Beamspring USB controller

[xwhatsit]

Well the basic principle stands for Model Fs, it's effectively the same thing except backwards (`high' is unpressed on Beamspring vs pressed on Model F).

The Model Fs typically have a 16x8 or Nx8 layout instead of the 23x4 matrix of the beamsprings. There is also a different connector (well, no connector, just a ribbon cable soldered on). I believe on some (all?) of the XT Model Fs the controller is non-detachable so that makes it entirely impossible unless you want to break out your hacksaw (I will not be a part of anything like that!). Basically what I mean is that the same technology is usable, but one would need to design a new controller PCB.

I've been wanting to do a Model F controller for a wee while but I still haven't done a beamspring Display Writer yet. Work pressure and wedding planning!

Those kishsavers look sweet. If I could get my hands on one I'm sure I could churn out something pretty quickly there. The weird Windows key thing is kind of freaky though, along with the midget spacebar.

[wcass]

that too, but i would love to finish what i started with the XTant.
http://deskthority.net/workshop-f7/brin ... t3047.html

i have everything but the controller done. 8x11 or 8x12 would work for me. here is the new steel plate and pad with barrels installed.

XTant Plate.JPG (97.81 KiB) Viewed 5497 times

i'm actually getting pretty good at DIY PCB's. I use the photoresist method and can get very fine detail.

[wcass]

Well the basic principle stands for Model Fs, it's effectively the same thing except backwards (`high' is unpressed on Beamspring vs pressed on Model F).

The Model Fs typically have a 16x8 or Nx8 layout instead of the 23x4 matrix of the beamsprings. There is also a different connector (well, no connector, just a ribbon cable soldered on). I believe on some (all?) of the XT Model Fs the controller is non-detachable so that makes it entirely impossible unless you want to break out your hacksaw (I will not be a part of anything like that!). Basically what I mean is that the same technology is usable, but one would need to design a new controller PCB.

I've been wanting to do a Model F controller for a wee while but I still haven't done a beamspring Display Writer yet. Work pressure and wedding planning!

Those kishsavers look sweet. If I could get my hands on one I'm sure I could churn out something pretty quickly there. The weird Windows key thing is kind of freaky though, along with the midget spacebar.

about the kishsaver, all of the caps are standard M/F compatible. the spacebar is actually the "large" one that can still be ordered from Unicomp.

i've been working on an M-60 design with a PCB (with diodes) as the bottom membrane - documented over on GH. that would work with an 8x8 matrix.

[Halvar]

@xwhatsit, @mr a_500: thanks a lot to your advice and ideas!

I don't see any shorts or misaligned pins on the large connector. I didn't have time to really go on with this today otherwise, will probably take apart and inspect the inner assembly/PCB tomorrow and check the USB part of the controller as well as I can. Some paths on the keyboard PCB look less than trustworthy, will have a look at that.

[Muirium]

i've been working on an M-60 design with a PCB (with diodes) as the bottom membrane - documented over on GH. that would work with an 8x8 matrix.

Very smart. I wondered how to put diodes in a Model M recently when, after a bolt mod, I found out how limiting the rollover is on my 122 key terminal. A custom membrane, playing controller duty too, is definitely a sound way to go. If I had the skills, I'd have a diode matrix and Soarer's controller in one of mine already!

Remember to show off your good work when it's done. I'm always after inspiration… to steal!

[xwhatsit]

@xwhatsit, @mr a_500: thanks a lot to your advice and ideas!

I don't see any shorts or misaligned pins on the large connector. I didn't have time to really go on with this today otherwise, will probably take apart and inspect the inner assembly/PCB tomorrow and check the USB part of the controller as well as I can. Some paths on the keyboard PCB look less than trustworthy, will have a look at that.

If you suspect some funniness with the keyboard's PCB itself, then you can try just unplugging the controller entirely from the keyboard, but then plugging it in via USB. This will quickly rule out any shortcircuits between columns etc.

With the controller plugged in by itself (no keyboard), you have a 50/50 chance of the controller either spamming heaps of keypresses, or sending nothing at all (depending on what happens during autocalibration). Then it's a question of whether you still get the strange USB device warnings... very odd.

[mr_a500]

You could always pinch one from some strange key you don't use. I treat my 3727 (which has a similar layout to a 3278) as close to an ISO layout, but with a wide backslash/pipe (marked as Backtab). So there is a vertical Return. That means I have a `spare' key to the left of the vertical Return, above the RShift. I just have it set up as an `extra' Return, but I don't think I've ever had much cause to use it. If your layout is the same you could swap the flyplate into your spacebar.

In theory the controller will pick up that previously unused sense pad just fine. Presumably IBM has it balanced just as well as the other pads. As you'd be remapping your board anyway it'd be easy enough to put the spacebar in a different place, or add an extra. I don't think the controller would have to do anything special to ignore the extra press either; experimenting with my double-key Return, the OS just deals with it; I press one key, roll over to the other, release the first, then release the second and it sees it all as one `press' and one `release'.

