Date   : Tue, 18 Mar 2003 18:34:30 -0000
From   : Seddon Tom <TSeddon@...>
Subject: Emulating Elite (more 6845 problems)

I'm writing a new emulator, and I'm confused about why Elite's mode split
point is 2 scanlines above where it should be. 

It works fine on BeebEm, but it looks like BeebEm induces a 2 scanline
delay, so the vsync start and end points are 2 scanlines after where,
judging from the registers, one might expect them to be. (Take a peek at
VideoDoScanline and the use of VSyncCount.)

So, the question is... does anyone know where this magic '2' come from? If I
force my emulator's vsync to last ((R3>>4)+2) scanlines (as a very quick
hack with the same net result in terms of when the CA1 IRQ occurs), Elite
looks fine, but now I have a magic +2 and I don't know why it's needed.

The strange thing is, the split point _is_ actually at the right point. The
Elite screen is pretty similar in timing to normal interlaced Mode 4 or 5.
R4 is 38, R6 is 31, R7 is 34, R5 is 0, R8 is 1, R9 is 7. So that's
(38+1)*(7+1)=39*8=312 scanlines, and R8&1 so +.5 making a grand total of
312.5. (By the way R3 is 0x24.) On each Vsync, SysVIA T1 is loaded with
14,622. That's 228.5 scanlines, accounting for the 2 tick delay for
reloading the T1 timer: 14624/60.0=228.5.

Let us assume that the vsync IRQ has just fired. The TV is on scanline
34*8+(R3>>4)=272. Now the screen is set up for Mode 4, T1 is set to 14,622,
and off it goes for 228.5 scanlines. There's 38.5 scanlines left in the
vblank, so the split point should occur at line 190, changing display to
Mode 5 for the scanner. R5 is 0, so this doesn't need to be considered.

And this it does. However, the border between the top and bottom of the
screen (the point at which the split _should_ occur -- it's the top line of
the status display) is at line 192. Hence the problem.

I've been searching the web and usenet and looking through 6845 datasheets,
but I haven't found anything that seems relevant. I found some stuff about
interlace sync and video register settings being in pairs of scanlines on
older 6845s, which sounds perfect except that the Beeb's is a 'newer' one
because it has programmable vsync widths. I also have a sneaking suspicion
I'm missing something totally and/or that I've got something terribly wrong
somewhere! Does anyone have any ideas?

--Tom

P.S. I have the same problem in Lunar Jetman, Imogen, Phantom and Revs. I'm
hoping all have the same root cause.