Introduction
Gallium nitride (GaN) technology has outperformed competitive technologies such as GaAs, SiGe, BiCMOS, and CMOS for power applications. GaN technology has also proved useful in low-noise applications due to built-in power handling capability, high linearity, and compact transceiver designs [Reference Rudolph1–Reference Zafar, Akoglu, Aras, Yilmaz, Nawaz, Kashif and Ozbay3]. A 60 nm GaN-on-Si process has shown better noise results up to 10 GHz at the device level compared to a 70 nm GaAs counterpart [Reference Ciccognani, Colangeli, Serino, Pace, Fenu, Longhi, Limiti, Poulain and Leblanc4].
Survivability and noise figure (NF) are two of the most important figures of merit (FoMs) for a low-noise amplifier (LNA) design. There have been reports of highly survivable and low NF X-band GaN LNA monolithic microwave integrated circuits (MMICs). More than 37 dBm survivability has been reported in [Reference Kim and Gao5–Reference Micovic, Kurdoghlian, Lee, Hiramoto, Hashimoto, Schmitz, Milosavljevic, Willadsen, Wong, Antcliffe, Wetzel, Hu, Delaney and Chow8] with an NF of better than 2 dB in all these designs. On the other hand, sub-1.8 dB NF LNA designs have been reported based on 0.09 and 0.25 µm GaN technology [Reference Kobayashi, Kumar, Campbell, Chen, Cao and Jimenez9–Reference Schuh and Reber13]. A high value gate bias resistor (RGB) technique was employed to achieve 42 dBm survivability [Reference Zafar, Akoglu, Aras, Yilmaz, Nawaz, Kashif and Ozbay3]. In another work, sub-1.2 dB NF was achieved owing to the choice of inductive source degenerated (ISD) HEMTs at both stages [Reference Zafar, Aras, Akoglu, Tendurus, Nawaz, Kashif and Ozbay14].
In the proposed work, the above-mentioned techniques, RGB and ISD at both stages, have been combined to achieve a low NF and robust LNA. Another focus of this work is bandwidth enhancement to report a 7.5–11.5 GHz broadband LNA. Along with other promising small-signal results, achieved sub-1.4 dB NF combined with 39 dBm survivability for 7.5–11.5 GHz is the best reported for GaN technology to date, to the best of the authors’ knowledge.
Reverse recovery time (RRT), a critical FoM for an LNA, refers to the time required for the gate bias to retrieve its steady-state value after the removal of an input stress. RRT depends primarily on the RC time constant and the trap phenomenon [Reference Zafar, Akoglu, Aras, Yilmaz, Nawaz, Kashif and Ozbay3]. The RC time constant is determined by RGB, gate bias capacitors (blocking and bypass capacitors), and the HEMT’s input capacitance (Cin). On the other hand, the trap phenomenon is inherent to the properties and fabrication processes of the GaN material. In [Reference Zafar, Akoglu, Aras, Yilmaz, Nawaz, Kashif and Ozbay3], it is shown that the trap phenomenon is a major contributor to RRT after a particular level of input stress. This work experimentally shows that for a fixed value of input stress, where the tarps are already excited, RRT is not a linear function of the RC time delay with changing value of RGB. This is a critical study in choosing RGB to address LNA’s robustness, especially for radar applications.
The fabrication process of the device and MMIC is discussed in the second section, while the HEMT parameters and LNA design are discussed in section “LNA design”. “Results and discussions” section covers the measured results of the fabricated MMIC, including small-signal, large-signal, survivability, and RRT. The conclusion is drawn in the last section.
Fabrication process
NANOTAM’s 0.15 µm AlGaN/GaN on SiC fabrication process is used for low-noise applications in the X-band. The epitaxial structure of HEMT, grown on a 3-inch SiC substrate by metal-organic chemical vapor deposition, consists of AlN nucleation, Fe-doped GaN buffer, GaN transition and channel layers, AlGaN barrier layer, and GaN cap layer. The next step is the fabrication of active and passive devices over the grown wafers. Ohmic contact formation (120 nm Ti/Al/Ni/Au stack), mesa isolation etching, a dielectric passivation layer, and gate formation (Ni/Au stack with 500 nm gate head and 150 nm gate foot) are related to active device fabrication. Mesa-etched regions are also used as high value resistances that have a density of 250 Ω/
$\square$. Passive device fabrication includes 80 nm TaN deposition to realize 30 Ω/
$\square$ thin-film resistor (TFR). It also includes metal-insulator-metal (MIM) capacitor formation with 220 nm thick dielectric layer of SixNy giving on-chip 270 pF/mm2 capacitance density. Backside process, the last step of fabrication, involves backside thinning of 100 µm, back-via holes opening through SiC substrate, and Au layer coating of 5.5 µm thickness. The back-vias serve the purpose of ground connections for microstrip transistors and circuits. Figure 1 shows the final sketch of the wafer, including active device and passive components.
