Showing papers by "Bart Nauwelaers published in 2004"

EVM Calculation for Broadband Modulated Signals

Abstract: We present a normalization that facilitates calculation of error vector magnitude (EVM) from measurements. We derive the definition of EVM for a common industry standard from a more basic equation. We compare EVM for various modulation types for a given average symbol power under simple distortion conditions.

Distributed dielectric charging and its impact on RF MEMS devices

Abstract: This paper gives a new insight into the problem of RF MEMS irreversible stiction due to dielectric charging. Previous reported works describing the phenomenon only account for a drift of the actuation characteristics as a whole as they consider uniform charge densities. We demonstrate how the spatial charge distribution in the dielectric layer can result in the failure of the devices. We emphasize the role of the variance of the distribution, a parameter neglected in the literature. Our model can account for a shift of the C-V actuation characteristics but also for a change in its profile. In particular, the pull-out window can be narrowed and even made to disappear as a result of the non-zero variance of the charge distribution. We identify the processing, the contact conditions and the distributed charge depths as variance- and thus failure-enhancer parameters.

Comparison of the effect of gate dielectric layer on 2DEG carrier concentration in strained AlGaN/GaN heterostructure

Abstract: The effect of surface passivation on undoped AlGaN/GaN heterostructures using SiO2, Al2O3, Ta2O5 and Si3N4 as a function of layer thickness is presented. It is found that all of the oxides caused decreased 2DEG carrier concentration with increasing thickness of the respective oxide layers between the gate and AlGaN layer. On the contrary, the 2DEG carrier concentration increased strongly with increasing Si3N4 layer thickness. An elementary polarization model was used to fit the behavior for all materials and thicknesses leading to quantitative results. The fitting suggests that the effect of the oxides and Si3N4 on the 2DEG carrier concentration can be explained by the differences between them with respect to charge accumulation at the AlGaN/dielectric interface. High temperature in-situ deposited Si3N4 especially shows interesting behavior by bowing measurements as it also adds strain which increases piezoelectric polarization charge in AlGaN layer, so that increases the charge density in the 2DEG.

RF-power: driver for electrostatic RF-MEMS devices

Abstract: The RF-power handling capability is an important characteristic for RF-MEMS switching devices. Apart from excessive heat dissipation, the power handling capability is mainly limited by the so-called self-biasing and/or RF-latching. These two phenomena result from the fact that the available RF-power from the source induces a non-zero electrostatic pulling force on the suspended structure. So far, self-biasing of RF-MEMS switches has always been studied assuming a perfect match of the device to the network in the ON-state (i.e. no reflection) and thus a fixed dc-equivalent rms voltage on the capacitor. If the RF-power exceeds a critical value, pull-in or self-biasing occurs. In practice, however, the assumption of the perfect match is not correct as the switch capacitance increases with increasing RF-power. This will cause a change in the reflected signal and thus a decrease in the dc-equivalent voltage source. This paper gives a new insight into the RF-power handling of RF-MEMS shunt switches and, per extension, of RF-MEMS shunt tunable capacitors. We analytically show how the negative feedback on the electrostatic force introduced by the capacitive mismatch changes the pull-in characteristics of the structure and can even stabilize it, totally avoiding the pull-in phenomenon.

Modelling of the RF self-actuation of electrostatic RF-MEMS devices

Abstract: This paper presents a study of the self-biasing of RF-MEMS shunt switching devices, and tunable capacitors per extension, taking into account the negative feedback on the electrostatic force introduced by the mismatch. The influence of the RF-power on the DC-actuation is included in the normalised analytical formulas developed for any shunt mismatch (capacitive or not) involving a moving armature. We demonstrate the possibility to overcome the pull-in instability by using the power-actuation and the importance of the mismatch to accurately predict the actuation characteristics. Our model can account for parabolic and close-to-linear dependence between DC-bias and RF-power at pull-in.

Nonlinear models of microwave power devices and circuits

Abstract: This paper presents an overview of activities on RF non-linear modeling of power devices and amplifiers throughout Europe.

Low loss coupled line filters with transmission zeros in multi-layer thin film MCM-D technology

Abstract: This paper presents the design and characterization of high-performance bandpass filters with transmission zeros, integrated in a multi-layer thin film MCM-D technology on high-resistivity Si. A mixed thin-film microstrip/coplanar waveguide transmission line has been used for the realization of the filters, as it presents lower loss than thin-film microstrip lines. The coupled line filters achieve an insertion loss of 2.3 dB at 25 GHz and 1.7 dB at 43 GHz with 3 dB bandwidths of 14% and 18% respectively. The use of transmission zeros provides a close in-band attenuation of more than 30 dB. The used topology allows changing the center frequency and the position of the transmission zeros more or less independently facilitating the design.

