Kailh Box Switch Stem Measurements and Possible Problems


#22

I guess the ones that measure 1.34 are still in spec. But there several in your batch of 90 that are over that 1.34 measurement. And your mean is higher on the x-axis

Seems x-axis is the main issue

What I know is that in 5 years of putting keycaps on, i have never actually heard a stem snap until this week. Something is wrong, whether it’s a batch issue or overall is yet to be seen. I don’t feel good taking chances on my collection.

As I’ve stated before, I’d be more than happy to re-buy the thousand or so box switches I plan to sell on r/mm if Kailh can make the stems thinner. I really like the Hako Royal, red, black, brown, and white. I have hako trues in my canoe. I’m not trying to be a hater


#23

So there is such a thing as TOO THICC?


#24

He could be on to something by measuring height, too. B/c kailh hasn’t stated anything about stem height in their specs. It might makes sense that they could also be (excuse me for a second, ladies) going in too deep.


#25

dang, took my epbt 9009 off my box jades and it looks like this set will be exclusively box switches now, as tapping my keycaps after placing them on some cherry switches had them blasting off like team rocket.


#26

Wew, Lad!


#27

Can you see any visible cracks on your EPBT 9009 caps or has the box jades just loosened the stem considerably?


#28

Hey all!

Currently on the plane to Seattle. But I wanted to make sure that everyone knew that I’m actively engaging with Kailh. I was hoping to have more solid information now, but I don’t have all the information I need yet.

Expect a post from me sometime this coming week.

Thanks for your patience while I work with Kailh to get all the information.


#29

Loosened the stems. No cracks


#30

Thank you, Mike.


#31

Thank you for taking the time to do all these measurements! Like you said, we need more data!


#32

Please design Kailh BOX Silents…and have fun in the Northwest!


#34

More measurements; this time I measured Hako True switches.

All measurements are in millimeters.

Hako Trues x-axis length y-axis length x-axis thickness y-axis thickness
1 3.96 3.96 1.34 1.11
2 3.98 3.99 1.36 1.08
3 3.96 3.97 1.34 1.08
4 3.97 3.97 1.33 1.12
5 3.98 3.96 1.34 1.10
6 3.97 3.97 1.33 1.10
7 3.98 3.96 1.35 1.08
8 3.97 3.97 1.33 1.12
9 3.97 3.97 1.33 1.10
10 3.97 3.97 1.33 1.08
11 3.93 3.95 1.33 1.08
12 3.96 3.98 1.31 1.09
13 3.98 4.00 1.34 1.09
14 3.96 3.97 1.31 1.09
15 3.98 3.99 1.33 1.09
16 3.98 3.95 1.35 1.10
17 3.98 3.97 1.33 1.14
18 3.96 3.94 1.31 1.11
19 3.96 3.96 1.33 1.11
20 3.97 3.99 1.36 1.11
21 3.98 3.97 1.34 1.11
22 3.97 3.97 1.35 1.12
23 3.97 3.96 1.34 1.10
24 3.98 3.97 1.34 1.09
25 3.97 3.96 1.31 1.09
26 3.97 3.96 1.31 1.08
27 3.98 3.94 1.31 1.08
28 3.96 3.96 1.32 1.07
29 3.99 3.97 1.34 1.09
30 3.97 3.96 1.35 1.10
31 3.97 3.97 1.34 1.14
32 3.96 3.93 1.30 1.10
33 3.97 3.96 1.30 1.08
34 3.97 3.94 1.31 1.12
35 3.96 3.98 1.30 1.10
36 3.99 3.97 1.35 1.12
37 3.98 3.97 1.34 1.10
38 3.97 3.96 1.31 1.10
39 3.96 3.95 1.36 1.09
40 3.99 3.97 1.35 1.10
41 3.99 3.96 1.31 1.10
42 3.98 3.98 1.36 1.10
43 3.97 3.98 1.35 1.08
44 3.98 3.95 1.32 1.07
45 3.97 3.97 1.34 1.09
46 3.98 3.97 1.33 1.10
47 3.98 3.99 1.35 1.10
48 3.99 3.97 1.32 1.10
49 3.97 3.96 1.33 1.08
50 3.98 3.97 1.36 1.10
51 3.98 3.97 1.34 1.10
52 3.99 3.99 1.36 1.10
53 3.97 4.00 1.35 1.09
54 4.00 3.97 1.32 1.10
55 3.99 3.97 1.34 1.08
56 3.98 3.97 1.33 1.09
57 3.97 3.95 1.31 1.09
58 3.97 3.97 1.33 1.09
59 3.97 3.96 1.34 1.09
60 3.97 3.95 1.31 1.08
61 3.98 3.98 1.32 1.10
62 3.97 3.97 1.34 1.08
63 3.97 3.97 1.31 1.08
64 3.99 3.97 1.34 1.10
65 3.98 3.96 1.35 1.11
66 3.98 3.97 1.36 1.11
67 3.96 3.93 1.32 1.11
68 3.96 3.94 1.30 1.10
69 3.99 3.97 1.36 1.13
70 3.98 3.98 1.37 1.10
71 3.98 3.96 1.31 1.11
72 4.00 3.97 1.34 1.11
73 3.98 3.97 1.35 1.10
74 3.97 3.97 1.34 1.12
75 3.98 3.97 1.35 1.07
76 3.97 3.96 1.36 1.11
77 3.97 3.97 1.35 1.10
78 3.96 3.95 1.30 1.10
79 3.96 3.94 1.32 1.10
80 3.96 3.95 1.30 1.11
x-axis length mean y-axis length mean x-axis thickness mean y-axis length mean
3.97 3.97 1.33 1.10
x-axis length median y-axis length median x-axis thickness median y-axis length median
3.97 3.97 1.34 1.10

