GH DESIGN CHALLENGE: force vs displacement measurement device for keyboards/sws

[mkawa]

frankly, we've been living on witchcraft and dubious looking datasheets for far too long. empiricism is required for us to proceed. this thread is to discuss the bits that are required for such a device, as well as a rough idea of what their tolerances and dfm (design for manufacturing) requirements (3d printable parts, off-the-shelf micros, etc.) should be.

note that i am aware that some have built nice one-offs of such devices before. imsto at one point posted some pretty sweet pictures of one based on a mouse optical sensor and basic jeweler's scale. iirc limmy had an analog version he designed. however, my feeling is that we have the expertise on the forum now to come together and design a kittable device that is reproducible, can be calibrated, and can be manufactured with a minimum of salvaging for a reasonable cost.

but, why? the variety of switch designs, variants, mounting configurations, and so on and so forth has become dizzying. just think, 5 years ago, there wasn't even a forum to discuss keyboards on the english internet. yet we now live in a world where logitech, one of the largest input device vendors in the world has made the mechanical-ness of their flagship keyboard its primary selling point (ironically monopolizing the switch output of an entire company for half a year, BUT ANYWAY). this is fantastic, as its attracted more attention to the hobby than ever before. however, as lampson might say, it's become a bit of a success disaster; we now have more opinions than ever on what keyboards feel like and which ones we like best (or we're on our way to becoming opinionated curmudgeons there). however, we have incredible trouble quantifying the metrics and intervals along those metrics that contribute to our like or dislike of a keyboard.

this is a problem for a few reasons. first, it makes it difficult for us as a community to express to manufacturers what we want more/less of. second, it makes it difficult to direct ourselves in what to push the envelope on. people have recently been going crazy over the korean-manufactured drop-in springs for cherry switches. however, no one can quite express exactly how they're different from the OEM springs in cherry switches, nor in what way they want replacement springs to be different from the springs provided by cherry (is it that we actually want progressive springs? different behaviors over a specific frequency bandwidth? !?!?!? no one knows, frankly). finally, i have noticed many complaints recently about defective switches. how can we determine when a switch is defective if we don't even know what the average operating parameters are supposed to be? we have many many samples of these devices (a hundred per board!) and yet we still don't know what the reasonable manufacturing variation is or should be.

what i suggest let's build open source plans for a device that can used to reliably measure the relevant properties of a keyboard switch. this holds value to us for two reasons as well. not only will we get a reproducible device out of it that can help us quantify our tastes, but it's also a great project for us to work on as a community that is both practically realizable, and yet not just another revision of a keyboard with an incrementally new feature-set.

[WhiteFireDragon]

Good speech. Boy that was a long read lol.

It won't be easy to build something that will measure the force curve of switches. You can probably make something simple that can measure distance pressed vs force to get a lot of data points, then extrapolate to get a nice smooth curve. This sounds like more trouble than it's worth though. It's hard to quantify something that is such a subjective matter, and that's what switch preference really is. Even if you can get all the data, then people would have to know how to interpret what the graphs mean.

[SmallFry]

Nickles?

[bavman]

[SmallFry]

I LOL'd so hard at that!

[asura]

Simplest solution - high resolution digital scale + digital callipers.

Set the switch on the scale and tare.
Rack in the callipers until the scale twitches and tare.
Rack in the callipers using either their reading in mm (yes mm not inches!) or the weight change on the scales in g (yes g not oz) recording both as you go, and you're done.

The only problem with this method is, what's the actuation point?

So let's solder up a couple of wires, a cell and a L.E.D, that way we can get the distance and "weight" needed for actuation as well.

[bavman]

In all seriousness here is my idea. It will probably be not as accurate as something that can be measured electronically, but its dirty cheap and if you keep good data should give consistent results.

Explanation:

-On top of the keys rests a little basket with a known mass.
-In the basket gets dropped weights (maybe by 1-5 gram increments)
-to the basket is a taut string tied to a spring on a stand (see picture below)
-attached to the string is a pointer that is pointing to a ruler
-When weight is added to the basket, it pulls down on the string causing the pointer to move down a certain distance which can be measured by looking at the ruler

Problems:

-The spring will add to the total force since stretching it out will take more weight. This can be solved by calculating the spring constant and accounting for it.
-Travel distance is 4mm so unless we can make an accurate ruler and pointer, it will be hard to take a lot of measurements (maybe limited to 4 or 8 total measurements through the travel distance of the key)
-Something would have to be in place to keep the string perfectly still relative to the ruler. In other words this probably won't work well if the string is just dangling in air.
-???

