them from a Chinese eBayer, $42.50 USD for 87 switches including postage to Australia; I doubt they're legit. If they are, then I'll be a very happy man! Most other sellers have $60 - $80 USD for postage alone on their listings...

[Hypersphere]

<snip> ... <snip> ... <snip>

Someone recently made the point about impulse — that's what matters: the "catapult" effect of your finger being fired downwards when the force drops, and it's not about the shape of the whole curve so much as the magnitude and suddenness of the drop in force.

I think I am the one who brought up impulse in connection with pointing out that on HaaTa's plots, "total force" should actually be "work" in units of energy (e.g., kcal or Joules). Impulse can also be regarded as a change in momentum. To visualize it, we would need a plot of force vs. time instead of force vs. distance (displacement), where the area under the force vs. time plot would be the "impulse".

[Daniel Beardsmore]

I think it was you, yes.

This should illustrate what I was thinking:

Graph theory (or someone extending Tube lines while stoned)

Graph theory.png (31.76 KiB) Viewed 6449 times

What would a switch with the pink or purple curves feel like? Purple increases the magnitude of drop (requiring more work) but with a greater wall at the start of the peak, while pink involves more priming but no greater drop. That is, can we prove whether this type of curve is truly non-viable, or whether it's only the force levels that are the problem.

What would be amazing is a machine with a mechanism under each key that modulates the force under computer control, say using electromagnets. You could draw in the curve you want into the computer and the keyboard would simulate it for you. (This could even work as a real keyboard.)

[Hypersphere]

I think it was you, yes.

This should illustrate what I was thinking:

Spoiler:

Graph theory.png

What would a switch with the pink or purple curves feel like? Purple increases the magnitude of drop (requiring more work) but with a greater wall at the start of the peak, while pink involves more priming but no greater drop. That is, can we prove whether this type of curve is truly non-viable, or whether it's only the force levels that are the problem.

What would be amazing is a machine with a mechanism under each key that modulates the force under computer control, say using electromagnets. You could draw in the curve you want into the computer and the keyboard would simulate it for you. (This could even work as a real keyboard.)

What is being plotted? Is it force vs displacement?

[Daniel Beardsmore]

yes, just regular force curves, downstroke only.

[Hypersphere]

When interpreting force vs. displacement curves, I think it is easy to confuse ourselves by sometimes equating displacement with time. This is another reason why it would be great if we could generate true force vs. time curves for keyboard switches.

[Daniel Beardsmore]

This is one of those things where you end up going crazy trying to judge things at and beyond the extremities of the human brain's ability to objectively compare. I've just compared: NTC KB-6251 (white Alps), Model M (1996), Fujitsu FKB-4700 (Peerless) and now a Majestouch (MX blue).

My conclusion would be that MX blue is sufficiently tactile when you compare its low switch weight and overall smoothness — the tactile peak is lower, but you don't have to press the keys as hard, so it balances out nicely.

I would still rate it the highest for quality of feel of those, although it's just a bit light. Subjectively it feels the smoothest, but that in part relates to the weight. Pressing keys on the Model M slowly produces audible scratchiness, although I can't feel it; I can't test this with white Alps as you can barely press the key before it jumps.

I actually really like Peerless switches — they're a nice medium weight and have a quiet, dignified sound. If I had one with Windows keys, it would almost be a contender but for is excessive size. (In the case of mine, I'd have to lubricate the wider keys and reseat the space bar spring so that you can actually press the space bar at all.) The Model M is also very nice. White Alps, my feeling is that the high initial force and very early jump point means that they're awkward: Cherry place the tactile peak some way into the travel, while the Model M is a linear-based design, so both of them allow for a build-up of force, while white Alps requires the keys to be struck hard, and hard-strike keys tend to result in missed keystrokes where you didn't smack them with enough force (a huge problem with my Acer keyboard, and my BTC 5149 with its age-stiffened rubber).

The only thing the MX design can't seem to achieve is a the exact tactility of MX blue without sound.

[Menuhin]

This is not helping, I thought I settled my inner clicky debate to only get either blue alps or model F, now I'm tempted to give mx blues a chance ;-;

How has this debate developed? Modern Cherry MX blues will have a chance..?!
Perhaps Gatistotles, Zealistotles, or Cherristotles, for those who like more defined tactile bump and more pronounced double clicks. Even those hate the Cherry MX blues like to give these Aristotle mods a try.

The above statement still is arguing for the MX switch design - at least that is easy to do mods and swap parts.
In a parallel universe, if complicated Alps design has gained majority of the market, there has to be a standard they stick to before other manufacturers can clone or produce swappable parts. But in our reality, the design of Alps has never stabilized - it keeps on evolving for whatever user experience related or cost related reasons - evolving is perhaps a good thing but non-compatibility and without a standard seems not a smart move.
I like vintage complicated "pine" Alps, but any "bamboo" Alps seem de-appetizing despite I wanted to like them; my memory about Matias switches is blurry, based on my experience with an Apple USB keyboard and anyway not so impressed. Through the course of keyboard history, MX design has now gained major market dominance, but it doesn't mean Cherry is doing a good job producing its modern switches: digging into the forums just a little bit, you will read mostly about vintage MX and mods, e.g. ErgoClears and Zealios, etc.