Figure 1. Sketch of NANOTAM’s fabrication process [Reference Zafar15].
LNA design
HEMT’s selection and characterization
The first and foremost step in designing an LNA is the choice of a suitable HEMT. The deciding factors in this choice include the periphery and topology of HEMT. The choice of the periphery of the HEMT depends upon the maximum available gain (MAG) and minimum noise figure (NFmin). In our two-stage LNA design, we have chosen 4 × 75 µm ISD HEMT for both stages. The reason for choosing this periphery includes its MAG value of >10.5 dB and promising NFmin value of <0.78 dB in the frequency band of interest from 7.5 to 11.5 GHz, as evident from Figure 2a
Figure 2. (a) MAG and NFmin; (b) 
$\Gamma_{opt}$ and 
$\Gamma_{in}$ of 
$4\,\times\,75~{\mu}$m ISD HEMT, denoted by Gopt and Gin, respectively, for 12 V, 200 mA/mm.
. The parallel step is the topology selection, ISD or common source (CS). Compared to CS topology, ISD HEMTs have optimum noise impedance (
$\Gamma_{opt}$) as the conjugate of input impedance (
$\Gamma_{in}$), allowing the LNA designer to match the input reflection coefficient (IRC) and NF simultaneously [Reference Zafar, Aras, Akoglu, Tendurus, Nawaz, Kashif and Ozbay14]. Figure 2b shows 
$\Gamma_{opt}$ and 
$\Gamma_{in}$ of the selected 
$4\,\times\,75~{\mu}$m ISD HEMT, denoted by Gopt and Gin, respectively. An added advantage of ISD implementation in HEMT is avoiding the resistive network for stabilization towards the gate side, thus keeping NF at a low value [Reference Zafar, Aras, Akoglu, Tendurus, Nawaz, Kashif and Ozbay14]. Figure 3 shows fT and fmax of 
$4\,\times\,75~\mu$m HEMT determined to be 46.5 and 63.9 GHz by extrapolating the current gain and MAG, respectively
Figure 3. fT and fmax of 
$4\,\times\,75~{\mu}$m HEMT as extrapolation curves for current gain and MAG.
.
Figure 4a and 4b show load pull measurements of 
$4\,\times\,75~\mu$m ISD HEMT at 8 and 11 GHz respectively. The closeness of optimum impedance points for power and gain from load and gain circles leads to the option of the simultaneous match for output power and output reflection coefficient (ORC).
Figure 4. Load pull contours for 
$4\,\times\,75~{\mu}$m ISD HEMTs at (a) 8 GHz; (b) 11 GHz for 12 V, 200 mA/mm.
MMIC design and simulations
Figure 5 shows the schematic diagram of the proposed LNA, and Figure 6 shows the microphotograph of the fabricated LNA MMIC having the dimensions of 2.9 mm × 1.3 mm. The design starts with the stabilization of HEMTs, where the first stage HEMT is stabilized for frequencies greater than 2 GHz using TLbd, Rs, Cbd1, and Cbd2. For lower frequencies, stability is ensured by TLi2, TLibg, and Cibg. Second-stage HEMT is stabilized by adopting a similar procedure. As mentioned in Section “HEMT’s selection and characterization,” no resistive component is used towards the gate side of the HEMT, ensuring a low NF in the final design. Although the contribution of noise from the second stage, according to the Friis formula, is small, it has the role of deciding the final NF of a multi-stage LNA [Reference Zafar, Aras, Akoglu, Tendurus, Nawaz, Kashif and Ozbay14].
Figure 5. Schematic diagram of the proposed LNA.
Figure 6. Microphotograph of the fabricated LNA.
