F62+F77 orders now open! New Kishsaver+Industrial Model F Keyboards

pandrew

24 Mar 2020, 04:06

tentator wrote:
24 Mar 2020, 00:10
Nah, I'd say forget it, many attempted and it was never possible so far.. i think something related to capacitive sensing missing in qmk... surely it's a wet dream of many here since qmk is the de facto standard.. do you know qmk already?
I just started reading a bit about QMK. I actually think QMK is quite possible. So far I don't see a reason why it wouldn't work. Capsense code would have to be reimplemented from scratch, because QMK is GPL2(+) and xwhatsit is GPL3(+). Yes, I know, they are compatible, but if we want main QMK to take the contribution, and it to show up in the main configurator tool, it's likely they won't accept a GPL3-only contribution. But I don't think implementing capsense is all that hard. Talking to the 74595 shift registers is simple. Talking to the DAC that generates the voltage threshold can't be that hard either. But it wold take at least a few days to implement. Testing it and tuning it could also take a while.

Spacesaver

24 Mar 2020, 07:29

Do any of you have experience operating the ibm-capsense-usb-util 0.9.0 software on a mac (version 10.15.3 Catalina)? I have installed it and when I start it up it tells me that I have to choose between two controllers. (I have my F77 plugged in. Maybe it's also trying to access my Unicomp?). No matter what I choose, the program freezes up and I am unable to use it. I wonder if it is compatible with Catalina. Any guidance would be very much appreciated.

BTW, I'm a total novice here.

xxhellfirexx

24 Mar 2020, 08:01

Spacesaver wrote:
24 Mar 2020, 07:29
Do any of you have experience operating the ibm-capsense-usb-util 0.9.0 software on a mac (version 10.15.3 Catalina)? I have installed it and when I start it up it tells me that I have to choose between two controllers. (I have my F77 plugged in. Maybe it's also trying to access my Unicomp?). No matter what I choose, the program freezes up and I am unable to use it. I wonder if it is compatible with Catalina. Any guidance would be very much appreciated.

BTW, I'm a total novice here.
It's a known bug. If you choose the wrong controller, the program freezes. If it does, just go cmd + opt + esc and kill the program. Then choose the other controller in the list.

Spacesaver

24 Mar 2020, 08:53

xxhellfirexx wrote:
24 Mar 2020, 08:01
It's a known bug. If you choose the wrong controller, the program freezes. If it does, just go cmd + opt + esc and kill the program. Then choose the other controller in the list.
That worked. Thanks!

Now that I have lowered the voltage threshold from 122 to 110, I am picking up several keys that were not registering. But there are several more that will not actuate--the < key, M, the spacebar, two of the arrow keys, and about half of the modifiers. They won't register in the software.

Am I at the point where I need to remove and try to reseat all of these keys?
They all feel okay to me, except for the spacebar, which doesn't click.
Also, has anyone else needed to set the voltage threshold that low?

Dikkus

24 Mar 2020, 10:02

Spacesaver wrote:
24 Mar 2020, 08:53
xxhellfirexx wrote:
24 Mar 2020, 08:01
It's a known bug. If you choose the wrong controller, the program freezes. If it does, just go cmd + opt + esc and kill the program. Then choose the other controller in the list.
That worked. Thanks!

Now that I have lowered the voltage threshold from 122 to 110, I am picking up several keys that were not registering. But there are several more that will not actuate--the < key, M, the spacebar, two of the arrow keys, and about half of the modifiers. They won't register in the software.

Am I at the point where I need to remove and try to reseat all of these keys?
They all feel okay to me, except for the spacebar, which doesn't click.
Also, has anyone else needed to set the voltage threshold that low?
I tried setting my voltage that low when I was having problems with a few keys, and basically all the keys ended up being registered because of it.

User avatar
Redmaus
Gotta start somewhere

24 Mar 2020, 10:26

Hey Ellipse, would it be possible to get a full pic of the red case? Thanks :)

User avatar
tentator

24 Mar 2020, 10:56

pandrew wrote:
24 Mar 2020, 04:06
tentator wrote:
24 Mar 2020, 00:10
Nah, I'd say forget it, many attempted and it was never possible so far.. i think something related to capacitive sensing missing in qmk... surely it's a wet dream of many here since qmk is the de facto standard.. do you know qmk already?
I just started reading a bit about QMK. I actually think QMK is quite possible. So far I don't see a reason why it wouldn't work. Capsense code would have to be reimplemented from scratch, because QMK is GPL2(+) and xwhatsit is GPL3(+). Yes, I know, they are compatible, but if we want main QMK to take the contribution, and it to show up in the main configurator tool, it's likely they won't accept a GPL3-only contribution. But I don't think implementing capsense is all that hard. Talking to the 74595 shift registers is simple. Talking to the DAC that generates the voltage threshold can't be that hard either. But it wold take at least a few days to implement. Testing it and tuning it could also take a while.
I can definitely volunteer for testing, and know qmk since a while.. even if I think qmk code currently became quite chaotic to understand for me.. quantum bla bla etc..

