XT/AT/PS2/Terminal to USB Converter with NKRO

[fohat.digs]

Sorry, I forgot to report back.

I bought an ethernet connector on ebay for under $3 which included a double female coupling and a short double-ended cable.

I cut the cable and spliced it to an AT plug from an old keyboard.

Untested in "real life" since I do not currently own a terminal keyboard, but continuity checks out.

[YakMN]

Can I get some help with macros? I looked through all the examples in the download, and through the docs. I have a couple issues I can't seem to solve:

1: If I include these macros, the .sc compiles, and appears to load, but after unplugging and replugging the usb, the keyboard will just have some kind of default map. (This is an AT model F).
# Shift-SysReq = control-alt-delete
  macro EXTRA_SYSRQ SHIFT
    PUSH_META CLEAR_META all
    SET_META LCTRL
    SET_META ALT
    PRESS DELETE
    POP_ALL_META
  endmacro

# SysReq = force quit app on mac (Command-Option-esc)
  macro EXTRA_SYSRQ
    PUSH_META CLEAR_META all
    SET_META LGUI
    SET_META ALT
    PRESS ESC
    POP_ALL_META
  endmacro

---
I don't know if it matters, but I have mapped EXTRA_SYSRQ to PRINTSCREEN, and SCROLL_LOCK to EXTRA_SYSRQ in the default layer, prior to this point in the .sc.

Problem 2: These macros, if included, don't kill the load. They just don't work, as far as I can tell. Neither one works to replicate the mac key combo in question. In a text editor, if I hit them, I don't see any characters get inserted, but in the terminal, "5~" shows up if no shift is used (nothing shows if shift is down). This appears to happen regardless of whether the macros are pointd at PRINTSCREEN or not, so maybe this is some kind of "normal" behavior in a mac?

----
macroblock

# Shift-PrintScreen = screen capture of entire screen on mac (cmd-shift-3)
  macro PRINTSCREEN SHIFT
    PUSH_META CLEAR_META all
    SET_META LGUI LSHIFT
    PRESS 3
    POP_META
  endmacro

# PrintScreen = screen capture part of screen on mac (cmd-shift-4)
  macro PRINTSCREEN
    PUSH_META CLEAR_META all
    SET_META GUI SHIFT
    PRESS 4
    POP_META
  endmacro
 
endblock

----

I'm sure it's something dumb I'm doing, so any advice would be appreciated.

--

[Soarer]

The only thing I can spot right now is that using the non-handed codes with set_meta will set both, for example 'set_meta shift' will set left and right shifts together. But, I'm not sure if that matters to the mac, and you've got only handed modifiers in the cmd-shift-3 macro already. (In the trigger, non-handed means 'either', so that part is fine as well).

Can you post your whole config, as an attachment or in a [code] block, please? I'll try and take a better look later

[YakMN]

Yeah, I was trying all kinds of things to se if it worked one way or the other.

Thanks for taking a look!

I have a couple things commented out, because I haven't actually done the full ANSI mod yet.

[Soarer]

I just changed the set_meta commands to use handed modifiers and tried it out... the last three work on the mac, and ctrl-alt-del works on windows... so, I'm puzzled why it's not working for you (The screenshot functions work with multiple modifiers pressed, so I don't think the handed vs non-handed modifiers made any difference).

What else could it be... are you using a KVM? Or a low-speed hub? Some remapping software?

