In addition, the Magnetics 2010 Conference will be running concurrently at the same venue.  Magnets are a strategic material without which many motors would simply not operate.  In the ever-changing motor industry, there is always a new design that seeks to make an enhancement over previous solutions, or introduce a new solution to old problems.  Declining prices for Neodymium Iron Boron magnets over the last few years have created a number of novel design shifts which have been instrumental in bringing more varieties of permanent magnet machines into the forefront of motion control and mechatronic technology.  To the point where over the last two years a resurgance of permanent magnet rotor designs have been created to improve the energy denisty and lower the cost of specialty motors in washing machines and air conditioning compressors.

This last development, combined with the forecast increase of hybrid electric car sales coming this year, are expected to increase the sale of permanent magnets by 10-15 percent by 2011.  That’s a staggering jump in a market that is almost exclusively supplied by China.  And there is no assurance that China can meet the forecast production.

The US Department of Commerce usually has a say in the sale of products or businesses to foreign countries that are deemed to be strategic or sensitive technology.  In fact, I got stuck in a situation where my employer was told specifically that we could not sell a CNC controller to a Korean customer.  That’s pretty small potatoes compared to controlling the supply of permanent magnets which influences billions of dollars worth of electric motors manufactured and sold all over the world.  So it strikes me as a little odd that the sale of Magnequench to its current owners, Neo Materials, was completed without a much discussion. leaving the US without a domestic magnet supplier.

There will surely be a lot of discussion about this situation at the conference, and I will be in attendance to get the latest information on the subject.  So look forward to a review of the conference in an upcoming post.

And there are so many technologies, all focused on solving application problems but balancing the economics of development cost and manufacturing scalability.  Where would the Oui or iPhone be without accelerometers that are really inexpensive?  Fax machines without G3 communications chips,  or $49 printers without stepping motor chips and ink jet controls?  All benefits of high volume economy of scale.

Industrial control systems have generally required chip technology, but in numbers of chips considered too small to merit custom designed solutions.  But the Rockwell Control Logix concept breaks the partitioning of applications by applying the same control processor to all kinds of control equipment, variable frequency drives, programmable controllers, HMI’s, you name it.

Is there an ultimate chip?  A chip solution that does everything?  Not really.  But the wizards of the microcircuitry world keep coming up with new architectures.  New approaches to existing applications that offer price or performance attributes that will hopefully trigger lots of new designs that result in breakthrough products.

Recent trade press is buzzing about a new processor that combines the logic solving capability of FPGA (Field Programmable Gate Arrays) with ARM (Advanced Risc Machines).

FPGA excels in the ability solve logic, and has scaled up to massive numbers of gates and tremendous processor speeds to solve enormously complex applications.  Even applications requiring real time operation like motor control can be solved through FPGA with proper attention to detail.   Applications that were considered impossible a few years ago are now within the range of these processors.

But using gate arrays may not be the most efficient way to do motor control.  Hard real time motor control requires a great deal of analog processing to monitor conditions in the real world (like voltage and current) and the ability to respond to dynamic changes through complex programming and mathematical models.  Much easier for ARM processors with super efficient instruction sets and single cycle multiplication and division.  In some designs 16 channels of high resolution A to D converters and direct PWM capabilities.

So combining the two technologies seems like the ideal solution for a huge range of industrial control applications.  And if you get it all in one processor, wouldn’t that be great?

I  can’t wait to see what new product developments take place in the next few years with this kind of processing power available.

And, in fact, the industry is getting there. The current estimates are that solar power is costing about the same as peak demand consumer power, $.23/kWh. And with the current wave of investment and scale up, something which the semiconductor industry has always done well, there is serious forecasting that the cost of solar electricity will continue to fall.

The mechatronic connection is really interesting. It was something I wasn’t really paying attention to. The solar tracking application. You need 2 axes of motion, elevation and azimuth, to follow the sun in its daily course and maximize the electricity coming out of the solar panel. Most vendors claim up to 40% increase in the amount of electricity generated by solar power. But when you get right down to it, this is a very difficult application to do, because even though the duty cycle is very low, it requires decent accuracy and low cost. A very tough combination.

Some solar trackers use two motors and gear reducers to move 12 or 16 solar panels in a large frame. The typical solar panel is 15″ wide and 48″ long and weighs about 33 to 40 pounds. 12 units would be a payload of 480 pounds. Pretty serious amount of load.

