A Survey of GaN HEMT Technologies for Millimeter-Wave Low Noise Applications

This article presents a set of measured benchmarks for the noise and gain performance of six different millimeter-wave (mm-wave) gallium nitride (GaN) high electron mobility transistor (HEMT) technologies fabricated at four different foundries in the United States. Measurements of the GaN transistors were collected on two independent noise parameter (NP) systems from 8–50 GHz and 75–110 GHz. The resulting raw NPs were stitched together to yield ultra broadband 8–110 GHz smoothed NPs. Several comparisons and summaries of the minimum noise figure and associated gain are reported for the six different GaN technologies. This work seeks to provide an initial database for noise and gain of GaN HEMTs at mm-wave frequencies to quantify progress on technology in the future.


I. INTRODUCTION
Gallium nitride (GaN) high electron mobility transistors (HEMTs) offer superior performance for radio frequency (RF) power amplifier (PA) applications due to their high voltage and current capabilities at higher frequencies [1].Recently, GaN HEMTs have become attractive for low noise amplifier (LNA) applications.Compared to high-performing low-noise technologies including gallium arsenide and indium phosphide, GaN HEMTs offer considerably higher breakdown voltages while maintaining good noise figure (NF) and gain [2].The high breakdown voltage is particularly important for transceivers as GaN-based LNAs can withstand higher input power levels.This eliminates the need for RF input limiters which increases the receiver's overall NF and size and complexity of the chips [2], [3], [4].
Several high linearity Ka-band GaN LNAs for 5 G applications have been shown in literature.In [5], a 100 nm GaN on Si LNA demonstrated a gain of 19.5-22.5 dB from 23-30 GHz with a 0.4-1.1 dB NF and OP1dB of about 20.8 dBm.Another work in the same technology has shown a gain of 26 dB from 33-38 GHz with a 2 dB NF and OP1dB of 20 dBm [6].A 90 nm GaN LNA demonstrated a gain of 14.3-24.4dB from 18-44 GHz with a 1.5-2.5 dB NF [7].A 150 nm GaN on SiC LNA measured a gain of 15-17 dB from 25-31 GHz with a 2.2 dB NF and OP1dB of 17.4 dBm [8].A previously reported LNA in 40 nm GaN had also shown 1-1.6 dB NF at 30-39.3 GHz with OP1dB of 11 dBm [9].
A critical aspect of first-pass design success of mm-wave GaN-based LNAs is the accuracy of the noise models employed in the circuit simulators.Accurate models require precision measurements of the GaN HEMT technologies.However, as the frequency of interest increases into the mmwave range, measurement systems which can collect noise data become rare.This becomes evident by investigating the measurement and modeling of the individual transistors used for GaN-based LNA designs in the literature.In [3], a 90 nm GaN HEMT was characterized and modeled across a lower 2-50 GHz frequency range and was used to project modeled performance and design the LNAs from 75-83 GHz and 91-95 GHz.The 63-101 GHz GaN-based LNA in [10] was designed using small-signal models extracted from S-Parameter measurements up to 225 GHz, but the noise models were extracted from on-wafer noise parameter (NP) measurements from 8-50 GHz.provided in Section III.In Section IV, a thorough summary and comparison of the mm-wave GaN HEMT measured noise and gain performances is presented.Section V provides conclusions for this mm-wave GaN HEMT survey.