About the GPLv2/v3 thing: I actually see both licenses in qmk here:
https://github.com/qmk/qmk_firmware
and also in tmk which is the father of qmk which derived from it has both licenses here:
https://github.com/tmk/tmk_keyboard/tre ... k_core/doc
In the end I think it's contributions in diverse licenses put together but that should not matter much, what do you think?

tent

Spacesaver

24 Mar 2020, 14:28

Dikkus wrote:
24 Mar 2020, 10:02
I tried setting my voltage that low when I was having problems with a few keys, and basically all the keys ended up being registered because of it.
That didn’t happen on my unit until I reached 98 or 97. (I’m using a Mac. It appears that different computers require different voltage thresholds.) When I reach 98 or 97, even the inactive keys register multiple keys. But I can’t get them to register only one.

I will go through the inactive switches and see if removing and reseating does the trick. The spacebar doesn’t click presently, so that’s a good candidate for reseating anyway.

User avatar
consolation

26 Mar 2020, 01:00

When I first got my F77 the software registered a bunch of non existent keys in column 15, they were set as "released" in the layers; I needed to change all of them to "ignored." Otherwise, after half hour of use, random keys would start popping up. I presume those are the ground points for the capsense? Additionally, the electronics are pretty temperature sensitive, so on day 1, putting the unit in the hot water cupboard for couple hours, to bring the thick steel case to ambient temperature, made a bunch of random issues disappear. Threshold 127 is the sweet spot on my unit. Lastly, any remaining issues have been fixed by the firmware that pandrew posted on the previous page, denounce 6 in my case.
I thought that my keys were good out of the box, but reseating all of them when I took the blanks off to put terminal ones on- moved the smooth to whole new level.
My space bar lagged a bit, just need to loosen the clips a touch with a screw driver.

On a side note, I just realised that I got my board 24 hours before courier deliveries were limited to essential services only, for the next month at least. Sigh, dodged a bullet there. I swear, if it managed to get around the globe to me, only to be stopped at the depot down town, I'd have tunneled in there.

PS. My board seems to prefer some USB ports over others, on some it sometimes need to be re-detected on wake up. I'm guessing it's a power management thing? I think, it might be an issue with 3.1 over 2.1&3.0 ports, but haven't had the time to tinker. It's running without any issues atm, and it's bit of a madhouse here due to the plague lockdown.

Ellipse

26 Mar 2020, 04:00

Thanks pandrew for fixing webwit's issue! I'd like to encourage everyone to "beta test" the firmware on their new/old Model F keyboards and share their findings so it can be the default for all future Model F's that go out.

Thanks consolation for posting your new Model F photo with interesting keys as well as your technical findings with your F77.

If anyone has any Model F tips and tricks please do share them!

For my setup (tested on a Lenovo laptop from several years ago), the F77 wouldn't activate all keys unless the threshold was 122 or so but this definitely varies by USB port and computer I am sure.

Those always pressed points on the matrix are built in to the keyboard capacitive PCB. IBM had them built into original Model F keyboards as well. My guess is they were used for calibration of the key press threshold. The xwhatsit software says these always-pressed keys should be noted as "pressed" in the layout file which is what I did in the layout files for each keyboard that has gone out. That is interesting that you had to set them to ignored to prevent random key presses. Can anyone else confirm this behavior with their Model F keyboard (new or old)? Agreed that waiting for the keyboard to go to room temperature is sometimes needed before starting to use the keyboard.

https://imgur.com/0IR9Uao
https://imgur.com/rHYQq1e

Redmaus yes here are photos of the F62 and F77 red cases.
https://imgur.com/xsmAzcA
https://imgur.com/xjnlzJ6
https://imgur.com/fkZvatK
https://imgur.com/w4cNg4M
https://imgur.com/TKxUAKy


