Bringing the IBM PC XT into the 21st Century

[wcass]

might the controller be using a "bridge" to measure capacitance? In that case, there would be a reference cap. the silver chip would generate an AC signal and adjust resistance to balance the bridge between the reference cap and the switch (unknown cap) and from the value of the resistance, we can calculate the capacitance of the switch.

C.unknown = C.known x (R1 / R2)

[Icarium]

I'm not following.

[Soarer]

The capacitance is given in one of the patents as: 0.4pF open, 4pF closed.

[wcass]

I don't know if a bridge is used to as the controller, but here is a short reference on a capacitance bridge:
http://conradhoffman.com/cap_bridge.pdf

it is a very simple circuit normally using pots for resistors so that you can very acurately dial in the value of an unknown capacitance. but we don't need 1% accuracy, so if we "pre-set" the resistors to be balanced at ... 2pF, then the meter would be able to say if the unknown capacitor (switch) is higher or lower (closed or open).

this is all just speculation as i suck at even basic electronics.

[bazemk1979]

I don't know if a bridge is used to as the controller, but here is a short reference on a capacitance bridge:
http://conradhoffman.com/cap_bridge.pdf

it is a very simple circuit normally using pots for resistors so that you can very acurately dial in the value of an unknown capacitance. but we don't need 1% accuracy, so if we "pre-set" the resistors to be balanced at ... 2pF, then the meter would be able to say if the unknown capacitor (switch) is higher or lower (closed or open).

this is all just speculation as i suck at even basic electronics.

you think you suck at basic electronics, what about all of us who keep our fingers crossed that if this modification does come up sucsessfull, hoping to only have to disassemble, maeby do slight adjustments and solder the new parts....

[wcass]

i just bought an El-Cheapo brand capacitor tester/meter on eBay. i'm hoping it will be here by this weekend. between now and then i will be making some more single test switches with .5oz, 1oz, and 2oz copper and drawing up a ten-key in KiCAD. i am on vacation all of next week - not going anywhere, just working on this project.

now that the project has an official name - XTant - let's finalize the key layout; this is a list of ideas discussed earlier in this same thread:
• uses the same case, controller, back plate, barrels, springs, hammers as the original IBM Personal Computer keyboard
• limited to the 83 matrix positions we know are supported by the controller
• converts to ISO or ANSI enter and left shift key (like the M)
• double-wide backspace can be split into two single keys (like the M)
• specific key assignment are 100% up to you
• uses keycaps that are readily available
• lock LEDs and backlighting are optional

the most likely barrel positions will allow all of the following (just needs a different keycap set):

Akimb02-ANSI.JPG (54.46 KiB) Viewed 8971 times

Akimb0-ISO.JPG (54.85 KiB) Viewed 8971 times

[Icarium]

I think we should go with a symmetric layout.

[AKIMbO]

I like "Akimbo2-ANSI"...but I may be a bit partial.

[wcass]

I like "Akimbo2-ANSI"...but I may be a bit partial.

All three configurations can be done with the same PCB and barrel plate. Only the Backspace, Enter and left Shift keycaps are different. You could even mix/match these - ISO enter with ANSI shift and split Backspace. I will go with the ANSI configuration myself, but will split the Backspace so that i can put my Esc in the tilde/grave spot and put the tilde/grave between Backspace and +/=

I think we should go with a symmetric layout.

If we did that then we would not be able to do an ISO Enter or an ANSI left shift. I think a symmetric Model F will come in time - and it will be easier because of the research we are doing right here.

[Ratfink]

One question about those layouts: how will the right shift key be stabilized? I count 83 keys on the last one, but wouldn't the right shift require an 84th barrel for stabilization (if we're using Unicomp right shift keys)?

That aside, I love the Akimbo-ISO layout. It can keep all the best parts of the original layout, while getting a hint of modernness.

[wcass]

Good eye, Ratfink. There will be a barrel there, but no pad underneath. This does mean that we will need extra barrels, but to me it seems to be the best option. The barrel frame will have at least some barrel holes with both the XT post and AT/122 notch so that either barrel type will fit.

[Ratfink]

Ah, okay, thanks. So will the kits contain any extra barrels needed, then? I don't have any keyboard I could take one out of at the moment.

[wcass]

Ah, okay, thanks. So will the kits contain any extra barrels needed, then? I don't have any keyboard I could take one out of at the moment.

probably. one keyboard would provide enough barrels for the entire group buy.

[msiegel]

You guys are awesome

I wish I had time to dive in to this too, but all I can do right now is cheer you on. Great work, guys!!

[dorkvader]

You could probably get the barrels made relatively cheaply.

I mean, who wouldn't want 5,000 extra new IBM model F barrels?

That said, one broken keyboard from ebay could provide a lot of spare barrels. Or people can steal one from another model F. Depending on price, it'd probably be $.50-.75/barrel.

[wcass]

The IBM board has much higher resting capacitance, but I think the problem with my switch is that there is not a big enough change with the hammer down. I will try making test switches with thinner PCB and heavier copper.

wcass PCB measured capacitance values:
1.6 pF without hammer, 2.6 pF with hammer

IBM PCB measured capacitance values:
Row 1 column I = 14.3 pF without, 16.8 pF with
Row 1 column J = 13.9 pF without, 16.9 pF with
Row 1 column K = 16.6 pF without, 19.2 pF with

cap_test_front.JPG (131.64 KiB) Viewed 8963 times

cap_test_back.JPG (126.54 KiB) Viewed 8963 times

[bazemk1979]

How is progress coming along wcass?

