Detect TV system

The following code detects the TV system. A game can use it to compensate for differences in clock rate among various NES models. It has been tested on NTSC NES and Famicom, PAL NES, and PAL famiclones using the Dendy chipset, as well as emulators in NTSC, PAL, and Dendy modes.

;
; NES TV system detection code
; Copyright 2011 Damian Yerrick
;
; Copying and distribution of this file, with or without
; modification, are permitted in any medium without royalty provided
; the copyright notice and this notice are preserved in all source
; code copies.  This file is offered as-is, without any warranty.
;
.export getTVSystem
.importzp nmis

.align 32  ; ensure that branches do not cross a page boundary

;;
; Detects which of NTSC, PAL, or Dendy is in use by counting cycles
; between NMIs.
;
; NTSC NES produces 262 scanlines, with 341/3 CPU cycles per line.
; PAL NES produces 312 scanlines, with 341/3.2 CPU cycles per line.
; Its vblank is longer than NTSC, and its CPU is slower.
; Dendy is a Russian famiclone distributed by Steepler that uses the
; PAL signal with a CPU as fast as the NTSC CPU.  Its vblank is as
; long as PAL's, but its NMI occurs toward the end of vblank (line
; 291 instead of 241) so that cycle offsets from NMI remain the same
; as NTSC, keeping Balloon Fight and any game using a CPU cycle-
; counting mapper (e.g. FDS, Konami VRC) working.
;
; nmis is a variable that the NMI handler modifies every frame.
; Make sure your NMI handler finishes within 1500 or so cycles (not
; taking the whole NMI or waiting for sprite 0) while calling this,
; or the result in A will be wrong.
;
; @return A: TV system (0: NTSC, 1: PAL, 2: Dendy; 3: unknown
;         Y: high byte of iterations used (1 iteration = 11 cycles)
;         X: low byte of iterations used
.proc getTVSystem
    ldx #0
    ldy #0
    lda nmis
nmiwait1:
    cmp nmis
    beq nmiwait1
    lda nmis

nmiwait2:
    ; Each iteration takes 11 cycles.
    ; NTSC NES: 29780 cycles or 2707 = $A93 iterations
    ; PAL NES:  33247 cycles or 3022 = $BCE iterations
    ; Dendy:    35464 cycles or 3224 = $C98 iterations
    ; so we can divide by $100 (rounding down), subtract ten,
    ; and end up with 0=ntsc, 1=pal, 2=dendy, 3=unknown
    inx
    bne :+
    iny
:
    cmp nmis
    beq nmiwait2
    tya
    sec
    sbc #10
    cmp #3
    bcc notAbove3
    lda #3
notAbove3:
    rts
.endproc

What to change

• PAL NES needs different music pitch from NTSC and Dendy.
• NTSC needs different music speed from PAL NES and Dendy.
• PAL NES needs different raster effects from NTSC and Dendy.
• NTSC needs different scrolling and sprite movement speeds from PAL and Dendy.
• PAL NES needs OAM DMA done first in vblank, while NTSC benefits from doing it last if using timed code to work around the DMC DMA glitch.
• If you want to take advantage of PAL's longer vertical blank to load more tiles into CHR RAM, Dendy needs different detection of the end of rendering.