Malander Engineering LLC

Disclaimer

This project was completed by the founder of Malander Engineering LLC as part of a graduate program. It is an excellent example of the benefits that can be seen from bringing in engineering help to assist in the successful completion of a project.

In addition a record of this work is publicly available in his masters thesis which was published and is available at the library of the University of Colorado at Colorado Springs.

Everything being shared here is also available there.

The amount of detail shared here is generally not permitted for engineering projects completed outside of a University environment. However, the founder participated in many projects of equal and greater magnitude, and went into business with the intention of sharing that advantage with others.

Summary

The project started with a process in which the characterization of one device could be completed by one person, if they worked non-stop for more than 500 years.

Thanks to our workflow automation, by the end of the project we had successfully completed multiple iterations of the design and characterization process.

Characterization and failure analysis was completed on many devices across multiple design iterations by two people, working part time for less than two years.

A more detailed overview of the project is presented below.

The Situation

A company that designed Micro-Electro-Mechanical Systems (MEMS) needed to characterize their devices. The devices were electrically actuated micro mechanical switches, which are similar to a relay. They needed to have these switches tested repeatedly until they failed by:

Using a manually controlled variable power supply connected to the control pads of a switch (via micro probes) to open and close the switch.
Connecting an ohm-meter across the signal pads to measure resistance before, during, and after actuation of the switch.
Using a current source to test if the current capacity of the signal path met, or exceeded, the design specification, while the switch was closed.
Recording all of the voltages applied, as well as the resistances and current capacities that were measured.

The Initial Process

Initially, the process of testing a MEMS device to failure at that company was done by hand, with each iteration consisting of the following steps:

Record the initial conditions, then turn up the power supply slowly until the resistance seen on the ohm-meter drops dramatically, and record the result.
Pass the design specfied current through the signal pads to see if the path between them could handle that level of current, and measure the result.
Lower the voltage of the power supply slowly until the resistance jumps up, ideally back to its initial value, and record the resulting voltage and resistance as in the previous steps.
Repeat the process until the device fails, then compile the results.

Although this process typically would take only a few minutes by hand. The desired lifetime for each device was in the tens of millions in terms of open and close cycles. As a result, testing even one such device to failure in this manner would take a number of years (approximately 500 years).

The Observation

After performing the above process for a couple of days, and successfully collecting relatively few data points, it seemed clear that unless an automated approach was taken, we would probably never reach the desired cycles to failure of approximately 90 million.

On top of that the prescribed process introduced a lot of variables, such as:

A lack of uniformity in voltage ramp rates
Uncertain sensitivity to closing (switch actuation) voltages
etc.

These complicated the process of engineering a better device based on the data collected. The more consistency one attempted in these areas, the longer the process would take. An automated solution could at once eliminate most, if not all, of this variability.

The Approach Taken

As a result, a proposal was created and presented that they allow us to create an automated test system. The opportunity was granted, with much enthusiasm from the engineers waiting for the data. The steps then taken were:

connect all of the electronic test equipment to a PC, using a GPIB card and cabling.
Use HP VEE to programmatically implement the test procedures.
A custom circuit board was created to extend their capabilities, as the electronic test equipment available to test the switches was not quite capable of doing all the needed measurements without human intervention.
Since the end result of the testing was to be an Excel spreadsheet, Microsoft Visual Basic (VB) was used to create components that could write to a spreadsheet. These components were integrated into the HP VEE system, so the measurements could be written directly into the spreadsheet.
In addition, VB was used to perform the statistical analysis and generate the charts used by management and the engineers to determine how best to improve the design in subsequent iterations.

The Result

The result was that measurements could be taken much more quickly and in a consistent manner. In fact, the results came in so quickly that we were actually able to test devices to failure, which generally occurred in the 1.6-2.6 billion cycle range.

In a real sense the test system created could achieve the original service that was requested and verify that a given device did indeed have a lifespan of greater than 90 million switching cycles, even while carrying the desired current. While at the same time collecting and preprocessing all the data requested by the engineers.

This success gave rise to new opportunities that were beyond the scope of the company's initial request. The main one being that there was now a need for failure analysis to be performed on the switches after they failed.

The company again looked to us for help, and we were able to use various resources in the lab to successfully complete failure analysis.

Micro-Electro-Mechanical
System Characterization