# IQ imbalance on quadrature signal

Hi I am attempting to develop a reasonable IQ balance mechanism for a quadrature signal coming from a software defined radio.I can balance out the amplitude and phase differences at one frequency quite well, about - 80dBm. However, over the typical span of my spectrum scope (192khz), there are differences at different frequencies. Now I have determined that the differences are quite linear. I could get away with doing a linear regression on a few points taken over the frequency range. But.. I am correcting the I and Q signals before the fast fourier transform. In other words, in the time domain. How do I bring this to the frequency domain? I have the real and imaginary amplitude and frequencies of course before I plot them. Can I use the same equations on the frequency values? Thanks

• If the phase offset is a linear function of the frequency (in the frequency domain), it indicates a fixed time delay (in the time domain).a; b; c; d; etc. You can correct a linear phase shift by a small delay in time before the FFT. – David Cary Jan 3 '15 at 1:35
• Yes I have determined however that the delay is a fraction of a sample so this complicates things considerably. – Tom Jan 4 '15 at 0:28