Separating Amplitude and Phase Imbalance in an IQ Image Rejection Test

An IQ conversion chain rejects an image by combining paths with the intended amplitude and phase relationship. A small mismatch can leave an unwanted sideband even when each path has reasonable gain. Diagnose both balance quantities rather than changing one gain setting until a single spectral line becomes smaller.

Identify the complete quadrature paths

The Marki Microwave IQ mixer primer relates image rejection and sideband suppression to amplitude and phase balance. The final result includes the surrounding quadrature network and signal paths, not only the mixer package.

Draw I and Q from their source or mixer outputs through filters, cables, gain stages, and combining or digital processing. Label polarity and the sign convention that determines which sideband is wanted.

Measure a controlled tone

Use a known tone and LO setting to identify the wanted and image frequencies unambiguously. Measure their power ratio with sufficient analyzer dynamic range and an input level that keeps the conversion chain linear.

Confirm that the apparent image is not a source spur or analyzer artifact. Change the tone offset in a controlled way and verify that the unwanted component moves as the expected image should.

Separate the two adjustments

Measure relative I/Q amplitude and phase at defined points, or use a supported calibration method that estimates both. Record the uncorrected values before changing compensation.

Adjusting amplitude can improve rejection when amplitude error dominates, but it cannot generally eliminate a phase error. A correction optimized at one frequency may also leave residual mismatch elsewhere because path delay creates frequency dependent phase error.

Verify over bandwidth and operating state

Sweep the required signal offsets and repeat at relevant RF frequencies, gain settings, and temperatures. Include the final filters and interconnects. A calibration made with short equal laboratory cables may not represent the installed paths.

Check that improved image rejection does not come from reducing the wanted signal or overloading another stage. Record wanted power, image power, and any other spurs separately.

Retain separate calibration coefficients for operating states that show materially different imbalance; interpolation between them also needs a check at intermediate frequencies.

Save the quadrature mapping, reference planes, measured imbalance, calibration settings, and residual image across the operating band. A satisfactory result explains which errors were corrected and where the correction remains valid. One deep image null at a convenient tone does not establish broad band rejection, nor does a well matched nominal component list prove that the assembled I and Q paths retain the required amplitude and phase relationship.