| Transmission lines are developed into an irreplaceable role in modern design for the achievement in impedance matching, filtering and interconnect. However, exceeding loss is introduced by higher operating frequency of millimeter wave, which makes the quality factor of the ordinary structures cannot meet the engineering need. Therefore, we need more superior structures to obtain better performance. Moreover, as the result of the CMOS process metal density demands as well as the characteristic impedance of the transmission lines, the design of millimeter wave transmission lines in CMOS technology is much more complicated, which makes it more difficult to analysis the performance of the millimeter wave transmission lines. This is why we need a directed and measurable way to assess the transmission lines. Besides, the design of millimeter wave transmission lines in CMOS technology costs too much time and memory, which can be resolved by proposing a model to lead the optimization direction.As for the above questions, this article proposed four structures of transmission lines to start with. They are designed in CMOS technology aiming at lower loss and higher quality factor. Second, distributed parameters are obtained by parameter extracted method are used to analysis the performance of the transmission lines. Third, in this article, we use two kind of neural network to model the relationship between the dimensions and the distributed parameters. At last, the process of the millimeter wave transmission lines with high performance is launched, after which we get the test results, compared with the theoretical analysis, embedded the data, and finally comes to the conclusion that the theory in this article is verified.The concrete achievements are as follows:First, in this paper, design the structure of four good performance of transmission lines are proposed, which are standard CPW, GCPW, SCPW, and Finger-SCPW. Then, distributed parameters are obtained by parameter extracted method are used to analysis the performance of the transmission lines(a) Standard CPW proposed in this article obtains the superior performance of the quality factor of12.84, with the standard CPW structure quality factor is usually5in the literature research. Federico put forward a new simulation method of CPW in which acquired a quality factor of12, which is the most superior performance at that time.(b) GCPW loss and the CPW is almost no difference, but quality factor of the CPW increased by65%.(c) SCPW structure design, however, get a larger cross section in order to satisfy the characteristic impedance and the principle of CMOS design metal density. SCPW’s cross-sectional area is twice as much as that of CPW, resulting in increasing loss. However, due to the existence of the periodic floating metal of the slow wave structure, quality factor Q compared with CPW also has a32%improvement.(d) FCPW, similar to SCPW, sectional area increase leads to increasing loss, due to the superiority of FCPW structures, even if the cross-sectional area is9.5times that of CPW, the quality factor Q still get a10%improvement.Second, this article uses BP Neural Network and Elman Neural Network to establish the model, to form the relationship between the sizes of the transmission line with distributed parameter RLGC, giving a direct relationship of the equivalent circuit parameters of the transmission lines, which provided reliable prediction model in the following-up study.Third, the optimized structures of millimeter wave transmission lines are processed. After the embedded simulation results will be compared with the measured results, the simulation results are consistent with measurement results, which verified the theory of this article. |
PDF Full Text Request