Lots of TI-99/4A Keyboards

[Eudimorphodon]

... Probably off topic here, I spent the weekend paralyzed by indecision how to proceed with building out something to test one. I did a lot of reading about the protocol spoken by the Datanetics strobed parallel keyboards as used by the Apple I/II (and a bunch of other early computers/terminals/TV Typewriters), torn whether to use an Arduino Nano, a Blue Pill (I bought a bag of those *years* ago and never ended up using them for anything), or a bare AVR 324 to scan it with. (The last on that list has enough I/O I might not need any support chips, the Nano would need a little help, not sure about the Blue Pill, it's *close*.) It'd still be nice if these had more keys, but they should make a decent if slightly-awkward ASCII parallel keyboard.

 

[acgs]

These keyboards do not have a membrane. Here is a picture showing the many variations of keyboards they went through over the course of the TI 99/4A's lifetime, the membrane keyboard was made by Mitsumi, these are Hi-Teks. These Hi-Teks have the same "wiping brush" mechanism as any other HI-TEK keyboard

Ah. They looked like Mitsumi's to me. I knew you were looking to replace a Hi-Tek, but I didn't realize you knew these were the same brand.

Although if I'm going to be completely fair, this board also has some improvements over the TRS-80 version; for instance, it has a molded upper plate that would *probably* make it a little more resilient to dust or spills verses the open top grid of the older board.

Hopefully that's a plus that will be matched with a serviceable quality, even if they are thinner.

Both of those are from the early units that cost much more than the tan machines that these boards were for, and include nicities like metal mounting plates that, yes, these Hi-Teks don't have. On the flip side, the sturdy waffle-iron grid of plastic probably means the HI-TEKs don't really *need* the metal parts.

The Mitsumi's were for the tan end of the product line too. I had thought they were the only design being used by that point. But then why were all of these Hi-Tek's still kicking around unused? I guess TI was scrambling and didn't know what to do before everything ended.

 

[Eudimorphodon]

I knew you were looking to replace a Hi-Tek, but I didn't realize you knew these were the same brand.

FWIW, I wasn’t really looking to “replace” anything, I ordered some just to experiment with, but maybe with a side mission of seeing if there was any possibly of using parts to repair other Hi-Teks. (The Model I’s keyboard is very, very prone to the stems splitting at the corners. I’m not sure how motivated I am to actually tear one to bits, though; I see springs and tiny parts going *everywhere*…)

But then why were all of these Hi-Tek's still kicking around unused? I guess TI was scrambling and didn't know what to do before everything ended.

I’m curious what these actually cost when new. The lore is that even after the cost reductions a TI 99/4A was still more expensive to make than a VIC-20 or C64, but I’ve never seen an itemized breakdown. These keyboards, mainly for having fewer keys, probably have significantly less plastic than a VIC-20 keyboard, but all those little metal wipers seem like they’d be more expensive than Commodore’s conductive pads…

Anyway. After work I decided to see if I could scan one of these. I grabbed a dusty Arduino Nano and wired it like this:


The 8 bit columns are one side, the 7 rows are on the other. I used this for the pinout: http://www.mainbyte.com/ti99/keyboard/keyboard.html

As that reference shows the pinout freely scrambles the rows and columns, which is annoying. I ignored the bit numbering and just wired mine as linearly as possible so the matrix is going to read differently. Only just started on the driver software, my Arduino C++ is rusty and I have no interest in letting AI do the work for me, but what I have so far works: it scans all seven rows and outputs over serial an 8 bit byte for each row that has keys held down. My plan is to have it prompt me to press every unshifted key in sequence and automatically assemble the 6x8 matrix corresponding with its current wiring. (The 7th row only has the caps lock on it.)

 

[Hak Foo]

Right now, a generic 'mechanical' keyboard will sell for about 30USD on up, and the comparable cheapest rubber-dome-membrane boards will be like 10USD. So we can say in the worst-case scenario, quantity of 1, the premium would be 20 USD in today's dollars.

But I'd argue it was probably even smaller, as the Commodore board is better built than today's KWJIBO-brand rubber-dome. It has distinct springs for each key, the caps are higher profile, and you have a hard PCB and the lock switch.

