geekhack
geekhack Community => Keyboards => Topic started by: intelli78 on Tue, 29 April 2014, 22:47:36
-
Gentlemen, I present to you keyboard science, the like of which the world has never seen the like of which.
Ok, so VesperSAINT and I were talking about the dome weighting on the FC660C in this thread (http://geekhack.org/index.php?topic=45475.msg1312516#msg1312516). I have owned 3 different Leo FC660Cs, and based on my experience with them, I believe that the later models have substantially heavier weighted domes than the earlier ones, even approaching RF 55g territory. I have seen other users make similar comments on GH, so I wanted to resolve this.
So, I did some real keyboard science.
Hypothesis
H0: There is no difference in dome weighting between my older model and my newer model
H1: The dome weighting on the newer model is heavier than that of the old model.
Procedures and Data Collection
I currently have 2 FC660Cs in my possession. A blank one I purchased new from EK in Feb 2014 and have used as a daily driver since then. Serial #944. The second one, printed, I purchased from another user who got it as part of an early batch last year, and used it for 2-3 months. Serial #419. So, to the best of my knowledge, they have been used an approximately equal amount, canceling out any potential difference due to "break in."
I popped off a bunch of keycaps on both keyboards, like this:
(http://i.imgur.com/5Bc7HXc.jpg)
(http://i.imgur.com/Xm0Kl08.jpg)
Perfect for using nickels (5 grams each). The home row is nice and flat, and with surrounding keys popped off, there is no interference with the distribution of weight.
I gently added nickels on top of each of these five keys until the domes collapsed, and recorded the weight in grams that finally triggered the collapse. When I was reaching the peak # of nickels, I gently slid them on from the side so as to not cause extra force due to downward acceleration. I did 10 trials for each of these five keys, and as my final data point, used the mode (hell yeah :) ).
Then, I did the same for the other four easily accessible keys on the home row (D, G, J, L) , for a total of nine data points per keyboard. Could have done more, obviously, but I was getting bored.
From this summary data, I could see that there was an approx. 5g difference in means in the direction I expected.
(http://i.imgur.com/0ovLp35.png)
Analysis
Tested equality of variances with f test :p
(http://i.imgur.com/BLPR7Qi.png)
Accepted the null hypothesis, variances are equal, did t-test for means
(http://i.imgur.com/Zd8A1Bp.png)
I rejected the null hypothesis with a two-tailed test at p=0.05, there was a statistically significant difference in dome weighting between the two models, the later one being heavier.
Discussion/Conclusion
The data supported and confirmed my hypothesis, however my sample size was super small and there are tons of potential confounding variables here. However, the best analysis I can do with what I have suggests the new Leo's are really different. I would like to compare these results to a similar nickel test on an RF 55g. Even though the newer model had a mean depression force of 57.2g, I think my RF 55g (which is not handy right now) nevertheless feels heavier.
Good news: if any of you guys want to repeat the nickel test using the procedure I outlined above, I can collect your data and add it to my analysis and we can try to corroborate the hypothesis, and/or get a sense for what serial numbers have the new, heavier domes :)
SCIENCE, *****es
(http://media.wbur.org/wordpress/16/files/2013/09/0927_Walter_White_cog1-592x323.jpg)
-
A+ on your lab report! Well done sir.
-
Aww shi.....I love statistics.
-
Aww shi.....I love statistics.
I love methods, hate the stats.
That said, come on people, we need an n of at least 30 for each model of keyboard so that we can hit statistical significance. And yes, that means that you should ideally be running a total of 30 tests on different switches on the keyboard in order to eliminate outliers and confirm statistical significance.
Let's science this *****.
-
Maybe when I'm not lazy I'll dump all my spring data.
-
Aww shi.....I love statistics.
I love methods, hate the stats.
That said, come on people, we need an n of at least 30 for each model of keyboard so that we can hit statistical significance. And yes, that means that you should ideally be running a total of 30 tests on different switches on the keyboard in order to eliminate outliers and confirm statistical significance.
Let's science this *****.
I have a blank sn #623. I can test with Canadian nickels that weigh 3.95 g. Or see if I can go to bank and get me some US change ;)
I'll report back with my findings
-
I can't lie...just did this **** in stats, and I absolutely hate it, but you've definitely put it in terms I can understand.
-
I can't lie...just did this **** in stats, and I absolutely hate it, but you've definitely put it in terms I can understand.
There's a reason I tried to stick with qualitative work for my Master's thesis and other research designs.
-
[quote author=drinkbleach
I have a blank sn #623. I can test with Canadian nickels that weigh 3.95 g. Or see if I can go to bank and get me some US change ;)
I'll report back with my findings
[/quote]
Cool!
-
****, now I want to get a FC660C to play too. This is really cool, thanks for sharing.
-
Serial # 681 (printed) reporting in; currently acquiring nickels, but I'll set aside some time this weekend to test as described by Nubbinator (30 tests on different switches on the keyboard) so I can add to your data pool. Oh and thanks for sharing, this is really interesting stuff!
-
I love stats! Maybe you can do a chi squared test for homogenity between all the claimed 45g topre boards? I've heard people say they don't all feel the same.
-
Question: Would it not have been better to use the same keycaps for tests on both boards, so you could eliminate any variance in cap manufactured weight?
A slightly heavier (thicker?) cap could be the difference between say a 50cn and a 55cn actuation, particularly if the true point of actuation is somewhere around 51cn.
Like the stats work, but I agree we need a larger sample size for statistically significant data. If someone feels like buying me a 660C, I'll happily contribute :P
-
wow, these findings definitely match up with my experience. i got my 660c a month or two ago, and the "45g" is definitely noticeably heavier than my "45g brown".
-
Dye Sub PBT arriving on Monday, will report nickel findings then
-
Dye Sub PBT arriving on Monday, will report nickel findings then
Don't leave us hanging! :D
-
Nothing quite so comprehensive as the OP's, but here are my results:
5 keys tested (D, G, J, L, Right Arrow)
FC660C (Serial Number 452): 13 nickels
RealForce 55g: 14 nickels
I'm actually quite surprised by these results. For an advertised 45g board, my nickel test indicated that the 660C requires an average of 60-65g per switch, whereas the 55g RealForce requires 65-70g.
I'd normally be skeptical of these numbers, but the test isn't particularly hard to perform: 1) stack nickels on a switch until it actuates 2) count nickels.
-
Nice report. I hate heavier switches. On the bright side I'll never go smaller than TKL so the leo is not a acquisition target.
Damn, many many moons ago I got A+ on both advanced stats courses in college.
Now reading this I can only vaguely remember what I learned, yet I can't make sense of the whole thing. So sad :(
-
Update: I made a video to demonstrate the difference in typing sound/volume: http://geekhack.org/index.php?topic=60984.0
-
Science!
-
I'm slowly working on gathering weighting data on my Leopold... so far the keys have been coming in at 55 - 60g collapse, 25-30g reset. They definitely feel heavier than my HHKB but I won't know for sure until more science ;)
I am wondering if I should fux with some pennies to get a more fine-grained measurement* but it may take kind of a ridiculously high stack to activate the switch :))
* http://www.usmint.gov/about_the_mint/?action=coin_specifications