By the way, McDonald is the greatest restaurant providing the bestest food for human beings in modern history.

[Sigmoid]

Pure modular conductive buckling spring is possible — photos of the prototypes are on Reddit somewhere (basically, instead of a membrane and hammer, there's a conductive strip inside the shell that the spring makes contact with, and the spring forms part of the circuit). The aspect that reduces its desirability is that it can't take Cherry keycaps without being extra tall. IBM keycaps have a gap in the keycap mount to make room for the spring, which allows for reduced height, but you can't do this with a Cherry style mount. With IBM-style keycaps it would be fine.

It's not that it's impossible... but the real trick in the Model F isn't just the fact that it uses capsense, it's the whole assembly: spring, barrel and the large, satisfyingly 'clappy' flipper all together. (BTW Model F's do not have a membrane, they have a PCB.)

Even the almost identical Model M feels significantly inferior to the F's, so I wonder what a discrete switch that did away with the flipper feels like (probably nothing like an F).

There's a workshop thread on an interesting project resurrecting the Model F technology, btw...

[Hypersphere]

I recall the first time I tried a contemporary mechanical keyboard after having used an IBM Model M for several years. The board was a Filco Majestouch Ninja with Cherry mx Blues and the classic beige doubleshot keycap set from Originative. What a letdown! The highly praised Filco felt and sounded like a toy. It was like typing on rice crispies.

However, now, after having experienced IBM Model F switches, the membrane buckling spring switches in my Model M and Model M SSK feel scratchy.

Lately, I have been experimenting with Alps switches. In general, I prefer clicky Alps (white [pine], blue, or click-modded linear or tactile Alps). However, I have found that the typing experience with Alps is greatly influenced by the chassis in which the switches are installed -- the best I've found thus far is the Northgate Omnikey 101.

Overall, the switch I come back to is 45g Topre. I have these in a HHKB Pro2, RF87UB, and Novatouch. My favorite is the HHKB -- it helps me get my work done efficiently and reliably with minimal distraction. Topre seems to make the best connection from my brain to the document I am writing at the time.

I am typing this on an AEK with Orange Alps, so I cannot be held responsible for whatever is getting transmitted to this thread.

[zslane]

I can't think of a time (over the course of 35 years) when I wanted my keyboard to be clicking loudly when I type. Having said that, I really thought I'd find MX blues to be super loud and irritating (after auditioning them in a small switch tester), but it turns out they actually yield a very soft click. In fact, the click is nearly drowned out by bottoming out and upstroke noise. Still, the only reason I would toy with clicky switches is to try and capture some of the essence of typing on an electric typewriter, just for the occasional fun of it.

After having spent a year and a half with MX reds, I find that I am now rather addicted to silenced Topre (uniform 45cN). I love them with SA keycaps installed, and I am hard-pressed to imagine a combination I would like more. I know that SA doesn't work for everyone and that Topre purists hate the added wobble that comes with MX-stemmed keycaps, but I'm a huge fan of the combination.

[Chyros]

This is the scheme I was talking about by the way, Alex drew it up:


It illustrates very well why Cherry MX is considered such an inherently "linear" switch IMO. Whether or not other switches are super tactile is another story of course — and there's sharp, broad, strong and weak tactility — but this kind of absolutely unchanged key travel isn't that common to find in other switch designs AFAIK.

[Hypersphere]

The mixed units in the IBM graph is humorous -- "force" in grams and key travel in inches. Makes it tricky to convert the area under the curve into electron volts ....

[Chyros]

The mixed units in the IBM graph is humorous -- "force" in grams and key travel in inches. Makes it tricky to convert the area under the curve into electron volts ....

Yeah, well, you know, if everyone just used normal units for everything...

[Hypersphere]

The mixed units in the IBM graph is humorous -- "force" in grams and key travel in inches. Makes it tricky to convert the area under the curve into electron volts ....

Yeah, well, you know, if everyone just used normal units for everything...

Yes, then life would be less interesting.

I like to keep in mind the crash of a Mars orbiter back in 1999 that was caused by a miscalculation arising from mismatched units. The Lockheed Martin team was using English units and NASA was using metric units.

Such mashups of units extend to all sorts of realms:

Fig-Newtons2.jpg (57.45 KiB) Viewed 6318 times

[randomist]

Best in my personal opinion? No idea, I'd have to try them all first. So far Cherry MX are on top, with different colours for different purposes. Rubber dome are at the bottom of the pile, even Topre didn't feel that great to me, though I didn't get much of a try on them. From what I remember buckling springs were nice (haven't used any since the 90's). Scissor switches are somewhere just above rubber dome, maybe equal with my current Topre feelings. Still to try the rest.

As for what would be best for a commercial product? They're all in use in commercial products, so the real question is who is your target market, what will they pay and what do they want? Are you trying to carve out a niche and offer something different from the mainstram masses? Or do you want to give the consumer something familiar? (Familiar being Cherry or Kailh in most cases).

EDIT: The NASA/Lockhead story is particularly amusing wnen you know that the issue was pointed out at least twice before the calculations were needed and both NASA and Lockheed ignored it. Well 1 lb is pretty much 1 N, right?

[fohat]

if everyone just used normal units for everything ....