II. BROADBAND NOISE PARAMETER MEASUREMENTS
Critical transistor figures of merit for LNAs include NF and small-signal gain.In this survey, the NPs of mm-wave GaN transistors are analyzed to shed light on their potential performance in LNA applications.The standard representation of a transistor's noise factor is defined as where F min is the minimum noise factor, R n is the equivalent noise resistance, and s,opt is the optimum noise reflection coefficient [16].In the remainder of this survey, the minimum NF, NF min = 10 log 10 F min , of the mm-wave GaN transistors is analyzed.The associated gain is defined as the available gain evaluated at the optimum noise reflection coefficient [17] where the output reflection coefficient is NP measurements were collected for each mm-wave GaN HEMT technology at AFRL's RF measurement laboratory.The transistors were characterized on independent test benches which operate across broad 8-50 GHz and 75-105 GHz ranges.Both systems utilize the Maury Microwave "cold-only" NP measurement formulation [18] in the Automated Tuner System (ATS) Software.Fig. 1 illustrates the first test bench used to characterize the NPs across the 8-50 GHz frequency range.Components of this measurement system includes an Agilent 346 C Option:K01 1-50 GHz noise source, the Maury Microwave MT7553N50 noise switch module, a Maury MT984AU 8-50 GHz automated mechanical impedance tuner, a Maury MT7553B03 noise receiver module, an Agilent E5273 A source measure unit (SMU), and a Keysight N5273B PNA-X.The PNA-X  is used as both a vector network analyzer and a NF meter in this system.FormFactor 50 GHz ACP probes and GGB 50 GHz picoprobes were used for probing the transistors.The W-band (75-110 GHz) NP measurement system is displayed in Fig. 2. The components of this system include a QuinStar QNS-FB15LW W-band noise source, two Flann 27333-3E W-band switches, a Maury MT979AL 75-110 GHz automated mechanical impedance tuner, a Maury MT7553M10 noise receiver module comprised of a W-band LNA and a mixer down-converter, two Virginia Diodes, Inc. WR-10 VNA extension modules, an Agilent E5273 A SMU, and a Keysight N5273B PNA-X.Again, the VNA and NFM features of the PNA-X are used in this measurement system.GGB WR-10 picoprobes were used to probe the transistors on the W-band system.
The following method was employed to synthesize NPs across the ultra broadband 8-110 GHz frequency range.First, the raw NPs were measured on both independent test benches and the resulting ATS files were saved to disk.The raw NP data files were then merged and loaded into the Maury ATS software.Then, the complete data sets were processed using a standard ATS procedure to generate smoothed NPs across the 8-110 GHz frequency range.    of the raw and smoothed NPs resulting from this technique.Here, the raw data collected on both the 8-50 GHz and the 75-110 GHz NP test benches are illustrated with black circles and the resulting smoothed data are illustrated with the solid black line.
It is clear that this approach provides a continuous set of NPs which can be interpolated to the 51-74 GHz frequency range where no raw measurements were collected.These resulting NPs can also be used for model extraction and validation across the entire broadband 8-110 GHz frequency range.

III. MILLIMETER-WAVE GAN HEMT TECHNOLOGIES
A total of six different mm-wave GaN HEMT technologies were characterized using the NP measurement systems outlined in Section II.Each transistor technology is described in the following subsections.

A. BAE SYSTEMS GAN09
BAE Systems developed a scaled device process targeting towards emerging applications at W-band [19].The key features of this device technology are epitaxial structures based on vertically-scaled barrier layers and 90 nm T-gates with low parasitic capacitance.The epitaxial structure consisted of a 5 nm thick In 0.04 Al 0.7 Ga 0.26 N barrier layer with a 1 nm thick AlN spacer layer and a 50 nm thick GaN channel.Structures contain an Al 0.04 Ga 0.96 N buffer layer to provide back-barrier carrier confinement which offers improvement to output conductance and device f max .These devices typically have I max of 1.8 A/mm and peak g m exceeding 700 mS/mm.Measured S-Parameter data show f T /f max of 107/227 GHz respectively, demonstrating the feasibility of this technology to provide sufficient gain for power amplification up to high W-band frequencies.Devices were characterized for power performance at frequency of 94 GHz using a Maury passive load pull system at AFRL.At quiescent bias of V d = 10 V and I d = 200 mA/mm, GaN09 microstrip (MS) devices without an on-wafer pre-match demonstrated 3.3 W/mm with 30% PAE and 5.2 dB associated gain [19].

