I'm driving an ADL5802 RF mixer with a 5400BL15B050E RF balun. The balun turns a single ended signal into a differential signal on balanced lines for the mixer. In my attempt to match the balun datasheet land pattern, the balanced lines for a short distance are not right next to one another (though they do have the same length). However, I understand that differential signals should generally be placed immediately adjacent. Here's the current PCB layout, showing the mixer (left), blocking caps (middle) and baluns (right). The white squares can be ignored as they are just component courtyards
And here's the land pattern from the balun datasheet (same pin orientation as my pcb layout):
Am I better off ignoring the balanced line exit angles from the balun and keeping the lines next to one another, or respecting the exit angle and having some temporary separation between them? I could, of course, also move the miters closer together to minimize the separation a little.
Although its not shown since I haven't done the copper pour yet (or the via fence), these tx lines are grounded coplanar waveguides with width 0.34mm and gap 0.1524mm (I've used this for the differential spacing after the caps as well). The height to the next layer ground plane is 0.1702mm. The substrate is FR408 Oshpark 4-layer with a dielectric of about 3.64 at 6GHz, which is my max signal frequency. I've simulated the GCPW (grounded coplanar waveguide) with OpenEMS as well as the miters and blocking cap ground cutouts for a single ended line. I haven't actually simulated the effect of the second line on Z0, but I think it should be the same since each gcpw just sees a conductor 0.1524mm away.
I contacted the manufacturer who indicated tightly coupling the lines right out of the pads is fine (shown in the blue circle in the image below). So, that's what I'll use.