[wcass]

i hooked the original matrix to the controller and got results very similar to my test switch (does not register with hammer or registers after hammer removed). i may have damaged the controller when i cut the matrix off. i may break out another XT to test with.

i have a plastics fabrication company near where i live, so i sent them a DXF of the top plate. i will likely have them make a barrel frame in 1/16" clear acrylic to test my measurements for fit. i would also be able to mount LEDs in this to test back lighting.

i'm also working on trace routing.

[Icarium]

Hm...could be a couple of other things as well, those switches are quite picky. When I took mine apart I hooked it up to a computer and it was sending all kinds of keypresses just from me handling it.

[Parak]

One thing I've noted from my F122 is that the controller pretty much requires to be grounded to the metal backplate. If it's not, the capacitive pads (which are testable with a finger, even), won't work.

[Icarium]

Just the controller or could it be about the pads? We should really just find the engineer who built these...

[Soarer]

The backplate does act as a capacitor to ground for each of the underside pads. This has no primary effect on the theoretical operation, but serves to reduce parasitic capacitances and make it more stable.

[wcass]

the local plastic fabricators wanted $100 just to look at my barrel frame/top plate drawing, but i was able to order it from Pololu for $35 including shipping. i hope to have it in a week.

I feel much more confident with drafting, fabrication, and financing than electronics. i even think that i could DIY etch a full keyboard PCB, but not sure i have the skills to get it to work. this will not stop me from trying, but should anyone want to take over switch discovery and testing, i would be happy to ship you any parts you might need.

[AKIMbO]

I'm getting pretty excited about this. I hope the PCB/switch issue can get resolved...that seems to be the biggest hurdle with this project.

[bazemk1979]

wcass how are things coming along?

[wcass]

i ordered a barrel plate last week. it is laser cut acrylic (plexiglass) 1/16" thick. this will allow me to check to see how the barrels and keys fit in the case with the new layout. it will come flat; to put the curve in it i will cook it like in this video:
http://www.youtube.com/watch?v=s23lMNN1lyE

i will also mount one or two LEDs and see how that looks. i'm going for very subtle (invisible in a well lit room) outline around the key groups.

i opened up another XT to work with a virgin controller. i tested with just the adapter/controller/matrix/backplate and a single hammer. tap the hammer on a key and it registers the one key. then i soldered one of my test switches to the controller (parallel with the switch matrix) and tried it out; keyboard erred. i removed the test switch and tested again and it is still throwing garbage. so the controller is very temperamental and does not recover well. without knowing just how the controller works, i'm at a loss. again, should anyone feel up to the challenge and want to figure out, i would be happy to lend you support and give you all the credit.

i want to use capacitive switches, but may have to switch over to M type switches. if i have to use a membrane, i can make it NKRO by making the bottom sheet a PCB and mount diodes at the top of each column, away from the hammers. i found a diode array equal to 5 discrete diodes; very small, and pretty cheap.
http://www.digikey.com/product-detail/e ... ND/1964793

[wcass]

the test acrylic barrel plate is in. it is still flat and covered with protective paper, but i could not resist checking it out with barrels and keys. i think it looks pretty good.

fit-test-1.JPG (115.95 KiB) Viewed 8714 times

fit-test-2.JPG (118.08 KiB) Viewed 8714 times

fit-test-3.JPG (109.9 KiB) Viewed 8714 times

fit-test-4.JPG (109.74 KiB) Viewed 8714 times

fit-test-5.JPG (105 KiB) Viewed 8714 times

fit-test-6.JPG (103.66 KiB) Viewed 8714 times

it is too tight across the top - but that is as expected (after the curve is added it should be better). i only have four 1.25 wide keys, so must keep moving them around to get a feel for the fit. i think i need to order a set of keys.

I also picked up another XT off eBay, so i will be able to offer the needed barrels with the kits.

[bazemk1979]

awesome wcass!

[wcass]

I took out the Dremel and removed the two metal PCB retainer studs from an XT back plate. This so i could use it as a mold to curve the new acrylic top plate. I stripped the paper off and cooked it in the oven for 5 minutes sandwiched between the back of the back plate and the top of the top plate, then let it cool for about 10 minutes.

The barrels fit tighter than they did when the sheet was flat.

fit-test-07.JPG (112.49 KiB) Viewed 8687 times

And you can't really tell what the keys will look and feel like unless the springs and hammers are in there, right? I'm testing both M and F type hammer/springs. Left side of keyboard is M hammer/springs, right side is F hammer/springs.

fit-test-08.JPG (104.04 KiB) Viewed 8687 times

Gaps at the top, bottom and sides looks good on the left.

fit-test-09.JPG (98.62 KiB) Viewed 8687 times

fit-test-10.JPG (101.34 KiB) Viewed 8687 times

The alignment looks good here too. With the keys off you can see the barrels and the backplate through the clear acrylic.

fit-test-11.JPG (98.03 KiB) Viewed 8687 times

The right side of the navigation cluster could move about .1 inch to the right. this might be a scaling error; 99.4% would not be noticed between adjacent keys, but from left side to right would be .1 inch small. i will have to measure to know for sure.

fit-test-12.JPG (104.45 KiB) Viewed 8687 times

Yes, i have typed on this. Both M and F sides function normally. I will talk about key feel in a future post.

[andrewjoy]

I can see this working well, people will be able to use custom coloured acrylic will look nice.

Hope that this one make it feel more like a model M"ushy"