G. Kompa

A 6W Uneven Doherty Power Amplifier in GaN Technology

In this paper the design of a 6W uneven GaN Doherty power amplifier is presented. The Doherty PA is designed to achieve high efficiency for modulated signals with high peak to average power ratio used in modern wireless communication systems. The Doherty amplifier has been designed using two equal sized GaN devices for the main class AB and peaking class C amplifiers. An uneven power divider is used at the input to deliver more input power to the peaking amplifier than the main amplifier. The measured maximum output power of the realised uneven Doherty is 38 dBm with 60% of peak power added efficiency (76% of drain efficiency). The power added (drain) efficiency is higher than 52% (62%) up to 6 dB of back off, or 42% (45%) up to 10 dB of back off.

Design and stability test of a 2-40 GHz frequency doubler with active balun

An ultra-broadband PHEMT MMIC frequency doubler is designed. It comprises a FET active balun stage and a push-pull doubler in common-source configuration. Due to the purely lumped-element concept, the required chip area is very low. Using broadband compensation and matching networks, a nearly flat output power performance up to a limit of 2/3 f/sub T/ with acceptable conversion loss is predicted in circuit simulations and confirmed with broadband waveform measurements. The stability of the entire circuit is investigated including hidden instabilities and arbitrary passive input and output load conditions based on a direct stability circle calculation.

A 40GHz large-signal double-reflectometer waveform measurement system designed for load-pull applications

A new large-signal measurement system combining waveform and load-pull characterization is presented. The setup comprises a double-reflectometer testset and a microwave transition analyzer (HP 71500A) as fundamental and harmonic receiver. The system calibration can be of TRL or modified SOLT type. For the first time, this setup permits a complete description of the harmonic amplitude and phase spectra under arbitrary loading conditions of the DUT. As an application example, an AlGaAs HEMT was. characterized for different reflective fundamental frequency output loads.

Quasi-monolithic integration of high-power GaN-based HEMTs for high-frequency applications

We report on the quasi-monolithic integration of unpackaged high-power AlGaN/GaN HEMTs into a silicon basis, which finally will be the backbone and main part of an integrated circuit. III/V compound semiconductor devices fulfil special tasks which Si devices are unable to perform. For demonstration purposes we embedded high-power nitride semiconductor HEMTs in silicon substrates with planar electrical contacts using micromachining techniques. This enables integration of both unpackaged devices of different materials (e.g. high-power nitride semiconductor HEMTs) and devices made by thin film technology (e.g. resistors, inductors, etc), leading to a system-on-package on a cost-effective silicon substrate.

Generalized Modeling of GaAs Mesfets and Modfets Based on Highly Accurate Broadband Measurements

A recently defined THLR method for vector network analyzer calibration is used to extract the error-parameters of coax to microstrip transitions with an accuracy hitherto unattained. The scattering coefficients of GaAs FETs such as MESFETs and MODFETs, installed in hybrid technology, are measured and error-corrected up to 40 GHz. A new parameter extraction method is discussed to get reliable bias-dependent model parameters. These are represented in 3D graphs covering the gate-forward, pinch-off, saturation, and non-saturation region as well. Thus analogue and digital operations can be described.

Volterra-Series-Based Distortion Analysis for Optimization of Out-of-Band Terminations in GaN HEMT Devices

This letter focuses on the critical device-level linearity issues resulting from out-of-band terminations for reliable distortion characterization in future Universal Mobile Telecommunications System-Long Term Evolution (UMTS-LTE). Using Volterra series technique, the key distortion sources arising from the envelope and harmonic components in 0.5-mm GaN HEMT were identified using commercial and in-house bias tees. With the designed in-house bias tee, the baseband performance, in comparison with the commercial bias tee, is tested through drain-bias sensing. In reference to the commercial bias tee, up to 99.3% reduction in drain modulation is achieved using the in-house bias tee. Memory-effect characterization of GaN HEMT exemplified the implications of baseband and second-harmonic load terminations, which was theoretically confirmed through Volterra series technique. Using the in-house bias tee, under two-carrier wideband code-division multiple-access excitation, up to 47-dBc 3rd-order intermodulation ratio (IMR3) is achieved at 13.5-dB backoff. This has resulted in a 5-dB IMR suppression together with the minimization of intermodulation-distortion asymmetry, confirming the possibility to achieve the 3rd Generation Partnership Project linearity specification at the device level.

Investigation of IMD3 in GaN HEMT based on extended volterra series analysis

This paper mainly focuses on providing theoretical justification for possible GaN device linearity improvement, interpretating key physical origins of IMD3. Based on bias dependent S-parameter measurement data of field-plate-free 8x125 mum GaN HEMT, IMD3 is modelled using classical Volterra series theory. Device diagnosis is hence carried out, by means of this technique, for efficiently localizing the distortion behaviour. Further, device linearity is shown to improve by appropriately tuning gate-drain feedback capacitance by taking advantage of field-plate technology proving the analysis to be a powerful tool for developing GaN HEMT technology. Further, with the intension of understanding IMD nulling, Volterra analysis is extended to 5th-degrce nonlinearity through which an insight into the distortion cancellation mechanism is obtained.

A systematic approach to a reliable neural model for pHEMT using different numbers of training data

A systematic approach is presented to achieve a reliable neural model for microwave active devices with different numbers of training data. The method is implemented for a small-signal bias depended modeling of pHEMT with different numbers of training data. The errors for different numbers of training data have been compared to each other and show that by using this method a reliable model is achievable even though the number of training data is considerably small. The method aims at constructing a model which can satisfy the criteria of minimum training error, maximum smoothness (to avoid the problem of overfitting), and simplest network structure.

Heat transfer and thermal stress analysis in the new generation quasi-monolithic integration technology (QMIT)

Static heat transfer and thermal stress analysis for the new generation quasi-monolithic integration technology (QMIT) have been performed using a three-dimensional finite element simulator. To confirm the simulation results, white-light interferometry measurement along with a Peltier element and a Pt-temperature sensor have been used. It has been shown that thermal resistances of 11/spl deg/C/W and 8.5/spl deg/C/W are possible using 200 /spl mu/m electroplated gold heat-spreader and diamond-filled polyimide on the backside of the active device, respectively. This promises successful realization of the high frequency circuits containing power active devices using the novel QMIT. Simulation and measurement results demonstrate a great decrease of thermal stress in the new generation QMIT in comparison to the earlier concept which extremely improves life-time of the packaging. A remarkable agreement between calculated and measured results was found.

Optimization of broadband drain modulation in GaN HEMT devices

This paper focuses on the optimization of broadband envelope termination for reliable device characterization for future UMTS-LTE systems. Drain bias modulation is suppressed in 8times500 mum GaN HEMT using compensation network overcoming the parasitic effects of the DC feed, generally ignored during broadband measurements. Drain modulation index (DMI) metric is defined for quantifying the extent of drain voltage modulation as a function of carrier spacing and up to 80% reduction in DMI is achieved after drain bias compensation. Further, based on the experimental analysis, a simple bias tee is designed which can overcome improper DC feed conditions. Under 2-carrier CW and W-CDMA excitation, maximum of 19 dB and 18 dB IMD3 suppression together with the minimization of IMD asymmetry, were achieved using in- house bias tee, emphasizing the importance of broadband bias network design.