About 50% duty cycle vs. 25% duty cycle passive mixer
I do not see why a non-sinusoidal pumping a mixer can cause any "improvement" and why.
A balanced mixer essentially operates as a multiplying switch driven by the local-oscillator power. Any such mixer generates due this process a spectrum of linear combinations of LO, RF and IF signals. Desired ones are selected by filters, others are suppressed. Using pulses instead of sinusoidal signals generates more harmonics which may be supported by adjusting the duty cycle.
According to this patent, B. Butaye, M. Camus, "New RF sampling mixer with enhanced gain and good linearity", FR 06 10288, applying less than 1/4 duty cycle LO removes the overlapping of the "on" state of both channels, and reduces the averaging window of the RF voltage. But this improvement is limited by rise and fall times of the LO driving. In theory, a gain of -1.9dB can be obtained when using 1/4 duty cycle LO. However, I couldn't duplicate this -1.9dB gain with even ideal LOs.
Well, it sounds interesting. In fact " gain - 1.9 dB" means a loss , and a good mixer has a conversion loss of ~6 dB, so the improvement looks negative.
I have more experience with mm-wave mixers where the duty cycle is not possible to vary. With a DC bias, both diodes in the balanced pair are driven by DC current to "open", so less LO power is needed.
In my opinion, a good RF mixer must use filters to select one side band and one useful harmonic while the other spectrum is rejected. By phasing certain rejected parts of the spectrum, a real reduction of conversion loss in upconverters can be achieved but I do not think the duty cycle is its culprit.
I presume the OP is referring to N-path mixers/filters as used in CMOS for radio frontends. It's quite different from the case of a double balanced diode mixer. In this case there are four switches in two pairs, each switch being driven with a 0.25 duty cycle, and each pair puts out Q and I terms. Don't expect to get any benefits out of driving a double balanced mixer with D=0.25.
My bad...I was referring to a 50% duty cycle LO-driven passive mixer, which would come at about -8dB conversion gain. So -1.9dB would be quite an improvement.
Yes, you are right on it!
According to the frequency multipliers theory regarding their efficiency vs duty-cycle can be seen that there is dependency of harmonic level vs duty-cycle of the signal, phenomenon that most probably affect also the switching mixers.
From the plot below can be seen that 2nd harmonic has highest amplitude at 25% duty-cycle, and a minimum of amplitude at 50% duty-cycle