As a note here is the link to the previously posted photo album of all mailed keyboard variations that is updated whenever there is a new variation. https://imgur.com/a/gw8FOl0

User avatar
Redmaus
Gotta start somewhere

26 Mar 2020, 06:37

Ellipse wrote:
26 Mar 2020, 04:00
Redmaus yes here are photos of the F62 and F77 red cases.
https://imgur.com/xsmAzcA
https://imgur.com/xjnlzJ6
https://imgur.com/fkZvatK
https://imgur.com/w4cNg4M
https://imgur.com/TKxUAKy


As a note here is the link to the previously posted photo album of all mailed keyboard variations that is updated whenever there is a new variation. https://imgur.com/a/gw8FOl0
Nice. I'm thinking red case + black keycaps :mrgreen:

Thank you for the images!

Dikkus

26 Mar 2020, 08:08

I've been testing the new firmware for a few days now and it seems to work perfectly.

As an aside, does anyone else happen to have rod stabilizer frustrations? Both my Unicomp and my IBM SSK have pretty flawless stabilizers - no binding at all and they're quite smooth - but my F62's stabs feel pretty rough. Especially the left shift. I'm wondering if there's a trick to it or something or if I simply just have bad stabs. Hopefully I don't have to take apart the keyboard again.

Ellipse

26 Mar 2020, 16:25

Someone previously suggested adding a small rubber/felt pad with a hole in the middle on the stabilizer insert side for those with stabilized key issues. Another suggestion is to have the top of the stabilizer insert be 1-2mm above the top of the barrel.

User avatar
Twst

27 Mar 2020, 16:03

Package arrived! Big thank's to Joe!
20200326_190606.jpg
20200326_190606.jpg (247.7 KiB) Viewed 1714 times
Shipping was suprisingly fast, even with a delay with the postal service and considering the times we are in - it took only 9 days shipping all the way to the north of Norway.
20200326_191009.jpg
20200326_191009.jpg (1.27 MiB) Viewed 1714 times
20200326_191211.jpg
20200326_191211.jpg (309.97 KiB) Viewed 1714 times
Gave the F62 a split left shift and it has already been put to work at the office.
20200327_074532.jpg
20200327_074532.jpg (250.83 KiB) Viewed 1714 times
Keys are supersmoooth and both the feel and sound is perfect. Even better than my XT and AT.
The F62 in paticular has such great heft and is easily the best keyboard I have owned.

Have them both flashed with pandrew's firmware fix, and have no issues at all. The F62 is running with threshold at 131 at work on my Dell, but I had it at 129 on the Thinkpad at home. So it defently behaves somewhat differently across computers.

User avatar
Twst

27 Mar 2020, 19:08

While I wait for the dye subbed caps, the F77 will have to do with what was lying around, mostly caps from an AT ;)
20200327_184456.jpg
20200327_184456.jpg (3.67 MiB) Viewed 1660 times
Had a smaller Unicomp spacebar from a Unicomp Classic - so I added a extra barrel+flipper and got a extra key. Working without a problem.
Have the threshold settled in on the F77 at 130.

minnusing

28 Mar 2020, 03:54

Hi! Just to confirm, if I'm interested in the HHKB keys, I would need to order the HHKB 6 Key Set ($24), the F1-F12 12 Key Set ($24), and the HHKB Front Print 27 Custom Keys ($4 x 27 = $108) for $158?

User avatar
LightningXI

28 Mar 2020, 05:30

minnusing wrote:
28 Mar 2020, 03:54
Hi! Just to confirm, if I'm interested in the HHKB keys, I would need to order the HHKB 6 Key Set ($24), the F1-F12 12 Key Set ($24), and the HHKB Front Print 27 Custom Keys ($4 x 27 = $108) for $158?
That sounds right if you are trying to get the accurate legends for all of the HHKB layout keys.

User avatar
tentator

28 Mar 2020, 16:47

Dear Keyboard Community,
prepare to make a monument to user "pandrew".. I will leave the announcement to him.. but after only one night of crazy remote programming and testing with me since he phisically doesn't even have the keyboard, he was able to make the impossible, finally possible!!
I'm so happy and delighted by the result, typing with it right now here on my F62... and it works just flawlessly!

