Daskeyboard Pro 4 and Matias Tactile Pro with Bouncing Issues

User avatar
Inxie

Daskeyboard Pro 4 and Matias Tactile Pro with Bouncing Issues

Unread post by Inxie » 26 Dec 2018, 22:31

I am seeing this beeing an issue on modern mechanical kkeyboards, as if those manufacturing them forgot the importance of debouncing. I am typing this on a Daskeyboard 4 Pro currently. The bounce above I didn't intentionally type. Further bounces I'll try my best to correct.

The worst offender I have is the Matias Tactile Pro, which has gotten so bad that nearly every letter is bounced into oblivion (I wish I was kidding, I have a video on YouTube demonstrating how I could NOT log into a Unix box because I couldn't actually see how bad the password was being typed). The Das Keyboard isn't *that* bad, but does get finicky at times. I tried to contact them over 4 years ago when this developed I was met with not only rudeness on the companies part, but the rep not even knowing what debouncing was. This seems to be the case, because while every switch will bounce, how well the microcontroller in the keyboard deals with it is important.

Take my Dell AT101W for example, I have never had that keyboard have a single bouncing issue once, and the Matias is supposedly cloning the alps switches, yet there's a huge difference (and I know, black vs white Alps, but even then, I've heard no complaints from Alps white owners?). The Das Keyboard I have was ordered with MX Blues, and it's taken it 4 years to develop bouncing issues, but surely the firmware in the keyboard should have been programmed for this, since the AT101W is hitting over 20 years old and has never once exhibited this problem?

Both companies won't honor fixing this issue, and Matias going so far to tell me that "mechanical keyboards are unreliable" - There words at the time which was what prompted me to get the Das Keyboard in the ffirst place.

Anyway, maybe someone has insight in this issue regarding modern keyboards? I am sure there is software to get around this, as I've tried one, but it interferes with gaming, and more than once this issue has made me game fairly poorly (as if I'm already crap anyway at some of these games, lol)

User avatar
Muirium
µ

Unread post by Muirium » 27 Dec 2018, 01:18

Chatter is an issue with Matias boards, unfortunately. I don’t know any simple way around it. There’s a fair bit of discussion on the topic here:

viewtopic.php?f=2&t=19813

You’d think this was fundamental stuff, that couldn’t be allowed to fail so routinely. And, while right, you’d be wrong.

User avatar
abrahamstechnology

Unread post by abrahamstechnology » 27 Dec 2018, 01:32

Does your Tactile Pro have Matias switches or is it one of the old Alps SKBM models?

User avatar
Inxie

Unread post by Inxie » 27 Dec 2018, 02:46

abrahamstechnology wrote: Does your Tactile Pro have Matias switches or is it one of the old Alps SKBM models?
That I'm not entirely sure. The keyboard is 3 hours away at my old house I was evicted from. I could try to get it back maybe in a month or so, but the keyboard was manufacturered around 2012.

User avatar
Inxie

Unread post by Inxie » 27 Dec 2018, 02:54

Muirium wrote: Chatter is an issue with Matias boards, unfortunately. I don’t know any simple way around it. There’s a fair bit of discussion on the topic here:

viewtopic.php?f=2&t=19813

You’d think this was fundamental stuff, that couldn’t be allowed to fail so routinely. And, while right, you’d be wrong.
Tell me about it, because *THIS* is pretty bad. This is the video of me attempting to use the Matias: https://www.youtube.com/watch?v=7VdfoNXeAFs

Clearly that's crap. I baby these keyboards, clean them monthly and put a cloth over them when not in use.

This was in 2013, a year after the keyboard failed. You can clearly see here I wasn't lying, they tried to basically claim we aren't buying reliability, but the feel, what utter horrible customer service, and only offer a 25% discount on another "new" one, rather than fix the faulty one. That's when I got the Das Keyboard, because why in the hell would I buy from them again?

