nRF52832 external XL LF Clock

[lesockz]

Hello,

is it possible to use a external clock e.g. a digital output of a TXCO. As source for the LF clock. So the same clock source as used for the host uC can be used for the nRF52832.

Greets, Andi

[MartinBL]

Something like this maybe:

    NRF_CLOCK->LFCLKSRC = (CLOCK_LFCLKSRC_SRC_Xtal << CLOCK_LFCLKSRC_SRC_Pos) | 
                        (1 << 16) | // Set bit 16 in LFCLKSRC to enable (use with rail-to-rail external source)
                        (1 << 17); // Set bit 17 in LFCLKSRC to enable use of external source instead of Xtal (SRC needs to be set to Xtal)

I'll report to the SDK team that the defines are missing.

[Tmac]

We have a design with a 32MHz clock avaliable, its a digital clock rail to rail from 0-1.8V. We would like to use this signal to drive the HFXO and save the cost of the Xtal an synchronize the nRF52 to the external clock. Can you shed some light on why this should not be possible to do with the nRF52 as stated below. I should think that most xtal oscillators would work with an external clock directly or maybee we need to add some level shifting.

---

PS: For those who might wonder, external clock source for the 32 MHz clock is not supported. It used to be on the nRF51, but the clock module has been redesigned on the nRF52.

---

[Håkon Alseth]

@Tmac,

It's because it is sensitive to noise on the radio end, where the 32M signal is used to generate the 2.4G signal. A line that is routed between 2 ICs or more will pick up noise and then provide a clock that will degrade the RF part.

Cheers, Håkon

[Tmac]

Thanks for the quick feedback. So functionally it should work, but there is a risk of performance degradation, right? What kind of noise is the issue here? maybee its possible to filter the clock signal close to the nRF52 XC1 pin. If XC1 doesn't have to be driven rail to rail, and high frequency noise is the issue, we could add a simple R-C LPF close to the XC1 pin cutting off even below the 32MHz. This would attenuate the 32MHz to some degree and result in a sinewave input to the XC1 pin. It would also attenuate any high frequency noise coupled into the clock signal. Any idea if this would be feasible.

[Mickael]

HI there, just to double check: all of this should also apply to NRF52840, right?