I've had like 3 hours of sleep in the past 36 hours so this is the best I could come up with right now. Its very crude, but might be made into something workable for cheap.
I think if you repeat this multiple times, you could extrapolate a decent force vs. displacement graph
As for actuation point, just keep the keyboard plugged in and open word, and once it starts going like: "BBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBB" then you know your key has actuated

[mkawa]

Simplest solution - high resolution digital scale + digital callipers.

Set the switch on the scale and tare.
Rack in the callipers until the scale twitches and tare.
Rack in the callipers using either their reading in mm (yes mm not inches!) or the weight change on the scales in g (yes g not oz) recording both as you go, and you're done.

yes, this essentially all we need. the challenge is a) designing a reproducible combination of these two devices. b) dumping output to a recorder c) mechanical design of a jig that combines all these things to potentially accept whole keyboards. if the micro meter is attachable to a servo, we can even get frequency data, giving us a switch dynamometer. cool, eh?

if you're stuck on nickels, you're missing the point of this entirely. the point of this is to design a device that can plot force curves empirically. nickels kinda-sorta-give-you-and-idea of what force is required to hit the "down" tactile point (if one exists). nickels can't capture hysteresis, can't tell you how force ramps up as the switch is actuated, have terrible repeatability, etc. etc. etc.

the age of nickels has passed. we can do this. we have the technology, the engineering talent, the drive, and the funding*

*we may have to scrounge for the funding.

[CPTBadAss]

I got 2 ideas. One's cheap and probably will get the job done. One's a bit pricier but we'll get some graphs I believe. And if anyone reads this, please double check my math.

• Cheap: Triple beam balance + length measuring tool (not sure what yet) + stop watch + rig to put weights onto switches
  
  We create a rig that can hold a switch and buy precision X gram masses like they use on triple beam balances. We put the masses onto the switch until it actuates.
  
  
  Setup the triple beam balance like this. The pan's travel will make it easier to measure out a travel distance. Zero the scale. We weigh the masses on this bad boy. This verifies the weights and causes the pan to travel. Take the measurement of the height of the pan from the bottom datum/surface to the bottom of the pan for the potential energy height. We also time how long the travel takes from top to bottom. This gives us h (height), t (time), and m (mass). Then we use the conservation of energy, accerleration, and F=ma.
  
  
  
  Where
  
  • g = 9.81 m/s or 32.2 ft/s
  • v = final velocity
  • v₀ = initial velocity
  • F = force
  • a = acceleration
  
  We find v, then sub it into the acceleration formula to get a. Then solve F=ma
  
  We can then have 3 points per graph: Zero, Point of Actuation, and Bottom Out. However, the graph might look a little sad with this method.
  
  Variables for inaccuracies include the stopwatch, height measurement tool accuracy, and the accuracy of the masses and balance itself.
  
  But this method is only good as long as we build a rig that gives us repeatability. Maybe like a lego/knex enclosure. Someone else get on that!
  
  How's that for keyboard science? 
  
  
• Expensive: Setup a Lego Mindstorms rig that can measure force. Set that up to plot a force/deflection diagram. Haven't thought this one through as well.

[mkawa]

is there a lego mindstorms compatible strain gauge?

[CPTBadAss]

is there a lego mindstorms compatible strain gauge?

Now that would be epic. I honestly have no idea.

[mkawa]

inexpensive, commercially available strain gauges

http://www.digikey.com/catalog/en/partgroup/cea/13674

[CPTBadAss]

Next we need a breadboard or something to hook it up to to take readings. But where exactly would you get your measurements? Hook the gauges up to the springs/stem?

[mkawa]

it almost seems like what we want is a large is a big racheted caliper with a strain gauge attached to the column of the caliper. the trick in this situation would be to make the caliper's column flexible enough under sheer stress that the strain gauge measurements aren't just noise and yet the caliper's length measurements are still accurate enough to measure between 0 and 4mm with a enough sig figs to be relevant.

finally, imo this device should have serial output. hand plotting is a PITA!