The nice thing is, you can use a lot of gear reduction to make a smaller motor do the job. But on average the systems I have seen add more than $200 per solar panel to the installation. And that’s about half again the cost of the solar panel to begin with. So it really impacts the return on investment for solar power.

And that means we need a better mechatronic solution to do this job at a much lower cost. So we need to run a contest for the best solar tracker. You can do it any way you want, 1 panel system or multi-panel system. There are some new solar energy systems where mirrors are used to create solar concentrators to increase the light density on the surface of the silicon photovoltaic device. But we just have get a better solution out into the marketplace.

See? There really is a mechatronic connection in this solar energy stuff.

Now here’s the hard part: It doesn’t matter how good a job you do on the mechatronic part if we don’t get the government to change its policy on home financing. Currently, none of the green energy technologies qualify for FHA lending. This means that you have to do your solar retrofit for cash. You can’t get a second on your house or roll it into a new home finance package. So you have to come up with $20K- $25K to get a system installed. That’s the reason we don’t have a million home off the grid yet. Write your congressman or woman.

Various stainless steels and nickel alloys are available as sheath materials for specific application requirements. Diameters are available from 0.020 to 0.500 in.

Termination options are application-driven and range from standard male plugs to screw-cover terminal housings. These units are field-bendable for easy installation. Standard calibration types are K, S, T, E, N, R, S and B. Temperature tolerances are available as standard or special limits-of-error. The thermo-element design can either be single or duplex for multiple outputs from the same point.

Durex Industries
www.durexindustries.com

By Dr. Jens Kummetz,
Marketing Application Development,
Dr. Johannes Heidenhain GmbH

The semiconductor industry continues to demand tighter precision and faster operating speeds from machines in order to satisfy growing demands on quality, production, and size reduction. Linear motors are becoming more important in such highly dynamic applications that use one or more feed axes. The benefits of direct drive technology are low wear, low maintenance, and more throughput.

However, this increase in throughput is possible only if the control, the motor, the machine frame, and the position encoder fit one another. Direct drives place rigorous demands on the quality of the measuring signals. High quality signals reduce vibration in the machine frame, stop excessive noise exposure from velocity-dependent motor sounds, and prevent additional heat generation, allowing the motor to realize its maximum mechanical power rating.

Design of direct drives
The decisive advantage of direct drives is the very stiff coupling of the drive to the feed component without any other mechanical transfer elements. This configuration allows significantly higher gain in the control loop than with a conventional drive.

Direct drives do not need an additional encoder to measure speed. Both position and speed are measured by the position encoder: linear encoders for linear motors, angle encoders for rotating axes. Because there is no mechanical transmission between the speed encoder and the feed unit, the position encoder must have a correspondingly high resolution to enable exact velocity control at slow traversing speeds.

The velocity is calculated from the distance traversed per unit of time. This method—which is also applied to conventional axes—represents a numerical differentiation that amplifies periodic disturbances or noise in the signal. The combination of the significantly higher control loop gain used with direct drives and insufficient encoder signal quality can reduce drive performance.

Signal quality of position encoders
Modern encoders have either an incremental, which means counting, or an absolute method of position measurement. In the encoder, path information is transformed into two sinusoidal signals with 90° phase shift. Both methods require that the sinusoidal scanning signals be interpolated in order to attain a sufficiently high resolution. Inadequate scanning, contamination of the measuring standard, and insufficient signal processing can lead to a deviation from the ideal sinusoidal shape. As a consequence, during interpolation periodic position error occurs within one signal period of the encoder’s output signals. These position errors are referred to as “interpolation error.” On high-quality encoders the error is typically 1% to 2% of the signal period.

During interpolation, position errors can occur within one signal period of an encoder’s output signals.

The interpolation error can produce several effects:
—Generation of heat and noise. If the frequency of the interpolation error increases, the feed drive can no longer follow the error curve. The current components generated by the interpolation error
increase motor noises and additional heating of the motor.

A comparison of the effects of linear encoders with low and high interpolation error on a linear motor illustrates the significance of high-quality position signals.  The LIDA linear encoder, for example, generates barely noticeable disturbances in the motor current: the motor operates normally and develops little heat.

