About the stability of PWM mode

[YaoYaoLin]

void init_servo_timer()
{
 u16 tmp_period = 0;
 u16 CCR1_Val = 0;
 
 pulse_num_1ms = 0;
 pulse_period = (1/((float)pulse_num_1ms));
 tmp_period = (u16)(pulse_period * 36000);
 CCR1_Val = (tmp_period / 2 );
 
 TIM_DeInit(TIM1);
 /* (TIM_Period = TIM1->ARR) = 72MHz / prescaler */
 TIM_TimeBaseStructure.TIM_Period = (tmp_period); 
 TIM_TimeBaseStructure.TIM_Prescaler = (2 - 1); /* prescaler = 2 */
 TIM_TimeBaseStructure.TIM_CounterMode = TIM_CounterMode_Up;
 TIM_TimeBaseStructure.TIM_ClockDivision = TIM_CKD_DIV1;
 TIM_TimeBaseStructure.TIM_RepetitionCounter = 0x00;
 TIM_TimeBaseInit(TIM1, &TIM_TimeBaseStructure);
 TIM_ITConfig(TIM1,TIM_IT_CC1,ENABLE);
 TIM_Cmd(TIM1, ENABLE);
 
 TIM_OCInitStructure.TIM_OCMode = TIM_OCMode_PWM1;
 TIM_OCInitStructure.TIM_OutputState = TIM_OutputState_Enable;
 
 TIM_OCInitStructure.TIM_Pulse = CCR1_Val; /* TIM1-> CCR1 */
 TIM_OCInitStructure.TIM_OCPolarity = TIM_OCPolarity_High;
 TIM_OCInitStructure.TIM_OCIdleState = TIM_OCIdleState_Reset;
 TIM_OC1Init(TIM1, &TIM_OCInitStructure);
 TIM_CtrlPWMOutputs(TIM1, ENABLE);
 
}
 
void TIM1_CC_IRQHandler(void)
{ 
 TEST_LED = ~TEST_LED;
 if ( TIM_GetITStatus(TIM1 , TIM_IT_CC1)) 
 {
 TIM_ClearITPendingBit(TIM1 , TIM_IT_CC1);
 
 pulse_cnt ++;
 if(am.dir)
 am.now_addr --; 
 else
 am.now_addr ++;
 
 if(pulse_cnt >= am.speed_cnt)
 {
 pulse_cnt = 0;
 if(am.now_addr == am.set_addr)
 {
 am.run = 0;
 am.speed_cnt = 0;
 }
 }
 }
 TEST_LED = ~TEST_LED;
 
}
// I put output_servo_am_pulse() function in 1ms loop 
void output_servo_am_pulse(void)
{ 
 if((am.h_run | am.run))
 {
 if(am.speed_cnt)
 {
 pulse_cnt = 0;
 TIM1->CNT = 1;
 TIM1->ARR = (u16)((36000 / am.speed_cnt));
 TIM1->CCR1 = (TIM1->ARR >> 1 );
 }
 }
 else
 {
 TIM1->ARR = 0;
 TIM1->CCR1 = 0;
 }
 
}

I was confusing why the pulse sometimes ARR should be reload (pulse 5.5us)but it didn't, but from the picture of waveform , I can still see the compare interrupt(TEST_LED 2us) still work.

By the way, I am using stm32f103xxx.

 

Can anyone help me for this problem? Thanks a lot.

[waclawek.jan]

Is your logic analyzer fast enough to display a pulse 1 cycle long? And is the PWM output pin set fast enough in GPIO_CRx.MODEx to output a pulse 1 cycle long?

Btw. if you use only 50% duty cycle, it's more elegant to use Toggle mode of the timer.

JW

[YaoYaoLin]

As you see the top of my logic analyzer its sample rate is 2MHz, that means it shows every 1/2MHz( = 500ns).

Do you mean this configuration?

GPIO_InitStructure.GPIO_Pin = GPIO_Pin_8
GPIO_InitStructure.GPIO_Mode = GPIO_Mode_AF_PP;
GPIO_InitStructure.GPIO_Speed = GPIO_Speed_50MHz;
GPIO_Init(GPIOA, &GPIO_InitStructure);

Yeah, I know about Toggle mode. The important thing is I want the interrupt appears at negative edge trigger of the pulse. So that's why I pick PWM1 mode.

By the way, PWM mode is more alternative I thought.

The new things I found was, if I show TIM1(pulse generate) and TIM5(1ms loop) separately, then everything happened accurately. But when I put them together, then TIM1 and TIM5 affect each other, for instance, 1ms later become 1.001ms ~ 1.004ms, that would increase about 1~4us, 2us is large enough to make the interrupt of TIM1 delay.

Thanks for reply, WYL