[YakMN]

Hmm. I am not using a KVM. I did have it going into a dell monitor as the USB hub, but I tried it just now going right into the mac mini, and it did the same behavior. I tried something new just now, and found out it's not so much the content of the 2 macros I said couldn't be there, but the fact there are 3 or more macros, period. Not that it actually does the force-quit command on the mac if it happens to be one of the macros left in. (oh wow, force quit dialog just opened. now. about 5 mins after I pressed the key. weird. )

Before that weird thing just now with the force-quit, I was thinking it is maybe a memory issue on the teensy? Here is the report when loading the config:
---
scwr v1.10
scwr: looking for Soarer's Converter: found
scwr: reading file: 160 bytes: ok
scwr: sending info request: ok
device: ok
protocol version check: converter=1.00, scwr=1.00: ok
settings version check: converter=1.01, file=1.01: ok
settings length check: max=506, file=158 bytes: ok
scwr: sending write request for 158 bytes: ok
device: ok
device: ready
scwr: sending 60 bytes: ok
device: ok
device: ready
scwr: sending 60 bytes: ok
device: ok
device: ready
scwr: sending 38 bytes: ok
device: ok
device: complete
----

[YakMN]

So something is definitely wonky here. I think some key presses are going missing, and now, every once in a while, the force quit dialog is coming up, AND the cursor is turning into the take screenshot cursor. at same time. I thought it was maybe time related, so I timed it. But it's irregular. Now I'm thinking maybe it's related to the number of keypresses after pushing the macro-key.

There doesn't seem to be a certain number of strokes that cause the macro to fire either.

[Soarer]

Ahhh. Yeah, that looks like a board with an AT90USB162 chip on it - 512 bytes EEPROM, and only 512 bytes RAM. I'm afraid that's not enough for either the layer or the macros Probably also explains the default map after reconnecting it, and the other odd behaviours. Most people are using the Teensy 2.0 with the ATMEGA32U4 chip (and 2.5KB RAM), and don't even come close to filling it up, so there are probably some bugs when the RAM is full.

Layers in particular take more space in RAM than in EEPROM - you can monitor it with the 'scinfo' command. Also, if you have hid_listen running while you 'scwr' a config, you'll see messages about whether allocation was OK or not.

The only real solution though, is to get a bigger chip.

[YakMN]

I'm having a hard time reading what's on the chip, but it does look like ATMEGA32x4. (I can't read what's in the x position).

I have had this for a while, and it was a testbed for some other projects. I'll order a new one and see if that works better. as long as it's not my precious AT F!

Thanks for help Soarer, and, if I forgot to mention it before, for the AWESOME mod. !!

[YakMN]

oh for the love of... I had picked the wrong .hex file when I started out: I loaded "Soarer_at2usb_v1.12_atmega16u4.hex", but should have loaded "Soarer_at2usb_v1.12_atmega32u4.hex". sheesh. can't fix stoopid...

works perfectly now. Just going to see if I can wipe off this shame. oh, the humanity... <looks at bright object> Ooh, pretty! now where was I...

[Soarer]

It was late, but damit, I should've realised... a) I don't provide a hex for a '162, and, b) few people have the kit required to load the almost 16K hex onto the 16K chip (i.e. without the help of a bootloader). Glad you figured it out regardless of my poor advice!

[fohat.digs]

Just going to see if I can wipe off this shame. oh, the humanity...

I have made mistakes much more stupid than that.

[YakMN]

Is there anyway to wire up the 3 LEDs on my AT F so that I cause one to light up if one of my Fn keys is held down? As is it, I can't use numlock or scroll lock lights (neither does anything on a mac). I don't even like wasting a key on caps lock, I only keep it there because I like to see the light turn on. But it would actually be useful if I could make the other 3 light up when you had a Fn key held down.

[Soarer]

There isn't, sorry.

[YakMN]

np, thanks for quick reply. Maybe I just need to do a blinkenlights mod on top of this one

[nightspirit]

I just made one of these converters for my 122-Key Terminal IBM Model M (8P8C Connector), and I seem to have run into a couple problems. The first is that I don't seem to have NKRO. I'm not sure if I did something wrong, but I try the "hold both shift keys and type quick brown fox" test and I get "THEQUIKROWFOJUPSOERTHELAYDOG". And there are some of my keys (11 of them) that don't register with xev, specifically all 10 keys on the left hand side of the keyboard and top-right button on the number pad. Have I done something wrong, or is this all normal? Sorry if this any of this has been discussed, this is an awful long thread to sift through.
EDIT: Noticed that there's another key that doesn't seem to work. It's the one labelled "{ }" that exists between Field Exit and quotation.

