So with trepidation I install one of my precious 2518 line register ICs.  I have several of them, all dated to 1973.  I got very lucky finding these with acpsurplus on ebay.  I grabbed all they had, figuring there would be some attrition due to age and some mistakes on my part.  I just need *one* of the eight I’ve bought to work.  Can’t be that hard, right?

Anyway, I put it in there , and follow the directions in the Construction guide to connect my 330ohm ‘jumper’ to each of the bus ‘ports’ (pins) on the memory board to see if I can produce the same characters I was producing earlier by connecting the jumper to the resistors.   The difference here is that where before I was simply forcing certain bits high or low on the 2513 and thus it was spitting out whatever character the resulting ASCII code produced, this time I’m actually invoking the line register and getting the TVT to ‘store’ these characters for eight lines at a time.

I confirm with my jumper wire that I am getting… nothing.

None of the bus lines I probe work.  Something is wrong.  I try some elementary probing around trying to understand what is (or isn’t) happening.  The best I can come up with is that either my 2518 is bad, or we have perhaps a short or something not connected.

I decide to remove the Memory board and do some more cleaning between IC pins and such, to make doubly sure we have no accidental bridges.  When I plug everything back in and power up again, this time I have a screen full of < symbols instead of @.    If I remove the 2518 line register, it goes back to @.    As I fiddle around with the board, testing continuity, making sure the sockets are ok, each time I power up things change a little.  Sometimes I’ll get columns full of letters.

After scratching my head, I turned to Chuck on vcfed.   Since we know the line counter is derived from the outputs of IC5 on the timing board (a Signetics 8288), we should be seeing pulses on pins 2, 9 and 12.  We should also see signals on IC7, pins 3 and 8.

I decide to do a general checkover of the Timing board and discover that the socket at IC9 is bad.  I was supposed to have a pulse at pin 17 but did not; with the socket replaced we’re good.   But we’re still not getting any joy – pin 4 on the 2518, which should have a pulse, is solidly high according to my logic probe.  Why?

It turns out that the problem is pretty elementary – bus pin 17 on the memory board was not making direct contact with the attached trace.  Pin 17 is the line 1, 13, 25 transfer pin, so without a connection there we’re not going to see much.  I’m flummoxed as to how this could be, since there was no obvious break – it looked like the pin was solidly in there.    Anyway, I correct this and solder the crap out of it, and now we’ve got a connection, and now we’ve got some action when I connect my jumper.  Sure enough, the characters displayed onscreen are changing into different letters as they should be!  Wahoo!

So that’s a good result.  Now comes the bear: I need to create some jumpers that connect to various clocks on the Timing board for the next test.  Gingerly I take the unit apart and jumper in a few of the clocks recommended.  When connected to B1, they should produce different ‘stripes’ of characters on the screen.  For example, connecting clock ‘R’ to B1 should give me a screen that is half @s and half blanks.  The Q clock should give me two vertical columns, or ‘stripes’ as the guide calls them, with blanks in between.  That seems to be what I get, except I’m getting A’s instead of @s in some places.

I’m not sure what the deal is there, but I decide to persist a bit further and try connecting the 0, I, H, G, F and E clocks to B6 through B1 respectively.  This should produce a full screen of all the characters the 2513 can produce.  I connect, power up and:

Cool!  I mean, not quite what I should have, but we’re definitely getting there.  Still gotta address those vertical bars in my character output, but I can see we’re a good chunk of the way there.

Unable to find a problem with the IC sockets or anything amiss with bridged IC pins or broken board pins, I decide to swap ICs around to see what happens.  The TVT uses 4 8288s, so swapping them around in theory shouldn’t produce any change if they’re all working.  However, when I swap two, I do get a slightly different pattern.  I also notice that some characters in the middle of the screen are changing on the fly.  After subbing in another 8288 and swapping around with no change, I decide to look at the 2518 on the memory board.  On a lark, I swap it with another that i have.  Sure enough, the screen changes.. but it’s worse.  So I grab yet another.  I’m getting ABC as I should but then it all goes bad from there.

