Is LSE required for STM32WLE5x Sub GHz radio?

[CLI.2]

We designed a customized board based on STM32WLE5JB SoC without the 32k LSE crystal to save cost. However the Sub-GHz radio doesn't work - LL_RCC_HSE_IsReady() always returns 0 and no radio interrupts generated. Is it because the LSE not being soldered? I didn't see the LSE is required for sub-GHz radio in the documents.

[Phoenix]

LSE is mandatory for LoRaWAN. I‘m sure I read it in the docs. But CubeMX as well pointed it out in the comments as well when you configure LoRaWAN. Saw it there just today.

These hints and the desired pinout are the main reasons I like to start from that tool before designing a schematic for a new board

[Tesla DeLorean]

Perhaps you can cite that.

The use of a TCXO based HSE for the radio/mcu is generally preferred as being on frequency significantly improves the link budget computations.

TCXO are harder to get these days given a fire at AKM who made die used in this type of clock.

The LSE in STM32 designs is frequently the only source that can be powered to maintain the RTC, and be used in low power sleep/standby type operation. This might be important for LoRaWAN if it sets specific timing dead-lines for responses, or subsequent communication.

An LSE with high accuracy could be used to quantify and calibrate a HSI/MSI clock of more dubious accuracy/stability.

[Phoenix]

You are right, it is not mandatory but recommended.

AN5042 is only about HSE selection for STM32WL

[Tesla DeLorean]

Ok, I seem to recall RAK Wireless or Seeed opting for crystals vs TCXO based on cost/availability in current crop of E5 modules.

Semtech wants better than 30 ppm under all conditions.

https://www.mouser.com/datasheet/2/761/DS_SX1261-2_V1.1-1307803.pdf

https://www.mouser.com/pdfdocs/an120014-xo-guidance-lora-modulation.pdf

https://tech-journal.semtech.com/expert-series-testing-devices-featuring-lora-how-to