S-Video from the VIC

[rga24]

Regarding the loading on the chrominance output (pin 2) of the 6561, in the original composite video circuit in the VIC-20 CR, it is connected via a ferrite bead FB7 and a 220pF capacitor C13 to the wiper of a 1K preset resistor R10, which is also connected to a 47uF decoupling capacitor C17.

This is the point where the luminance signal and the chrominance signal are mixed, and it is also the only load on the chrominance pin. Because it is connected to the wiper of the preset, the resistive load depends upon the position of the wiper; it is possible to short the chrominance output to ground.

I worked out what position of the wiper would present the smallest load on the chrominance pin; with the preset at the 775 ohm position, the 775 ohms pulling down would equal the 775 ohms pulling up (245 ohms in the preset in series with 510 ohms in R9), so the effective load would be half that, i.e. 387.5 ohms. I choose the next E6 value lower than this for the load on the chrominance signal in the buffered version of the S-Video mod, hence the 330 R resistor.

[rga24]

By the way, in your CR VIC, do you have a 180 ohm resistor in the FB8 position on the board?

Hi a4000bear,

I just had a look inside a VIC-20 CR, position FB8 is actually occupied by a 240 ohm resistor (red yellow brown), which is marked on Commodore's circuit diagram as R41. It is connected to pin 3 of the VIC chip and to the R9 / R10 junction. This is part of the luminance circuit.

Do you have a 180 ohm resistor in that location? That would be like the two prong VIC-20's video circuit. I'll check my other VIC-20 CRs.

[Mike]

(Its the ratio between the level of the colour signal and the level of the colour burst that determines how much colour appears on the monitor.)

That sentence really didn't deserve to be put in parenthesis.

The data sheet mentions the chroma output of the VIC chip already *is* buffered, albeit with a high impedance. Unless the signal catches some extra noise, I'd then expect no different outcome regardless whether the signal is buffered for low impedance, or not: even if the voltage division at the 75 Ohm input of the monitor leads to a smaller received signal than normal, that gets compensated through the normalisation against the (also smaller) colour burst.

BTW, rga24, your buffer circuit adds a DC offset on the chroma signal. Shouldn't there be another coupling capacitor on the emitter pin to output?

Until now, I had deferred the S-Video mod on my VIC-20. But the cross-hatch patterns on my VIC-20 are now the quality limiting factor, when VFLI pictures are displayed, so this mod is the next one in the pipeline ...

[rga24]

The data sheet mentions the chroma output of the VIC chip already *is* buffered, albeit with a high impedance. Unless the signal catches some extra noise, I'd then expect no different outcome regardless whether the signal is buffered for low impedance, or not: even if the voltage division at the 75 Ohm input of the monitor leads to a smaller received signal than normal, that gets compensated through the normalisation against the (also smaller) colour burst.

BTW, rga24, your buffer circuit adds a DC offset on the chroma signal. Shouldn't there be another coupling capacitor on the emitter pin to output?

Until now, I had deferred the S-Video mod on my VIC-20. But the cross-hatch patterns on my VIC-20 are now the quality limiting factor, when VFLI pictures are displayed, so this mod is the next one in the pipeline ...

Hi Mike,

In practice I don't think normalisation actually occurs; that's why a two prong VIC-20 has a lower colour saturation level than a VIC-20 CR. Regardless of which chip type you use (6561E or 6561-101), the larger chrominance amplitude in the VIC-20 CR composite output produces more saturated colours and more chrominance interference. The more saturated colours would not happen if the TV were normalising to the colour burst level. Possibly it normalises the signal to the sync pulse level and assumes everything is to standard after that, or maybe it doesn't normalise at all.

Yes, the buffer circuit introduces a DC shift because the emitter follower bias point is between 0V and 5V. This is true of the original composite connection too. Composite and S-Video decoders are AC-coupled and establish their own black level by clamping. A capacitor could be added at the output for the chrominance circuit, but I have not found this to be necessary.

My main concerns when building the circuit were to match the loading on the chrominance pin of the VIC chip as closely as possible to the original circuit, and then provide an amplified version of the chrominance signal to drive 75 ohms as per the luminance signal.

Now I'm working on a mod for the original two prong VIC-20, and after that I'd like to do a version of the mod without the filter components in the chrominance path - i.e still using an emitter follower but passing as much of the chrominance bandwidth as possible. I think this could help sharpen up transitions between colours.

