STM32WB55: Power supply ripple/pulsation causing BLE TX to fail

[Aku]

We've switched to new DC-DC(LM3281) for main power supply and found that BLE TX stopped working correctly and not passing packet test anymore.

After deeper investigation we found that our DC-DC in ECO mode has ripple/pulsation of 60mV/50uS, which causing this issue.

Switching DC-DC into PWM mode solves the issue, but power consumption raise makes this option undesirable.

Issue can be reproduced on Nucleo board with Transparent VCP test firmware and CubeMonitor-RF in packet test mode.

Questions:

• How clean power supply should be and where this information can be found?
• Is there any way to expose radio phases that requires clean power supply to GPIO?
• Is it bug in core2 firmware or expected behavior?

[Aku]

@Remi QUINTIN​ Looks like we need your help again ;)

[Remi QUINTIN]

We are internally brainstorming about the possible causes of such artefact.

VDD ripple/pulsation of 60mV/50uS could (to be confirmed) have an impact on the LSE stability leading to RTC timers being drifted and then BLE scheduling getting corrupted => TX stopped working correctly ...

Now using an LDO instead of a DC-DC could solve the issue. For sure the PWM mode seems also OK.

Anyhow we are continuing our discussions to understand how such noise can arise and propagate in such ECO configuration.

Could you add a picture of just the power schematics?

The RF block is controlled by the wireless stack with no way to export any information related to the RF activity phases outside the M0+ core.

[Remi QUINTIN]

Adding an LDO to filter out the ripples coming from the DC-DC converter may also be a possible solution.

[Aku]

We've checked LSE first, it's super stable.

We use 12pf xtal with LL_RCC_LSEDRIVE_HIGH. Also CSS enabled for both domains, there were no issues.

Capturing LSE clock with signal analyzer and counter shows no issues.

Full schematic in attachment: DA6 (Power) provides power to MCU.

Core2 already have ext PA control ability, is it silicon or software? Maybe somethings like that can be exposed to GPIO?

[Aku]

Couple new findings:

• Only rising edge causes issues, issue appears if dv/dt > 4V/ms (LM3281 got >15V/ms)

[Aku]

Correction: clean VDDRF is not enough.

[Remi QUINTIN]

Can you capture a picture of the spectrum around the channel frequency used for the communication?

The question behind is whether the DC-DC converter is impacting the power or the RF area via spurious harmonics.

[Aku]

Those are previous test samplesThose are new:

Packet test is failed, there is a small spikes after Carrier Test that can be seen, and then silence.

[Jonny]

Hi Aku,

I'm working on a project that uses the same micro and we did the same changes on supply due to electronics shortage and I'm having a problem with the BLE connection like you.

Have you solved the problem somehow?

Thanks for your help

Jonny