A real transceiver corrupts its own signal before it ever reaches the air. It leaks its oscillator, it never matches its I and Q paths perfectly, and both drift with temperature. Digital correction, driven by calibration, is how the radio undoes its own errors.
A direct-conversion (zero-IF) transceiver is compact and cheap, which is why it dominates — but it adds characteristic errors. DC offset and LO leakage: the local oscillator bleeds into the signal path, producing a DC term and a spike right at the carrier. I/Q imbalance: the in-phase and quadrature paths are meant to be identical and exactly 90° apart, but gain and phase never match perfectly, which creates an image — a mirror copy of the signal at the opposite frequency.
Left alone these degrade everything downstream: the EVM rises, the constellation skews and offsets, and energy leaks where it should not. At high order modulation they are the difference between a link that closes and one that does not.
Rather than trim the analogue perfectly — expensive and still drifting — the modern approach measures the impairments and corrects them digitally. DC offset is removed by estimating and subtracting the DC term (and nulling the LO leakage). I/Q imbalance is compensated by a correction that applies the inverse gain and phase, restoring the balance between the two paths and cancelling the image.
The same applies on transmit: predistorting the I and Q so the imperfect modulator produces a clean signal. It is cheap, flexible and done in the same baseband DSP that already handles the waveform — the cost is the estimation and the correction logic, not new analogue.
The impairments are not constant. They shift with temperature, supply voltage, frequency and component ageing, so a one-time factory trim is not enough. Calibration tracks them — at power-on, periodically, or continuously in the background — and updates the correction coefficients so the compensation stays valid across the operating envelope.
Some systems calibrate against a known internal tone; others estimate blindly from the live signal's statistics. Either way, the correction is only as good as the calibration behind it: fix the numbers once and let them drift, and the image and offset creep back exactly when the part heats up under load.
Impairment correction sits in the digital front end next to the other baseband DSP, and it is foundational: CFR, DPD, beamforming and the receiver's own synchronization all assume a clean, balanced signal to work on. It is unglamorous and almost always necessary — the quiet layer that lets a cheap, integrated transceiver behave like a far more expensive one.
For a modem or radio design, the work is the estimation and calibration: how to measure DC offset and I/Q imbalance accurately and cheaply, how often to recalibrate, and how to apply the correction without adding latency the rest of the chain cannot afford.
In a direct-conversion transceiver, chiefly DC offset and local-oscillator leakage (a spike at the carrier) and I/Q gain and phase imbalance (which creates an image of the signal). Both degrade EVM and spectral performance and drift with temperature and ageing.
The in-phase and quadrature paths of a transceiver are meant to be identical and exactly 90 degrees apart, but their gain and phase never match perfectly. The mismatch creates an image — a mirror copy of the signal at the opposite frequency — which corrupts the constellation.
Digitally, by applying the inverse gain and phase correction so the two paths are rebalanced and the image cancels. On transmit, the I and Q are predistorted so the imperfect modulator produces a clean signal. It is done in the baseband DSP, driven by calibration.
Local-oscillator leakage into the signal path in a zero-IF receiver, plus device mismatches, produce a DC term and a spike at the carrier. It is estimated and subtracted in the digital domain, with the LO leakage nulled.
Because the impairments drift with temperature, supply, frequency and ageing, a one-time factory trim is not enough. Calibration at power-on, periodically, or continuously updates the correction coefficients so the compensation stays valid as conditions change.
Digital correction is cheaper, more flexible and re-tunable than trimming analogue precisely, and it can track drift. The analogue stays imperfect and the baseband DSP undoes its errors, which is how a cheap integrated transceiver reaches the performance of a far more expensive one.
Where I/Q imbalance shows up.
KNOWLEDGEThe DSP next door.
KNOWLEDGECFR and DPD assume clean I/Q.
KNOWLEDGEPer-element calibration.
PRODUCTWhere correction lives.
KNOWLEDGEAnother RF cleanliness problem.
Measuring DC offset and I/Q imbalance accurately and cheaply, and recalibrating as the part drifts, is the quiet work that makes an integrated transceiver perform. Tell us the transceiver and the band.