Parallel Meshing Refinement in Electronics Desktop

September 14, 2022 Jared Hansen

Introduction:

Every year Ansys creates multiple updates to the different softwares. Sometimes these updates are quality of life improvements, sometimes fixes to prior bugs and/or errors, and sometimes they are general enhancements to the entire tool. Today we are going to present simulation benchmark information around a new beta feature in 2022 R2 called Parallel Mesh Refinement (PMR). We have created a microstrip differential pair created in HFSS and will be simulating it at multiple lengths in attempt to compare simulation times between using PMR and not using the new feature. Before we begin, we will show you where to find this new feature within the Electronics Desktop.

Where to find beta features:

In the design ribbon shown in figure 1, we can click on general options. This will open a dialog box found in figure 2. Within General go to “Desktop Configuration” and then at the bottom click on “Beta Options.” This will open a beta options selector with all the available beta options. For this blog we will be focusing on PMR only. Checking/Unchecking the box next to PMR will enable/disable it as shown in figure 3.

Figure 1. Finding General Options

Figure 2. General Options

Figure 3. Beta Options Selector

Simulation Setup:

For this simulation we have created a differential microstrip with an FR4 substrate. The conductors are copper, and the geometry is created for a differential impedance of 90-ohm. Figure 4 shows the cross section of the differential microstrip. We are using waveports set at each end of the microstrip with the same ground as reference. For the analysis setup we are using a broadband mesh with default settings, a low frequency of 1 MHz and a high frequency of 20 GHz, and a max delta S of 0.02. Figure 5 depicts the analysis setup used for each run. For the frequency sweep we are using an interpolating sweep with a sweep of 0-20GHz, and a default setting of 0.5% error tolerance. We are varying the length of the microstrip in order to increase the mesh counts and potential time of simulation. Table 1 depicts the different lengths used and figure 6 shows an example of the microstrip at a certain length.

Figure 4. Differential microstrip cross section

Figure 5. Analysis setup.

 

Run Number

Length (Mils)

1

1500

2

3000

3

3500

4

4000

5

4500

6

5000

7

5500

8

6000

Table 1. Length of each differential microstrip for each simulation

Figure 6. Example of differential microstrip length (6000 mils)

Simulation Results:

Simulation results are shown in table 2 while also being shown graphically in figure 7.

Run Number

Length (mils)

PMR Time (seconds)

Without PMR Time (seconds)

Approximate number of solved elements without PMR

1

1500

299

455

10,000

2

3000

838

825

51,000

3

3500

1190

1182

73,000

4

4000

1060

1891

120,000

5

4500

2205

7289

360,000

6

5000

2337

1758

110,000

7

5500

2349

3121

180,000

8

6000

2112

4169

230,000

Table 2. Simulation Results

Figure 7. Results of simulation

Discussion:

The data shows that PMR almost always gives a benefit is nearly equal to a non-PMR run. There appears to be a significant benefit in solve time when mesh counts are higher than 100k elements and would expect for this to continue as mesh element size increases. It is worth noting that the writer’s intent was just comparing with a single, rather simple model. Additional investigation will need to be performed on more complex models with non-ideal geometry as well as port sizes that are greater than two. For now, this data provides some interesting insight into this new beta feature that could provide a simulation speed boost to an area (meshing) where it has been difficult to see speed increases in the past. The writers would also love to express interest in seeing how our customers results with this new feature and any feedback with PMR. We imagine that Ansys will continue to refine this beta feature for additional usage and are excited to see how this may improve simulation times for more complex models moving forward.

 

About the Author

Jared Hansen

Manager, Electronics Engineering Solutions

Follow on Linkedin More Content by Jared Hansen
Previous Article
Happy Halloween: Jack-o'-Lantern Simulation
Happy Halloween: Jack-o'-Lantern Simulation

We’re getting into the Halloween spirit here at RandSim, and I thought I’d take the opportunity to play aro...

Next Article
New Home for Rand Simulation Solutions Blog
New Home for Rand Simulation Solutions Blog

We truly enjoy writing blog posts and sharing them. But that’s not all we do. Our new blog site lives on th...

×

Have Questions?
Just Ask.

First Name:
Last Name:
Organization:
Country
Comments:
Thank you!
Error - something went wrong!