Interlaced picture demo/test (160x384)

[candle]

Rybagss: i'm using 37" lcd from lg, and picture on Your demo programs is stable in wide range of settings,only on the very end of scale lcd get dumb, and try to deinterlace whole picture and displaing it double height, and loosing sync

there is no diffrence if You flip the fields when interlacing - i have to disable interlace to see what id does to see any effect at all

on test pattern i got 8 lines though so the algorith works

[Rybags]

 

 

Moving along with my new approach. Forget this trial/error stuff and getting half-baked results that work only for some people.

 

Thanks to our Antic bug, we can almost create or suppress Sync signals at will in the "offscreen" area that's outside the normal 240 scanlines we can normally use.

So, may as well go the whole hog and create something as close to "by the book" as possible.

Halfway there... this is just doing the one field continuously, but I've managed to get rid of all of the distortion you normally get.

 

It all comes down to timing. Several subroutines with NOPs everywhere for fine-tuning. Jump one position in later, or pad in an extra NOP and the whole thing introduces distortion again. Luckily our Ataris run quick enough that one or two cycles timing difference doesn't make a huge impact.

 

Now to get the other field taken care of... put the two together and hopefully get a 100% stable interlaced display that should work on virtually all TVs/monitors.

 

 

"The Book" - it's from a doc called "VBI-625-PAL.PDF". I've traced over some of the area to make it easier to see.

[Rybags]

 

Coming soon...

 

Half-height 8x8 characters (oh so cute, but a little hard to read). 1st pic is photo of it on a TV, second is snapshot from capture card.

 

I've also done an 8x16 mode which is standard height, using a custom character set with Serif characters.

 

The happy medium, I think for a reduced height mode might be something like 8x10 or 8x12 characters although that means more RAM is needed for the character sets (both above mentioned modes need 2 character sets in RAM).

 

8x12 characters would = 40x40 at fullscreen

8x10 characters would = 40x48 at fullscreen.

[ClausB]

Very nice.

something like 8x10 or 8x12 characters although that means more RAM is needed for the character sets (both above mentioned modes need 2 character sets in RAM).

Doesn't ANTIC already have an 8x10 mode for extenders, which does not need 2 sets?

[Rybags]

The 8x10 mode isn't useful for doing the reduced height characters.

 

This method above uses the VScrol trick to reduce the characters to 8x4... resultant is 8x8 since we've doubled the vertical res by inducing interlace.

 

The cost here is only 2 character sets since you just replicate the top half of the character into the bottom.

 

ie - each alternate Gr. 0 line shows rows 4-7 then 0-3. For 12 pixel-height you need 6 rows per field which means you also need 2 more character sets since you only get 2 unused bytes per character.