iq mixer - harmonics and unwanted frequencies

Thread Starter

Omri Shachak

Joined Mar 27, 2018
I am having difficulties understanding the unwanted frequencies analysis in REALITY and not in theory.
i am aware that whenever you use a mixer, for example RF and LO mixed to get IF, the IF frequencies will ideally be consisted of (f_lo+f_rf) and
(f_lo-f_rf). In my design i am interested only in the (f_lo-f_rf) part.
I am working with hmc1056 from analog devices(Link at the bottom) , and i am wondering if i need to put a low pass filter in the outport IF, in order to get only (f_lo-f_rf). i don't understand why in the data sheet of the component they write that the IF frequency band is DC to 4GHZ, does that mean the if the f_rf = 10 GHz and f_lo=10GHz than in the IF port i will not get the (f_lo+f_rf) = 20GHZ? so should i put a low pass filter or not after all ?
Another question concerning the harmonics in the mixer. Let's assume i can use any quantity of different reflection less filter from the mini circuits X series in order to diminish the harmonics. where would it be useful to put them? in the IF port or also at the RF and LO ports as well ?

thanks in advance.


Joined Nov 6, 2017
The way I read this is that their 90 degree hybrid is designed to operate in this range (8-12G), as is their mixer presumably, not that you wouldn't see output spectrum outside of [-4G, +4G]. Note that the image rejection specs fall off towards the edges of the 8-12G range, with gain and phase mismatch fairly centered in it. This the likely motivation for specifying valid IF output as DC to 4GHz and would not imply that higher frequencies are removed/absent; it may simply mean that spectra outside of this range are likely to behave poorly as a (de)modulated output.

The next question of where to place filters depends a bit on your signal. Does your input have spectra at n*fc and/or m*LO that are concerning? This depends on the response of the mixer to harmonics of the LO as well (i.e. if the harmonics of the LO are strongly attenuated in the multiplier, then there is already less concern). For the LO itself, you can filter it before sending it to the mixer, but if the mixer has a gain stage for the LO to drive it to the rails anyway, an external filter wouldn't eliminate harmonics of the LO into the mixer's multipliers.

In short, I think this means:
- filter at the input if spectra exists at higher harmonics and would be problematic (may need to characterize LO harmonic recip mixing gains)
- filter at the output to remove: IF image, LO + harmonic spurs, LO harmonic modulation outputs

Good luck,