Working with microwaves

My desk

Today my work has been repairing electronic boards with RF (radio frequency) signals at up to 13GHz. This combines the finesse of a watchmaker with the crudeness of heat guns and the magic of RF.

The RF signals travel between semiconductor integrated circuits (chips) on the board along co-planar waveguides. The magic is that, at these frequencies, it is no longer true that the potential is equal at all points on a conductor. Far from it. A conductor has significant inductance and capacitance and can act as a receiving or transmitting antenna. You can try shorting a microwave signal to nearby ground and the result will be unpredictable - the signal can even end up being larger as a result!

Surface-mount microwave devices typically have a hidden paddle underneath that has to be soldered to the ground plane. To do this you apply solder paste and then raise the temperature of the ground plane by sitting the board on a hot-plate whilst applying a hot-air soldering gun to the component taking care that adjacent components don't get fried. It takes several minutes for the heat to penetrate to the paddle. If you overdo the solder paste then you get shorts and have to start again.

Touching up the soldering of the replacement part, after the heat gun treatment,  requires a steady hand and magnification.  Many of the parts are have fine pitch or 0402 package (about as big as a grain of sand 1mm x 0.5mm x 0.2mm).

0402 package

Then there is the test equipment. I don't have limit-less funds and microwave test equipment is expensive so I make do. I have access to a Spectran HF-10105v4 handheld spectrum analyser, good for up to 9GHz (on the left in the picture). This is a frustrating piece of kit as the attendant PC software doesn't always work and the device itself overheats and then stops working if left on for more than about 10 minutes.  I own a desktop spectrum analyser / signal generator (Marconi 2945) but this only goes up to 1GHz. To generate higher frequencies I use a home built generator based on the ADF4350 chip that works up to 4400MHz. This, too, is a bit perhaps-a-tronic so, all in all, the testing can be frustrating. With RF you cannot probe the circuit to see what the signal is without disturbing it. But some idea can be gained by using small coax cable with the last cm stripped. An 0402 (d.c. blocking) capacitor is soldered on the end of the inner. On the far end of the 0402 is a whisp of solder with a sharp point and this is the probe. You hold the coax making contact with the braid outer to provide a sort of ground, and the other end plugged into the Spectran analyser.  This arrangement will give a relative signal level repeatable to within about a 5dB range. Until the Spectran overheats...

Test "probe" with 0402 capacitor at the end