GaN HEMT Performance --- Measurements and ... - CiteSeerX

0 downloads 13 Views 88KB Size Report
indep(Pdel_contours_p) (0.000 to 32.000). P del_c o nto urs. _p m2. Fig. 6 Simulated Pout vs. Pin for Vds=48 V. Fig. 5 10 mm device. Simulated load pull for.
Proceedings of the 36th European Microwave Conference

GaN HEMT Performance – Measurements and Simulations of a 3.6 mm Device from Cree Asher Madjar 1 , Zygmond Turski2, Yifei Li2 1

ECE Department, Temple University, Philadelphia, PA 19122, e-mail: [email protected] 2 F&H Applied Science Associates, Mt. Laurel, NJ 08054

ABSTRACT — The premise of GaN HEMT devices to offer superior power amplification performance in microwave frequencies lead to an extensive research in this field, and to numerous publications emphasizing individual performance records such as, current & power densities, linearity and break down voltage. The performance of these, primarily experimental devices, is often reported in the terms of their intended applications, offering either the saturated output power, or 3dB compressed output power, or OFDM and CDMA waveform–specific output power, making the comparison among the devices, and a comparison with other technologies such as LDMOS & GaAs rather difficult. The authors are engaged in the design of a wideband (10:1 BW) 50W @ 1dB compression point linear power amplifier operating up to 2GHz. As a step in this process we determine the performance of an actual, 3.6 mm periphery, preproduction GaN HEMT from CREE. The device performance is simulated using CREE model, and it is compared with the measurements we have taken, both reported in this paper. The results obtained exhibit good convergence with the model, and are presented using the more common performance parameters such as power and efficiency at 1dB compression point, and the linearity in terms of IP3. Index Terms — GaN, HEMT, wide band gap, power amplifiers.

I. DESIGN OBJECTIVES This work is carried out at F&H Applied Science Associates, Mt. Laurel NJ, whose objective is to commercialize the Wide Band Gap (WBG) SiC and GaN technologies. The first design objective is the development of a highly efficient 50 W, 20 dB gain, linear, wideband amplifier operating over the 200MHz to 2,000 MHz bandwidth. The future objective will be to extend the frequency of operation to 4,000 MHz. Currently many similar requirements are met using well established LDMOS technology, except for the bandwidth, efficiency & linearitywithout linearization. In fact, LDMOS amplifiers cannot achieve the required bandwidth, they demand an elaborate cooling system, and there is a substantial trade-off between efficiency and linearity. Here is where the attractive features of GaN HEMT come to play; GaN high breakdown voltage and high current density yield a much smaller device, which reduces the parasitic capacitances and inductances, thus allowing for a wider bandwidth. The high operating temperature of GaN reduces the complexity of the cooling system. These advantages are well documented in the literature ([1]-[5]). Also, the reportedly higher linearity of GaN HEMT may require lesser linearization complexity, and in some applications, it may eliminate the need for linearization altogether. These considerations lead the authors to a selection of a GaN HEMT for this project. In close cooperation with CREE we embarked on the task of

2-9600551-6-0  2006 EuMA

characterization and evaluation of a CREE pre-production 3.6 mm periphery GaN HEMT. The results presented in this paper follow basic definitions, intrinsic to the device itself, and not dependent on any specific modulation format or waveform. Thus, we operate in Class A, define the output power at the 1 dB compression point, and the linearity in terms of third order intermodulation level (IMD3) at several power levels (and in particular at the 1 dB compression point). The results of this work confirm the many benefits of GaN, as well as reveal some existing material & processing problems that need resolution. These results also helped us determine the required periphery size, and the expected performance of the actual 60 Watt device to be developed.

II. CREE DEVICE DESCRIPTION The Cree GaN HEMT employs GaN on SiC technology. The well known, excellent thermal conductivity of SiC reduces the thermal resistance of the device. The gate periphery of the tested device is 3.6 mm (10 fingers @ 360 micron length each). At the present time, the maximum DC drain voltage is restricted to 35 V, but in the near future it is expected that the process will support the desired 48V. The performance of this device sets the reference for the process upon which a whole group of GaN devices of various sizes and power capabilities will be developed. Based on our measurements we estimate at this time that a 10 mm periphery GaN HEMT operating at VDS = 48V will produce the desired, 60 W output power

III. SIMULATED PERFORMANCE A CREE-developed, large signal device model was put to test. We have used this model to estimate the device’s performance, and the load characteristic for maximum output power and for maximum efficiency. In particular, the 3.6 mm periphery GaN HEMT was simulated in class-A operation, with the gate voltage, VGS = -1.5 V and drain voltages, VDS = 35 V and 48 V, respectively. The load pull has been simulated at the frequency of 1 GHz, yielding output power of 41.18 dBm (12.58W) and efficiency (PAE) of 32%, at 1 dB compression point [Fig. 1]. Similar simulation at 48V yields output power 44.12 dBm (25.8W) and efficiency (PAE) of 39%, at 1 dB compression point [Fig. 2].

IV. MEASURED DATA AND COMPARISON TO SIMULATION In order to verify the simulated results, a set of measurements was taken using the 3.6 mm device. The device was placed on a test jig with 50 Ohm input and output microstrip lines. Gate and drain bias of VGS = -1.5V and VDS = 35V, respectively, were supplied via two bias-Ts. Two triple-stub tuners were attached,


September 2006, Manchester UK

one at the Gate Port, and the other at the Drain Port. The input side tuner was pre-adjusted to conjugately match the input impedance to 50 Ohm. The output tuner was pre-adjusted to present a Zopt impedance to the drain. Both were tuned at the experiment frequency of 1 GHz. Two dual-directional couplers, one at the input and the other at the output provided us with the ability to manually tune the input tuner for minimum reflection, and the output tuner to maximize the output power. The test jig was cooled using a small fan. The following two tests were conducted, and subsequently, the efficiency was calculated. 1. A single-tone measurement of the output power vs. input power, and determination of the 1dB compression point [Fig. 3]. 2. A two-tone measurement to determine the 3rd. order distortion level [Fig. 4].

