Atari 2600 Jailbars After UAV Mod - Clean Up Output From TIA COL Pin?

[ctrl_alt_rees]

Hey all,

 

Something interesting to get your teeth into.

 

I'm modding my NTSC Heavy Sixer with the excellent UAV mod from The Brewing Academy. I'm really, really happy with it - in composite mode. I hooked up S-Video and while it looks great on a CRT, there are faint jailbars (regular vertical lines across the screen) being picked up on my capture device.

 

If I disconnect the "color in" from pin 9 of the TIA (COL), however, they're gone and I'm left with a perfect, crisp B&W image. I hooked up my oscilloscope to pin 9 and sure enough the signal is a mess. So I've tried adding capacitors of various values in various places but I've reached the limit of my (very limited) electronics knowledge.

 

I have already read through a few threads here with similar issues but none matched mine exactly - wrong system, different region etc so none of the suggested fixes applied in my case. I'd really love to get to the bottom of this as I'm working on some documentation and YouTube videos on various UAV mods. Of course the winning suggestion will get full credit!

 

Many Thanks for reading and I look forward to the replies. I'm very willing to try anything and everything at this point.

[+DrVenkman]

Is the second image composite or S-video? It honestly looks pretty good as-is. Most jailbars issues with UAV installation come from grounding problems. Often noise from other system components (DRAM chips in 5200's and A8 computers for instance). The signal you're getting from pin 9 isn't really messy - that's because of the way the TIA chip generates an image, literally line-by-line for every frame of the image in real-time. You might try looking at the same signal using a color test program, preferably one that shows the same color down the the entire screen at once. 

[ctrl_alt_rees]

1 hour ago, DrVenkman said:

Is the second image composite or S-video? It honestly looks pretty good as-is. Most jailbars issues with UAV installation come from grounding problems. Often noise from other system components (DRAM chips in 5200's and A8 computers for instance). The signal you're getting from pin 9 isn't really messy - that's because of the way the TIA chip generates an image, literally line-by-line for every frame of the image in real-time. You might try looking at the same signal using a color test program, preferably one that shows the same color down the the entire screen at once. 

Thanks for the reply. Interesting stuff.

 

Good idea about hooking up the oscilloscope while there's a solid colour on the screen, maybe it'll look a bit tidier. I appreciate that the 2600 is built on 1970s technology and was only ever intended to display over RF (especially as mine is an early Heavy Sixer) so maybe I'm asking too much - and if I am that's totally fine, I can live with it as-is.

 

I should've explained the images a bit better. Both images were taken from the 2600 hooked up to a RetroTINK-2X into my Elgato Game Capture HD. They're both s-video.

 

The only difference is that the second one was taken with the "color in" input on the UAV disconnected. It's just proof that the banding only appears once the chroma signal is introduced. To be fair you'd need to click to enlarge the image and then look *really* closely to see it - I've certainly seen a lot worse on other modded consoles.

 

Even the composite output is a huge improvement on the cheap ebay composite mod I had before but if I can improve the s-video just a little bit further I'll be happy as Larry as I'd like to get the best possible quality for game captures for my YouTube channel.

[+DrVenkman]

Okay, so I broke open one of my 2600's and took some scope readings (pardon the phone pics - didn't have a USB stick handy to do straight screencaps from the scope).  

 

Remember, pin 9 is the video Color data, which has to follow your local (analog) television signal spec. So I pulled up a color test program on my Harmony Cart and measured the output.

 

 

When you zoom out far enough, you'll see not just the clock-by-clock modulation of the signal for the color, but also the  horizontal blank and the vertical blank as well. But if you're familiar at all with how the 2600 draws on the screen, every bit of this is definable by the program running. The "kernel" used by the programmer is the heartbeat of the program, so to speak, and it has to conform to the basic limits of the analog signal the system was intended for (NTSC, PAC, SECAM ...). There's some wiggle room because analog TV's tended to be more forgiving than today's digital stuff. 