[Soarer]

Model Ms aren't NKRO, and the converter can't change that.

You'll need to load a config with some remaps  to get all the keys working - there's a couple in the configs folder that work with the 122-key 'boards (legacy.sc, halvar.sc). Check the docs for how to compile (scas) and load (scwr) configs, then tweak one of those to suit your needs. Finding out which codes you need to remap is described in Initial Translation on the Config page, but most are pretty obvious if you start with halvar.sc!

[nightspirit]

Okay, thanks for the clarification, I'll look into that.

[JBert]

The first is that I don't seem to have NKRO. I'm not sure if I did something wrong, but I try the "hold both shift keys and type quick brown fox" test and I get "THEQUIKROWFOJUPSOERTHELAYDOG".

That's not an NKRO test, it's just some random test for your controller. How often do you press both shift keys? It's only good as a "newbie test".

You have to press 3 key combinations, then see which ones don't work. If literally any 3-key combination works, try with 4 keys, and so on.

Not that it will help you much as the model M is 2KRO if I recall correctly.

[xwhatsit]

I'd have to remove Out from the report as well It's a pain because each one of these grows the descriptor by 22 bytes, which might cause its own problem at some point. I'll just have to keep an eye on that table for now, until I can think of a good way to phrase a bug report such that it gets some action (if they see it as just one DIY project that's acting up, it won't get fixed!).

This has, predictably, bitten my USB beam-spring converter, now that I've implemented NKRO.

I'm surprised it took me a few days to notice too, being a Linux guy who uses the backslash/pipe key frequently!

I have solved it in the same manner you have, by putting padding bits into the descriptor where the problematic keycodes would normally be.

I'd suggest reporting it; looking around the internet (I had no idea what was causing it at first! I even stripped down my beamspring to see if there was any dust on the backslash sense pad ), I see many others with weird backslash issues under Linux, some from keyboards that seem to be gaming keyboards or otherwise a bit `special' and thus probably implement a bitmapped keyboard report for NKRO. If you don't report it I'll have to get motivated to do so but it appears you have a bit more of a handle on the underpinnings within usb-core and usb-input (well that's my excuse anyway ).

Here's an example: http://forum.corsair.com/v3/showthread.php?t=119654

[killmaster]

Any idea if this works with a Teensy 3.0?

[Soarer]

It won't. Teensy 3.0 has a totally different CPU architecture for a start. But the real killer is that it runs at 3.3 volts, and can't interface directly to our 5 volt keyboards.

[killmaster]

Okay, thank you!

[Soarer]

I'd have to remove Out from the report as well It's a pain because each one of these grows the descriptor by 22 bytes, which might cause its own problem at some point. I'll just have to keep an eye on that table for now, until I can think of a good way to phrase a bug report such that it gets some action (if they see it as just one DIY project that's acting up, it won't get fixed!).

This has, predictably, bitten my USB beam-spring converter, now that I've implemented NKRO.

I'm surprised it took me a few days to notice too, being a Linux guy who uses the backslash/pipe key frequently!

I have solved it in the same manner you have, by putting padding bits into the descriptor where the problematic keycodes would normally be.

I'd suggest reporting it; looking around the internet (I had no idea what was causing it at first! I even stripped down my beamspring to see if there was any dust on the backslash sense pad ), I see many others with weird backslash issues under Linux, some from keyboards that seem to be gaming keyboards or otherwise a bit `special' and thus probably implement a bitmapped keyboard report for NKRO. If you don't report it I'll have to get motivated to do so but it appears you have a bit more of a handle on the underpinnings within usb-core and usb-input (well that's my excuse anyway ).

Here's an example: http://forum.corsair.com/v3/showthread.php?t=119654

Yeah, one of us really ought to get round to reporting it

I think It helps a lot that other keyboards are falling prey to it - even if they don't claim to support Linux (which in most cases is probably just laziness / lack of testing).

