Comments? email to
There are lots of nice low-noise FETs out there now, and most are very affordable. That's good news for us; my favorite device for 33 through 9cm is the Avago ATF-54143 E-PHEMET. The E-PHEMET is an enhancement-mode pseudomorphic HEMET...a mouthful for sure, but in simpler terms, it's the kind of GaasFet that is biased on with a positive voltage, eliminating the need to have a separate negative bias supply, or to float the source leads above DC ground, which often causes stability problems.
3456 (9cm) info
This device worked right off the bat for me; was 100% stable,
had 18 db gain at 23cm, good input match, and a noise figure well under 1 db.
The schematic and parts list below were derived from the
Avago datasheet; I
changed some of the values a bit to enhance performance at 23 cm, and used 1206
size resistors. Other builders using this same board report noise
figures under .5db. Recent tests (2010) comparing boards made using Ro4003 and
the original FR4 material show no significant differences in performance. If you
have a NF meter, best results (under .4 db reported) can be achieved with
adjustments to the position of the source lead jumpers.
This
close-up shows proper component placement for L band. Note the position of the
shorting jumpers; about 2.5mm from the source leads.
The widest lead on the GaasFet is positioned to the upper right.
For 13cm, changing L1 to 2.7nh and C1 to 3.9pf gets the job done nicely. Resistor R5 is not used on this band.
The placement of the shorting straps must also change to 1 mm from the source leads so that good input match can be maintained (see inset below).
These jumpers control the amount of degenerative feedback,
and their proper placement has a significant effect on gain and input match.
I lack the equipment necessary to make NF measurements on 13cm;
however, the gain and input return loss measurements seem to be in line with the
data sheets.
For 3456, here are the necessary component changes:
- C1 = 1.5pf
- C2, C5, C4 = 3.9pf
- L1 = 33nh
- L2 = 10nh
- R5 = not used
The other change to make is to cover the entire length of the
source traces with the shorting strap (see inset below).
This eliminates most degenerative feedback, and along with the
above component changes, optimizes the LNA for 9cm. As with the 13cm LNA, my
only measurements were of gain and input return loss, but these measurements
were very close to the ones in the Avago application note, which also lists the NF below
1db.