XTant - Bringing the IBM XT into the 21st Century: Part 2

[wcass]

During the dark age here at GeekHack (July 2012) like many of the displaced, I ended up at Deskthority. There, I put forth the idea of replacing just the matrix, pad, and top metal plate of an XT - re-using an unmodified case, barrels, springs, keys, and controller - then using a Soarer's Converter for USB connectivity and key mapping. I made lots of progress; decided on a layout, name, designs, and then hit a problem that could not be overcome. I found that once I removed the controller I could not re-attach it and get it working again reliably. Capacitive controllers are very sensitive things - completely different from that of a conductive switch controller.

But there have been some recent developments that make me sure that a replacement F controller is imminent. For one thing, xwhatsit made a beam spring controller and open-sourced it. For those that don't know, beam spring switches are very similar to F switches (see here and here), so the controller should be almost the same. Then there was that stockpile of Kishsavers that was found.

So rather than revive a long dead thread over on Deskthority, I'll be posting progress and updates here. If you haven't read part one, you can find it here. The plan now is to replace the pad, top plate, and PCB including the controller. You will need to drill 10 holes in the donor bottom plate but there will be a template for that.

Just to let you all know i haven't been completely idle the last year, I offer you this:
new 1mm thick neoprene pad:


new steel top plate (thanks the_beast):


PCB design (the matrix anyway):
 XTant PCB.pdf (63.99 kB - downloaded 581 times.)

PS: Remember where I said "Capacitive controllers are ... completely different?" If you would like to know HOW different, this one post by xwhatsit explains it very well. If you want to get your head blown off completely, read the whole discussion between xwhatsit and dfj.

[bcg]

nice 

[fohat.digs]

Awesome work.

Is that plate flat? What does it use for a PCB?

[0100010]

OST.   

[mkawa]

capacitive controllers are tricky because they're so analog, basically. scanning a diode and switch matrix is a simple matter of pulling your matrix up and pulling out the binary values of the switches. capacitors are fundamentally analog little things, and the capacitance you can get out of a keyswitch is absolutely tiny. further, measuring capacitance is very very tricky because they are charge wells. they store electrons and then eject those electrons when they "fill up". so, to sense a capacitance, you have to insert electrons, then pull them out, and in the case of a keyswitch, there are very few electrons going in and then coming out. it's quite difficult to perform this with modern microcontrollers (analog computing went the way of the dodo, i'm afraid).

so, xwhatsit's controller is pretty cool, but it's also not as clean as he wants it to be, or possibly other people for new modern keyboard designs.

there is another modern application that deals with scanning lots of charge wells, but unfortunately it points out the problem with minimizing the circuit. image sensors are basically the same thing. each pixel device is a photon-collecting charge well. basically, every photon that hits the device causes some number of electrons to accumulate in the device. to pull an image out of this, you have to very quickly sense the number of electrons in each pixel element (all 24+ million of them in modern sensors) and output that as a discrete value. however, therein lies the rub. the modern solution to this is dedicated chips that do nothing but hold a whole crapload of pixel elements and very fast scanning and charge sensing. further, image data is very lossy. you're collecting some number of photons in each charge well, and it's very difficult to determine exactly how many for quantum reasons, so if you make a bit of a mistake somewhere, heck, everywhere; it's not that big a deal, as long as the mistakes are properly random (which is turns out they almost always are).

[wcass]

Is that plate flat? What does it use for a PCB?

Yep, flat - at the moment. The right tool for putting a curve in it is a slip roll. A local sheet metal shops should have one. A less elegant option would be to use a bending break. Harbor Freight has an 18" one for under $50. The PCB will be new. Matrix and controller will be together just like the original.

There is a PDF attachment under the pictures in the first post that shows the matrix - that's as far as I can get until a controller goes public. The first page is the bottom copper layer of the PCB; the second page is the top copper layer of the PCB; the third page overlays the two. This layout copies some of the design "features" of the original matrix that I don't completely understand, but must assume the designers put in there for good reason. There is no ground plane fill on the PCB, only a ground trace around the columns on the back. I can only guess that less would allow too much noise but more would kill the signal. Also, all row traces cross all column traces exactly one time (usually at right angle) but for the most part keep as far away from each other as possible. You can see that this requires adding "tails" to trace routing on the original XT PCB most notably on the bottom and left side.

[mkawa]

the harbor freight brake is crap (naturally). slip rolls are getting somewhat rare, i've found. you will probably need the plate stamped before it gets rolled though right? i'm like 90% sure that IMS has a slip roll, and will do fairly short runs. for the stamping you need a die machined. that's kind of a pain.