B. HRL LABORATORIES T3
HRL's T3 GaN HEMT is a commercially available 40 nm T-gate monolithic microwave integrated circuit (MMIC) foundry technology fabricated on a double heterostructure epitaxy consisting of an AlGaN cap, an AlN barrier, a GaN channel, and an AlGaN backbarrier, with a two-dimensional electron gas (2DEG) channel charge of 1.6 ×10 13 cm −2 and electron mobility of 1000 cm 2 /V• s.The source and drain ohmic contacts are defined by selective-area n+ GaN MBE re-growth with a fixed spacing of 1.5 μm and a contact resistance of 0.1 -mm between source-drain metal electrodes and the 2DEG.The typical dc and RF performance of the T3 GaN HEMT in production include 0.85 -mm on-resistance, 1.4 A/mm maximum dc drain current density, 817 mS/mm peak transconductance, 72 V three-terminal breakdown voltage, f T of 140 GHz, and Ka-band saturated output power of 2.7 W/mm at V d of 14 V [20].
C. HRL LABORATORIES T3. 5 HRL's T3.5 GaN HEMT is a 40 nm T-gate MMIC technology fabricated on the identical AlGaN/AlN/GaN/AlGaN double heterostructure epitaxy that is described in the T3 section.The T3.5 differs in the aggressive and adjustable scaling of the source-drain length, also defined by selective-area n+ GaN MBE ohmic regrowth.The regrowth scales from 1.5 μm to 0.15 μm (150 nm) on the same wafer, offering a continuum of FET access lengths that can be tailored to the desired application.T3.5 is a method of fabricating previously reported advanced T4 devices [21] in a fraction of the cycle time in a manner completely compatible with the T3 fabrication process.T3.5 essentially offers an additional degree of design freedom-the source-drain length-with its resulting performance advantages.The T3.5 devices reported in this article have a source-drain length of 0.25 μm (250 nm) with a centered 40-nm T-gate.The typical device characteristics are as follows: 2.3 A/mm maximum drain current density, 1.2 S/mm peak transconductance, 0.35 -mm on-resistance, 15 V three-terminal breakdown voltage, 230 GHz cut-off frequency (f t ), 500 GHz maximum oscillation frequency (f max ), 3.1 W/mm saturated output power with 45% PAE at 94 GHz [22].Recently, HRL published record performance at W-band of 56 % PAE with 780 mW/mm associated output power density [23].

D. HRL LABORATORIES T3L
HRL's T3L GaN FET is a graded-channel AlGaN/GaN device with the engineering of GaN device linearity and the peak electric field [24].The device is fabricated as part of HRL's GaN MMIC technology.The device channel consists of a GaN layer and an Al-composition-graded AlGaN layer.The scaled graded-channel GaN offers the spreading of the electron carriers to form a quasi-3DEG.This was confirmed by measured capacitance vs. voltage curves of a graded-channel GaN HEMT structure [25].An AlGaN barrier layer is grown on top of the graded AlGaN layer to reduce gate leakage.The graded-channel AlGaN/GaN HEMTs were grown by MOCVD on SiC substrate.The graded-channel AlGaN/GaN HEMTs show electron mobility of 1650-1000 cm 2 /V• s, depending on carrier density of 5×10 12 -1.2×10 13cm −2 , comparable to scaled conventional AlGaN/GaN HEMTs.The ohmic process was done with selective re-growth of n+ GaN by MBE, showing an ohmic contact resistance of 0.1 -mm.With the source-to-drain spacing of 1.1 μm, the on-state resistance was about 0.9 -mm.The 50 nm gate-length T-gate was fabricated with a mini field-plate for additional electric field engineering.The Imax was about 1.6 -1.8 A/mm.HRL's T3L GaN recently demonstrated a record 45 % power added efficiency at 94 GHz [26] and 2 W W-band PA MMICs [27].

E. NORTHROP GRUMMAN CORPORATION GAN09
Northrop Grumman's (NG's) 90 nm GaN technology (GaN09) used for this study has been designed for V-and W-band operation.NG's GaN09 technology utilizes a 90 nm T-gate without field plates.The epi-profile includes a backbarrier and a scaled Schottky barrier to minimize short channel effects.Parasitic resistances are reduced with the use of regrown n+ GaN for Ohmic contact formation.Typical Ohmic contact resistance is < 0.1 -mm.Although discrete transistors were characterized, a full monolithic microwave integrated circuit process which includes metal-insulator-metal capacitors, thin film resistors, plated lines with airbridges, a 50 μm thick silicon carbide substrate, and slot thru-wafer-vias were used for fabrication.

