Pretty good guess actually... The rest of the board is pretty simple stuff I think... Maybe just pullups... Someone who is better than me at squinting at old circuitry can take a peek! Mike On Dec 15, 2015 2:18 AM, "tony duell" <ard at p850ug1.demon.co.uk> wrote:

Yes. The former. You can see it clearly on a photo on Ebay where someone offers a set of 24 of these; the lighting reveals the thick film resistor elements running across the package, showing up as humps in the lacquer coating. That sounds right, and for pullup resistors that would be a plausible value. It matches the numbering used on surface mount components. paul

On Tue, Dec 15, 2015 at 4:59 AM, tony duell <ard at p850ug1.demon.co.uk> wrote: Maybe, but Selectrics aren't exactly fast devices; there's a whole lot of potential 'no, wait, I'm not ready!' conditions. Would they all be ORed onto one pin? Confirmed off-list; it's a resistor network; 14 pins, seven resistors. True and possible. But even if they're ALL inputs, only seven of them? That was my conclusion too. The old Western I/O ads I've seen definitely refer to it as having an 'ASCII' or 'parallel' interface. Assuming they only ever made the two models; I suppose it *could* be some variant of RS232, with very non-standard pinouts - but the ads are specific; they made a smart terminal with 6800 CPU & serial interface, and a dumb printer with an 'ASCII parallel' interface. And that's all I have to go on, beyond prodding the hardware. That would be well weird. Still trying to work out what exactly it was intended to hook up to; a standard parallel port with a special cable, using only four of the data lines, and a driver, to drive it nybble-wise? Maybe I could use business records to try to track down the former owner of 'Western I/O' and just ASK him, if he's still alive, but that would be cheating :-) I've gone over the connector again and we have ten signal pins plus a ground plane... that's *just* enough for 8 data bits, a strobe, a ready/wait line... but that Allen Bradley pull-up pack is only 14 leads, 7 lines... I'd expect to see 8 data lines all going to the same place if it was anything resembling standard Centronics but with a weird pinout. So I'm scratching my head still over just exactly what it was supposed to hook up to. That's true and possible. I'm in two minds on this thing: - intention was to rip all this out and convert it to a full I/O serial terminal, using an Arduino-based setup that Lawrence Wilkinson has already built and tested: https://www.flickr.com/photos/ljw/sets/72157632841492802/with/9201494189/ - all the keyboard contacts are already in there, Western I/O just cut the IBM wires off when they ripped the IBM guts out and converted it printer-only. I'd like to figure out the interface that's presently in it, just to check out the mechanism, and for that 'ah ha!' moment :) - but I don't want to spend any significant time on it if I'm just going to rip it all out. - but, although the Western I/O conversion 'butchered' a perfectly good IBM 2970, it IS a rare representative of that era, when all kinds of Selectric conversions were commonplace. So perhaps, as a nod to that era, it should be left as-is, as a preserved example? What say people? I've seen posts on old lists where people have referred to buying these back in the day - converted Selectrics I mean - and seeing 'mountains' of them in warehouses. They were once common. Where have they all gone? Is mine the *only* survivor from those mountains of 3rd-party backstreet conversions? Does anyone else have any? I've just spent a few hundred bucks with one of the few mechanical Selectric gurus left standing - a local guy here in NZ who did an amazing job, several broken and seized bits fixed, the mechanism is now like new and works perfectly in typewriter mode - so it's going to end up working, one way or another! Mike http://www.corestore.org 'No greater love hath a man than he lay down his life for his brother. Not for millions, not for glory, not for fame. For one person, in the dark, where no one will ever know or see.'

On Tue, Dec 15, 2015 at 1:21 PM, Fred Cisin <cisin at xenosoft.com> wrote: I guessed that might be the case... any suggestions for what were common pinouts and signals used? I can analyze 'backwards', testing possible suggestions, as fast or faster than I can do it 'forwards', trying to recreate a schematic from examination of the hardware. It sounds a lot like what we have here. But I'm not certain a modern standard parallel port can ever be trivially adapted to drive this thing... I thought Centronics dated back to early 1970s - not always in the standard 'modern' form, but in general principles with same signaling and strobing of data. Thanks for input! Mike http://www.corestore.org 'No greater love hath a man than he lay down his life for his brother. Not for millions, not for glory, not for fame. For one person, in the dark, where no one will ever know or see.'

