Reverse Engineering Vs Modeling in ADS (21)

In this section, let’s learn about Reverse Engineering Vs Modeling in ADS and A vs. B modeling. These are two important techniques for analyzing and comparing circuit or system performance. They can help engineers understand the behavior of existing designs and optimize and improve them based on this.

Specifically, reverse engineering is to deduce the working principle and structure of an existing circuit or system by analyzing it without complete design documents. ADS provides a variety of tools to support the reverse engineering process. A vs. B modeling is a method to compare the performance and behavior of two different designs (A and B) to determine which design is better. In ADS, A vs. B modeling can be achieved through parameter scanning and simulation.

We open the previous TestBench.

In this TestBench, we called the S parameter file.

Now, we want to replicate the design corresponding to the simulation result, that is, we want to create an equivalent circuit that can match the current circuit.

From the figure, the function we implemented is a simple low-pass filter, but the actual components may not be like this. It may be a combination of any passive devices or even active devices.

We click in the TestBench window , create a new schematic, and then build a circuit like this:

Save it as the template “Reverse_Engineering_Template”. Most of the components have been used before, and you can try to build them yourself. Among them, represents the measurement equation (Measurement Equation), which can be found in “Simulation-S_Param”. We define two measurement equations in it. The first is “deltas21=mag(S21)-mag(S43)”, which is used to calculate the amplitude difference between S parameters S21 and S43 to characterize the transmission response; the second is “deltas11=mag(S11)-mag(S33)”, which is used to calculate the amplitude difference between S parameters S11 and S33 to characterize the loss of the circuit. We can define more measurement equations according to our needs. We also created a new variable “matchtol” (matching tolerance) with a value of 0.001 (corresponding to -60dB matching) to assist in optimization operations. Since we defined the measurement equation, the GOAL we optimized is also defined based on the measurement equation.

The optimization settings are as follows:

The definition of GOAL is as follows (the same applies to OptimGoal2):

After looking at the images and graphs we obtained earlier, we guessed that this circuit can be implemented with a fifth-order filter. The input and output load impedance of both circuits is 50 Ohm.

Then we click on Execute Simulation. Then we insert some plots (remember to check “Group Delay” in the parameter tab):

In the first four figures, the red curve is the reference curve, and the blue curve is the curve corresponding to the equivalent circuit. The last two figures represent the dB value of the measurement equation. If the two designs match, the dB value of the measurement equation will be lower than -60dB. As you can see, our current design is still far from the reference design. We click on Optimize on the top-level design page (since the process is relatively random, we can try several times). When the Error is very small, we can change the algorithm to “Gradient”:

Then we can find that Error has become 0, indicating that the match has been successful:

Update the optimized data into the design.

For more complex designs, our matching process is the same. However, when there are too many goals to meet, the optimization process may take a long time. We can optimize a few goals separately first, and when the match is almost the same, we can update the design first, and then enable all goals for optimization.

We can see that at this time, the dB values ​​of the two measurement equations are both less than -60dB, indicating a good match.