3

I just watched an interesting video on the imperfections of video output on the Commodore 64.

Among other things, he seems to be saying the commonly observed 'jail bars' effect is a result of other things going on in the VIC-II, like DRAM refresh, interfering with the video signals.

If I'm understanding this correctly, the only effective way to prevent this would be to have the VIC-II output a digital signal, and have a separate chip (with no other responsibilities) convert it to analog.

Is that an accurate assessment? Or are there other ways Commodore could have solved this problem (if, perhaps, they had the luxury of a little more development time)?

1
  • 1
    rwallace, when you post a link into your questions, it's easy to do so by making part of your text a hyperlink. Just select that text then click on the Hyperlink icon and paste the link into the dialogue box that appears. It makes posts easier to read and hides long and unsightly links. See my edit on your question here for an example.
    – TonyM
    Commented Jun 13 at 13:32

3 Answers 3

6

Of course it is only one opinion how it should have been done. There's also infinitely many other ways how it could have been done.

It is true that all the signals affect the video output, maybe through supply pin or ground pins or coupled in other means inside the video chip or outside it. The RAS and CAS signals for DRAM are not the problem themselves, but these signals make the DRAM chips to gulp a largish spike of current, so it will show as a ripple voltage on supply and/or ground pins.

Often chips have multiple supply and ground pins so that some can be reserved for noisy digital supplies and some for clean analog supplies.

Video DAC chips generally have a pin for separate clean reference current or voltage, so even if buffer supplies have some ripple, the analog signal that needs to be buffered is generated stably.

So yeah with better chip design, PCB design, onboard power supply design, supply bypassing and filtering, the video output could have had less noise from other signals.

It will not require a digital bus to another DAC chip - the digital bus already exists inside the chip and the DAC used to generate luma and chroma too. The DAC just works in a way that can directly output the chroma signal by summing the quadrature color carrier reference sine/cosine signals in correct proportions for selecting a specific color phase.

If the DAC were made with similar design, it would not improve much if the VIC-II were just split into two chips, it would see same ripples on supplies for example, unless it were designed to have a better supply.

5

I've not studied the internal layout of the VIC-II chip in detail, but I strongly suspect that power and/or ground has to go through some rather thin strips of metal (or worse, silicon) en route to both some circuitry whose power demand is highly variable and the circuitry that generates analog video output levels. A well designed chip and system would ensure that there was a reasonably good connection between a reasonably substantial external capacitor and any video output circuitry, and--more importantly--that any portion of that connection that was shared with other circuitry that could represent a large and time-varying load was exceptionally solid. If the problem with the VIC-II power supply feed were external to the chip, they could be fixed by adding external capacitance between the power and ground pins. The fact that addition of capacitance isn't a recognized fix would suggest the problem is probably with the ground routing internal to the chip.

2
  • Ground...or power. Or both. Commented Jun 16 at 0:51
  • @JerryCoffin: Since jailbars are from what I've seen usually light rather than dark (at least on my C128), I would think ground is more likely the problem than power. I think NMOS of that era would usually have a chip layer devoted to serving passive pull-ups, which could maybe be a problem because it's not a metal connection, but at the same time it would from what I understand be a relatively solid plane.
    – supercat
    Commented Jun 17 at 0:02
3

You can certainly build mixed signal (digital and analog) electronics with very low crosstalk. It's not so much a matter of architecture as a matter of getting a lot of fiddly details right. That takes time and money.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .