Bringing the IBM PC XT into the 21st Century

Icarium · 09 Jul 2012, 19:59

Damn, I need another.

dorkvader · 10 Jul 2012, 05:33

Icarium wrote:Damn, I need another.

After I heard about this from Wcass, I bought one almost immediately. (used lightly in box: not a bad deal, IMO) The only thing that had been holding me back was the layout.

I'm pretty happy about this coming to fruition: Those IBM keyswitches are pretty excellent.

Next, we need a flexible PCB project to replace the bottom membrane in an M (or unicomp) to achieve NKRO ('cause you can solder diodes to it).

Icarium · 10 Jul 2012, 19:24

Really? Where have you found one so quickly?

dorkvader · 11 Jul 2012, 01:16

There was one on ebay. I'm really surprised with how little used it looks, though the legends have this ugly yellow border around them.

I won the auction about 16 hours after he Pm's me about this project. I got really lucky again (all my model F's have been really good finds off ebay)

It's also super easy to take apart. I do need to get a nut driver in that size, since it's just about the only size I don't yet have and using pliars to loosen them is sort-of a pain.

Icarium · 11 Jul 2012, 07:12

Yeah, the XT is way easier to take apart then the AT. None of the sticky foam shit.

JBert · 11 Jul 2012, 09:04

It is far harder to reassemble though, thanks to the greater pressure between the front- and back-plates of the inner assembly. The AT ones simply slide together, whereas my XT needed some clamps and a few hits with a hammer to get all the hooks in their slots.

wcass · 11 Jul 2012, 19:44

Hay, checking preferences here; would you prefer to re-use the slot/tab closure method or go with a nut/bolt?

AKIMbO · 11 Jul 2012, 20:00

wcass wrote:Hay, checking preferences here; would you prefer to re-use the slot/tab closure method or go with a nut/bolt?

Nut/bolt. The slot tab was a PITA to align the spacebar stabilizer properly.

Ratfink · 11 Jul 2012, 20:16

I'd say nut/bolt. The slot/tab thing is scary to assemble/disassemble.

JBert · 11 Jul 2012, 21:25

I once posted a tutorial on Geekhack how you could keep the spacebar in place when you're sliding the plates together. I thought I added it to the end of my "mod your model F spacebar" tutorial.

But yes, nut/bolt is preferable if there's room for it, then you can at least fasten each nut separately and even choose how much pressure you want to put on the barrels.

wcass · 11 Jul 2012, 23:26

the new version will have a Model M space bar and stabelizer - and the plates will bolt together.

dorkvader · 12 Jul 2012, 00:46

This is for mounting the metal plate to the PCB?

I'd say nut-bolt. I wasn't going to try it unless I had to because I'm afraid of drilling the PCB, but in this case, we can have bolt holes in the PCB and plate already, so all we'd need are the bolts/nuts. (It'd make sense to run a GB for them and distribute with the plates/PCB's. Hardware can get pretty inexpensive if purchased in bulk)

Also, I feel that nut/bolt is ultimately a better solution, and won't take that much more effort.

wcass · 14 Jul 2012, 05:47

If you have not already done so, please vote for your favorite name for this keyboard. Thanks
http://deskthority.net/keyboards-f2/nam ... t3194.html

I thought I would bring up LED lighting for lock indicators and for subtle back lighting. I want to make this perfectly clear that all LED lighting will be optional. The kit instructions will include a How-To, but will probably not include any of the parts required to do this unless there is lots of interest in it. Total cost for this is under $10 - probably closer to $5

Around the parameter of the keyboard case is a few 1/10th inch stand-offs. I plan to use 1/10th thick acrylic sheet as an LED lens to emit light from the edge between the case and the keys. For the pictures below, I placed a mouse on one end of a 10” x 1.5” strip of 1/10th inch thick acrylic. These pictures were taken without flash – the reflection is from the room’s overhead lighting.

: standoff.JPG (97.8 KiB) Viewed 7001 times

: red LED.JPG (90.47 KiB) Viewed 7001 times

: blue LED.JPG (92.59 KiB) Viewed 7001 times

I plan to make another set of test switches this weekend; wish me luck.

bazemk1979 · 15 Jul 2012, 15:50

keeping fingers crossed for you bud!!!!!!!!!!

wcass · 15 Jul 2012, 22:37

DIY PCB take 2

For my next attempt I found a glossy corporate magazine with thicker paper. I followed the same procedure for prepping and ironing and the transfer results were again unsatisfactory, so I did not bother etching and just cleaned off the toner that did stick.

Working under the assumption that the problem lay with my printer or toner, I decided to give the local office supply store a try. I printed from KiCAD to PDF and took it to Office Depot to be printed on “glossy text” paper (total cost $0.39 for 2 pages). The first thing I noticed was that the pad fills were free from the streaks that I had with my cheap Brother printer. I prepped and ironed the same as the other attempts. While ironing, I noticed that the paper shrank much more than the magazine paper I had used – perhaps > 1%. I also noticed the pattern bled through while ironing the paper much more than with my printer, toner, and magazine paper. But the end results speak for themselves. Testing for continuity, only one of the pads was found to be “bad” and this was because of incomplete etching – the pad is still touching the ground. Next time I need to use the Scotch Brite or an old tooth brush to remove more of the paper.

While drilling through holes i discovered that small drill bits burn out quickly. As I only had one bit smaller than 1/16", that is all the update for today. Smaller bits are available at the hobby shop near my work, so will likely finish the test switches tomorrow.

: PCB top.JPG (96.21 KiB) Viewed 6975 times

: PCB bottom.JPG (90.71 KiB) Viewed 6975 times

: PCB xray.JPG (87.1 KiB) Viewed 6975 times

Icarium · 15 Jul 2012, 23:17

Well, I am very curious to see if this will work. Thanks for trying.