[Mike]

It would be interesting to know the exact output impedance of the VIC's chroma output. It might be in the order of 5 kOhm, i.e. comparable to a TTL common-emitter output stage.

Furthermore, I suppose the signal form not to be sinusoid, but rather having a rect shape with lots of unwanted harmonics. That would attribute to that many monitors mistake the higher harmonics for luminance information, i.e. cross-talk in the composite signal. When I have time next week I can check this with a scope.

Yes, the buffer circuit introduces a DC shift because the emitter follower bias point is between 0V and 5V.

It doesn't necessarily matter. The older monitors might even have their own decoupling capacitor in the input stage.

However, a non-0 DC offset on chroma might signal to the monitor/TV that a non 4:3 screen format should be used, even though that one hasn't been specified as a standard.

[rga24]

It would be interesting to know the exact output impedance of the VIC's chroma output. It might be in the order of 5 kOhm, i.e. comparable to a TTL common-emitter output stage.

Furthermore, I suppose the signal form not to be sinusoid, but rather having a rect shape with lots of unwanted harmonics. That would attribute to that many monitors mistake the higher harmonics for luminance information, i.e. cross-talk in the composite signal. When I have time next week I can check this with a scope.

Yes, the buffer circuit introduces a DC shift because the emitter follower bias point is between 0V and 5V.

It doesn't necessarily matter. The older monitors might even have their own decoupling capacitor in the input stage.

However, a non-0 DC offset on chroma might signal to the monitor/TV that a non 4:3 screen format should be used, even though that one hasn't been specified as a standard.

Hi Mike,

I read somewhere that the luminance output was open drain, i.e. that it needed a pullup resistor, and the chrominance output was open source, i.e. that it needed a pulldown resistor. The datasheet for the 6567 VIC-II gives pullup and pulldown resistor values of 500R and 1K, on page 17. I wonder if the VIC is similar. Page 10 of the 6560 VIC datasheet refers to pin 3 (luminance) as an open drain output and pin 2 (chrominance) as a high impedance output buffer. In the NMOS technology of the VIC, that is highly likely to be a source follower. There are references to application note No. 1 which it would be good to find.

http://www.zimmers.net/anonftp/pub/cbm/ ... index.html

In the next version of the VIC-20 CR S-Video mod I will use 1K as a pulldown resistor, omit FB7 and C13 and drive the 47uF capacitor and the rest of the emitter follower circuit from that.

No idea where you get the notion about DC offsets on the chrominance signal being for widescreen signalling. That would make it S-Video specific, for one thing. My TV uses this:

http://en.wikipedia.org/wiki/Widescreen_signaling

[Mike]

Page 10 of the 6560 VIC datasheet refers to pin 3 (luminance) as an open drain output and pin 2 (chrominance) as a high impedance output buffer. In the NMOS technology of the VIC, that is highly likely to be a source follower.

A source follower provides a low output impedance (all we would need ). It is more likely either a common-source, or even a cascode output stage.

No idea where you get the notion about DC offsets on the chrominance signal being for widescreen signalling. That would make it S-Video specific, for one thing.

Exactly also from Wikipedia, in the article about S-Video:

The S-Video connection also has general provision for widescreen signaling through a DC offset applied to the chrominance signal; however, this is a more recent development, and is not widely supported.

[rga24]

A source follower provides a low output impedance (all we would need ). It is more likely either a common-source, or even a cascode output stage.

It may not be a large enough transistor to source current for a 75 ohm load. A small, relatively high on resistance n-channel FET in source follower mode could fit the description of a "high output impedance buffer".

Common source would be more like the luminance output, open drain.

I agree with your point about the colour waveform. Chrominance is supposed to be band-limited to about +- 1MHz around the colour subcarrier in composite video, but in S-Video there is no need for such band limiting. I think FB7 and C13 are supposed to achieve this in the VIC-20 CR in combination with the resistors.

Chrominance interference occurs at colour edges; where the frequency content of the QAM waveform has side bands above and below the colour subcarrier frequency. No doubt the 6561 simply switches from one phase output to another. In S-Video, without the filtering, this should produce sharper colour edges.

The S-Video connection also has general provision for widescreen signaling through a DC offset applied to the chrominance signal; however, this is a more recent development, and is not widely supported.

I'm trying to find a reference for that. Interesting!

[a4000bear]

Do you have a 180 ohm resistor in that location? That would be like the two prong VIC-20's video circuit. I'll check my other VIC-20 CRs.

It turned out that one CR VIC had the 180 ohm, the other had a 240 ohm resistor. Seems as though Commodore were tinkering with the circuit quite a bit.

The 'two prong' version also had a 180 ohm resistor.

[a4000bear]

The data sheet mentions the chroma output of the VIC chip already *is* buffered, albeit with a high impedance. Unless the signal catches some extra noise, I'd then expect no different outcome regardless whether the signal is buffered for low impedance, or not: even if the voltage division at the 75 Ohm input of the monitor leads to a smaller received signal than normal, that gets compensated through the normalisation against the (also smaller) colour burst.

Yes, thats what I found when modifying all my VICs without a buffer. Colour on the screen was just as saturated after the mod compared to before.
With the modification, I got 0.2V p/p on the burst, a fair bit lower than it should be, but not low enough to cause any real problems.

For those cases where people have found the colour level low (as it appears on the screen) I am now convinced they have a bad VIC chip. Both new VICs I recently received have low colour. This was before I modified them. One was worse than the other, although they were both perfectly watchable.

I looked at the chroma signal on these bad chips on a CRO, and it looks like the signal that triggers the PAL switch is breaking into the output. Alternate lines and alternate bursts have a high/low DC level on them. Looking at the Wiki page for the S-video mod, there are some screenshots posted showing the effect of low colour. There are severe 'venetian blinds' in his pictures. I wonder if he has the same problem, only much worse, as I was only getting mild venetian blinds in blue areas with the worst chip.

[rga24]

Yes, thats what I found when modifying all my VICs without a buffer. Colour on the screen was just as saturated after the mod compared to before.
With the modification, I got 0.2V p/p on the burst, a fair bit lower than it should be, but not low enough to cause any real problems.

For those cases where people have found the colour level low (as it appears on the screen) I am now convinced they have a bad VIC chip. Both new VICs I recently received have low colour. This was before I modified them. One was worse than the other, although they were both perfectly watchable.

I looked at the chroma signal on these bad chips on a CRO, and it looks like the signal that triggers the PAL switch is breaking into the output. Alternate lines and alternate bursts have a high/low DC level on them. Looking at the Wiki page for the S-video mod, there are some screenshots posted showing the effect of low colour. There are severe 'venetian blinds' in his pictures. I wonder if he has the same problem, only much worse, as I was only getting mild venetian blinds in blue areas with the worst chip.

I think the issue with low colour saturation has to do with the behaviour of display devices.

In composite displays, the video circuit has a sync pulse to normalise everything to. Evidently the chrominance burst does not get normalised separately from this, hence you can have high colour saturation on a VIC-20 CR and low colour saturation on a two prong VIC-20.

In some S-Video displays, the luminance and chrominance evidently get normalised separately, for example in the displays a4000bear is using. Here, regardless of the chrominance amplitude, the colour saturation is normalised with reference to the colour burst. In other displays this may not happen, for example the displays Bacon of Gothenburg was using, at the bottom of the wiki page.

http://sleepingelephant.com/denial/wiki ... deo_output

It does appear that a 75 ohm load causes the p-p voltage coming out of the chrominance pin to drop more than the original load. Whether this matters or not possibly depends on what kind of display you have, and whether you prefer the simplest possible mod.

The PAL VIC's colours vary quite a bit on alternate lines, you can see this from mixing colours of similar luminance on alternate lines.

If you set up a definable character with alternating 11111111 00000000 lines, and another one with alternating 0000000 11111111 lines, and use multicolour characters, you can mix pairs of colours using the screen colour and the auxiliary colour. Some colours show much bigger phase errors than others, which suggests to me that what the VIC is outputting are not exact PAL colour phase shift pairs to begin with.

[a4000bear]

I think the issue with low colour saturation has to do with the behaviour of display devices.

In composite displays, the video circuit has a sync pulse to normalise everything to. Evidently the chrominance burst does not get normalised separately from this, hence you can have high colour saturation on a VIC-20 CR and low colour saturation on a two prong VIC-20.

In some S-Video displays, the luminance and chrominance evidently get normalised separately, for example in the displays a4000bear is using. Here, regardless of the chrominance amplitude, the colour saturation is normalised with reference to the colour burst. In other displays this may not happen, for example the displays Bacon of Gothenburg was using, at the bottom of the wiki page.

http://sleepingelephant.com/denial/wiki ... deo_output

It does appear that a 75 ohm load causes the p-p voltage coming out of the chrominance pin to drop more than the original load. Whether this matters or not possibly depends on what kind of display you have, and whether you prefer the simplest possible mod.

The PAL VIC's colours vary quite a bit on alternate lines, you can see this from mixing colours of similar luminance on alternate lines.

If you set up a definable character with alternating 11111111 00000000 lines, and another one with alternating 0000000 11111111 lines, and use multicolour characters, you can mix pairs of colours using the screen colour and the auxiliary colour. Some colours show much bigger phase errors than others, which suggests to me that what the VIC is outputting are not exact PAL colour phase shift pairs to begin with.

I'm not so sure about the display device bit.
I certainly did get low apparent colour in two cases, but then, when I changed the VIC chip over to one that was already giving full colour in another computer, I got full colour in the new machine, implying it was the VIC chip.

Can anyone out there who had low colour try a different VIC chip to see what happens?

I wonder if there is a psychological aspect too?

When I first tried out those two VICs I recently received, I did not notice anything unusual about the colour levels. But when I modified them, straight away, I noticed the colour was low. This was confirmed when I compared them to the VIC I had modified long ago, which had full colour levels.

This was when I tried swapping VIC chips, and found the low colour was caused by the VIC chip itself.

To confirm that the S-video mod was not to blame, I reversed the mod on one of the VICs, and noted the same thing. Low colour with the original VIC chip, good colour with the other chip. It was at this point I realised that the low colour level I have now was the same colour level I had before the mod. I just did not notice it.

So why did I not notice low colour before?

I don't know. But perhaps having a nice clean smooth image made me more aware of other deficiencies in the display. Perhaps there is also a subconscious tendency to keep the colour turned down low on a composite image due to the poor quality, and that with a nice, clean S-video image, one might feel less constrained and try to turn up the colour only to find there is little or no adjustment range left.

I think questions we need to ask are:

Is this variation in colour levels normal in different VIC chips? (I have noticed no correlation with this and chip versions)

How common is it?

Or is this actually a fault certain chips have developed?

For those out there who have multiple machines and multiple VIC chips, (especially those not modified for S-video) it would be very interesting to swap chips and see the results.

[rga24]

I have one more set of results to report. Yesterday I did a second S-Video mod on another VIC-20, this time omitting FB7 and C13 from the circuit and using a 1K pulldown resistor instead of a 330R.

Both machines are VIC-20 CRs, the purpose of the first mod was to copy the chrominance signal path as closely as possible to the original composite circuit but in a separate signal from the luminance, and the purpose of the second mod was to remove any high pass (band pass?) filtering due to the ferrite bead (inductor) and capacitor, and to use a 1K load on the chrominance output instead of 330R.

330R simulates the loading effect of the 1K preset and 510R pullup resistor in the original circuit. The ZTX300 transistor amplifier is still there, connected to the chrominance output via a 47uF capacitor as per the original circuit. A 47uF decoupling capacitor connected to 1K8 and 2K7 bias network resistors works out as a 3.1Hz high pass filter.

Well... the colour saturation in the first mod was entirely satisfactory, and didn't need any adjustments on the colour control after the TV had been connected to a composite VIC-20.

The colour saturation on the second mod was much higher! Especially on the colour blue, the colours were so saturated I actually had to turn the colour control down on my 21 inch Sony TV.

The circuit in the second mod actually provides a higher degree of colour saturation than even stock VIC-20 CRs. This is because the 1K load on the chrominance output loads it somewhat less than the 330R equivalent in the original circuit, and also because the 2.19MHz high pass filter in the original circuit (provided by C13 220pF and the 330R load) is missing here. I'm going to take some measurements of the waveforms using an oscilloscope, and hope to post some videos later.

My Sony TV does not appear to be normalising to the colour burst.

Richard

[a4000bear]

The colour saturation on the second mod was much higher! Especially on the colour blue, the colours were so saturated I actually had to turn the colour control down on my 21 inch Sony TV.


Richard

Very interesting. One thing, which I may not have mentioned is that the variation in colour saturation that I have found also seems to affect blue and yellow the most.

Are your VIC chips socketed? I would love to see what happens when you swap them between the two machines.

[mikem]

I have been following this thread with the hopes that someone has tried and succeeded in getting an s-video out on the original NTSC Vic-20 (gold/brown badge)

Can the instructions be adjusted?

If I was smarter I may know how to do this myself, but no deal wit that

Thanks