[CPTBadAss]

I'm so stoked for this wcass. I really enjoyed typing on this keyboard when I tried it at Keycon. Read through that whole XTant Part 1 thread and I'm excited for this thread ^__^

[wcass]

the harbor freight brake is crap (naturally). slip rolls are getting somewhat rare, i've found. you will probably need the plate stamped before it gets rolled though right? i'm like 90% sure that IMS has a slip roll, and will do fairly short runs. for the stamping you need a die machined. that's kind of a pain.

I don't think it needs to be stamped at all. I'm locking the top to the bottom plate using bolts instead of tabs, so i don't need to turn the edges down. I'll use 10 bolts very near where the 10 tabs were, to it should maintain the feel.

[mkawa]

are you planning on drilling out the bolt-holes? tricky to get the geometry right prior to slipping. stamping is the preferred solution, but for short runs a drill press will work fine.

[fohat.digs]

I'm locking the top to the bottom plate using bolts instead of tabs, so i don't need to turn the edges down. I'll use 10 bolts very near where the 10 tabs were, to it should maintain the feel.

Does this mean that you are pulling the plates into curvature by torquing down bolts, or will it stay flat?

[wcass]

The bolt holes on the top were laser cut with the barrel holes. The back plate will require a hand drill, but i plan on including a paper template and instructions on getting those holes drilled within allowed tolerance.
 
The top plate will be curved using a tool like this:
http://www.youtube.com/watch?v=O6dSihaHW_M
 

[mkawa]

why not have the_beast's guy laser out the bolt holes then send it over to the slip roll shop? this would be basically plug and play, and his rates are extremely reasonable for 16ga 304

[wcass]

why not have the_beast's guy laser out the bolt holes then send it over to the slip roll shop? this would be basically plug and play, and his rates are extremely reasonable for 16ga 304

That is exactly what i did. This plate is 18 gauge CRS. i thought it would be easier to roll and take paint better, but stainless would be an option. I'm just trying to find a local shop that will roll it for me. South Florida is full of HVAC shops. I wish i had more time during the day to call them all.

[wcass]

The top plate went through a slip roller at Merlin Industries. Thanks David! It took all of about 1 minute and 50 seconds of that was setting up the machine.

I'm planning on using the DPH controller for this. The guys developing this controller want it to work in all of the IBM model F boards that use a 30 pin ribbon cable to connect - 122, KishSaver, UnSaver, and AT. So all i have to do is add a 30 pin connector and get HaaTa to include it as an option. To that end, the XLS file has the "default" key assignments.

I neglected to embed the font used in the PDF attached to the OP. It is supposed to look 8-bit retro like this.

[The_Beast]

Very cool! Glad you got it rolled out!

[mkawa]

why not have the_beast's guy laser out the bolt holes then send it over to the slip roll shop? this would be basically plug and play, and his rates are extremely reasonable for 16ga 304

That is exactly what i did. This plate is 18 gauge CRS. i thought it would be easier to roll and take paint better, but stainless would be an option. I'm just trying to find a local shop that will roll it for me. South Florida is full of HVAC shops. I wish i had more time during the day to call them all.

my understanding is that cold-roll's machinability is due to the fact that it's yielded once. for strength you'd want to slip roll hot-roll sheet. this would effectively give you cold-roll, but it could also distort the lasering you'd doing prior to rolling.

stainless is non-ferritic and i thought i knew what that meant but just spent the last hour learning what it actually means, which is damned fascinating crystallization metallurgy. for what it's worth though, powder paint is basically a pigmented and potentially filled polyster substrate. the polyester can be metal filled, cut with another polymer, or what have you. however, the paint never bonds with the metal (unlike anodization, which is the formation of a pigmented oxide layer), so it doesn't really matter that much what metal you start with, as long as it is inert up to about 3-400F and can be abraded with a reasonable material (Si carbide usually) to increase surface area and promote granularization

[Aer Fixus]

The top plate went through a slip roller at Merlin Industries. Thanks David! It took all of about 1 minute and 50 seconds of that was setting up the machine.

I'm planning on using the DPH controller for this. The guys developing this controller want it to work in all of the IBM model F boards that use a 30 pin ribbon cable to connect - 122, KishSaver, UnSaver, and AT. So all i have to do is add a 30 pin connector and get HaaTa to include it as an option. To that end, the XLS file has the "default" key assignments.

