Question about LO leakage cancellation for upconverter

[Hawaslsh]

Hello all,

I am attempting to construct a circuit which will both cancel one of the upconverted sidebands as well as suppress any leaked LO. I've had good success in some ideal simulations but I had a few questions I wanted to ask before moving forward simulating some real components.

To build up some context I will quickly show how i built up the circuit. Above is an ideal mixer in Microwave office to which I am feeding an input IF of 50 MHz and an LO of 2.3 GHz (sorry about the mixer ports being labled differently, the ideal mixer won't work if I feed the IF). From the output spectrum on port 3 you can see the two upconverted sidebands: 2.25 and 2.35 GHz, and a significant LO leakage at 2.3 GHz. I am interested in isolating the upper sideband.

To cancel out the lower sideband I used a Hartley style topology. I inserted two ideal quadrature hybrid couples, one at the input, and one for the LO. An ideal 3dB power combiner is used on the output to add the two mixer outputs together. From the output spectrum at port three you can see the lower sideband, 2.25GHz, was canceled quite significantly.

To cancel out the leaked LO i used an ideal 3dB power splitter to split the LO. Half the LO was used to drive the mixers and the other half was fed into an ideal phase shifter and variable attenuator. I used another ideal combiner to combine the phase shifted and amplitude adjusted LO signal back into the mixer signal path. The output spectrum shows the LO, 2.3Ghz, being canceled quite well.
My first question is: when I've looked at other LO canceling circuits, most people use a directional coupler to combine the phase shifted and amplitude adjusted signal back into the mixer signal path rather than a power combiner as I have done above. Is there a specific reason its usually done with a directional coupler?

Secondly, why wasn't I able to cancel the LO leakage when I tried to inject the phase shifted and amplitude adjusted LO backwards into the mixer signal path like the picture above? I tried adjusting the phase shift and attenuation but wasn't able to cancel the LO on the output. In fact I saw no change the LO leaking at port 3 regardless of the phase shift or attenuator settings. Perhaps I simply didn't get the right values, but is the topology above possible? The topology seems somewhat undesirable since I would be splitting my signal of interest, but seemed odd I couldn't get it to work.

I noticed the same issue when I tried to use a directional coupler to inject the phase shifted and amplitude adjusted LO into the signal path. I was able to get LO cancellation when I injected the signal towards the output port (image above), but not when I injected the signal back towards the mixers. (As a note, the S-parameter block in Microwave Office lists the ports non-standard like for their 4-port s-parameter block. This resulted in port 4 coupling to port 2 and port 3 coupling to 1 as the dashed arrows suggest. Not the standard look for a directional coupler, sorry)

Any other criticisms or thoughts on the up-converter topology are welcome.
Thanks in advance.

 

[BMR]

A BPF saw filter after at port 3 would solve the component count? LO cancellation is also generally done inside the radio Rx/Tx Soc. What's the dBc requirement you are looking for? From the graphs it seems like it's around -53 dBc. Is that leakage level enough? The directional coupler in most cases are used to monitor certain frequency bands before they enter the radio front end. Although I'm not very familiar with the coupler topology you have described.

 

[biff44]

if you need good LO suppression, use a subharmonic mixer instead. i.e. drive the LO at 1/2 the desired frequenecy, and let the mixer "double" it. most subharmonic mixers keep the LO frequency circulating inside of the diode ring, and it does not leak out very much

 

[BigBoss]

You do some Phase Related mistakes. If you check the Phase relationships, it will give you the right result.

--- Updated Aug 20, 2022 ---

Mixers and Combiner must be "idealized" because they are not ideal. If you click second parameters, you will see the Isolation between the ports. For instance I've made them -60 dB for between each port, the mixers became almost ideal.
Also, Power combiner is not ideal too. Isolation between ports is dependent to insertion loss.
Briefly, you should check the components' parameters because They are not ideal and should be arranged to get a meaningful results.