AKIMbO · 16 Jul 2012, 03:02

Wow, those turned out 1 million times better than the previous.

dorkvader · 16 Jul 2012, 03:34

I'd go over the upper right one (on the bottom PCB side) with a razor to make sure there's no continuity where there shouldn't be. Other than that, it looks really nice! Maybe make the pad slightly smaller, thus increasing the space between. It should still work just fine, and would probably be easier to print.

wcass · 16 Jul 2012, 03:41

SUCCESS!!!! maybe

i just could not stop myself from wiring up one today. The laptop screen is all "2222222222222222222"

: 2222222.JPG (114.56 KiB) Viewed 6963 times

Here is a close up of the controller and converter. The "row" and ground wires are soldered to the controller. the "column" wire is loose so i can change it easily. Even with poor alignment it still registers. [edit - it registers even without a hammer, so still have work to do.]

: switch.JPG (129.75 KiB) Viewed 6963 times

Ratfink · 16 Jul 2012, 04:40

Just a thought: perhaps the copper is the wrong thickness. That might have the potential to cause oversensitivity, but depending on how sensitive this thing is, there are many variables that could affect it. The wires getting close to each other between the two boards, maybe.

Also, I see you still have it connected to the original key part; connecting a second capacitor might make it think a key is pressed. This actually seems like the most likely problem to me, now that I think about it a bit more.

wcass · 16 Jul 2012, 05:48

Ratfink wrote:Also, I see you still have it connected to the original key part; connecting a second capacitor might make it think a key is pressed. This actually seems like the most likely problem to me, now that I think about it a bit more.

I thought the same thing so i chopped off the old key matrix and no effect.

this test was with 1 oz copper. i have a small piece of .5 oz and 2 oz.

JBert · 16 Jul 2012, 10:07

wcass wrote:Here is a close up of the controller and converter. The "row" and ground wires are soldered to the controller. the "column" wire is loose so i can change it easily. Even with poor alignment it still registers. [edit - it registers even without a hammer, so still have work to do.]

Have you tried to connect the extra pad before the controller is powered?

If I remember right, the first thing the controller does is calibrate the pads' sensitivity values in the assumption that no key is pressed. If you hook and unhook pads to different places in the matrix while the controller is active, the controller suddenly has to deal with different capacitance on a given crosspoint in the matrix and might think you're holding a key.

wcass · 16 Jul 2012, 16:47

JBert wrote:Have you tried to connect the extra pad before the controller is powered?

If I remember right, the first thing the controller does is calibrate the pads' sensitivity values in the assumption that no key is pressed. If you hook and unhook pads to different places in the matrix while the controller is active, the controller suddenly has to deal with different capacitance on a given crosspoint in the matrix and might think you're holding a key.

Thank you so much. This kind of insight is why this community is so valuable.
I'll tack the second pad to one of the pins when i get home and post the results.

Ratfink · 16 Jul 2012, 17:08

JBert wrote:If I remember right, the first thing the controller does is calibrate the pads' sensitivity values in the assumption that no key is pressed.

I'm not so sure about that. I just unplugged mine, pressed and held the F key, and plugged it back in, and it was spitting out F's like no tomorrow. I released the key and it stopped. Maybe it's averaging all the keys and using that for the baseline value, in which case having all the keys cut off would have the same effect as still having them connected. Just speculation; I could very well be wrong.

__red__ · 16 Jul 2012, 17:11

You can also send your gerbers to china and get 10 boards for 90c each. Predictable, repeatable etc...

Red

Ratfink · 16 Jul 2012, 17:21

__red__ wrote:You can also send your gerbers to china and get 10 boards for 90c each. Predictable, repeatable etc...

If you don't mind waiting a month to continue testing, that is.

Icarium · 16 Jul 2012, 20:17

JBert wrote:
wcass wrote:Here is a close up of the controller and converter. The "row" and ground wires are soldered to the controller. the "column" wire is loose so i can change it easily. Even with poor alignment it still registers. [edit - it registers even without a hammer, so still have work to do.]
Have you tried to connect the extra pad before the controller is powered?

If I remember right, the first thing the controller does is calibrate the pads' sensitivity values in the assumption that no key is pressed. If you hook and unhook pads to different places in the matrix while the controller is active, the controller suddenly has to deal with different capacitance on a given crosspoint in the matrix and might think you're holding a key.

From when/where do you remember that?

JBert · 16 Jul 2012, 20:53

There was some thread a while ago on GH where MSiegel wanted to rebuild a model F keyboard controller.

I believe I picked it up from one of the following IBM patents:
http://www.google.com/patents?id=vh8QAAAAEBAJ&pg=PA1
http://www.google.com/patents?id=kCo8AAAAEBAJ&pg=PA3

I must admit though that I've only skimmed the documents, as it seems it works still different because those documents talk about reference capacitors. Also, it's not because it got patented by IBM that it saw immediate use.

wcass · 17 Jul 2012, 03:41

i tacked the loose wire to column E, so along with row 1 i should be getting 4's. I booted up, opened notepad and plugged in the keyboard with no hammer on the test switch. I did not immediately get 4's, so i placed a hammer on the test switch and ... nothing.

the test switch registers without a hammer when connected after boot, but when connected before boot does not register even with a hammer.

so what does this tell us? am going to have to measure the capacitance of the original switch?

Icarium · 17 Jul 2012, 07:39

Hm...strange
on my controller there is a little silver, square chip labeled 8273565\nIBM 52\n185040L01 which seems to be responsible for checking whether keys are pressed or not. I don't think there is a start-up calibration process because I think it doesn't have enough wires going into it but I can't be sure. It seems to be connected to two capacitors however which might be used as a reference.