I got horribly lost looking at that code I kept thinking I had a handle on it, but then I'd see something that proved my notions wrong. It must be tracking scan code state somewhere, after that table is used, but I don't think I found it.

[fohat.digs]

I have gotten a question that I cannot answer.

Are any or all of these contacts available to reassign and use?

It seems that I have heard that some are not, such as the left side of numpad "0" and the bottom of numpad "Enter"

Can anybody say for sure which of these 5 could be activated, and where to find them in the table?

Thanks!

[Halvar]

Have you tried hid_listen? If something shows up there when you short-circuit the contact, then you know for sure that the contact works, and you also get the code it sends, which you can look up in the table.

[fohat.digs]

Thank you for the reply, but I was hoping to get a definitive answer from someone who knows, rather than learning for myself by trial and error.

For one thing, I do not currently have a keyboard disassembled, and would prefer to avoid the time and pain of experimentation.

If I knew what I was facing in advance, it would make the process smoother.

[Soarer]

All of those should work fine, although I haven't tried them myself. Well, apart from 03 and 05 which are standard on 122s!

01 = EXTRA_INSERT
02 = EXTRA_PAD_PLUS
03 = PAD_MINUS
04 = INTERNATIONAL_1
05 = EUROPE_1

The other two normally hidden pads that are wired are:
top half of ISO enter = BACKSLASH
left half of backspace = INTERNATIONAL_3

[winged mantis]

Im having trouble getting the firmware to work with my keyboard. I have a Dell AT101W connected to it. The hex file installs on the teensy just fine, but after the 3 led's blink and windows installs the drivers, no characters appear when I type.

[Soarer]

See the 'Trouble' page

Probably a problem with your wiring, and maybe hid_listen can give you a clue whether something is not connecting well, or if you have data and clock swapped over.

[winged mantis]

Edit: Nevermind, wired the data and clock wires to the wrong places

[vivalarevolución]

Sooooooo........

I'm looking into making a converter, and I am somewhat unsure of what parts I need.  I need this for a 122-key Model F.   I read through the first 17 pages of this thread, downloaded the documents, and finally got tired of all the reading.  Here is what I think I need:

1.  breadboard
2.  Teensy w/ pins
3.  5-pin socket
4.  Mini USB cord to connect to computer

Will I need anything else to make this work?  With the breadboard, is necessary for me to solder anything?

I probably will have more questions as I go along, because I am neither a programmer nor electronics fiend.  I also need step-by-step instructions and lots of diagrams and pictures to help me learn, along with the deep, reassuring voice of a guru behind my shoulder.

[fohat.digs]

I'm looking into making a converter, and I am somewhat unsure of what parts I need.  I need this for a 122-key Model F.   I read through the first 17 pages of this thread, downloaded the documents, and finally got tired of all the reading.  Here is what I think I need:

1.  breadboard
2.  Teensy w/ pins
3.  5-pin socket
4.  Mini USB cord to connect to computer

Will I need anything else to make this work?  With the breadboard, is necessary for me to solder anything?

I probably will have more questions as I go along, because I am neither a programmer nor electronics fiend.  I also need step-by-step instructions and lots of diagrams and pictures to help me learn, along with the deep, reassuring voice of a guru behind my shoulder.

You won't need a socket unless you don't want to do individual wires or cut the original cord. If you cut the original cord, you will need to solder the cut ends at the Teensy.

These will take care of you since you will have pins on both ends that you need to connect. Color-coded, too.

http://www.ebay.com/itm/181165458233

Otherwise, you should be good except that you will want to fasten the Teensy somehow, but even a cable tie would do that.

[Soarer]

Yes, I would suggest those wires as well; breadboards aren't ideal for the long-term. With them, fitting the Teensy inside the case, and getting a suitable length USB cable for your needs, you can have a reasonably permanent (but reversible) conversion up and running pretty quickly, and without soldering. An external converter is obviously more flexible if you have multiple keyboards you want to use with it, but usually requires at least some soldering and making mechanically sound (i.e. a box etc).