Image

User avatar
Inxie

Unread post by Inxie » 27 Dec 2018, 02:55

I kept that email, because that was absolute proof that Matias doesn't care about your keyboard, they care about your money rather than deliver a quality product.

I am more than happy to forward that email to anyone who wants proof that this happened if proof is every required. Matias failed to honor the repair on my keyboard, and forced my hand.

Polecat

Unread post by Polecat » 27 Dec 2018, 05:43

Long time Alps user, going on three decades, but just got my first Matias-switch keyboard a few weeks ago. Blue or early white Alps switches last pretty much forever in my experience. The quality of whites got worse as time went on, and the changes are documented here in great detail. There's a lot more to it than complicated/simplified, pine/bamboo, or slots/no slots. I used a Monterey K104 for probably ten years, with early white SKCM, and in that time it literally never missed a keystroke or had a single chatter. And it was well used already when I got it. Still has the $6.00 price tag on the bottom from the thrift store where I bought it. I only stopped using it because I wore the legends off the keycaps and because it was badly yellowed. I really wouldn't expect Matias or any other modern switch to match the performance of something from thirty years ago, especially when we expect to get them for 20 cents a pop. Just my two cents.

User avatar
Inxie

Unread post by Inxie » 27 Dec 2018, 13:41

Polecat wrote: Long time Alps user, going on three decades, but just got my first Matias-switch keyboard a few weeks ago. Blue or early white Alps switches last pretty much forever in my experience. The quality of whites got worse as time went on, and the changes are documented here in great detail. There's a lot more to it than complicated/simplified, pine/bamboo, or slots/no slots. I used a Monterey K104 for probably ten years, with early white SKCM, and in that time it literally never missed a keystroke or had a single chatter. And it was well used already when I got it. Still has the $6.00 price tag on the bottom from the thrift store where I bought it. I only stopped using it because I wore the legends off the keycaps and because it was badly yellowed. I really wouldn't expect Matias or any other modern switch to match the performance of something from thirty years ago, especially when we expect to get them for 20 cents a pop. Just my two cents.
I have problems with that though in regard to the "wouldn't expect matias or any other modern switch to match the performance".

1. Matias shouldn't be flat out falling on it's face in one year, with a customer service rep being rather rude and insulting to worm their way out of warranty repair.
2. Most of these chatter issues could have been fixed in the microcontrollers firmware, as they should have good debouncing algorithms anyway (or caps, electrically). It's like the threshold is far too small.

User avatar
swampangel

Unread post by swampangel » 27 Dec 2018, 14:32

Inxie wrote: I have problems with that though in regard to the "wouldn't expect matias or any other modern switch to match the performance".

1. Matias shouldn't be flat out falling on it's face in one year, with a customer service rep being rather rude and insulting to worm their way out of warranty repair.
2. Most of these chatter issues could have been fixed in the microcontrollers firmware, as they should have good debouncing algorithms anyway (or caps, electrically). It's like the threshold is far too small.
I agree with you.

Early Matias boards had problems - fine. But it's his problem to fix, not just kick it down the road and make it the user's responsibility.

User avatar
Muirium
µ

Unread post by Muirium » 27 Dec 2018, 15:24

Fun idea: debounce takes many milliseconds to implement per keypress. That’s just how it works: the controller takes a wait and see approach, buffering the keystroke until it’s sure. So maybe they’re cutting back on debounce for gaming ULTRA FAST RESPONSE RATE cred.

User avatar
Inxie

Unread post by Inxie » 27 Dec 2018, 15:44

Muirium wrote: Fun idea: debounce takes many milliseconds to implement per keypress. That’s just how it works: the controller takes a wait and see approach, buffering the keystroke until it’s sure. So maybe they’re cutting back on debounce for gaming ULTRA FAST RESPONSE RATE cred.
I'm fully aware of how debouncing works, but the time required would certainly not affect gaming, we'd be talking time that's still shorter than the input lag on your LCD monitor.

