Low Noise Amplifiers for 2304, 3456, 5760, and
10368 MHz using the
ATF-36077 PHEMT
by
Al Ward
WB5LUA
INTRODUCTION
The Hewlett-Packard ATF-36077 PHEMT device is
described in a series of low noise amplifiers for 2304, 3456, 5760, and 10368
MHz. Single stage amplifiers are described for 2304, 3456, and 5760 MHz while a
two stage amplifier is described for 10368 MHz. The goal for these amplifiers
was to establish a common printed circuit board size that would ultimately
allow all LNAs to be built into a common aluminum waterproof enclosure.
LNA DESIGN
All LNAs were designed for ER=2.2 dielectric
material. The 10368 MHz LNA was designed for .015 inch thickness material to
minimize radiation losses while the lower frequency LNAs were designed for .031
inch thickness material. I have used both Taconics TLY-5 and Rogers D5880
material with very good success. All LNAs make use of plated through holes to
obtain good high frequency grounding of the PHEMT devices.
All 4 LNAs use microstripline matching except
for the noise match of the 2304 and 3456 MHz LNAs where a wire inductor was
used for lower loss. Quarterwave bias decoupling lines are used to provide gate
and drain bias to each stage. 50 W resistors are used
along with the bias decoupling lines to provide low frequency terminations for
the devices. On the 3 low frequency units a resistor in series with the drain
is used to improve stability.
Each LNA will be described separately in the
following sections
2304 MHz LNA
The 2304 MHz LNA including component
placement is shown in Figure 1. The LNA uses a wire inductor for the input
noise match and a microstripline match for the output with a small resistor in
series with the drain to lower gain and improve stability.
Figure 1. 2304 MHz LNA showing component
placement
Although the ATF-36077 is capable of
"device only" noise figures of nearly 0.2 dB at 2 GHz, the losses of
the input network limit actual LNA noise figures to around 0.4 dB. I described
another 2304 MHz LNA1 that used wire inductors for both the noise
match and the input bias decoupling line. This technique allowed the LNA to
achieve a noise figure of 0.4 dB or slightly less. Since this LNA was designed
to be mounted in a standard enclosure I included some additional input etch
which raised the noise figure slightly. The prototype LNAs I built were mounted
in a box made from brass strips 1 inch tall. The SMA connectors were soldered
directly to the printed circuit board and the brass walls. I was able to
achieve 0.4 to 0.5 dB noise figures on all units built to date. Associated gain
is 15 to 16 dB typically. Bias point is Vds =2 volts and Ids = 15 mA.
The schematic diagram for both the 2304 MHz and 3456 MHz LNAs is shown in Figure 2.
Figure 2. Schematic Diagram for 2304 and 3456 MHz LNAs
The parts list is shown in Figure 3.
|
C1,C2,C4 |
8.2 to 10 pF chip capacitor |
|
C3,C5 |
1000pF // .01uF chip capacitor |
|
L1 |
2304 MHz , length = .5 inch 3456 MHz, length = .3 inch .007"diameter with .075" soldered to input etch and .025" soldered to gate lead. Adjust for best NF |
|
Q1 |
Hewlett-Packard ATF-36077 PHEMT |
|
R1,R2 |
50 W chip resistor |
|
R3 |
10 to 27 W chip resistor (effects gain and stability) |
|
R4 |
For operation from a power supply voltage of 5 volts, R4 = 200W - R3 - R2 |
The 3456 MHz LNA including component
placement is shown in Figure 4. 
The LNA design is very similar to the 2304 MHz
LNA. It also uses a wire inductor for the input noise match and a
microstripline match for the output network. A small resistor in series with
the drain is used to lower gain and improve stability.
5760 MHz LNA
The 5760 MHz LNA including component
placement is shown in Figure 5. The LNA uses microstripline matching networks
for both the input noise match and output gain match. Some resistive loading in
the drain circuit is required for unconditional stability, according to the
computer analysis. However, several LNAs have been built without this resistor
with no problems experienced.
The prototype 5760 MHz LNAs were installed in
brass housings similar to the other LNAs. Noise figures of 0.7 to 0.8 dB were
obtained along with 12 to 13 dB gain. Somewhat lower noise figure could
possibly be obtained by cutting off the input 50 W
microstripline and placing the input connector adjacent to the input blocking
capacitor. Bias point is Vds =2 volts and Ids = 15 mA.
