USB Typewriter Conversion Notes
Michael Schippling  Jan 17, 2011

Some details and (possibly constructive) criticism of the
USB TYPEWRITER(TM)  REV 2.4 conversion kit
by JACK ZYLKIN at usbtypewriter.com


Introduction

The conversion kit in question is available from Jack Zylkin on Etsy.com: http://www.etsy.com/shop/usbtypewriter and it's a deal at about $70. The kit contains two circuit cards and necessary components to add a USB keyboard interface to a mechanical typewriter. I found that it took about 3 days to do the whole conversion, including some reverse engineering and redesign time (detailed in the Complaints Department below) -- J.Z. also sells completed typewriters for a reasonable labor markup if you are less inclined to hack. Since the keys available on your selected instrument may not match up to the extended set on contemporary computer keyboards some of the modern character set is not available to the USB Typewriter user, but most of what you need is there. Aside from a few quibbles at the end of this page this is an excellently weird project to undertake...

If you know how to assemble and solder circuit boards you should have no trouble building the cards. Download the Design Files from the usbtypewriter.com site to get a look at the schematic and layout. Assembly instructions are also on the site. Just follow the schematic and put the named components in the locations indicated by the silkscreen on the boards -- you will need the schematic to figure out the appropriate resistor component values. Once assembled, test the thing by plugging in the controller chip and USB cable and initiating the calibration procedure as described.

A caution that may not be necessary: I would leave the controller chip out of the socket and in some nice static free container when doing anything but actually USING the keyboard. My experience with other controller chips and FET input op-amps indicates that they can be destroyed by inappropriately grounded soldering irons and whanging around in noisy environments. The Atmega used here may be an exception, but why take chances?

Here's what I had to start with before installation (note: use "View Image" on photo files for much larger versions):

typewriter


Installation

Installing the boards and, more importantly, all the switch contacts necessary is an exercise in inventiveness. How you go about it will depend on the actual typewriter you have to deal with. Unfortunately this is the weakest part of the online instructions, so I'm going to give you the benefit(?) of my experience here. I'm sorry to report that this is the most intimate I have been with a mechanical typewriter even though I'm old enough to have written all my college papers on one, so my knowledge of typewriter nomenclature is less than limited. Thus I'm going to make it up as I go along.

Key Contacts

The main labor-intensive component of the USB Typewriter is making and installing the key contact switches. J.Z. invented a method involving hammering resistor leads and sticking them on a cross-bar one by one -- which is why the kit contains a bunch of (in my case) 100 ohm resistors to be sacrificed. This seemed like a PIA to me so I looked for a "easier" solution.

Fortunately, my Corona Four portable typewriter had a spring-loaded plate underneath all the key connecting-rods, and each key-rod had a nice little nubbin that contacted the plate. The nubbins were clear of the plate when the keys were not pressed.

Nubbins, a perfect place for the switch contacts:

key contacts

Through some trial and error measurement procedures I found that my keys were .188 inch apart. I layed out a circuit board with contact strips on that spacing which could be glued to the spring-plate. A contact strip gets hit by a  key-nubbin when a key is pressed with a slight wiping motion so it should make a good contact. It remains to be seen how long the contacts will last since they are just thin copper with a layer of solder tinning applied. In any case the strip you see in the photo is about 8.5" long and .75" wide, with the rainbow colored wires soldered on the edge out of the way of the nubbins. The contact board is glued to the spring-plate using Goop (very sticky silicone sealant), which I used on the wiring elsewhere, and greatly prefer to hot-melt glue. When complete, the whole assembly is flipped over to the right and mounted as seen in the last photo on this page.

An image of the contact board layout is here. I used a laser printer to make an iron-on mask and then cut strips from the board after it was etched. I found that I had to diddle with the DPI setting of the image (using PhotoShop) until I got the contact strips to match up to reality, by repeatedly printing on regular paper and holding it up to the nubbins for comparison. If your keys are on a different spacing you will have to do more diddling to scale the image correctly.

Here's a picture of how the contacts and nubbins line up before assembly:

contact alignment

Also note that there are many more shift-register chips on the Sensor Board than necessary, which is why you can cut it to length. It doesn't matter which inputs (outputs actually) you connect to any key-contact -- you can skip whole chunks if your wiring doesn't reach -- as long as you have all the chips populated up to the end of your line. The configuration procedure will sort out what's connected to what.

Other Switches and those "D" inputs

The least clear part of the instructions deals with the non-keyboard-contact switches for which tiny (and very delicate -- I broke 3 of the supplied 4 before they even got near the typewriter) reed switches are supplied. You can see the wiring for these switches on my typewriter in the last image. The reed switches and magnets need to be mounted near the operational components of keys which are not on the key-contact strip such as RETURN, SHIFT, SPACE, and BACKSPACE. Since this is a matter of mechanics and taste I will leave the details as an exercise for the reader. There are however a few devils...

