Power control

(2.2)

A 100% η implies that the entire supply power is delivered to the load. However, this is practically impossible to achieve. The best way to improve the efficiency is the use of circuit techniques such that both voltage and current waveforms do not exist simultaneously. Switching amplifiers use this approach to achieve efficiencies up to 80%. However, as a trade-off, linearity needs to be compromised for better efficiency.

When comparing PAs with different input power levels, PAE (Power Added Efficiency) is a commonly used metric.

(2.3)

2.2.5 Power control

One of the many power saving schemes, especially in cellular systems, is the use of power control circuitry. When the mobile system is near a base-station, a decision logic at the output of the PA senses that high output power levels are not required and the control circuitry controls the amount of bias/supply to reduce the power levels. The reverse operation is performed when the base-station is at a distance away

η 𝑜𝑤𝑒𝑟 𝑑𝑒𝑙𝑖𝑣𝑒𝑟𝑒𝑑 𝑡𝑜 𝑡ℎ𝑒 𝑙𝑜𝑎𝑑 𝑜𝑤𝑒𝑟 𝑑𝑟𝑎𝑤𝑛 𝑓𝑟𝑜𝑚 𝑡ℎ𝑒 𝑠𝑜𝑢𝑟𝑐𝑒

from the mobile system. The battery life is thus improved, but at the expense of extra circuitry.

Chapter 3

Broad Band Design Techniques

To design a broad band amplifier, many specs other than bandwidth, such like output power, linearity, efficiency, etc., will be compromised in order to expand the operation bandwidth. In this chapter, several broad band design techniques are presented. Each of them has their unique pros and cons, depending on their specific application.

3.1 Previous work on broad band power amplifiers

In broad band power amplifier design, a key element is maintaining a flat gain over the band while also providing a good input VSWR. Commonly this is addressed using the balanced amplifier approach, as illustrated in Figure 3.1, whereby the input is reactively matched for gain sloping (equalization) and quadrature couplers provide a good match when two similar amplifiers are placed between them.

INPUT

OUTPUT

Figure 3.1 A simplified diagram of the balanced amplifier

Another approach often used is the distributed amplifier, as illustrated in Figure 3.2. The distributed amplifier has the advantage of large bandwidth, excellent flatness, and low input VSWR. But it has low gain, low efficiency, and requires a relatively large size.

I N P U T

O U T P U T

Figure 3.2 A simplified diagram of the distributed amplifier

Also used for broad band amplifiers is the feedback amplifier, as illustrated in Figure 3.3. This approach commonly employs negative feedback in a combination of series current and shunt voltage type to provide input match and gain equalization. This approach yields a relatively small size but at microwave frequencies the gain is low and the efficiency is compromised when resistive feedback is used.

OUTPUT

INPUT

Figure 3.3 A simplified diagram of the feedback amplifier

3.2 Compressing Trajectory Dispersion Method [8]

Shunt C and series L disperse a trajectory in smith chart with increasing frequency. In other words when using these matching elements in a low pass network, the higher frequencies will rotate and transform more than the lower frequencies, which spreads the trajectory relative to frequency in a clockwise direction. In Figure 3.4, a two-element low pass network is charted on a Z0 = 25 normalized Smith Chart. The normalized impedance of 25 ohms is calculated from the geometric mean of the system load and source impedance, 50 to 12.5 ohms respectively. The constant Q curve of 1.75 is derived from the resistive ratio of 50/12.5 from the equation

1 Q

²

R

ratio (3.1)

Figure 3.4 Z0=25, Q=1.75; node1-node2 shunt C=6.1pF;

node1-node2 series L=3.8nH; red line marks the trajectory dispersing ranging from 800MHz to 1000MHz

On the other hand, high-pass matching networks consisting of shunt inductors and series capacitors will transform the lower frequencies more than the higher frequencies. In Figure 3.5, a two-element high pass L-network transformation from 50 to 12.5 ohms is demonstrated on a 25-ohm normalized Smith Chart. It is easy to see that the trajectory , ranging from 800MHz to 1000MHz, of lower frequencies dispersing more than the higher ones.

