Silicon-Germanium Front-End Electronics for Space-Based Radar Applications

The primary goal of the research under this Fellowship was to leverage the unique capabilities of SiGe BiCMOS technologies to develop low-power, highly integrated microwave remote sensing instruments to lower costs and improve the feasibility of future remote sensing missions. The initial research focus of this fellowship was to investigate the response of an X-band SiGe HBT power amplifier to cryogenic temperatures and ionizing radiation, as the first step in developing a temperature-invariant and radiation-tolerant driver PA for space-based X-band radar systems. The amplifier under test unexpectedly became quite unstable when operated at 78 K, while it was quite stable at room temperature. This was caused by the increased transconductance of the SiGe HBTs with cooling, and prevented linearity measurements of the full PA at cryogenic temperatures. Instead, the linearity of a PA core was measured across temperature, and decreased output power was observed as expected. Separate PA chips were exposed to 3 Mrad(SiO2) doses of 10 keV X-ray irradiation as well, and little performance degradation was observed as expected due to the built-in total-dose radiation tolerance of SiGe HBTs. After the first year, the focus of this research shifted to exploring the benefits of SiGe HBTs for use in 18.7 GHz snow-sensing radiometers. Proposed satellite-based snow and ice measurement concepts require hundreds of highly-integrated ultra-low power radiometer receivers. Integrated SiGe receivers would enable cost reductions and improve the feasibility of such mission concepts. Two SiGe LNAs were designed to explore the attainable noise performance of SiGe LNAs, leveraging best-in-class SiGe HBTs with peak fT / fMAX of 300 / 500 GHz to achieve unprecedented noise performance at 18.7 GHz with very low power consumption. To further improve NF, a localized backside etching (LBE) module was utilized to remove the conductive silicon substrate under the inductors and suppress substrate-induced losses. This LBE improved the noise figure of the LNAs by an average of 0.12 dB across the desired band. The first LNA used a single-stage common-emitter topology with no input DC bias on-chip and achieved a noise figure (NF) of 1.10 dB with 8.6 dB of gain and only 5 mW of power consumption. The second LNA used a cascode topology with on-chip input biasing and DC blocking and achieved an NF of 1.48 dB with 13.9 dB gain and only 10 mW of power consumption. Both of these NF values are record for SiGe LNAs at this frequency, and the results are competitive with the best published III-V LNA results. These LNAs show that best-in-class SiGe technologies now offer competitive performance to the best III-V technologies and are worth consideration for use in exceptionally performance-demanding scientific applications. The next focus was the development of SiGe front-end circuits for small satellite-based G-band humidity sounding radiometers. Small satellites such as CubeSats are increasingly attractive platforms for hosting G-band radiometers due to their relative affordability and replaceability along with their potential to be deployed in distributed constellations of 15-20 sensors which would enable rapid high-resolution global observations. Manufacturing radiometers for such constellations poses a challenge, as InP and GaAs semiconductor technologies usually are not high-yielding and are limited to wafer sizes of 100 mm or smaller, making these chips expensive and not optimally suited for medium-volume production. Silicon based technologies, however, have low across-wafer performance variability and are readily accessible through multi-project wafer (MPW) services, unlike most submillimeter-wave InP and GaAs processes. The ability to fabricate small-sized chips with shared mask costs makes the MPW development model well suited for economically fabricating silicon-based radiometer components for these envisioned CubeSat constellations. Emerging SiGe BiCMOS technologies are increasingly compelling platforms for performanceconstrained millimeter-wave receivers. Commercial foundries currently offer SiGe HBTs with peak maximum oscillation frequencies (fMAX) of up to 500 GHz, and near-THz demonstrations of SiGe HBTs at room-temperature are expected soon. Standalone SiGe LNAs with measured noise figures as low as 5.8 dB at 140 GHz have been demonstrated, and numerous low-noise G-band SiGe receiver front-ends have been recently published. Furthermore, SiGe HBTs have excellent 1/f noise, inherent total ionizing dose radiation tolerance, and significant single-chip integration potential, making them well-suited