Consistent LO Leakage & Distorted Cosine in Loopback / IQ Replay

[skrt8]

I’m testing loopback and IQ replay on a bladeRF 2.0 micro (A4) using Soapy/GR. My flowgraph sends a simple cosine at 30 MSPS → bladeRF sink → bladeRF source, with DC removal on/off.

Observation:

Sometimes the cosine comes back clean, other times it is distorted.

In both cases, a strong LO leakage spike is always present at the center frequency.

Same behavior whether using live loopback or replay from file.

Question:

When samples are sent into the AD9361 FIFOs in the Quartus project, is there any chance of sign extension / DC bias being introduced internally? Or is the LO leakage purely expected behavior of the AD9361 front-end?

I want to confirm if the HDL wiring into the AD9361 core (FIFO → IQ interface) guarantees zero DC injection, so I can rule out HDL issues and focus only on analog/LO causes.

[emmascott63]

From what I’ve seen, the LO leakage is pretty normal on the AD9361. The distortion showing up only sometimes makes me think it’s more about buffering or sample alignment than the HDL wiring itself. I’d lean toward checking levels and making sure there’s no clipping, rather than worrying about sign extension.