TV Typewriter Composite Out Issues

[AgentOrange96]

Hello all,
I am working on a TV Typewriter build using boards I got from @kalinchuk's store. Since I don't yet have a TV to use with it, I'm testing with just the composite output for now.
Because the exact crystal needed for the timing board is unobtainable, my timing isn't exact. I have a similar crystal in parallel with a variable capacitor. Where I expect to have 60Hz, I am getting 59.5Hz or 60.2Hz. I cannot seem to get anywhere in the middle, which I don't know why. I know NTSC color runs slow, but it's still closer than that.
I also ran into an issue where the composite output is too weak to drive a composite device. Thankfully, I have an old composite mod amplifier for an Atari 2600 laying around, so I put that on and it solved that issue.
Installing the Memory board, I see the following behavior:

On a more modern mini monitor (EM713) that has composite input, nothing happens.
On a Retrotink 2X Mini, I see the LED turn yellow (seeing a signal) momentarily, but it cannot keep sync. I got it to light up the above mentioned mini monitor over HDMI a couple times, but not for long enough to see anything.
Finally, I have a cheap backup camera monitor. On that, it consistently flashes the whole screen at about 2Hz or sometimes about 4Hz. I can't tell if it's the backlight or the screen turns white. But it seems like it's getting bursts of data.
Unfortunately, I do not have a CRT that takes an analog input right now. I feel like something dumb would be more willing to accept a signal.

The best info I have for the TVT including schematics can be found in the build guide.
I will need to set up a better work space and break out the oscilloscope to get a better idea of what the signal is doing. But if anyone has any other suggestions in the meantime, I'd greatly appreciate it! Thanks!

 

[AgentOrange96]

I hooked up the scope. It seems I see wave forms at 15.656kHz but at two different amplitudes. 1.94V and 3.53V. This is after going through the little Atari comp mod board.

I'm not sure if this is normal. I don't really know what I'm looking at. For reference, the scale is 1V/div with 0 offset. So the low sits at 31mV.

 

[AgentOrange96]

I ordered both an old composite monitor and a television. I feel like it may just be a timing issue and these digital displays are all too picky.

But also, the variable capacitor I have on the timing crystal has a minimum of 2.5pF. So perhaps it's going from 0-2.5pF and that's why I see the jump. I believe I have some capacitor kits around. So depending on what values I have, I may just experiment with my fixed value caps to see if I can dial it in. That'll have to be an activity for this evening though.

 

[AgentOrange96]

Bad news: I do not own any capacitors smaller than 2.5pF. So that's a no-go for now.

 

[Hugo Holden]

It appears that no attention was paid to generating a standard video signal output stage.

If the video output comes directly out of the 7401 Nand gate, shown on the schematic, with the resistors there too the output impedance of that circuit is too high to drive a standard VDU with a 75 Ohm input resistance. It would drive the input with it switched to unterminated, if there was that option on the VDU, sometimes there is a switch on the VDU near the input connector. But that is not as ideal because of the cable capacitance and the high frequency resolution would suffer somewhat.

If you wanted it to drive a 75 Ohm coaxial cable that was terminated at the VDU end in the normal manner with a 75 Ohm resistor (this gives best image detail) it would requide a transistor added in an emitter follower configuration to get the output impedance down. Or an OP amp buffer using an IC like the AD8056 would work very well if a + & - 5V supply was available.

 

[AgentOrange96]

It appears that no attention was paid to generating a standard video signal output stage.

If the video output comes directly out of the 7401 Nand gate, shown on the schematic, with the resistors there too the output impedance of that circuit is too high to drive a standard VDU with a 75 Ohm input resistance. It would drive the input with it switched to unterminated, if there was that option on the VDU, sometimes there is a switch on the VDU near the input connector. But that is not as ideal because of the cable capacitance and the high frequency resolution would suffer somewhat.

If you wanted it to drive a 75 Ohm coaxial cable that was terminated at the VDU end in the normal manner with a 75 Ohm resistor (this gives best image detail) it would requide a transistor added in an emitter follower configuration to get the output impedance down. Or an OP amp buffer using an IC like the AD8056 would work very well if a + & - 5V supply was available.

