Differential pair to single ended component - issues

[TheUnnamedNewbie]

So, I am making a high-speed test PCB, aimed at 56 and 112 Gbit/s data. Some of the connections have to be with a 100 ohm differential pair. However, we want to use coaxial connectors to connect our test circuit to the oscilloscope.

Because the performance is critical, I wanted to make a kind of 'two single ended to differential' component that I design and verify in my EM simulator and then input as 'component.

But altium seems to get confused about the (in this case intentional) short circuits between the microstrip lines and the differential lines. If I remove the connection between them, I can route the differential pair as expected, but with the connection I can't. I've tried setting it as net tie but that didn't seem to help either. (note, the seeming disconnects of the lines is because of the renderer of altium, they are in fact connected. notenote the shape seen here isn't the final shape, this is just what I have now when trying to get altium to work. I'll see how I end up importing the geometry later.)

[ajb]

You need to set the component type as Net Tie (and "No BOM" if there's no component placed on the board).  Net ties allow a short circuit to be created within the component while still treating the two nets as separate and subject to normal short circuit and clearance rules in the rest of the board.

[TheUnnamedNewbie]

You need to set the component type as Net Tie (and "No BOM" if there's no component placed on the board).  Net ties allow a short circuit to be created within the component while still treating the two nets as separate and subject to normal short circuit and clearance rules in the rest of the board.

I have tried setting it to Net Tie (No BOM) and that didn't seem to fix it. I will play some more with the settings and report back later.

[ajb]

Looking at your picture, it looks like the copper region abuts but does not overlap the pads, I wonder if that's the problem.  Whenever I've done net tie components (or even used regions for odd-shaped pads) the pad actually overlaps the region, which maybe is required to establish the net tie properly?

[TheUnnamedNewbie]

With the help of T3sl4co1l I have managed to get this working. I think in the end the pads were too small and DRC rules and such were interfering with the differential layout algorithms. extending these pads fixed the problem.