Never miss a story from RF Engineering. Subscribe for in-depth analysis and articles.
The Team RF "μWaveRiders" blog series is a showcase for Cadence AWR RF products. Monthly topics will vary between Cadence AWR Design Environment release highlights, feature videos and spotlights, and software tips, tricks, and customization. Subscribe to receive email notifications about our latest μWaveRiders blog posts.
Optimization automatically adjusts selected parameter values of a design to meet goals established by the user. An example of a bandpass filter where passband and stopband goals are constructed for both gain and return loss is shown below:
Parameter values selected for optimization in this case would include the capacitor and inductor values of the circuit elements used to model the bandpass filter.
Optimization algorithms create an error function that compares measured results against the optimization goals. Parameter values are adjusted in order to minimize the error function cost, as shown below:
An issue prevalent in the optimization process is for the algorithm to only find the local minima error function cost and not the best solution, which is the global minima cost. Several different optimization methods are available in the AWR Design Environment platform, where the main objective is finding the global minima as efficiently as possible. Different classes of circuit designs lend themselves better to certain optimization methods when the objective is quickly finding the lowest optimization error function cost. The AWR Design Environment Simulation and Analysis Guide includes an Optimizer Selection table with recommendations on the type of circuit design best suited for each optimization method.
AWR V22.1 software includes a new Pointer-Hybrid optimization method which uses a combination of optimization methods, switching back and forth between methods to efficiently find the lowest optimization error function cost. The optimization algorithm automatically determines when to switch to a different optimization method, making this a superior method over manual selection of algorithms. This method is particularly robust in regards to finding the global minima without getting stuck in a local minima well.
An example of an optimization problem shows optimization costs versus optimization iteration plotted for the same design.
This plot compares the optimization cost versus iteration for the new Pointer-Hybrid method against a few other optimization methods. For examples such as this one, where there are many optimization variables and more than one local minima, the Pointer-Hybrid approach shows that optimization goals can be met with fewer iterations over other methods.
Note that these results apply to one particular design; for other designs, a different optimization method may be more efficient. In general, Pointer-Hybrid is a good first choice optimization method for most design classes.
In addition to the new Pointer-Hybrid optimization method, AWR V22.1 software simplifies the optimizer selection process. The list of optimization methods found in previous software versions has been reduced, not by eliminating optimization methods but by combining methods that had minor variations. For instance, single and parallel thread optimization methods are now combined, with an option to select the number of parallel jobs (optimization spread amongst multiple compute nodes). In addition, optimization methods in the selection list are now ordered according to the generally recommended method.
When opening projects developed in previous versions of AWR software, optimization methods are automatically mapped to the current optimization methods available.
By: Brian Avenell (Sr. Principal Product Engineer)Cadence AWR R&D - U.S.
For questions, general feedback, or suggestions for future blog topics, write to firstname.lastname@example.org.