Knowles Acoustic Case Study
How does a designer of MEMS microphones reach one billion shipped units and go from concept to finished GDSII in about two weeks?
When you make audio transducers for more than 60 years, eventually you can design almost anything—even a heat–tolerant microphone that fits into electronic assembly flow like any other component.
It’s not easy, but Pete Loeppert, Vice President of R&D for Knowles Acoustics, leads a team that figured out how to increase the high–temperature survivability of microphones. This temperature–robustness has played a key part in productizing their MEMS designs, which now account for over $100 million per year of revenue.
Why a MEMS Microphone?
Since 1946, the parent company, Knowles Electronics, has specialized in microphones and receivers for hearing aids. The Knowles Acoustics division has built a new business around designing and manufacturing MEMS microphones used in mobile phones and consumer electronic devices, and is making inroads to the entire microphone market. The division’s business revolves around its answer to a thorny electronic problem.
Traditionally, microphone suppliers stack up individual components and assemble microphones one at a time. Most are cylindrical, about 6mm in diameter by 1-2mm high, and inexpensive enough to go into everything from phones to toys.
The problem is that traditional microphones are heat–sensitive, which precludes the use of lead-free solder and the option for surface mounting into circuits. So, most manufacturers of high–volume products resort to some kind of offline task, such as hand assembly or a special insertion machine, at the end of the mainstream assembly line to get around the fact that the microphones won’t tolerate high temperatures or reflow soldering.
“High–end EDA tools are better suited to Manhattan-style geometry and keeping things ‘square,’ but everything we do in MEMS is circular. With L-Edit, I can go from concept to finished GDSII in about two weeks. There’s never been anything as easy to use for this as Tanner tools.”
Knowles’ answer is the SiSonic™ MEMS microphone, batch–produced on silicon wafers and assembled like any IC, except for an air pocket that allows sound waves to vibrate the diaphragm. The SiSonic™ is reflowable, so an assembler can place it on a circuit board with a chip shooter like any other component. Suddenly, microphones fit in with normal assembly flow.
The World of Toroidal Geometry
Loeppert describes the difference in design paradigm: “In MEMS there are no circuits per se, so we don’t deal in schematic versus layout. MEMS in our group is all about drawing complex polygonal and curved structures.
“I used other high-end tools for years because we were doing circuit work in the background, but when the task is mainly to render complex geometries in these microphones, those tools have a lot of overhead, and they’re awkward. Most of what we design and make is circular, based on tori—toroidal elements and toroidal sections—and there are almost no right angles in our designs. You’ll find some circular symmetry, but most of the design looks like modern art.”
In the late 1990s Knowles Acoustics moved entirely to L-Edit for MEMS design. In particular, Loeppert cites the flexibility of manipulating thousands of repeating elements through L-Edit’s hierarchical architecture.
“L-Edit has saved me a great deal of time in creating all sorts of shapes that are parametrically driven. In fact, I always have my circles, pie wedges and tori instanced slightly differently in every layer, because I use them as primitives. I might have been able to do this in a tool like AutoCAD, but I’d have wasted hours going back and forth between mechanical design and EDA tools. I want to create and analyze designs in one environment, then send them out for photomask fabrication.”
Design Meets Scripting
Loeppert’s designs are not especially small—the die size is about 1mm—but they are intricate, and drawings with 10-12 million objects are common. While MEMS design flow today does not lend itself to direct object generation through standardized libraries, L-Edit’s scripting function makes creating and managing thousands of parametric objects extremely easy.
High-end tools have highly specialized scripting languages, but users of Tanner tools can write scripts using ordinary C/C++ code.
“L-Edit’s scripting function is very flexible, and I use it often to create primitives with a one or two-page script,” notes Loeppert. “For example, in MEMS we often etch holes through our wafers and we can not have die intersecting the edge of the wafer, so we need to array our die in a circular pattern. I start with an instance of a die and use L-Edit to make that into a rectangular array. Then I use a script that I have written to clip the rectangular array to fit within the wafer extents leaving a few millimeter exclusion zone around the edge. This is a big time-saver for us.”
Loeppert takes advantage of L-Edit’s scripting function for other tasks as well: I’ve made mapping programs for die bonding pick-and-place equipment. The script goes through the database of cells in the layout and automatically generates a wafer map—this is particularly important when working on a matrix design that has lots of different designs in it. I create the maps and assign different letters to different design styles, then tell the die-bonding engineers to pick a particular letter out of the map. Being able to generate the map automatically saves an enormous amount of time.”
Getting to the Fab
DRC is also different in the world of MEMS because there are few set rules. Loeppert has to create his own rules, or work them out on the fly with the fab. The cross-sectioning tool in L-Edit helps with this because it allows Loeppert to visualize his designs in the third dimension of stacked layers.
Loeppert exports to GDSII for handoff to his fabs. He notes two limitations peculiar to many GDS tools and explains how L-Edit helps him get around them: “First, the fab can have instances only at 90/180/270/360–degree, and when I rotate things, I don’t always end up with these particular angles. I wrote a script for L-Edit that scans the database for any acute or obtuse angles and flushes them out for me. I ungroup each one, change it to be rectilinear, then later regroup it as it was originally.
“Also, L-Edit supports user-controllable fracturing of polygons. This allows me to set the limits for the various mask fabrication vendors. We’ve never had a tool–related problem. Even when we encounter limitations in our vendors’ systems, L-Edit lets us easily overcome them. We have a very smooth interface with the fab.”