Figure 3.5 Z0=25, Q=1.75; node1-node2 shunt L=5.1nH;

node1-node2 series C=8.2pF; red line marks the trajectory dispersing ranging from 800MHz to 1000MHz

Unlike a low-pass L-network, the higher frequencies are transformed less than the lower frequencies. If the low-pass trajectory of Figure 3.4 were overlaid onto Figure 3.5, the two trajectories would form the letter X. Exploiting this relationship by combining these dispersion

effects can leverage a broad band transformation. A broad band band-pass network is illustrated in Figure 3.6, a 50 (node1) to 3 (node5) ohm transformation. With the Smith Chart normalized to the geometric mean, it is easy to see that low pass nodes 1-2-3 are symmetrical in Q to the high pass nodes 3-4-5. Combining these two networks’ halves folds and compresses the trajectory into a condensed 3-ohm driving point load.

Figure 3.6 50 to 3-ohm transformation; Z₀ = 12.5, Q = 1.75; node1-node2 Shunt C

= 6.10pF; node2-node3 Series L = 3.85 nH; node3-node4 Shunt L = 1.32 nH;

node4-node5 Series C = 32.3 pF.

Compare to the simple two low pass L-network, as depicted in Figure 3.4 and Figure 3.5, the combination of Low-Pass and High-Pass L-Network obviously compresses the trajectory around node5 and thus broadening the bandwidth, shown in Figure 3.6.

3.3 Multiple-Q Method (Chebyshev broad-banding technique)

As a rule in broad band transformations, maintaining a lower Q (quality factor) curve for a given transformation by increasing the number of n-sections will yield a higher bandwidth. However, there is the limitation that using more than a four-section matching network will not yield greater bandwidth. To further extend the bandwidth, one of the effective ways is the multiple-Q method, also called Chebyshev broad-banding technique.

With the complexity of the multiple Q curve network, deriving a design from a Smith Chart alone would not be an intuitive process.

Tables 3.1 and Tables 3.2 were derived by optimization with an ADS simulator utilizing a gradient optimizer.

Table 3.1 Q curves, 2-section network, per transformation ratio. The Q curves are numbered from the outer most Q1 towards the inner Q3.

Table 3.2 Q curves, 3-section network, per resistive transformation ratio. The Q curves are numbered from the outer most Q1 towards the inner Q3.

Q1 Q2 Q3 Trans. Ratio 0.92 0.60 0.65 1.67 1.37 0.90 0.61 5.00 1.53 1.00 0.60 7.14 1.68 1.08 0.58 10.00 1.89 1.21 0.53 16.67 2.06 1.32 0.50 25.00 2.20 1.41 0.49 33.33 2.41 1.52 0.45 50.00 2.57 1.61 0.44 66.67 2.80 1.74 0.41 100.00

As illustrated in Figure 3.7, the transformation is mostly symmetrical with two Q curves, an outer curve (Q1 green) and an inner curve (Q3 magenta). However, node 5 falls at a higher impedance than the 3-ohm target in order to center the fish shaped trajectory at Z = 3 + j0 and so therefore a third Q curve (Q2, cyan) is defined at node 4.

Figure 3.7 Z0 = 12.3, SWR = 1.12 @ ZS = 3; N1-2 Shunt C = 6.9 pF; N2-3 Series L = 4.4 nH; N3-4 Shunt C = 30.4 pF; N4-5 Series L = 0.99 nH.

In Figure 3.8, there is a three-section transformation, the trajectory fits into a 3-ohm 1.01 SWR circle. Three Q curves are adequate for defining the three section network since the trajectory is small and circular in shape, unlike in Figure 3.7, here no impedance offset is needed at node 7.

Figure 3.8 Z0 = 12.2, SWR = 1.01 @ ZS = 3; N1-2 Shunt C = 4.30 pF, N2-3 Series L = 6.22 nH; N3-4 Shunt C = 16.57 pF, N4-5 Series L = 2.42 nH; N5-6 Shunt C = 42.05 pF, N6-7 Series L =0.63 nH.

As mentioned above, Table 3.2 was derived from optimization.

Here Q curves are provided for resistive transformation ratios of 1.67:1 (50 ohms to 30 ohms) to 100:1 (50 ohms to 0.5 ohms).

3.4 Dummy bias technique

This broadband technique is suitable for implementing in the gain block or the PA driver and has shown the excellent gain flatness.

Figure 3.9 Circuit diagram of dummy bias

This technique use smaller RF choke (L) and much larger bias resistor (R), hence causing more power consumption.

3.5 Bridged-T network

To deduce the Bridged-T network, an evolution view of circuit analysis is present. Starting from a simple L type LC low pass filter example shown in Figure 3.10, the 3dB bandwidth is about 1.35GHz.

Figure 3.10 A simple L type LC low pass filter

To extend the 3dB bandwidth, one more inductor is added and formed a T type low pass filter, the 3dB bandwidth extended to about 1.46GHz.