F. TELEDYNE SCIENTIFIC N-POLAR GAN
Teledyne Scientific developed a nitrogen-polar (N-polar) GaN HEMT for RF power applications [28].The cross-section of the planar HEMT device is illustrated in Fig. 7.The HEMT epitaxy was grown by MOCVD on a 100-mm SiC substrate [29].The average sheet resistance of the structure obtained using Lehighton after deposition of an additional 14 nm of PECVD SiNx is also plotted in Fig. 7 and indicates 285 /sq with a standard deviation of 5 /sq.The device fabrication starts with the n++ GaN MOCVD regrowth process after hydrogen silsesquioxane (HSQ) patterning of the HEMT source-drain region for protection during regrowth.The film, composed of an unintentionally doped GaN and a highly Si-doped n++ GaN layer, is grown at a total thickness of 80 nm yielding a sheet resistance of 100 /sq.Following Ohmic regrowth, ion implant of As+ and He+ species is used for device isolation.Prior to gate metal patterning, a 6.5 nm MOCVD SiNx gate dielectric is deposited, which is reduced to a nominal thickness of 5 nm following the gate metallization pre-treatment etch.A tri-layer E-Beam resist process is then utilized for patterning the gate metal electrode at a gate length of 80 nm.The HEMTs are oriented such that channel transport is perpendicular to the substrate miscut steps, ensuring that the device current flows in the highest mobility direction.For HEMT device passivation, a PECVD SiNx film is used at a thickness of 120 nm.Finally 0.6 μm of Au Ohmic overlay metallization is patterned, which also serves as probe pad metal.
Teledyne's N-Polar GaN HEMT has exhibited excellent RF power performance, including a highly uniform 100 mm wafer demonstration of 43% PAE and 2 W/mm at an 8 V drain bias at 94 GHz [23].

IV. MEASURED RESULTS
The noise and gain performance of the mm-wave GaN HEMT technologies are next investigated.The various GaN technologies were measured with two different VNA calibration algorithms depending on whether the samples included on-wafer calibration structures.The WinCal line-reflect-reflect-match (LRRM) algorithm [30] was used to measure transistors which did not include any on-wafer calibration structures.The reference planes of these measurements were therefore at the probe tips of the test benches.The multiline thru-reflect-line (mTRL) calibration algorithm [31] was used for measuring transistors which included on-wafer calibration structures.The reference planes of measurements collected using the mTRL calibration algorithm were at the ends of the thru line "launch" structures and are typically closer to the intrinsic device.
The mTRL calibration algorithm is widely regarded as the most accurate for on-wafer measurements [32].This technique was utilized whenever possible.However, not all mm-wave GaN chips included on-wafer calibration structures.Therefore, special care was taken when analyzing measurements using different calibration algorithms as high frequency calibration is heavily affected by choice of the calibration algorithm [32], [33].In this survey, the authors are careful not to directly compare the measurement results obtained using the two different calibration algorithms.The measured results in the remainder of this article will therefore be reported in two categories: measurements collected using the LRRM calibration algorithm and measurements collected using the mTRL calibration algorithm.It is also noted again that the LRRM calibrated measurements on the W-band NP test bench were collected with GGB WR-10 Picoprobes.
The minimum NF NF min and associated gain G assoc of the mm-wave GaN HEMT technologies were analyzed by collecting NP measurements using the methodology outlined in Section II.This manuscript reports the best performing device out of the small sample size collected from each foundry.The standard definitions for the measured NPs were reported in Section II.In order to assess the physicality of the measurements, the Wiatr-Pospieszalski (WP) parameter [34], [35] was computed and used as a criterion for filtering out inaccurate NP measurements.

