STM32WL Minimal RF circuit for LoRaWAN endnode

[Julian1]

hey there!

I'm planning to design my own PCB based on the STM32WL55 (900MHz range) and I have the NUCLEO reference design to look at, but find that here's a lot of RF paths in there, and I just want to know if there's any discussion anywhere on how much of all this is actually needed.

ideally I would like to get rid of those RF switches and multiple paths, but I'd like to understand a bit more about what would happen if I did. And I'd also like to do that as fast as possible with as few PCB iterations as possible. :)

1) does the Rx path really need to be isolated from the TX paths with an RF switch rather than just linked up? (with impedance matching of course!) if so, what are we achieving with the switch?

2) What would the penalty be (assuming LoRaWAN!) if I was to just go with one PA output? I'm guessing that depending on what one I picked, either I'm transmit power limited or I'm limited in how much I can reduce battery draw during good quality link conditions... but I also need to understand what could happen to the device on the network and what would need to be modified with the existing cubeWL firmware if I choose to reduce the design this way?

Ideally, any overview description of the design requirements would be good. As it is I haven't found anything beyond the one application note which is a bit light on details about only using one TX path, and doesn't really seem to talk much about the RX path switch beyond saying you need one.... and the reference design of the nucleo board... everything else I found focuses solely on the bits inside the RFSOC itself.

[TZhou.2]

Hi, does it support NFC? thanks

[Julian1]

I think you're in the wrong place?

but to answer your question directly: I guess you could add an NFC IC to a design with the STM32WL...

[YBOUV.1]

Hi @Julian​ ,

You can get rid of the RF switch. By tightening together Rx and Tx, the sensitivity you will achieve will be degraded.

Choosing one of the PA output depends on your requirements. As you said, if you choose the Low Power PA, ouput power will be limited to 15dBm @IC ouptut. But if you choose the 'high power PA', the efficiency for output power lower or equal to 15dBm will not be as good than for the 'low power PA'.

In term of cubeWL, it supports this configuration, but you will have to update the BSP accordingly,

[Julian1]

thanks so much for your reply @YBOUV.1​ 

Do you have any idea of the kind of sensitivity loss I should expect if I leave out the rx path switch?? Obviously the received signal will split down two paths this way, so my receiver pin will only see half the signal... but is the sensitivity loss likely to be worse than 6dB in practice?

And has any of this actually been tried to your knowledge, or does every implementation you know of just use a switch here?

Still, good to know that at least there's not a likelihood of blowing up the Rx input pin with directly coupled transmit signal.

[YBOUV.1]

Hi,

I would expect around 3dB degradation in sensitivity