#35

Damn looks like trues are concerned by the issues too…


#36

I measured around 50 box jades and my findings were very similar to the tables above.


#37

Here are some measurements I took of the box switches I have.

Box Yellows (Bought Sept. 27, 2017)
stem # x-axis thickness y-axis length x-axis length y-axis thickness
1 1.35 3.98 3.99 1.10
2 1.37 3.98 3.99 1.10
3 1.36 3.98 4.00 1.10
4 1.37 3.97 3.99 1.10
5 1.36 3.93 3.96 1.08
6 1.35 3.93 3.98 1.10
7 1.34 3.96 3.99 1.10
8 1.35 3.95 3.98 1.10
9 1.36 3.93 3.96 1.08
10 1.34 3.98 4.00 1.10
11 1.35 3.95 3.98 1.10
12 1.35 3.96 3.99 1.10
13 1.34 3.95 3.98 1.10
14 1.34 3.97 4.00 1.10
15 1.34 3.97 3.99 1.10
16 1.34 3.95 3.99 1.10
17 1.34 3.96 4.00 1.11
18 1.35 3.97 4.00 1.11
19 1.34 3.99 3.99 1.10
20 1.35 3.96 3.98 1.10
21 1.34 3.96 3.99 1.11
22 1.34 3.97 4.00 1.11
23 1.33 3.96 3.98 1.10
24 1.34 4.00 3.98 1.11
25 1.34 3.95 3.98 1.10
26 1.34 3.97 4.00 1.10
27 1.35 3.94 3.96 1.08
28 1.34 3.95 3.99 1.11
29 1.34 3.95 3.95 1.08
30 1.34 3.96 3.99 1.10
31 1.34 3.95 3.99 1.10
32 1.35 3.93 3.96 1.08
33 1.34 3.98 4.00 1.10
34 1.35 3.97 3.99 1.11
35 1.35 4.01 4.00 1.10
36 1.34 3.98 3.99 1.11
37 1.34 3.97 3.99 1.11
38 1.35 3.99 4.00 1.11
39 1.33 3.96 4.00 1.10
40 1.34 3.93 3.94 1.08
41 1.34 3.98 4.00 1.11
42 1.34 3.97 4.00 1.11
43 1.34 3.95 3.99 1.10
44 1.34 3.97 3.99 1.11
45 1.34 3.98 4.00 1.11
46 1.34 3.94 3.96 1.08
47 1.33 3.96 3.99 1.10
48 1.35 3.95 3.99 1.10
49 1.34 3.98 3.99 1.10
50 1.34 3.99 3.99 1.11
51 1.35 4.00 4.00 1.11
52 1.34 3.97 3.99 1.11
53 1.34 3.96 3.98 1.10
54 1.33 3.94 3.97 1.09
55 1.35 3.98 3.98 1.10
56 1.34 3.97 4.00 1.11
57 1.34 3.97 4.00 1.11
58 1.34 3.97 3.98 1.10
59 1.34 3.97 3.99 1.11
60 1.35 3.96 3.99 1.10
Mean 1.3440 3.9643 3.9867 1.1008
Median 1.34 3.97 3.99 1.10
Box Navies (Bought Feb. 20, 2018)
Stem # x-axis thickness y-axis length x-axis length y-axis thickness
1 1.32 3.99 3.99 1.13
2 1.32 4.00 3.99 1.14
3 1.32 3.99 4.00 1.12
4 1.32 1.98 3.99 1.13
5 1.33 3.99 3.97 1.13
6 1.32 4.00 3.99 1.14