After fulfilling the unconditional stability condition, the design continues with input, inter-stage, and output matching networks. The stability networks further serve the purpose of gate and drain bias as well as matching, leading to a compact design. TLi1 and Cibl complete the input matching network, while Co1 and Cobl are further components for output matching network. Inter-stage matching network is finalized with Cit1, Cit2, and TLit1. LSD connected to the source terminal of HEMT represents the source degeneration. Finally, RGB is 5 kΩ mesa resistor in the gate bias line of the first stage to amplify the survivability of LNA. The scope of the design of this paper includes small and large-signal measurements, as well as robustness. Measurements are focused on small-signal, noise, output power at 1-dB compression point (OP1dB), output third-order intercept point (OIP3), survivability, and RRT.
Electromagnetic (EM) simulations are accomplished using PathWave Advanced Design System (ADS) from Keysight Technologies.Footnote 1 All matching networks are optimized simultaneously due to their effect on each other because small periphery HEMTs have little reverse isolation. Figure 7 shows that EM simulations result in small-signal gain >16.2 dB, NF <1.37 dB, IRC 
$ \lt -$8 dB, and ORC 
$ \lt -$10.5 dB in the desired frequency band.
Figure 7. Simulated and measured small-signal gain, reflection coefficients, and NF.
Results and discussions
Small-signal and noise measurements are carried out using PNA-X from Keysight Technologies, ZVA40 from Rohde & Schwarz,Footnote 2 and DC & RF probes from the GGB industries.Footnote 3 Maury’sFootnote 4 tuners, FormFactor’sFootnote 5 probe station, and IVCAD software of Amcad EngineeringFootnote 6 are utilized for large-signal measurements. We used the DSOS204A oscilloscope and the E8257D pulsed signal generator from Keysight Technologies for the recovery time measurements.
The bias condition for all the measurements is 12 V, 200 mA/mm under the base temperature of 25 ∘C. The small-signal, noise, and linearity measurements are performed with continuous wave (CW) signals, while survivability measurements are performed under pulse conditions having a 250 µs pulse period and 25 µs pulse width.
Small-signal and noise results
Figure 7 demonstrates the small-signal and noise characteristics of the fabricated MMIC. Small-signal gain greater than 15.3 dB and NF less than 1.36 dB are achieved for the bandwidth 7.5–11.5 GHz. A minimum NF of 1.15 dB is measured at 9.9 GHz. IRC and ORC are less than −8.4 and −11.1 dB, respectively, in the whole frequency band of interest. The measured and simulation results are in close agreement with each other. The promising values of NF and IRC at the same time are indicators of the advantage of ISD implementation in HEMT, as discussed in Section “HEMT’s selection and characterization.”
The value of obtained NF depends on the matching of the optimum noise impedance values and the resistive losses of the matching networks. To obtain the broadband response of NF across the 4 GHz bandwidth, the matched impedance values may slightly differ from the optimum values, which leads to a difference between NFmin of HEMT and the obtained NF of LNA. Moreover, as explained in Section “HEMT’s selection and characterization,” although the ISD HEMT topology keeps NF at a low value by avoiding the resistive network towards the gate side for stability, there is still a resistive contribution due to lossy components of the matching networks mentioned in Section “MMIC design and simulations” for the stability of both stages. This is another reason for the deviation of the NFmin and the NF value. Furthermore, according to the Friis equation, although small, there is a contribution to the overall NF of the LNA from the second stage. In the proposed LNA, we have achieved an NF of less than 1.4 dB in the entire frequency band of interest compared to the individual HEMT having an NFmin of <0.78 dB from 7.5 to 11.5 GHz.
Large-signal and survivability measurements
OP1dB and OIP3 values, indicators of the linearity, are measured to be 23 and 33 dBm, respectively, at 11.5 GHz, as evident from Figure 8 and 9. Survivability is a measure of the ability of an LNA in the receive chain to handle high input power levels. LNA is subjected to input stress levels of 37, 38, and 39 dBm in steps at 10 GHz, and small-signal gain is measured after each step, as shown in Figure 10a–c, respectively. LNA survives an input stress level up to 39 dBm without significant degradation in the gain.
Figure 8. OP1dB of fabricated MMIC at 11.5 GHz.
Figure 9. OIP3 of fabricated MMIC for 12 V, 200 mA/mm.
Figure 10. Gain before and after input stress levels of (a) 37, (b) 38, and (c) 39 dBm at 10 GHz.
