Switch lifetime testers?

[obra]

Is there a Geekhack best practice or standard for switch lifetime testers? I was contemplating rigging something up with a solenoid and an Arduino, but if this has already been done, I'd love to learn what works and what doesn't.

[SpAmRaY]

It's been talked about a bunch but I don't think I've ever seen anything actually done.

Very interested in anything you might come up with.

[obra]

It's been talked about a bunch but I don't think I've ever seen anything actually done.

Very interested in anything you might come up with.

Well, the first solenoids show up tonight

[jacobolus]

To get meaningful results you probably need to test a pretty big sample of switches across different batches. I don’t think too many hobbyists can do it effectively, even if they could build a capable tester. Seems like the kind of thing worth talking to MassDrop or MK.com or WASD or somebody about, someone who handles a lot of switches of different types and could afford to build a rig that could test a bunch of switches at a time.

[obra]

To get meaningful results you probably need to test a pretty big sample of switches across different batches. I don�t think too many hobbyists can do it effectively, even if they could build a capable tester. Seems like the kind of thing worth talking to MassDrop or MK.com or WASD or somebody about, someone who handles a lot of switches of different types and could afford to build a rig that could test a bunch of switches at a time.

I'm not looking for scientific results. Building a tester that can exercise switches to a specific # of presses _will_ tell me a lot about the build quality of a switch and how the spring degrades with a relatively small sample size.

The secret (and business-time justifiable) plan is actually to build a full test rig for a keyboard. Which is more on the order of a few keystrokes per key.

[Oobly]

Another option is a cam mounted to a motor, but it requires some form of roller bearing mounted to the keycap. I've been considering making something like this to wear in new MX switches. It should be pretty fast (even at just 1000rpm you'll get almost 17 presses per second). You can mount different switches at different heights to get the movement to match the bottom out distance.

The advantage of a solenoid, though, is that you can simulate bottoming out forces and I'm sure that will put a bit more strain on the springs. On the other hand, the cam will simulate slight off-angle presses, so they may give slightly different wear results.

I'd be interested in more than just the spring fatigue, but also how the tactility and smoothness change with wear.

If you do complete this project, PLEASE post your testing methods and results here. I suspect it would be very interesting and useful information for most of the users of this site.

[vvp]

Hmm, it looks like 2 ms should be enough to return linear 65g switch from fully pressed position to force free position. I assumed stem mass of about 0.26g (?). So one probably can go to 100 presses per second. One press takes less than 2.6 mJ of energy without friction. So friction loss should be smaller than that. It should not melt when driven by solenoid. But if cam is used then we get additional friction loses at the touch point between cam and switch stem and that may be a problem.
At 100 presses per second, it should take about 2.3 days till the switch fails. (IRC the switches should last 2e7 presses.)
I'm curious to see the results of this experiment (if there will be any).

[0100010]

There was a thread a couple months ago that has a video from some tech / geek site that had some guys visiting the Cherry factory in Europe.  Part of that video featured the testing of MX switches on I believe a pnuematic solenoid setup - to some tens of millions number.

[obra]

There was a thread a couple months ago that has a video from some tech / geek site that had some guys visiting the Cherry factory in Europe.  Part of that video featured the testing of MX switches on I believe a pnuematic solenoid setup - to some tens of millions number.

Yep, I've seen similar things. One of them at a small keyboard maker in Shenzhen. (They were testing their Gaote switches)

[CPTBadAss]

I'm pretty sure I recall a thread where QWER had a machine which tested it. I'm pretty sure it was a cam setup like Oobly mentioned and had little arms which actuated keys on a keyboard. I can't find the thread right now though .

[mkawa]

mount a piston type solenoid above the switch driven by a relay or FET that smashes the key down every time the key registers an actuation (since it's an SPST, this is pretty straightforward). start the little perpetual motion machine and wait until it stops. time a couple cycles to determine number of live cycles.

note that this is a MTBF measurement device and not a torture test total failure deal.

[rowdy]

Is this just testing a raw switch?

Or is there any consideration given to a keycap that would suddenly find itself squashed between a suffering switch and a relentlessly pressing artificial finger?

[jacobolus]

mount a piston type solenoid above the switch driven by a relay or FET that smashes the key down every time the key registers an actuation (since it's an SPST, this is pretty straightforward). start the little perpetual motion machine and wait until it stops. time a couple cycles to determine number of live cycles.

This isn’t actually that interesting to me.

What I’d much rather see is a force curve for the switch’s stroke measured every 100k presses up through 10-50 million. It would be very illuminating to see some kind of precise demonstration of how a switch feel changes as it wears with heavy use.

Also maybe some kind of oscilloscope measurement showing how chattery a switch is when pressed and released at each of those milestones. I’m curious whether heavy use ends up affecting the switch contacts differently in various types of switches.

[mkawa]

i think switch actuation force has a much higher variance than you think it does. combine that with measurement error on an economical small force strain gauge and hysteresis and actuation force measurements over time aren't going to show clear trends. i think the best reasonable goal is to measure variance over time. variance will decrease over time as the switch wears, as will hysteresis. switch chatteryness is easy. take a metric for pulse length (switches chatter from brand new, so switch on time is always going to have a duration), which you'll need to debounce the switch for actuation anyway, and again keep track of variance.

[mkawa]

http://web.mit.edu/imoyer/www/portfolio/applegantry/index.html

[jacobolus]

i think switch actuation force has a much higher variance than you think it does. combine that with measurement error on an economical small force strain gauge and hysteresis and actuation force measurements over time aren't going to show clear trends.

I think you’re mostly misunderstanding me. I want to show the plots of force vs. press distance at various amounts of use. For example, on Alps switches the tactile leaves should start deforming a little bit after a few million presses, so the tactile clicks aren’t going to be as crisp, and on MX switches, the little plastic bumps will start wearing down.

i think the best reasonable goal is to measure variance over time. variance will decrease over time as the switch wears, as will hysteresis.

What do you mean by variance in this context? You think the switches become more uniform after heavy use?

Why do you think hysteresis will decrease? How do you quantify hysteresis? (And which type of switch are we talking about?)

switch chatteryness is easy. take a metric for pulse length (switches chatter from brand new, so switch on time is always going to have a duration), which you'll need to debounce the switch for actuation anyway, and again keep track of variance.

I think it’s much more useful to just show a picture (or a few) of an oscilloscope plot when the switch is pressed and released.