[NLRS] PA3GIE 10GHz high power SSPA

[Dr. Gerald N. Johnson]

Barry is kind of busy preparing to head south for the winter next week.

What he told me was that Jeffrey was playing with a packaged GaN 
amplifier chip from Cree, and made a drastic change in the load tuner 
and suddenly lost half his power out while doubling the supply current. 
He fried the part with too great an SWR when the part was only rated at 
5:1 SWR survival. Before frying the part it was not very linear, for the 
25 watt part, P1 was 5 watts and gain was down 10 dB at 25 watts out. 
Barry said: "I was rather disappointed how he optimized the output with 
an EH WG tuner, connected through a Coax WG transition and a relatively 
long coax jumper." I have noticed that the few Cree parts are pricey 
looking at DigiKey who carries some.

The bare chips from Triquint have a bit better SWR survival, but between 
being small and physically unprotected appear to be significantly 
difficult to use in the hamshack that's not equipped with a controlled 
atmosphere reflow oven and wire bonding apparatus along with a supply of 
tin/gold solder foil and gold bonding wire. I think there may be other 
options for mounting probably somewhat compromising maximum power 
capabilities. Packaged parts probably will include relatively narrow 
band matching that is not a part of the bare chip.

The TGF2023 family of parts from Triquint are made up of a group of 
transistors on the same wafer. The basic device has one gate connection, 
two through holes to the back side for the source, and bar for the drain 
connection. The twice that power part has two of those scrunched 
together with three through holes and two gate pads. The higher power 
parts add more until the 90 watt part has 16 of the small parts in 
parallel sharing adjacent source through holes. And they want two bond 
wires to each gate pad and two for each drain position. I think at 10 
GHz that straps rather than wires may work better moving the matching to 
the PC board rather than using the bond wires as inductors in the 
matching network. I've not done it so I may be more full of it than I 
wish to be. The 90 watt part is about .8mm x 4.6mm x .1 mm. I'm thinking 
Sn/Bi solder for low temperature chip mounting (at a cost of poorer heat 
transfer, but trying to keep the solder film thin because Sn/Bi 
soldering paste is quite reasonable priced and available) and either 
silver epoxy or that Sn/Bi soldering past for connecting copper foil 
"broad" conductors to the gate pads and the drain bar. Flip chip using 
the top side source through holes and the drain bar for heat sinking and 
board connection might have merit too, depending on the distance of the 
heat inside the part from the two surfaces.

MACOM tech just announced some packaged GaN parts, but they are only 
good to 3.5 GHz and about 25 watts, not a market that I perceive as 
neglected though the GaN parts have some nice characteristics like a 28 
volt supply to raise the load impedance to a few ohms at 90 watts 
instead of a fraction of an ohm at 6 volts supply and power added 
efficiency upwards of 55% compared to 10% in the parts we use now at 10 
GHz (that have three stages and three times the package gain of the 
unpackaged chips). And GaN on SiC has a reputation of being tougher than 
nails standing absurd device temperatures without damage. But Jeffrey 
was able to do damage probably by voltage break down. Carborundum has a 
long history of being tough as one of the toughest abrasives in industry.

Its a field that isn't standing still.

There is an interesting article on Triquints web pages about the TGF2023 
25 watt part titled: "Compact Highly Integrated X Band Amplifier using 
Commercialy Available Discrete GaN FETs." It looks like an IEEE article 
but their copy doesn't show that.

73, Jerry, K0CQ

On 10/29/2012 10:58 AM, S. Earl Jarosh wrote:
>
>
> Barry,
>
> Maybe you can touch on the highlights or provide some synopsis at Aurora for
> the rest of us??
>
>
> S. Earl Jarosh, N0HZ
> Cell:  612.868.1313
> Off:   763.545.3275
> Home:  763.546.7897
> Fax:   763.546.7897
> earl at moneycenters.com
>