San Jose State University Department of Electrical Engineering ELECTRICAL ENGINEERING SENIOR PROJECT Microwave Amplifier Design (part 1) by Steve Garcia Jaime Cordoba Inderpreet Obhi December 15, 2003
San Jose State University
Department of Electrical Engineering
ELECTRICAL ENGINEERING SENIOR PROJECT
Microwave Amplifier Design (part 1)
by
Steve Garcia Jaime Cordoba Inderpreet Obhi
December 15, 2003
Objective The goal of our senior design project was to design and build a prototype single stage Microwave power amplifier operated at 2.4 GHz with a linear region of operation up to our desired output of 1 watt or 30 dBm. This project was chosen because of its apparent complexity and the RF design experience that would be gained by the end of the project. Project Specifications: As a design team, we came to the conclusion that we wanted to design our amplifier to operate in the 2.4 GHz ISM (Industrial, Scientific, and Medical) Band. Typical 2.4 GHz Microwave amplifiers on the market were researched and comparable target specifications were proposed. The overall target specifications of the amplifier design are as follows:
• Operating frequency @ 2.4GHz • Output power of 1 watt (30 dBm) • Obtain a gain of +10 ~12db • 1dB compression >= 30dbm • Cost < $ 100
Application The applications of our proposed device include many products in the field of microwave communications. One of the important applications of a Microwave power amplifier is in the output stage of a transmitter where a signal needs amplification before it is transmitted. A high power amplifier is needed for transmitting a signal through an antenna and a medium. The Microwave power amplifier amplifies the input signal after the signal has been modulated in the transmitter. The High power amplification step is necessary for every application of antenna transmission.
Figure 1: General Transmitter Output Stage
Data Signal
High Power RF Amp
Modulated Signal
Mixer RF output Transmitted
Local Oscillator
Design Methodology A power amplifier is an amplifier that takes a low or intermediate level signal and significantly boosts its power level. At low frequencies this might be a trivial design which would only involve the careful choice of a DC bias circuit designed for maximum power output. But our chosen amplifier design is designed to operate at the microwave frequency of 2.4 GHz. When operating at this frequency, transmission line theory comes into the picture. The high frequency and short wavelength of microwave energy make for difficulties in analysis and design of microwave components and systems. Matching of the input and output of the transistor must be considered and designed around. A typical block diagram of a single-stage RF amplifier is shown below.
Figure 2: General Microwave Amplifier Topology This was the basic topology that we adhered to through our design procedure. The basic design flow for this topology is as follows:
• Choose an Microwave Transistor based on design specifications • Design a DC Biasing circuit for desired operation: Class A, Class B, Class C, or Class AB • Design the Input and Output Matching Circuits based on the desired type of amplifier: Low-Noise
Amp, High-Gain Amp, or High-Power Amp Because our design is that of a high-power device, there is a more specific design flow to follow when designing a high-power microwave amplifier that is illustrated in the book, Microwave Circuit Design Using Linear and Nonlinear Techniques by Vendelin, Pavio, and Rohde. This design method is based strictly on using small-signal S-parameters for the design of M1 (the input matching circuit) and M2 (the output matching circuit). It also addresses the proper form of biasing for maximum output power. The design flow is as follows:
• Obtain transistor static IV curves • Using IV curves, design the bias circuit for the transistor in order to achieve Class A operation • Obtain the transistor S-parameters for the determined bias values • Using IV curves, define the optimum load line for maximum output power and determine the
resistance value, RL, corresponding to this load line • Design M2 for maximum output power using the value of RL
RF AMP
Output
Matchng Network
M2
Transistor
Biasing Network
Input Matchng Network
M1
• Determine S11’ with M2 at output and design input matching circuit, M1, for zero reflection. The IV curve of the transistor is the starting point for the design. An example of one is shown below: Figure 3: FET Transistor IV Curve From the IV curve, the usable regions of ID and VDS can be determined. For Class A operation, the DC operating point should be centered in these usable regions. This determined operating point will define the VGS and VDS in the bias circuit design. In order to obtain maximum power from this device, we must define a load line that fully spans these usable regions. The slope of this load line will determine the large-signal load impedance, RL, for maximum transistor output power. In other words, RL is used as the goal impedance to be presented to the device’s drain terminal by the output circuit in order to achieve maximum RF power output. This value of RL defines the output matching circuit, M2. The Smith chart is then used to define the lossless output circuit M2 by matching RL to 50Ù at the design frequency. With M2 designed, the input matching circuit, M1, can then be designed using S11’. M1 should be designed for zero input reflection. To do this, the Smith chart is used by matching S11’ to 50Ù at the design frequency. This design method should result in a high power amplifier with little input reflection, ÃIN. It should be observed that the output match will be poor because it is intentionally mismatched in order to achieve maximum RF power generation. In other words, the output match is optimized on RL rather than the device's S22.