This thread is best read backwards these days, or via the index and search in the docs!

[vivalarevolución]

Ooooffff.   I'll see what I can do.  So many of the terms thrown around for this converter are so foreign to me that I spend half the time looking up the meaning of the term.

[fohat.digs]

This thread is best read backwards these days, or via the index and search in the docs!

Soarer's documents are really very thorough.

Wcass's programming is post #560 (I have memorized that one).

You know the saying "RFM" you will save some grief if it is part of planning rather than troubleshooting.

[vivalarevolución]

This thread is best read backwards these days, or via the index and search in the docs!

Soarer's documents are really very thorough.

Wcass's programming is post #560 (I have memorized that one).

You know the saying "RFM" you will save some grief if it is part of planning rather than troubleshooting.

RFM?  Recency, frequency, monetary value?

[fohat.digs]

RFM?  Recency, frequency, monetary value?

"Read the frakking manual."

[IonutZ]

Could I upload the binary to a Teensy 3.0?

[fohat.digs]

Could I upload the binary to a Teensy 3.0?

Post # 922

[fohat.digs]

here is the problem:

After completely disassembling and putting an F-122 back together, I get a screen full of random garbage. I get dozens of "Calculator" (which is F14 remapped) and those are mostly full of "5s" with a few other numbers mixed in.

When I could get it all to hold still for a few seconds, Aqua key test reported a few hits and a lot of misses.

I will take it apart again, of course, but this is clearly an electrical problem.

[SOLVED] - NOT

After completely disassembling it, there appeared to be a  VERY  few  TINY  specks of dust on the mylar sheet that goes between the metal back plate and the capacitive PCB. I wiped all 3 pieces with alcohol and put it back together. Everything worked properly - for a while.

[UNSOLVED]

After hooking it back up to the original cable and outboard Teensy, it happened again. After some experimentation, I realized that the problem might be the cable. I can't tell whether it is not making good contact at the PCB, or the male end of the 240 DIN plug, or in the wiring, but when I did a direct connection to a loose Teensy with loose wires, it worked.

So I use it for an hour or 2 with no problem, then it starts spitting out spurious keystrokes. Sometimes there is a pattern, for example pressing "W" produces "2W" or "W22" and I am getting a lot of "5s" and sometimes I suddenly get a dialog box or something that would normally indicate that I had pressed something special.

Once it starts, it gets rapidly worse, as if it were heating up or something. A reboot usually fixes it temporarily.

I like this board and it seems nice and clean otherwise. I would like to make this a daily driver.

Is there an electrical component that may be failing? Would a bad connection produce extra keystrokes?

Thanks for your help.

[Soarer]

How very strange! Was it working OK before you took it apart? Or didn't you use it for long enough before to reveal the problem?

The prime suspect from an aging perspective is the power smoothing capacitors - if they aren't effective (particularly the one near the capsense chip) then the capacitive sensing could well become unreliable.

To get keypresses requires a decent clock signal; I don't think a bad connection could cause extra keypresses.

[fohat.digs]

How very strange! Was it working OK before you took it apart? Or didn't you use it for long enough before to reveal the problem?

The prime suspect from an aging perspective is the power smoothing capacitors - if they aren't effective (particularly the one near the capsense chip) then the capacitive sensing could well become unreliable.

To get keypresses requires a decent clock signal; I don't think a bad connection could cause extra keypresses.

I had tested it in the beginning with Aqua Key Test and it turned out good. I probably did not use it for an extended period. Also, I tend to ignore occasional isolated spurious keystrokes because my typing is sometimes sloppy when I am going fast.

Now I hooked it up with different wires and different Teensy applications (external box with cable and adapter, and loose Teensy wired direct) and got the same result, so clearly (heartbreakingly) the problem is in the keyboard somewhere. I get a variety of different characters, so that indicates that there is not an isolated problem "out in the field" somewhere.