User avatar
Muirium
µ

Unread post by Muirium » 27 Dec 2018, 16:13

I’m not talking about gaming. I’m talking about marketing.
Cherry wrote:CHERRY MX SPEED Silver is the perfect high-precision mechanical switch with extremely short reaction times and high switching frequencies.
Everyone’s at it these days. As if every gamer’s scourge was that laggy keyboard in front of them! Marketing is all about perception, not practicality.
Last edited by Muirium on 27 Dec 2018, 16:20, edited 1 time in total.

User avatar
Inxie

Unread post by Inxie » 27 Dec 2018, 16:20

Oh ok, so it's like the Electronic Arts of keyboards. Got it. ;)

User avatar
Muirium
µ

Unread post by Muirium » 27 Dec 2018, 16:25

It’s like the world. You know, the little round place we all live. Marketing is more real than reality.

User avatar
depletedvespene

Unread post by depletedvespene » 27 Dec 2018, 16:27

Inxie wrote: Oh ok, so it's like the Electronic Arts of keyboards. Got it. ;)
Not exactly. Long ago(*), Electronic Arts was a good company.





*: And far away, in a different age, when I was a dumb young guy.

User avatar
Inxie

Unread post by Inxie » 27 Dec 2018, 16:29

depletedvespene wrote:
Inxie wrote: Oh ok, so it's like the Electronic Arts of keyboards. Got it. ;)
Not exactly. Long ago(*), Electronic Arts was a good company.





*: And far away, in a different age, when I was a dumb young guy.
"Was", but isn't anymore. Now it's loot boxes and marketing.

User avatar
Muirium
µ

Unread post by Muirium » 27 Dec 2018, 16:35

Anyone who survives: marketing.

User avatar
Inxie

Unread post by Inxie » 27 Dec 2018, 16:46

I should start replying to customer service emails as a marketing agent, and spin it around like an Italian mafia offer they can't refuse.

User avatar
Muirium
µ

Unread post by Muirium » 27 Dec 2018, 17:06

Marketing reigns supreme for passive consumers. It’s the only channel they know or care about.

Customer service is a different beast. You know well, surely, that your job is to appear caring and helpful while not caring and offering the barest minimum of help. A customer waylaid is a customer retained.

User avatar
XMIT
[ XMIT ]

Unread post by XMIT » 27 Dec 2018, 17:51

Wow, that's extra terrible. Sorry to hear that customer support nightmare story from Matias.

If you had sent that e-mail to XMIT Keyboards about one of our boards, the response would basically have been:
- We're really sorry for the trouble.
- Please return your keyboard for a full refund, if we're not able to service it.
- If you'd like, you may purchase the next revision of the keyboard at a discount, or service parts at cost.

The quip about your job was uncalled for. I can tell "Barb" was having a bad day.

Debouncing is a pretty fundamental property of mechanical switches. The only way to really quell the bounce time is mechanically, meaning, to make the switch components as light (with as little mass) as possible, and as stiff (as high a spring constant) as possible, so they just plain bounce less. I wonder if anyone's experimented with little elastomeric dampers in the switch itself to give it more reliable properties, but then again, elastomers don't tend to last.

With key switches we sort of expect that they won't be held down for any less than a few milliseconds, so, a keyboard controller could in theory be super sensitive to key presses, not try to debounce, and always hold a key down for what would otherwise be the debounce time.

If I were Matias - meaning, if I were in the business of sourcing switches for a keyboard design that I owned, including possibly having them manufactured - I would characterize them really thoroughly before agreeing to buy a million of them.

I'm a little bit obsessed with inherently bounce-free designs. That's why I love Hall effect sensing so much. But optical, capacitive, and inductive sensors are also bounce free; maybe consider a keyboard with one of those?

(Too bad magnetoresistive materials are still way too expensive for keyboard designs. Imagine a magnetic sensing keyboard that was based on passive rather than active components! Wouldn't that be something?)

User avatar
Inxie

Unread post by Inxie » 27 Dec 2018, 18:24

XMIT wrote: Wow, that's extra terrible. Sorry to hear that customer support nightmare story from Matias.