ID
VDS
useable region
useable region
Operating Point
Load Line
It should be noted that this design method for high power works even though small-signal S-parameters are used as a design basis. Ideally, large signal S-parameters should be used because power amplifiers are inherently large signal components. This is because they operate into the power saturation area of the IV curve which is, in most cases, a non-linear region. Ideally, a set of large-signal S-parameters would include these non-linearities and would characterize a transistor for high power applications but unfortunately device manufacturers do not provide them. This is because the measurement of large-signal S-parameters is difficult and is not properly defined. Fortunately, small-signal S-parameters are suitable for use in large-signal amplifier design when they are operating in Class A. Although this design procedure may seem straight forward, the actual design process turned out to be much more involved. Design Environment: Microwave Office It was obvious from the start that our amplifier would need to be designed in the software environment if we actually wanted to build it. There are several software packages in the industry that are used for the design and simulation of RF circuits. The one that we chose to use was Applied Wave Research’s Microwave Office. The primary reason for this choice was that we could obtain our own trial copy which gave us much more flexibility in the design process. Microwave Office is one of the top three industry standard RF design and simulation packages which also made it very attractive. Learning the use and capabilities of the software through the design process turned out to be very time consuming but the experience gained with the software will no doubt be invaluable in an RF career. We also used Agilent Technology’s Advance Design Software for parallel design and comparison in the early stages of design. This was done primarily to gain a little experience with comparable software. Design Phase I: Obtaining a Non-Linear Device Model The first step in the whole design process was to choose a transistor. We chose the Filtronic LP1500 transistor in a P100 package which is a 1 Watt power PHEMT. The LP1500P100 is a packaged Aluminum Gallium Arsenide / Indium Gallium Arsenide (AlGaAs/InGaAs) pseudomorphic High Electron Mobility Transistor (pHEMT). This transistor was chosen because it met all of the requirements for our target specifications. The most unexpected problems that we encountered when we started our design was that there are no perfect non-linear models for microwave transistors. The non-linear model given for the LP1500 does not generate the actual S-parameters of device test data that they also supply. Filtronics supplies, on their website, a Cutice-Cubic Non-Linear Device Model is given along with a set of actual S-parameters of a tested LP1500P100 device at a specific DC bias. But when we compared this test data with the S-parameters that are generated by the non-linear device model in Microwave Office simulation, the values did not match at our design frequency of 2.4 GHz. We found the solution to this problem by consulting a professional in the design field. We were advised to optimize the non-linear device model for our design frequency by adjusting various parameter values in the non-linear model. The Curtice-Cubic non-linear device model that we obtained from Filtronics has a total of 36 parameters. The optimization procedure of the non-linear model involved changing arbitrary values one at a time, running S-parameter simulation,
and comparing the results of simulation with that of the given device test data. This was an especially time consuming procedure because it is a hit-and-miss method. We first had to determine which of the 36 device model parameters made the most significant changes in the S-parameter simulation. Then we had to adjust them while watching all 4 S-parameters. This was quite frustrating because changing one parameter might bring one of the S-parameters values come with in the desired range but would, at the same time, make another S-parameter deviate further from the desired value. We realized how hard it was to bring every device model S-parameter close to that of the actual test data values, so we concentrated on optimizing the input and output matching S-parameters; S11 and S22 respectively, while allowing more error within the through parameters; S21 and S21. This decision was made because obtaining the right matching in the simulation environment would ensure good matching in the realized amplifier circuit. In other words, designing around a device model with S11 and S22 close to the actual device values should guarantee a properly matched circuit when it is actually built. It should be noticed that optimization would not necessarily be needed if the amplifier circuit were only to designed and tested in the software environment. Below is a graphical representation of the actual device Test Data S-parameters (Blue Trace), the non-optimized model (Green Trace), and our final optimized device model (Pink Trace). Notice the optimization at our design frequency, 2.4 GHz of S11 and S22.
