ESP32 MCPWM Input Capture Problem

[steve.kim]

I am testing the input capture function contained within MCPWM.
I am testing an example source MCPWM(mcpwm_basic_config_example.c).
In order to know the frequency values entered in CAP 0,
the remainder has been disabled and only the pwm0 and Capture0 functions have been enabled.

*Test content
1. I entered a frequency of 2 kHz into the CAP 0 pin.
2. The output results are output alternately between 500us and 9500us, rather than continuously output to 500us as shown below.
3. Also, if I input the input frequency to 20 kHz, the value will be output strangely. I think it's almost impossible to measure.

* Questions
1. Is the example source working normally?
2. Doesn't it work normally with high frequency inputs?
If so, what are the measurable frequency limits?
(I would like to measure up to 200kHz)

* Mainly Function Setting
static void disp_captured_signal(void *arg)
{
uint32_t *current_cap_value = (uint32_t *)malloc(sizeof(CAP_SIG_NUM));
uint32_t *previous_cap_value = (uint32_t *)malloc(sizeof(CAP_SIG_NUM));
capture evt;

while (1) {
xQueueReceive(cap_queue, &evt, portMAX_DELAY);

if (evt.sel_cap_signal == MCPWM_SELECT_CAP0) {
current_cap_value[0] = evt.capture_signal - previous_cap_value[0];
previous_cap_value[0] = evt.capture_signal;
//How was the fomula made?
current_cap_value[0] = (current_cap_value[0] / 10000) * (10000000000 / rtc_clk_apb_freq_get());
printf("%d us \n", current_cap_value[0]);
}
}
}

static void IRAM_ATTR isr_handler()
{
uint32_t mcpwm_intr_status;
capture evt;
mcpwm_intr_status = MCPWM[MCPWM_UNIT_0]->int_st.val; //Read interrupt status
if (mcpwm_intr_status & CAP0_INT_EN) {
evt.capture_signal = mcpwm_capture_signal_get_value(MCPWM_UNIT_0, MCPWM_SELECT_CAP0);
evt.sel_cap_signal = MCPWM_SELECT_CAP0;
xQueueSendFromISR(cap_queue, &evt, NULL);
}
MCPWM[MCPWM_UNIT_0]->int_clr.val = mcpwm_intr_status;
}

static void mcpwm_example_config(void *arg)
{
//1. mcpwm gpio initialization
mcpwm_example_gpio_initialize();

//2. initialize mcpwm configuration
printf("Configuring Initial Parameters of mcpwm...\n");
mcpwm_config_t pwm_config;
pwm_config.frequency = 2000; //frequency = 1000Hz
pwm_config.cmpr_a = 50.0; //duty cycle of PWMxA = 50.0%
pwm_config.cmpr_b = 50.0; //duty cycle of PWMxb = 50.0%
pwm_config.counter_mode = MCPWM_UP_COUNTER;
pwm_config.duty_mode = MCPWM_DUTY_MODE_0;
mcpwm_init(MCPWM_UNIT_0, MCPWM_TIMER_0, &pwm_config);
mcpwm_capture_enable(MCPWM_UNIT_0, MCPWM_SELECT_CAP0, MCPWM_POS_EDGE, 0);
MCPWM[MCPWM_UNIT_0]->int_ena.val = CAP0_INT_EN; //Enable interrupt on CAP0, CAP1 and CAP2 signal
mcpwm_isr_register(MCPWM_UNIT_0, isr_handler, NULL, ESP_INTR_FLAG_IRAM, NULL); //Set ISR Handler

vTaskDelete(NULL);
}

Output Result:
500 us
9500 us
500 us
9500 us
500 us
9500 us
500 us
9500 us
500 us
9500 us
500 us
9500 us
500 us
9500 us
500 us
9500 us
500 us
9500 us
500 us
9500 us
.
.
.

[holopaul]

Hi,

I think i'm getting the same results as you did. Have you find a solutions for this?

[Milan991]

Hi,

I have the same problem. Is there any update on this?

@steve.kim Have you created issue on ESP-IDF githbub repo with same question?

[ESP_houwenxiang]

Hi,

I have the same problem. Is there any update on this?

@steve.kim Have you created issue on ESP-IDF githbub repo with same question?

Hi,
The default scheduling time for our task is 100 milliseconds. However, the frequency of the interrupt is too high (depending on the frequency of the signal), the queue is very easy full. So `xQueueSendFromISR` will fail. This means you will miss several measurements. but the counter of CAP0 is always increasing. `period = current measure - last measure ` get a wrong result.

[ESP_houwenxiang]

You can calculate the time between two adjacent triggers in the interrupt, then turn off the interrupt. Because you don't need to measure the frequency of the signal all the time. re-enable the interrupt when you want to measure the signal.

[Milan991]

Hi @ESP_houwenxiang,

I've managed to capture distance between rising edges of 125 us, after increasing frequency even further i could not get lower than this value. Is this the limit of capture mode?

Thank you very much for explanation, it had helped me a lot.

[CityLark]

Dude I have the same question. why 125us? the timer frequency capture uses is 8khz? why it's not 80M as same as the apb clock 8khz is totally not enough for a coder capture!

[CityLark]

dude! I've solved this problem

0e7fc16e3569522c4051c93399fdaf7.png (11.68 KiB) Viewed 6814 times

!!!

It's because those calculations in the example code uses integer division so the detail of the number is discarded, use some float calculations and It will be smooth as you wish