[CPTBadAss]

We had setups at school that had strain gauges hooked up to breadboards which outputted graphs to Excel. So I know it can be done. I just wish I remembered more about the software side.

[mkawa]

strain gauge -> encoder -> micro platform -> USB

i consider that the easy part. the mechanical design is more opaque to me.

[CPTBadAss]

I shall ponder this dilemma good sir.

Edit: I have the world's worst sketch. Please bear with me as I explain.



We make a rig that looks like one of the aluminum keyboard plates (Borrowed from WhiteFireDragon's IC.) This will hold the switch and cap.

We have a cantilever beam which has the strain gauge mounted to it. The strain gauge is hooked up like mkawa said (the software side is a bit beyond me).

We use the carriage bolt/handle setup from a vice. We slowly turn the handle until it actuates the switch. The software generates a plot of deflection vs force. Then we win. 

[mkawa]

can we make this vice out of PLA plastic? how does that stuff deal with compressive forces?

ps, that sketch is lolwin. love the frame most

HAH

[haskellelephant]

I shall ponder this dilemma good sir.

Edit: I have the world's worst sketch. Please bear with me as I explain.

Show Image

We make a rig that looks like one of the aluminum keyboard plates (Borrowed from WhiteFireDragon's IC.) This will hold the switch and cap.

We have a cantilever beam which has the strain gauge mounted to it. The strain gauge is hooked up like mkawa said (the software side is a bit beyond me).

We use the carriage bolt/handle setup from a vice. We slowly turn the handle until it actuates the switch. The software generates a plot of deflection vs force. Then we win.  :p

I second this design, seems very reasonable. All you would need is something like this: http://www.phidgets.com/products.php?category=3&product_id=1131_0 and
a vice like thingamabob.

[CPTBadAss]

can we make this vice out of PLA plastic? how does that stuff deal with compressive forces?

ps, that sketch is lolwin. love the frame most

HAH

I would recommend not making either the arm or the vice out of the printer. Polymers have  plastic deformation and creep which I'm afraid would show up over time in this rig. In other words, you might apply enough pressure over time to cause the polymer vice to warp and not rebound. The same could happen with the arm. And if either experience this creep or deformation, our strain gauge readings are bust.

If you want, I can ask about the SLA stuff on Monday at work. And I found these material properties for PLA.

And kawa, you get good ideas or good sketches outta me. NOT BOTH! 

[mkawa]

I shall ponder this dilemma good sir.

Edit: I have the world's worst sketch. Please bear with me as I explain.

Show Image

We make a rig that looks like one of the aluminum keyboard plates (Borrowed from WhiteFireDragon's IC.) This will hold the switch and cap.

We have a cantilever beam which has the strain gauge mounted to it. The strain gauge is hooked up like mkawa said (the software side is a bit beyond me).

We use the carriage bolt/handle setup from a vice. We slowly turn the handle until it actuates the switch. The software generates a plot of deflection vs force. Then we win.  :p

I second this design, seems very reasonable. All you would need is something like this: http://www.phidgets.com/products.php?category=3&product_id=1131_0 and
a vice like thingamabob.

yaaaaaaaaaaaaaaaa dat sensor

if PLA is too soft we could go with shapeways. they have a bunch of harder materials

[CPTBadAss]

I'm not worried about hardness, I'm worried about wear over time. The only material I like on Shapeways for this is the stainless. I would really not like the vice portion to be plastic. But let me ask around at work on Monday. Someone might know how the wear on printed parts is.

[mkawa]

keep in mind: the forces involved are very small, and the number of cycles over the lifetime of the unit doesn't need to be huge.

[CPTBadAss]

Sorry, I'm too used to having things over-engineered. 

[TotalChaos]

@Mkawa

Great speech!  Your brain seems to work exactly like mine.

It won't be easy to build something that will measure the force curve of switches. You can probably make something simple that can measure distance pressed vs force to get a lot of data points, then extrapolate to get a nice smooth curve.

We need something connected to a computer that will apply gradual ever increasing force and take samples 1000 times per second.  That will give us the data we need.

This sounds like more trouble than it's worth though. It's hard to quantify something that is such a subjective matter, and that's what switch preference really is. Even if you can get all the data, then people would have to know how to interpret what the graphs mean.

