Things are rolling along smoothly on my TV Typewriter build.  At last, I have reached the summit of this device – the Memory board.  This is the board – easily the most intricate and complex, and important.

The Memory board is where the magic happens.  It is host to the Signetics 2513 character generator ROM:

Chances are if you had any involvement with 8 bit computing back in the day, you’ve had or used this ROM.  It was used in the Apple I, the Apple II, the SWTPC CT-1024, early Atari games and so on for generating (upper case) text.   In fact, if you’re typing on an original Apple II, you’re producing exactly the same characters this TVT device does.

The memory board also hosts the all important ‘memory’, in the form of 6 Signetics 2524v shift registers.

These tiny (and rare) 8 pin chips can each handle 512 bits of information.  If my calculations are correct, 6 of them together produces the equivalent of about 384 bytes of memory.  Oooooooh!

Now, the TVT was a modular design, so you could add additional ‘pages’ of memory to handle more.  But to get to just 1KB you’d need three of these things, and because my prototype case is only about 4.5″ tall, it wouldn’t fit.  So I’m sticking with just the one – that’s good enough for a full page of characters.  Plus, it’s not like Page B will do you much good.  The two pages do not exist seamlessly for the user – ie, you can’t just type two pages worth of data on the screen, scrolling down as you go.  You have to manually flip a switch to go between pages.  When you reach the bottom of either page, the whole thing blanks and begins fresh.  Anything you typed before is gone!  As a terminal, this is pretty useless.  Don was questioned back in the day on the design by Lee Felsenstein, and his response was essentially ‘hey buddy, it’s for putting words on a TV screen!’.

And at any rate, supposing you did make use of both pages, dutifully typing out your 2 page essay on your TVT.  You had no way to back it up, unless you knew how to wire up a cassette recorder.  Or here’s another recommended method from the article:

LOL

Well, I guess technically it is a hard copy..

Again, Don Lancaster’s TV Typewriter is not an entirely practical device out of the box, really.  It was about the concept.  Further development was up to you (or him), depending on who got to it first.

Anyway, the Page A board is distinct from any subsequent boards you build – Page B etc don’t need to have their own character generator or associated circuitry, so a whole bunch of resistors, etc are left out.  But for Page A, it’s all on, and this board is a blizzard of ICs, resistors, diodes, caps and my least favourite thing in the universe next to mint-flavoured-anything: jumper wires.  Lots.  And lots.  (And lots).  Lots of jumper wires.  You need to be in a Zen mode to handle that.

And you need to pay very close attention.  Because the traces are tightly packed, there’s a lot of holes, and it’s easy enough to accidentally bridge something with solder or install into the wrong holes.  Definitely do not do this while tired!

The drilling was the worst though – it took me about an hour with a Dremel to get it all and even my poor eyes, which were crossing, missed a couple.  And let me tell you about drilling PCBs: it stinks.  Do not do this in a small room.  You should also wear a respirator.  Drilling PCBs produces a nasty, smelly, very fine glass dust.  It’ll stink up your room for an hour.  Trust me.  Do this in a garage.

Anyway, after that and about about five hours of soldering, here’s what we have so far:

Looking pretty good!   I received and am using those little Bakelite 100uf caps.  They definitely look period correct!  I’m also mixing and matching the other IC sockets as I suspect any hobbyist/prototype maker reaching into his parts bin would have back in the day.

I am waiting for a correct-looking 24 pin socket for the character generator and some blue 8 pin sockets I found for the shift registers to finish that end of it off.  In the meantime, I’ll Zen-out and install jumpers everywhere else it’s needed.

Regarding connecting all four boards together,  I’m relieved to report that they all can be snapped together, with a fair bit of finagling.  In this pic I haven’t pressed them in all the way – I really hate Molex connectors and the amount of force required to install or separate things with them.  It flexes the hell out of the boards and makes me cringe every time.  I’m amazed the pins are breaking off!

Anyway, here’s a beauty shot.  Beneath it is a shot of Don’s original.  We’re getting there!

Just for clarity, what Radio Electronics refers to as ‘the Mainframe’ is what you and I today would call the ‘motherboard’.  Just to clear that up in case it causes confusion among those who think of mainframes as something else entirely.

I ended up making three motherboards.  The first was my first-ever homemade PCB; and it showed.  I rejected it outright for overetching and bad cutting.  The second was much better, but had a few spots of copper remaining where I didn’t want it.  I decided to use that one to try doing ‘silkscreening’ on the backside using laser toner.  The third board is the one I’m starting with.  It benefitted greatly from my newly acquired knowledge about etching, cutting and so on, and looks really clean.  Let’s get started.

I should note I decided late in the game to forgo silkscreening altogether on my prototype.  I don’t know why I was so fixated on it.  As we can deduce from this photograph, the prototype does not have it:

So that lets me off the hook.  And anyway, the original silkscreening on the SWTPC kit boards was white, and since I don’t have a (expensive) printer that can print with white toner and I’m not willing to shell out to make an actual silkscreen, I’m going to leave the silkscreen off.