If at the same controller setting, the interpolation errors of the same encoder are increased through poor adjustment, significant noise arises in the motor current, which can cause more noise and heat generated in the motor.

—Dynamic behavior. Digital filters will smooth the position signals for direct drives. However, the additional phase delay caused by filtering in the speed-control loop must be kept to a minimum, otherwise the dynamic accuracy decreases.
Position encoders with optimum signal quality help to reduce the use of filters, which maintains the control bandwidth.

Position Encoders for direct drives
Linear encoders that generate a high quality position signal with low interpolation errors are essential for optimal direct drive operation in the electronics industry. Encoders that use photoelectric scanning are ideally suited for this task, since they can scan very fine graduations.

The right encoder will generate minimal disturbance in the motor current.

Encoders with optical scanning measure periodic structures known as graduations. The substrate material is glass, steel, or—for large measuring lengths—steel strips. These fine graduations—graduation periods from 40 μm to under 1 μm are typical—are manufactured in a photolithographic process. They have high edge definition and excellent homogeneity—a fundamental prerequisite for low interpolation error, and therefore for smooth operating performance and high control loop gain.

By the nature of their design, the measuring standards of exposed linear encoders are less protected from their environment. The manufacturer should therefore always uses tough gratings made in special processes.

Here’s a look at the heat generation of linear motor contolled with an encoder
TOP: With low interpolation error
BOTTOM: With high interpolation error
In the DIADUR process, hard chrome structures are applied to a glass or steel carrier. The AURODUR process applies gold to a steel strip to produce a scale tape with a hard gold graduation.

In the SUPRADUR process that we use, a transparent layer is applied first over the reflective primary layer. Then an extremely thin, hard chrome layer is applied to produce a grating. Scales with SUPRADUR graduations have proven to be particularly insensitive to contamination because the low height of the structure leaves practically no surface for dust, dirt or water particles to accumulate.

These production technologies ensure an enduringly high signal quality suitable for the use of direct drives in demanding applications.

Optimal scanning
The scanning method and the high quality of the grating share responsibility for low interpolation error. In single-field scanning, the output signals are generated from one scanning field. This large field and the special optical filtering through the structure of the scanning reticle and photosensor
generate scanning signals with constant signal quality over the entire range of traverse. Constant signal quality is necessary for:
—Low signal noise
—Low interpolation error
—High traversing speed
—Good control loop performance for direct drives
—Low motor heat generation

To put it simply, the imaging scanning principle functions by means of projected-light signal generation: two scale gratings with equal grating periods are moved relative to each other—the scale and the scanning reticle. The carrier material of the scanning reticle is transparent, whereas the graduation on the measuring standard may be applied to a transparent or reflective surface.

When parallel light passes through a grating, light and dark surfaces are projected at a certain distance. An index grating with the same grating period is located here. When the two gratings move in relation to each other, the incident light is modulated: if the gaps are aligned, light passes through. If the lines of one grating coincide with the gaps of the other, no light passes through. Photovoltaic cells convert these variations in light intensity into electrical signals. The specially structured grating of the scanning reticle filters the light current to generate nearly sinusoidal output signals.

Linear encoder scales made with SUPRADUR graduations tend to be less sensitive to contamination because they have few surface structures.

In the XY representation on an oscilloscope the signals form a Lissajous figure. Ideal output signals appear as a concentric inner circle. Deviations in the circular form and position are caused by position error within one signal period and therefore go directly into the result of measurement. The size of the circle, which corresponds to the amplitude of the output signal, can vary within certain limits without influencing the measuring accuracy.

In single-field scanning, the output signals are generated from one scanning field. This large scanning field, and the optical filtering through the structure of the scanning reticle and photosensor generate scanning signals with constant signal quality over the entire travel range.

On direct drives, deviations from the circular form cause acoustic noise, reduce control quality and increase heat generation.

This chart shows a selection of position encoders for direct drives and the maximum values of interpolation error with respect to the signal period.

Lower sensitivity to contamination
Production facilities and handling devices for the semiconductor industry demand high acceleration and compact designs. Such requirements usually mean exposed measuring systems that operate without friction and, because they operate without their own housing, can be designed to be very small and low in mass. Special scanning methods and production techniques provide tough protection against contamination even without sealing the encoder.