I know how to interpret the graph.  And that is good enough for me 

There are millions of ppl who live there entire lives in a state of confusion, never knowing that they want or what they need or even what they have.  But you can't let those ppl hold back the progressive march of Keyboard Science.  Just ignore those ppl.

[The_Beast]

After 4 minutes looking at this thread and 2 minutes thinking. What about a force gauge (shown above) and a micrometer?



Use the force gauge connect to a computer to log force and manually enter the micrometer numbers for a force/displacement graph?

[TotalChaos]

I'm not worried about hardness, I'm worried about wear over time. The only material I like on Shapeways for this is the stainless. I would really not like the vice portion to be plastic. But let me ask around at work on Monday. Someone might know how the wear on printed parts is.

It will only be exposed to forces of 0 to 300g, correct?

[Soarer]

It won't be easy to build something that will measure the force curve of switches. You can probably make something simple that can measure distance pressed vs force to get a lot of data points, then extrapolate to get a nice smooth curve.

We need something connected to a computer that will apply gradual ever increasing force and take samples 1000 times per second.  That will give us the data we need.

Other way round! Press the switch at a fairly constant speed and measure the force.

Something like a crank/piston to press it might give constant enough speed over the part that's really of interest. Angle of rotation of the crank gives a pretty accurate measure of distance. Switch sits on a hacked digital scale rigged up to a logging device. Job done bar the precision engineering  )

Seriously, the main challenge is simply moving the slider without adding vibrations, jitter etc. And to move the slider accurately requires much higher forces to be involved, so that the force of the switch acting against it is negligible.

[mkawa]

btw, this looks like a really nice turnkey force measurement device for us. we'll have to do a compressive to shear conversion mechanically, but mr badass's little sketch already does that, so we should be good. most importantly: look at the range and accuracy. perfecto!

http://www.phidgets.com/products.php?category=3&product_id=3132_0

[CPTBadAss]

I'm not worried about hardness, I'm worried about wear over time. The only material I like on Shapeways for this is the stainless. I would really not like the vice portion to be plastic. But let me ask around at work on Monday. Someone might know how the wear on printed parts is.

It will only be exposed to forces of 0 to 300g, correct?

I believe so but I'm not sure. Just keep me in line when I start to wander off into over-engineered land.   

After 4 minutes looking at this thread and 2 minutes thinking. What about a force gauge (shown above) and a micrometer?

Show Image

Use the force gauge connect to a computer to log force and manually enter the micrometer numbers for a force/displacement graph?

I think the point is that we take out any manual calculations but thats really not a bad idea at all. Like I said: over-engineered land  :p

[The_Beast]

I think the point is that we take out any manual calculations but thats really not a bad idea at all. Like I said: over-engineered land 

reading a micrometer isn't even that hard. It would be a $25 (depends on the micrometer/pressure pad is) force/displacement tool which is cheap.

[CPTBadAss]

I think the point is that we take out any manual calculations but thats really not a bad idea at all. Like I said: over-engineered land 

reading a micrometer isn't even that hard. It would be a $25 (depends on the micrometer/pressure pad is) force/displacement tool which is cheap.

Nono, I meant that I think we want to not manually input anything. I think the ideal is to have like a strain gauge or whatever tool record the measurements.

I agree reading a micrometer isn't hard.

[mkawa]

it's a better design exercise to have serial output of the displacement

[TheSoulhunter]

[CPTBadAss]

Wow...=claps=

I applaud your sketch sir. The only issue is I've had problems in the past with height gages sticking on the track. And how would you mount the switch to the scale? Build a lot of holders?

Edit: Also I asked about the SLA printers we have at work today. The advice was that if we're going to use my screw design, the flues need to be as thick as possible. The SLA printer uses ABS which was said to have "pretty good" wear qualitiies. If Lego Mindstorms are anything to go by, ABS would be a really nice material for this project. And yes, I know caps are made out of them but keycaps aren't exactly structural. And the smallest/finest dimensions you can make is .010 inches.

[jcrouse]

Wow...=claps=

I applaud your sketch sir. The only issue is I've had problems in the past with height gages sticking on the track. And how would you mount the switch to the scale? Build a lot of holders?