I just took a fly plate from the unused Clear key to add to the left of the space bar. Then I opened the Beamspring utility and set the key under left Shift as Spacebar. Sadly, it doesn't appear to work. Pressing just the left does nothing. Am I setting the right key?

(funny... I have Mac error sound set as DOOM shotgun blast- and after removing the Clear fly plate (before setting that key as Ignore), pressing the key to the left of it had that error sound going repeatedly, nonstop - as if I was repeatedly killing that DOOM shotgun guy. Oh how relaxing... )

[xwhatsit]

I'm not sure if you have the right node or not; however you can see what node is `high' or `low' if the background is highlighted in dark grey or not. You can see this by pressing any ordinary key. If that node that you've added a new flyplate at is valid, then you should see it light up dark grey when you press the key.

[mr_a500]

Yes, I noticed that after posting (nice feature). When I press the left spacebar, nothing lights up. Does that mean I'm screwed?

But... you said it would work! You said it would woorrkkkaaahh aahhhhah... (word disintegrates into manly sobs)

[xwhatsit]

Bugger! If it's a proper sense pad and hooked up to the matrix then indeed there's no reason why it shouldn't.

You haven't confused that pad with a half-pad (they're used as always-off (I think?) dummy nodes for calibration)?

[mr_a500]

Aw crap! It is a half-pad. Well that just sucks.

[Daemon Raccoon]

Hmm, the schematic in the download is missing libraries.

[xwhatsit]

Ah damnit!

I've uploaded v0.5 again (as v0.5a), see original post. It will still warn about the missing libraries (I don't think you'll want to set up the exact same directory structure as me to get the paths right), but it now has the library cache file so everything will work fine.

Thanks.

[xwhatsit]

Halvar: any luck with your controller? USB still behaving badly?

[Halvar]

I didn't find the time yet in the past weeks to sit down and tinker with it (or any keyboard related stuff really), especially as I want to take it apart further to have a look at the faceplate. Probably after the holidays. Thanks for asking! I haven't given up or anything, so much is clear...

[mr_a500]

Is there any progress on solenoid support? Clicks are not enough - I want beeps too!

If power is a problem, what about using 2 USB ports? (I got a mega-hub right beside my keyboard just pulsing with power )

[xwhatsit]

When I finish work for the year (today, yay!) and get some free time I hope I can do a Rev4 board (with similar technology to the Displaywriter).

Be warned that the solenoid doesn't `beep' though—it makes a very lovely `clack' instead! Complements the key `click' nicely.

[Parak]

Aw, right after I finally got around to ordering the pcbs that are now heading my way :p

[xwhatsit]

Probably sensible—can't make any promises—between Christmas lunacy, wedding planning, house painting, bathroom renovations, fianceé catnapping 6 kittens for rehoming, motorbike gearbox repairs, re-doing electrics on my Mk1 Mini, *maybe* sneaking in some sailing on a borrowed Laser... what else have I missed...

We'll see how successful I am in managing to knock off a Rev4 over my `break'

[Game Theory]

Wow down time is not your thing:) (or mine actually)

[webwit]

You gotta manage your PRIORITIES better!

[nourathar]

Hi guys,

for those who were wondering: I bought that IBM 5251 beam spring keyboard that has been around on Ebay for ages, and that did not sell because the price was sort of high. But then I figured that the last one or two beam spring keyboards I have seen on Ebay also went for such prices, so yeah. And it's not like these things ever pop up in Europe, it seems.

I just received it and it feels great: keyfeel is very smooth and quite a difference with anything else I've ever typed on. Can't check, but it seems fingers are just flying on this thing. And on first sight it is in great condition; all keys nicely aligned as far as I can see, but I have not opened it yet.

So the next step will be to get / build one of xwhatsit's converters and that's why I am posting. I feel comfortable soldering and compiling and flashing and such things, but soldering SMD is not something I know how to do.

Is there somebody out there who might have a half-assembled converter ?
Or someone who has the pcb's and would be willing to do the smd soldering on one more of them ?
Or is there anyone with an assembled converter I could buy ?

can't wait to connect this baby !

J.

[wheybags]

Damn you!!!!
I was hoping they'd eventually lower the price

[Soarer]

Weren't we all? They should've just auctioned it... would've got much the same price probably!

[xwhatsit]

Congrats nourathar!

If you feel comfortable with normal soldering, give SMD a go! The first time I tried it I hated it, but now I reckon it's easier than through-hole (no bending legs and clipping things off—also way easier when doing 32-pin µCs!

Starting off with an OSHPark board is very forgiving too, as they seem to be very robust and the silkscreen and ENIG finish really helps. If you make a mistake, maybe the worst you will do is bend some leads on an 75¢ SOIC-16 or something, but you're unlikely to damage the board itself.

Look up a Youtube video on `drag soldering'. Basically you just fill the thing with solder and bridge everything, then use solder wick to clean it up.

If you still don't want to give it a go after watching those Youtube videos (http://www.youtube.com/watch?v=b9FC9fAlfQE is a fantastic, if lengthy, introduction—if you can stand the accent!), and can't find anyone else, I have a preassembled Rev2 board sitting beside me on my desk ready to go.