A. LRRM CALIBRATED MEASUREMENTS
The first set of data include LRRM calibrated NP measurements of the three technologies in a coplanar waveguide (CPW) configuration without on-wafer calibration structures.These device configurations contain ground-signal-ground pads with short CPW transmission lines leading to the intrinsic transistors.The technologies in CPW configuration are the HRL T3, HRL T3.5, and the Teledyne N-Polar GaN HEMTs.The HRL T3 and T3.5 HEMTs were biased with V DS = 4 V drain supply voltage and the Teledyne N-Polar  GaN HEMT was biased with V DS = 8 V drain supply voltage.Fig. 8 presents the smoothed NF min measurements determined from the processed NPs across an ultra broadband 8-110 GHz frequency range and at a single quiescent bias.The associated gain of the CPW transistors as a function of frequency from 8-110 GHz is reported in Fig. 9.The transistor quiescent drain current used for these measurements was I D = 100 mA/mm for each GaN HEMT technology.It is noted that G assoc is not processed or smoothed during the raw NP post processing procedure.
The trade-off between minimum NF and associated gain is next analyzed for the LRRM calibrated NP measurements.This is enabled by measuring the NPs of each transistor at a single drain supply voltage and across a range of quiescent drain currents.The drain supply voltage of V DS = 4 V is used for the HRL T3 and HRL T3.5 GaN HEMTs and a drain supply voltage of V DS = 8 V is used for the Teledyne N-Polar GaN HEMT.The gate supply voltages are swept such that the transistors are biased at a quiescent drain current ranging from 25 mA/mm to 350 mA/mm in non-uniform  steps.Fig. 10 presents the processed NF min as a function of quiescent drain current for the two HRL and Teledyne GaN technologies at 30 GHz.Each technology exhibits excellent Ka-band minimum NF properties across a wide range of operating conditions.Fig. 11 reports the associated gain of each GaN technology as a function of quiescent drain current at 30 GHz.These results compare favorably to prior N-polar GaN NF reports including [36], where an NF min of 1.6 dB was reported at 30 GHz.
The W-band noise and gain performance is also investigated using the ultra broadband NP measurements collected on the CPW GaN HEMTs.Fig. 12 illustrates the smoothed NF min measurements as a function of quiescent drain current at 94 GHz.These results range from 1.1 dB to 2.1 dB NF mintwo excellent benchmarks at mm-wave frequencies.Fig. 13   reports the measured G assoc as a function of quiescent drain current at 94 GHz.The wide range of low NF min enables a good trade-off between high gain and NF for each of these technologies.
A summary of the LRRM calibrated NP measurements of the HRL T3, HRL T3.5 and Teledyne N-polar CPW transistors is reported in Table 1.Here, the measurements collected at the 150 mA/mm quiescent drain current are reported.The total device periphery, W G and gate length, L G , are reported in the table.It is evident from the measured results that each technology demonstrates excellent noise and gain performance at Ka-band and W-band frequencies.These results also provide a good benchmark for future low noise technology demonstrations at mm-wave frequencies.

B. mTRL CALIBRATED MEASUREMENTS
The next set of measured results include CPW and MS extended reference plane (ERP) transistors on die which included on-wafer calibration structures for the mTRL calibration algorithm.The three technologies in these device configurations were the BAE GaN09, the HRL T3L, and the NGSP GaN09 HEMTs.
The trade-off between minimum NF and associated gain is again analyzed across the quiescent drain current space.Each GaN HEMT technology was biased using a drain supply voltage of V DS = 4 V, and the gate supply voltage was applied such that the quiescent drain current ranged from 25 mA/mm to 350 mA/mm in non-uniform steps.The smoothed NF min as a function of quiescent drain current measured on the three ERP transistors at 30 GHz are reported in Fig. 14.Fig. 15 presents the measured G assoc at 30 GHz from the mTRL calibrated NP measurements of the ERP transistors.These measurements provide excellent benchmarks for Ka-band low noise performance of mm-wave GaN HEMT technologies.Again, the W-band noise and gain performances of the mmwave GaN HEMT technologies are analyzed.These metrics are extracted from the on-wafer mTRL calibrated 8-110 GHz NP measurements.Fig. 16 presents the smoothed NF min at 94 GHz as a function of quiescent drain current for the three ERP GaN HEMTs.The measured G assoc at 94 GHz of the ERP transistors are illustrated in Fig. 17 A summary of the mTRL calibrated measurements of the BAE GaN09, the HRL T3L, and the NGSP GaN09 ERP GaN HEMTs is presented in Table 2.The quiescent drain current of each ERP transistor was 150 mA/mm for consistency in the summary.Each ERP GaN HEMT technology exhibits  excellent Ka-band and W-band noise and gain performance.This summary also serves as a good reference for future mm-wave low noise technology developments and amplifier demonstrations.