I did notice this early on. The Atari 2600 is also not really designed to output composite, so when you pull the composite signal off the board you typically use a 3904 based circuit for this reason, such as this one. I previously had one just like this in my 2600, but swapped it out for one of RetroSix's Clean Comp Boards which grab the video straight off the TIA, the chip that generates the video signal for the console. The result was I just had one of those 3904 based composite mod boards laying around. That was able to solve that issue! But that was prior to this post. I have made progress though, so I'll update in a new comment.

 

[AgentOrange96]

So I got a beautiful CRT monochrome monitor in! Unfortunately it sustained shipping damage. The plastic standoffs for the board broke, which allowes it to move and ripped off a ground wire to chassis and pulled off the main power connector. Thankfully that was the extent of it and I got it all fixed up!

After plugging it in, I got a signal off the TVT! I did manage to adjust the VHold to get it mostly not rolling, although it was super finicky. And the horizontal hold was clearly off. Well, eventually I realized why. The TVT timing boards has jumpers to set the vertical and horizontal locations. This sets when the sync signals are generated. I hadn't populated them. There were no sync signals. Yeah that'll do it!

After populating those, even the LCD was happy to display a picture! Unfortunately, I still have issues. After adjusting contrast and choosing the better responding of the two crystals I have (neither exactly matches what is called for) this is what I was able to get:



The LCD looks similar. Definitely progress, but definitely something is still wrong. I'm not really sure what though.

 

[AgentOrange96]

I removed the character generation chips again to get an easier to understand output. This shows just white boxes instead of '@'

What's interesting is that the LCD now is happier than the CRT. Adjusting H-hold on the monitor doesn't do much to fix this. But I do know the monitor displays fine on my IBM XT via the composite out port on the CGA card.




I have no clue if the issues on both displays are related or not.

 

[AgentOrange96]

I was able to confirm sync was perfect. However, blanking was not. It was getting noise on it. I took IC7 out on the timing board and still saw the weird signal going on telling me it was elsewhere. Removing the timing board removed the noise telling me the issue was on the timing board.

Looking at pin 21 where blanking should be, I saw two issues. One Didn't even affect pin 21 but was pretty glaring:




With the short cleared and the pins soldered, I had a signal on my LCD! The monitor was still unhappy with horizontal sync. I switched back to the faster crystal and...



I'm gonna adjust the jumpers for horizontal and vertical position. But this is working!!!

 

[kalinchuk]

This is awesome! Good job with the debugging.

 

[AgentOrange96]

Okay, so here's where I'm at.

First of all, I designed a case and had Protocase fab it. There are some bugs of my own design I'm working around, but Protocase did an amazing job and it's beautiful!


Now onto where I'm currently stuck: The cursor. I haven't hooked up a keyboard yet, but I put the cursor board on. Initially it seemed to flicker, but based on symptoms I was able to track down a short. With that fixed, I see the cursor! Twice...


Yeah, so I assume this is a timing issue. And in fact, I suspect if I did have the keyboard hooked up, it'd overwrite BOTH selected characters. But I'm struggling to figure out between the schematics, description and board what exactly is causing this. Unless I really do just need the keyboard plugged in? But I don't think so.

If anyone else has any ideas, please do let me know. Thanks!

 

[AgentOrange96]

I have installed the switches and the keyboard and it really didn't make much difference. Nothing does anything.

I'm focusing right now on the double cursor issue first. I'm really struggling with it. There are effectively two counters. The first counts to 32 and goes high every 32 clock pulses. This indicates our column. The second counter's clock counts to 16 and goes high every 16 pulses of the column counter. This indicates our column. The cursor signal is driven off the row counter.

Since the row clock is the column counter, if the column counter were high two cycles in a row in error, I'd expect the row counter just wouldn't advance properly. But it seems to be? But if the row counter were high two cycles in a row, we'd get the cursor on two adjacent rows rather than two adjacent columns. But perhaps I'm misunderstanding something here.

I was also wondering if maybe the cursor was alternating between two positions for some reason, but filming at 60FPS shows both appearing and disappearing on the same frame, so I don't think it's that.