If you had sent that e-mail to XMIT Keyboards about one of our boards, the response would basically have been:
- We're really sorry for the trouble.
- Please return your keyboard for a full refund, if we're not able to service it.
- If you'd like, you may purchase the next revision of the keyboard at a discount, or service parts at cost.

The quip about your job was uncalled for. I can tell "Barb" was having a bad day.

Debouncing is a pretty fundamental property of mechanical switches. The only way to really quell the bounce time is mechanically, meaning, to make the switch components as light (with as little mass) as possible, and as stiff (as high a spring constant) as possible, so they just plain bounce less. I wonder if anyone's experimented with little elastomeric dampers in the switch itself to give it more reliable properties, but then again, elastomers don't tend to last.

With key switches we sort of expect that they won't be held down for any less than a few milliseconds, so, a keyboard controller could in theory be super sensitive to key presses, not try to debounce, and always hold a key down for what would otherwise be the debounce time.

If I were Matias - meaning, if I were in the business of sourcing switches for a keyboard design that I owned, including possibly having them manufactured - I would characterize them really thoroughly before agreeing to buy a million of them.

I'm a little bit obsessed with inherently bounce-free designs. That's why I love Hall effect sensing so much. But optical, capacitive, and inductive sensors are also bounce free; maybe consider a keyboard with one of those?

(Too bad magnetoresistive materials are still way too expensive for keyboard designs. Imagine a magnetic sensing keyboard that was based on passive rather than active components! Wouldn't that be something?)
Yea, Barb wasn't on my high list that day. lol. 5 years later there's not much I can do, but I feel it's time I talk about it though, considering Matias is really pushing themselves as an ALPS keyswitch source, when my experience with their switches has been rather poor, including the customer service.

For debouncing however, I had learned that there were two ways to reduce it or eliminate it. Capacitors on the switches, or average the input to the microcontroller through a very small amount of time (like 30-60ms). Based on how bad my Matias performs, I'm curious if it had debounce at all, because as you saw in the video, literally every key bounced. Surely there can't be any debouncing going on if it's that bad.

User avatar
vvp

Unread post by vvp » 27 Dec 2018, 18:57

add debouncing:

It is there everywhere, Including hall-effect, optical, etc sensors. It is just not called debouncing but hysteresis in those cases. But conceptually it is the same: either key press or key release will be postponed:
  • Analog sensors will need addition of some hysteresis in the software so that they do not flip randomly around the actuation location. That will lead to delay in the threshold cross detection. The lower limit of the delay is depending on the maximum speed of the moving keycap/stem.
  • Mechanical switches must be debounced which leads to time delay as well. Some firmwares debounce both key press and key release but only one of them needs to be debounced.
The result is about the same for the end user. There is some time delay added to either key press detection or key release detection. Or both.

Typical debounce time for mechanical switches is about 5 ms. One can definitely do much better with analog sensors. Of course, a group of users who care about time delay of 5 ms e.g. for key release is not big.

User avatar
Muirium
µ

Unread post by Muirium » 27 Dec 2018, 19:11

I’d like to see someone take an oscilloscope to some badly chattering Matias switches. What I expect to see is a lot of white noise at the individual switch level. Ever played with a flickery light switch? They can pulse out an impressive range of frequencies of wrong, for amazing lengths of time when you hit the sweet spot on the dimmer. Physical contact switches break down badly when they do.

Now, a controller has got to somehow figure out what to do with all that noise, while spinning plates. Controllers don’t get the luxury of watching one switch at length, they’re strobing a whole bunch of different ones in fast cycling waves. What to do when you’re bamboozled by noise?