I neglected to embed the font used in the PDF attached to the OP. It is supposed to look 8-bit retro like this.
(Attachment Link)
(Attachment Link)
(Attachment Link)

Looks very good. If ti was possible, I would love a few different models, though. If there was one that could fit into existing/modified Model M cases would be one of my dream boards. If extra models are something that are possible, we could create Kishsaver reproductions and Model F Ergodoxes.

Also, how much extra work would this be? Would this be viable? I'm not sure how long it takes to design one of these PCBs and how production costs would be effected adding more things to create.

[wcass]

... I would love a few different models, though. If there was one that could fit into existing/modified Model M cases would be one of my dream boards. If extra models are something that are possible, we could create Kishsaver reproductions and Model F Ergodoxes.

Also, how much extra work would this be? Would this be viable? I'm not sure how long it takes to design one of these PCBs and how production costs would be effected adding more things to create.

That is my thoughts too. Designing a PCB for this type switch should be much easier than other types of switches because there are no switch contacts - just a couple of geometric shapes on the PCB and lining up the barrel plate just right. But it does come at a cost; sacrificing an XT or worse an AT or 122 for parts. If this works as I hope, we might be able to re-use the case, frame, and back plate of any model M, but we must find a new source of capacitive flippies.

[Aer Fixus]

... I would love a few different models, though. If there was one that could fit into existing/modified Model M cases would be one of my dream boards. If extra models are something that are possible, we could create Kishsaver reproductions and Model F Ergodoxes.

Also, how much extra work would this be? Would this be viable? I'm not sure how long it takes to design one of these PCBs and how production costs would be effected adding more things to create.

That is my thoughts too. Designing a PCB for this type switch should be much easier than other types of switches because there are no switch contacts - just a couple of geometric shapes on the PCB and lining up the barrel plate just right. But it does come at a cost; sacrificing an XT or worse an AT or 122 for parts. If this works as I hope, we might be able to re-use the case, frame, and back plate of any model M, but we must find a new source of capacitive flippies.

In theory, those flip pads are just plastic mixed with some sort of capacitive material. Those could probably be made by a company somewhere. Same thing with the barrels. They can be taken from an existing board, or outsourced to a company with resources beyond what we have so we don't suddenly have XTs selling for $150 because their valuable parts. I expect a lot of people would be interested in this when it comes around so having a company make multiple thousands of these made and ordering springs from Unicomp wouldn't be that big of an issue.

[fohat.digs]

so we don't suddenly have XTs selling for $150 because their valuable parts.

That would be a good use for XTs, in my opinion. If you needed to harvest the barrels (aka chimneys) then remember that the alignment tabs are different on the ATs and 122s.

[mkawa]

they're almost certainly just conductive plastic. if we need to make a small tool we might be able to do so if we loosen tolerances (we will have dimensional and capacitive tolerances that are a function of injection density)

[JPG]

Can't wait to see where this goes. Already dreaming about a custom 60% (and other nice layouts) F keyboard.

[mougrim]

Any news?

[berserkfan]

Any news?

Mougrim!

I was wondering who necrobumped this important but esoteric thread and am pleased to see its you. So you're still alive! If Western propaganda is to believed most of Eastern Ukraine is now a wasteland of starving people!

[Harrowed]

Any news?

This, I have XT coming on my way and I don't want it to end up as just a collection piece.

[wcass]

Top plate is done.
Controller is done (use DPH or xwhatsit)
I just finished making gerber files for a PCB and they are getting peer reviewed.
Then I'll order one (or more) to be professionally made. Anyone want to to buy 82 inches² of an untested design?

[berserkfan]

Top plate is done.
Controller is done (use DPH or xwhatsit)
I just finished making gerber files for a PCB and they are getting peer reviewed.
Then I'll order one (or more) to be professionally made. Anyone want to to buy 82 inches² of an untested design?

Given my inability to even put one Model F back together I am not going to buy anything until everything is worked out. However I am quite keen on providing some funding to anyone involved in Model F revival. If you give me a pp address I'll tell the people who buy from me to gift to that address instead.

Since I rarely sell keyboards or expensive keycaps it won't be a whole load of money. But maybe 10 bucks here and 30 there and I think I can do my part.

[mougrim]

Any news?

Mougrim!

I was wondering who necrobumped this important but esoteric thread and am pleased to see its you. So you're still alive! If Western propaganda is to believed most of Eastern Ukraine is now a wasteland of starving people!

Things ain't that bad yet. Anysay I'm planning to move to the capital

Besides, I wanted to know what happened to Xtant

[wcass]

Progress pictures.