I try subbing in another 8288 and do another 8288 shuffle, and now I’m very close:

I really want that vertical line gone though.  Chuck suggests checking IC10 on the memory board – a 74165.  Chuck explains that since the characters are being formed on a 5×7 matrix, the 74165 is programmed to add two blank dots on one side of the character and one on the other.  Since the vertical lines appear to be happening on the 4th column of dots, we trace that to output pin 2 of the 2513, which feeds into pin 14 of the 74165.  He suggests something here might be amiss, maybe a short, and it’s causing the 74165 to pass on an extra vertical column of dots that shouldn’t be there.   Sure enough, with a bit of hunting I realize that a jumper that connects pin 14 of the ‘165 to the 2513 output hadn’t been properly soldered!  With that fixed, voila:

That’s almost perfect.  Granted, we still have an @ where there should be an H.  Why is that?  Everything looks good.  Chuck suggests that without a logic analyser or ‘scope, it’s going to be a real challenge to sort out.

I spend a full week trying to hunt it down, but to no avail.  Despite checking and rechecking the boards, the ICs, the sockets, jumpers… everything… I just can’t find it.  I speculate on causes and try my best to interpret what is going on.  I even consult the guru himself, Don Lancaster.  He gives me a few cryptic suggestions but nothing seems to change matters.   I finally decide, on Don’s advice, to leave it aside for now and move onto building the Cursor board.  We’ll have to hope whatever this is is just a problem with this particular test, and not a real problem with the unit.

Alright!

In my last post, I finally got my TV Typewriter to actually display something onscreen!  It wasn’t much, but seeing 16 rows of 32 well formed boxes on my screen made me feel like I’d reached low earth orbit, and was now on my way to the moon.

According to the TVT construction guide, the next step in our journey is to add the fabled 2513 character generator.  This is exciting because this is the point where the TV Typewriter actually starts putting recognizable stuff onscreen!  The instructions advise strongly to check all the power pins.  These ICs were rather exotic in their day (and pricey), and frying one in 1973 did not mean a quick call to Mouser.

Applying what I’ve learned so far, I check all the voltages and they’re all good.  I also apply extra caution and check all around for shorts caused by bad soldering or solder flux.  Where I see shiny flux where it shouldn’t be, I use a tiny jeweller’s flatblade screwdriver to scrape it out.  Not the ideal way but effective.

The construction guide advises the next step is to plug the board into the other two TVT boards, connect TV or Monitor, and fire it all up.  If all is ok, I should have a whole screen of @ symbols (this is the default binary code the address lines produce if nothing is trying to change them).  And we do!

Woohoo!  Now, I do see one little problem – there are some pixels activated that shouldn’t be.  I’m reassured by more experienced hands that this is likely just a short somewhere.

Still, the amazement isn’t wearing off.   This is a fairly complicated electronic device and I’ve got it working!  So many TVT builders before me did not even get past the planning stage, let alone get (mostly) working video output!

Ignoring the extra pixels for now, I now craft a ‘330 ohm jumper wire’ by attaching a 330 jumper to one end of a solid core #22 wire.  The goal here is to attach one end of the jumper to one of the +5V ‘ports’ on the bus, and then briefly touch the other end to the ‘signal end’ of each resistor tied to the 2513’s address lines.  The result should change the screen full of @s to As, Ds, Hs, Ps or blanks.  At least, according to the manual.  I try this out, and it sort f goes as I’m expecting, except when I connect R49, I get a B instead of P.   About 30 min of fooling around ensues.  I’m sending off what I’m getting to my vcfed friend Chuck, and he’s suggesting I’ve got something backwards, or that possibly the documentation is in error.  Eventually we figure out that B is what should be produced, not P, and that I’m suffering from a bit of PCB dyslexia and mixing up which pins are which on the 2513.  Once we account for that, we’re all good!