Both Teensies are set up with pull-up resistors (1K ohm, 1/4 watt), could I be "pulling up" too much juice into the clock or data circuit?

How would I find and check the capacitors? Should I be looking for a short circuit somewhere else?

I actually have another controller board, but that long line of connections off the flat gray cable is very intimidating.

[Soarer]

It would definitely be easier to replace the capacitors than swap the controller! I haven't got an open F 122 handy to point them out, but they'll be the same as in the XT I guess, silver, lead at each end (axial), little under an inch long... ahh, here, scroll right down... http://www.seasip.info/VintagePC/ibm_6110344_cswap.html. There's two of them worth replacing - one near the square silver capsense chip, and one near the cable connector. (I'm pretty sure the third one next to the two yellow capacitors is just for providing the reset signal to the CPU, and that seems fine).

Gotta admit I'm just guessing! But, I'm also thinking of what's the easiest thing to try first, when you haven't got lots of test equipment to hand.

Unfortunately they're not easy to test, so replacing them is the best bet. TBH, anyone into Model Fs should have a stock of them... they're 47uF, 16V, axial, electrolytics, e.g. http://www.ebay.com/itm/300351410410 (actually, I think the originals are tantalum, but electrolytics work fine).

They need to go in the right way round: the -ve side is marked, and goes to ground, which is on the end away from the cable connector for both of those caps (if your controller board looks like the one in those pics).

[fohat.digs]

Thanks!

We are talking about these 2, right? What about the 2 black upright ones, are they questionable?

[Soarer]

The black ones will be fine. It's only the ones that smooth power that have even a remotely hard life.

Nope... these two...

[fohat.digs]

Nope... these two...

Great, thanks!

I ordered those capacitors that you linked. Hopefully, from Canada, they won't take too long.

Maybe I will get lucky. Maybe I already am - I have worked and used several Model Fs without capacitor issues. It will be nice to have some spares on hand, just in case.

Are my symptoms, spurious key strokes, typical for a failing capacitor? What should I look for in one starting to go bad, and/or a completely dead one?

[Soarer]

I don't know if it would be 'typical', or have any statistics for failures of capacitors of this type (don't even know exactly what type they are!), but it's certainly a reasonable assumption that if either of the capacitors fails to an open circuit then the Model F's capacitive sensing could have trouble, and cause those symptoms.

It might be worth measuring resistance between +5V and GND to see if it seems off, or if it takes a different time to settle to a steady reading... you have a few to compare with

[fohat.digs]

All hail King Harry, perpetrator of the most bizarre and improbable problems imaginable (or not).

I think that I have found and fixed the most recent problem, and you are not going to believe this:

After painting several plates in different colors, including flat black and glossy black, I was not happy enough with the results to let them show as a bare "stealth" board. Looking at my options at the paint store, I decided on "hammered" black in a spray can. I was extremely pleased with the results - a nice light subtle textured finish that was "just right"

After reassembling the keyboard, I gave the front plate a couple of light taps with a hammer to seat the tabs fully. Since this left a couple of small marks on the end of the plate, I masked the keys and gave the end a light touch-up spray of paint. Since ordinary paint is not a conductor, I did not think about a bit of overspray on the PCB behind. But wait - the "hammered" paint must have some metallic dust or flakes in it to produce the desired effect, and, of course, that makes the paint conductive!

There are a few small parallel traces running down the side of the board, and there was a cloud of black about 5mm x 2cm that must have been partially shorting them together. I very gently rubbed the area with alcohol on a paper towel to remove the overspray. It seems to be working properly, it has been about 20 minutes now and I am typing this on it with no spurious key strokes.

Not one of the top 5 most obvious possibilities to troubleshoot, eh?

[Soarer]

Good theory, but I'm not convinced! There's a solder resist layer, so it shouldn't have directly shorted anything, and lines in parallel have some capacitance between them anyway.

But we shall see in about an hour I guess!