The specially structured grating of the scanning reticle  filters the light current to generate nearly sinusoidal output signals. An XY representation of the signals on an oscilloscope takes the form of a Lissajous figure. Ideal output signals appear as a concentric inner circle. Deviations in the circular form and position are caused by position error within one signal period.

Many exposed linear encoders operate with single-field scanning where only one scanning field generates the scanning signals. Local contamination on the measuring standard (such as fingerprints from mounting or oil accumulation from guideways) influences the light intensity of the signal components, and therefore of the scanning signals, in equal measure. The output signals do change in their amplitude, but not in offset and phase positions. They stay highly interpolable, and the interpolation error remains small. The large scanning field additionally reduces sensitivity to contamination. In many cases this can prevent encoder failure.

Thus, optical encoders with low sensitivity to contamination need an optimal scanning method, a large scanning field, and contamination-tolerant graduation.

Very small signal periods usually come with very narrow distance tolerances between the scanning head and scale tape. However, several varieties of encoders provide ample mounting tolerances in spite of the small signal periods. Within the mounting tolerances, therefore, changes in the signal amplitude remain negligible.

HEIDENHAIN Corp.
www.heidenhain.com

Both inductive sensors use “auto-configure” output technology, which auto-detects if the sensor has been wired for NPN(sinking) or PNP(sourcing), and switches the sensor to the appropriate mode without user intervention.

Eaton Corporation
www.eaton.com

We don’t have a major position in the display technology arena, most of the flat panel display technology comes from Asia. But organic LED may be the next emerging technology and it is still substantially in American hands.

Battery power storage and fuel cell development is a big technology effort in the US and development of small, power dense battery technology has “fueled” (pun intended) many new product breakthroughs like the iPhone and iPod. Scalable storage technology like the recently improved Lithium Ion battery is going to be the key to short range electric vehicles that operate at a few pennies per mile, unlike the hydrocarbon based vehicles that dominate the road today.

And like the classic economists would predict, all the electronics businesses behave in perfectly linear economies of scale. The more product made, the lower the price. The entry level laptop is now $500 and is far more powerful than the most expensive desktop available five years ago. Now that’s my idea of a great “engine of commerce”.

And the commerce cuts across many fields. Medical imaging would not be possible without PC-like products to calculate the volumetrically correct image to be rendered and manipulated in 3D. Mechanical design workstations that used to cost $50,000 can now be hosted on $1000 PC’s The DaVinci surgical robot, an extraordinary accomplishment of mechatronic excellence, would be impossible without semiconductors to control and power it. In fact, we run all our businesses with computers, because its cheaper and faster.

These are multi-billion dollar market segments that are all closely knit together in an industry infrastructure that requires a lot of support. And the factories of the current era of semiconductors and consumer electronics are multi-billion dollar facilities that host some of the most sophisticated technology ever conceived by man.

So it strikes me as somewhat irresponsible that we can have the megalithic technology structure hobbled by the inability to generate electricity, a business that has been with us for almost 100 years. How can we let the energy infrastructure fall into such a state of disrepair that in the summer in California, when demand for electricity is high because of air conditioning usage, that our semiconductor industry can have brown outs and plant shutdowns.

Its absurd. And it is irresponsible for politicians who hold the power to grant permits to build new powerplants, to deny the industry the ability to supply its customers. There haven’t been powerplant permits granted in the US in 30 years. No wonder we’re in trouble.

A former client of mine is in the power generation business and I have done work for his company. Powerplants in the US are among the cleanest burning anywhere. There are even a handful of plants that burn garbage that is carefully pre-screened, and some of those plants are even cleaner than the cleanest coal fired plants.

Then there is the nuclear option. 20 years ago the American Nuclear Regulatory Agency participated in demonstrations of the “pebble bed” reactor. The size of a waste paper basket and totally safe. It can’t melt down and it can’t become unstable because each “pebble” of nuclear material is encased in ceramic that is resistant to temperatures of 3000 degrees.

Let’s wake up the sleeping giant of American ingenuity and start putting to work the people, the technologies that will reinvigorate our economy, before its too late.

Wafer polishing machines must polish the slices of silicon to a flatness and perfection that can’t be measured by conventional means. Multi-axis robots handle silicon wafers in vacuum chambers without putting the tiniest scratch on the surface. Wafer cassettes with $250 to $500K worth of uncut chips have to be shuttled from process machine to process machine inspected and tested for defects.