Kr,
tent:wq

skully

28 Mar 2020, 18:02

pandrew wrote:
24 Mar 2020, 04:06
Capsense code would have to be reimplemented from scratch, because QMK is GPL2(+) and xwhatsit is GPL3(+). Yes, I know, they are compatible, but if we want main QMK to take the contribution, and it to show up in the main configurator tool, it's likely they won't accept a GPL3-only contribution.
We actually have a number of situations in which QMK is distributed under GPL3+, for example when using ARM processors. I haven't looked into any of the technical aspects yet, but I stumbled on this post when someone came on our discord asking about this project and wanted to clear up misconceptions about licensing.

If you have any questions about how to get this into QMK feel free to open an issue or hit up our discord. I don't have a lot of time for forums right now but this is a very cool project and I'd love to see it make it into QMK. :)

minnusing

29 Mar 2020, 00:47

LightningXI wrote:
28 Mar 2020, 05:30
minnusing wrote:
28 Mar 2020, 03:54
Hi! Just to confirm, if I'm interested in the HHKB keys, I would need to order the HHKB 6 Key Set ($24), the F1-F12 12 Key Set ($24), and the HHKB Front Print 27 Custom Keys ($4 x 27 = $108) for $158?
That sounds right if you are trying to get the accurate legends for all of the HHKB layout keys.
Thanks for confirming. The keys cost as much as a used HHKB :shock:

pandrew

29 Mar 2020, 07:36

tentator wrote:
28 Mar 2020, 16:47
Dear Keyboard Community,
prepare to make a monument to user "pandrew".. I will leave the announcement to him.. but after only one night of crazy remote programming and testing with me since he phisically doesn't even have the keyboard, he was able to make the impossible, finally possible!!
I'm so happy and delighted by the result, typing with it right now here on my F62... and it works just flawlessly!

Kr,
tent:wq
Hi all, let me just clarify, I'm working on a QMK port for the model F keyboards. It's not yet ready to be shared publicly, but if you want to help out to collect some data from your keyboard, let me know in a PM.
Also, tentator has been of great help, testing my first pieces of code, thanks!
skully wrote:
28 Mar 2020, 18:02
We actually have a number of situations in which QMK is distributed under GPL3+, for example when using ARM processors. I haven't looked into any of the technical aspects yet, but I stumbled on this post when someone came on our discord asking about this project and wanted to clear up misconceptions about licensing.
Doesn't matter anymore since I implemented all my code from scratch. Getting it into QMK shouldn't be a problem.

Andrei

User avatar
LightningXI

29 Mar 2020, 08:01

QMK for the New Model F would be fantastic -- assuming that we won't have issues with the voltage threshold concerns that are occasionally brought up by users every now and then.

pandrew

29 Mar 2020, 10:29

LightningXI wrote:
29 Mar 2020, 08:01
QMK for the New Model F would be fantastic -- assuming that we won't have issues with the voltage threshold concerns that are occasionally brought up by users every now and then.
If anyone is having issues on their board with finding a good voltage threshold, I would love to collect some data from your board, to understand what's making it hard. I can plot scope-like signal traces with my code. I PM-ed Spacesaver, but he hasn't responded yet.

mSSM

29 Mar 2020, 10:46

Is it possible to pay extra for shipping to get the keyboard right now with unprinted caps, and the printed caps later?

I have ordered both sets of keys but my fingers are itching to have the keyboard already....

User avatar
webwit
Wild Duck

29 Mar 2020, 12:39

pandrew wrote:
29 Mar 2020, 10:29
LightningXI wrote:
29 Mar 2020, 08:01
QMK for the New Model F would be fantastic -- assuming that we won't have issues with the voltage threshold concerns that are occasionally brought up by users every now and then.
If anyone is having issues on their board with finding a good voltage threshold, I would love to collect some data from your board, to understand what's making it hard. I can plot scope-like signal traces with my code. I PM-ed Spacesaver, but he hasn't responded yet.
So I wonder how IBM did it. I doubt they made millions of keyboards which failed under different temperatures etc. Maybe the hardware has less divergences. But maybe it did some kind of self-test at startup and auto-tuned itself? I wonder if something like that would be possible.

pandrew

29 Mar 2020, 13:35

webwit wrote:
29 Mar 2020, 12:39
So I wonder how IBM did it. I doubt they made millions of keyboards which failed under different temperatures etc. Maybe the hardware has less divergences. But maybe it did some kind of self-test at startup and auto-tuned itself? I wonder if something like that would be possible.
I don't know how IBM did it. The xwhatsit's calibration is not gonna work on the model F reproductions, because the always-pressed calibration pads give a much higher signal then real buttons. I don't know if the reproduction keyboards are different, or if the xwhatsit approach is flawed. I was thinking that one way of calibrating would be to assume that when the keyboard is plugged in no keys are pressed, and then to measure the signal level for each unpressed key. Add a static offset to the measured value, and then try to come up with 2-3 bins of thresholds. For each matrix scan do 2-3 actual scans with different thresholds, and based on that determine each key status. For temperature-drift during operation, we could periodically measure the calibration pads. But I'm not sure how I should adjust the thresholds based on drift. If the reference changed +10 steps, should I change my thresholds +10 steps? We could come up with a formula, but we'd need to measure the same keyboard in various temperature conditions.

