Zephyr nrfx_timer_compare

[BvdP4Celsius]

Hi, I am using Zephyr and want to use the nrfx_timer_compare functions.
It seems that the system is working with Timer4 and channel 0 to 4. However, the compare of channel 5 is never fired.

I used IRQ_DIRECT_CONNECT(TIMER4_IRQn, IRQ_PRIO_LOWEST, bus_timer_event, 0); and in my interrupt handler I test NRF_TIMER4->EVENTS_COMPARE[0] up to NRF_TIMER4->EVENTS_COMPARE[5].
All events (NRF_TIMER4->EVENTS_COMPARE[0] to NRF_TIMER4->EVENTS_COMPARE[4]) are detected and handled, but NRF_TIMER4->EVENTS_COMPARE[5] will never be set.

The initialisation looks like:

nrfx_timer_compare( &bus_timer, NRF_TIMER_CC_CHANNEL0, delay, true ); 
nrfx_timer_compare( &bus_timer, NRF_TIMER_CC_CHANNEL1, delay+417, true );
nrfx_timer_compare( &bus_timer, NRF_TIMER_CC_CHANNEL2, 2400, true ); 
nrfx_timer_compare( &bus_timer, NRF_TIMER_CC_CHANNEL3, 5500, true ); 
nrfx_timer_compare( &bus_timer, NRF_TIMER_CC_CHANNEL4, 13500, true );
nrfx_timer_compare( &bus_timer, NRF_TIMER_CC_CHANNEL5, 16300, true );

Any idea why the 5th channel will never be set?

[ovrebekk]

Hi Bob

There is no reason why CC5 shouldn't work. 

One thing I notice in your code is that you refer to your own event handler in the call to IRQ_DIRECT_CONNECT, rather than the interrupt handler of the nrfx_timer library. 

Could you change this to point to the nrfx handler instead, like shown here, and see if it works better?

Best regards
Torbjørn

[BvdP4Celsius]

Hi Torbjørn,

Thanks for the reply.

I changed it into:

{
    IRQ_DIRECT_CONNECT(TIMER4_IRQn, IRQ_PRIO_LOWEST, bus_timer_event, 0);
    
    const nrfx_timer_config_t bus_timer_config =
        {
            .frequency          = NRF_TIMER_FREQ_1MHz,
            .mode               = NRF_TIMER_MODE_TIMER,
            .bit_width          = NRF_TIMER_BIT_WIDTH_32,
            .interrupt_priority = 3,
            .p_context          = NULL
        } ;


    nrfx_timer_init( &bus_timer, &bus_timer_config, bus_timer_event );
    
    nrfx_timer_compare( &bus_timer, NRF_TIMER_CC_CHANNEL0, _measdelay, true );  
    nrfx_timer_compare( &bus_timer, NRF_TIMER_CC_CHANNEL1, _measdelay+417, true );
    nrfx_timer_compare( &bus_timer, NRF_TIMER_CC_CHANNEL2, 2400, true );          
    nrfx_timer_compare( &bus_timer, NRF_TIMER_CC_CHANNEL3, 5500, true );          
    nrfx_timer_compare( &bus_timer, NRF_TIMER_CC_CHANNEL4, 13500, true );         
    nrfx_timer_compare( &bus_timer, NRF_TIMER_CC_CHANNEL5, 16300, true );         
        
    nrfx_ppi_channel_alloc( &ppi_channel );
    nrfx_timer_enable( &bus_timer );
}

void bus_timer_event(nrf_timer_event_t event_type, void * p_context)
{
    switch( event_type )
    {
    case NRF_TIMER_EVENT_COMPARE0 : 
        ...
        break;
    case NRF_TIMER_EVENT_COMPARE1 : 
        ...
        break;
    case NRF_TIMER_EVENT_COMPARE2 : 
        ...
        break;
    case NRF_TIMER_EVENT_COMPARE3 : 
        ...
        break; 
    case NRF_TIMER_EVENT_COMPARE4 : 
        ...
        break;    
    case NRF_TIMER_EVENT_COMPARE5 : 
        ...
        break;    
    default:
        LOG_INF("Unknow event type %d",event_type);
    }
}

There is a continuous event generated with event_type = 0. It seems that the system is never clearing the interrupt flag?  

removing the nrfx_timer_init does not change anything. Removing the IRQ_DIRECT_CONNECT will result in kernel errors.

[ovrebekk]

Hi Bob

I still don't think it is correct to refer to the bus_timer_event in the call to IRQ_DIRECT_CONNECT. Did you try to replace this with nrfx_timer_4_irq_handler instead?

I tried to set up a simple example with all 6 compare events, and it seems to work fine here: 

Fullscreen 7824.main.c Download

/*
 * Copyright (c) 2016 Intel Corporation
 *
 * SPDX-License-Identifier: Apache-2.0
 */

#include <zephyr.h>
#include <device.h>
#include <devicetree.h>
#include <drivers/gpio.h>
#include <nrfx_timer.h>

/* 1000 msec = 1 sec */
#define SLEEP_TIME_MS   1000

/* The devicetree node identifier for the "led0" alias. */
#define LED0_NODE DT_ALIAS(led0)

#if DT_NODE_HAS_STATUS(LED0_NODE, okay)
#define LED0	DT_GPIO_LABEL(LED0_NODE, gpios)
#define PIN	DT_GPIO_PIN(LED0_NODE, gpios)
#define FLAGS	DT_GPIO_FLAGS(LED0_NODE, gpios)
#else
/* A build error here means your board isn't set up to blink an LED. */
#error "Unsupported board: led0 devicetree alias is not defined"
#define LED0	""
#define PIN	0
#define FLAGS	0
#endif

