• Sonuç bulunamadı

A 6-18 GHz GaN power amplifier MMIC with high gain and high output power density

N/A
N/A
Protected

Academic year: 2021

Share "A 6-18 GHz GaN power amplifier MMIC with high gain and high output power density"

Copied!
4
0
0

Yükleniyor.... (view fulltext now)

Tam metin

(1)

A 6–18 GHz GaN Power Amplifier MMIC with

High Gain and High Output Power Density

Batuhan Sutbas

*#1

, Ulas Ozipek

*#2

, Armagan Gurdal

*†3

, Ekmel Ozbay

*#4 *Nanotechnology Research Center (NANOTAM), Turkey

#Department of Electrical and Electronics Engineering, Bilkent University, TurkeyDepartment of Electrical and Electronics Engineering, Hacettepe University, Turkey

{1sutbas, 2ozipek}@ee.bilkent.edu.tr, {3agurdal,4ozbay}@bilkent.edu.tr

Abstract — A three-stage reactively-matched 6–18 GHz power amplifier MMIC design is presented. The design effort is focused on obtaining a low-loss output matching network for a high output power density. Active unit cells consist of an 8x125 µm transistor stabilized with a symmetrical parallel RC circuit. The wideband amplifier is fabricated using our in-house 0.25 µm GaN on SiC HEMT process. The fabrication technology details and overall device performance are reported. Experimental results show that the MMIC has a minimum gain of 22 dB and a maximum gain of 26.5 dB across the operation band. An average output power density higher than 3.3 W/mm with an associated average power-added efficiency of 22.5% is achieved. The MMIC demonstrates output power greater than 9.5 W at the center frequency. This design is distinguished from recent studies with its low-ripple high gain and high output power density.

Keywords — wideband, gallium nitride, silicon carbide, HEMT, reactive-matching, high power amplifier, MMIC.

I. INTRODUCTION

Aerospace and military applications require wideband power amplifiers for use in radars, electronic countermeasure and anti-jamming systems. Gallium nitride (GaN) based monolithic microwave integrated circuits (MMICs) are ideal for such applications thanks to the high electron velocity, wide bandgap and high breakdown voltage of the material. GaN devices fabricated on highly thermally conductive silicon carbide (SiC) substrates produce very high output power densities with high efficiency. However, there are fundamental limitations on the matching of transistor impedances, making the wideband design difficult [1].

Non-uniform distributed amplifiers have been used to incorporate transistor parasitics into artificial transmission lines achieving multi-octave bandwidths [2]. Some of the works combine distributed stages with reactively matched networks to sidestep low gain and limited output power capability of purely distributed topologies [3]. An alternative to the standard approaches was investigated in [4], using negative impedance matching components for broadband operation. Still, many of the wideband power amplifiers utilize reactive matching networks to avoid gain and power limitations of other techniques, at the expense of working at the design limits and having more complicated circuits [5]–[9].

This paper presents a 0.25 µm GaN on SiC high electron mobility transistor (HEMT) process developed at NANOTAM. The measured and characterized transistors are used to design

a 6 GHz to 18 GHz wideband three-stage power amplifier MMIC. Experimental results show that the fabricated MMIC achieves 22 dB to 26.5 dB gain in the band, more than 9 W output power and 32% efficiency at the center frequency using a 2 mm total gate periphery at the output stage. Despite having a moderate output power goal, the designed amplifier stands out with its low-loss and wideband output matching network which enables a higher power density with an associated higher efficiency compared to latest published works.

II. GAN HEMT TECHNOLOGY

The MMIC design is fabricated using our in-house 3-inch microstrip GaN on SiC technology. Ohmic contacts are formed using a Ti/Al/Ni/Au metal stack. Transmission line measurement (TLM) structures show a typical 0.25 Ω mm contact resistance. Transistors on the MMIC have T-shaped Ni/Au gates for optimized power performance. TaN sputtering is used to produce thin film resistors with a sheet resistance of 30 Ω/. The process uses two metal layers made of Au with thicknesses of 1 µm and 4 µm. 240 nm thick SixNy with

a dielectric constant of 7 is used as an insulator to form metal-insulator-metal capacitors. The SiC substrate thickness is 100 µm after back-via process. The substrate has a dielectric constant of 9.7.

Measured typical two-dimensional electron gas density of the AlGaN/GaN epitaxial layer is 1.08 × 1013cm−2 with

a corresponding average mobility of 2 × 103cm2V−1s−1 at

room temperature. Typical DC parameters are extracted from a measurement data set which contains more than a thousand transistors with gate peripheries ranging from 0.3 mm to 2 mm. Typical pinch-off voltage is −3.6 V and the typical knee voltage is 5 V. The average maximum drain current and peak transconductance densities are 1.03 A/mm and 315 mS/mm, respectively. Device breakdowns do not occur before at least 70 V which allows a safe operating drain voltage of 28 V for the MMICs.

