Modeling And Parameter Extraction For Milliltre-Wave Passive Decives Based On Heterogeneous Integration Process

Millimetre-wave heterogeneous integration technology integrates highperformance Ga As circuits and In P circuits with silicon-based circuits by heterogeneous interconnection and assembly.This technology breaks through the limitations of a single semiconductor material,offering an important direction for integrated circuit development called “More than Moore”.The photosensitive composite film multilayer interconnection process,which is based on silicon MEMS technology,is a popular three-dimensional heterogeneous integration solution in the industry.The process employs a photosensitive BCB photoresist as a dielectric material and a high-resistance silicon as a substrate,which is implemented by MEMS fabrication technology.BCB features a low cure temperature,low dielectric loss,and good compatibility with high-resistance silicon.Device modelling provides an essential connection between the IC designer and foundry engineers.Accurate millimetre-wave passive device models can shorten the development cycle and reduce labour costs as well as time costs.Passive devices are commonly used in millimetre-wave heterogeneous integrated circuits as inductors,filters,transformers,and power splitters.They undertake complex tasks such as transmission,filtering,impedance matching,and power synthesis.However,in the millimetre-wave band,the parasitic effects,coupling effects and dispersion effects of passive devices become increasingly complex,which makes it more difficult to establish high-precision wideband models.At present,few process manufacturers have provided models for inductors and transformers in the millimetre-wave frequency band.There are few models for heterogeneous integration passive devices.Modelling passive devices for a photosensitive composite film multilayer interconnection process in the millimetre-wave frequency range is thus important in engineering and academic practice.In this paper,the modelling and parameter extraction methods of on-chip passive components based on 3D heterogeneous integration technology are studied.The innovative research achievements are as follows:1)The effect of asymmetric error on the de-embedding precision of two-port onchip devices is analysed,and a de-embedding algorithm based on an asymmetrical equivalent circuit structure is proposed.Cross-finger capacitance and asymmetric transmission lines are fabricated by using photosensitive composite film multilayer wiring technology,and the effectiveness of the de-embedding algorithm is verified.2)A wideband CPW transmission line model based on the theory of fractional calculus is proposed in this paper.The fractional-order model consists of two fractionalorder elements and five integer-order elements,which are capable of accurately describing the behaviour of a T-line in heterogeneous integration circuits.The working bandwidth of the model is increased from 40 GHz for the original integer-order model to 110 GHz.The parameters of the fractional element values are extracted by a direct method.Not only in the frequency domain,but also in the time domain,The proposed model is verified as well as compared with the classical model.3)A novel fractional-order 1-? inductor equivalent circuit model is proposed to characterize chip inductors based on a photosensitive composite film multilayer interconnection process.The proposed model,which contains 6 integer-order circuit elements and 3 fractional-order elements,can describe frequency-dependent effects,including the skin effect,proximity effect and distributed effect.The three fractionalorder elements introduced in the above model are major contributors to improving the model's accuracy and bandwidth.A parameter extraction method with an accuracy of up to 40 GHz is described.4)A millimetre-wave slow-wave transmission line(T-line)implemented with photosensitive composite film multilayer interconnection process technology is proposed and modelled by an artificial neural network(ANN)method.An ANN is applied to map the numerical relationship between the geometric parameters of the slow-wave T-line and its S-parameters.The iterative process is optimized by using a 3-layer perceptron neural network structure and the LM-GDBP algorithm.The proposed model is applied to design a slow-wave coupler.Good agreement among the ANN results,electromagnetic(EM)-simulated results and measured results of the coupler is obtained in a frequency range from 20 GHz to 40 GHz.