coxial pitails for up to 6GHz

Hi all, How you guys make the coxial connection for debug when frequency higher than 5.8GHz. When I use the coxial pitials like the one mentioned in a maxim AN, It really sucks. When I soder two outer conductor of two coxial pitail on the pcb groud, and the two center pin on a short trace, I got about 10dB retuan loss. I think this error introduce by the test setup is not acceptable, the measure result is trustless, especially when I want to use this to tune the matching of the circuit.
Did you guys have any good idea on it? Thanks all in advance!
http://pdfserv.maximintegrated.com/en/an/AN3962.pdf

if you are not happy with 10 db RL, give up.

If you really need better, you should have either planned for surface mount pads to allow G-C-G coplanar probing, or had surface mount mechanical RF switches installed as test ports.

I think, the question makes no sense. You are talking about a PCB to coaxial transmission line transition. This means, the "pigtail" has to be soldered exactly to the end segment of a 50 ohm microstrip or similar transmission line. It can't be connected e.g. to the middle of a transmission line without causing mismatch.

So the first question is, what's the nature of the debug node? If it's a regular 50 ohm signal that's otherwise terminated by a 50 ohm resistor, you can e.g. place a suitable connector like U-FL or even a switched MS-162 (good for 10 GHz).

yeah, it does a 2mm length 50 ohm microstrip, I soder two pitail to each end segment of this 2mm microstrip line to check the reliability of this test setup. Then I check it with a full two port s parameter measurement, and it end up with 10dB return loss and 1.5dB insertion loss

To be honest, I don't see a any example of a good microcoax to microstrip transition in the linked Maxim application note. The paper doesn't clearly show to what the inner and outer conducter exactly connects, apparently not to a micro strip. I wonder how your test setup exactly looks like?

Of course, I can imagine a microcoax to microstrip transition, but it requires a bottom side ground with via fence or a signal through connect. There will be still a nonzero S11. An in-circuit SOL or SOLT calibration gives probably the best performance.