The next step according to the manual is to add the 2518 shift register.  This another rare IC, especially one dated to 1973, and I’m a bit nervous about plugging it in.  But after checking and rechecking, I go for it.  Now comes the real test – jumpering in several ‘clock’ wires from marked points on the timing board to bus ‘ports’ B1-B7, to try and get the TVT to produce a full screen of all available characters.  This should be interesting!

Having successfully gotten (I think) a tune into the TV using the TVT’s sketchy RF modulator, the next stage is to put some actual integrated circuits to work and see if we can get this this puppy to do something interesting.

The construction guide advises that I need to completely build up the Timing board and part of Memory Board A (Page A), just enough on the latter to try and get something onscreen. Here’s the Timing board all dolled up and ready for action!

At this stage of construction, we are not yet adding the character generator, just IC 10 (a PISO generator) and IC 11 (an open collector NAND gate). First, the TVT should establish 32 columns and 16 rows of boxes as the field where text would appear. Then, with everything sort of floating open, it should set each of those boxes to be fully on, meaning we’ll have 512 of them on screen together if it all works.

After a few very careful checks for solder bridges, I stack the two boards into the motherboard, and go for it. And….

Nada.

No boxes, nothing onscreen. I check pin 20 on the system bus (the composite video out pin) for voltage, and it has it. I then decide to connect a composite monitor directly and bypass the TV for now, just to be a bit more certain that what I’m not seeing isn’t due to the RF modulator playing games. But there’s nothing there.

On the advice of Chuck from VCFED, I start digging into the timing board circuitry. First, we want to make sure there’s an actual clock signal happening, otherwise nothing at all will function. We’re looking at the MC4024 ‘dual astable’ chip used in conjunction with the crystal to create our system clock. Based on pinouts and the TVT schematic, I should be seeing something on the XTAL (crystal) pins of that chip, pins 3 and 4. But all is silent. My logic probe indicates no pulse activity at all.

Mystified, I start probing around the chip with the logic probe. I’m really scratching my head. Is it a bad chip? Maybe.. these things are 40+ years old and failure is always an option. I swap one of my precious spares in. Same thing. I swap the other spare in. Nope. So I’m really scratching my head here trying to figure out what the heck is going on. Chuck’s concerned I might have gotten the wrong 4024 – there is a CD4024 made by TI and Fairchild – similar part number, but very different chip. I verify mine are Motorola MC4024s. Did I get a bad bunch? Certainly possible. But not likely. I get back to basics and look for voltage at pins 1 and 14. Without adequate power (+5V), nothing much is gonna happen. I find 5V at pin 14 but *not* at pin 1. Hrmm.

It turns out after an hour of puzzlement that part of the answer is at hand. As I look more closely, I realize that I have goofed! I failed to solder in one end of a jumper (dang blasted jumper wires!) that should have connected 5V to the trace leading to that pin. Soldering it in, I now have 5V.. but still nothing on screen when I power up. Argh! Thought that would do it.

More headscratching and investigation ensues, before it dawns on me that my IC socket might be a problem. Thus far, I have been conducting my voltage test from the socket pins on the backside of the board. Now it dawns on me to check from topside. Sure enough, although voltage is getting to the socket pin, it is *not* getting to the actual pin 1 of the MC4024. Aha! After testing things a bit, I realize something is broken in the socket. This was a hazard of using vintage sockets of dubious provenance. To test things out, I connect pin 1 and 14 via a jumper, and now I have something on both my monitor and TV:

Yay! Sort of. What’s onscreen *kind of* looks like boxes. But not nice, bright white boxes. It’s the same on composite… so it’s not an RF modulator issue. I’m going to get rid of this socket and replace it obviously since something’s borked on pin 1. I’m advancing, an inch at a time.

I keep probing around and find more problems. Pin 1 of IC9 (a 7402) on the timing board isn’t connected properly. I resolder that and now I have a signal there I didn’t, but still no change on screen.

Next I notice pins 3 and 4 on IC9 aren’t showing anything. The soldering looks ok and socket itself seems ok, so I swap to another 7402. Now i’ve got activity on those pins, but still no change onscreen. Agh! So frustrating!