I was not able to bend down the leading and trailing edge on my custom top plate to create the tabs that normally keep Model F key assemblies together. The next two pictures show my method of keeping the top and bottom plates held tightly; binder clips. They are very low profile if you remove the "handles" (easy to take off and put back on). You can get a box of these for cheap at any office supply store. They probably will break about as often as Model M rivets, but would be a lot easier to replace.


The binder clips alone keep the plates very tight. There is no wobble in the barrel at all. I'll use two bolts (right side of caps lock and left side of of backslash) to align the plates east/west/north/south.

[Techno Trousers]

Woah! Where did you buy, and how did you label those custom 1.25x navigation and arrow keys? Or were you able to convince Unicomp to dye sub them for you?

That is completely awesome.

[dorkvader]

Woah! Where did you buy, and how did you label those custom 1.25x navigation and arrow keys? Or were you able to convince Unicomp to dye sub them for you?

That is completely awesome.

Unicomp will do that. FN is new as well.

I also notice the backspace looks a little different.

[wcass]

Yes, the guys at Unicomp can be very accommodating. But it is a bit of a hassle for them because they must manually modify a dye-sub template to go on a different size key cap. If you want custom caps, it helps to ...

• Ask politely and know how to take "No" for an answer.
• Ask for it to be added to a larger order. This helps balance the higher labor cost of the special order.
• Do not complain if it comes out less than perfect. Special requests are manual and have a higher chance of misprint. It might take 2 or 3 tries to get one cap with only a small imperfection.

When the XTant layout was finalized, I ask if they could do this set and that there might be requests for more (I sent a picture and a link to the project on Deskthority). This specific set was purchased as product code CPSET ("Custom Printed Buckling Spring Set"). It cost me a little more than a standard printed set and I got fewer caps, but they still probably lost money on it.

And that backspace is pearl instead of pebble. Backspace and NumPad0 are the same size, so that cap was easy.

[Techno Trousers]

That's great. I've asked for a few truly custom printed caps from them and was always shot down. When XTant is ready, maybe we can do a big group buy order from them to make it worth their while.

[mougrim]

Progress pictures.

(Attachment Link)

I was not able to bend down the leading and trailing edge on my custom top plate to create the tabs that normally keep Model F key assemblies together. The next two pictures show my method of keeping the top and bottom plates held tightly; binder clips. They are very low profile if you remove the "handles" (easy to take off and put back on). You can get a box of these for cheap at any office supply store. They probably will break about as often as Model M rivets, but would be a lot easier to replace.
(Attachment Link) (Attachment Link)

The binder clips alone keep the plates very tight. There is no wobble in the barrel at all. I'll use two bolts (right side of caps lock and left side of of backslash) to align the plates east/west/north/south.

As I understand, F1-F12 buttons are accessible through fn layer, but how exactly?
 

[mougrim]

That's great. I've asked for a few truly custom printed caps from them and was always shot down. When XTant is ready, maybe we can do a big group buy order from them to make it worth their while.

Hmm... Anyone knows where we can order custom cases? I mean, why stop halfway?

[wcass]

That's great. I've asked for a few truly custom printed caps from them and was always shot down. When XTant is ready, maybe we can do a big group buy order from them to make it worth their while.

There is a GH member that is doing custom dye sub; nkzt. You might ask him for something unique.

As I understand, F1-F12 buttons are accessible through fn layer, but how exactly?

I use a two key combo. My Fn key is in the traditional Caps Lock location (Fn+shift = Caps lock). Fn+1 = F1 ... Fn+0 = F10 ... Fn+= = F12
Unicomp even has special key caps for folks that like this
http://pckeyboard.com/page/SBLE/AltFunNumKey

But, because the controller is fully programmable, you can put any key anywhere you want. Have the F keys to the left if you prefer (but you only have 10 spots there).

[dorkvader]

That's great. I've asked for a few truly custom printed caps from them and was always shot down. When XTant is ready, maybe we can do a big group buy order from them to make it worth their while.

Hmm... Anyone knows where we can order custom cases? I mean, why stop halfway?

Gotta get some sick CNC cases!


I recommend cutting down a shift key to make an "unstepped" function key for capslock.

[mougrim]

That's great. I've asked for a few truly custom printed caps from them and was always shot down. When XTant is ready, maybe we can do a big group buy order from them to make it worth their while.

Hmm... Anyone knows where we can order custom cases? I mean, why stop halfway?

Gotta get some sick CNC cases!


I recommend cutting down a shift key to make an "unstepped" function key for capslock.

Yeah, CNC could be nice!

Why cut shift key if you could get unstepped one from Unicomp?

[dorkvader]

Why cut shift key if you could get unstepped one from Unicomp?

