Kailh Box Switch Stem Measurements and Possible Problems

Hello, I am Pleanpo. I originally posted this on Reddit but thought it would be good to post it here too! (this post is the same on found on reddit and I will update this and the reddit post the same way)

I have been browsing reddit/discord/geekhack and came across a few posts/comments claiming that Kailh Box Switches are breaking keycaps in specific, GMK sets.

About a week ago I decided that I wanted to give Kailh Box switches a try and ordered the 17 switches tester off of KBDfans to see which one would suite me the best. After hearing about the controversy surrounding the caps I decided to test each switches stem thickness individually to see how they compare to competitors in the market.

The tool I used is a 6 In. Steel Digital Caliper made by General which has a accuracy of ±0.001 in./0.02mm. This was deemed accurate enough to test my findings based on the results that were found.

Below I have created a table of my findings and have included an Imgur album with photos of all the switches and thicknesses found. The table will follow the order of the album. All measurements were done in mm with the caliper being zeroed after each switch.

Switch X-Axis Thickness Y-Axis Thickness
Tealio 1.26 1.09
Gateron Black 1.25 1.09
Cherry Silent Black 1.27 1.09
Zealio 1.26 1.07
Kailh Box Navy 1.36 1.22
Kailh Box Blue 1.39 1.16
Kailh Box Jade 1.39 1.21
Kailh Box White 1.42 1.24
Kailh Box Brown 1.39 1.17
Kailh Box Yellow 1.40 1.19
Kailh Box Orange 1.41 1.10
Kailh Box Black 1.30 1.19
Kailh Box Red 1.38 1.14
Kailh Box Royal 1.34 1.15
Kailh Pro Burgundy 1.27 1.10
Kailh Pro Purple 1.27 1.10
Kailh Pro Light Green 1.27 1.10

These measurements show that the Kailh’s Box switch on average is larger than competitors by .1 mm on the x-axis and y-axis. I would like to point out that the Kailh Pro switches were not larger than competitors and fit in the threshold for size. This increase in size can be harmful and possible damaging to keysets. Finally, I would like to know what Kailh’s response to these findings will be and how they will handle this situation.

Thank you for reading, if you would like more measurements done on any of the switches show here or others leave a comments and I will be happy to post more data (if I have the switch).

EDIT: A discord user of the name mgsickler (Mike from Novelkeys) brought this file to my attention. It is the product specification for Dark Blue box switches (navies). In the document it states that for the x-axis 1.32 mm +/- .02 mm is the intended size and for the y-axis 1.10 mm +/- .02 mm is the intended size. These numbers do not match up to my findings but I only have one sample size for navies and it could be an outlier.

21 Likes

Already replied on GH and Reddit, but my findings so far;

Box blacks here on 2 boards.
KBD19x (compact 1800) with GMK 9009 R2, never removed, 16 cracked stems.
YAS (60%), GMK Honeywell bought secondhand, removed it twice myself, 20 cracked stems.

I’ve removed all caps now, will keep an eye on both this topic and the one on GH and go from there.

I have a feeling the outcome will be to look for other switches tho and swap it.

1 Like

I’ve been talking about it as well. You can see my reddit post here. I would link the GH thread but GH is not loading for me at the moment.

I’ve removed my GMK keycaps from my box switches.

I have BOX Navies with SP SA keycaps and no issues so far. From what I’ve read, it’s only an issue for Cherry profile keycaps. Anybody having this issue with SA keycaps?

There were some people reporting that their SA Godspeed keycaps were damaged by BOX switches. I believe that was the only SA set I saw listed, though

I’ve had no issues so far with Box switches and SA keycaps. It might be because SA keycaps have a loose fit.

Yeah, a few of my SA Godspeed Novelties caps got cracked stems on their first mounting on BOX Navies. They still fit fine even moving to other switches, but I was a little bit disheartened when that happened.

Mine is Filco SA. I’m very curious as to why it seems it’s only specifically the Godspeed keyset that has this issue and not all the other SA keysets. Apart from BOX switches having thicker than normal stems, is there something about the Godspeed keyset that makes it different from other SA keysets? It concerns me as I plan to buy more BOX switches and SA keycaps in the future.