Hope the keyboard turns out well!

[wcass]

there's also the hotplate method.
http://www.youtube.com/watch?v=Uov0SPHKcnk
http://www.hobbytronics.co.uk/hotplate-smd-soldering

[xwhatsit]

Must try the hotplate style one day. Now that I have built my toaster oven I'm reasonably happy with the results though!

The only issue with the hotplate is that I designed the board to have components on both sides. This works fine in the toaster oven (just flip over, place paste and new components, re-reflow—surface tension holds the other side on), but would be impossible I think with the hotplate.

In the future I will design single-sided component placement if possible. I only allowed myself to go double-sided as when I started I hadn't built my toaster oven so was hand-soldering.

[nourathar]

If you still don't want to give it a go after watching those Youtube videos (http://www.youtube.com/watch?v=b9FC9fAlfQE is a fantastic, if lengthy, introduction—if you can stand the accent!), and can't find anyone else, I have a preassembled Rev2 board sitting beside me on my desk ready to go.

Hi Xwhatsit,

thanks for that link; those eyes too !! It is going to be hard to fall asleep after seeing those..
But very clear indeed and he is pretty effective at dissolving my fear of smd soldering; it actually looks really easy, so I guess you're right and perhaps this is the time to try and do such stuff myself. That might come in very handy for other projects too...

I hope tomorrow I'll have the time to look at it a bit more in detail and find a good place to get the pcb's made (what is a good place in The Netherlands or Europe, anyone ??) and to order the parts.
Is there anybody else who would be interested to have some of these pcb's ?

Just now I opened the thing and it looks very good: in good shape, no shine at ll on the keys, quite clean, only the rubber sheet under the keys is falling apart, but I guess that is not strange after appr. 36/37 years.

Funny how my Northgate now seems miniature and lightweight in comparison!

thanks, J.

[xwhatsit]

Heh, yes, Dave is an acquired taste...

I'll plug OSHPark again even if you're in Europe (they're in USA); they have free shipping (if you can wait for the slow post). I'm in New Zealand, yet they're still easily the cheapest and best quality I can find. I could also make the PCB `shared' (it is open-source after all) so you can order it directly without having to muck around with gerbers etc.

Congrats on the board turning out well. My 5251 has seen better days unfortunately

Yes the PC-AT Model F I have sitting in front of my 3727 seems like a cheap toy.

EDIT: permalinks to Rev2 & Rev3 boards on OSHPark

[xwhatsit]

OK so it will be obvious to all my break over Christmas did not produce either a Displaywriter controller or a Rev4 controller.

However I still did do some useful stuff re. beamsprings.

I had a tiny board made at OSHPark to test out the 74AHC595 shift registers in DHVQFN16 package (2.5x3.5mm):

1972_hi.jpg (62.92 KiB) Viewed 5046 times

These will replace the 74HC4094 (SOIC-16 package, 10x6mm) shift registers used on Rev1–3 for column strobing. Aside from the massive size reduction (see below, compare size next to 0805 capacitors), there is a significant other benefit; 74 AHC series, instead of HC series CMOS.

The AHC series has very powerful drive and fast rise times (to allow ≥100MHz clock speeds). As we are relying on I = C(dV/dT) to detect the change in capacitance, we can increase the `gain' of the pulse out of the row by either reducing dT, or increasing dV. It's hard to increase dV; one could either reduce biasing voltage, which leads to the danger that you'll get a negative voltage when the strobe turns off—probably not good long term for the comparator, or you need strobe voltages higher than 5V, which is a bit tricky. However by switching to 74AHC-series shift registers for driving the strobe pulses, we can reduce dT. 74AHC has a brutally fast rise time (so much so that concerns around ground bounce have plagued it since introduction); on the scope I'm seeing at least 2–4X decrease in rise/fall times compared to the 74HC, even when testing at 3V (I used my STM32F4 Discovery board to test it out). At 5V the rise time significantly shortens even further. In practice it should be even better when compared to the 74HC, as the powerful drive (which is designed to overcome parasitic and input gate capacitance when driving other chips) should mean the rise times stay good even under load (i.e. driving a beamspring board).

The reason I'm bothering with this is less for the classic beamspring boards, but more for looking towards the Displaywriter and (somewhat similar) Model F boards, which don't seem to be laid out with quite as much care as the nice clean early beamspring boards. I'm told by others experimenting with the Model Fs that there is significant variance between individual keys; having a greater `gain' on the resulting row pulse hopefully means there is more tolerance for variations and hopefully still allows a single threshold reference.

This was also a good test to see just how fine-pitch I can get with my toaster oven reflow. As you can see, the chips have no leads per se, and the gap between each pin is about 0.25mm, at 0.5mm pitch.

photo 2.JPG (114.25 KiB) Viewed 5046 times

As it turned out, everything reflowed fine; there were a couple of bridges, but they were easily fixed with a soldering iron and some liquid flux.

Every pin scoped out fine when I drove the thing from my STM32F4 dev kit:

photo 1.JPG (260.78 KiB) Viewed 5046 times

So gives something to start with on Rev4/Displaywriter anyhow.