See attached waveforms from tantator's keyboard. There's a massive difference between CAL pad response, and the left control key pressed. (thresholds are on the Y axis, there's 4 pixels per step of threshold)
Attachments
out.png
out.png (6.44 KiB) Viewed 1120 times

User avatar
tentator

29 Mar 2020, 14:11

@pandrew: I'm going to test the board close to my microwave and again with the drill today! :)
@ellipse: would it be possible for you to send me a couple of xwhatsit controllers only over? I'd like to test qmk with pandrew also on some other model F boards and maybe even a beamspring I have...
I'm so happy and stunned this is working so nicely!
And btw it looks rock solid so far.

tent:wq

Spacesaver

29 Mar 2020, 20:44

pandrew wrote:
29 Mar 2020, 10:29
LightningXI wrote:
29 Mar 2020, 08:01
QMK for the New Model F would be fantastic -- assuming that we won't have issues with the voltage threshold concerns that are occasionally brought up by users every now and then.
If anyone is having issues on their board with finding a good voltage threshold, I would love to collect some data from your board, to understand what's making it hard. I can plot scope-like signal traces with my code. I PM-ed Spacesaver, but he hasn't responded yet.
I just got back to you.
There are several buttons on the bottom two rows that do not actuate at any voltage (V, M, <, spacebar, one or two of the arrow keys, and a few others). I've tried this on two different computers and have had the same result on both. The voltage threshold at which the rest of the buttons actuate is between roughly 107 and 97.

I have a 2nd F77 with no buttons that I'll set up in the next day or two to see if that one works differently.

User avatar
webwit
Wild Duck

29 Mar 2020, 21:35

pandrew wrote:
29 Mar 2020, 13:35
I don't know how IBM did it. The xwhatsit's calibration is not gonna work on the model F reproductions, because the always-pressed calibration pads give a much higher signal then real buttons. I don't know if the reproduction keyboards are different, or if the xwhatsit approach is flawed. I was thinking that one way of calibrating would be to assume that when the keyboard is plugged in no keys are pressed, and then to measure the signal level for each unpressed key. Add a static offset to the measured value, and then try to come up with 2-3 bins of thresholds. For each matrix scan do 2-3 actual scans with different thresholds, and based on that determine each key status. For temperature-drift during operation, we could periodically measure the calibration pads. But I'm not sure how I should adjust the thresholds based on drift. If the reference changed +10 steps, should I change my thresholds +10 steps? We could come up with a formula, but we'd need to measure the same keyboard in various temperature conditions.

See attached waveforms from tantator's keyboard. There's a massive difference between CAL pad response, and the left control key pressed. (thresholds are on the Y axis, there's 4 pixels per step of threshold)
Interesting. I bet if someone ever finds out, it turns out to be something really basic and simple. Those Model F were originally used in factory production lines, banks, doctors, etc. - really professional stuff. So it was out of the question the keyboard could drift under different circumstances. They would have gotten it right. Imagine a production line were a machine next to keyboard gets hot and the keyboard malfunctions, between periodically calibration. That would make IBM liable. Also both controller and the firmware are relatively simple compared to modern times, e.g. complex firmware wouldn't even fit. Maybe something that somehow normalizes the input instead of calibrating. I don't know, this is not my trade. But I bet something simple ;)

Ellipse

29 Mar 2020, 23:19

mSSM wrote:
29 Mar 2020, 10:46
Is it possible to pay extra for shipping to get the keyboard right now with unprinted caps, and the printed caps later?

I have ordered both sets of keys but my fingers are itching to have the keyboard already....
Yes this is possible but please check with me over PM for details for anyone who is interested because a number of keyboards (including maybe half of the unprinted keyboards) are in the second shipment of the early bird round, expected in 1-2 months.

Post Reply

Return to “Group buys”