I have a BOX switch tester, and three of the caps are now broken. They are Cherry profile, if I remember correctly.

O I really hope they address it! If they don’t I would be dissapointed :frowning:

Figured I’d add from my comment on the Reddit post:

I've got 700 Jades in boards currently, so I took a look to see if there's anything going on with those:

GMK RA R1 - no cracks seen at all, late 2017 Jades
GMK Plum - a couple cracks found :(, early 2018 Jades
SA Nuke Data - couple cracks with stress marks inside the stems, early 2018 Jades
SA Lime - several cracked stems but Lime is known for that to happen, early 2018 Jades
SA Carbon R3 - no sign of cracks or stress marks, early and mid 2018 Jades
DSA Hana - no sign of cracks or stress marks, early 2018 Jades
G20 Semiotic - no sign of cracks or stress marks, early 2018 Jades

So from my analysis, more recent production box stems may have more of an affect than older runs, GMK seems more susceptible to this issue perhaps due to tighter tolerances, and sets that are already “predisposed” for stem issues (Lime and Nuke Data both had numerous QC issues) may find these issues exacerbated by Kailh’s specs being based around a thicker stem than other switches.

1 Like

Well this is not good news at all! Just went over my GMK Sky Dolch set that I had on BOX Royals & thankfully no cracks. Although it did make them fit other switches a bit looser after being put on the Royals. Right now I have 2 boards with BOX switches & both are rocking SA caps that haven’t been damaged by them. My Nightfox has BOX Royals from their first run & I got SA Pulse from it’s first run on them, then my ACR60 has BOX Navys from late 2017 & MAXKEYS SA B&G from the same time period (newer run, has the fixed kerning). I have also used ePBT sets with both boards & no issues with the caps.

Guess I should consider myself lucky after reading this post. A lot of broken caps from a pretty small sample size, not looking to good for Kaihl. Shame too, BOX switches are really nice other than this one issue. Hope they fix it quickly & it would be nice to see them do something for those who have cap sets ruined by the BOX series. Although they do have it right on their spec sheets that the stems are a bit bigger than normal MX ones. :disappointed:

1 Like

From the volume of comments, there are certainly tolerance issues. However, most people don’t have this problem so I think it’s rare.

1 Like

@Plean Sorry I’m late! I took data from my batch of Royal Hako Clears:

I took the length of the x-axis and y-axis of each stem as well as the thickness of the x-axis and the y-axis.

To take these measurements I used the Neiko 01407A.

I measured the thicknesses under the assumption that the thickness on one side of the x-axis cross is the same as the other. I applied the same assumption when measuring the thickness for the y-axis.

below are my results.

Hopefully this helps.