I had a lovely place for the SPACE bar sensor at the back of the typewriter under the carriage, until I got to the final test and found that this was also the mechanism that moved the carriage after each character was typed. So I had to move the sensor to a new location that was directly activated by the space bar itself. The lesson learned is stay as close to the lever or key as you can since those old mechanical designers had a buncha tricks up their collective sleeves.

The thing that is not covered at all is: For what special keys do I need to do this? For some reason CNTL, ALT, and SHIFT keys, if you want to have them available, must be wired directly to the control board using the reed switch mechanism. This generally precludes them from being "normal" keyboard keys, unless you have some tricks not used by those original designers. I got lucky and was able to put a switch near the last key on the keyboard which was not used in the final character set: "@-cent", so I could use it as my CNTL key. YMMV.

This leaves us with a few more non-key-contact keys or whatnot. Specifically one needs SPACE and RETURN and probably wants BACKSPACE as well. Unfortunate this adds up to FIVE reed switches and only FOUR were supplied in the kit. Plus I had to add a sixth to get my CNTL without overloading my BACKSPACE. So my list of  special keys needing mechanical reed switches is:
  1. CNTL: A regular key, disconnected from contact strip, using a required reed switch;
  2. ALT: A non-key-contact key ...margin release... using a required reed switch;
  3. SHIFT: A non-key-contact key, using a required reed switch;
  4. SPACE: A non-key-contact key, using a reed switch, wired to the "D" auxiliary inputs;
  5. BACKSPACE: A non-key-contact key, using a reed switch, wired to the "D" auxiliary inputs;
  6. RETURN: The carriage return lever, using a reed switch, wired to the "D" auxiliary inputs.
Two things your should know about the "D" auxiliary inputs: The first is that they do indeed "swing both ways" in that the software will detect if they are normally open or normally closed and behave accordingly, so no-worries about which way you need to get the switches to work vis magnets near and far. The second, which is not mentioned, is that they are included in the regular keyboard scan such that they can be mapped to any key (except those durn Cntl-Alt-Shifts) during configuration.

Configuration

The program is pretty swell when it comes to configuration. Every key (except those durn Cntl-Alt-Shifts) gets mapped to it's contact and then the config is saved for later power up, so it doesn't really matter what you solder to what (except those durn Cntl-Alt-Shifts). What isn't mentioned in the documentation is exactly which keys are to be mapped, so I've included a dump of the configuration run here.

What I learned from running the configuration is that there are some keys for which I didn't have actual hardware to map, and that I did have one spare key on the contact strip that was not called for in the map "1/4-1/2" -- which are very pretty but apparently useless to a computer. So I decided to assign that key to TAB, although it could be anything.

Also not on my keyboard, and thus mapped to oblivion, were:
Another tidbit not mentioned anywhere and in need of reverse-engineering is what symbols are mapped when one uses the SHIFT key with numbers and other punctuation keys. The program goes along pretty much with the modern computer keyboard's idea of what should be where, which does not entirely match my typewriter's old-fashioned layout. Here's the list of mapped upper-case keys (lower upper):

A Final Look Under the Hood

The circuit boards are mounted using double-sided foam tape and the wires glued down with Goop:
final assembly


Complaints

I can't think of a good reason for the Cntl-Alt-Shift keys to be treated differently and in need of special inputs -- as you might guess from reading the foregoing. Perhaps there is a reason, but if there was none it would make the key-map selection much more straight-forward.

Another no-good-reason-to-me thing is the use of  M74HC595 shift register chips for the key-contact detection. These are serial-to-parallel devices and their use here requires that the whole typewriter chassis be wired to an input pin on the controller. In practice this seems to work fine, but it would be "better" if the entire case could be grounded. If that were the case (no pun left un-tended) then the key-contacts would work just like the D auxiliary switches -- switching to ground -- and any switch or key could be used in any function. The 74xx165 chip is a parallel-to-serial shift register that could implement this, however it may need pullup resistors which would complicate the sensor circuit board.

There's got to be a better way to do the key-contacts. Instead of rolling my own contact strip board I suspect I could have used standard .100 spaced edge-board-contacts and skipped a contact every once in a while while wiring.

In the REV2_0 documentation, the page named USBTYPE_S2_Schematic.pdf seems to be a copy in a different color of the USBTYPE_S2_TOP.pdf file. An actual schematic is in the REV1_0 bundle, but has been slightly modified for the 2_0 board.

There are two push-button switches on the REV_02 board, but only one seems to do anything. The top switch, labeled S1, is the one to use for configuration.

I needed six reed switches and broke three of the supplied four... fortunately I have hoards of just about everything arcane.

It would be nice if the upper-case punctuation was mappable during configuration so it could be made to match older keyboards.