The interesting factors might be scaling-related.
* Commodore was ordering a bazillion of the same board for the VIC and 64, and weren't they basically all from Mitsumi? TI ordered fewer boards and shopped them to multiple vendors, each with a discrete design. Did that push their unit costs up?
* For the TI Hi-Tek/Stackpole design, there was less bespoke tooling to pay down-- it didn't require a distinct "mounting plate" for the key switches to slot into.

 

[Eudimorphodon]

Commodore was ordering a bazillion of the same board for the VIC and 64, and weren't they basically all from Mitsumi? TI ordered fewer boards and shopped them to multiple vendors, each with a discrete design. Did that push their unit costs up?

Yeah, I just don't know how the costs would actually compare. Commodore during the PET era was kind of notorious for contracting out parts to different OEMs so if you get a PET with a model number like "2001n" or "4032" it might be a pretty random collection of case, motherboard, keyboard, and sticker variants. I know there were several different versions of the keyboards during the VIC-20 and C64's run, but I don't know if they were ever manufactured in parallel or it was strictly a sequential progression, and how many foundries actually turned them out.

More broadly speaking, I wonder if anyone has any numbers on how many of the Beige all-plastic 99/4As were actually made/sold compared to the black and metal ones. The latter seem to be much more common now, at least. Did TI ever end up bulldozing heaps of them into the desert, ala Atari?

 

[MicroCoreLabs]

Maybe they cleared out all of their inventory when they sold for $99. I was one of the beneficiaries...

 

[Ksarul]

The beige consoles were only sold for the last year or so of production, and for about half of that time they were already starting to wind production down. They also came in two different motherboard varieties: The earliest ones had the same motherboard as the Black and Silver consoles (copper fingers at the side port) and later ones that used a cost-reduced version of the motherboard (silver-colored fingers at the side port). Internal power supply switches are also a somewhat useful indicator--the older ones have a blue line to indicate they are in the ON position, and the later ones have an orange line instead. The two types tend to track the motherboard types for the most part but not always). The keyboards on the earlier style are mostly Hi-Tek, Futaba, or Alps. Later ones tended to have more of the Mitsumi keyboards. Towards the very end, parts were a mix of whatever happened to be in stock that day. The other differences are in the GROMs. Earlier boards almost always have the older GROM zero that allows third-party cartridges to run, whereas the later ones usually have the newer (v2.2) GROM zero that requires all cartridges to have a GROM to start. Note that at least one cartridge from Funware had a licensed GROM, and Parker Brothers cartridges used their own work-alike chips (PCROMs) to get around the v2.2 issue. Again, the motherboards at the end of production had a somewhat crazy mix of GROM zero chips, a condition even further muddied by the fact that TI would sell you either GROM zero for several years after production stopped, so people were regularly replacing v2.2 GROM zero chips on their own (that was a lot cheaper than buying a GROM Buster device from CorComp or Navarone).

On keyboards, TI ordered them from a lot of sources over a very short period of time. I'm not even sure the Mitsumi's were actually the last ones ordered either, as there are a small number of Black Mitsumi keyboards out there as well (with early 1983 assembly dates). It may be that they were ordered earlier but delivered later in the assembly process, or that different assembly locations were receiving different keyboards. My serial number database isn't tracking that data, although it might not be a bad thing to add it where I can get it.

 

[Eudimorphodon]

Apparently I’ve decided to try to force myself to make *something* interesting out of these because I bought another batch. Remember that weird listing mentioned earlier where you could buy one new hi-tek and get a molested Stackpole for “free”? The seller listed a six+six combo for $33:

https://www.ebay.com/itm/327319587991

My curiosity overcame my common sense, I want to see what the Stackpoles are like, and it’s kind of hard to beat about $4 a keyboard.

If anyone wants, I dunno, an ASCII parallel encoded keyboard or weird 6502 or Z80 kit computer in a few months I guess that’s what I’m going to have to do to justify this.

 

[Hak Foo]

You could probably also set up your favourite $3 MCU with QMK or similar firmware and turn them into minimalistic USB keyboards. There are a lot of ridiculously small layout options out there these days, the 99 board isn't that much worse than some of them.

 

[Eudimorphodon]

You could probably also set up your favourite $3 MCU with QMK or similar firmware and turn them into minimalistic USB keyboards. There are a lot of ridiculously small layout options out there these days, the 99 board isn't that much worse than some of them.