Again before embarking, I decided to check the appendix at the end of the construction guide for any last minute warnings about the mainframe.  Sure enough, there was one.  Diode D6 is shown backwards on the parts overlay!  Luckily, the overlay I borrowed from SWTPC.com had it corrected already:

I’m glad they caught that 40 years ago and that I didn’t have to spend a week or two puzzling over schematics.

Oh and did I mention the schematics themselves have errors?  Yikes!  Several corrections are made in the construction guide appendix.

Per the guide, I want to get molex connectors installed first and check fitment.  It takes minimal time to drill these out on the PCB and then solder them in:

The Dremel does a decent job, although it does have a tendency to wander.  Once you get used to this though it’s easy enough to adjust your entry point to get your holes where you want them.

In the event, I decided to drill out the entire board, carefully looking at the photos from that ebay auction I missed to make sure I drilled the right size holes (more or less).  I then couldn’t resist installing some of the capacitors:

Originally I was going to use these vintage looking blue Rubycons for the twin 1000uf caps, but along the way to getting here I found these mint condition Temple units.  They look much closer to the prototype’s design and are shorter than the Rubycons, so they fit better on the board.  In the end I decided to stack them just as Don did on the prototype.  There’s a practical reason for this – the 5000uf cap is huge and cuts uncomfortably into the space you need for the first 1000uf cap.  Stacking them frees up room.

I also switched from a modern 4700uf cap to a Mallory 5000uf unit.  These tall metal caps are way more authentic looking to the early 70s than the modern ‘sausage’ ones.  Unfortunately these caps are the ‘slot terminal’ type.  They have three contacts around the outside (for negative) and a fourth on the inside (for positive).  I had to jury rig it with 14 gauge wire to make it work with my board.

This prompted some tut-tutting online, as it’s not the way these bladed units are intended to be installed.  The soldered on wire terminals are weaker than the mounting plate it’s meant to use.  However, I did wiggle it a bit and found it to be fairly solid.  This could all be moot anyway – caps do not store for long periods well and even though these are all NOS, there’s a good chance they may be all dried up and will have to be replaced with modern.  Fingers crossed!

After the molex connectors were installed, I decided to try installing the Timing board to see how it fit:

Okay so it looks good, but unfortunately it turns out I was indeed a victim of scan skewing.  One of the molex connectors is skewed just slightly off to the right, so it is not straight in line with the others and requires an uncomfortable twisting of the back end of the Timing board to get it to mate.  I ended up removing the Timing board and adjusting the pins as I had them pointing outward somewhat rather than straight down.  This helped, and the Timing board now snaps much better into place!

That’s all we’re doing today!  Looking good so far!

The timing board is now nearly finished, save for jumper wires:

I’m pretty satisfied, although the real test will be how it cleanly it mates to the motherboard.  I recently got my hands on an original Mark-8 construction guide (thank you, Roy!) and found out just how much scanning distorts PCB artwork.  It doesn’t just change the scale, it actually skews the artwork.  This means in situations where you need parts to align, like the molex connectors that the boards stack onto, you need to check to make sure it all lines up, at least.  Being that these were my first ever homemade PCBs, I didn’t think to check that.  Ulp.

There are a few caps and a whole whack of jumpers missing; those will be installed later.  I’m not liking the new style caps at all — I think those will be revisited.  I’ll get into the whole caps affair in another post.

Having done a bunch of jumpers on my ASCII encoder, I know this will be the least fun bit of the whole project.  I will do them all on all four boards while I await the last parts.  For now, I’m sticking to main components (ICs, caps, diodes, resistors, connectors).

Also, as I mentioned, I’m using sockets for all of it.  This is one clear break from the prototype — from the photo it’s clear Don soldered all his ICs in directly.  For an experienced electronics engineer this is doable, but for a self-teaching novice using 40+ year old ICs, this would be the definition of insanity.  It’s just so much easier from a troubleshooting standpoint to be able to swap ICs.  So, that’s how that goes.  However, in keeping with the hobbyist/prototype theme, I’m just using what sockets I have lying around that look correct vintage, rather than trying to make every one the same.  I think that gives the unit a more earthy feel.

I’m also keeping an eye peeled for solder bridges, and even accidental bridges from the toner transfer/etching process.  The construction guide warns you about these, and indeed, they were the major thing that prevented my ASCII encoder from working right off the bat.

Anyway, we’re done here for now.. onto the ‘mainframe’ (motherboard)!

As I mentioned earlier, I had been procrastinating on starting my TVT build for months.  I had it in my head that I could not begin until I had everything, and could build the boards sequentially starting from mainboard up to all the others.  One reason for this was I wasn’t really sure how some components would fit.  Thus, I might want to drill holes larger than I had in the PCBs.  This is trickier to do when the boards have parts installed.

In the end though, I realized it could be a while before I found the remaining parts I needed, and if I didn’t get going soon with each passing day I might not ever.

