Raising The IQ Of Your MEMS-Based IC Design Flow

By Nicolas Williams and Qi Jing

Internet of Things (IoT) applications depend on smart objects that interact with the real world. So your IoT project is likely to contain ICs that integrate micro electro-mechanical systems (MEMS), such as accelerometers, pressure sensors, motors, and microphones that acquire data for analysis. These projects are finding their way into automobiles, phones, and medical devices as differentiating features. IoT projects are unique in that they often require analog, digital, RF, and MEMS design skills to create a system on a single IC.

But when adding MEMS to your chip designs, you might discover that your familiar EDA tools aren’t smart enough. Why? MEMS design requires the specification of unique shapes (see Figure 1) that are difficult to create and to verify using your standard IC design tools.

Figure 1: MEMS magnetic actuator, accelerometer, and rotary side drive motor.

The unique shape of MEMS is the root cause of many problems that can occur when you use standard IC layout tools. You might be looking for new techniques to address these problems. Here are the four key areas that are unique to MEMS design:

• Drawing complex shapes. You want to use curved objects that can be stitched together to form complex structures. IC layout tools require you to use rectangles, rectilinear polygons, or polygons with 45-degree angles to draw MEMS, which makes it difficult to create curves.
• Curve conversion. You must convert curved polygons to all-angle polygons for DRC and exporting to GDSII. The all-angle approximation must be as accurate as possible. You need a solution that converts curves based on the manufacturing grid, which automatically adjusts the number of vertices for maximum precision, regardless of curve size. Standard IC layout tools suffer from conversion errors that increase as the size of the curve increases.
• Passing design rule checks (DRC). You need a method that uses edge analysis in order to minimize false DRC violations. DRC in standard layout flows are designed for IC geometry that does not handle all-angle or curved geometry, resulting in many false errors.
• Checking connectivity. Visual techniques to highlight and inspect connectivity are more useful than standard IC layout tools that require you to comb through textual error reports in order to ensure connectivity.

Drawing complex shapes
The easiest way to manually draw MEMS shapes is to draw a set of curved objects that can be stitched together to form complex structures, such as a gimbal for a magnetic actuator (see Figure 2).

Figure 2: Complex MEMS shape stitched together from basic curved objects.

MEMS layout requires precision placement and alignment in order to create complex shapes. You will need all-angle rotation and alignment features. Snapping techniques allow complex alignment options such as snapping to a vertex or the midpoint of an edge allowing for precise assembly of complex MEMS. Sometimes moving objects is easier to accomplish using commands that specify horizontal and vertical coordinates, while preserving angles and edge length.

You typically want to group a set of objects into a block or master cell for reuse. You can then reference instances of this master cell to build a complex MEMS layout. If you make changes to the master cell, all instances in the design change unless you protect specific instances. This allows you to quickly update a set of changed cells.

Curve conversion
Standard IC tools convert curved polygons to all-angle polygons using a fixed number of vertices (typically 64). This means that as the size of the curve increases, the error rate increases. A more accurate technique is to use the manufacturing grid to scale the number of vertices to keep the error rate low. Even though edges are smoothed when fabricated, the error rate affects how the resulting MEMS structure performs. Table 1 shows the differences in vector count and error rate for three sample shapes, based on conversion technique.

Table 1: Comparing curve conversion techniques.

Passing design rule checks
Another challenge that MEMS shapes can present is dealing with false errors during DRC due to the all-angle conversion. For example, if you are using an extension rule (measurement between an object on one layer that has extended out of an object on another layer), false errors can occur. In this case, false errors occur when curved edges are approximated as multiple, small all-angle edges by breaking a long single edge into multiple small edges as seen in Figure 3.

Figure 3: False DRC errors caused by all-angle conversion.

A better approach is to employ edge analysis on the shapes to eliminate as many false errors as possible.

Checking connectivity
It is more intuitive to check the connectivity of your MEMS visually, instead of working through connectivity errors in textual reports. You want to select a node and then see all the geometry connected to that node through layers. Figure 4 shows two highlighted nodes that visually indicate bad connections. In this example, you can see a small gap caused by round-off.

Figure 4: Visually highlighting connectivity between two nodes.

Conclusion
When you design MEMS, it is important to examine the key limitations of standard IC tools and find solutions to the challenges of arbitrary shapes and structures. To learn about more about raising the IQ of your MEMS-based IC design flow, download the whitepaper “Meeting MEMS Design Challenges with Unique Layout Editing and Verification Features.” Click here to download Part 1, and here to download Part 2.

Qi Jin is a technical marketing engineer at Mentor Graphics.