Recovery time measurements
For RRT analysis using different values of RGB, LNA design from our previously published work [Reference Zafar, Akoglu, Aras, Yilmaz, Nawaz, Kashif and Ozbay3] with the new fabrication is used. RRT is a function of the RC time constant and the trap phenomenon. As explained in [Reference Zafar, Akoglu, Aras, Yilmaz, Nawaz, Kashif and Ozbay3], traps are excited after a particular input stress level, after which RRT becomes a dominant phenomenon and a major contributor to recovery time. Figure 11 depicts the variation of the forward gate current IGF with increasing input stress level for different values of RGB. As expected, it is evident that the value of IGF, an indicator of trap excitation, at 25 dBm input power (Pin) level is almost similar for RGB values of 1, 3.8, and 5 kΩ. Therefore, Pin = 25 dBm is a reasonable choice to observe the behavior of RGB value towards RRT for the same level of trap excitation.
Figure 11. Variation in IGF with change in the value of RGB.
RRT for an input stress level of 25 dBm and RGB values of 1, 3.8, and 5 kΩ is measured to be 60, 80, and >90 ms, respectively, as shown in Figure 12. The pulse conditions for the RRT measurements are 200 µs pulse width and 90 ms pulse period. It is evident that the LNA recovers fully for RGB values of 1 and 3.8 kΩ during the pulse relaxation time, while full recovery is not achieved for RGB value of 5 kΩ. This is the reason that the gain for 5 kΩ RGB LNA never touches its initial small-signal gain value of ∼20 dB.
Figure 12. RRT for Pin = 25 dBm and RGB values of 1, 3.8, and 5 kΩ.
The values of the DC block capacitor, bypass capacitor, and Cin are 0.4, 10, and 0.3 pF, respectively. RC time delay of 1, 3.8, and 5 kΩ RGB for 10
$\%$ to 90
$\%$ is calculated to be 23.5, 89.5, and 117.7 ns. It shows that RRT is not linearly dependent on the RC time constant with increasing values of RGB. It is experimentally concluded that the gate shifted to the negative value with increasing RGB takes much more time to return to its normal value than the expected RC time constant. Hence, after the trap excitation due to high input stress, RRT becomes a non-linear function of RGB.
It is worth mentioning that the use of an RGB of even greater value can enhance the ability of LNA to restrict the IGF, but at the cost of the shift in the gate bias voltage. Moreover, the factor that limits the choice of higher RGB is the ability of HEMT to withstand the negative voltage peak value before the breakdown [Reference Zafar15].
Comparison
Table 1 summarizes the state-of-the-art LNAs based on GaN and other competitive technologies with sub-1.8 dB NF, including this work. FoM is based on small-signal gain, OP1dB, bandwidth (BW), NF, PDC, and the center frequency (fo) as follows [Reference Xuan, Cheng, Gong, Zhang and Le28]:
\begin{equation}
FoM = 20~log_{10}~\left(\frac{Gain~[abs]~.~OP_{\textrm{1dB}}~[mW]~.~BW~[GHz]}{(NF~[abs] - 1)~.~P_{\textrm{DC}}~[mW]~.~f_o~[GHz]}\right)
\end{equation}Table 1. A comparison of designed MMIC with the recently reported GaN-based and other competitive technologies at X-band
* Estimated values from figures; **Simulation-based values.
This work reports competitive performance compared to GaN-based designs in the similar frequency band and with CMOS, GaAs, and SiGe technologies.
Conclusion
An X-band GaN-based LNA MMIC is fabricated using NANOTAM’s 0.15 µm technology. The advantages of ISD HEMTs in terms of easy stabilization to maintain low NFmin and simultaneous match condition for noise and IRC are exploited to design a two-stage LNA with a promising NF value of less than 1.4 dB. This NF value is close not only to the reported GaN-based designs but also to the SiGe and GaAs technologies recognized for the best noise performance. Along with achieving sub-1.4 dB NF for the frequency band of 7.5–11.5 GHz, LNA exhibits promising small-signal and linearity results. The design also covers the aspect of robustness by implementing high value RGB, acting as a limiter to survive high power levels up to 39 dBm. Recovery time measurements and their dependence on the different values of RGB are also analyzed in this work. It is concluded that after trap excitation due to high input stress, RRT becomes a non-linear function of RGB.
Acknowledgements
The authors are obliged to NANOTAM’s fabrication and measurement teams.
Competing interests
The authors declare none.
Salahuddin Zafar received MS degree in Electrical Engineering from NUST Islamabad, Pakistan, and PhD degree in Electrical and Electronics engineering from Bilkent University, Ankara, Türkiye, in 2012 and 2023, respectively. He has been working as a Senior Design Engineer in CESAT, Islamabad, and NANOTAM, Bilkent University, Ankara. His current research interests include the design and characterization of amplifiers for RF and microwave frequencies and the development of AlGaN/GaN on SiC-based MMICs.