Figure 3.11 A T type LC low pass filter

Based on the T type low pass filter, a capacitor is added between two terminators, after some adjusting, this capacitor significantly increased the 3dB bandwidth. This might be due to the zero included because of the new bridged capacitor.

Figure 3.12 A T type LC low pass filter added a bridged capacitor

The following will describe the use of the bridged-T input matching network that absorbs the FET input capacitance into a simple filter network to provide simultaneous gain flatness, good input return loss [9]. The bridged-T input matching network is a second-order

“all-pass” network shown in Figure 3.13. In Figure 3.14, it shows a simplified linear FET model.

Figure 3.13 The prototype of the bridged-T matching networks

Neglecting Ri, Cds, and taking into account Miller effect, we obtain the further simplified FET model of Figure 3.15, where Cin, is the equivalent input capacitance and RL is the load impedance presented to the FET stage.

Considering the circuit shown from Figure 3.13 to Figure 3.15, this bridged-T is a second-order all-pass network if L = Rs

2Cgs/2, C1=Cgs/4, R=Rs, [9]. When Ri >0 the network no longer is of the standard all-pass form but L, C1 and R can still be chosen such that the generator sees a pure resistive at all frequencies. The design equations for this condition are [10]

R = R

_s

(3.2) L = R/ω

c

(3.3) C1 = 1/2ω

c

R (3.4)

C2 = 2/ω

_c

R (3.5)

Figure 3.14 Simplified FET linear model

Figure 3.15 Further simplified FET model with equivalent input capacitance

Combining Figure 3.15 with Figure 3.13 and using Cin, as C2 and C as C1, we obtain the all-pass matching network of Figure 3.16.

Arranging L and C and let the generator seeing a pure resistive R, this network provides a broad band resistive match as well as gain equalization up to the frequency of resonance, ωc. Although the all-pass network itself has a flat amplitude response with no real cutoff frequency, the voltage across Cin (or C2) does decrease after ωc.

Figure 3.16 Bridged-T network with simplified FET model

The voltage (Vin) directly corresponds to the linear gain of the FET stage, since this is the control voltage for the voltage controlled current source in the FET model. For a given Vin, ωc

can be increased by

reducing R.

Comparing with prior broad band PA design methods, the Bridged-T input matching network is the most suitable method for our specs mainly due to the low Q characteristic.

Table 3.3 Comparison of Broad Band PA Design Method Broad band PA Design Method Comparison

\ Balanced PA Distributed PA Bridged-T input matching Pros flat gain

good VSWR

large bandwidth flat gain

compact size, low pass (Q) flat gain

Cons coupler bandwidth issue low gain, low PAE,

large size moderate noise figure

Chapter 4

Practical Broad Band PA Design

The major task of this thesis is to design and implement an extremely broad band RF high power amplifier with a GaN HEMT unmatched transistor. A NITRONEX product “NPTB00004” [11] is chosen because of cost effectiveness and non-linear model availability.

For a research-driven purpose, this device is suitable for the school lab to study or verify the circuit architecture since its package is SOIC-8, which is easy to solder on the PCB. A completely design and verification procedures are present in the following sections.

4.1 GaN HEMT large-signal modeling introduction

Recently, most effort in PA design has been focused on GaAs pseudomorphic HEMTs (pHEMTs), Si LDMOSFETs, and GaN HEMTs.

Models have been developed and adapted to these devices and share many common features because they are all field-effect structures. The focus of this section is to introduce the development of GaN HEMT models.

Here present one possible solution to the modeling applied to the GaN HEMT while acknowledging that there are many other viable solutions. There are two general approaches to HEMT (or other active device) modeling. One is table-based, the best known of which has been developed by Root [12]. The table data can either be measured or

simulated using 2-D physical simulators. A more recent version of this approach is the new x-parameter model formulation, which is based on significant small- and large-signal measurements [13]. This approach can be very accurate, but requires intensive measurement resources. To improve accuracy, the entire simulation space must be mapped using both large- and small-signal measurements including load–pull and linearity.

It is certainly desirable to have the largest possible measurement database from which to extract and verify any model, but these measurements can be time consuming and expensive. A properly formulated model based on physical equations allows a reduction in required measurements without a significant loss in accuracy. The second approach involves the description of the active device by closed-form physical equations, the parameters of which can be extracted from measured data. This is the approach chosen to support the NITRONEX device model used in this thesis. The model described here uses various formulations, combined in such a way as to allow parameter extraction using a minimal set of measurements. A well-known application is found in Angelov (or Chalmers) model [14].

