R
Member
A mixer that should reject the image by 30 dB but only manages 15 dB is usually telling you about IQ imbalance, not about the mixer core itself. The bench sequence below separates amplitude error from phase error before you touch anything else.
First, drive the mixer with a single tone and look at the wanted sideband and the image sideband on a spectrum analyzer. The image rejection ratio is your baseline; write it down at several frequencies, because an error that is flat across frequency and one that grows with frequency point at different causes.
Second, measure the I and Q paths separately. Feed the same tone into each path and compare amplitude and phase at the IF port. A 0.5 dB amplitude mismatch alone costs roughly 25 dB of rejection, and a few degrees of phase error costs a similar amount, so small numbers here matter a lot.
Third, check what is frequency-dependent. Cable length mismatch, balun imbalance, and filter group delay all show up as phase error that drifts with frequency. If rejection is fine at one frequency and poor an octave up, suspect the surrounding network, not the mixer.
Fourth, separate LO leakage from the image. LO feedthrough sits at the LO frequency, not at the image frequency; confusing the two leads to tuning the wrong trim. Use a narrow span and marker measurements to keep them apart.
Finally, calibrate in the order that sticks: amplitude trim first, then phase trim, then re-check across temperature if the application needs it. A calibration that only holds at room temperature on one channel is not a fix.
What do you check first when image rejection degrades: the IQ balance, or the surrounding balun and filter network?
First, drive the mixer with a single tone and look at the wanted sideband and the image sideband on a spectrum analyzer. The image rejection ratio is your baseline; write it down at several frequencies, because an error that is flat across frequency and one that grows with frequency point at different causes.
Second, measure the I and Q paths separately. Feed the same tone into each path and compare amplitude and phase at the IF port. A 0.5 dB amplitude mismatch alone costs roughly 25 dB of rejection, and a few degrees of phase error costs a similar amount, so small numbers here matter a lot.
Third, check what is frequency-dependent. Cable length mismatch, balun imbalance, and filter group delay all show up as phase error that drifts with frequency. If rejection is fine at one frequency and poor an octave up, suspect the surrounding network, not the mixer.
Fourth, separate LO leakage from the image. LO feedthrough sits at the LO frequency, not at the image frequency; confusing the two leads to tuning the wrong trim. Use a narrow span and marker measurements to keep them apart.
Finally, calibrate in the order that sticks: amplitude trim first, then phase trim, then re-check across temperature if the application needs it. A calibration that only holds at room temperature on one channel is not a fix.
What do you check first when image rejection degrades: the IQ balance, or the surrounding balun and filter network?
