I designed an IQ transceiver board incorporating the ADL5380 as downmixer operating at 5.8 GHz (center LO) ISM band.
I had manufactured some prototypes (~20) and experienced some weird behavior:
Receiving an externally generated signal at f.e. 5.81 GHz, the mixer shows an IQ amplitude imbalance of 1 to 6 dB, resulting in an image rejection of about 30 to as low as 12 dB, which is way below the expected figures.
At 5.8 GHz the datasheet shows an IQ imbalance in the range of 0.05 to 0.15 dB.
Does someone know, where the problem could be originated?
I'm operating the mixer nearly identical to the reference design with 3 exceptions:
1) I'm using the HHM1595A1 balun, as I already used them for the upmixer.
2) I'm using a 4p7F RF cap for DC decoupling, which should be fine for 5.8 GHz.
3) Instead of a transformer, I'm using a fully differential opamp in unity gain configuration. Vocm at R-C-filtered input CM voltage (10k from pos/neg to 100nF cap). Rfb = Rg = 330 ohms each, which results in a differential input impedance of round about 500 ohms.
However, for testing purposes, I already fully removed any load in a first step and just added a 499 ohm differential termination resistor in a second step - each step resulting in unchanged behavior.
As each PCB shows an imbalance way above the stated specs, the error must be originated in the design.
Is there any potential obstacles when using the ADL5380, which can drastically reduce the image rejection? (Maybe besides the output loading, as I already removed any load.)
Thanks for all help.