Accurate RF electrical characterization of CSPs using MCM-D thin film technology

Abstract: The thin film multilayer multichip module-deposited (MCM-D) technology of IMEC is used for characterising the RF electrical performance of two types of chip scale packages (CSPs). The measurement technique called MCM-on-package-on-MCM (MoPoM) enables accurate measurements and de-embedding in the gigahertz (GHz) range of frequencies. Wafer processing of the MCM-D technology allows for several design structures to be integrated on a single mask. The packages chosen are a 120-pin plastic ball grid array (PBGA) and an 80-pin polymer stud grid array (PSGA). Lumped element models extracted from measurements and three-dimensional simulations show good agreement with the measurements up to 6 GHz for the BGA and the PSGA. The electrical performance of the packages is compared at 1.8 GHz (GSM), 2.4 GHz (Bluetooth), and 5.2 GHz (HiperLAN) and at 5.2 GHz both the packages exhibit a return loss of lower than -10 dB and hence cannot be used in most cases without design improvement. We also show that the influence of encapsulant is significant while transmission line detuning due to the package is not significant at microwave frequencies. We also briefly mention about the crosstalk effects. We demonstrate the significant degradation in the performance of a 5.2 GHz MCM-D low noise amplifier (LNA) after packaging. A significant improvement in package performance is observed by conjugate matching the package interconnects.

Extending on-die wiring hierarchy with wafer level packaging concepts

Abstract: Wafer level package (WLP) redistribution layer interconnects can be effectively used for countering on-die global wiring challenges. To demonstrate this, we fabricated WLP test chips with two configurations namely IMPS and microstrip transmission lines in WLP layers. Experimental results indicate superior electrical performance. Near speed of light propagation is observed with insignificant signal distortion. We compare the WLP layers interconnect performance with a representative 70nm node upsized global interconnect. Details on the WLP technology are also provided.

Compact broadband resistance model for microstrip transmission lines

Abstract: This work presents a closed form broadband resistance model for microstrip transmission lines. The proposed model is compact and scalable, and is within 7% of experimental values. The model uses a non-discontinuous analytical function valid in all bands ranging from DC to microwave frequencies.

A surface micromachined tunable film bulk acoustic resonator

Abstract: This paper reports on the design, modeling, fabrication and measurement of a novel-structured film bulk acoustic resonator (FBAR) that allows frequency tuning by MEMS actuation. FBAR's are micromachined frequency controlling devices working in RF regime. For many applications, a small range of tuning is desired to cope with drifts from different origins. To realize this functionality, it has been suggested to place tunable elements such as variable capacitors or inductors in the circuit. A conventional approach, in which an external element is used, would introduce parasitics and might seriously degrade the quality factor of the system. In contrast, our work integrates the piezoelectric resonating film and the tuning element to build a compact structure. By reducing possible parasitics and electrical resistance, this structure enables frequency tuning while maintaining a high quality factor.

Influence of RF measurement uncertainties on model uncertainties: practical case of a SiGe HBT

Abstract: Measurement based models, like small-signal equivalent circuit models of microwave transistors, are often extracted under the assumption of having perfect measurements. In this work we study how the uncertainties of S-parameter measurements affect the uncertainties of model element values in the practical case of a SiGe HBT.

A New Optimisation-Based Methodology for Determination of Small-Signal Equivalent Circuit Model Parameters for a Si/SiGe HBT Process

Abstract: An accurate and simple method for determining the small-signal equivalent circuit model parameters of Si/SiGe heterojunction bipolar transistors is presented. This method differs from the previous ones in that it overcomes the often less accurate parasitic elements’ extraction by finding out the proper equivalent circuits for the dummy structures, used to assess the access interconnects’ contribution, and incorporate those in the complete circuit of the model. The method is based on optimization and requires the S-parameter measurements of the dummy structure and the active device. It is shown that it can be easily implemented in commercial software. To validate our methodology, we determined the model elements of a Si/SiGe HBT in the active region. Excellent results were obtained up to 20 GHz. I. Introduction A small-signal equivalent circuit model is a physics based model. This means that it reflects the important physical characteristics of the device. There are two main methods to extract the model parameters. The first way is a direct parameter extraction method [1-3]. These methods suffer from the problem that the extracted element values may be frequency dependent [4], and therefore often need an additional optimization step. Also, the intrinsic part is virtually limited to eight elements, as we have to determine all values from four complex S-parameters. This implies for example that the distributed effect of the base can not be represented. The second method to determine equivalent scheme model elements is optimization, meaning that numerical search algorithms are used to minimize the error between the measured and some simulated error function. The drawback of this technique is that it can yield non-physical element values. In this work, we developed an optimizationbased methodology. Separating the optimisation of the extrinsic element values from the intrinsic part reduces the potential problem of non-physical values. Note that in case of HBT’s, the majority of the modelling papers are focused on GaAs or InP based HBT’s. Due to the semi-insulating substrate, the determination of their parasitic element values is simple and straightforward. As this paper concentrates on Si/SiGe HBTs, special attention to the extrinsic network is required. This paper is organized as follows: firstly, we determine the extrinsic parasitic network using the Sparameter measurements of dummy structures Next, we optimize the intrinsic elements while fixing the extrinsic element values, obtained in the previous step. Experimental results are presented in Section III. Finally, conclusions are drawn in section IV.