4.2 Bridged-T broad band design with large-signal model

In this thesis, NITRONEX GaN HEMT transistor NPTB00004 is used to design the Bridged-T network extremely broad band power amplifier. The vendor provides the large-signal model, a kind of Angelov model, executed in non-linear simulation, generating results such like compressed output power, harmonics, efficiency and linearity predicting.

43

4.2.1 Bias consideration

The goal of this thesis is to optimize to output power, therefore, the bias will be set to achieve max output power, rather than linearity.

Based on the datasheet [11], the drain-source voltage Vds should be set at 28V, and the quiescent gate-source voltage, Vgsq, optimized for CW power and efficiency, should be regulated so as to get the quiescent drain current around 50mA.

Figure 4.1 ADS Schematic of bias sweeping

Figure 4.2 Result of bias sweeping

By the simulation data shown in Figure 4.2, the quiescent gate-source voltage, Vgsq, getting around 50mA quiescent drain current at Vds 28V is about -1.6V. Under such bias condition, the PA will operate close to class-AB mode.

4.2.2 PA architecture

Intending to design an extremely broad band PA of operation frequency ranging from several tenth of MHz up to GHz, the most suitable input matching network architecture is the Bridged-T network, which is mentioned in 3.5 of Chapter 3. To enhance the gain flatness, the resistive feedback from the drain to the gate is also implemented in this design. Considering to the easy implementation and adjustability, a structure of LC sections (several pieces of microstrip section following grounded capacitor) is carried out as the output matching network.

Vg VDD

INPUT

OUTPUT

OMN

IMN

LC Sections Bridged-T Network

Figure 4.3 Diagram of the Bridged-T network extremely broad band PA

4.2.3 PCB layout

A four layer PCB with 50 Ω grounded-CPW line was fabricated with FR-4, and soldered manually. The ground through-holes (via) are specifically arranged for minimize the grounding parasitic effect and enhance the heat transportation. The two screw holes fit the precise location of the heat sink used in this case. To deserve to be mentioned, the grounded-CPW line is draw with mini-meter rule nearby, which is convenient to accurately locate the parallel grounded capacitors.

Figure 4.4 PCB layout of the Bridged-T network extremely broad band PA

Figure 4.5 PCB assembly of the Bridged-T network extremely broad band PA

4.2.4 Non-linear simulation with large-signal model

After settling down the Bridged-T network, the non-linear simulation can be executed to get the PA load impedance with good broad band performance. A topology of three LC sections was adopted to become the PA output matching network due to easy design and simple implement. Tedious iteration between input and output matching adjustment made the simulation result contenting.

Figure 4.6 ADS Schematic of the Bridged-T network extremely broad band PA

By means of ADS non-linear simulation template, some non-linear characteristics of the power amplifier, suck like compression output power or harmonics, can be extracted from NPTB00004 Angelov model. By the way, the impedance of grounded-CPW line in the PCB layout (Figure 4.4) calculated by TXLINE calculator is 49.885 Ω, the microstrip with the same width calculated by TXLINE calculator is 49.0697 Ω. Therefore, the microstrip model is used in the simulation (Figure 4.6) for simplicity since the difference between them is trivial.

From Figure 4.7 to Figure 4.9, the design details of the bridged-T input matching network, resistive feedback and the output matching network are revealed.

Figure 4.7 The detail of the Bridged-T network in the ADS Schematic

Figure 4.8 The detail of the resistive feedback in the ADS Schematic

Figure 4.9 The detail of the output matching network in the ADS Schematic

Figure 4.10 shows the gain flatness simulation data of this extremely broad band power amplifier. It drops after 1.3GHz and maintains excellent flatness before the rolling off. The transducer power gain was deduced from 3dB compression point and still be around 12dB in the pass band.

Table 4.1 ADS non-linear simulation data @P3dB

Transducer

Figure 4.10 The transducer power gain @ P3dB

Figure 4.11 Fundamental output power, 2^nd and 3^rd harmonic vs. frequency@ P3dB

200.M 400.M 600.M 800.M 1.00G 1.20G 1.40G

0.000 1.60G

Fundamental @ P3dB and Harmonic, dBm

4.3 Verification and measurement

The most essential task of all is to measure, verify as well as tune

the soldered “NPTB00004 Bridged-T Networkextremely broad band PA”

PCB in this work. Both small and large signal measurement data are collected and put in order.

4.3.1 Bias measurement

The actual bias setting is a little different from the simulation.