for distributed networks of next-generation spaceborne radiometers. Our first G-band focus was to investigate the attainable NF of SiGe LNAs for this application. We designed two LNAs to achieve this goal. These amplifiers leveraged the same best-in-class HBTs used in the 18.7 GHz LNA designs along with careful consideration of the transistor core layouts and optimization of the noise figure tradeoff space for low-gain transistors to obtain unprecedented noise performance at this frequency for SiGe LNAs. The first amplifier used entirely cascode stages and achieved a peak gain of 24 dB at 158 GHz with a measured midband noise figure of 8.2 dB. The second amplifier, a six-stage common emitter design, achieved 17.2 dB of gain and a mean noise figure of 8.0 dB across 165-200 GHz while consuming only 16.1 mW of DC power. The measured gain, linearity, and power dissipation of these amplifiers are highly competitive with similar published G-band LNAs, although the NF is somewhat higher than the best demonstrated results using InP HEMT and GaAs mHEMT amplifiers (NF < 4 dB at 183 GHz). SiGe amplifiers cannot yet match the noise performance of GaAs and InP amplifiers, but it is expected that the performance gap will shrink as SiGe HBTs continue to advance toward THz frequencies. As SiGe HBTs continue to improve, the integration capabilities, 1/f noise, and manufacturing potential of SiGe technologies will warrant serious consideration for use in millimeter-wave scientific radiometers. In order to use SiGe LNAs in practical systems, however, packaged SiGe amplifiers must be demonstrated. Negar Ehsan at GSFC is leading an IRAD titled “MM-wave Radiometer Front End Development” under which they are packaging pHEMT LNA MMICs, and we are collaborating with her under this program to package the G-band SiGe LNAs. The common-emitter G-band LNA was modified for a larger chip size and different padframe to be compatible with the existing pHEMT LNA packages. These amplifiers have been recently (late July 2015) flip-chip bonded onto quartz interposer boards with integrated chip-to-waveguide transitions at Georgia Tech. Gold studs were manually placed onto each of the bondpads of the SiGe LNA using a ball bonder, and the chip was attached to the quartz board using ultrasonic flip-chip bonding. These chips have been shipped back to GSFC where they will be inserted into existing waveguide packages and characterized. If these packaging efforts are successful, these will be, to our knowledge, the first waveguide-packaged G-band SiGe LNAs to date. The primary obstacle in this packaging work is preventing cavity resonances in the waveguide package, which may be difficult to suppress due to the relatively large size of the LNA chip along with the relatively high dielectric constant of the 300 um thick silicon chip. This work is ongoing, so the conclusions of this work are not yet known. The next step in this G-band SiGe radiometer development was to design the first known singlechip down converter for G-band radiometers to benchmark versus existing multi-chip radiometer receivers. Multi-chip radiometer receivers are best-suited for use in one-off instruments, but pose challenges for producing many receivers for CubeSat constellations. Multi-chip modules require many millimeter-wave packaging transitions, which tend to be lossy and are difficult to identically repeat. Furthermore, the InP and GaAs components typically used in G-band radiometers are not high-yielding and do not offer optimal economy-of-scale benefits. InP processes in particular typically have poor fabrication properties and high performance variation, which makes InP circuits very expensive and poorly suited for mass production. An integrated SiGe radiometer would be capable of rapid assembly and will have improved SWaP versus existing millimeter-wave radiometers. Integrating all receiver blocks on a single chip significantly reduces chip-to-chip transition losses, simplifies module assembly, and maximizes the repeatability of the packaging—all highly beneficial characteristics for mass-production of millimeter-wave receivers for CubeSat constellations. The G-band SiGe downconverter developed under this work is targeted for compatibility with the Hyperspectral Microwave Atmospheric Sounder (HyMAS) under development by MIT/LL and GSFC. This application uses an RF bandwidth of 172-184 GHz, a fixed local oscillator (LO) frequency of 77 GHz, and an IF bandwidth of 18-30 GHz to enable hyperspectral microwave atmospheric sounding. The developed downconverter consists of an optimized G-band SiGe LNA which contains image-reject filtering, a double balanced active Gilbert cell mixer, and a 77154 