ALL MEASUREMENTS TAKEN IN MILLIMETERS

Hako Royal True x-axis length y-axis length x-axis thickness y-axis thickness
1 3.97 3.97 1.36 1.09
2 3.97 3.99 1.34 1.08
3 3.97 3.97 1.34 1.09
4 3.97 3.98 1.34 1.1
5 4.00 3.99 1.34 1.1
6 3.97 3.99 1.36 1.1
7 3.97 3.95 1.37 1.08
8 3.95 3.97 1.35 1.09
9 3.98 3.96 1.35 1.09
10 3.98 3.96 1.35 1.08
11 3.98 3.96 1.34 1.09
12 3.97 3.96 1.38 1.1
13 3.97 3.96 1.35 1.1
14 3.98 3.98 1.35 1.1
15 3.98 3.98 1.36 1.09
16 3.97 3.97 1.36 1.13
17 3.95 3.95 1.36 1.14
18 3.97 3.95 1.35 1.1
19 3.97 3.99 1.35 1.1
20 3.99 3.97 1.34 1.1
21 3.96 3.96 1.34 1.09
22 3.99 3.97 1.35 1.09
23 3.97 3.96 1.35 1.1
24 3.98 3.96 1.35 1.12
25 3.97 3.96 1.35 1.08
26 3.98 3.97 1.34 1.11
27 3.97 3.97 1.35 1.1
28 3.99 3.96 1.35 1.1
29 3.97 3.97 1.34 1.09
30 3.97 3.97 1.34 1.09
31 4.00 4.00 1.34 1.1
32 3.98 3.97 1.35 1.09
33 4.00 3.96 1.35 1.07
34 4.00 4.00 1.34 1.1
35 3.98 3.97 1.34 1.08
36 3.98 3.96 1.35 1.08
37 4.00 4.01 1.34 1.07
38 3.98 3.99 1.35 1.08
39 4.00 3.98 1.34 1.09
40 3.98 3.96 1.34 1.08
41 3.97 3.97 1.34 1.1
42 3.98 3.97 1.34 1.08
43 3.98 3.96 1.35 1.08
44 3.97 3.96 1.34 1.08
45 4.00 3.97 1.35 1.08
46 3.98 3.96 1.35 1.12
47 3.97 3.96 1.34 1.08
48 3.97 3.96 1.35 1.1
49 4.00 3.98 1.34 1.08
50 3.96 3.97 1.35 1.09
51 3.97 3.97 1.35 1.08
52 3.97 3.97 1.35 1.12
53 3.96 3.97 1.35 1.08
54 3.97 3.97 1.35 1.08
55 3.96 3.96 1.36 1.08
56 3.96 3.96 1.34 1.09
57 3.96 3.96 1.34 1.08
58 3.98 3.97 1.35 1.08
59 3.96 3.96 1.34 1.08
60 3.97 3.97 1.34 1.08
61 3.97 3.98 1.35 1.09
62 3.97 3.96 1.34 1.08
63 3.98 3.98 1.34 1.08
64 3.98 3.99 1.35 1.08
65 3.97 3.97 1.33 1.08
66 3.97 3.96 1.35 1.08
67 3.99 3.97 1.35 1.1
68 3.97 3.99 1.35 1.08
69 3.97 3.98 1.35 1.12
70 3.97 3.97 1.34 1.09
71 3.97 3.97 1.34 1.08
72 3.97 3.96 1.35 1.09
73 3.97 3.98 1.34 1.1
74 3.97 3.97 1.36 1.13
75 3.99 3.97 1.35 1.08
76 3.97 3.98 1.35 1.1
77 3.99 3.98 1.34 1.09
78 3.98 3.98 1.34 1.09
79 3.99 3.97 1.35 1.07
80 3.96 3.96 1.35 1.08
81 3.97 3.96 1.34 1.07
82 3.97 3.96 1.35 1.09
83 3.97 3.96 1.34 1.07
84 3.97 3.97 1.34 1.09
85 3.97 3.96 1.34 1.1
86 3.97 3.97 1.35 1.12
87 3.98 3.97 1.34 1.07
88 3.97 3.98 1.35 1.11
89 3.96 3.96 1.34 1.08
90 3.96 3.97 1.33 1.07
Mean x-axis length Mean y-axis length Mean x-axis thickness Mean y-axis thickness
3.974888889 3.97 1.346666667 1.090666667
Median x-axis length Median y-axis length Median x-axis thickness Median y-axis thickness
3.97 3.97 1.35 1.09

Note on Data: Assume 2 sig figs, so 1.1 is actually 1.10. I forgot to adjust that earlier because I was lazy. My b.

Edit: Thanks @frostbyte for catching a typo on the table. I’ve fixed it.

I can provide more data on more switches (box and non-box) this weekend when I have time.

6 Likes

I have a batch of Hako trues waiting to be used and I would really like to know if there are problematic as well.

Guess I’ll be keeping an eye on this thread as I have BOX Navies waiting to be installed in a board.

Wouldn’t your measurements suggest that every switch in the batch is out of spec? Kailh rep is posting on GH showing their measurements being right on spec.

I haven’t seen the Kailh rep’s comments yet, can you post the link to the thread here?

If my measurements suggest that than it may be so for these 90 switches. I’m planning on doing more stem measurements, box and non-box, this weekend as well.

1 Like

Maybe you can post a link to your measurements in this thread:

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