 

So here zoomed in far enough, you can see 16 repeating cycles on pin 9, followed by a period of COL going high again (I didn't measure that period). Were I to zoom out the horizontal scale, I would see the image you get on your scope in your first post.

 

 

That "mess" you see is the where your scope is trying to capture a signal that's not really following the same predictable frequency at the range you're capturing - think of a blurry photo; the exposure is too long to freeze the motion, if you get the analogy. And the COL signal is definitely not a simple, predictable single frequency. It's the NTSC colorburst frequency, being modulated to display the contents of whatever the program is trying to show on that particular scanline.

 

 

But zoomed out from my first photo, you can see the 15.75 khz horizontal frequency (my measurements are quick and dirty with manually-set cursors; I could've zoomed in closer and gotten more precise but you get the idea). Similarly, when I zoom out in the time base even further, you can see that this program's kernel seems to be creating a 59.9 Hz progressive image instead of a 29.97 Hz interlaced image. Not sure whether that's a measurement artifact because I'm connected to a 60 Hz flat panel screen or if I just missed a drop in the signal in the middle of all that data.

 

 

 

 

 

EDITED TO ADD SOME MORE STUFF:

 

So I got curious and set the test pattern cart to show different patterns. Take a look at this one:

 

 

Now look at the COL signal between that 15.75 khz horizontal trace:

 

 

Whaddaya know - 10 pulses on the scanline, each one corresponding to one of 10 vertical lines.  Now check the next test pattern:

 

 

And we see exactly what we should - a single pulse right in the middle of the scanline:

 

 

And finally, if you simply let the scope try to track the main frequency (the "messy" images above), you'll see exactly what you should: the base 3.57 MHz NTSC colorburst frequency that's being modulated by the TIA chip to produce the image, scanline by scanline down the screen:

 

 

Edited February 15, 2020 by DrVenkman
Add more scope traces

[ctrl_alt_rees]

10 hours ago, DrVenkman said:

Okay, so I broke open one of my 2600's and took some scope readings (pardon the phone pics - didn't have a USB stick handy to do straight screencaps from the scope).  

 

Remember, pin 9 is the video Color data, which has to follow your local (analog) television signal spec. So I pulled up a color test program on my Harmony Cart and measured the output...

Woah, you've really gone above and beyond, thanks so much for the detailed info! It's going to take me a while to process this... Still, the main takeaway for me is that my TIA seems to be "normal" and there's not a lot I can do about it.

 

As you can probably tell I'm new to the oscilloscope and in the process of teaching myself electronics. I tend to just hook it up and hit "auto" and see if the signal looks "clean" or look for obvious things like 15.75 / 3.57 MHZ (I'm in the UK but a lot of my consoles are NTSC). Definitely not using it to its full potential but it's still more useful than a multimeter when trying to work out if a signal is in the right ballpark. This is all really useful, thanks again!

[Danjovic]

The color component produced by TIA is a sawtooth waveform instead of  sinusoidal. Yo might try to add a tuned L-C filter (bandpass filter) centered at color carrier frequency but that solution can cause color artifacts due to inherent phase delay of the filter. 

 

 

[+DrVenkman]

If you want to have more fun with your scope, compare the several LUM outputs of the TIA with SYNC; or as here, the Chroma output of the UAV (yellow) versus Luma (purple) taken while displaying a color bar test pattern.  And with the Chroma signal hidden, look how wonderfully sharp the Luma level transitions are from the UAV.

 

 

 

[+-^CrossBow^-]

On 2/16/2020 at 6:19 PM, DrVenkman said:

And with the Chroma signal hidden, look how wonderfully sharp the Luma level transitions are from the UAV.

Which is why the B&W images are so clear! 

 

To be fair I think the jailbars are a game by game thing. Because most 2600 games I don't see these, but Pitfall is really one of the worst at showing them. 

 

BTW you see the same thing on a UAV modded 7800 as well. 2600 will still show some slight jailbars on some games with again Pitfall being the worst offender while 7800 games do not exhibit this. Pretty much shows it has to be an issue specific to how the TIA is generating the combined color signals. I think the only real way to not have this, is to RGB mod the 2600.