I'm using an active differential Gilbert cell mixer with a tuned load for frequency doubling. The two input ports of the mixer have the same signal. However at the output, i get a large up-converted signal swing at the '+' termainal  and a very small swing at the '-' terminal. The circuit is in CMOS and i simulated with Spectre. Aren't the amplitude of the outputs to be the same?

Has anyone encountered this problem before? Is it a  problem with the simulator or the transistor models? Or anyone has a physical explanation for this?

Thanks for your suggestion.

I tried driving the mixer harder with unreasonable large swing . It didn't help and in fact, the output swings became even more unbalanced. I also get the same unbalanced output swings when i use BJTs.

What can be the explanations? Or how can i simulate the gilbert mixer as doubler?

I think because you are using exactly the same frequency, then its reasonable that the double dalance mixer has single-ended outputs which are unbalanced, depending on the exact phase delays of the two signals (lo tier and gm input). You can see the effect if you introduce phase delay between the tweo signals... assuming delays in "lo" or "rf" path are equal, the 45 deg error will give you balanced output on + and - outputs. Normally you get balanced + and - outputs when the mixer is driven with signals which are not EXACTLY the same.

What is important is that the doubler output (ie 2*freq) DIFFERENTIAL amplitude is correct, and this will be irrespective of lo or rf path delay

hello has all,
please I work on a subject of fine point technology: MIXER DUAL-GATE (two transistors cascodé) in 60Ghz  and I will like to understand his theory and how its functions, more I want to know in which area I must operate my transistors so that they function (NMOS-OL : switch and NMOS-RF: saturation or linear)???.

thus thank you has all and I a help of you waits.

Hi,

I agreed with yang and ACWWong. In fact if the mixer is a switching mixer with square wave LO and RF, the switching mixer will only work (i.e. as a frequency doubler with 50% duty cycle output) when the LO and RF has a quadrature relationship. It is easier to understand this intuitively by thinking your transistors as switches.

The Gilbert cell will work closer like a switching mixer as you increase your LO and RF input power and therefore will not help you by increasing the LO and RF power. So if you want a clean frequency doubler, it is best to have a phase shifter/quadrature generator before the mixer. Feeding it with in-phase LO and RF will not do.

Regards,
chase