Joint TX and feedback RX IQ mismatch compensation for integrated direct conversion transmitters

A Hybrid Scheme for TX I/Q Imbalance Self-Calibration in a Direct-Conversion Transceiver

“…In fact, due to the wide use of DSPs, it has become easier to build RX IQ imbalance schemes within a DSP. However, this kind of scheme is not robust for TX I/Q calibration because the performance in a TX is strongly dependent on RX calibration results [18,24]. To overcome the TX I/Q imbalance issue, a calibration scheme in a direct-conversion transceiver is shown in Figure 1.…”

“…When performing TX I/Q calibration, the Q path is set to zero. Different from the most used loop-back methods proposed in [18,24], in order to eliminate the extra I/Q imbalance caused by the RX LO, our loop-back signal was squared by the squaring circuit (u 2 in Figure 1) and directly output into the low-pass filter (LPF) in the I path without passing through the RX LO. Then, the squared signal was filtered by the LPF, converted by the analog-to-digital converter (ADC) and output into the digital domain.…”

“…However, in order to extract the inherent characteristics of the signals, the blind method requires high-complexity calculation blocks, thus usually requiring external digital signal processing (DSP) for the calibration of the whole system [14][15][16]. On the other hand, training sequence methods are easier to implement, and are widely used for power-on calibration (self-calibration) since they can implement a high-precision I/Q imbalance estimation by sending a designed training sequence [17][18][19]. Furthermore, any training sequence method for a frequency-independent I/Q imbalance calibration can be extended to calibrate a frequency-dependent I/Q imbalance by utilizing a finite-impulse response (FIR) filter in the system, since the frequency-dependent I/Q imbalance can be further compensated by changing the tap coefficient of the FIR [20,21].…”