GHz frequency doubler for the LO to emulate subharmonic mixer operation. The integrated subsystem has a simulated peak conversion gain of 16 dB with a 3-dB bandwidth of 162-192 GHz and a noise figure of less than 8.7 dB across the design bandwidth. The optimal simulated LO power is 0 dBm, the input-referred P1dB is -35 dBm, and the DC power dissipation at peak drive is only 85 mW. This hardware has been fabricated and returned to Georgia Tech in early August 2015, and will be measured over the next two months along with all subcircuit breakouts. It is expected that these circuits will shift somewhat from simulation, but will achieve excellent performance and show the potential of integrated SiGe downconverters for use in next-generation constellations of Gband radiometers for CubeSats. The last major piece of this research is the development of an IF chipset for this HyMAS application. MIT/LL, under an ESTO ACT award, is developing a 52-channel hyperspectral microwave receiver module to channelize 18-30 GHz IF data from G-band radiometer front-ends and promises significant improvements in atmospheric humidity profiling. Each channel contains reflective LTCC filters, and many very low-power and high-isolation amplifiers are required to provide buffering between the channels and implement the 52-way power division network effectively. Additionally, the detectors are currently implemented manually using discrete diodes. This module would greatly benefit from the low power SiGe LNAs and high-responsivity monolithic detectors which we have designed under this research. The LNAs were designed for all-around performance with minimal power consumption across the 18-30 GHz bandwidth. The amplifiers were implemented in IBM’s BiCMOS8HP technology, a 0.13um process which offers SiGe HBTs with peak fT and peak fMAX of 200 and 285 GHz, respectively. The SiGe LNA design achieved a simulated peak gain of 19 dB at 23 GHz with a 17-30.5 GHz bandwidth. The noise figure is less than 3.7 dB across the band, the output-referred one-dB compression point is better than 2.5 dBm, and the power consumption is only 33 mW. The detector was designed using a common-emitter transistor biased in Class B for low 1/f noise, and a large transistor was used to achieve the 18-30 GHz bandwidth. A reference voltage output is included. The simulated responsivity is 10 kV/W, the noise equivalent power is less than 3 pW/sqrt(Hz), and the 1/f noise corner is less than 1 kHz. An integrated LNA-detector chipset was also designed, which achieves a simulated noise equivalent power of less than 50 fW/sqrt(Hz) with over 1 MV/W responsivity and a similar 1/f noise corner to the detector. The size of the integrated chipset is only 1.87 mm by 0.85 mm, which would enable significant miniaturization of MIT/LL’s hyperspectral receiver module. It is expected this LNA, detector, and integrated chipset will achieve excellent performance and demonstrate the ability of custom-fabricated SiGe circuits to significantly improve the implementation and enhance the performance of demanding hyperspectral receiver modules. Two additional studies relating to this research have been identified as potential topics of study, if time permits, prior to my graduation next summer. First, we may package the 18-30 GHz detector and integrated LNA-detector chipset and investigate their response to transient single-event radiation effects (SEE). SEE are a major concern in spaceborne systems, and the response of RF SiGe circuits to SEE events has only begin to be explored. These circuits are well-suited for this experimentation, as the detector output is a baseband signal and transient signals can be easily measured at the output using a high-speed oscilloscope. This would fit well with my research path and would potentially shed light on best practices for designing SEE-resistant radiometer IF backends. The final study would be completing an investigation into the options for implementing G-band avalanche noise sources using the standard device offerings in the 500 GHz IHP SG13G2 SiGe BiCMOS technology. To my knowledge, no monolithic diode-based noise sources have been demonstrated—all existing radiometers use discrete diodes for simplicity and flexibility. Test structures for numerous potential noise source elements consisting of HBT emitter-base junctions, collector-base junctions, open-base HBTs, and Schottky diode variants have been fabricated and are awaiting characterization. Monolithic SiGe noise sources implemented using standard SiGe BiCMOS process offerings would enable significantly enhanced functionality and make the possibility of a future integrated G-band radiometer system even more compelling.