REFERENCES 1. R. J. Trew, G. L. Bilbro, et. al, “Microwave AlGaN/GaN HFETs”, IEEE Microwave Magazine, March 2005, pp. 56-66. 2. Y. F. Wu, A. Saxler, et. al, “30 W/mm GaN HEMTs by field plate optimization”, IEEE Electron Devices Letters, vol. 25, pp. 117-119, Nov. 2004. 3. R. J. Trew, “AlGaN/GaN HFET amplifier performance and limitations“, 2002 International Microwave Symposium Digest, Seattle, pp. 1811-1814. 4. R. J. Trew, “SiC and GaN Transistors: Is there one winner for microwave power applications?”, Proceedings IEEE, pp. 1032-1047, June 2002. 5. L. F. Eastman, et. al, “Undoped AlGaN/GaN HEMTs for microwave power amplification”, IEEE Transactions Electron Devices, pp. 479-485, March 2001.

Plots of the measured output power vs. input power at 1 GHz depicted in Fig. 3 are for VDS = 35 V. Comparing Figs. 1 and 3 we can see that the simulated and measured output power levels agree quite well, with the small discrepancy being attributed primarily to the thermal effects. The power added efficiency at the 1 dB compression point is about 32% for Vds=35 V, however, in the future device that will operate at VDS = 48 V we can expect an efficiency of more than 40% for the class-A operation. The third order intermodulation was measured by applying twotones to the input at a very small frequency separation (5 MHz). The third order intermodulation product is depicted in Fig. 4 vs. input power of each tone. Also depicted are the output power for each tone and the Peak Envelope Power (PEP). We can see that at the 1 dB compression point (16 dbm input power per tone), the 3rd order intermodulation product is more than 20 dB below the signal level.


Validation of a non-linear model for the 3.6 mm device adds credibility to the next simulation of a 10 mm device needed for the 60 W output power. The simulated load pull results for a 10 mm device operating at VDS = 48V class A, and a frequency of 1 GHz are depicted in Fig. 5. The simulation indicates that the device is capable of delivering about 60 W. The simulated Pout vs. Pin at 1 GHz for VDS = 48 V is depicted in Fig. 6.

VI. CONCLUSION The ongoing GaN HEMT reliability improvements rapidly propel the process towards the 50V operating point mark. Improvements beyond the 50V mark are expected to continue as well. At this point it is not too soon to start considering practical applications of a GaN HEMT. Our work is directed towards this objective, and in this paper we report some preliminary measurements performed on an actual, 3.6 mm periphery device operating at 35V. Our measurements of the output power, linearity and efficiency shad more light on the present device performance (VDS = 35V), validate the model and stipulate the device performance at VDS = 48V.


43 42 41

Pout: dBm


Pdel_contours_p PAE_contours_p

m2 m1


40 dBm @ 1dB compression point

38 37 36 35 34 33 14

in d ep (P AE _ con t ou rs_p ) (0.0 00 to 44 .00 0) in d e p (P d el _co n to u rs_ p ) (0 .00 0 to 3 8.0 00 )

m1 indep (m 1)=2 m 1= 0.386 / 169.84 6 leve l=3 2.2658 55, num ber= 1 im pe danc e = Z 0 * (0.4 46 + j0 .0 71

Fig. 1.







Pin: dBm

m2 indep (m 2)=3 m 2= 0.56 1 / 158 .68 5 leve l= 41.184 543, num ber= 1 im pe danc e = Z 0 * (0.2 90 + j0 .1 7 3

Fig. 3 Measured single tone output power vs. input power at VDS = 35 V

3.6 mm device. Simulated load pull at Vds=35 V


output power: dBm


Pdel_contours_p PAE_contours_p

m1 2

30 PEP IMD3 Pout per Tone




in d ep (PAE_con tou rs_p ) (0.000 to 40.000) in d ep (P d el_con tou rs_p ) (0.000 to 36.000)

m1 indep(m 1)=2 m 1=0.431 / 150 .715 level=39.580347, num ber=1 im peda nce = Z 0 * (0.420 + j0.21 8

-10 11

m2 indep(m 2)=3 m 2= 0.426 / 151.467 level=44.128193, num be r=1 im pedan ce = Z 0 * (0.424 + j0.21 1







input power per tone: dBm

Fig. 4 Measured third order intermodulation vs. input power for Vds=35 V

Fig. 2.3.6 mm device. Simulated load pull at Vds=48 V

Pout vs. Pin

51 P1dB=48.65dB


Pout dBm

Pdel_contours_p PAE_contours_p

50 m2 m1

48 47 46 45 44 43 42

indep(PAE_contours_p) (0.000 to 38.000) indep(Pdel_contours_p) (0.000 to 32.000)

m1 indep(m1)=4 m1=0.757 / 178.637 level=35.023177, number=1 impedance = Z0 * (0.138 + j0.012











Pin: dBm

m2 indep(m2)=3 m2=0.734 / 177.055 level=48.750256, number=1 impedance = Z0 * (0.154 + j0.025

Fig. 5 10 mm device. Simulated load pull for Vds=48 V

Fig. 6 Simulated Pout vs. Pin for Vds=48 V