Total Force

User avatar
HaaTa
Master Kiibohd Hunter

17 Sep 2016, 08:53

As I'm sure you're all aware, comparing switches by actuation force, tactile force, bottom-out force is rather inconclusive when comparing different switches. But if we can't use those what can we use?

After thinking long and hard, really what we want is the Total Force of the press.
But what does that mean?

For those that remember integration from high school.

Image
Image

For those who don't remember what that means. We just multiply the force at every point between 0 mm and the bottom-out position (e.g. 4 mm). More simply, this can also be read as the area under the force curve.

Image
Image

Using this, we can derive our new Total Force unit.

Image
Image


Bonus points, anyone who remembers Physics will notice that this is also the same unit for Torque (doesn't mean the same thing though).

Now that we have a measure to compare switches, let's use it! :mrgreen:
(remember, you can click on the images to get the interactive graphs)

Image
Image
Image
Image

I leave the analysis to you guys, but notice how the difference between 45g and 55g is much higher than between 30g and 45g.

One of the nice things about how I've implemented my force gauge, I can use numpy (Python library for scientific analysis) to add whatever sort of analysis I want. I just need to come up with the algorithm to apply it to switches in general (that's tricky, but it can be done).

The measurement I use is the average across the 4 different presses rather than just measuring the first one.
Last edited by HaaTa on 17 Nov 2016, 08:58, edited 4 times in total.

face

17 Sep 2016, 09:33

Nice idea, and very cool that you actually tested stuff with a force gauge!
If I may add one thing: although they have the same dimension/unit, what you are calculating though multiplying force times distance is work, not torque. Torque and work have this relationship: W = T*phi, where phi is an angle (and thus dimensionless).
Maybe this helps understanding the meaning of this analysis better in an intuitive way - the harder it is to press a switch, the more work your fingers will have to do over the course of a keystroke. The more area under the curve, the more work you do.

I have thought about it and would like to add a few more points to your analysis.
First, I think you can compare very well switches of the same kind with this. You have to take other things into account with other switches though... for this reasons:

i) Unlike topre not every switch needs bottoming out. So you have less travel, less work (assuming same force curve).
iI) You are dissipating energy in vibration, temperature, sound etc when bottoming out or especially when hitting an o-ring on cherry switches

The force curve is very dependent on friction, but this is of course covered when measuring force. Yet, measuring cherry switches with a lot of friction will be bugged, because: if you measure the force only in very small increments, slowly, you will have a lot of static friction, unlike when pressing the key down fast.

Please take this as an positive answer to your post from an very interested viewer.

User avatar
HaaTa
Master Kiibohd Hunter

17 Sep 2016, 09:46

Haha, yep. It's not torque, just amusing nonetheless.

Not sure if you've used Topre, but they do not require bottoming out. For contact switches I have a second measurement "Actuation Force". Which while interesting, isn't super accurate because no one stops perfectly at actuation. Once I start measuring Cherry switches, you'll see this in those graphs.

I'm not using a stepper motor here. This is a smooth press at a constant rate and I'm recording the data in real-time (somewhere around 8000-10000 samples per second iirc). The data here is averaged quite because it's much more accurate coming from the ADCs than the calipers are.

The Total Force measurement isn't really the best way to measure friction, but it does work well to compare non-similar switches as a way of categorizing them.

Friction is a deviation from the expected/ideal force curve. And depending on how fast you press, the effect of friction will be different. I'm thinking about computing a semi-ideal force curve for each switch, then computing the overall deviation and using that as a measure of friction.
It won't be for a while, but I'll add speed control to the force gauge so I can test how the variance of speed affects friction.

User avatar
HaaTa
Master Kiibohd Hunter

17 Sep 2016, 10:20

I showed this force curve a long while back. But here's an example of Actuation Force.

Image

Bottom out force really isn't factored into the calculation of Total Force. Since this event happens so suddenly, I have to stop the force sensor a soon as possible (so not to destroy it). This means there are next to no data points (you can see what I mean if you click on the graph).
Even when I calculate the bottom-out position, I take the data point just before bottoming out as I found that was a more consistent place to measure.

face

17 Sep 2016, 11:47

HaaTa wrote: Not sure if you've used Topre, but they do not require bottoming out. For contact switches I have a second measurement "Actuation Force". Which while interesting, isn't super accurate because no one stops perfectly at actuation. Once I start measuring Cherry switches, you'll see this in those graphs.
Very interested to see them!

