I am using the IQ demodulator LTC5584 in SDR. I have measured imbalances. The amplitude imbalance seems to be OK, the measured ration was 0.2dB. The phase imbalance is issued, I have measured ~18deg, the datasheet guarantees ~1deg. The graph is below. I am using the Eval board DC1670A. LO frequency is fixed: 307MHz (LO port is matched as well as RF port). 
Meas procedure: I used CW signal at the input of IQ-demod, the IQ-outputs were sampled at dual ADC, sampled data loaded to file, calculated convolution function between I and Q data and fitted to sin() - function. As you can see, the imbalance is constant, it does not depend on the BB frequency.
Do you see any reason for such high phase imbalance? Is there any requirements on the harmonics content at the LO-port?
