STM32WL RX noise figure - adding an external LNA

[Albi G.]

Hi,
I am current reading the material to develop a LoRa application. My PCB is wicked small, therefore the chip-antennas wont have a good ground plane and therefore bad efficiency.

While reading about the RX-path, I stumbled upon AN5457 where in figure 83 on page 64/75 the LNA-noise figure is depicted as a miserable 8dB.

This generated the urge to place an external LNA in front of the receiver chain. Here is my reasoning, assuming external LNA-parameters to be v=13dB, NF=0.8dB, Ic=5mA:
- the v=+13dB compensate for my bad antenna
- the resulting overall noise-figure of ~1.66dB gains ~6dB SNR
- therefore, overall the link-budget increases by 6dB+13dB = 19dB.
- all this while only doubling the RX-current consumption (from 4.8mA to ~10mA).
- with 19dB higher signal (or even just with the 6dB better SNR alone) the LoRa-datarate can be increased (quadrupled), thus reducing the RX-on-time by 4 and/or saving massively during the power hungry TX.

Now, I wonder, if my assumptions were true, it would be a no-brainer to include a much better LNA directly into the STM32WL. From a nanojoule-per-bit-perspective, there is literally no reason ever why the designers of the STM32WL would be contempt with a 8dB noise figure.

I therefore want to double check and would be very thankful if someone could review my assumptions and confirm that adding an external LNA is actually the low hanging fruit it seems to be.

Thank you

[Mystical]

Are there any updates on this? If an external LNA could actually make this much of an improvement, an official response would be very useful. 

[Albi G.]

I wish there was an answer. I honestly have iced the project due to lack of support and questionable design decisions regarding the limited output power when VDDIO is restricted to 1.8V.

If the team behind the ICs is not explaining how to get decent performance, there is probably no decent performance to be found.

I think there is the underlying problem that the link-budget of LoRa is already that high, that the motivation to get the last few dB out of the system design is pretty low.

I wish ST would care.