Nonlinear Modeling of Si/Si Ge HBT Using ANN

Abstract: We present a large signal model for Si/SiGe HBTs using an Artificial Neural Network (ANN). The ANN is used to model the DC non-linearities of the intrinsic device. In this way, physical phenomena such as nonideal leakage currents and the Kirk effect can be modeled without time-consuming extraction. Capacitive nonlinearities are modeled by the well-known relationship between the capacitance and the junction voltage, ignoring the diffusion capacitance. By comparing ANN model results to measurements, we show that a good agreement for DC and nonlinear characteristics is obtained.

A physics-based VLSI interconnect model including substrate and conductor skin effects

Abstract: The goal of this work was to model the influence of the substrate and conductor skin effects on the RL line characteristics of on-chip interconnects. The proposed modelling approach is based on the skin-effect theory results for rectangular wires and the complex integration technique for perturbation series per unit length impedance of silicon substrate. It is shown that the calculated frequency-dependent distributed inductance and the associated resistance are in good agreement with the results obtained from rigorous full-wave solutions and CAD-oriented analytic modelling technique.

Large-signal time-domain behavioral model of a packaged device for high efficient wireless circuit design

Abstract: This work presents fur the first time a largesignal time-domain behavioral model of a packaged pseudomorphic High Electron Mobility Transistor (pHEMT) for 5 GHz Wireless Local Area Network (WLAN) applications. The methodology together with some important issues concerning high efnciencylhigh frequency wireless applications is discussed. The model's accuracy is evaluated using large-signal measurements.

High frequency modelling approach of on‐chip interconnects considering conductor and substrate skin effects

Abstract: Simple and accurate high frequency modelling approach of on‐chip interconnects on a lossy silicon substrate, that considers conductor and substrate skin effects, is presented. The closed‐form formulas for the frequency‐dependent series impedance parameters are obtained using a closed‐form integration method and the vector magnetic potential equation. The proposed frequency‐dependent inductance L(f) and resistance R(f) per unit length formulas are shown to be in good agreement with the electromagnetic solutions.

38.5 GHz MCM-D phased array antenna with electronically configurable bearnwidth and sidelobe level

Abstract: The proposed active antenna, to be realized in the MCM-D technology of IMEC, consists of 2 concentric subarrays of CPW-fed aperture coupled microstrip patches. By feeding the two suharrays with different amplitude and phase the beamwidth and sidelobe level of the antenna can be controlled. Simulations show that the beamwidth can he controlled between 55 and 140 degrees and the sidelobe level between -10 and -25 dB. The phase and amplitude control is done at 2.4 GHz using a QAM-like approach without the use of phase shifters. The applications targeted are point-to point radio links at 38.5 GHz and indoor broadband wireless networks.

Artificial neural network to statistically model the variation in small signal equivalent circuit model parameters for a Si/SiGe HBT process

Abstract: On a single wafer and between different wafers, there are variations in device characteristics due to processing non-uniformity and non-reproducibility. In order to capture these variations, we could extract a complete equivalent circuit model for each device. Since this is a timeconsuming process, we developed an alternative approach based on an Artificial Neural Network (ANN). This ANN takes measured quantities as inputs, and generates the model parameters as outputs. To keep the complexity of the ANN down to a reasonable size, we limited this mapping to the most sensitive elements. These are determined by performing a sensitivity analysis on a reference device. To demonstrate the method, we applied it to a Si/SiCe HBT process. Results show that the ANN very well predicts model parameters and that it is B very good statistical model. Index Terms De-embedding, small-signal modeling, artificial neural networks, Statistical Database, SiCe HBT.