Are there any caveats to implementing power amplifiers (GaN type) at home?
Since these packages have leads, is it possible to just deadbug them on top of a heat sink they are properly mounted to (flange mount part)? (say a 2-20GHz part)
I imagined directly soldering (or using super high conductivity silver epoxy) RF connectors (the bulkhead type) to the leads of the package and the body to the heatsink, which is a flangemount ground, so that the total length between the GaN amplifier and the SMA/3.5mm connector is only the length of the lead that the amplifier is supplied with. Then have tombstone decoupling capacitors from the part leads as close to the package as possible, decoupling to the heatsink body. Like standing up a 0402 thats glued/soldered to the heatsink on one end (a finicky operation), then being up against the body of the amplifier and then being soldered to the lead.
Then running power and ground to the chips by cut out strips of copper (like floating planes), which would then be connected to a power supply. (or copper braid).
The evaluation board just says to use 50 ohm connections, but how important is this if you make the distance between the rf connector and the chip as small as humanely possible (i assume ~ 1mm). I would file the flange mount RF connectors so that the center pin (approximately 1mm protrusion) is at the same level on a flat plane as the lead of the chip, which could also probably be trimmed, if the power leads are bend upwards and the ground and power are floating planes (perhaps supported by little standoffs). Kind like a highway overpass.
The epoxy I have in mind is MG Chemicals high conductivity silver epoxy, which has a very low resistance, or perhaps lower melting point solder (heat the heatsink up on a hot plate to do the attachment for the decoupling capacitors).. or just decouple them to a floating plane.
The wavelength of 20GHz is 1.5cm, but if the distance between the rf connector and the chip is <1mm, does the trace impedance really matter? I have not inspected the package in person, but I can't see why 0.5mm distance between the chip and the connector is unreasonable. I would simply assemble it under optics, then make a custom shield to put over it (I have alot of hand tools and i got good at forming copper etc).
If the heatsink clearances are a problem I could use a dremel to dig holes under the nuts of the rf connectors.
It really does seem to be a PCB with a single part and some decoupling.
It seems like a pain in the ass to make this, but i can see it happening and given the cost of all this rf stuff, id like to save anywhere I can and use the tools i got. I am pretty handy with metal working.
Alternatively, I'm not sure if this is any better, can I solder hardline directly to the chip? If the legs of the IC are bent just right, and the NC pins are cut off, it might be possibly to push a piece of hardline directly onto the surface of the chip. Since the hardline is 50 ohm, that would be a good solution? The center pin of the hardline could be recessed slightly, by a small cutting tool, and conductive epoxy could make the bridge (the IC pin would push directly into the hardline)... so it would theoretically be better then the PCB?
I can imagine cutting the inside of the hardline out with a tiny high speed bur/mill.. but is this really relevant? It seems like something that would come into play when the frequency is like, less then an order of magnitude then the impedance discontinuity lenght.
Home
Help
Search
Login
Register