Muhammad Imran Nawaz received his BS degree in Electrical Engineering from NUST, Islamabad, Pakistan, in 2005. He completed his MS degree in Electromagnetic Field and Microwave Technology from NPU, Xi’an, China, in 2014, and his PhD in Electrical and Electronics Engineering from Bilkent University, Ankara, Türkiye, in 2025. He has been working as a Design Engineer in CESAT, Islamabad, and the Nanotechnology Research Center (NANOTAM), Bilkent University, Ankara. His current research interests include the design and characterization of sub-system modules for RF and microwave frequencies and the development of AlGaN/GaN on SiC-based MMIC amplifiers.
Erdem Aras received the BS and MS degrees in Electrical and Electronics Engineering from Bilkent University, Ankara, Türkiye, in 2017 and 2020, respectively. He is currently pursuing his PhD degree in the same department. His doctoral research focuses on radio frequency (RF) and microwave monolithic integrated circuit (MMIC) design, with an emphasis on high-efficiency, low-noise, and high-power amplifier architectures. He is also with the Nanotechnology Research Center (NANOTAM), where he works as an RF Design Engineer. At NANOTAM, he has been involved in the development of gallium nitride (GaN) on silicon carbide (SiC) high-electron-mobility transistor (HEMT) technologies operating up to 40 GHz. His research interests include high-power MMIC design, nonlinear device modeling, and thermal and electrical performance enhancement techniques in GaN HEMTs.
Gizem Tendurus received her BS and MS degrees in Electrical and Electronics Engineering from Başkent University, Ankara, Türkiye, in 2017 and 2022, respectively. She has been working as a senior RF Design and Application Engineer at Nanotechnology Research Center (NANOTAM), Ankara. Her expertise covers process characterization in S, C, X, and Ku band technologies, including DC, small-signal, large-signal, and noise measurements. Her design experience includes amplifier MMIC development with a strong focus on phase shifters, switches, and low-noise amplifiers. She has also been actively involved in designing, testing, and packaging of GaN HEMTs specifically designed for radar and electronic warfare applications. With a solid foundation in both circuit design and measurement systems, she offers a comprehensive approach to high-performance RF and analysis.
Emirhan Urfali completed his BS in Electrical and Electronics Engineering from Başkent University, Ankara, Türkiye in 2019, and has since worked as an RF design engineer at NANOTAM, Ankara. His expertise covers process characterization across S-, C-, X-, and Ku-band, which includes DC, small-signal, large-signal, and noise measurements. His design experience includes amplifier MMIC development using cluster matching techniques, with a strong focus on stability and large-signal analysis. He has also been involved in designing, testing, and packaging of GaN HEMTs tailored for radar and electronic warfare applications. With a solid foundation in circuit design and measurement validation, he brings a comprehensive approach to advancing high-performance RF and microwave systems.
Ahsan-Ullah Kashif received his MS degree in Material Physics and Nano Technologies and PhD degree in Semiconductor Physics from Linköping University, Linköping, Sweden, in 2005 and 2010, respectively. He has been working as a Senior Researcher at CESAT, Islamabad, and as an Adjunct Professor at the IIU, Islamabad, Pakistan. He has an enormous amount of experience in the field of RF communication, from RF device level design to system level applications. His current research interests include design, development, and characterization of power amplifiers, receivers, and RF frontends for different communication bands.
Ekmel Ozbay received MS and PhD degrees from Stanford University, Stanford, CA, USA, in Electrical Engineering, in 1989 and 1992. He worked as a postdoc at Stanford University and later as a scientist at Iowa State University, Ames, IA, USA. He joined Bilkent University, Ankara, Turkey, in 1995, where he is currently a full professor in the Physics and Electrical and Electronics Engineering Departments. In 2003, he founded the Nanotechnology Research Center (NANOTAM), Bilkent University, Ankara, Türkiye, where he leads a research group working on nanophotonics, nanoelectronics, nanometamaterials, and GaN-based devices. He is the 1997 recipient of the Adolph Lomb Medal of OSA and the 2005 European Union Descartes Science award. He worked as an editor for Optics Letters, PNFA, SPIE JNP, and IEEE JQE journals. He has published 600+ articles in SCI journals. His papers have received 34000+ citations with an h-index of 89. He is also the CEO of a spin-off company: AB-MicroNano Inc., which is founded to commercialize the technologies developed in NANOTAM.
Open access