On Mon, Dec 14, 2015 at 9:15 PM, Fred Cisin <cisin at xenosoft.com> wrote: My first experience was in the mid-1980s. Someone gave me an ancient tank of a Centronics-brand printer - 132 columns and 2 print heads! One head got the left side of the paper, the other head got the right side. Unlike the later "Centronics Interface", this one had an internal edge connector for input - 40 or 44 pins (I used a standard 44-pin protoboard but I can't remember if I had to cut it down or not) I interfaced it to the user port on my Commodore 64 and wrote my own handler to trap device #4 and squirt out the data through the user port. Yep. About the only exceptions I can think of are the Amiga 1000 (proprietary but similar parallel port pinout on a DB25M) and DEC minicomputers which leaned towards the Dataproducts-type interfaces that required a gate or two (and probably a cable pin swabber) to talk to "modern" Centronics printers - one example was the DC37 on the back of a VAX-11/730... you had to add somewhere between 1-3 inverters to make that talk "Centronics" but once you did, you could just tell the VMS line printer process to squirt chars out the parallel interface and it would "just work". -ethan

There were just about as many parallel interface versions as devices that used them back then..... Nothing 'common' really... The idea of 7 or 8 data lines, a strobe, and a ready signal was certainly arround back then, but the timing, polarity (active high or low) and timing were not standardised. A couple of examples that I can see without getting up are the Facit 4070 paper tape punch and the HP9866A thermal printer. Those were both around in the early-mid 1970s and are rather different parallel interface. OK, what I would do to get some idea is focus on those 7475 chips. Get the pinout. The most obvious use for them on this board is as the character input latch. IIRC each is really 2 2-bit latches, so 2 enable/clock pins on each chip. So : 1) Are the 4 clock pins linked together (if so, it loads a character at a time [1]), or are they in pairs or what 2) Where do the D inputs go? Are any of them linked together, or do 7 or 8 of them go to the interface connector? If the latter, then those are the data inputs. [1] Before anyone suggests you could use them as a sort-of shift register and load half a character into one, then copy it into the second one while loading the other half character, remember the 7475 is a transparent latch, not an edge-triggered flip-flop making this a very difficult thing to do. If you can identify the data lines on the connector you are getting there. See if you can trace the other pins to inputs or outputs. -tony

----- Original Message ----- From: "Mike Ross" <tmfdmike at gmail.com> Sent: Tuesday, December 15, 2015 3:41 PM ----- Reply ----- Tony just makes everything sound more complicated than it is ;-) Actually I suspect it might indeed quite possibly work with a standard parallel port without any conversion at all other than a rewired cable and maybe a very minor change or two on the board. For the most part a parallel port is a parallel port; data and strobe going out, strobe acknowledge and printer status coming in, nothing more complicated than that and timing is rarely an issue. In the Selectric there is the R/T code conversion of course but that's done for you already and it should be simple to find the data bit order. I had a contract years ago that involved adding a serial or parallel interface to Olivetti typewriters and it's really not rocket science; I even still have some info on interfacing to Selectrics but unfortunately not relevant to yours. Your choice of course but I'd take Brent up on his generous offer to reverse engineer that interface; you may find it's a much simpler project than you think. (another) mike

On 2015-12-15 10:21 PM, Mike Ross wrote: Useful information about interfacing to selectric I/O and maintenance can be found on bitsavers http://bitsavers.informatik.uni-stuttgart.de/pdf/ibm/typewriter/selectric/&… manuals contain all the mechanical adjustments and one of the manual includes wiring diagrams and timing charts. The most useful ones would seem to be 241-5308 IO SelectricRefMan.pdf, CEInstructionManual.pdf, and S225-1726-7 IOseleMnt Nov70.pdf. These are the manuals that we used when servicing Selectric terminals. Paul.

On Wed, Dec 16, 2015 at 8:01 PM, Brent Hilpert <hilpert at cs.ubc.ca> wrote: Amazing work Brent! I've wired the thing up in accordance with your schematics and here are the results: On power-up the line we believe is Strobe is high; all others are low - and I'm monitoring the printer side of the interface here. I cat file.name > /dev/lp0 The printer prints a character; Linux is waiting. The line we presume to be Busy/Ready flickers briefly high as it is printed. I toggle the local/com switch from com to local and back to com: another character is printed. Linux waits. I can sometimes continue this process one character at a time by toggling the local/com switch. At other times toggling the switch sends Linux straight back to the command prompt. The characters printed are pretty exclusively semicolons underscores and 8s. The carriage never advances; all characters are printed at the same spot. Further observations: - If I initiate the print with the Strobe line disconnected Linux returns to the command prompt instantly and nothing is printed. - If I disconnect the Strobe line after printing has started Linux returns to the command prompt instantly after the com/local switch is cycled - If I disconnect the Busy line prior to starting to print nothing is printed until I connect the Busy line - If I print a character by cycling the local/com switch with the Busy line disconnected a *second* character is printed when I reconnect it. - Busy flickers high every time a character is printed. The status of the Strobe line never visibly changes; it always appears high. Might put a scope on those... There's clearly something funky going on with signaling - timing or active high vs. active low. At no time does the printer *ever* print more than one character without some manual intervention. http://www.corestore.org 'No greater love hath a man than he lay down his life for his brother. Not for millions, not for glory, not for fame. For one person, in the dark, where no one will ever know or see.'