// Get a reference to the TIMER1 instance
static const nrfx_timer_t my_timer = NRFX_TIMER_INSTANCE(4);

// Interrupt handler for the timer
// NOTE: This callback is triggered by an interrupt. Many drivers or modules in Zephyr can not be accessed directly from interrupts, 
//		 and if you need to access one of these from the timer callback it is necessary to use something like a k_work item to move execution out of the interrupt context. 
void timer1_event_handler(nrf_timer_event_t event_type, void * p_context)
{
	static int counter = 0;
	switch(event_type) {
		case NRF_TIMER_EVENT_COMPARE0:
			printk("Timer 1 callback 0. Counter = %d\n", counter++);
			break;
		case NRF_TIMER_EVENT_COMPARE1:
			printk("Timer 1 callback 1. Counter = %d\n", counter++);
			break;
		case NRF_TIMER_EVENT_COMPARE2:
			printk("Timer 1 callback 2. Counter = %d\n", counter++);
			break;
		case NRF_TIMER_EVENT_COMPARE3:
			printk("Timer 1 callback 3. Counter = %d\n", counter++);
			break;
		case NRF_TIMER_EVENT_COMPARE4:
			printk("Timer 1 callback 4. Counter = %d\n", counter++);
			break;
		case NRF_TIMER_EVENT_COMPARE5:
			printk("Timer 1 callback 5. Counter = %d\n", counter++);
			break;		
		default:
			printk("Unknown timer event!\n");
			break;
	}
}

// Function for initializing the TIMER1 peripheral using the nrfx driver
static void timer4_init(void)
{
	nrfx_timer_config_t timer_config = NRFX_TIMER_DEFAULT_CONFIG;
	timer_config.bit_width = NRF_TIMER_BIT_WIDTH_32;
	timer_config.frequency = NRF_TIMER_FREQ_1MHz;

	int err = nrfx_timer_init(&my_timer, &timer_config, timer1_event_handler);
	if (err != NRFX_SUCCESS) {
		printk("Error initializing timer: %x\n", err);
	}

	IRQ_DIRECT_CONNECT(TIMER4_IRQn, 0, nrfx_timer_4_irq_handler, 0);
	irq_enable(TIMER4_IRQn);
}

// Function for scheduling repeated callbacks from TIMER1
static void timer4_repeated_timer_start(void)
{
	nrfx_timer_enable(&my_timer);
	nrfx_timer_extended_compare(&my_timer, NRF_TIMER_CC_CHANNEL0, 500000, 
                                0, true);	
	nrfx_timer_extended_compare(&my_timer, NRF_TIMER_CC_CHANNEL1, 1000000, 
                                0, true);
	nrfx_timer_extended_compare(&my_timer, NRF_TIMER_CC_CHANNEL2, 1500000, 
                                0, true);
	nrfx_timer_extended_compare(&my_timer, NRF_TIMER_CC_CHANNEL3, 2000000, 
                                0, true);	
	nrfx_timer_extended_compare(&my_timer, NRF_TIMER_CC_CHANNEL4, 2500000, 
                                0, true);
	nrfx_timer_extended_compare(&my_timer, NRF_TIMER_CC_CHANNEL5, 3000000, 
                                0, true);
}

void main(void)
{
	const struct device *dev;
	bool led_is_on = true;
	int ret;

	dev = device_get_binding(LED0);
	if (dev == NULL) {
		return;
	}

	ret = gpio_pin_configure(dev, PIN, GPIO_OUTPUT_ACTIVE | FLAGS);
	if (ret < 0) {
		return;
	}

	// Initialize TIMER1
	timer4_init();

	timer4_repeated_timer_start();

	while (1) {
		gpio_pin_set(dev, PIN, (int)led_is_on);
		led_is_on = !led_is_on;
		k_msleep(SLEEP_TIME_MS);
	}
}

I get the following log output, as expected:

*** Booting Zephyr OS build v2.7.99-ncs1-1  ***
Timer 1 callback 0. Counter = 0
Timer 1 callback 1. Counter = 1
Timer 1 callback 2. Counter = 2
Timer 1 callback 3. Counter = 3
Timer 1 callback 4. Counter = 4
Timer 1 callback 5. Counter = 5

Please ignore that it says "Timer 1" in the log, I based the code on an existing sample using timer 1 instead of timer 4. 

Best regards
Torbjørn

[BvdP4Celsius]

Hi Torbjørn,

I implemented your suggestion and was able to narrow down the problem. In the original code, there is a statement under NRF_TIMER_EVENT_COMPARE4 event, that changed the NRF_TIMER_CC_CHANNEL5: nrfx_timer_extended_compare( &bus_timer, NRF_TIMER_CC_CHANNEL5, 16300 + rand(), 0, true );  

This will be a source of the problem, but only if the log level is LOG_LEVEL_DBG. When I change the log level to LOG_LEVEL_ERR it seems to work....

[ovrebekk]

Hi Bob

It is good to hear you found the problem. I didn't quite understand what the problem was though...

Do you mean the problem is in one of the Zephyr drivers, or in your own application code? 

If you think there is a problem in our code we should follow it up, and see if this is something that can be fixed in a future update. 

Best regards
Torbjørn