7 1.32 4.00 3.98 1.13
8 1.32 3.99 3.99 1.12
9 1.32 3.98 3.98 1.13
10 1.32 3.99 4.00 1.13
11 1.32 3.98 3.99 1.13
12 1.32 3.98 3.98 1.13
13 1.32 4.01 4.00 1.13
14 1.32 4.01 3.99 1.14
15 1.32 4.00 4.01 1.13
16 1.32 3.99 3.98 1.13
17 1.32 3.99 3.99 1.12
18 1.32 4.00 4.00 1.13
19 1.32 4.00 3.99 1.14
20 1.32 3.97 3.98 1.13
21 1.32 3.97 3.99 1.12
22 1.32 4.01 4.00 1.14
23 1.32 3.98 3.98 1.14
24 1.33 3.99 3.99 1.12
25 1.32 3.97 3.99 1.13
26 1.32 3.97 4.01 1.13
27 1.32 4.00 3.99 1.13
28 1.32 3.97 3.99 1.13
29 1.32 3.99 4.00 1.13
30 1.33 3.98 4.00 1.13
31 1.32 3.98 3.99 1.12
32 1.33 4.00 3.99 1.13
33 1.33 4.00 4.00 1.13
34 1.32 3.98 3.99 1.12
35 1.32 3.99 3.99 1.13
36 1.32 3.99 3.99 1.12
37 1.32 4.00 3.98 1.13
38 1.33 4.00 3.99 1.13
39 1.33 4.00 3.99 1.12
40 1.33 4.00 4.00 1.13
41 1.32 3.98 3.98 1.13
42 1.32 4.00 4.00 1.13
43 1.32 4.00 4.00 1.13
44 1.33 3.98 3.99 1.13
45 1.33 4.01 4.00 1.13
46 1.33 3.99 4.00 1.13
47 1.32 3.98 3.98 1.13
48 1.33 3.98 3.99 1.13
49 1.33 3.99 4.00 1.14
50 1.32 3.99 4.00 1.13
51 1.32 4.00 3.99 1.13
52 1.33 3.98 4.00 1.13
53 1.32 3.99 4.01 1.13
54 1.32 3.99 4.00 1.13
55 1.32 3.99 3.98 1.14
56 1.32 3.99 3.99 1.13
57 1.33 4.00 3.96 1.14
58 1.33 3.97 3.99 1.13
59 1.32 3.98 4.00 1.13
60 1.33 3.99 3.98 1.13
Mean 1.3228 3.9565 3.9913 1.1300
Median 1.32 3.99 3.99 1.13

#38

@Owlbert Thanks so much for those measurements! I’ll add them to my list.

@Nebulant Did you have a chance to put that data into an excel or google sheet?


#39

Interesting. I guess I’ll need to measure some switches as well.

Some questions I’ve had about the situation.

  1. Are these “new” switches as in they’ve never had keycaps on them?

It seems semi plausible to me that depending on what kind of keycaps were used, the POM cruciform mount could be warped just by inserting a keycap.

I’m still a bit of a novice when it comes to materials science…but

ABS (105) seems to be not as hard as PBT (112-120) or POM (69.0-121). Compared using Hardness, Rockwell R.