Yeah, I was actually considering programming one to do that for a laugh (I have some Pi Picos not otherwise occupied), the main bummer is there's just a paucity of modifier keys. The shift keys aren't independently addressable, so all you get is shift, ctrl, and fctn. I guess you could sabotage the lock in the alpha lock key to make it an alt, but there's also no dedicated backspace/delete. You're getting *this* close to a chording keyboard at that point, and these don't have diodes to prevent phantom keys so once you're asking for more than a pair of keys to be held down at once... there there be dragons.

They should make pretty good ASCII-encoded keyboards for Apple I clones, CP/M machines, and TV Typewriters, but of course there's a lot less call for that...

 

[MicroCoreLabs]

Mount two of the TI keyboards side by side, rearrange some of the keycaps, and rename the rest to make a Frankenstein XT keyboard.

Extra credit for having three or four keyboards to use as macro's for common commands such as DIR, FORMAT, etc... To make life interesting make every other macro key "DEL *.*" or "CTL-ALT-DEL" to encourage the user to be careful with keypresses. One key could be a random "Surprise me" macro... maybe a random compliment - or insult - or joke.

 

[acgs]

Mount two of the TI keyboards side by side, rearrange some of the keycaps, and rename the rest to make a Frankenstein XT keyboard.

I never realized before that old surplus TI keyboards were used to make Microsoft's "Natural" keyboards.
That explains a lot...

 

[Eudimorphodon]

I don't know if anyone else bought any of these to play with yet, I'd be be curious what you're planning to use them for if you have. Anyway...

I stuck a working prototype of a keyboard scan routine that goes with the wiring scheme I took a picture of earlier (with one small change) up on GIT.

The code is a trash fire in terms of efficiency (the basic digitalRead/Write routines in Arduino are notoriously inefficient), but it's still able to scan the board plenty fast. (There's no explicit debounce but because of the imposed 1ms scanning interval it seems like that's not necessary.) The wiring scheme doesn't try to copy how a TI-99 sees the matrix, what this does is just start at pin #1 on the keyboard plug and wire the rows and columns linearly.

(I'll update with a diagram if anyone actually wants to try this. Basically put the 8 data columns on pins D2-D9, the 7 rows are on D10+A0-A5. You probably have it right have if the Alpha Lock is on D10 and the other six row lines in the main 8x6 matrix are on A0-A5.)

You can type on the keyboard and it comes out through the standard Arduino software serial port a 9600 baud. Converting to a parallel ASCII encoding should be trivial, but an Arduino Nano doesn't have enough pins internally so it'll need an output shift register or something. The code implements repeat if a key is held down for about a second; if multiple keys are held down simultaneously it'll output them all, subject of course to crosstalk limitations.

 

[Eudimorphodon]

FWIW, Here is an updated prototype Arduino sketch that is the result of reading the manual for an old Jameco keyboard kit that another user on this forum has connected to their Poly-88 computer. It implements what I think should be authentic 2-key rollover like a real (cheap) 1970's keyboard would do, and appears plenty performant for a real-world application. (I don't think replacing the serial output with output latch/strobe generation will be a problem.)

Probably won't work on this much more before I put together/repair something I can actually use an ASCII-encoded keyboard on, although I suppose switching gears to doing a PS/2 (or XT?) scancode generator might be worth a laugh. The way these TI boards are constructed the shift lock key is on a row all by itself; it would be pretty easy to add up to 7 more keys to that row. Could solder some pushbuttons to some perfboard to make alt/esc/del/ins/whatever keys, albeit awful ones.

 

[Eudimorphodon]

(In case anyone is still interested in buying these keyboards...)

I got my order of a mixed batch of a few more of the Hi-Tek keyboards plus an equal number of "hacked" Stackpole keyboards. Note for anyone interested in this deal:

• They aren't kidding about the Stackpole keyboards being hacked. It looks like it'll take about a dozen repairs to the PCB board to restore it to the original matrix.
• Other than that the boards look pretty brand new. I don't know if they're NOS, pulls, what, but they're shiny.
• I think the "Hi-Tek" keyboard in my TRS-80 is actually a Stackpole? It looks a lot more like the stackpole than the Hi-Tek. (Maybe Hi-Tek was making them like this in the late 1970's and they licensed this design to Stackpole while iterating a new for their own products?)
• Haven't actually thrown them on a scale, but I think the Stackpoles might weigh almost twice as much as the Hi-Teks. The plastic is *much* thicker, especially in the keycaps. (I posted that comparison shot between the TRS-80 Keycap and the Hi-Tek 99/4A keycap that showed how much thicker the TRS-80 plastic was; the 99/4A stackpole looks just as thicc as the Trash 80.)
• The keycaps are compatible. (IE, if you really wanted to you could buy the 2-for-1 deal and transfer the heaver stackpole keys onto the Hi-Tek.)