Chuck suggests that I need to check pin 45, the character clock pin. I check it, but there’s no pulse there, so that indicates no dots are getting shifted out to the screen.

His next suggestion is to really check out those Signetics N8288s. These are very old divide by 12 counters that were a Signetics-only part and went extinct not long after their introduction. Chuck warns that the manufacturing processes used on these were not quite dialed in back in the early 70s, and it’s quite possible these have just degraded and died over the course of 40 something years.

I probe these ICs carefully following clock signals around. The probe’s speaker changes tone and the pulse speeds up or slows down depending on where I’m probing, as the clock is being divided progressively further and further. I’m scratching my head again – it looks like the 8288s are working just fine. However that proves to be incorrect. One of the 8288s, in IC2, is getting an inconsistent, flaky signal at pin 5 and 6. I notice if I press on the pin itself with my logic probe with a little bit of pressure, I get signal. As soon as I let the pressure off, nothing. I decide to (again) replace a socket, hoping the replacement works, as my stock of vintage sockets is dwindling fast.

That doesn’t solve the problem although now those pins are getting a much better signal. I keep digging around and eventually find IC9 is still not working correctly. I pull the board out and shine a bright light from on side so I can see the traces clearly on the other. It turns out I’ve accidentally bridged pins 2 and 3!

I make the correction, inspect the board one more time, and then do a literal Plug’n’Pray hoping I’ve found it. To my amazement and delight, this happens!

The RF modulator has dialed itself out a bit again but it’s clear from what I can see that we’ve finally got something to light up on screen. Connecting to the much clearer composite, it’s confirmed! I now have my boxes!

Wow. That was epic!

So I’ve learned a few things. One, if you’ve built it yourself, assume ‘user error’ is at fault before assuming your ICs or components are the issue. With traces, IC pads and the like packed in so tightly together, it is very easy, especially for a novice like me, to accidentally connect things that aren’t meant to be connected with solder, or not even solder them at all! Double, triple and quadruple check your work. Employ some form of illumination under your boards to really verify that you’ve got everything in order.

Further, be wary of sockets – especially vintage sockets. Even if unused, the quality of these varies and they do degrade over time, depending on how they are stored. Never assume because everything looks okay that it is. Check each and every pin on your IC with the pins on your socket and make sure they are actually connected!

And of course if you’re lost, it’s good to have the ‘phone a friend’ option. The vintage computing forums and email lists are fantastic for this. Everyone is so helpful and kind (and understanding!). Chuck’s expertise was invaluable here as it told me what to look for. It’s all a learning process.

I’m so impressed this thing is actually doing something! I honestly didn’t think I’d get this far. Okay, on to the next step!

Okay so now comes the scary part.  The TV Typewriter has its own built in DC power supply, connected to the wall by a line cord soldered into the board.  That’s a little intimidating for someone who, until recently, had never built anything electronic in his life!  We’re talking live, exposed AC connections.  One wrong move under power and.. ouch!

The goal for this first major step in construction is to install the transformer, protection diodes, and any other circuitry related to the power supply.  We then want to plug in and switch on and do some voltage tests.

Now, I’m going to confess, I am no electronics engineer.  I did not understand the transformer wiring at all.  I got some help from friends on vcfed.org with this, as it is not as simple as just connecting the post marked 12V to the PCB point marked 12V.  You have to tie certain posts together and then connect them, per the schematic.  And prior to that, they had me do some quick voltage tests, to make sure Signal had labelled things correctly in the first place.

Anyway, this is what the completed wiring looks like:

Another problem that tripped me up was the orientation for diode D6.  Orientation of diodes here is critical – the diodes provide protection for the system and make sure different voltages aren’t crossed.  A mistake here could cause damage to the whole unit!

