Electronic – Soldering RG-405 coaxial cable directly to PCB pads meant for U.FL connector

coaximpedance-matchingmicrowaveRFsoldering

Is there an ideal method to solder a bare cable to a PCB for microwave (5-6GHz) signals?

The U.FL connectors that are normally used in this application are terrible, and there aren't many (or any) manufacturers making U.FL ends for this kind of cable anyway.

As far as I can tell, we're getting near-ideal data rates with the current methods. I'm a little concerned about reflections at the transition between PCB and cable, because I can't be sure the impedance is correct across the junction. Others have voiced concern about RF "leaking" at the joint – the center conductor is unshielded where it is soldered to the PCB.

I was able to increase the received power on the analyzer by making a minor modification in my soldering technique. I would interpret that as making an overall improvement – more of the power getting radiated, right?

I'm attempting to be sure our production techniques are giving us the best results I can, but after hours of searching I can't find anything much better than this – Can I solder a RG316 coax directly to a PCB instead of using the intended SMA connectors?.

EDIT:
I've added a photo of the current "target" connection. The resist layer has been etched away at the 'neck' to provide greater structural support than bare PCB pads. The coaxial cable diameter is .080", for size reference.

Described solder joint to U.FL pads on PCB

Best Answer

Crimped connectors achieve very close dimensional tolerance. This means once the manufacturer has 'tuned' the PCB layout, the PCB connector, and the cable connector and its crimp, it should be fairly repeatable.

There are two problems with soldering to the board.

a) It's different geometry to the connectors

b) It's not as repeatable a geometry as for the connectors

You can tune out out the difference in geometry, to some extent. It's likely that the excess of inner will introduce a small series inductance. A rule of thumb that's remarkably good for a quick 'back of envelope' calculation is 1nH per mm of wire. Put 2nH series inductance into your simulator, and see what change in S21, but more importantly S11, it gives you.

This can be tuned out with a sniff of parallel C, to make a low pass filter. Enlarging the PCB pads would be one way of doing this. My quick back of envelope (I'm not going to fire up a simulator) suggests that an LC filter that's matched and uses an L of 2nH will have a corner somewhere in the mid GHz region, which is not good for your prospective frequency range.

Once you've tuned any excess of L or C by measurement of your soldered connection, you then have to make the geometry as repeatable as you can, perhaps stripping the coax and assembling with a jig, and inspecting under a microscope after soldering.