Re: Trying to make complete sense of the video scanner, HBL, and the floating bus

[Vladimir Ivanov <vladitx@XXXyahooXXX.com>]

On Sat, 11 Feb 2012, BLuRry wrote:

> Sather does an awesome job at talking it up from the hardware side, but 
> after re-reading his notes a couple of times I still can't quite make 
> sense of what exactly happens during HBL, except when he talks about 
> what happens on the left-most fringe, where the previous byte in memory 
> affects the left edge (e.g. the upper-right pixel at $2027 affects the 
> middle-left pixel 64 rows down at $2028).

As pointed, this is fixed after II+. I think it was very noticeable on 
Lode Runner while climbing a rightmost ladder.

> So, for starters, I assume that the video scanner draws the screen from 
> Row 0 ($2000), then row 1 ($2400), and so on up to row 191.

If you are talking about the vertical count register, it starts before and 
finishes after the visible part. And part of this visible part is the 
Apple II screen, the rest are just blank lines.

> Each clock cycle, a byte is read from ram and that byte remains 
> available via the floating bus.

In the //e case - one byte is read from main and one byte from aux bus, on 
the same cycle.

> At the end of each row, HBL is started for the next 25 cycles.  The row 
> is immediately incremented, and the scanner starts reading the next row 
> at the start address minus 25 bytes, so after writing row 0 the HBL 
> starts the scanner at $2400 - 25, or $2E37.  The scanner continues 
> reading bytes, but video doesn't get output until 25 cycles later at 
> $2400.

Sather has provided detailed tables of video addresses scanned, for both 
text and graphics modes - you can easily check whether you "walk" them 
right.

> Once the scanner has finished row 191, VBL starts but the scanner 
> effectively repeats this process over again for the last 70 rows as if 
> it was re-scanning them again.

Ditto as above.

I choose to made a short program that precalculates scanner addresses into 
a table, to save some time on calculating effective address from HCNT / 
VCNT.

> I get kind of confused on these points, because when I re-read Sather, 
> he talks about the HBL and VBL accounting for the left, right, top and 
> bottom margins.  But effectively, there are only left and bottom margins 
> so to speak, because Y is incremented at the very end of the visible 
> scanline, and VBL stops the moment it begins rendering row 0.  Am I 
> reading too much into this?

As already pointed, there are actually margins. What you call HBL/VBL 
contains front porch, sync and back porch+burst.

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

> All things being said and done, if I am right about the behavior of the 
> scanner and floating bus then JACE should have worked for money munchers 
> and the rainbow example a couple of iterations ago -- provided I didn't 
> blatantly fat-finger something (ok, so I really shouldn't rule that out 
> either. ;-)

I have a weak memory - does Money Munchers work on //e or only on II/II+?

II/II+ and //e have different scan patterns during HBL, IIRC.