S-parameter measurement results of 8x125 µm devices show a maximum available gain of 10.5 dB at 18 GHz with a 30 GHz cut-off frequency and an extracted 65 GHz maximum oscillation frequency. Load-pull measurements of the same devices show power densities more than 4.2 W/mm at 6 GHz, more than 4.4 W/mm at 12 GHz and more than 3.8 W/mm at 18 GHz when gain of the devices are compressed by

Proceedings of the 1st European Microwave Conference in Central Europe

(2)

3.5 dB. The associated power-added efficiencies and power gains are approximately 15.9 dB, 8.6 dB, 5.3 dB, and 55%, 41%, 29%, respectively. Measured output power contours at 3.5 dB compression from a typical 8x125 µm device are shown in Fig. 1. Here, source impedances are also optimized for maximizing the power gain.

III. MMIC DESIGN

The design goal for this application is to achieve a small-signal gain more than 21 dB with ripple less than 6 dB from 6 GHz to 18 GHz. The mean input and output return losses are expected to be at least 10 dB and 5 dB, respectively. The large-signal requirement is to obtain an average saturated output power higher than 6 W with at least 20% power-added efficiency using a 28 V voltage supply.

Although a distributed topology seems suitable for the multi-octave bandwidth amplifier, the pre-design attempts have resulted in a very high gain below 9 GHz and a gain drop after 15 GHz with unacceptable ripple. Moreover, in order to reach the output power goal, the amplifier has required too many active devices in both the driver and output stages which might have caused issues with the overall MMIC yield. Therefore, a three-stage reactively-matched structure is adopted. The interstage and output matching networks are designed for maximizing the output power which naturally produces a high gain at the lower band. Small-signal gain ripple requirement is satisfied by the input matching network design with intentional mismatch at the lower frequencies.

The input and driver stages each have one active device while the power stage combines two of them in parallel. All of the MMIC stages use 8x125 µm transistors with a drain-source spacing of 3.5 µm. The transistor geometry is carefully selected to accommodate both gain and power requirements at 18 GHz. T-shaped gates of the transistors have source-connected field plates for higher breakdown voltage and improved output power capability. For even-mode stability, parallel RC circuits are used in series at the transistor gates. The stability circuit is designed to preserve symmetry for the incoming RF signal to have a uniform distribution over all transistor fingers. The active unit cell is optimized for unconditional stability on itself in order to precautiously shut down possible oscillations emerging from loops in the bias circuits. For odd-mode stability, odd-mode resistors are added between the output transistor gates and drains. The resistor values are selected to ensure that the negative resistance oscillation conditions [10] are totally avoided even if a phase difference of 10◦ between

the output transistors is introduced. The oscillation conditions are also checked for admittance parameters imposing the same phase margin.

The bias circuits are incorporated into the matching networks and two shunt capacitors are used for each line to maintain good RF isolation in the entire band. Width of the lines are determined considering the maximum current carrying capacities. A shunt RLC circuit is used at the input network to achieve an input reflection coefficient of −10 dB despite the intentional mismatch at the lower band. The first

(a)

(b)

(c)

Fig. 1. Measured load-pull transistor output power contours at: (a) 6 GHz; (b) 12 GHz; (c) 18 GHz.

(3)

Fig. 2. Microscope image of the fabricated MMIC with test transistors.

interstage network between the input and driver stages aims a gain and output power balance. On the other hand, the second interstage network is designed to present optimum load and source impedances for maximum output power with an emphasis on higher frequencies. Loss of the output matching network is less than 0.4 dB below 15 GHz and becomes 0.6 dB at 18 GHz. Low-loss of the output matching network enables a high output power density. The initial design is completed using table-based models developed for passive matching components and the final layout is verified with full electromagnetic simulations taking cross-coupling and additional parasitics into consideration.

The fabricated die photograph of the MMIC is shown in Fig. 2. Test transistors with and without the stability cell are placed in the bottom left corner for process monitorization and model verification purposes. The overall circuit size is approximately 8.5 mm × 2.5 mm. Future design goals include achieving a smaller chip size by folding the input stage under the interstage network and meandering the output transmission line.

IV. EXPERIMENTALRESULTS

Small-signal and large-signal measurements of the fabricated MMIC are completed on-wafer at room temperature. All transistors on the MMIC are biased with 28 V drain voltage and about 120 mA drain current, operating in class-AB mode. Oscillatory behaviors are not observed, transistors draw the same quiescent current throughout the measurements.

Fig. 3 shows the simulated and measured S-parameters of the MMIC, illustrating the good agreement between them. The measured small-signal gain is at least 22 dB from 6 GHz to 18 GHz with an average of 24 dB. Although the simulation expectations for the gain ripple is less than 3.5 dB, the measurements show a gain expansion at 13 GHz up to 26.5 dB, possibly as an outcome of slightly changed transistor models. As a result, the measured gain ripple is 4.5 dB and the fabricated MMIC behaves marginally better than the simulation expectations, demonstrating a wider operation bandwidth up to 18.7 GHz. Averages of the measured input and output return losses are about 10 dB and 8.5 dB, respectively.

5 10 15 20 Frequency (GHz) -20 -10 0 10 20 30 |S 21 |, |S 11 |, |S 22 | (dB) Sm. (|S 21|) Meas. (|S 21|) Sm. (|S 11|) Meas. (|S11|) Sm. (|S 22|) Meas. (|S 22|)

Fig. 3. Simulated (solid) and measured small-signal gain, input and output return losses of the wideband amplifier.

Fig. 4 shows the output power and efficiency of the MMIC measured at 6.5 dB gain compression from 6 GHz to 18 GHz with 1 GHz measurement steps. The maximum output power is more than 9.6 W with an overall average of 6.6 W. The associated power-added efficiency average is about 22.5%. In contrast to the simulation expectations, the MMIC is underperforming below 8 GHz, we suspect that the driver stage is highly compressed, leaving a smaller compression margin for the output stage and limiting the performance. Fig. 5 illustrates an example of the input power sweep recorded at the center frequency of 12 GHz. The transducer gain at low input drive coincides with the small-signal gain measurements. Performance comparison with recently reported GaN based wideband power amplifier MMICs is tabulated in Table 1. Here, average output power and average efficiency across the operation bandwidth are considered. Although our amplifier design targets a rather modest output power, a normalization with the total gate periphery used at the output stage shows that our work achieves a power density better than most of the recent studies. Moreover, the presented amplifier is more attractive with high average power-added efficiency and low gain ripple. It is also notable that power measurements of this chip are performed at a much lower compression level.

(4)

Table 1. A performance comparison with recent GaN based wideband power amplifier MMICs

Ref. Year No. of Type of Bandwidth Chip Size SS Gain Comp. PStages Matching (GHz) (mm2) (dB) (dB) (W) (%) (mm) (W/mm)out PAE OGP Pdensity

[2] 2016 EL 2 Distributed 6 – 18 7.8 × 2.7 13.5 – 18 >8 25.7 14 13.6 2.1 [3] 2015 EuMIC 3 Mixed 6 – 18 5.0 × 5.0 22 – 28 >5 10 15 4.8 2.1 [5] 2018 EuMIC 2 Reactive 6 – 18 5.0 × 5.0 13 – 23 >11.5 8.3 17.5 3.2 2.6 [4] 2015 MTT 2 Non-foster 6 – 18 3.4 × 2.6 13.5 – 19.1 >7.5 4.7 17 1.5 3.1 [6] 2017 EuMIC 3 Reactive 6 – 17 3.8 × 1.8 17 – 29 >12 5 13 1.6 3.1 [7] 2018 MTT-S 2 Reactive 6 – 18 7.6 × 6.2 17.5 – 25.5 5 40 21.5 12.8 3.1 [8] 2013 MTT 3 Reactive 6 – 18 5.5 × 3.5 18 – 27 >12.5 10.6 18 3.2 3.3 [9] 2017 EuMIC 3 Reactive 6 – 18 2.6 × 3.6 26 – 31 >12.5 17 25.5 3.12 5.5

This work 3 Reactive 6 – 18 8.5 × 2.5 22 – 26.5 6.5 6.6 22.5 2 3.3

6 8 10 12 14 16 18 Frequency (GHz) 28 30 32 34 36 38 40 42 Output Power (dBm) 0 5 10 15 20 25 30 35 PAE (%) Pout(dBm) PAE (%)

Fig. 4. Measured output power and power-added efficiency parameters of the wideband amplifier at 6.5 dB compression levels.

-10 -5 0 5 10 15 20 25 P n ava lable (dBm) 18 19 20 21 22 23 24 25 26 Ga n (dB) 0 5 10 15 20 25 30 35 40 P out (dBm), PAE (%) Gt(dB) Pout(dBm) PAE (%)

Fig. 5. Transducer gain, output power and power-added efficiency of the wideband amplifier versus swept input power at 12 GHz.

V. CONCLUSION

Our in-house 0.25 µm GaN on SiC HEMT technology and design considerations for a wideband power amplifier covering C-, X- and Ku-bands are presented. The fabricated MMIC achieves a small-signal gain more than 22 dB with low-ripple from 6 GHz to 18 GHz. Experimental results show an average output power higher than 6.6 W with an associated power-added efficiency of 22.5% using 2 mm total gate periphery at the output stage. The three-stage amplifier is

distinguished with its high power density. Future efforts will focus on decreasing the overall circuit size and improving the output power below 8 GHz.