Edit: Also I asked about the SLA printers we have at work today. The advice was that if we're going to use my screw design, the flues need to be as thick as possible. The SLA printer uses ABS which was said to have "pretty good" wear qualitiies. If Lego Mindstorms are anything to go by, ABS would be a really nice material for this project. And yes, I know caps are made out of them but keycaps aren't exactly structural. And the smallest/finest dimensions you can make is .010 inches.

Cost would be too high but how about a depth micrometer instead kf a height gauge.

[Soarer]

I reckon a couple of things...

1. force should be measured 'in motion', to avoid static friction effects.
2. measurement should be at least semi-automatic, to allow many measurements to be averaged feasibly.

So here's my idea drawn up roughly... the crank allows 4mm travel, and position can be measured fairly easily by a simple sensor. Just set the flywheel turning (not too fast, perhaps geared down) and let the logger do the work of taking readings over multiple presses of the switch



The best way I think to keep it smooth would be to have a strong-ish spring pushing the piston back up, and maybe just use the switch itself to hold the piston vertical.

[okooko]

Use a digital vernier caliper for height measurements
should have sufficient resolution for this anyway

[The_Beast]

Use a digital vernier caliper for height measurements
should have sufficient resolution for this anyway

Kawa want's a "serial output of the displacement", INPUTING NUMBERS IS FORBIDDEN!!!!!!!!!

[okooko]

Use a digital vernier caliper for height measurements
should have sufficient resolution for this anyway

Kawa want's a "serial output of the displacement", INPUTING NUMBERS IS FORBIDDEN!!!!!!!!!

I'm pretty sure Mitutuyo digimatic calipers have a PC link.
Mate, I'm a slacker, I would never manually enter numbers lol, if there was a better solution

[CPTBadAss]

Use a digital vernier caliper for height measurements
should have sufficient resolution for this anyway

Kawa want's a "serial output of the displacement", INPUTING NUMBERS IS FORBIDDEN!!!!!!!!!

IT IS VERBOTEN. We want lazy keyboard science!

And Soarer, I think I like your idea better than my own. Other than the fact that strain gages (or the other sensors that were posted) seem perfect for this project, I really like that idea. Can you get the scale to collect information? I imagine it would just spit out a continuous stream of masses that we would have to decipher later.

[Soarer]

My contraption wouldn't need that, since the distance can be determined from the rotation angle of the crank. The 4mm travel needn't be accurate either, as long as the thing is calibrated before use

[okooko]

http://www.tricor-systems.com/products/switch-testers/921b-force-displacement.htm
http://www.tricor-systems.com/products/switch-testers/921b-force-graph.htm


how much?

edit:
http://www.tricor-systems.com/products/switch-testers/switch-tester-921xy.htm
http://www.tricor-systems.com/products/switch-testers/force-distance-tester.htm

[mkawa]

a-hah! i was thinking about a crank today as well soarer, but for some reason i was thinking that a very small gear and a hall effect sensor would be the way to go. (hall effect sensors are cheap iirc)

[Soarer]

I like the crank because it should be fairly easy to make it operate very smoothly, without adding vibrations. The details are obviously still to be worked out - the sensor just needs to produce a decent number of pulses somehow, and give some absolute reference (e.g. the zero angle mark). It could be easier with two sensors, who knows. The noisiest part of it would be whatever holds the piston vertical.......

[TheSoulhunter]

Wow...=claps=

I applaud your sketch sir. The only issue is I've had problems in the past with height gages sticking on the track. And how would you mount the switch to the scale? Build a lot of holders?

30 seconds in Paint
Never had problems with sticking height gauges, but I mainly used the ones with cranks...
For the holder: Just one where you can easily swap the switches, no need for multiple ones.

[TotalChaos]

I am surprised nobody figured out a way to integrate a 1200DPI lasermouse sensor into the project.  The sensor need not move.  It could just point at something that does move.  Seems like it would have 1/1200th of an inch resolution.

[Bullveyr]

Depends on how much precision you guys need, 1200 CPI sounds a lot but in reality there is quite some deviations.

[mkawa]

i'm bringing this back from the dead because now that i have the 3d printer we can actually build this device

[CPTBadAss]

I can't believe I didn't even think of that.

* CPTBadAss is excited