But, I have no idea which formulation of POM Kaihua is using, so that it’s hard if the ABS is actually less hard or not.

Modulus of Elasticity (again from the above links)

  • ABS 2.21 GPa
  • PBT 0.420-21.9 GPa
  • POM 0.586-11.7 GPa

I’d say from those numbers that ABS isn’t as elastic (more brittle) than PBT and POM. So…ABS keycaps aren’t as durable as PBT or POM and are more susceptible to cracking it seems.

  1. Do any keycap manufacturers actually publish their keycap mounting dimensions? Cherry has in the past it seems (https://deskthority.net/w/images/c/c8/Cherry_MX_1×1_keycap_specification_(1983).pdf). If I’m reading the schematic correctly, I’d say 1.17 +/- 0.02 mm? It has to be less than the stem mount or the keycaps will fall off, lol.

  2. Is it possible the keycap fabs screwed up the melting temperatures and then how it cooled afterwards? Again, I’m not an expert in this sort of thing. Brittle ABS is a thing when dealing with 3d printing at least.

  3. What to do about the situation? Is Kaihua entirely at fault? Is GMK making faulty keycaps? TBH, I don’t know.
    Getting Kaihua to manufacture things to their spec shouldn’t be too difficult (as long as keycap insertion warping isn’t a thing). Getting Kaihua to make smaller stems might be doable, but that would require new tooling for every stem Kaihua makes.
    Really what would be better: Get keycap manufacturer to give dimensions and tolerances for their keycaps. If they are out of spec, that’s the manufacturer’s screw-up.

I’ve asked Kaihua to check all the molds of all the stems (it seems like they had already started doing this, so at least Kaihua isn’t ignoring the situation).

  1. Do keycaps fail consistently? As in, for a BOX switch that is known bad, does it consistently destroy keycaps?
    Will a BOX switch that is measured to 1.32 mm in the X-Axis destroy any keycaps?

Really, what this situation reminds me of is Matias Alps switches vs. Signature Plastics DSA PBT Alps mount keycaps. Matias tightened the stem mount slightly to better hold onto keycaps (so they wouldn’t fall out during shipping); whereas SP decided to make their Alps mount as tight as possible. And at least in 2014 when I noticed this (when working on the Infinity 60%) tried to get SP to adjust the molds so they would better accommodate Matias switches. Long story short, SP refused to do anything about the situation.

The more data we can get on all the different parts (proper datasheets and possibly even material specs for plastics used) the easier it is to focus on the problem at hand.


#40

Here are my results form measuring Box Jades I bought from KBDFans on July 19.

