I'm designing a receiver in which we'll use an ADF4351 or ADF4355-2 (not sure yet) to drive the LO input of an ADL5380. Our application will need to cover from 500MHz to 4 GHz and there's no guarantee about which RF subband filters the customers will implement prior to the RF input. We were concerned about odd-harmonic mixing and the implications it would have regarding LO harmonic filtering.
We set up a test using two Vector Signal Generators locked in phase with a shared clock reference and outputting tones, one set to 500 MHz and the other set to 1500 MHz, and we combined their outputs with a combiner. That way we simulate a LO with a varying thrid harmonic level. We injected that combined signal as LO to an ADL5380 evaluation board we have. The Q output was loaded with a 50 ohm load and the I output was connected to a spectrum analyzer. We inyected a test signal through the RF port at 1520 MHz using a third vector signal generator, this running on its own internal reference.
We first set the 500 MHz level to a fixed 0 dBm and started increasing the level of the 1500MHz tone from -60 dBm upwards. The input RF signal level was set to 0 dBm at 1520 MHz. The spectrum analyzer was set to look from 10 MHz to 100 MHz.
We saw a 20 MHz tone at aproximately -19 dBm independently of the 1500 MHz tone level until we set its level above the 500 MHz tone (in that case the 1500 MHz would be the fundamental and the 500 MHz tone a subharmonic). When that happens the 20 MHz tone goes up to -6 dBm. We then set the 1500 MHz tone back to -60 dBm and set the RF input signal to a 0dBm 520 MHz tone, getting -6dBm 20 MHz signal at the I output.
That means that either our test is flawed in some way or the ADL5380 mixes whatever RF falls within the thrid LO harmonic band to baseband with only a 13 dB rejection. I've searched and it looks like the ADL5380 internally converts the LO to a square current wave to operate the mixer core. Is this thrird harmonic mixing a direct consequence of that? We've also looked that no matter how pure the LO is its leakage to baseband has harmonics at the same levels. In that case harmonic content at the LO input woudn't matter and no LO filtering would be neccesary, so a direct broadband interface to the ADF4351 would suffice. Is that right?
Thanks for your help
EDIT: According to CN-0245, which I read in the past but didn't remember, proper LO odd harmonics filtering is required for IQ balancing. I will include a filter bank that will suppress the third harmonic to below -60 dBm in each octave, but I'm still concerned about the third harmonics mixing. In our test we fed the ADL5380 with the purest tone we can generate here (from agilent E4438c's which have 3rd harmonic at -75 dBc).
Mensaje editado por: David Marcos