That’s what I’m getting at: even debounce has a limit. When switches are hot garbage and chatter like a skeleton’s teeth in an amusement park spook ride, there comes a point when there’s nothing you can do. Delay key presses so long htey flal uto fo rode?r

User avatar
vvp

Unread post by vvp » 27 Dec 2018, 19:38

If a (mechanical) switch is scanned with digital input then the firmware does not see all the changing frequencies of a bad switch. It only reads sometimes 1 and sometimes 0. Firmware will finish debouncing routine when only 1 or only 0 are read for a given period of time.

In a way, the debouncing routine works like a low pass filter. All the high frequencies are filtered out. Low frequency is passed and interpreted as key presses/releases at zero cross points.

My experience indicated that when a mechanical switch does not debounce in 6 ms then one should just replace it. I had some bad switches which stopped to debounce well within 6 ms. I have configurable debounce time on my keyboard so I just increased it easily to about 8 ms. It worked better (but not perfectly) for about a week. Then I needed another 2 ms increase. Then I needed another increase but these were not really helping. My conclusion would be that if a switch misbehaves then it may make sense to increase the debounce time to 8 or maybe even 10 ms. If it is not enough then the switch must be replaced. The experience is with Gatheron brown and white switches.

User avatar
XMIT
[ XMIT ]

Unread post by XMIT » 27 Dec 2018, 19:42

To be super pedantic, when I refer to bounce I refer to mechanical bounce, which is strictly speaking not the same thing as hysteresis. With bounce there is an amount of time that the switch is in an indeterminate state. With hysteresis there is a gap between the "make" and "break" points.

User avatar
vvp

Unread post by vvp » 27 Dec 2018, 20:00

Yes, agreed. But the consequence is the same. A delay for the user. And how the user will perceptive is the only thing which matters.

That said, it is good you are trying analog switches. You should definitely achieve smaller delay times. And for some things analog switches would be great. E.g. they could work as analog joystick axe when some modifiers are pressed. Then one can merge an analog "track point" with a standard key ... without needing a "nipple" between keys like some laptop keyboards have. That would be an interesting experiment.

User avatar
Inxie

Unread post by Inxie » 27 Dec 2018, 20:11

I've always judged my keyboards based on weather I could jam out with Native Instruments FM8 and play along with a song using the keyboard vs a MIDI keyboard. The AT101W does have a slight considerable delay compared to the DasKeyboard that is perceivable when you press a key to when a sound is produced by the soft synth. It's about 20ms from what I can tell, whereas my DasKeyboard is about 10ms.

User avatar
Muirium
µ

Unread post by Muirium » 27 Dec 2018, 21:21

I’m loathe to eye roll at a fair decent point, but. Playing a midi instrument with a computer keyboard, no matter how nice the keyboard, is like playing a piano with a pair of spoons. Hook up something real, for the love of music!

Capsense and other analog techniques could indeed make a better back up pointing device than a trackpoint. But piano keyboards require a good bit more finesse. And, ah, the whole layout thing…

User avatar
abrahamstechnology

Unread post by abrahamstechnology » 27 Dec 2018, 21:24

For raw switches I'm switching from Matias to Hua-Jie mainly because of the slider issue, but also for the noise and keyfeel. Hopefully Hua-Jie doesn't have chattering or other weird issues.

User avatar
Inxie

Unread post by Inxie » 27 Dec 2018, 21:25

Muirium wrote: I’m loathe to eye roll at a fair decent point, but. Playing a midi instrument with a computer keyboard, no matter how nice the keyboard, is like playing a piano with a pair of spoons. Hook up something real, for the love of music!

Capsense and other analog techniques could indeed make a better back up pointing device than a trackpoint. But piano keyboards require a good bit more finesse. And, ah, the whole layout thing…
Hey this is just when you need to write something and the MIDI keyboard is packed up (since my keyboard is a Yamaha DX7 and that's a bit of a heavy beast).

You can map out two octaves and a half, rather easily with a computer keyboard, and the keys do line up properly even. But like I stress, I only do this primarily if I'm writing something, or just want to quickly jam out to a song and learn the notes. I've memorized where each key corresponds to which note by now.

Post Reply

Return to “Keyboards”