The materials are among the most exotic in the world; gold interconnects, high purity copper, pure silicon crystals that are grown in furnaces at temperatures in excess of 2600 degrees Fahrenheit. Processes that require chemicals with extraordinary purity, some of which are the most corrosive acids on the planet.

The shiny little sliver of crystalline silicon can embed the intelligence of man-years of programming, sense the position of an actuator to millionths of an inch, measure current in a conductor or regulate lethal amounts of voltage or current in power semiconductor applications. Pretty amazing stuff.

And when it comes to controlling motion, many solutions are available. Digital Signal Processors have been one of the key technologies for controlling motion because of their ability to process mathematical models of analog events. Field Programmable Gate Arrays that host huge arrays of logic gates can read encoder feedback and execute control tasks at incredible speed. Microprocessor based motor controls and servo systems have been around for a decade or more. And the latest generation of microcontrollers offers to integrate the power of the DSP with the flexibility of a microprocessor and multitasking needed to support network communications to other devices.

What’s a bit strange is that we keep solving the same problems over and over using different platforms. Why haven’t we found the ideal solution? At a certain point, its the same mathematical model of a real world phenomenon that we are trying to run. Isn’t it? Or are we solving different problems and needing to find better hardware solutions? Or are the tools evolving to make treatment of the complexities easier?

A little bit of all of these. And maybe that’s what makes it so interesting.

It is somewhat ironic that as we move into the e-tainment era of the 2010’s, surrounded by e-media delivered by ever more powerful portable electronics, that the US semiconductor industry is at least the size of, and by some accounts, a much larger enterprise than the auto industry.   The Department of Commerce shows semiconductor manufacturing at $90B for 2002 and computer manufacturing at about $88B, some of which of course is overlapping.  If you start adding all the flat screen display, cellphones, well, you get the picture.  Semiconductors enable so many products that we take for granted, it is hard to estimate the impact.

US auto sales have been falling since 2002, but the Department of Commerce data lists auto sales at $90B.  Similarly, to get a better picture of the overall auto industry you have to add in trucks and tractors and all the other gasoline and diesel powered vehicle segments that make up the overall industry.  The largest user of glass and carpet is, you guessed it, the auto industry, not residential construction.  So cars are still important in the US economy.

But when you consider the role of control systems in the manufacture of electronic products, the solutions are not the same. Semiconductor equipment companies are among the largest users of motion control across all industries.  Some motion control companies have made their fortunes supplying board level motion solutions to operate the most sophisticated equipment in the world. Just like 50 years ago in the auto industry.

The odd part is, that the semiconductor industry has not taken to the classical controls solutions of the automotive type suppliers very well.  And it took me a while to get that message.  Partly, I think there is a preference for the PC as a platform of control as a cultural proposition.   Computers are the dominant consumer of the electronic components, so people in the industry tend to use what’s closest, most familiar, and in some cases, lowest cost.

But there are much more compelling issue involved.  The semiconductor industry did not grow up with relays as a control technology, so ladder logic as a programming environment isn’t really significant.  Many people in the industry are more familiar with C language programming or other languages that don’t depend on electrical conventions.

But the biggest challenge to control systems isn’t the control, its the data.  Process data is crucial to finding out if you have a $25,000 wafer of silicon that’s good, and can continue in its manufacturing process, or bad and needs to be scrapped out.  When making the platter for a hard disk drive, it takes many steps to build the magnetic layer on the aluminum disk.  Each step is monitored for temperature, time, chemical concentration and pressure so that the process will operate correctly over hundreds of thousands of operations.  Data is the key to process integrity.   So control systems in the semiconductor world have to be data centric, not control centric.

Its a different world.

By Mitsuhiro Nakamura
Agilent Technologies, Inc.

Recently, various devices using MEMS technology such as pressure sensors, accelerometers, and RF MEMS have been commercialized. Additionally, new devices such as silicon microphones, are rapidly evolving. The MEMS market started with the automotive industry and has been expanding to consumer products such as cellular phones.

This MEMS market expansion also applies pressure on manufacturers to lower their costs per device. However there are few opportunities for cost reduction. The limiting factors include:

• Low yields due to the precision process
• Slow throughput due to application of the physical stimulus.

