Hi there !!
I'm using a routine to check if the ship and the bullets hit a wall.
However, this routine is not 100% accurate (sometimes you die when you fly very close to a wall)
Maybe this is because r#18 (horizontal screen offset) and r#23 (vertical screen offset) are used on VBLANK and at the linesplit (y=20) this routine.
Also the page is switched every 16 steps.
The sprites are 16x16 and I want the collision detection to occur exactly in the middle of the sprite.

This is my routine, I would love some help to perfect it, or speed it up !!! THANX !!

CollisionCheck:                ;collision with foreground check

;set X-coordinate
        LD      A,(vdp_8+10) ;horizontal screen offset
        ADD   7 
        AND   15
        RRA                           ;devide by 2
        LD     B,A                   

        LD     A,(X_coord)      ;x coordinate of whatever needs collision detection
        RRA                           ;devide by 2 (screen 5 is used, and 2 pixels are defined in 
        ADD   B                      ;1 byte, so devide by 2)
        LD     L,A                   ;store X

;set y-coordinate
        LD      A,($F3E1)        ;current page
        AND   %00100000    ;check which page is displayed (0 or 1?)
        RLCA                         ;page 1? then a=128, page 0? then a=0 
        RLCA                         ;add 128 bytes to y-coordinate when page 1 is used 
        LD      C,A

        LD    A,(vdp_8+15)    ;vertical screen offset
        LD    B,A

        LD     A,(Y_coord)      ;y coordinate of whatever needs collision detection
        ADD   B                      ;add vertical screen offset
        RRA                           ;devide by 2 (screen 5 is used, and 2 pixels=1 byte)
        ADD   4                      ;chech the middle of this sprites Y
        ADD   C
        LD     H,A                   ;store Y
        CALL  VDPREA           ;now check the color value at these coordinates
        IN    A,($98)
        AND   15                    ;color from 0-15

        SUB   4                      ;color=>4 ?
        RET   C
 	YOU  DIE !!
 

Actually ising colors is a very bad limitation !!!

This implies that you have to reserve teh first colors to the background and that you
cannot reuse background colors in the main layer!

Probably the best thinh should to use a direct access to teh "map" of tiles you have
in ram tarher that reading vram.

It is less accurate (as each collision happens lokking at the tile as a single object)
but you can use all the colors everywhere you like on the screen

Quote:

Actually using colors is a very bad limitation !!! This implies that you have to reserve teh first colors to the background and that you cannot reuse background colors in the main layer!

true, but thats no problem for me.
I'm always using different color in the background/foreground.

Quote:

Probably the best thing should to use a direct access to the "map" of tiles you have in ram rather that reading vram.

Ok, but I have no "map" nor "tiles".
I'm just using screen 5 copies. 16 big copies in 16 steps. No tile copies used. No "map" used.

Quote:

It is less accurate (as each collision happens lokking at the tile as a single object)

And thats the problem.
Vertically speaking the collision detection is perfect, looking at exactly the middle of the sprite.
Horizontally speaking I want the collision detection to be pixel accurate, but it's not.
Sometimes the right edge of the sprite detects the collision, sometimes the left edge. And I want the exact middle of the sprite to detect collision.

I presume it has to do with page switching and r#18 (jumping whole screen 16 pixels to the right).
How should I take the fact that r#18 and the current page is set at lineint in account???
The collision detection is outside of the interrupt.....