So... I guess in short, if you want to repair actual 99/4As with these keyboards the Stackpoles might actually be better quality, you'll just have to repair the cut marks. When I feel particularly bored I'll try fixing one and see how it goes. From what I can tell they're all cut identically... I kind of regret not asking the seller what in the world they were using them for.

 

[Eudimorphodon]

... Probably should stop responding to this, but in the slight chance anyone cares, here's what you get if you get some of those "hacked" Stackpole keyboards:

The stackpole and hi-tek keyboards look almost alike; the Stackpoles are *slightly* lighter colored, and while both types have textured keytops the sides of the Stackpole keys are much shinier. Here's the two next to each other, Stackpole on top:



As I said, the plastic for the keys is much thicker. The stems are compatible enough to swap keys, but the angles of the stems are *slightly* different, so if you mix keycaps on a single board between this and the differences in shine it won't look quite right:


It looks like they cut the traces on the boards with a combination of a drill bit and a wheel cutter. Here's what one of the drill-bitty holes looks like, and it shows what I did to fix it; I sanded off the solder mask, blobbed some solder on the trace ends, and laid a single strand from some stranded wire across the gap.




Here's a shot of the whole back of the board after I fixed it and covered the patches with nail polish. The worst part to repair is under that blob in the upper middle of the board, there are four traces cut with a wheel cutter at an angle. All the boards I've looked at have exactly the same hacks.



And here's the whole mess hooked up to the test arduino scanner, showing all the keys work. I did discover a bug in my scanning software that causes you to get an extra press if you release a shift key before the key you're shifting. I've fixed it.



Anyway. If you're willing to spend about an hour bridging the cuts the keyboards are serviceable enough. Pretty good deal for $3 each, I guess.

 

[seat safety switch]

At least it's a rigid PCB instead of some awful membrane thing. I guess that's probably why they could hack them up in the first place, too.

Thanks for the details! I was tempted to get one but the shipping to Canada is strong.

 

[hlt1]

I got a tray of these. Cost per keyboard was a bit more than the old $3.95 archer/radio shack keyboards on closeout long ago, but reasonable for modern times.
The Stackpole keyboard trace cuts were the same as others here have observed. it looks like they were isolating control keys for whatever reason. Maybe a ZX81 upgrade? The cuts are highlighted here:



Correction, $2.95! I guess they had a closeout sale on the 277-1017 Stackpole variant and replaced it with the 277-1023 Mitsumi type. The older Stackpole model was the better deal but back then it probably looked dated and they couldn't get rid of them. I recall TI99 power supplies going for about the same price. It was junk unwanted surplus.

 

[Eudimorphodon]

At least it's a rigid PCB instead of some awful membrane thing. I guess that's probably why they could hack them up in the first place, too.

They’re very… nostalgic, I’ll say that. As “mechanical” keyboards go they are probably pretty mediocre compared to a modern Cherry or Alps, but they have a satisfying 70’s “clunk” to them. They would be really good for a TV typewriter or other retro-primitive terminal, other than the proviso that a true 70’s board would probably have “ASCII encoded” key legends. (IE, that layout where the double-quote is over the number 2, etc.)

 

[Eudimorphodon]

I got a tray of these. Cost per keyboard was a bit more than the old $3.95 archer/radio shack keyboards on closeout long ago, but reasonable for modern times.

Out of curiosity, what did you get them for?

The Stackpole keyboard trace cuts were the same as others here have observed. it looks like they were isolating control keys for whatever reason. Maybe a ZX81 upgrade? The cuts are highlighted here:

I would kind of love to know what they actually were doing (or intending on doing) with these. The cuts are completely consistent between boards (I assume they used a template), but none of them seem to be touched beyond the cuts.

BTW, if anyone else wants these you might want to move sooner than later, they've been slowly ratcheting up the prices on the auctions; I paid $60 for the tray of un-molested Hi-Tek's, it's $90 now, and likewise the 50/50 tray has gone from $33 when I bit to $46.