Clock inaccuracy of nRF52832 custom board

RE: Clock inaccuracy of nRF52832 custom board

barky — Wed, 04 Jul 2018 08:31:54 GMT

Radio was not working because pins DCC and DEC4 were shorted. From typical application example i made a mistake, it looked for me that those pins were together...

RE: Clock inaccuracy of nRF52832 custom board

Kenneth — Tue, 05 Jun 2018 08:09:26 GMT

Hi, to use the softdevice api you will first need to enable the softdevice.

RE: Clock inaccuracy of nRF52832 custom board

barky — Fri, 01 Jun 2018 12:00:03 GMT

int main(void)
{
    /* Configure leds. */
    nrf_gpio_cfg_output(LED_3);

    //Initialize HFCLK 
	uint32_t p_is_running;
    
    (void)sd_clock_hfclk_request();
    do { 
        sd_clock_hfclk_is_running(&p_is_running); 
    }while(p_is_running==0);

    while (true)
    {
        nrf_delay_ms(500);
        nrf_gpio_pin_set(LED_3);
        nrf_delay_ms(500);
        nrf_gpio_pin_clear(LED_3);
    }
}

I tried this code and it stuck, when using softdevice.

From infocenter: "parameter p_is_running results in 1 if the external crystal oscillator is running, 0 if not."

Moreover, return value of "sd_clock_hfclk_is_running" is 0 or "NRF_SUCCESS" if successful.

I found issue, quite relevant: Case ID: 117750 ,where DC-DC power caps were cold soldered, maybe my case is similar, even though i checked all connections and values of passive components (i am useing internal LDO configuration).

RE: Clock inaccuracy of nRF52832 custom board

Kenneth — Thu, 31 May 2018 09:22:20 GMT

If the softdevice is enabled, then you need to use:

sd_clock_hfclk_request();
while(!sd_clock_hfclk_is_running());

...

sd_clock_hfclk_release();

RE: Clock inaccuracy of nRF52832 custom board

barky — Wed, 30 May 2018 19:22:47 GMT

Yea it is...

I used blinky example with with s132 softdevice, added lines:

 NRF_CLOCK->EVENTS_HFCLKSTARTED = 0;
 NRF_CLOCK->TASKS_HFCLKSTART = 1;

 nrf_delay_ms(300);
 // Wait for the external oscillator to start up
 while (NRF_CLOCK->EVENTS_HFCLKSTARTED == 0) {}

And its still stuck...

I have the LF_clock also on the board, maybe configuration with it are messing with software.

RE: Clock inaccuracy of nRF52832 custom board

Kenneth — Wed, 30 May 2018 16:06:11 GMT

Is the softdevice enabled?

RE: Clock inaccuracy of nRF52832 custom board

barky — Wed, 30 May 2018 15:55:26 GMT

As i mentioned earlier, the clock became accurate after placing crystal in correct position (90 degrees).

For example, each bit lenght of uart message is now 104uS as supposed to, but program halts on:

//while (NRF_CLOCK->EVENTS_HFCLKSTARTED == 0) {}

loop. So crystal kinda works.

As i read in other threads, maybe there is wrong capacitors placed on DEC pins...

Any suggestions?

RE: Clock inaccuracy of nRF52832 custom board

dgerman — Tue, 22 May 2018 16:36:40 GMT

It happens. Good that you mentioned it here.

RE: Clock inaccuracy of nRF52832 custom board

barky — Tue, 22 May 2018 16:15:30 GMT

Looks like the case can be closed. Sorry for my frivolity.

Haven't every made this mistake, people remember to look in datasheets if layout of part is not drawn from the bottom view, not the top!

Therefore XTAL X1 and X2 were connected to board GND pins.

RE: Clock inaccuracy of nRF52832 custom board

barky — Tue, 22 May 2018 10:45:35 GMT

[quote userid="2111" url="~/f/nordic-q-a/34423/clock-inaccuracy-of-nrf52832-custom-board/132703"]

No need to set the frequency, as there is only one crystal supported (32MHz).

Does the radio work? Are you able to transmit and receive data

[/quote]

Radio doesn't work. I tried beacon example, and it is not advertising.

I have connected 32MHz crystal like this:

The crystal is "LFXTAL055595" with 10pF load capacitance. I added 12 pF load capacitors. Maybe i fried the crystal while soldering. Will try resoldering another one.

RE: Clock inaccuracy of nRF52832 custom board

Kenneth — Tue, 22 May 2018 09:43:57 GMT

No need to set the frequency, as there is only one crystal supported (32MHz).

Does the radio work? Are you able to transmit and receive data.

RE: Clock inaccuracy of nRF52832 custom board

barky — Tue, 22 May 2018 07:56:37 GMT

Well yes, the frequency will shift, because of the probe capacitance, but waveform should have been visible.

I will continue actively debugging the clock. I added lines to code, but program halts, meaning that it is stuck in a while loop:

 // Start the external high frequency crystal
 NRF_CLOCK->EVENTS_HFCLKSTARTED = 0;
 NRF_CLOCK->TASKS_HFCLKSTART = 1;

 nrf_clock_lf_actv_src_get();
 // Wait for the external oscillator to start up
 while (NRF_CLOCK->EVENTS_HFCLKSTARTED == 0);

Isn't it needed to set the clock frequency e.g. 32MHz or 16MHz somewhere?

RE: Clock inaccuracy of nRF52832 custom board

AmbystomaLabs — Mon, 21 May 2018 14:01:39 GMT

You can never measure clock at a crystal with probes. The probes will always load pull the crystal. The easiest way to measure frequency accuracy is with a calibrated spectrum analyzer with the nRF producing CW. Also you can route P_Clk out through gpio and measure that.

RE: Clock inaccuracy of nRF52832 custom board

barky — Mon, 21 May 2018 07:21:23 GMT

Thanks you for the tip.

Problem stays the same. I tried to measure the waveform on 32MHz crystal, but all i see on osciloscope is 100mV DC on both X1 and X2 pins.

RE: Clock inaccuracy of nRF52832 custom board

Kenneth — Wed, 16 May 2018 19:08:02 GMT

Hi, have you started the external crystal oscillator by calling NRF_CLOCK->TASKS_HFCLKSTART = 1;?

http://infocenter.nordicsemi.com/topic/com.nordic.infocenter.nrf52832.ps.v1.1/clock.html?cp=2_1_0_18_2#topic