VDD is still the same set at 28V. However, in order to get the quiescent drain current around 50mA, the gate voltage (Vg) must adjust to -1.4V, instead of the bias simulation result -1.6V.

Figure 4.12 The picture of NPTB00004 Bridged-T Network extremely broad band PA PCB

4.3.2 Small signal measurement

As depicted in Figure 4.13, the small signal measurement data are

quite similar to the simulation data. Input return loss and output return loss match fairly well, only the gain of the measurement data is somewhat

larger than the simulation data in the pass band. It is observed that the gain roll off slightly after 1.2GHz, significantly rolling off after 1.4GHz.

Figure 4.13 S-parameter simulation and measurement data comparison

4.3.3 Environment setup for large signal measurement

Figure 4.14 shows the setup of NPTB00004 Bridged-T Network extremely broad band power amplifier large signal measurement. A Broad band PA driver made of Mitsubishi GaAs FET MGF0951P [15] is used to drive this extremely broad band power amplifier. A 50W 30dB attenuator is to protect the spectrum and avoiding extra non-linearity produced by the spectrum itself.

Figure 4.14 Diagram of large signal measurement setup

Figure 4.15 Picture of large signal measurement setup

4.3.4 Large signal measurement

The large signal measurement data describe in the following pages, which are measured under the environment setup depicted in 4.3.3 of this chapter. The measurement items include output power, 2^nd harmonic, 3^rd harmonic, gain and PAE. The measurement frequency ranges from 150MHz to 1.2GHz.

Figure 4.16 Large signal measurement data@ 150MHz

Figure 4.17 Large signal measurement data@ 300MHz

Figure 4.18 Large signal measurement data@ 600MHz

Figure 4.19 Large signal measurement data@ 900MHz

Figure 4.20 Large signal measurement data@ 1200MHz

To summarize the miscellaneous measurement data to a more concise presentation, a chart in Figure 4.21 extracts the information showing up from Figure 4.16 to Figure 4.20.

Figure 4.21 Summary of all the large signal measurement @ Pin 27dBm

Table 4.2 Raw data of Figure 4.21

4.4 Result analysis and discussion

From Figure 4.21 and Table 4.2, a performance overview of the NPTB00004 Bridged-T Network extremely broad band power amplifier can be taken in at a glance. It shows that the output power can exceed 37dBm (except 150MHz) while input power as high as 27dBm in the pretty wide operation frequency range. Within the operation frequency ranging from 150MHz to 1200MHz, the maximum output power, driving by 27dBm input power, could be as high as 38.2dBm, while minimum

Freq. (MHz) Pin (dBm) Pout (dBm) Gain (dB) 2nd Harmonic (dBm) 2nd Harmonic (dBc) 3rd Harmonic (dBm) 3rd Harmonic (dBc) Id (mA) PAE (%)

150 27 36.55 9.55 27.06 -9.49 19.16 -17.39 370 43.62

300 27 37.72 10.72 25.08 -12.64 24.55 -13.17 420 50.30

600 27 38.2 11.2 25.65 -12.55 24.33 -13.87 450 52.44

900 27 37.3 10.3 26.65 -10.65 20.14 -17.16 420 45.67

1200 27 37.26 10.26 26.44 -10.82 5.27 -31.99 360 52.79

output power reaches 36.55dBm. Also under such driving condition and frequency range, PAE all exceeding 40%, 2nd harmonic being around -11dBc, 3rd harmonic ranging from -13.17dBc to -31.99dBc and the compressed gain is about 10dB.

Table 4.3 Achieved performances of this work @ Pin 27dBm

Categories Min Max

Frequency range (MHz) 150 1200

Gain (dB) 9.55 11.2

Output Power (dBm) 36.55 38.2

PAE (%) 43.62 52.79

2nd harmonic (dBc) -12.64 -9.49 3rd harmonic (dBc) -31.99 -13.17

Though the output power can exceed 37dBm, from Figure 4.10 to Figure 4.14, we can observe that non-linearity or compression imperceptibly begin at 22dBm or so, depending on the frequency. It is just as well to use in constant envelope modulation such like FSK or GMSK, which remains constant amplitude and the data is modulated

Though the output power can exceed 37dBm, from Figure 4.10 to Figure 4.14, we can observe that non-linearity or compression imperceptibly begin at 22dBm or so, depending on the frequency. It is just as well to use in constant envelope modulation such like FSK or GMSK, which remains constant amplitude and the data is modulated

在文檔中 應用於極寬頻之射頻高功率放大器設計與實現 (頁 37-0)