C. SMALL-SIGNAL MODEL ANALYSES
An extensive investigation of the measured NPs of both the CPW and ERP GaN HEMT technologies was presented in the previous section.To further aid in the analysis of the mmwave GaN transistors, small-signal models were extracted and analyzed to shed light on the noise performance.Small-signal  equivalent circuit (SSEC) models were extracted from the S-Parameter measurements included with the NP measurements.This work employs the small-signal model developed by Lu et al. [37].However, the analysis is not constrained to this particular SSEC model and could be accomplished with the various others that exist in the literature.
The first step in developing the SSEC model was extraction of the extrinsic parasitics.Cold field effect transistor (FET) measurements were employed for these purposes in a similar fashion to [37].The pinch-off cold FET measurements were collected at varying gate supply voltages according to the threshold voltages of each mm-wave GaN HEMT technology.Next, the extrinsic parasitics were fed into custom SSEC extraction codes.These codes optimized the extrinsic parasitic components in order to ensure a quality fit of the SSEC S-Parameter simulations compared to the measured S-Parameters.This optimization process also ensures that the intrinsic components of the SSEC are flat versus frequency in a fashion similar to the work by Jarndal in [38].An example of an extracted SSEC model compared to the S-Parameter measurements of a transistor is presented in Figs.18 and   mm-wave GaN HEMT technologies exhibited similar accuracy compared to the measurements.
The noise performance of the mm-wave GaN HEMT technologies is now investigated using the well-known Fukui minimum noise factor model [39], where K f is a fitting factor, f is frequency, C GS is the SSEC intrinsic gate-source capacitance, R G is the SSEC extrinsic gate resistance, R S is the SSEC extrinsic source resistance, and G m,i is the SSEC intrinsic AC transconductance.The pertinent SSEC intrinsic and extrinsic parameters extracted from the measurements of each mm-wave GaN HEMT technology were fed into the analytic model for minimum NF.For this work, a fitting factor of K f = 1.6 was employed to ensure good agreement between the calculated minimum NF and the measured equivalents.This fitting factor was used for evaluating each mm-wave GaN HEMT technology.The fitting factor used in this work could be useful for future evaluations of mm-wave GaN HEMT technologies as it is different than what was reported in [39].A summary of the SSEC model parameters and the Fukui model minimum NFs are reported in Tables 3 and 4. A 100 mA/mm quiescent drain current was used for evaluating Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.
It is clear from these two examples that NP measurements collected on GaN HEMT technologies in the upper mm-wave frequency ranges are rare in the literature.This article seeks to provide quantitative benchmarks for the noise and gain performances of mm-wave GaN HEMT technologies.Broadband NP measurements from 8-50 GHz and 75-110 GHz were collected on a total of six different mm-wave GaN HEMT technologies fabricated at four independent foundries in the United States.The raw NPs collected on the individual test benches were merged and processed together to yield ultra broadband 8-110 GHz smoothed NPs.The article is organized as follows.Section II briefly describes the NP systems at the Air Force Research Laboratory (AFRL) Sensors Directorate RF measurement laboratory.Descriptions of the six different mm-wave GaN HEMT technologies are

FIGURE 1 .
FIGURE 1.The 8-50 GHz NP system used to characterize the mm-wave GaN transistors.Photo by AFRL.

