The funny business using Hires with 240 scanlines - an inside look

[Rybags]

As many of us know, if you build a display list which has any 320 pixel mode (e.g. GR. 0 or 8 ) occupying the last "displayable" scanline of the display, you end up with the annoying distorted display.

 

 

So, I decided to have a bit of a deeper look into things. Problem is, I've got an oscilloscope but it seems to have stopped working.

 

Next best solution: use the PC's sound input. Luckily, my nice new(ish) motherboard can do modern hi-fi sound, so with the help of Audacity and a hacked up cable, I took some samples of the composite video signal from my 130XE.

 

Of course, it would have been better to sample the Luma component only, but the cable I have made up only has the Composite connected up at the moment.

 

 

I've taken four samplings. Then, edited each one so that they are in sync with each other. To reduce influence from chroma, I set the colours to 04 (background), 02 (border) and 00 (text)

 

The display present for each sample is the standard Gr. 0 screen, with colours set as just describe, and the following exceptions:

- first one is standard.

- second one is with an extra 23 blank lines inserted at the top of the display list

- third one is with an extra 24 blank lines at the top. This is the one we all know causes the distortion of the display

- the fourth one is as the third, except I've changed the last line of the display to Antic Mode 4 (aka Gr. 12) - doing so creates a "stable" display

 

 

Looking at the first samplings (at minimal magnification). You see the overall story. VBlank occurring around the 1.00 second mark. Note the differences in each sample, due to the differing screen architecture of each setup.

 

The second and third samplings show the business end (Vblank). Look at the third one down. Very strange.

 

At the maximum magnification I've provided, you see the story (as best as my sampling can tell). The VBlank signals being generated here look way different in the distorted screen than all the others. It's as if ANTIC is suffering some major confusion.

Edited November 23, 2007 by Rybags

[emkay]

My guess:

It's not ANTIC having this problem.

As we know, the base resolution of GTIA is at the "colour clocks" , based on NTSC systems. Hires seems to be a special command that ANTIC sends to the GTIA.

We also know that the "special command" is lost when changing to a GTIA mode and back, which makes it possible to have 3 resolutions in one scanline.

At the end of the playlist, GTIA is ordered to stop displaying. This works timing correct by using standard colour mode.

When using hires in the last line, the "end" command seems to get interrupted and GTIA displays "garbage".

 

BTW: On PAL TVs you can use this to add more visuable lines (more than 240) onscreen, because the screen starts to roll.

[Rybags]

It could be either, I guess.

 

Normally, it sends (on AN0-AN2) either "010" for "Horizontal Blank and clear 40 character mode" or "011" for "HBlank and set 40 character mode".

 

Maybe, while it's busy in preperation for the VBlank, it accidentally sends the "set 40 character" command instead of the clear 40 char. There is the "001" for VBlank command, but the doco seems unclear whether this is sent at the end of the "visible" display, or at the actual time of VBlank.

 

The curled white bit at the bottom of the screen assumes the colour of whatever text is set to (PF1), so maybe that's the case.

 

Agreed, GTIA seems to override the hires command once you establish one of the GTIA modes on a scanline, then "forgets it" (since there seems to be no way to revert to nomral hires on a line).

[UNIXcoffee928]

Ha, that error gave me a neat idea. If you ran the signal through a comb filter & saw-wave generator appropriately, for the frequencies that make up the bottom error fill (maybe hardware, maybe software, maybe can't be done...), you could make it into a 3D shelf. See pic.

 

This is just a guess, and could be completely erroneous... but I like the shelf!

 

I guess that you could hack something up that took the second part of a saw-wave's duty-cycle for the diagonals, and make any screen have a 3D shelf like that. This is just a wild guess, from playing with analog synth waves, but maybe it could be done with video output, I'm sure that someone here knows if this would work or if it's just plain wrong.

 

 

Just get a new oscilloscope under warranty to fix the old one...

 

= )