Calibration (Wootility/Firmware)

[twin fiber]

What impact does the current implementation of dynamically calibrating the full travel distance of a switch have on the keyboard’s performance?

Wouldn’t manually calibrating the full travel distance of a switch by pressing each switch completely offer more accuracy (and be less resource intensive) than the current calibration: On startup at the switches’ resting positions then dynamically during use?

Wouldn’t that also offer more accuracy than an option to “select switch type” which would use the general magnetic flux range of a switch instead of the individual magnetic flux ranges the user’s switches have?

Manual calibration would require the user to press each switch completely. Each switch’s magnetic flux range would be recorded and saved by the keyboard until the user initiates a manual calibration again.

An option to manually calibrate individual switches would be imperative for QoL (per key switch type selecting for the “select switch type” option).

I don’t yet understand why Steelseries and Wooting opted for calibrating on startup because it doesn’t seem necessary (based off Rakka and SayoDevice’s “raw data”. Both seem to read the exact same with days in between without recalibrating). I assume there’s a major disadvantage or limitation to infrequently calibrating which I’ll be happy to learn about! 😅

Dynamically calibrating the last 0.5mm of a switch’s potential travel. Implemented to account for the KS-20T’s reduced travel distance.

[sleek arrow]

Would definitely like to hear more from devs on this one. Many of us are switching to 3.5mm switches and tolerances on them are probably not 0.1mm in variations. Having individual key calibration would be the way to go for extreme accuracy.

[devout cape]

im unsure what "tolerances on them are probably not 0.1mm in varations" means in the context of hall effect switches

[twin fiber]

I think he means that Jade switches can be set to 40 different actuation points despite their 0.5mm decrease in travel distance. So if you set Jade switches to 1mm in Wootility it’ll actually be 0.875mm in practice.

Whereas if you set the Lekker switches to 1mm in Wootility they’ll actually actuate at 1mm in practice.

[twin fiber]

Calibration (Wootility/Firmware)

[twin fiber]

The reason for my heap of questions:

To make the KS-20T (Jade) switches compatible¹, a firmware update was released in which all switches are considered to have a total travel distance of 3.5mm until a switch exceeds that, and its total travel distance is updated immediately (dynamic calibration²). This was implemented because it is impossible for the board to differentiate between the KS-20 (Lekker) and KS-20T (Jade) switches, as the initial magnetic flux of both is identical (102±15 Gauss). Manual calibration would make it possible for the board to differentiate between switches, as it would record the initial and final magnetic flux of each switch. It also allows the board to measure the individual magnetic fluxes the user’s switches have, instead of using a predefined final magnetic flux value for that switch type (905±80 Gauss for the KS-20 (Lekker) switches), which could increase accuracy.

1: The KS-20T (Jade) switches have always been compatible. You had to set your actuation point within the total travel distance of the switch (3.5±0.2mm) and couldn’t use DKS to its full potential, as it requires 3.6mm of total travel distance, but it was definitely compatible. A presumably better implementation would be manual calibration or switch type selecting that would use the predefined final magnetic flux value of the KS-20T (Jade) switches.

2: Dynamic Calibration, as implemented, is great when you either use KS-20T (Jade) switches or press each switch completely before use (essentially manually calibrating). If you don’t, the actuation point(s) you set won’t be accurate as long as you haven’t pressed a switch completely, which could be an annoyance or feel like an inaccuracy.

[sleek arrow]

@twin fiber When do you press each switch completely before use? upon plugging in the device?

[twin fiber]

With the current implementation? Never, as it’s not “needed” but would technically be best for optimal performance.
With Manual Calibration? Once, when you’d first plug in your board, then again only when you switch out the switches as each switch’s magnetic flux range varies slightly. 🙂

[sleek arrow]

Haha I don’t know why but your posts sounds like official features.

It got me believing they’re official changes.

I suck at reading.

[twin fiber]

😂, gotta mind the would could and should.

All good bro, I hope it’ll be an official feature one day

[twin fiber]

The HEMU Loong66 has manual calibration exactly like I suggested and from this: https://youtu.be/ut7_7qf1_N8?si=sRCRRBEY5DIdOLk5 seems to have a significant reduction in dead zone to the Wooting 60HE. Not sure if the manual calibration has anything to do with it, but it does show that it is entirely possible.

[twin fiber]

Doesn’t seem that thorough though. I think a full press and hold would be more thorough than this. 🤷‍♂️

[twin fiber]

bump

[fading oak]

bump

I'll just put in here what I said in the other thread about manual calibration

For end of press calibration, I can see merits to manual calibration.

But for the start of the press, I don't see any good reason. The key is in that position 99% of the time the keyboard is on, so it's super quick and easy to calibrate the current resting position of the signal. With manual calibration, you're somewhat locked into the calibration being used, so it's much easier for it to mismatch the current behaviour.
In one of the beta builds, we had the calibration saving across power cycles, and it was just not good.

To properly utilise manual calibration (for the start of press), you have to do so much work to make sure that the value is still valid, that it just doesn't make sense to do it when it's so quick and easy to calibrate it at runtime.
Automatic calibration is also fairly invisible to the user, whereas manual calibration, by definition, requires people to do something, so I'd say that's also a pretty big con. I find it really hard to imagine a manual calibration system producing better results than our automatic process. So you as a user have to go through additional effort for virtually no gain?

[fading oak]

You cannot determine if it's Jade or not from just the signal

The magnet is stronger and results in virtually the same signal to Lekker switches when fully pressed

When it comes to the Jades handling, manual/dynamic calibration whatever is completely irrelevant

If you use the analog monitor tool, you'll see that the Jade's display 4.0mm when bottomed out. The issue with them is primarily a visual thing in Wootility when it comes to the range shown (and ideally adjustments to the linearisation), which is why we're going to be adding configuration for the used switches.

I don't remember if we publicly said anywhere that we are going to do this, but this is what will be done to get Jades to work on the same level as Lekkers.

2: Dynamic Calibration, as implemented, is great when you either use KS-20T (Jade) switches or press each switch completely before use (essentially manually calibrating). If you don’t, the actuation point(s) you set won’t be accurate as long as you haven’t pressed a switch completely, which could be an annoyance or feel like an inaccuracy.
Do people exist that never bottom out? 😛

1: The KS-20T (Jade) switches have always been compatible. You had to set your actuation point within the total travel distance of the switch (3.5±0.2mm) and couldn’t use DKS to its full potential, as it requires 3.6mm of total travel distance, but it was definitely compatible. A presumably better implementation would be manual calibration or switch type selecting that would use the predefined final magnetic flux value of the KS-20T (Jade) switches.

DKS works fine. '3.6mm' is primarily a visual thing. In reality, how it works is more on percentages. So 3.6mm/4.0mm is 90%. So for the Jades, it'll be closer to 3.15mm where that happens

[twin fiber]

Reply to your answer in #1187862357720977530

Oh, you meant it wasn’t relevant to Woot because of its calibration, that’s what I though! 😅 Yeah, manual calibration would only be for bottom out as that can only be known by the board “recording” a key press.

The benefits are the board can support any travel distance switch without having to make a “switch select” preset for every distance switch and/or non-switch like travel limiters. During use/dynamic calibration basically already does what manual calibration would but since it is only the last 0.5 of a switch, it is hard to notice during use. I think it’d benefit accuracy too as instead of guessing, it’d be known to the board. Also wouldn’t require replugging the board if a switch was pressed during start up.

Manual calibration would only be beneficial if the user calibrates once then only ever again if reinitiated. The two other boards that have it dynamically calibrate during use too, it’s only really to set switch travel.

[twin fiber]

You mean the updated firmware treats Jade’s bottom out as a fully presses Lekker? If the Jades actually have stronger/different magnets what was the reason for dynamically calibrating during use?

[twin fiber]

With DKS or a 0.1 actuation point I’d hope so. It’s something manual cal would eliminate.

[twin fiber]

Yeah travel distance within Wootility is % rather than distance. Before Jades were supported it was 90% of a Lekker’s travel which Jades couldn’t reach iirc.

TL:DR: Benefit of manual calibration = adaptability, consistency and potentially accuracy too.

[twin fiber]

Have a bit of experience since the last 3 days. It takes half a minute to thoroughly press each key. Dynamic calibration essentially already is manual calibration. This board dynamically calibrates during use too, presumably so manually calibrating again isn’t needed. If implemented similar to this board, I strongly believe it is a far “better” approach than anything else. It’s more about saving the switch data so calibrating on start up or manually calibrating a second time isn’t necessary. Which maybe could already work if the switch data isn’t lost the moment the board loses power. 🤷

[twin fiber]

Gonna bump this thread again.

The three boards that have manual calibration are all rated higher for accuracy by reviewers. It can be a coincidence, but on paper it makes sense. Instead of guessing the bottom out value like on Woot’s boards (currently), it is known with manual calibration.
Instead of using the values recorded on startup, use the values recorded during a manual calibration and use those.
If it wasn’t apparent by now; I really think it’s the best way to do calibration. The current implementation is not a jack of all trades and adding support for different switches with different travel distances and flux tolerances sounds like the story of sisyphus to me.