So I decided not to go sequentially, and picked a board to start with.  The Timing board seemed like a nice one – fairly straightforward.  I had all the parts necessary except the crystal, which I’m working on.

The first bit of business was getting the molex connectors for the device’s bus in place.  Initially I couldn’t quite visualize how all this went together.  There is a female molex connector that sits on the top, non-copper side of the PCB.   Like so:

It is soldered in on the bottom.  The connectors are arranged in such a way that pins installed and soldered beside their own pins underneath line up and press into the connector beneath.

My concern primarily was how to ensure the pins were strong enough to sustain the force necessary to slide in.  Throughout the course of the build and testing, and later on as the unit aged, I’d probably need to repeatedly assemble and disassemble.

Initially I thought about molex headers.  I’d mount them such that the plastic casing around them was on the same side as the female molex connectors.  But this wouldn’t work — the molex connector overhangs the hole where the bottom pins come up.  Nothing could fit under there.  I ended up buying headers with the pins spaced incorrectly on purpose, just so I could pull the pins out one by one and solder them like so to the PCB:

Once you see it in person, it starts making sense.  The long individual pins basically slide through their hole and butt up under the lip of the molex connector up top.  It doesn’t matter if they make contact with the molex connector’s own pins – they’re all on the same trace anyway.  Then it’s just a matter of getting the pins aligned properly and soldered in.

I should mention that one other worry I had about this setup was that these pins would be vulnerable to breakage.  There’s really not much holding them to the board.. just the thin slice of copper trace they’re soldered to.  But a quick test proved they were more than strong enough to hang on through several plugging/unpluggings.  I did have to repeatedly adjust to get them nice and straight and spaced evenly, but eventually I got the first row done:

Yay!  Now just.. oh.. several dozen more to go!

Okay, this is where the rubber hits the road, sort of.  With the encoder tested out working, now we begin the laborious process of rewiring this Mohawk Data Sciences (actually Microswitch) key-to-tape keyboard.  This is essentially the exact same keyboard Don Lancaster used in his prototype.  It’s not clear if he used the earliest version of his ASCII encoder with it, although to me that would make sense.  Interestingly, the encoder’s mounting points seem to match the keyboard’s depth.

Anyway, here’s the hot mess we are presented with, as it was cut from its machine:

Actually in this pic I’ve already started removing wires.  I need to remove all of them, because the encoding scheme used here was probably EBCDIC or baudot or who knows.. whatever Mohawk used.  We need ASCII.

Once fully desoldered, I figured I’d start working my way from the top of the keyboard down.  This is where care and attention must be paid, join the wrong wire, and you’ll get the wrong result.  Early on I decided on a scheme whereby, looking at the back, I would use the left post for each key to connect to the pin marked specifically for that key, and the right post would connect to the pin corresponding to the last three digits.  So for example, keys 0-7 would all be chained together on the left:

And so on for each group of keys corresponding to the appropriately marked pin on the encoder.  Then on the last key in that group, a wire would connect to the encoder’s pin.

Now where it all gets tricky is when we start connecting the pins corresponding to the last three bits.   To give an example, keys “, 2, B, R, :, J and Z all use 010.    So these are the keys that have to be connected together on their right post, and then that line would go to the pin marked 010 on the encoder.  Get it?  You really gotta keep your wits about you.  If I’d been smart, I would have first marked all the keys on their backsides so I didn’t have to constantly flip the keyboard over and check which key I was connecting to.  Further, as you connect more keys, the wiring gets progressively messier and more dense, which makes it harder to tell where you’re going.  An even smarter strategy might have been to use color coded wires to make seeing the links easier – but I was determined to use the keyboard’s original wires to keep as close to prototypical era as I could.

In the above pictures you can see this taking shape.  To make things a tad easier, I did resort to using a Dymo labeller to label the wires that end up connecting to the encoder.

Finally when this was all done, I connected each wire to the appropriate pin on the encoder.  I declined to use a connector for this, which would have allowed easy removal.  I just didn’t have one and didn’t want to wait.  I figured this would be stronger anyway.  Here’s the final result:

I can clean this up a bit for the install, but yeah.  What a spaghetti mess-o-wires!

I regret I don’t have a video to show the testing.  I don’t have a tripod for my camera.  Essentially on testing we had some fairly positive results, however as expected some keys turned out to be wrong.  What I had expected to happen had in fact happened: I had gotten some keys mixed up.  In some cases I goofed and put a wire that should have been on the left post to the right instead.  In others I simply mixed up which key belonged on the 3 digit binary line, which confused the encoder.  Naturally I didn’t catch this of course until I had desoldered and resoldered the keyboard connections to the encoder three times, convinced I had an undiagnosed encoder problem.  But eventually I figured them out, traced the wires and connected them to the right places, removed some solder bridges, and now it works perfectly!  Wow!  So chalk up a successful project there!  This gives me hope we’ll have good sailing during the actual TVT build.