Sure, the key is to keep the signal clean as soon as it leaves the microphone. Start with a low‑latency adaptive filter that tracks the channel’s impulse response in real time – that helps you cancel out echoes and multipath. Pair that with a lightweight spectral‑weighting noise gate that targets storm‑related broadband hiss; you’ll need a fast decision window so you don’t introduce delay. Use a robust modulation scheme like OFDM with subcarrier reallocation; it’s tolerant of deep fades and allows you to shift power to the subcarriers that are still clear. Finally, implement a quick‑look packet‑retransmission scheme for any dropped frames—reducing latency is critical, so keep the recovery window under 20 ms. With those layers, you’ll see a noticeable jump in clarity without breaking the real‑time constraint.

OFDM can fit inside a 10 MHz LTE slice or even a 5 MHz 5G NR slice if you keep the subcarrier spacing at 15 kHz or 30 kHz. Just shrink the number of active subcarriers to match the available spectrum and use adaptive coding on the remaining ones. That way you don’t waste bandwidth and you still get the resilience OFDM offers. For packet‑retransmission, keep it lightweight: use a sliding‑window ACK scheme and limit the window size to a few packets. Drop the “go‑back‑N” style logic and switch to selective repeat. That keeps the pipeline moving because you only re‑send the packets that truly failed, and you can do it in a separate thread so the audio stream keeps flowing. Timing is critical; aim for a retransmission window of 10–15 ms so the delay stays below the 20‑ms real‑time threshold.

Just a couple more tweaks: run a quick voice‑activity detector so you can mute the noise gate when there’s no talk, that cuts down on false positives during storms. Also consider a lightweight interference suppression filter that hunts for the specific lightning‑induced spikes; a high‑pass with a notch at the mains frequency can shave a few dB off the hiss. Finally, keep a small buffer for the codec so you can swap in a higher‑order mode if the channel improves, but stay under that 20 ms headroom.