0 1.0
1.0
-1.0
10.0
10.0
-10.05.0
5.0
-5.0
2.0
2.0
-2.0
3.0
3.0
-3.0
4.0
4.0
-4.0
0.2
0.2
-0.2
0.4
0.4
-0.4
0.6
0.6
-0.6
0.8
0.8
-0.8
LP1500 S11Swp Max
18GHz
Swp Min0.5GHz
2.4 GHzr 0.11x -0.18
2.4 GHzr 0.11x -0.37
2.4 GHzr 0.13x -0.23
S[1,1]LP1500 Test Data
S[1,1]Optimized Package Model
S[1,1]Factory Package Model
0.5 5.5 10.5 15.5 18Frequency (GHz)
LP1500 S11
-5
-4
-3
-2
-1
0
2.4 GHz -1.738 dB
2.4 GHz -1.854 dB
2.4 GHz -2.164 dB
DB(|S[1,1]|)LP1500 Test Data
DB(|S[1,1]|)Optimized Package Model
DB(|S[1,1]|)Factory Package Model
1.0
1.0
2.0
0.2
-0.2
0.4
-0.4
0.6
0.6
-0.6
0.8
0.8
-0.8
1.2
1.2
1.4
1.6
1.8
0.1
-0.1
0.3
-0.3
0.5
0.5
-0.5
0.7
0.7
-0.7
0.9
0.9
-0.9
LP1500 S22
Swp Max18GHz
Swp Min0.5GHz
2.4 GHzr 0.83x -0.35
2.4 GHzr 0.73x -0.12
2.4 GHzr 0.79x -0.35
S[2,2]LP1500 Test Data
S[2,2]Optimized Package Model
S[2,2]Factory Package Model
0.5 5.5 10.5 15.5 18Frequency (GHz)
LP1500 S22
-22
-19
-16
-13
-10
-7
-5
2.4 GHz -15.28 dB
2.4 GHz -13.55 dB
2.4 GHz -12.92 dB
DB(|S[2,2]|)LP1500 Test Data
DB(|S[2,2]|)Optimized Package Model
DB(|S[2,2]|)Factory Package Model
1.0
1.0
2.0
0.2
-0.2
0.4
-0.4
0.6
0.6
-0.6
0.8
0.8
-0.8
1.2
1.2
1.4
1.6
1.8
0.1
-0.1
0.3
-0.3
0.5
0.5
-0.5
0.7
0.7
-0.7
0.9
0.9
-0.9
LP1500 S12
Swp Max18GHz
Swp Min0.5GHz
2.4 GHzr 1.05x 0.04
2.4 GHzr 1.03x 0.05
2.4 GHzr 1.06x 0.04
S[1,2]LP1500 Test Data
S[1,2]Optimized Package Model
S[1,2]Factory Package Model
0.5 5.5 10.5 15.5 18Frequency (GHz)
LP1500 S12
-40
-30
-20
-10
2.4 GHz -30.89 dB
2.4 GHz -30.11 dB
2.4 GHz -29.12 dB
DB(|S[1,2]|)LP1500 Test Data
DB(|S[1,2]|)Optimized Package Model
DB(|S[1,2]|)Factory Package Model
0
15
30
45
60
75901
05
120
135
150
165
-180
-165
-150
-135
-120
-105 -
90
-75
-60
-45
-30
-15
LP1500 S21Swp Max18 GHz
Swp Min0.5 GHz
Mag Max20
5Per Div
2.4 GHzRe -0.04Im 5.07
2.4 GHzRe 0.84Im 4.24
2.4 GHzRe 1.36Im 6.79
S[2,1]LP1500 Test Data
S[2,1]Optimized Package Model
S[2,1]Factory Package Model
0.5 5.5 10.5 15.5 18Frequency (GHz)
LP1500 S21
-10
0
10
20
30
2.4 GHz 12.72 dB
2.4 GHz 14.09 dB
2.4 GHz 16.81 dB
DB(|S[2,1]|)LP1500 Test Data
DB(|S[2,1]|)Optimized Package Model
DB(|S[2,1]|)Factory Package Model
Design Phase II: IV Curve With an acceptable device model obtained, the actual design process could begin. The curvetracer in Microwave Office was used to generate the LP1500 IV curves from which an operating point and load line could be determined.