FIGURE 2 .
FIGURE 2. The W-band (75-110 GHz) NP system used to characterize the GaN transistors.Photo by AFRL.

FIGURE 3 .
FIGURE 3. Raw NF min (black circles) collected on the 8-50 GHz and 75-110 GHz NP measurements systems and the smoothed, continuous NF min (solid line) processed from the raw NP sets.The frequency bands of the two independent NP measurement systems are indicated with the vertical dashed lines.

FIGURE 4 .
FIGURE 4. Raw R n (black circles) collected on the 8-50 GHz and 75-110 GHz NP measurements systems and the smoothed, continuous R n (solid line) processed from the raw NP sets.The frequency bands of the two independent NP measurement systems are indicated with the vertical dashed lines.

FIGURE 5 .
FIGURE 5. Raw | s,opt | (black circles) collected on the 8-50 GHz and 75-110 GHz NP measurements systems and the smoothed, continuous | s,opt | (solid line) processed from the raw NP sets.The frequency bands of the two independent NP measurement systems are indicated with the vertical dashed lines.

FIGURE 6 .
FIGURE 6. Raw ∠ s,opt (black circles) collected on the 8-50 GHz and 75-110 GHz NP measurements systems and the smoothed, continuous ∠ s,opt (solid line) processed from the raw NP sets.The frequency bands of the two independent NP measurement systems are indicated with the vertical dashed lines.

FIGURE 7 .
FIGURE 7. Cross-section of the Teledyne N-polar HEMTs reported (not to scale) and sheet resistance across wafer after deposition of 14 nm of PECVD SiNx.

FIGURE 8 .
FIGURE 8. Measured NF min of the HRL T3 (plus signs), HRL T3.5 (squares), and the Teledyne N-Polar (inverted triangles) GaN HEMT technologies as a function of frequency from 8-110 GHz.The LRRM calibration algorithm was used for measuring each CPW transistor.

FIGURE 10 .
FIGURE 10.Measured NF min at 30 GHz of the HRL T3 (plus signs), HRL T3.5 (squares), and the Teledyne N-Polar (inverted triangles) GaN HEMT technologies as a function of quiescent drain current.The HRL transistors were biased with a 4 V drain supply voltage and the Teledyne N-Polar GaN HEMT was biased with an 8 V drain supply voltage.The LRRM calibration algorithm was used for measuring each CPW transistor.Measurements which did not satisfy the WP parameter were excluded.

FIGURE 11 .
FIGURE 11.Measured G assoc at 30 GHz of the HRL T3 (plus signs), HRL T3.5 (squares), and the Teledyne N-Polar (inverted triangles) GaN HEMT technologies as a function of quiescent drain current.The HRL transistors were biased with a 4 V drain supply voltage and the Teledyne N-Polar GaN HEMT was biased with an 8 V drain supply voltage.The LRRM calibration algorithm was used for measuring each CPW transistor.Measurements which did not satisfy the WP parameter were excluded.

TABLE 1 .FIGURE 12 .
FIGURE 12. Measured NF min at 94 GHz of the HRL T3 (plus signs), HRL T3.5 (squares), and the Teledyne N-Polar (inverted triangles) GaN HEMT technologies as a function of quiescent drain current.The HRL transistors were biased with a 4 V drain supply voltage and the Teledyne N-Polar GaN HEMT was biased with an 8 V drain supply voltage.The LRRM calibration algorithm was used for measuring each CPW transistor.Measurements which did not satisfy the WP parameter were excluded.

FIGURE 13 .
FIGURE 13.Measured G assoc at 94 GHz of the HRL T3 (plus signs), HRL T3.5 (squares), and the Teledyne N-Polar (inverted triangles) GaN HEMT technologies as a function of quiescent drain current.The HRL transistors were biased with a 4 V drain supply voltage and the Teledyne N-Polar GaN HEMT was biased with an 8 V drain supply voltage.The LRRM calibration algorithm was used for measuring each CPW transistor.Measurements which did not satisfy the WP parameter were excluded.
. The three ERP GaN HEMT technologies exhibit excellent noise and gain performance across a wide range of quiescent drain currents at the V DS = 4 V drain supply voltage.

