timer6 strange behaviour

[manes6969]

Posted on August 20, 2011 at 10:21

Hi

I wrote a delay routine using timer6. R2 folds the number of milliseconds to delay. It works fine , however only from the second time on. the first time there is no delay , the program just 'flies' through the code without a delay. This is my routine (in assembler) delayms        ;delay r2 milliseconds                 stmfd    r13!,{r0,r1,r14}                                  ldr    r0,=TIM6_CNT        ;start counting from 0000         mov    r1,#0x0000         str    r1,[r0]         ldr    r0,=TIM6_PSC        ;prescaler = 59904 ; gives 1ms/count         ldr     r1,=59904         str    r1,[r0]         ldr    r0,=TIM6_ARR        ;r2 = # milliseconds         str    r2,[r0]         ldr    r0,=TIM6_CR1        ;enable counter,no reload, single puls mode         mov    r1,#0x000d         str    r1,[r0]                  delayl        ldr    r0,=TIM6_SR        ;wait for UIF         ldr    r1,[r0]         cmp     r1,#0x0000         beq    delayl                 ldr    r0,=TIM6_SR        ;clear UIF flag         mov    r1,#0x0000         str    r1,[r0]         ldr    r0,=TIM6_CR1        ;stop the timer         mov    r1,#0x0000         str    r1,[r0]                 ldmfd    r13!,{r0,r1,pc} Does anyone know what the reason could be for this behaviour ? Any help is welcome ! Ronny Suy

[Tesla DeLorean]

Posted on August 20, 2011 at 16:36

Without putting it on the bench, I'd be concerned with the status register test and making it more specific, then perhaps look at the loading/reloading of registers.

What frequency are you running the timer at? It I were just spinning in a loop like this, I'd probably just look at using a free running timer, or cycle counter, and waiting for a timeout delta to expire. The core has a 32-bit cycle counter that should permit times of about a minute to be handled, down to tens of nanoseconds.