Why not better composite video like C64?

[llopis]

Hi everybody,

I've read almost every thread talking about the s-video mod and I understand its advantages. However, I don't have any s-video devices and I'd rather enjoy the VIC-20 on my Sony 14" CRT like all my other retro computers.

So having said that, is there a reason why the composite video output can't be more like the C64? I see that both the VIC and VIC-II chips have similar kinds of outputs (sync and luminance on one pin, color on the other), and yet the C64 manages to generate really good a composite video signal. Is there any technical reasons we can't duplicate the circuit that combines those two signals to achieve similar results?

Just in case it matters, in my particular case I'm talking about PAL systems.

Thanks.

--Noel

[Mike]

Hi, Noel,

for sure it should be possible to improve upon the video circuitry for a better composite signal.

Commodore made lots of changes to that circuitry over the course of time. Just take the time and look at the differences between the video output circuitry for the two early ('2-prong' power plug) and the later CR (DIN power plug) mainboard revisions! The original version is much more complex and derives the setpoint value for the amplifier stage with a diode. It also supposedly has a better video quality than the later CR variant.

In the CR variant the chroma component is overmodulated (more or less) and this is what tears the pixels apart.

For most people, applying one of the several S-Video mod seemed to be easier than redesigning the video stage. You could still try doing the S-Video mod and then externally mixing Chroma and Luma again, by a capacitive coupling, with C=220..470 pF.

If you make a redesign of the video stage and find a way to put it into the RF box of the video chip, that would be nice.

Greetings,

Michael

[llopis]

Hi Michael,

I thought of doing the combination afterwards, but I was convinced it would be worse than the dedicated circuit the VIC 20 has inside. But maybe not! I'll try that.

Also, technically, I have an adapter for composite to SCART that has an s-video input, but I suspect it does the same thing and just combines the signals into a composite one (just guessing since I don't have a single s-video source, they're not that common in Europe).

I'm planning on making a video about this whole process, so I'll definitely share what I end up doing.

Thanks.


--Noel

Hi, Noel,

for sure it should be possible to improve upon the video circuitry for a better composite signal.

Commodore made lots of changes to that circuitry over the course of time. Just take the time and look at the differences between the video output circuitry for the two early ('2-prong' power plug) and the later CR (DIN power plug) mainboard revisions! The original version is much more complex and derives the setpoint value for the amplifier stage with a diode. It also supposedly has a better video quality than the later CR variant.

In the CR variant the chroma component is overmodulated (more or less) and this is what tears the pixels apart.

For most people, applying one of the several S-Video mod seemed to be easier than redesigning the video stage. You could still try doing the S-Video mod and then externally mixing Chroma and Luma again, by a capacitive coupling, with C=220..470 pF.

If you make a redesign of the video stage and find a way to put it into the RF box of the video chip, that would be nice.

Greetings,

Michael

[nbla000]

If I'm not wrong there is something already done for that on the web.

[llopis]

If I'm not wrong there is something already done for that on the web.

If someone find it, please post a link here. I'd love to check it out at some point. I'll probably do the s-video mod anyway just to test, but it would be nice to have options.

By the way, is it normal to have different VIC chips have pretty different video qualities on the same board?

The first VIC chip I tried, part of the problem is that multicolor mode wasn't working, so some games were just unplayable (I saw at least one other post about this, so I'm guessing that's not an uncommon fault).
I tried a different VIC chip and it had multicolor working properly, but letters were much less crisp with some slight ghosting.
So I tried a third one, and that one looked the best: crisp and multicolor working.

Do I just have horrible luck, or is it normal for those chips to have those variations? A bit like the C64 SID chips maybe?

[nbla000]

Clear video for Vic-20, there is also for C64/C128
https://www.etsy.com/listing/775553994/ ... -modulator

And there is an universal version too:
https://www.etsy.com/listing/789437205/ ... -filtering

[llopis]

Clear video for Vic-20, there is also for C64/C128
https://www.etsy.com/listing/775553994/ ... -modulator

And there is an universal version too:
https://www.etsy.com/listing/789437205/ ... -filtering

Interesting! Those seem to be specifically for RF signals, but there's no reason it can't be used for composite out I imagine.
Thanks!

