Abstract:A practical formulation for EM‐based design optimization of high‐frequency circuits using simple polynomial surrogate functionals is proposed in this paper. Our approach starts from a careful selection of design variables and is based on a closed‐form formulation that yields global optimal values for the surrogate model weighting factors, avoiding a large set of expensive EM model data, and resulting in accurate low‐order low‐dimension polynomials interpolants that are used as vehicles for efficient design opt… Show more
“…A similar approach to develop polynomial surrogate models is proposed in [32]; however, our new approach differs in three aspects: the surrogate model formulation, the calculation of weighting factors, and the surrogate order determination. For the surrogate model formulation, [32] implements the Nth order surrogate model by using an element-wise power operator, which creates some redundant terms, while our new formulation exploits the multinomial theorem, allowing us to expand a polynomial raised to an arbitrary power including all cross terms and no redundant terms. For the weighting factors calculation, the approach in [32] calculates simultaneously all weighting factors available for each surrogate model order.…”
Section: Introductionmentioning
confidence: 99%
“…For the surrogate model formulation, [32] implements the Nth order surrogate model by using an element-wise power operator, which creates some redundant terms, while our new formulation exploits the multinomial theorem, allowing us to expand a polynomial raised to an arbitrary power including all cross terms and no redundant terms. For the weighting factors calculation, the approach in [32] calculates simultaneously all weighting factors available for each surrogate model order. In contrast, the proposed approach in this work automatically calculates the weighting factors by assuming that lower-order surrogates are fixed or by calculating all weighting factors simultaneously, and the selection between both manners is based on the conditional number of the system matrix.…”
Section: Introductionmentioning
confidence: 99%
“…In contrast, the proposed approach in this work automatically calculates the weighting factors by assuming that lower-order surrogates are fixed or by calculating all weighting factors simultaneously, and the selection between both manners is based on the conditional number of the system matrix. Finally, the order of the surrogate model can be different for each simulated frequency point, while in [32] and [33] the same surrogate model order is used for all simulated frequency points.…”
A general formulation to develop EM-based polynomial surrogate models in frequency domain utilizing the multinomial theorem is presented in this paper. Our approach is especially suitable when the number of learning samples is very limited and no physics-based coarse model is available. We compare our methodology against other four surrogate modeling techniques: response surface modeling, support vector machines, generalized regression neural networks, and Kriging. Results confirm that our modeling approach has the best performance among these techniques when using a very small amount of learning base points on relatively small modeling regions. We illustrate our technique by developing a surrogate model for an SIW interconnect with transitions to microstrip lines, a dual band T-slot PIFA handset antenna, and a high-speed package interconnect. Examples are simulated on a commercially available 3D FEM simulator.
“…A similar approach to develop polynomial surrogate models is proposed in [32]; however, our new approach differs in three aspects: the surrogate model formulation, the calculation of weighting factors, and the surrogate order determination. For the surrogate model formulation, [32] implements the Nth order surrogate model by using an element-wise power operator, which creates some redundant terms, while our new formulation exploits the multinomial theorem, allowing us to expand a polynomial raised to an arbitrary power including all cross terms and no redundant terms. For the weighting factors calculation, the approach in [32] calculates simultaneously all weighting factors available for each surrogate model order.…”
Section: Introductionmentioning
confidence: 99%
“…For the surrogate model formulation, [32] implements the Nth order surrogate model by using an element-wise power operator, which creates some redundant terms, while our new formulation exploits the multinomial theorem, allowing us to expand a polynomial raised to an arbitrary power including all cross terms and no redundant terms. For the weighting factors calculation, the approach in [32] calculates simultaneously all weighting factors available for each surrogate model order. In contrast, the proposed approach in this work automatically calculates the weighting factors by assuming that lower-order surrogates are fixed or by calculating all weighting factors simultaneously, and the selection between both manners is based on the conditional number of the system matrix.…”
Section: Introductionmentioning
confidence: 99%
“…In contrast, the proposed approach in this work automatically calculates the weighting factors by assuming that lower-order surrogates are fixed or by calculating all weighting factors simultaneously, and the selection between both manners is based on the conditional number of the system matrix. Finally, the order of the surrogate model can be different for each simulated frequency point, while in [32] and [33] the same surrogate model order is used for all simulated frequency points.…”
A general formulation to develop EM-based polynomial surrogate models in frequency domain utilizing the multinomial theorem is presented in this paper. Our approach is especially suitable when the number of learning samples is very limited and no physics-based coarse model is available. We compare our methodology against other four surrogate modeling techniques: response surface modeling, support vector machines, generalized regression neural networks, and Kriging. Results confirm that our modeling approach has the best performance among these techniques when using a very small amount of learning base points on relatively small modeling regions. We illustrate our technique by developing a surrogate model for an SIW interconnect with transitions to microstrip lines, a dual band T-slot PIFA handset antenna, and a high-speed package interconnect. Examples are simulated on a commercially available 3D FEM simulator.
“…Additionally, the required time to evaluate and even to train any metamodel becomes, for practical purposes, insignificant as compared to the time required to collect the measurement data. On the other hand, it has been demonstrated [13], [14] that both ANN and polynomial functional surrogates perform better than SVM and Kriging surrogates in cases with a very limited amount of training data, while polynomial surrogates exhibit better performance than ANN only in cases with low-dimensionality and small regions of interest, Then, we propose a neural modeling approach to efficiently approximate the effects of a HSIO post-silicon receiver equalizer with a very reduced set of testing and training data, and possibly a large number of knobs. The resultant metamodel, obtained from the proposed inexpensive method, could later be used as a fast coarse model in a space mapping approach [7], [8] to find the optimal equalizer settings that maximize the actual HSIO performance.…”
Section: Machine Learning In Post-silicon Validationmentioning
As microprocessor design scales to the 10 nm technology and beyond, traditional pre-and post-silicon validation techniques are unsuitable to get a full system functional coverage. Physical complexity and extreme technology process variations severely limits the effectiveness and reliability of presilicon validation techniques. This scenario imposes the need of sophisticated post-silicon validation approaches to consider complex electromagnetic phenomena and large manufacturing fluctuations observed in actual physical platforms. One of the major challenges in electrical validation of high-speed input/output (HSIO) links in modern computer platforms lies in the physical layer (PHY) tuning process, where equalization techniques are used to cancel undesired effects induced by the channels. Current industrial practices for PHY tuning in HSIO links are very time consuming since they require massive lab measurements. An alternative is to use machine learning techniques to model the PHY, and then perform equalization using the resultant surrogate model. In this paper, a metamodeling approach based on neural networks is proposed to efficiently simulate the effects of a receiver equalizer PHY tuning settings. We use several design of experiments techniques to find a neural model capable of approximating the real system behavior without requiring a large amount of actual measurements. We evaluate the models performance by comparing with measured responses on a real server HSIO link.
“…In particular, application of space mapping methodology is considered in Rodrigues et al for automated design of microwave filters, whereas Hassan et al describe space mapping for low‐cost design centering of microwave circuits. In Rayas‐Sanchez et al, EM‐based optimization of microstrip and SIW interconnects is presented using low‐order polynomial surrogates. Application of adaptively adjusted design specification technique and sequential approximate optimization is considered in Bekasiewicz and Koziel in the context of design optimization of photonic directional couplers.…”
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.