Relevant schematic and excerpt from the build guide are here:

 

[AgentOrange96]

Okay, I misunderstood the counters. We can actually just think of the whole thing as a 9-bit counter that counts up to 512. The cursor appears if the MSB is 1.

However, this still seems wront because then the cursor would be on for half the screen rather than for a single count.

So I must still be missing something here.

 

[AgentOrange96]

I'm even more confused too. So the blinker signal is effectively used to disable the character ROM. It controls the enable pin on that. When disabled, character ROM just outputs 1. If we only do this for the top line, we get a line above the letter.

Except, I don't see how it's turned on only for that line either. It comes out of the 512-bit counter as the MSB and then is ANDed with a 4Hz signal to make it visibly flash.

As I understand it, half the screen should alternate between its given character and a white box. The other half should be normal. But that's neither the intended nor observed behavior. So I'm definitely lost here.

 

[syzygy]

I don't want to disrupt the flow, but when I read this,

First of all, I designed a case and had Protocase fab it

I looked them up as I had not heard of them. Mouth watering! Too afraid to ask the price, but did you use their CAD?

 

[daver2]

If you have an oscilloscope, can you take a look at the enable pin of the character generator to see what you are getting.

You may need to look at the 'start of frame' signal as a reference with a second oscilloscope channel as a trigger.

The 'blinking' should be slower than the frame rate.

I agree that the character should alternate with a solid box if the cursor is present. However, I also remember an option of a line or block depending upon a link. But, of course, my memory may be playing tricks with me...

Dave

 

[AgentOrange96]

I don't want to disrupt the flow, but when I read this,


I looked them up as I had not heard of them. Mouth watering! Too afraid to ask the price, but did you use their CAD?

It was expensive, but very well done! Yeah, I used their Protocase Designer tool.

 

[AgentOrange96]

If you have an oscilloscope, can you take a look at the enable pin of the character generator to see what you are getting.

You may need to look at the 'start of frame' signal as a reference with a second oscilloscope channel as a trigger.

The 'blinking' should be slower than the frame rate.

I agree that the character should alternate with a solid box if the cursor is present. However, I also remember an option of a line or block depending upon a link. But, of course, my memory may be playing tricks with me...

Dave

Unfortunately, I did not have time to get a scope on this last night. However, I'm looking into it tonight. But I have something better than a scope! (Hopefully) I have an Analog Discovery from my university days, which has a 16-channel logic analyzer. Now if only I had enough jumpers to use all 16 channels.



From IC4, we see that the counter is working properly. Interestingly, we're only seeing short pulses for the clock.


And after I also probed the cursor, we actually only see a short pulse, which while I don't understand, is what we want to see. I think.

 

[AgentOrange96]

Okay, I have a theory. I was really confused by how the output of IC4 was hooked up to the cursor. Why is there a capacitor between them? That's weird, right?Perhaps we're using Q3/Do to charge or discharge that capacitor. So the capacitor is ultimately pulsing the cursor on. Perhaps it's taking too long to charge or discharge. Scope time! But I already have the Analog Discovery out, so I'll use that.

Orange is Q3/D0 from IC4, and blue is Pin 9 of IC5, which is the other side of that capacitor:

So we do get a brief pulse, but it's very brief.

Here, orange is the cursor signal and blue is still Pin 9 of IC5:


What's interesting though is that now I'm only getting the cursor ONCE.

Moving IC4 back on the board and removing the logic analyzer, I see a double cursor again and this with the same scope placement as before:

 

[AgentOrange96]

I probed across the capacitor C2, and I get two out of phase signals going on.




Very weird. I do think something to do with the capacitance and timing is at play here, but I'm still trying to figure out exactly what.

That secondary waveform seems to be the trigger for the enable pin somehow, seen in blue below:


Checking the frequency:

That's suspicious. This must be the 60Hz signal coming from the timing board somehow since if it were coupling to grid power it'd be much tighter (whereas I know the crystal is slightly slower than spec) and it'd be sinusoidal. This actually has me wondering how difficult would it be to make a PAL TV Typewriter. Anyway, I'm not sure if this 60Hz signal is supposed to be here or where it'd be coming from. I'll have to go back and study this schematic a little more. Maybe I have something shorted on the board that I haven't found yet.