TABLE 2 .FIGURE 14 .
FIGURE 14. Measured NF min at 30 GHz of the BAE GaN09 (diamonds), HRL T3L (circles), and the NGSP GaN09 (pentagons) GaN HEMT technologies as a function of quiescent drain current.Each transistor was biased using a 4 V drain supply voltage.The on-wafer mTRL calibration algorithm was used for measuring each ERP transistor.Measurements which did not satisfy the WP parameter were excluded.

FIGURE 15 .
FIGURE 15.Measured G assoc at 30 GHz of the BAE GaN09 (diamonds), HRL T3L (circles), and the NGSP GaN09 (pentagons) GaN HEMT technologies as a function of quiescent drain current.Each transistor was biased using a 4 V drain supply voltage.The on-wafer mTRL calibration algorithm was used for measuring each ERP transistor.Measurements which did not satisfy the WP parameter were excluded.

FIGURE 16 .
FIGURE 16.Measured NF min at 94 GHz of the BAE GaN09 (diamonds), HRL T3L (circles), and the NGSP GaN09 (pentagons) GaN HEMT technologies as a function of quiescent drain current.Each transistor was biased using a 4 V drain supply voltage.The on-wafer mTRL calibration algorithm was used for measuring each ERP transistor.

FIGURE 17 .
FIGURE 17. Measured G assoc at 94 GHz of the BAE GaN09 (diamonds), HRL T3L (circles), and the NGSP GaN09 (pentagons) GaN HEMT technologies as a function of quiescent drain current.Each transistor was biased using a 4 V drain supply voltage.The on-wafer mTRL calibration algorithm was used for measuring each ERP transistor.

FIGURE 18 .
FIGURE 18.A comparison of the measured (red solid lines) and the SSEC simulated (blue dashed lines) S-Parameters of a GaN HEMT biased at 4 V drain supply voltage and 100 mA/mm quiescent drain current.
19.Here, the measured S-Parameters are plotted with solid red lines and the SSEC simulated S-Parameters are plotted with dashed blue lines.These comparisons provide confidence that the extracted SSEC model accurately represents the measured data.The SSEC models extracted for the complete set of

FIGURE 19 .
FIGURE 19.A comparison of the measured (red solid lines) and the SSEC simulated (blue dashed lines) S 21 and maximum gain of a GaN HEMT biased at 4 V drain supply voltage and 100 mA/mm quiescent drain current.

TABLE 4 . SSEC Model Parameters and Fukui Noise Calculations of the On-Wafer mTRL Calibrated ERP Transistor Measurements each
GaN HEMT technology.The simple Fukui model indicates that NF is minimized when the transistor's gate-source capacitance, gate resistance, and source resistance are minimized while maximizing the intrinsic AC transconductance.In the case of the LRRM calibration measurements of the CPW transistors, the HRL T3.5 GaN HEMT exhibited the lowest combination of C GS , R G , and R S while simultaneously exhibiting the highest G m,i .Out of the mTRL calibrated measurements of ERP transistors, the BAE GaN09 and NGSP GaN09 HEMTs exhibited an excellent minimization of the C GS , R G , and R S parameters while maintaining good G m,i .These model results provide first-order quantitative insight into the mm-wave noise and gain performances of the GaN HEMT technologies that were analyzed in this survey.This work provided a set of measured benchmarks for noise and gain performance of mm-wave GaN HEMT technologies used in Ka-band and W-band frequency ranges.In particular, six different mm-wave GaN HEMTs were measured on two NP test benches from 8-50 GHz and 75-110 GHz.The resulting raw NP measurements were stitched together to yield smoothed NPs across an ultra broadband 8-110 GHz.Comparisons of the smoothed minimum NF and associated gain were presented for the transistors measured with LRRM and on-wafer mTRL calibrations.Excellent minimum NF and associated gain were observed in the various GaN HEMTs characterized in this article. V.CONCLUSION