LFSYNTH additional current and accuracy

[rgrr2]

Hi

in our device, HFXO is running permanently.  There is an external LFXO running permanently as well.

Now I'm wondering what would be the additional current draw if we switch from NRF_CLOCK_LF_SRC_XTAL to NRF_CLOCK_LF_SRC_SYNTH.  Any data available about this?

Because our application depends on an accurate LFCLK: how actually is LFSYNTH generated from HFCLK?

Thanks for help

Hardy

[Simonr]

Hi Hardy

According to the PS, the HFXO clock consumes 250µA by default and adding the LF SYNTH (specced at 100µA run current) you'd get ~350µA there. The LFXO will only consume ~0.25µA so the difference between the two would be almost ~99µA seeing as the HFXO is running regardless.

As for the LFSYNT accuracy, I'll need more details to know what you've got running that uses the HFXO as well and ask one of our experts what sharing the HFXO will have to say for the initial 8ppm + the HFXO tolerance that is stated in the spec. The LFXO is rated at a frequency tolerance of +-250ppm.

Best regards,

Simon

[rgrr2]

Hi Simon

thanks for the information.  Because current consumption is essential in my application: where can I find those 100uA in the PS?

Regards

Hardy

[Simonr]

Hi

Sorry, I was looking at the nRF52832 PS, which includes the ILFSynth of 100µA. This has been removed from newer product specs, as you generally can't add these current consumption values to find the correct current consumption, since current consumption might be subsets of others, etc.

That's why we've chosen to have a separate chapter for current consumption in newer PS's. I'm sorry that the LFSYNT is not added here though, and that seems to be a mistake from our side. I have reported it internally, so we'll be looking into this on our side.

Regardless, we did some quick measurements here on our side (which should be taken with a grain of salt, and are not official measurements), but they showed that the LFSYNT draws ~375µA when running if from the HFRC, and noticeably more, ~614 when running from the HFXO. So if current consumption is essential in your application, I highly recommend going for the LFXO crystal as that provides the best combination of low power consumption and high accuracy for your low-frequency clock.

Best regards,

Simon

[rgrr2]

Hi Simon

thanks again.  In the application both LFXO and HFXO are yet permanently running.

But out problem is, that temperature drift of the LFXO is to high.

So the ideas are

• use some drift compensation formula, which is crystal dependent (so another manufacturing step to consider)
• use LFSYNTH (so LFXO can be switched off at all) because the temperature drift of the HFXO is much smaller (small enough at least)
• use HFXO and some TIMER to measure the LFXO drift from time to time

Any more hints?

Thanks

Hardy

[Simonr]

Hi

I see. If a large temperature deviation is expected, you might want to adjust the loading. The drift is indicated by the "Parabolic coefficient" parameter in the datasheet. Seeing as the LFSYNTH draws notably more current than the LFXO, I would still suggest using the LFXO, and if necessary you can add either a timer from HFXO or some further steps to compensate for the expected drift. You might also be able to find a 32.768 kHz crystal that has less temperature drift than the one you're looking at.

Best regards,

Simon