ACKNOWLEDGMENT

The authors would like to thank Dr. ¨Ozlem S¸en, ¨Omer Cengiz and Halil Akcalı for helpful comments and M. ¨Omer Akar for providing support in device load-pull measurements. The authors wish to thank the process team at Bilkent University NANOTAM for the MMIC fabrication.

REFERENCES

[1] R. M. Fano, “Theoretical limitations on broadband matching of arbitrary impedances,” J. Franklin Inst., vol. 249, pp. 57–83, Jan. 1950. [2] J. Kim, H. Park, S. Lee, J. Kim, W. Lee, C. Lee, and Y. Kwon,

“6–18 GHz 26W GaN HEMT compact power-combined non-uniform distributed amplifier,” Electron. Lett., vol. 52, no. 25, pp. 2040–2042, Dec. 2016.

[3] P. Dennler, S. Maroldt, R. Quay, and O. Ambacher, “Monolithic three-stage 6–18 GHz high power amplifier with distributed interstage in GaN technology,” in Proc. Eur. Microw. Integr. Circuit Conf., Paris, France, Sep. 2015, pp. 29–32.

[4] S. Lee, H. Park, K. Choi, and Y. Kwon, “A broadband GaN pHEMT power amplifier using non-foster matching,” IEEE Trans. Microw. Theory Techn., vol. 63, no. 12, pp. 4406–4414, Dec. 2015.

[5] E. Oreja-Gigorro, E. D. Pascual, J. J. Sanchez-Martinez, M. L. Gil-Heras, V. Bueno-Fernandez, A. Bodalo-Marquez, and J. Grajal, “A 6–18 GHz GaN on SiC high power amplifier MMIC for electronic warfare,” in Proc. Eur. Microw. Integr. Circuit Conf., Madrid, Spain, Sep. 2018, pp. 85–88.

[6] G. C. Barisich, E. Gebara, H. Gu, C. Storey, P. Aflaki, and J. Papapolymerou, “Reactively matched 3-stage C-X-Ku band GaN MMIC power amplifier,” in Proc. Eur. Microw. Integr. Circuit Conf., Nuremberg, Germany, Oct. 2017, pp. 93–96.

[7] M. Litchfield and J. J. Komiak, “A 6–18 GHz 40W reactively matched GaN MMIC power amplifier,” in IEEE MTT-S Int. Microwave Symp. Dig., Philadelphia, PA, USA, Jun. 2018, pp. 1348–1351.

[8] U. Schmid, H. Sledzik, P. Schuh, J. Schroth, M. Oppermann, P. Bruckner, F. van Raay, R. Quay, and M. Seelmann-Eggebert, “Ultra-wideband GaN MMIC chip set and high power amplifier module for multi-function defense AESA applications,” IEEE Trans. Microw. Theory Techn., vol. 61, no. 8, pp. 3043–3051, Aug. 2013.

[9] H.-Q. Tao, W. Hong, and B. Zhang, “6–18 GHz 13W reactive matched GaN power amplifier MMIC,” in Proc. Eur. Microw. Integr. Circuit Conf., Nuremberg, Germany, Oct. 2017, pp. 357–360.

[10] R. G. Freitag, “A unified analysis of MMIC power amplifier stability,” in IEEE MTT-S Int. Microwave Symp. Dig., Albuquerque, NM, USA, Jun. 1992, pp. 297–300.

Şekil

Fig. 1. Measured load-pull transistor output power contours at: (a) 6 GHz;
Fig. 2. Microscope image of the fabricated MMIC with test transistors.
Table 1. A performance comparison with recent GaN based wideband power amplifier MMICs

Referanslar

Benzer Belgeler

Should terrorism be defined as a legal concept or, is it preferable to define the criminal acts that take part in the actions taken by terrorist organizations separately

These studies (e.g., Beck, 2000; Giddens, 2000) view globalization as leading to major socioeconomic transformations that generally produce insecurities for domestic/societal

Analysis of normal breast tissue revealed that Chk2 mRNA was abundantly expressed in all cases analysed (Figure 2a), but steady- state levels of Chk2 mRNA were down-regulated in 16/

In this chapter we will construct the notion of continuous space which is defined by modifying the first axiom of diffeological spaces (see page 5 in [6]) and later it will be

We have developed a mathematical model which con- currently selects product design, generates process plans and develops production plans to minimize the sum of processing cost,

We asked four observers to compare the glossiness of pairs of surfaces under two different real- word light fields, and used this data to estimate a transfer function that captures

However, the complexity of the structure and the fact that the piezoelectric we use is a ring rather than a full disk necessitate the use of finite ele- ment analysis to determine

Following fiber amplification, these pulses are used in micromachining titanium surfaces and compared to a picosecond and a nanosecond laser... Thus, the entire setup is