from Nikolai Golovchenko
; ;d0, d1, d2, ...dn  counters ;d0 = 0..3  least significant 2 bits ;d1,...dn = 0..255  more significant 8, 16, 24, ... bits ; ;Total Delay = ;= d0+4*d1+4*256*d2+4*256*256*d3...+4*256^(n1)*dn + overhead ; ;Overhead depends on a number of counters, see below ;First preincrement all counters except d0 ;overhead = n cycles incf d1, f incf d2, f ... incf dn, f ;2 bit delay (d0  1 cycle resolution) ;overhead += 4 cycles comf d0, w andlw 0x03 addwf PCL, f Delay64Mx Delay256Kx nop Delay1Kx nop Delay4x nop ;8 bit delay (d1  4 cycle resolution) ;overhead += 2 cycles decfsz d1, f goto Delay4x ;8 bit delay (d2  1024 cycle resolution) ;overhead += 3 cycles decf d1, f ;change d1 to 255, so previous loop ;(from Delay4x) will take ;255*41=1019 cycles ;we need to add 5 more cycles to get 1024. ;4 cycles are in this loop and ;another 1 cycle is a nop above decfsz d2, f goto Delay1Kx ;8 bit delay (d3  262144 cycle resolution) ;overhead += 3 cycles decf d2, f ;previous two loops (from Delay4x) will take ;1019+255*10241=262138 cycles ;we need to add 6 more cycles to get 262144. ;4 cycles are in this loop and ;another 2 cycles are above  two nops decfsz d3, f goto Delay256Kx ;8 bit delay (d4  262144*256 cycle resolution) ;overhead += 4 cycles decf d3, f ;previous two loops (from Delay4x) will take ;262138+255*2621441 cycles ;we need to add 7 more cycles to get 4*256^3. ;5 cycles are in this loop and ;another 2 cycles are above  two nops nop decfsz d4, f goto Delay64Mx ;at this point we have a 34 bit one cycle resolution delay! ;Total Delay = overhead + 0..1.7e10 cycles ;overhead = 4+4+2+3+3+4 = 20 cycles ;
And the same for a SX chip:
; ;d0, d1, d2, ...dn  counters ;d0 = 0..3  least significant 2 bits ;d1,...dn = 0..255  more significant 8, 16, 24, ... bits ; ;Total Delay = ;= d0+4*d1+4*256*d2+4*256*256*d3...+4*256^(n1)*dn + overhead ; ;Overhead depends on a number of counters, see below ;First preincrement all counters except d0 ;overhead = n cycles inc d1 inc d2 ... inc dn ;2 bit delay (d0  1 cycle resolution) ;overhead += 5 cycles mov w, /d0 and #$03 add PC, w Delay256Kx nop Delay64Mx nop Delay1Kx nop ;8 bit delay (d1  4 cycle resolution) ;overhead += 2 cycles Delay4x decsz d1 jmp Delay4x ;8 bit delay (d2  1024 cycle resolution) ;overhead += 3 cycles dec d1 ;change d1 to 255, so previous loop ;(from Delay4x) will take ;255*42=1018 cycles ;we need to add 6 more cycles to get 1024. ;5 cycles are in this loop and ;another 1 cycle is a nop above decsz d2 jmp Delay1Kx ;8 bit delay (d3  262144 cycle resolution) ;overhead += 3 cycles dec d2 ;previous two loops (from Delay4x) will take ;1018+255*10242=262136 cycles ;we need to add 8 more cycles to get 262144. ;5 cycles are in this loop and ;another 3 cycles are above  three nops decsz d3 jmp Delay256Kx ;8 bit delay (d4  262144*256 cycle resolution) ;overhead += 6 cycles dec d3 ;previous two loops (from Delay4x) will take ;262136+255*2621442 cycles ;we need to add 10 more cycles to get 4*256^3. ;8 cycles are in this loop and ;another 2 cycles are above  two nops jmp $+1 decsz d4 jmp Delay64Mx ;at this point we have a 34 bit one cycle resolution delay! ;Total Delay = overhead + 0..1.7e10 cycles ;overhead = 4+5+2+3+3+6 = 23 cycles ;
This is probably too small for SX :)
Code:
+The PIC delay routine above works correctly only if d3 and d4 are zero (the same is probably true for the SX version as well). The problem lies in how the routine is extended by adding another loop. Basically, every time a loop is appended, it adds at least 2 cycles, which must be subtracted. Sometimes, it can be done by moving the jump target by 2 nops down, but when there are no more nops, it can be done by subtracting one from d1, which subtracts 4 cycles. So d1 is decremented, a new loop adds 3 cycles and subtracts 4, which is 1 overall. So the jump should include an additional nop. The corrected PIC version is below: ; preincrement counters except d0 incf d1, f incf d2, f incf d3, f incf d4, f ; Let Delay = 0 at this point comf d0, w andlw 0x03 addwf PCL, f nop Delay2 nop Delay1 nop ; Delay = 4 + d0 Delay0 decfsz d1, f goto Delay1 ; Delay = 6 + d0 + 4 * d1 decf d1, f decfsz d2, f goto Delay2 ; Delay = 9 + d0 + 4 * (d1 + 256 * d2) decfsz d3, f goto Delay0 ; Delay = 11 + d0 + 4 * (d1 + 256 * d2 + 256^2 * d3) decf d1, f decfsz d4, f goto Delay1 ; Delay = 14 + d0 + 4 * (d1 + 256 * d2 + 256^2 * d3 + 256^3 * d4) ; This can be extended like this: ; ; decf d1, f ; decfsz d5, f ; goto Delay2 ; ; decfsz d6, f ; goto Delay0 ; ; decf d1, f ; decfsz d7, f ; goto Delay1 ; ; decf d1, f ; decfsz d8, f ; goto Delay2 ; ; decfsz d9, f ; goto Delay0
See also:
file: /Techref/microchip/delay/xb1crnmng.htm, 7KB, , updated: 2005/3/28 21:03, local time: 2020/10/24 22:38,

©2020 These pages are served without commercial sponsorship. (No popup ads, etc...).Bandwidth abuse increases hosting cost forcing sponsorship or shutdown. This server aggressively defends against automated copying for any reason including offline viewing, duplication, etc... Please respect this requirement and DO NOT RIP THIS SITE. Questions? <A HREF="http://www.piclist.com/techref/microchip/delay/xb1crnmng.htm"> PIC Microcontoller Delay Method  Delay X cycles (nx) exclusive</A> 
Did you find what you needed? 
PICList 2020 contributors:
o List host: MIT, Site host massmind.org, Top posters @20201024 Bob Blick, RussellMc, madscientistatlarge, Jason White, Neil, David C Brown, Sean Breheny, Harold Hallikainen, James Cameron, Allen Mulvey, * Page Editors: James Newton, David Cary, and YOU! * Roman Black of Black Robotics donates from sales of Linistep stepper controller kits. * Ashley Roll of Digital Nemesis donates from sales of RCL1 RS232 to TTL converters. * Monthly Subscribers: Gregg Rew. ongoing support is MOST appreciated! * Contributors: Richard Seriani, Sr. 
Welcome to www.piclist.com! 
.