Yeah, I used Topre, though only the HHKB 45g switches. Sure, they don't need to bottom out, but because of the "dent" after the actuation point they suck the fingers and most people can't actuate them without bottoming out.

What would be very interesting is a force curve for hypersphered topre switches. Maybe you get the fingers on one anytime.

User avatar
XMIT
[ XMIT ]

17 Sep 2016, 13:03

This is force dot distance, not just force times distance. Two vectors yielding a scalar. This is better know as work. Torque, a vector quantity, is force cross distance.

I agree that work is the most meaningful metric of how to measure the effort required for a key press. This nicely explains why the 55g Topre switch is so laborious to press.

Next up is quantifying scratchiness. For this I think measuring the standard deviation of a force curve would be best since scratchiness should look like noise. Also mushiness, which should look like hysteresis, and/or a curve where bottoming out does not look very vertical.

These curves are beautiful HaaTa, keep up the good work. I was skeptical about the Topre force curves on EliteKeyboards.com but you have reproduced them exactly.

face

17 Sep 2016, 14:10

XMIT wrote: This is force dot distance, not just force times distance. Two vectors yielding a scalar.
Yes, but since in this case the force only acts (or whatever to say in english, which wasn't my best subject in school) in the direction of the travel (you could say because it is a one-dimensional view only too), force dot distance becomes force times distance.

User avatar
XMIT
[ XMIT ]

17 Sep 2016, 17:28

face wrote:
XMIT wrote: This is force dot distance, not just force times distance. Two vectors yielding a scalar.
Yes, but since in this case the force only acts (or whatever to say in english, which wasn't my best subject in school) in the direction of the travel (you could say because it is a one-dimensional view only too), force dot distance becomes force times distance.
Sure, the two are the same in value, but one is still a scalar and one is still a vector. It just so happens that when the force and distance are in the same direction, the magnitude of the torque vector is exactly the work scalar.

(Your English is fine!)

User avatar
Wodan
ISO Advocate

17 Sep 2016, 17:52

I'm having a "hard" time hiding my excitement about your recent work Haata. This force measuring device brings a good amount of science to this subject. So much potential and such brilliant results ...

There is one assumption I recently made when I was trying to measure the "smoothness" of switches with a lavalier mic.
workshop-f7/the-scratch-o-meter-t14495.html#wrapper

My conclusion is based on the assumption that Gateron switches have a "weaker" crosspoint contact construction than Cherry switches. It is impossible to me to verify that with my means. If you have the time, I would really love to see a force curve measurement of the identical slider+spring in different MX-based switches (Cherry, Gateron, Kailh etc.) and see if the force curves differ. That's the best way I can come up with to measure the influence the crosspoint contacts have on the force curve of a MX switch.

face

18 Sep 2016, 12:01

XMIT wrote: Sure, the two are the same in value, but one is still a scalar and one is still a vector. It just so happens that when the force and distance are in the same direction, the magnitude of the torque vector is exactly the work scalar.

(Your English is fine!)
Thanks, happy to hear that :)

User avatar
adhoc

19 Sep 2016, 08:45

XMIT wrote: I agree that work is the most meaningful metric of how to measure the effort required for a key press. This nicely explains why the 55g Topre switch is so laborious to press.
Indeed, when I saw the graph I thought to myself "yeah, makes sense". 55G switches did not feel only ~25% harder to press at all, they were actually fatiguing during fast typing sessions. But they do feel amazing. Heh.



What happened to Topre 45G '91 OEM switches? There doesn't seem to be an actuation point or am I blind? Does this mean this is what we can expect with our beloved Topre boards after a couple of decades of use?

User avatar
Chyros

19 Sep 2016, 10:07

Probably also explains why Alps SKCM neon green feels stiffer than other Alps switches. Idem SKCM brown.

Post Reply

Return to “Gallery”