Rather Wideband Mixer Design Concern

[WideBandwidth]

Hi,

I'm trying to build a wideband mixer (first mixing stage of a spectrum analyzer I'm building) that takes an RF input from DC-3 GHz and spits out 3.5 GHz using an LO that sweeps from 3.5 GHz-6.5GHz. I'm trying to see if this is possible using a single-balanced configuration, but wanted some advice on the design.

My current vision is as follows:
An LO balun feeds two series connected diodes. The center-tap of the diodes is a combined RF-IF port that connects to a diplexer to separate out the RF and IF.

The LO balun consists of a microstrip-slotline transition, and the two slotline conductors are used as the 0deg and 180deg ports that connect directly to the diodes. On this front, I've tested back-to-back microstrip-slotline transistions and am getting around 1 dB loss from around 2-8 GHz for an individual transition...not too shabby! I've also tried a single transition hooked up to the diodes with the center tap going straight to a spectrum analyzer, and am getting >30 dB LO isolation with the LO sweeping from 3.5-6.5 GHz. So far so good.

On the RF-IF diplexer front, I have a suspended stripline design simulated and ready to fabricate. It achieves something like 50 dB isolation between the RF and IF frequencies (at least it does in EM simulation) using a 7th order elliptic filter (for the RF) and a 4th order coupled line filter (for the 3.5 GHz IF).

Now to the actual question: If I just hook up the LO balun+diodes to the diplexer, I am suspecting I'm not going to have a good time. The reason being is that if we consider the path that the RF and IF frequencies will see, they are expecting the LO ends of the diodes to be a virtual ground, as we would have if the LO balun was an ideal transformer. This would allow the RF and IF frequencies to see the two diodes in parallel.
But a microstrip-slotline transition is not an ideal transformer, and doesn't even look like a balun below 2 GHz in this case. As a result, the impedance seen by the RF and IF ports, which should ideally just consist of the diodes' dynamic impedance, will be frequency dependent due to the LO balun's own output impedance. This would naturally harm the conversion loss and cause it to flop all over the place with respect to frequency. I should note that in the particular case of a microstrip-stripline balun, the output port will look like indeed a short at very low frequencies, since the slotline portion is terminated in a short (see attachment). However, this only holds near DC, since transmission line effects will eventually turn that short into an open.

So how would one go about solving this balun issue in a single-balanced mixer design? Is it even solvable given these frequency ranges? Also please correct me if I am making a mistake in my reasoning.

Thanks.