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18Voltage (V)
IV Curve of LP1500
050
100150200250300350400450500550600650700
IVCurve (mA)CurveTrace
From this IV curve, the operating point was designed. The DC bias values turned out to be VDD=8.5 Volts and VGS= -0.5 Volts. The maximum power load line was determined and has a RL value of 22 ohms. RL was then translated to the Smith Chart and matched to 50 ohms using lumped elements.
This determined the output match M2. M1 was then designed using S11’ on the smith chart as described earlier. The final lumped element circuit is shown below.
0 1.0
1.0
-1.0
10.0
10.0
-10.0
5.0
5.0
-5.0
2.0
2.0
-2.0
3.0
3.0
-3.0
4.0
4.0
-4.0
0.2
0.2
-0.2
0.4
0.4
-0.4
0.6
0.6
-0.6
0.8
0.8
-0.8
Smith ChartSwp Max
1
Swp Min-1
RL= 22 ohms
Match to 50 ohms
CAPQ
ALPH=FQ=Q=C=ID=
1 0 GHz0 100 pFCdcblock1
CAPQ
ALPH=FQ=Q=C=ID=
1 0 GHz0 100 pFCdcblock2
CAP
C=ID=
1.49 pFC2
IND
L=I D=
1.64 nHL2
I NDQ
ALPH=FQ=Q=L=
ID=
1 0 GHz0 100 nHLrfchoke2
DCVS
V=ID=
8.5 VVDD1
INDQ
ALPH=FQ=Q=L=
ID=
1 0 GHz0 100 nHLrfchoke1
DCVS
V=ID=
-0.5 VVGS1
CAP
C=ID=
10.34 pFC1
IND
L=ID=
1.17 nHL1
1
2
3
SUBCKT
NET=ID=
"LP1500 Package Model" S1
PORT
Z=P=
50 Ohm2
PORTFN
Tone=Ang=Pwr=Freq=
Z=P=
1 0 Deg16 dBm2.4 GHz50 Ohm1
Because the circuit was to be built on a substrate, a mixed lumped element microstrip design had to be generated based on the above design. This is because every connection between all of the elements are in actuality microstrip pieces and therefore had to be considered and modeled in the design. This was a very long and drawn out process but it finally yielded a working circuit. This is the final circuit in Microwave Office along with the S-parameters and 1 dB compression point determination:
CAP
C=ID=
M1cap pFC1
IND
L=ID=
M1ind nHL1
INDQ
ALPH=FQ=
Q =L=ID=
1 0.00796 GHz7 1000 nHLrfchoke1
MSUB
Name=ErNom=
Tand=Rho=
T=H=
Er=
RO1 3.38
0.0021 1
0.669 mil32 mil
3.38
MLIN
L=
W =ID=
L1 in mi l
W 1in milTL1
1 2
3
MTEE
W3=
W2=W1=
ID=
160 m i l
W1in mil120 m i lTL2
MLIN
L=W =ID=
500 mil160 milTL3
MLIN
L=
W =ID=
L4in mil
W 4in milTL4
12
3
MTEE
W 3=W 2=W 1=ID=
120 mil120 milW 4in milTL5
MLIN
L=W=ID=
500 mil74.5 milTL7
DCVS
V =
ID=
-0.4775 V
VGS1
INDQ
ALPH=FQ=
Q =L=
ID=
1 0.00796 GHz7 1000 nHLrfchoke2
RES
R =ID=
22.5 OhmRD1
MLIN
L=W =ID=
L1 milW 1 milTL8
12
3
MTEE
W3=W2=W1=ID=
160 milW 1 m i l120 milTL9
MLIN
L=
W =ID=
500 mil
160 milTL10
MLIN
L=W =ID=
L4 milW 4 m i lTL11
1 2
3
MTEE
W 3=W 2=
W 1=ID=
120 mil74.5 mil
W 4 milTL12
MLIN
L=W =ID=
500 mil74.5 milTL14