A recent study (item 1 in the Appendix) estimates that 80% of the total production cost is attributed to the device packaging process and how defective chip inflow to the packaging process can contribute to cost increases. Therefore, we will discuss how to evaluate MEMS elements at the on-wafer stage in order to lower the total production cost.

Lowering Production Cost

Testing MEMS elements in the earliest stages of the manufacturing process can help contribute to lowering overall production cost. Key considerations are:

• Prompt product quality improvement by fast process feedback
• Production cost reduction by removing defective chips before package integration.

In particular, testing MEMS at the on-wafer or die level is critical for lowering mass production cost. When testing the MEMS movable part at the wafer or die level, the input and output of the MEMS device need to be considered.

Input and Output Considerations

There are two input methods that drive MEMS devices. One is to apply a physical stimulus such as pressure or acceleration, and the other is to apply an electrical signal. The movable part of sensors is also driven by the bias voltage applied. Applying an electrical signal as the input stimulus is the best method in terms of the speed, repeatability, accuracy and usability, while applying physical stimulus is better when duplicating the device’s operating behavior.

There are also two different methods to measure the output of MEMS devices. One is with a direct displacement measurement with a laser interferometer, and the other is an electrical measurement using test signals. The electrical measurement is applicable for electrostatic capacitance or piezoelectric resistance. Though the direct measurement with a laser interferometer is straight forward, the electrical measurement is superior in terms of repeatability, accuracy and usability.

Both test throughput and yield are critical in the mass production process, as are measurement speed and repeatability of the test instrumentation used. Test equipment usability should also be considered for lowering production cost because the usability affects the ease of maintenance for a test system, which determines the production line up time. Thus, electrical testing to characterize MEMS wafers or dies is preferable to physical stimulus test for lowering production cost (Figure 1).

MEMS wafer characterization process.

MEMS Capacitive Sensors

Common methods for detecting the displacement in MEMS sensors, such as pressure sensors, accelerometers, and silicon microphones, utilize piezoelectric or capacitive techniques. We will examine the electrical test of capacitive sensors as an example.

The capacitive sensor can be modeled as shown in Figure 2. The distance between electrodes is changed by the physical stimulus such as pressure, acceleration, or sound waves. The change in distance can be read electrically as the electrostatic capacitance change.

Capacitive sensor diagram.

The capacitive sensor has two mechanical characteristics — the static response (static characteristic) and the dynamic response (dynamic performance) against the input of the physical stimulus. Static response is a fundamental characteristic and is defined as the capacitive sensor displacement when a static physical stimulus is applied. The dynamic performance is described as the response when a dynamic physical stimulus is input. The dynamic performance is expressed as the frequency response of amplitude and phase, often represented by parameters such as resonance frequency, Q-factor, and 3 dB bandwidth.

Evaluation by Electrical Measurement

As previously mentioned, the physical stimulus input can be replaced with the electrical stimulus input, which we will now examine.

The static physical input can be replaced by applying DC voltage bias. Therefore, the static characteristic can be evaluated by measuring the electrostatic capacitance by sweeping the DC bias voltage. This measurement is generally referred to as a Capacitance-Voltage (C-V) measurement. The important specifications and features of the test instrumentation are impedance range, measurement accuracy, frequency range, measurement speed, repeatability, and DC bias voltage range.

The capacitance of the most capacitive sensor is approximately 0.5 to 1 pF at the neutral position. Consequently, the test equipment needs to be capable of measuring electrostatic capacitance accurately in the order of 0.1 pF. The four-terminal pair method is recommended for the most accurate impedance measurement.

Note that an AC test signal is used for the impedance measurement. If the test frequency is set as low as the electrodes of the device, the voltage of the test signal can actuate the electrodes. When the electrodes are moving, electrostatic capacitance cannot be measured correctly. Therefore, the test frequency should be set much higher than the mechanical operating frequency of the device under test. Generally, the operating frequency of the MEMS device is in the low kHz range, so a 1 MHz test frequency is adequate.

As superior impedance measurement repeatability enables the tightening of the guard band in the testing process, it also helps to improve yields. Measurement repeatability is important when characterizing small mechanical displacements, as is device processing accuracy. In the case of a capacitive sensor with capacitance of 1 pF, the measurement repeatability should be less than 0.1%, which means that 1 fF or less of the repeatability is recommended.

