Best practices for reducing latency when streaming I/Q data from BladeRF into Python?

[mikejohnson]

Hi everyone,

I’m working with a Nuand BladeRF SDR and trying to stream I/Q samples into Python for real-time signal processing. I’m currently using the standard libbladeRF + Python bindings, but I’m noticing latency spikes and occasional buffer overruns at higher sample rates.

What are the best practices to improve performance especially in terms of buffer sizing, threading, or using alternatives like C++ preprocessing before Python?

Has anyone here optimized a stable low-latency pipeline for continuous I/Q streaming?

Any suggestions or example setups would be really appreciated.