Complicating matters further, this is one of the (many) examples where the schematic is at odds with the PCB layouts themselves.  More experienced eyes than mine looked at the schematics and were convinced the orientation of several diodes were wrong.  However, the errata provided after the construction article came out suggested that only D6’s orientation was shown wrong on the PCB silkscreens.  Indeed, the scanned copy of the construction guide that I was relying on initially had a correction from the original owner:

After much debate on the forums, I decide to go with the errata and only change the orientation of D6.   Although I usually defer to the experts, everything I have read indicates that the PCBs are correct, minus this particular error.  So I install all the pieces.

One other note: I took this opportunity to correct a situation I wasn’t comfortable with.  In trying to keep the look of the ‘new’ TVT vintage, I had chosen some Mallory 5000uf caps.  These are big silver can caps rather than the coloured ‘sausage’ caps common today.  The caps I had picked had little metal tabs on them, and to make it fit on the mainframe I had been forced to rig the tabs up with wire and then solder the wires into the board.  They looked kinda sketchy.  Anyway, I later found some screw top GE 5000uf caps, which I was able to install much more securely.  They are silver also and look perfect for the role.  So that’s what you see installed in the photo.

With everything connected, I decide to go for a test.  I’m more than a little nervous here and go a bit crazy, setting up the unit on the concrete porch outside my (steel) door.  I’m going to set the power switch to on and connect the cord behind the door.  That way if it goes snap, crackle, pop, I’ve got more than adequate protection!  I realize this is of course a little paranoid, but this is my first time messing with something that uses live AC power that I  built.

Anyway, after checking and rechecking, I plugged the AC line cord in, braced for explosion.  But none comes!  I wait a good 5 minutes.  Nada!  No smoke, no fire!  No scary crackling sounds!  Woohoo!

Getting more brave, I pull the unit into the house and onto my old tile floor.  I plug in again, still standing a few feet away.  No problem!  Now I man up, and grab my DMM to check voltages.  According to the manual, I should have 5V on pins 58 and 59 of the ‘bus’, -5V on pin 57, -12V on pin 56, and +12V at a spot on the mainframe board set aside to provide keyboard power.  My results are very close!  +5V on 58 and 59 exactly (thank you 7805 voltage regulator!), then -5.45V on pin 57, -12.4V on pin 56, and +13V on the keyboard +12V point.  A little out of spec, but given the machine is not under load and that the variance is so small, the power supply is given a clean bill of health!  Yahoo!

I know this is kind of a minor thing – after all this is just a power supply at this point, but the fact that I built A POWER SUPPLY THAT DIDN’T BLOW UP is kind of awesome nonetheless!  Thank you to all those that helped along the way!

Following the instructions of the TV Typewriter construction guide, the first thing to do after recreating my TVT boards (again and again) is to drill them out and work on aligning the stack connectors.  I knew this was going to be a problem from the get go — to test alignment as well as make sure scale wasn’t being lost during scanning or printing, I actually printed the artwork onto transparencies and then compared to the originals in the guide.  The issue I ran into was the stack connectors – the ‘bus’ that the four TVT boards connect via.  A quick study of the alignment of these using the transparencies revealed that between boards, they didn’t quite line up.  I supposed this could have been due to 40 years of humidity acting upon the paper of the guide, although I also have two copies of the Mark-8 construction guide and there is no difference between the artwork in those dimensionally or otherwise.  It could also simply be errors – I had read some contemporaneous accounts of TVT builders that complained of issues with alignment.  Anyway, after carefully considering the issue, I decided I could just ‘make it work’ by carefully adjusting the pins as I installed them.  Altering the artwork seemed like a fools errand – it might fix the connector problem but risked distorting everything else.

Anyway, after making the boards, I’ve drilled them and began removing the Molex connectors from my first, erroneous board set, and installing them on the ‘new’.  I also installed my new transformer, as well as the control switches (bottom) for test fitting.  So far so good, although the transformer sits a little too close to the board stack.

In terms of alignment vertically.. yeah.. it’s not pretty.  But I’ve confirmed it’ll work.  Now to solder in all 24 molex connectors and 240 pins (again.. ugh)!!

Okay, so when we last left off, I had discovered an unfortunate mistake in the original PCBs I had recreated for my TV Typewriter project: they were too small, a victim of the distortions that happen when ‘line art’ is scanned.

