nRF52 on custom board can receive BLE packets, but not send

[SenseJo]

Hello, we have a custom board with an nRF52840, for which we are developing using NCS / Zephyr.

While reading a button and blinking the on-board LEDs works fine (more peripherals remain to be tested), Bluetooth (Low Energy) behaves strangely:
When flashing the observer sample from zephyr/samples/bluetooth/observer, we can see that the board can receive BLE packets with reasonable RSSI values.
However, when flashing the beacon sample, we can't find the board (using nRF Connect for Android). The log output usually is okay.

Just sometimes, with vanilla Zephyr, just after "Beacon started", we expire an assertion fail:

Bluetooth initialized
Beacon started, advertising as CB:A6:83:F2:75:8F (random)
ASSERTION FAIL [0] @ WEST_TOPDIR/zephyr/subsys/bluetooth/controller/ll_sw/nordic/lll/lll_adv.c:1044
        prepare_cb: Actual EVENT_OVERHEAD_START_US = 46783
[00:00:01.003,662] <err> os: r0/a1:  0x00000003  r1/a2:  0x00000000  r2/a3:  0x00000002
[00:00:01.003,692] <err> os: r3/a4:  0x200005f8 r12/ip:  0x00000010 r14/lr:  0x000006a1
[00:00:01.003,692] <err> os:  xpsr:  0x41000028
[00:00:01.003,692] <err> os: Faulting instruction address (r15/pc): 0x000006ac
[00:00:01.003,753] <err> os: >>> ZEPHYR FATAL ERROR 3: Kernel oops on CPU 0
[00:00:01.003,753] <err> os: Fault during interrupt handling
[00:00:01.003,784] <err> os: Current thread: 0x20000c58 (unknown)
[00:00:01.345,825] <err> os: Halting system

However, I am not sure this is related, because when this assertion fail does not occur, and also with NCS, we can't find the board anyways.

Since we don't have an external 32 kHz crystal on the board, we have CONFIG_CLOCK_CONTROL_NRF_K32SRC_RC=y in the boards _defconfig file. We also tried CONFIG_CLOCK_CONTROL_NRF_K32SRC_SYNTH=y instead as well.

We have tried lots of different combinations of supplying the nRF52 with power, since our board has exposed the relevant pins to jumpers.
In the boards Kconfig file, we currently select SOC_DCDC_NRF52X and in addition, we also had SOC_DCDC_NRF52X_HV when the chip was supplied in high voltage mode.

Here is the MCU page of our schematics:

Any ideas on this are appreciated!

[ketiljo]

Is it correct that you're using a 40 MHz crystal? 

This needs be 32 MHz, nothing else will work. You're missing the crystal load caps. Which, for a crystal with a CL value of 8 pGF, should be 12 pF each. 

See the reference schematic: 

https://infocenter.nordicsemi.com/topic/ps_nrf52840/ref_circuitry.html#unique_1496623802

Note that _all_ components here, except where noted, are mandatory. You're missing the 15 nH inductor in series with L302. The matching netwotk on the ANT pin must be there are well.

[SenseJo]

Hi ketiljo,

thanks a lot for your prompt response.

Sorry about the confusion about the crystal, it's just the label in the schematics which is wrong, but our boards are populated with 32 MHz crystals.

We soldered some load caps to the crystal and an inductor in series with L302 on a board, but unfortunately, nothing changed.

This is our matching network on the ANT pin:

[ketiljo]

What's the spec on the crystal? The caps needs to match the CL value of the crystal as follows:

Ccap = CL * 2 - Cpin, where Cpin = ~4 pF

[SenseJo]

Thanks again for your response!
We checked the caps on the crystal again. They have the right capacity (12 pF).
But it seems one of them wasn't soldered properly – after we resoldered it, we can now find the chips advertising packets!

Thanks again!