[tokra]

I saw those some time ago, but do not understand what they would do for the VIC-20. It claims "better than S-Video" and it seems there is a S-Video output on this one as well. Does someone know more about this device and what it does EXACTLY? I am looking for the best possible video-output for a project I am working on, so this might be worth a try...

[llopis]

I did a couple experiments and I have some interesting findings (not what I wanted but hey).

First of all, this is the output of my VIC 20CR unmodified (just with the mods that came from the factory):

So I removed FB7, added a 0.1uF capacitor and hooked that up directly to pins 4-5 so it gets combined with the luminance. I figured that's probably what an outside s-video to composite circuit would do since that's what I would have to do anyway. And this was the result:

The crosshatch pattern is gone, but now I have some serious color ghosting there! It looks like the chrominance signal isn't in sync with the luminance signal. Maybe that's because the luminance one through through a couple more stages and the chrominance one just skips all of that? Not sure. I tried changing the 0.1uF cap for a 470nF cap and it made pretty much no difference either.

So then I hooked up an s-video cable directly to the board to avoid cutting any tracks (0.1uF + 100 ohm resistor) and then plugged in the s-video cable into my cheapy s-video/composite to SCART adaptor and I got this:

Wow! That's much, much better! I'm pretty sure that my Sony 14" Trinitron CRT doesn't support s-video internally, so the cheap adaptor must be generating a composite signal, but somehow it does it a lot better than what I could do directly on the board of the VIC 20! I wish I knew what it did so I could keep the composite out.

Unless someone has some suggestions on what to try to generate a better composite, the final quality improvement is significant enough that I'm going to make the s-video mod permanent.

[Mike]

The crosshatch pattern is gone, but now I have some serious color ghosting there! It looks like the chrominance signal isn't in sync with the luminance signal. Maybe that's because the luminance one through through a couple more stages and the chrominance one just skips all of that? Not sure. I tried changing the 0.1uF cap for a 470nF cap and it made pretty much no difference either.

I mentioned a capacitor with 220 or 470 picofarads to connect Chroma to Luma to form the Composite signal. You used nearly 1000 times that value, which at the carrier frequency of Chroma acts like a short - which you don't want.

Anyhow, the last picture is about as good as you could expect with the S-Video mod. Note the white 'frames' around the characters are likely the result of picture sharpening circuitry inside your Sony. Also, blue on white is a rather uncritical colour combination to begin with. The cross-hatch pattern at the left and right edge of the display window (white to cyan) is gone. That's far more important.

Here's a program to test all combinations of foreground/background on a single screen: https://dateipfa.de/.Public/denial/tool ... r_test.zip (for unexpanded VIC). Could you provide a full format screenshot of the output (preferably with the checkerboard pattern - press SPACE to switch the patterns)?

[llopis]

I mentioned a capacitor with 220 or 470 picofarads to connect Chroma to Luma to form the Composite signal. You used nearly 1000 times that value, which at the carrier frequency of Chroma acts like a short - which you don't want.

Right. I was going by the s-mod described here, that's why I tried 0.1uF: http://sleepingelephant.com/denial/wiki ... deo_output
I tried 0.1uF directly into luminance, and 0.1uF + resistor + 470pF but they were identical.

Anyhow, the last picture is about as good as you could expect with the S-Video mod. Note the white 'frames' around the characters are likely the result of picture sharpening circuitry inside your Sony. Also, blue on white is a rather uncritical colour combination to begin with. The cross-hatch pattern at the left and right edge of the display window (white to cyan) is gone. That's far more important.

Yeah, it looks pretty good. It's just a shame that I couldn't get that from a composite out signal. But oh well, I'll take it. It's not that clear in the static pictures, but not having the crosshatch pattern makes games SO much better visually!