You can't. This is the closest they have: Note the CTRL key is still stepped
http://pckeyboard.com/page/Buttons/CTCL

[mougrim]

Why cut shift key if you could get unstepped one from Unicomp?

You can't. This is the closest they have: Note the CTRL key is still stepped
http://pckeyboard.com/page/Buttons/CTCL

Oh, they have non-stepped Caps Lock! I totally should get it for my Model F - as for now I use extra Ctrl key in place of Caps Lock.

[dorkvader]

Why cut shift key if you could get unstepped one from Unicomp?

You can't. This is the closest they have: Note the CTRL key is still stepped
http://pckeyboard.com/page/Buttons/CTCL

Oh, they have non-stepped Caps Lock! I totally should get it for my Model F - as for now I use extra Ctrl key in place of Caps Lock.

No they don't have a non-stepped capslock. That's why I advocate cutting up another key. The capslock in that image is the same key as "ctrl" normally is" and will not fit where capslock is on the keyboard (CTRL is center stemmed)

The only key that will fit there is a very rare japanese "shift" key.

[wcass]

I thought it was a Portuguese right shift, but anyway dorkvader is correct; non-stepped 1.75 BS caps are unobtainium.

Fun fact: if you could perfectly split a buckling spring cap down the center of the stem; you could make a perfect non-stepped 1.75 by joining the right side of a right-shift with the left side of any 1.5 cap.

[dorkvader]

I thought it was a Portuguese right shift, but anyway dorkvader is correct; non-stepped 1.75 BS caps are unobtainium.

Fun fact: if you could perfectly split a buckling spring cap down the center of the stem; you could make a perfect non-stepped 1.75 by joining the right side of a right-shift with the left side of any 1.5 cap.

You could get it on some international layouts, (I guess Portuguese is one?) but the one in particular I was thinking of was this:
http://www.kbdmania.net/xe/?mid=photo&search_target=title&search_keyword=quest&document_srl=1874062

Which is the right shift on a 5576-c01 keyboard (that unicomp uses the case / moulds from as their endurapro first and now all their keyboards.) I really want one for the convex modifiers.

I think the reason it works is that IBM decided making a dedicated keycap for that location was cheaper than making a new membrane for just those boards.

More info on the mod is available here

[mougrim]

Why cut shift key if you could get unstepped one from Unicomp?

You can't. This is the closest they have: Note the CTRL key is still stepped
http://pckeyboard.com/page/Buttons/CTCL

Oh, they have non-stepped Caps Lock! I totally should get it for my Model F - as for now I use extra Ctrl key in place of Caps Lock.

No they don't have a non-stepped capslock. That's why I advocate cutting up another key. The capslock in that image is the same key as "ctrl" normally is" and will not fit where capslock is on the keyboard (CTRL is center stemmed)

The only key that will fit there is a very rare japanese "shift" key.

On AT Model F Caps Lock IS center-stemmed and unicomp one will fit on mine But yes, it's no good for XT

[dorkvader]

On AT Model F Caps Lock IS center-stemmed and unicomp one will fit on mine But yes, it's no good for XT

Also on 3179 "model C" and PC/XT. The issue is they aren't "standard" layout at all. I think you would need a different PCB for centered capslock like those keyboards have.

As you should know, on the AT model F, the key-where-capslock-is-on-today's-keyboards (labelled "CTRL") is stepped just like on the modern IBM "enhanced" keyboard and is not centered. The key labelled "caps lock" is in the right control position and is center stemmed and stepped on both sides. Neither one helps us here.

I even went and got my AT model F out of it's box to double check.

[wcass]

OK, this is officially now a done thing. It works.

I will be refining some steps and am waiting for a few more parts, but i plan to offer "kits" in the Marketplace before too long.

I have mine working with xwhatsit's IBM Capsense USB controller - you can see the controller in a post from me about 20 hours ago in his topic. The DPH capsense controller should also work, but has not yet been confirmed. Both controllers are intended as replacement controllers for IBM boards, so the wire and mount holes on both are where IBM put them. My matrix board is designed to fit directly over the holes on the controller. Just feed wire through both boards and solder on both sides.

One of the final things for me to decide is how to hold the space bar stabilizer against the barrel frame. I'm thinking of using a stamping tool to bend a small tab cut into the barrel frame. This would be much more durable than the solution I used for an IBM AT keyboard space bar conversion; glued on tabs. However, I don't have access to tool design/simulation software. So maybe i'll have them welded or brazed on.

[geniekid]

*applause*

[JPG]

Very nice!

[esoomenona]

.