In extreme brief as nearly 3am and I've been hacking late on a 3277 :) 1. The mechanism has just been extensively serviced by an expert and works perfectly in local typewriter mode. 2. Possible bad contacts had occurred to me and will be investigated. 3. Yes there's another board that drives the mechanism! I'll get pics. Mike On Dec 18, 2015 2:08 AM, "Paul Berger" <phb.hfx at gmail.com> wrote:

On 2015-12-18 5:00 PM, Brent Hilpert wrote: There are 7 only 6 are actually used to select the character the 7th is to maintain odd parity check contacts below the selection mechanism are wired so that the signal goes true when an odd number of magnets are picked, I don't remember the polarity of the signal off the top of my head, but I doubt this adapter checks them. Of the 6 that are actually used to select the character, two are for tilt (T1 and T2) three for positive rotate (+1, +2 and +2A) and one for negative rotate (-5). There are also two magnets for shift shift up and shift down, one each for functions installed up to four: Tab, Carriage Return, Line Feed, and Back Space ( not all typers have all the functions installed) and finally there is a magnet to release the cycle clutch. The basic operation is pick the selection magnets and then the cycle clutch feed back contacts tell you when you can release the magnet and set up the next character, properly timed you can have it print at full speed without the cycle clutch latching up, but that is hard on the mechanics they seem to survive better at a slower speed. The shift magnets operate on their own releasing the shift clutch to make a half turn to rotate the ball 180 degrees. The function magnets release a clutched cam on the operational shaft the initiates their function, there are also feedback contacts for these functions and for shift. Paul

On 2015-Dec-18, at 12:13 AM, Mike Stein wrote: On 2015-Dec-18, at 8:17 AM, GerardCJAT wrote: They're drawn using the drawing segment of the old Mac AppleWorks / ClarisWorks application suite. The symbols are my own 'symbol package', so to speak. It's not something you want to take up with: - Apple has EOL'd AppleWorks and no longer supports it. Currently I have to maintain a MacOS 10.4 system to continue to maintain the schematics I've drawn over the past 20+ years of use. They can be PDF'd for presentation on current machines and of course printed, but then they're static. I hear BareBones Software produces a current program that can draw & manipulate the original files, I need to check that out. - It's not a native schematic-drawing program, so it doesn't do things such as automatically maintain connections as you manually organise elements. - By the same token, it of course doesn't produce things like netlists, which would be nice to feed into simulations. I need to move to something newer but I haven't evaluated current schematic-drawing programs, and I'm a little concerned they won't provide the degree of control or finesse I like

On 2015-Dec-14, at 2:02 PM, Mike Ross wrote: Well, I vote for doing a little more work to get it working as is. If it's all there, it 'should' work as intended and really doesn't look to be very complex. Chances look good it will be adaptable to a centronics port. Pin 9-11 & 21-24 connect to the pull-up resistors and head in the the direction of the 7475 DFFs - that's likely the 7 data bits. Traces from the DFFs look like they then head to the proms which are likely 3 * 256*4. That leaves 13,17,19 for some combination of control signals such as print-strobe-in, ready-out, fault-out. I notice there is an MC14490 hex contact debouncer there. You say this is not a keyboard-send conversion, so I'd guess those debouncers may be picking up mechanical contacts that indicate end-of-operation, to produce a ready signal. If reverse engineering really isn't your shtick, send along reasonable-resolution photos of both sides of the board and I'll work on a schematic. Could do it with the photos you provided earlier except the solder-side photo isn't complete.

Why not? It's all the host really needs to know -- can I send another character or not. [...] Why 8 data bits? ASCII (which we are assuming this is) is a 7 bit code. A number of older printers did indeed only have 7 parallel data lines. I reckon it would take me a couple of hours at most..... In some ways I agree with doing that (other than using a a million more components than you need...). On the other hand this board is part of the history of the unit, so I would keep that if at all possible I think. -tony

[Printer's Ready signal] True it's nice, but back then such things were often handled by indicators on the printer itself (possibly even by a flashing combination of indicators, so that something like On-line, paper-out and 12cpi all flashing together means 'RAM failed') OK, the Centronics interface does have a paper-out line that most printers implemented, but AFAIK if a printer asserted paper-out then it also asserted busy (or would do when the buffer was full) so that the host really only _had_ to check for busy. It is certainly possible to make a parallel printer interface with 7 or 8 data lines, a strobe and a ready line. Only. -tony