The schematic diagram and parts list are
shown in Figures 6 and 7.
Figure 6 Schematic Diagram of 5760 MHz LNA
|
C1,C2, |
1 pF chip capacitor (1-2 pF OK) |
|
C4 |
10 pF chip capacitor |
|
C3,C5 |
1000pF // .01uF chip capacitor |
|
Q1 |
Hewlett-Packard ATF-36077 PHEMT |
|
R1,R2 |
50 W chip resistor |
|
R3 |
10 to 27 W chip resistor (effects gain and stability) |
|
R4 |
For operation from a power supply voltage of 5 volts, R4 = 200W - R3 - R2 |
Figure 7. Parts list for 5760 MHz LNA
10368 MHz LNA
The 10368 MHz LNA was designed as a 2 stage
LNA in an attempt to lower "low frequency" gain which became a
problem when mating an earlier design single stage unit with a waveguide input
configuration. With a single stage LNA, gain peaking in the 7 GHz frequency
range coupled with the rolloff of the high pass nature of WR-90 produced some instabilities.
The 2 stage 10368 MHz LNA is shown in Figure
8. The LNA uses microstripline matching throughout the amplifier. The input
network provides the noise match. Both the output matching network and the
interstage
matching network were design for reasonable
gain at 10368 MHz while providing some low frequency rolloff.
Figure 8. 10368 MHz LNA showing component
placement.
The prototype 10368 MHz LNAs were installed
in brass housings. I used .75 inch wide brass for the 4 walls. With the higher
gain of the 2 stage amplifier, the effect of the housing became more noticeable
especially when attaching a cover to the box. I made use of a .5 inch wide
piece of brass as a divider down the middle of the amplifier. This is shown in
Figure 8. The .5 inch wide piece of brass is placed .85 inch in from the side
of the box. This makes it slightly offset from being run directly down the
middle of the box. The divider is soldered to the walls of the box in such a
way that it will make contact with the cover and be about .2 inch above the
printed circuit board etch. This divider tends to break up the waveguide effect
of the enclosure which tends to help propagation of unwanted waves through the
box., causing undesired feedback. I also made use of some absorber material as shown
in Figure 8.
Some tuning of the initial design was
necessary. I used .04 by .1 inch stubs cut from some transistor leads. Place 1
stub as shown on the input line. See Figure 8. This may necessitate some tuning
depending on the type and quality of the SMA connectors you have in your junk
box. Use 2 more stubs to widen the output etch as shown on Q1.
Noise figures of 0.7 to 1.0 dB were obtained
along with 23 dB gain. Bias point is Vds =1.5 volts and Ids = 15 mA per device.
The schematic diagram and parts list are
shown in Figures 9 and 10..
Figure 9. Schematic Diagram of 2 stage 10368
MHz LNA.
|
C1 |
0.6 pF chip capacitor (0.5-1 pF OK) |
|
C2,C3 |
1 pF chip capacitor (1-2pF OK) |
|
C4,C5,C6,C7 |
1000pF // .01uF chip capacitor |
|
Q1,Q2 |
Hewlett-Packard ATF-36077 PHEMT |
|
R1,R2,R3,R4 |
50 W chip resistor |
|
RD1 RD2 |
For operation from a power supply voltage of 5 volts, RD1=RD2 = 180W |
Figure 10. Parts list for 10368 MHz LNA
Biasing
I use passive biasing in most of my
amplifiers with good success. I use a regulated 5 volt source which also feeds
the dc-dc converter which generates the negative voltage .. I use a
potentiometer off the negative source to set the proper gate voltage required
to sustain 10 to 15 mA drain current. The resistance between the 5 volt power
supply and the drain of the device is calculated based on a 3 volt drop to the
drain. The actual dc bias point of the device is not that critical. Vds can be
anything from 1.5 to 2 volts and drain current can be between 10 and 15 mA. DC
to DC converters and various passive and active bias schemes are covered in
other articles.23.
Downeast Microwave offers both single and
dual stage power supplies for powering up FETs. (Part numbers PPS-1 and PPS-2)
A potentiometer is used to set the gate voltage. The important thing to
remember is to set the gate voltage at about -0.2 v before applying the drain
voltage or before installing the bias resistors on the board. The worst thing
you can do to a fet is to apply a high enough negative voltage to the gate that
it "pinches off the drain" and causes no current to flow. With
several volts negative on the gate and 5 volts positive on the drain, the
device is history, even before it even thought about operating!. It is best to
pre-set the gate voltage to where there is some drain current being pulled
through the drain bias resistors. The gate voltage can then be set for optimum
drain current which should coincide with maximum gain and nearly minimum noise
figure. A good test for an LNA is to determine if optimum gain and noise
performance occur coincidentally within the range of the specified bias point.