Caution must be exercised when the DC voltage bias sweep measurement is performed. Capacitive sensors have hysteresis characteristics based on the amount of electrical charge being inducted to its electrodes. This hysteresis is one of the parameters to be evaluated by a C-V measurement.

Precision LCR meters are generally used for these types of measurements.

Dynamic Performance

The dynamic physical stimulus can be replaced with the AC voltage applied by the electrical measurement. The characteristic of the movable part as a portion of the electrode can be modeled as shown in Figure 3. The mechanical characteristic of the movable part is reflected to the measured impedance at a lower test frequency than the mechanical operating frequency. Thus, measuring the impedance of the device can illustrate the frequency response of the movable part. The four-terminal pair method with the impedance measurement is recommended to achieve the most accurate measurement.

The movable part being driven by the AC voltage applied has electrostatic attraction between electrodes. Because the electrostatic attraction is proportional to the square of AC voltage applied, it generates a second distortion to the current flowing into the electrodes.

An impedance analyzer obtains the impedance value by the measured vector value of the fundamental element of voltage over that of the current, so the second distortion may cause measurement error. Applying DC bias voltage to the AC test signal is a good way to solve this problem. When the amplitude of the AC voltage is adequately smaller than that of DC bias voltage, the second distortion of the current is negligible so that a valid measurement can be performed. This method allows for a quick and easy evaluation of the frequency response of the device, except when the electrode is at its neutral position.

Impedance versus frequency profile is the fundamental measurement for characterizing the dynamic performance of the device and can be obtained with an impedance analyzer. The dynamic performance of the device can be derived from the measurement results, which represents the performance at the position of the electrodes driven when a DC bias voltage is applied. The dynamic performance at any position can be obtained by varying the DC bias voltage. The level of AC test voltage needs to be set smaller than that of DC bias voltage.

Note that the measured impedance profile has both impedance, representing the dynamic performance, and the electrostatic capacitance, representing the electrode displacement. An equivalent circuit model is shown in Figure 3. To determine the dynamic performance of the device itself, electrostatic capacitance can be subtracted. The electrostatic capacitance can be obtained from the measured impedance value at a higher test frequency than the operating frequency of the device.

Characteristic of the movable part of the electrode.

Leakage Measurement

Besides characterizing static and dynamic performance of MEMS devices, leakage measurement is also an effective consideration for quality management. The leakage measurement between the electrodes enables the early detection of device defects.

A pico ammeter such as a semiconductor device analyzer or high-resistance meter is generally used. If parametric testing is required due to monolithic-type MEMS devices containing transistors and MEMS elements in one chip (e.g. at the die level), a semiconductor test system can be used.

However, a high-resistance meter can be sufficient for leakage test in terms of cost, simplicity, and quick operation. The required performance for the MEMS leakage test is that an instrument has the ability to measure the resistance at 100 GΩ or more accurately.

Setup and Configuration

For on-wafer measurements, configuring the probe station and probe card with the test instrument also need to be considered. The shape of the probe card depends on the device under test. However, for the precise measurement by the four-terminal pair method, the cabling from the device to the probe and also the card design are important.

The impedance measurement requires the ability to compensate for the measurement errors caused by cable extension and the parasitic impedance of the probe card from the measurement data. Compensation has to be performed at the end of the probe, using supplied impedance standard substrates from the probe station vendor.

The above considerations and compensation procedure are similar to that of a FET gate insulator measurement. For additional information, refer to items 2 and 3 in the Appendix.

Conclusion

As we have discussed, making on-wafer impedance measurements at the earliest stages in the manufacturing process can be very effective in lowering the production cost of MEMS devices. Using high-performance test instruments with impedance measurement techniques that are accurate, fast, and repeatable are required to characterize the small mechanical displacements of these MEMS devices.

Agilent Technologies, Inc. www.agilent.com

[edit] Appendix

1. . The MEMS Test Community — http://www.memunity.org/on-wafer_testing.htm
2. . Application Note: “Agilent Evaluation of MOS Capacitor Oxide C-V Characteristics Using the Agilent 4294A,” Literature Number 5988-5102EN.
3. . Application note: “Agilent Technologies Impedance Measurement Handbook,” Literature Number 5950-3000.