Switch x-axis length y-axis length x-axis thickness y-axis thickness
1 3.96 3.98 1.30 1.11
2 3.96 3.98 1.32 1.09
3 3.97 3.99 1.35 1.10
4 3.96 3.95 1.33 1.10
5 3.95 3.97 1.35 1.06
6 3.97 3.98 1.31 1.12
7 3.98 3.96 1.32 1.12
8 3.99 3.97 1.34 1.08
9 3.97 3.96 1.32 1.11
10 3.99 3.97 1.33 1.10
11 3.97 3.97 1.34 1.09
12 3.98 3.95 1.34 1.11
13 3.97 3.97 1.33 1.13
14 3.99 3.97 1.31 1.07
15 3.97 3.98 1.37 1.11
16 3.95 3.98 1.33 1.08
17 3.99 3.98 1.34 1.08
18 3.96 3.98 1.29 1.08
19 4.00 3.97 1.30 1.10
20 3.97 3.96 1.35 1.10
21 3.95 3.97 1.31 1.11
22 3.97 3.96 1.37 1.09
23 3.97 3.96 1.33 1.10
24 3.96 3.98 1.25 1.12
25 3.98 3.96 1.31 1.10
26 3.97 3.96 1.31 1.09
27 3.99 3.95 1.34 1.11
28 3.97 3.95 1.32 1.10
29 3.98 3.97 1.35 1.09
30 3.98 3.98 1.37 1.12
31 3.95 3.96 1.34 1.11
32 3.98 3.97 1.33 1.10
33 3.99 3.96 1.33 1.10
34 3.98 3.95 1.32 1.10
35 3.96 3.96 1.34 1.10
36 3.98 3.98 1.34 1.11
37 3.97 3.98 1.32 1.08
38 4.00 3.99 1.34 1.08
39 3.96 3.96 1.35 1.11
40 3.97 3.95 1.33 1.11
41 3.98 3.98 1.32 1.10
42 3.96 3.98 1.33 1.08
43 3.97 3.97 1.35 1.09
44 3.96 3.97 1.35 1.08
45 3.96 3.99 1.33 1.10
46 3.95 3.95 1.36 1.11
47 3.96 3.98 1.33 1.10
48 3.99 3.97 1.32 1.08
49 3.96 3.97 1.33 1.11
50 3.99 3.96 1.30 1.10
51 3.96 3.97 1.33 1.11
52 3.99 3.98 1.32 1.11
53 3.98 3.96 1.33 1.10
54 3.97 3.98 1.33 1.09
55 3.98 3.97 1.31 1.10
56 3.98 3.97 1.32 1.11
57 3.95 3.96 1.32 1.09
58 3.96 4.00 1.29 1.09
59 3.96 3.95 1.32 1.10
60 3.97 3.98 1.33 1.10
61 3.96 3.97 1.36 1.10
62 3.96 3.98 1.33 1.09
63 3.97 3.98 1.32 1.07
64 3.96 3.96 1.33 1.10
65 3.98 3.97 1.32 1.12
66 3.97 3.97 1.34 1.10
67 3.95 3.96 1.29 1.10
68 3.96 4.00 1.33 1.10
69 3.96 3.98 1.31 1.09
70 3.95 3.98 1.33 1.08
71 3.95 3.95 1.33 1.09
72 3.99 3.96 1.36 1.12
73 3.98 3.97 1.36 1.09
74 3.97 3.98 1.35 1.10
75 3.98 3.99 1.31 1.08
76 3.98 3.96 1.31 1.14
77 4.00 3.97 1.35 1.13
78 3.97 3.97 1.35 1.09
79 3.96 3.99 1.31 1.08
80 3.98 3.96 1.32 1.08
81 3.98 3.96 1.34 1.10
82 3.98 3.98 1.32 1.10
83 3.95 3.97 1.35 1.09
84 3.96 3.95 1.31 1.07
85 3.98 3.99 1.35 1.10
86 3.99 3.99 1.31 1.10
87 3.98 3.99 1.36 1.08
88 3.98 3.97 1.35 1.11
89 3.97 3.96 1.33 1.12
90 3.99 3.97 1.32 1.09
91 3.98 3.96 1.34 1.09
92 3.99 3.97 1.32 1.10
93 3.98 3.97 1.30 1.11
94 3.97 3.97 1.33 1.09
95 3.98 3.98 1.34 1.11
96 3.98 3.99 1.36 1.08
97 3.96 3.94 1.34 1.12
98 3.98 3.99 1.31 1.10
99 3.96 3.97 1.31 1.12
100 3.97 3.97 1.32 1.07
MEAN 3.972 3.971 1.329 1.098
MEDIAN 3.974 3.972 1.329 1.099
STDEV 0.012 0.012 0.019 0.015
CV 0.31% 0.30% 1.46% 1.35%

#41

I talked briefly to Mike from Novelkeys at the Seattle meetup yesterday and he said that he had talked to Kailh about it around 2am on Friday morning and that they were investigating and would likely make some sort of official statement soon.

He also mentioned something about it possibly being related to Dare-U being the only people buying box switches for a while and how they may have asked Kaihua for a tighter fit?

Side note, Mike’s just like actually the best person ever.


#42

Gonna get a good caliper from my buddy at work tomorrow so I can add the measurements of my batches of BOX switches to this thread. I only have Navys & Royals in boards at the moment & haven’t had any issues beyond the Navys making a GMK & ePBT set of mine fit loosely on regular MX switches. However I do have some samples of Hako Trues, Hako Clears, BOX Jades, BOX Burnt Oranges, BOX Pale Blues, & BOX Yellows, so I can add some info beyond just the Navys & Royals. I’ll post the results here once I get the measurements taken & recorded! :metal: