exxos's Atari, Amiga & retro forum

Badwolf wrote: 16 Oct 2024 21:14

ijor wrote: 16 Oct 2024 17:13 Also there is a technical difference between frequency and bandwidth. The (maximum) bandwidth is just the pixel clock, not divided by 2, and multiplied by the color resolution.

I can't see why it wouldn't be divided by two. The pixel clock goes through a full cycle for each pixel, but the frequency that the pixels can go from black to white is half that. Ie. the clock goes hi-lo-hi-lo to make the luma go hi-lo (at the maximum rate).

I think you are wrong. It is true that the clock toggles at double the frequency of the data, or which is the same, the data toggles at half the frequency of the clock (except on DDR, of course). But the video data still toggles at the pixel clock frequency. It is the clock that toggles at double the pixel clock frequency. But since the clock itself is not transmitted here, the maximum frequency is the pixel clock/2.

To put it in numbers. The frequency at mono is ~32 MHz, let's assume just 32 MHz for this purpose. The period is 32/2 = 31.25ns. Right? Well, pixels can toggle each 31.25ns, while the clock toggles at each 15.625ns (64 MHz).

PS: Note that I said that the bandwidth is not divided by 2. The frequency yes. That's why I said they are technically two different things.

ijor wrote: 16 Oct 2024 21:54 To put it in numbers. The frequency at mono is ~32 MHz, let's assume just 32 MHz for this purpose. The period is 32/2 = 31.25ns. Right? Well, pixels can toggle each 31.25ns, while the clock toggles at each 15.625ns (64 MHz).

PS: Note that I said that the bandwidth is not divided by 2. The frequency yes. That's why I said they are technically two different things.

I don't quite follow those numbers because I'm least au fait with mono mode, but with a pixel clock of 32MHz, the pixels can toggle at every 31ns, yes. Which would represent a wave of period 62ns. Or 16MHz.

But you're right. These would be square waves rather than sine waves, so the actual frequency components (and thus real bandwidth) required to describe them accurately would need to be several times higher. So my initial conclusions were probably wrong on that front.

In fact, thinking about it, to describe a single active pixel in the middle of an otherwise inactive row, the necessary harmonics of the signal would be extremely high frequency indeed.

BW