DCVS
V=ID =
15.78 VVDD1
MLIN
L=
W=ID=
500 mil
160 milTL16
MLIN
L=W=ID=
500 mil160 milTL15
VIA
RHO=T=H=D=
ID=
1 1.4 mil32 mil50 milV3
VIA
R H O =
T=H =
D =ID=
1
1.4 mil32 mil
50 milV4
CAPQ
ALPH=FQ=
Q =C=ID=
1 0 GHz0 1000 pFCdcblock3
CAPQ
ALPH=F Q =
Q=C=
ID=
1 0 GHz0 1000 pF
Cdcblock4
CAP
C=ID=
1000 pFDCblock1
CA P
C=ID=
1000 pFDCblock2
IND
L=ID=
M2ind nHL3
MLIN
L=W=
ID =
60 mil120 mil
PAD0806
MSTEP$ID= TL17
MLIN
L=W =
ID=
60 mil120 mil
PAD1
MLIN
L=W =ID=
60 mil120 milPAD2
MSTEP$ID= TL18
MLIN
L=W=ID=
60 mil120 milPAD3
MLIN
L=W=
ID=
60 mi l120 mil
PAD4
MLIN
L=W =ID=
60 mil120 milPAD5
MLIN
L=W=ID=
60 mil120 milPAD6 MSTEP$
ID= TL19
MLIN
L=
W=ID=
60 mi l
120 milPAD7
MLIN
L=W=ID=
100 mil50 milPAD9
MSTEP$ID= T L21
MLIN
L=W=ID=
100 mil50 milPAD10
MSTEP$ID = TL22
CA P
C=ID=
M2cap pFM2cap1
MLEF
L=W=
ID=
L6in milW6in mil
TL6
MLEF
L=W =ID=
60 mil120 milTL23
1
2
3
SUBCKT
NE T =ID=
"LP1500 Package Model" S1
PORT
Z=P =
50 Ohm2
PORT_PS1
Ang=PStep=
PStop=PStart=
Z=P=
0 Deg0.2 dB
30 dBm0 dBm
50 Ohm1
L1in=167.7
L4in=10
L4=32.17L1=480.8
W4in=240
W1in=423.2
W 1=238.8
W4=59.79
M1ind=58
M1cap=39M2cap=0.5
M2ind=1.8
L6in=2687
W6in=165.1
1 2 3 4 4.8Frequency (GHz)
Magnatude of S Parameters
-5
0
5
10
15
-40
-30
-20
-10
0
2.4 GHz -28.12 dB
2.4 GHz -10.3 dB
2.4 GHz -2.763 dB
2.4 GHz 14.56 dB
DB(|S[1,1]|) (R)LAYOUT
DB(|S[2,1]|) (L)LAYOUT
DB(|S[1,2]|) (R)LAYOUT
DB(|S[2,2]|) (R)LAYOUT
0 1.0
1.0
-1.0
10.0
10.0
-10.0
5.0
5.0
-5.0
2.0
2.0
-2.0
3.0
3.0
-3.0
4.0
4.0
-4.0
0.2
0.2
-0.2
0.4
0.4
-0.4
0.6
0.6
-0.6
0.8
0.8
-0.8
Smith Chart S ParmSwp Max
2.8GHz
Swp Min
2GHz
2.4 GHzr 0.32x -0.98
2.4 GHzr 1.10x -0.67
S[1,1]LAYOUT
S[2,2]LAYOUT
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18Voltage (V)
IV Curve and Dynamic Load Line at 1dBC
050
100150200250300350400450500550600650700
8.657 V 308.9 mA
IVCurve (mA)CurveTrace
Dynamic Load Line (mA)Final Layout
-10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18 20Power (dBm)
Gain vs Input Power
6
7
8
9
10
11
12
13
14
15
16
13.21 dBm 13.55 dB
-10 dBm 14.56 dB
DB(PGai nSP[PORT_1,PORT_2,1])POWER_LAYOUT
1 dB compression of gain at input power of 13.21 dBm
-10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18 20Power (dBm)
P1dB Compression
5
10
15
20
25
30
35
13.5 dBm 29.24 dBm
AMtoAM[PORT_2,1] (dBm)power
Which, in physical layout looks like this:
Compression Point at 29.24 dBm
This layout was translated to the appropriate file for the circuit milling machine after which the circuit was built and tested.
Input Outpu
VDD
VG
GND