Here's a program to test all combinations of foreground/background on a single screen: https://dateipfa.de/.Public/denial/tool ... r_test.zip (for unexpanded VIC). Could you provide a full format screenshot of the output (preferably with the checkerboard pattern - press SPACE to switch the patterns)?

Ah, great! I was looking for a color test program. I'll try that one and report again. Thanks.

[llopis]

More interesting findings!

Here's the color test pattern with the unmodified video:

And here it is with the s-video out (going into that SCART adaptor):

Clearly much better.

BUT... I did something completely by accident while I was trying the different configurations and I got this through the regular composite out:

And here's a detail of the BASIC prompt to compare to the earlier screenshots:

That looks even better than the s-video signal!! There's a hint of the crosshatch pattern along the left edge, but the characters are sharper and without any ghosting!

So, what was this accident? It was this!

I put a 0.1uF cap on the chrominance signal connected to the far side of the adjustable resistor, AND a 470nF cap going directly to the composite output. So it's adding BOTH chrominance signals, the one that was processed along with the luminance and the one coming out straight out of the VIC 20!

It wouldn't have occurred to me to do that, but I accidentally left the 0.1uF cap connected when I connected the other one and I was shocked at the results.

Does anyone have a good insight as to why this results in such a good image quality?

[Mike]

[...] That looks even better than the s-video signal!! There's a hint of the crosshatch pattern along the left edge, but the characters are sharper and without any ghosting! [...]

That mostly serves to show how bad the original circuit really is in producing the composite signal.

With older CRTs and the S-Video mod in place, there's often a bit of "colour bleed" to the right (about one hires pixel) and with PAL, also to the next line - the latter comes from the 64 µs delay line PAL uses to weed out phase errors that on NTSC lead to hue changes with terrestrial broadcast. The horizontal bleed happens because the decoder takes at least a full period of the Chroma signal to fully detect a phase change. Newer CRTs try to compensate for this by delaying the Luma signal for a few tenths of nanoseconds, so both are mostly in sync.

Your two-capacitor arrangement seems to nudge the delay between Luma and Chroma more in the direction suited for your Sony, indeed.

If you have a +24K RAM expansion at hands, try out this picture: https://dateipfa.de/.Public/denial/tools/md.zip (and maybe also post a screenshot of it?) - this one really needs some kind of improved video circuit.

[Noizer]

More interesting findings!
[...]
That looks even better than the s-video signal!! There's a hint of the crosshatch pattern along the left edge, but the characters are sharper and without any ghosting!
So, what was this accident? It was this! [...]

Amazing discovery! This adaptation certainly means that one don't actually need an S-video capable input / device, but that the composite connection is sufficient. This would mean a lot for those who did not dare to apply a modification until now.
It would be really great if you could give a summary of how to modify it in a post.
BR

[llopis]

If you have a +24K RAM expansion at hands, try out this picture: https://dateipfa.de/.Public/denial/tools/md.zip (and maybe also post a screenshot of it?) - this one really needs some kind of improved video circuit.

Here are the results. And since you mentioned that it could be a result of just syncing with whatever the Sony CRT needed in particular, I tested it with an LCD as well.

Are those horizontal lines correct? I keep getting them consistently, so I imagine it's some side effect of the way the graphic is displayed (either that or I have another faulty VIC chip).

S-video on the LCD. Blargh! That looks horrible. The smearing of colors is VERY noticeable on the LCD.

S-video on the CRT. Much better.

My composite mod on the LCD. OKish.

Composite mod on the CRT. Looks the sharpest of all of them, although I don't know what it's supposed to look like.

Just for completeness, the startup screen with S-video on the LCD:

And with the composite mod on the LCD:

That looks pretty good even if I say so myself!

I was doing all of this as part of a video for my channel, so it'll all be documented there. I'll post with a link here and I'll also post a final picture of the circuit and an explanation of how to do it for reference.

Cheers!