It’s all good.  As I mentioned also, I had managed to run into some actual vintage 1973 board stock.  Now re-doing my work was no longer a matter of desire – it was necessity, just to get that inch or so closer to a full replica.

If you read my post about my 2-5-2-2 toner transfer process you’ll see the first board I worked on – a redo of the TVT mainframe board.  But the TVT gremlins weren’t going to let me off that easy — it turned out, after etching, that it was still the wrong size!  Apparently printing onto magazine paper causes some kind of weird dimensional loss.  I suspect it may be because the inks used to print on magazine paper are water based and perhaps the heat from the laser printer causes a bit of that to boil off and thus shrink?  Not sure.  Anyway, after some trial and error I figured out the loss was about 1% all around.  Adjusting for this and reprinting gave me the right dimensions, and finally the correct size board.

Now, I mentioned before that these vintage copper clad sheets, which I was lucky to find, were only 0.03″ thick.  That’s a problem, because most PCBs commonly are around 0.06″ thick.  And we want that thickness, because the TVT has some heavy components, like the transformer, and we don’t want any flex.  So what I decided to do was etch my PCB pattern on one board, and then completely etch off the copper on another, and then cut to size and sandwich them together with epoxy.  This seemed straightforward enough, and so I trundled ahead — made my first mainframe board and then epoxied together.  And it looked good until  I noticed that bubbles were forming internally.  And no matter what I did, I couldn’t make them go away.  The result was a splotch effect that made the board look obviously glued together.

Nope.  Don’t want that.  It was suggested to me that I buy a book press.  Apparently you want to apply a huge amount of pressure to prevent bubbles from forming.   I couldn’t find one though – not one that wouldn’t incur a fortune in shipping fees from the U.S., which was the only place I could find them.

So I played around.  On the mainframe board, I actually ripped it apart (the epoxy is suprisingly easy to remove once exposed), cleaned and retried, this time waiting a little while before bonding the two boards.  This helped, but there’s still some discoloration.  I must have tried four different ways on five different boards before realizing there wasn’t much I could do.  My final trick was to switch to contact cement.  This made a world of difference – on my cursor board, it left it looking mostly like an original piece of 0.06″, save for a couple of splotches or ‘birthmarks’, as I like to call them.  I decided to accept that and move on.  A couple of boards I did – my Memory and Timing boards, I actually started drilling right after bonding, and this produced a kind of ‘craquelure’.   Here’s a picture of the ‘new’ timing board, next to the old one as I’m stripping the part.  Notice the ‘cracks’.

Oh well — over time they faded and now, frankly, I don’t notice them.  On my final board, the Cursor, I waited a full 15 minutes after applying the cement to both boards and then attached them.  This made for a much cleaner look, albeit with some ‘birthmarks’.  I decided to live with those.  This TVT will live inside a case anyway.  I doubt anyone but the most discerning will bother to note the splotches.   Alright, now to actually follow the directions this time and start building.  Here comes the mainframe!

This week I put in my order to Signal Tranformer for a new 24-1A transformer.  Believe it or not, despite the passage of 40 years, Signal will still custom make just about anything they previously made, for a price.  The cost of the transformer itself is about $103USD – not cheap, but according to Signal it will be identical in every way to the originals used in the 70s, which would be handy since the TVT instructions call for a 24-1A as the preferred unit.

The boards meanwhile are pretty much done.  Here they are (please forgive the lighting:

And stacked together:

Now I just have to hope that when I apply power, it works.  I should note that I kind of failed to follow the instructions – the instructions have you build the unit in a particular order in order to safely test things as you go.  For example, the instructions for the mainframe have you build and test the power supply before installing other components like the RF modulator.  This is to minimize the risk of serious damage if something is wired wrong.    I could help myself perhaps by dismantling those parts.. but I think with a good, thorough checkover (several times) I’ll feel pretty confident about plugging in the mainframe at least and working my way up.  Cannot believe I’m so close to being finished!  What a fun ‘little’ project!