If the LNA is oscillating, then as one increases drain current and drain
voltage, the amplifier performance, i.e. gain, will peak at a much lower bias
point. This signifies that the device is oscillating and going into a self-bias
mode because of oscillations.
Enclosures
My preferred choice for an enclosure is
normally a box made from brass strips.. This allows gold plated SMA connectors
and the printed circuit board to be soldered directly to the brass. This
provides a good RF tight box. Unfortunately when the box surrounds the circuit
board, it begins to look like a piece of waveguide when viewed from either the
input or output connector. The best solution would be to design the circuit
board to be very narrow as viewed from the end. This will tend to look like a
piece of higher frequency waveguide and will tend to rolloff lower frequencies
and consequently attenuate signals that are propagating through the box and not
the circuit.
Solutions to the ‘waveguide effect"
problem include the use of dividers to break up this phenomena. I described
this technique in the section on the 10368 MHz amplifier. I have also used this
same technique for the lower frequency amplifiers on occasion. If you come
across an amplifier that does not seem to like having a cover placed on it,
then try this divider on top of or slightly offset from the centerline of the
circuit. The top of the divider should hit the lid and the bottom should be
about .2 inches above the microstrip. The divider is parallel to the side
walls. Be careful not to drop it on the circuit when finding the sweet spot.
Another solution is the use of absorber along
at least one of the side walls. This tends to minimize unwanted reflections off
the side walls which are the primary source of trouble. It was brought to my
attention by Bill Janssen, K7NOM, back in 1995 that suitable microwave absorber
can be found locally. In fact, Bill says there is a large supply of this stuff
alongside most highways in the form of old rubber from truck tires. Sure
enough, it does have some microwave absorption properties!. So if you get
desperate, check the highways, but be careful! Bill attributes the absorbing
properties to the carbon added to improve the ultra-violet resistance.
An alternative to the brass box approach is
the aluminum housing approach that I mentioned earlier. I designed the
amplifiers to fit inside an aluminum housing that is available from Downeast
Microwave, part number "Rose-S". Steve makes a special pallet out of
.25 inch thickness aluminum that spans the distance between the input and output
connectors. It makes a tight fit in the enclosure and the SMA connectors thread
into the edge of the inner pallet. The circuit board is then installed onto the
aluminum pallet. I use small 4-40 hardware to hold the board down to the pallet
I have also found that the liberal use of
conductive epoxy insures that the ground plane of the printed circuit board
makes good RF connection along the entire surface of the aluminum pallet. I
have used "Circuit Works Conductive Epoxy" part number 2400 with good
success. It is a 2 part mix and once the tubes have been opened once, it
appears that shelf life is not very long, so it is best to do several
amplifiers at one time. Steve at Downeast Microwave uses shim stock to help
enhance the RF connection between the pallet, circuit board, and the inside
surface of the enclosure which then mates with the RF connectors.
Based on several amplifiers built to date it
appears that the noise figure of the units built in the aluminum housings are
within a 0.1 dB of the brass box units. I have also noticed that none of the
aluminum housing units seem to have any stability issues when installing the
covers. This could be due to the fact that the sides of the aluminum pallet and
therefore the sides of the printed circuit board are suspended in the housing.
As I mentioned before, it is the side walls of the amplifier that give rise to
the "waveguide effect" problem. With the aluminum housing units,
there is a gap or discontinuity between the side edge of the board and the
inside edge of the housing.
Closing
Hopefully, these LNAs will provide a boost to
both your terrestrial and moonbounce systems as they have to mine.
A.J.Ward
09-09-97
References
1. A.J.Ward, "Low Noise Amplifier for
2304 MHz using the HP ATF-36077 PHEMT Device", Proceedings of Microwave
Update ‘94, pg.8-10
2. A.J. Ward, "Simple Low Noise
Preamplifiers", QST, May, 1989, pg 31-36
3. Using the ATF-10236 in Low Noise Amplifier
Applications in the UHF through 1.7 GHz Frequency Range, Hewlett Packard
Application Note 1076, publication number 5963-3780E(3/95)