LO leakage in direct conversion - calibration routine?

Direct-conversion TX has LO leakage 30dB above spec before calibration. What's your production calibration flow? One-time at factory, or periodic re-cal in the field?

30 dB above LO leakage before cal is common on direct-conversion TX; the production question is whether you cal once at cold, over temp, or per-channel.

A flow that usually sticks:
1. Factory: measure LO feedthrough into the TX port (or image of DC offset) at a fixed mid-band tone with PA in a known bias; solve I/Q DC offsets + gain/phase imbalance; store per-unit (or per-SKU) coefficients.
2. Verify residual leakage against mask at band edges, not only mid-band — LO and baseband filters walk.
3. Decide temp: if leakage drifts out of mask over the product temp range, add a coarse temp LUT or a short power-up re-cal; if it stays inside, one-time is enough.
4. Guard against PA AM-AM / supply sag changing the apparent leakage after the cal tone — cal at the same average power you ship.

One-time mid-band only works when LO leakage is dominated by static baseband offsets. If mixer isolation or PA feedthrough dominates, you need RF-side isolation work before more digital cal. What LO architecture (on-chip vs discrete) and PA class are you shipping?

— rfforum 编辑部 / Editorial Team