From single-chip to smart to super-smart

Author : Josep Montanyà Silvestre, Baolab Microsystems

13 March 2013

NanoEMS technology

The diminishing price-tag of motion sensors provides opportunity for these sensors to become ubiquitous.

The meteoric growth of motion sensor deployment in mobile phones and tablets has led to eye-watering price erosion. Consequently, whilst creating a profitability headache for many MEMS manufacturers, the diminishing price-tag provides opportunity for these sensors to become ubiquitous - enabling electronic gadgets, devices and accessories with motion and position awareness, adding to the appeal of existing products and opening the door to some exciting new concepts.

For the trend to advance, the ever decreasing price requires a fundamental change in the way sensors are traditionally made. Baolab’s NanoEMS technology uniquely supports the fabrication of the entire set of motion sensors in standard CMOS fabs, resulting in a much reduced cost, thus enabling the continued proliferation.  Even more significant is that NanoEMS allows all of those sensors to be simultaneously fabricated on a single chip - together with the supporting analogue (signal conditioning) and digital (DSP and intelligence) electronics, thus paving the way for smaller, smarter motion sensor modules at a much reduced cost.
There are three key problems limiting the potential ubiquitous proliferation of motion sensors -- size, power consumption and complexity -- which drive manufacturing cost. NanoEMS is a radical new approach that shrinks all the motion sensors and integrates them together with control electronics onto a single chip in the standard CMOS wafer process, used to fabricate the majority of high-volume cost-sensitive chips today. To understand the significance of this breakthrough, we will describe the problems posed by traditional MEMS approaches and show how NanoEMS systematically removes the issues and provides a clear pathway to smaller, cheaper, lower power intelligent solutions that will drive exciting new consumer product innovation.

Discrete complexity
Four different motion sensors are required to provide sufficient position and motion information: compass (absolute 2D angular position); pressure sensor (altitude); accelerometer (linear acceleration); and gyroscope (angular acceleration). Traditionally, each type of sensor is manufactured as discrete devices using specific, non-standard, manufacturing processes or materials, and each having its own ASIC for signal conditioning and conversion to a digital interface. Consolidating these devices is highly complex using traditional techniques. Even though a single CMOS control ASIC could unify and optimise the electronics required for all four sensors, most manufacturers do not have access to all the specific sensor processes and typically form partnerships to offer all devices. Furthermore, bringing MEMS devices together into the same package requires complex, specialist, multi-chip packaging that increases cost and size, and impacts reliability.

Incompatible processes limit integration
Traditional MEMS sensor devices (accelerometers, gyroscopes) are manufactured in specific MEMS fab lines, which enable the mechanical structures to be formed on top of a silicon substrate by depositing additional materials and using micromachining to define the structures, or within cavities in the wafer that are created using costly, non-standard additional processes such as isotropic Deep Reactive Ion Etching (DRIE) to produce vertical-sided cavities.

Heading sensors (three-dimensional electronic compasses) are traditionally created using either the Hall effect, or a magneto-resistive material deposited onto the silicon to detect the Earth’s magnetic field. To measure the field in three dimensions, separate sensing elements are required – one for each dimensional component of the field, and additional complexity is required in the device construction. In some cases, special devices called magnetic concentrators are bonded to the chip to rotate the X and Y (horizontal) components of the magnetic field into the z (vertical) dimension so they are perpendicular to the plane of the chip. Alternatively, two separate sensor chips are arranged in horizontal and vertical planes within the device package to measure the magnetic field in three dimensions. Such techniques result in mechanical alignment difficulties, which lead to cross-axis measurement inaccuracies.

Gyroscopes are typically created using resonant structures, requiring the MEMS to reside in a partial vacuum. For traditional MEMS, this requires use of special manufacturing techniques, like wafer fusion bonding and wafer-level, eutectic-bonded, hermetic capping to create a vacuum within the package. The use of chemical getters required to mitigate outgassing add further to the cost of hermetic packaging: these chemically absorb gases that may be released from the silicon and other package materials, during periods of elevated temperature experienced during package processing or solder reflow.

A mixed-signal ASIC is required to accept the small signals from the MEMS or magnetic sensor devices, condition them with low-noise amplifiers and gain stages, convert to digital then apply further digital signal processing conditioning and provide a digital interface. The ASIC is a CMOS device and typically has to be manufactured separately from the MEMS devices and stacked with the MEMS in the final package using complex, multi-chip modules or specialist, wafer-bonding methods.

NanoEMS: a brief overview
The Baolab NanoEMS technology uses the existing metal layers in a CMOS wafer to form the MEMS structure using standard mask techniques. 
The Inter Metal Dielectric (IMD) is then etched away through the pad openings in the passivation layer using vHF (vapour HF), releasing the metal MEMS structures. The low-cost isotropic etching uses equipment that is already available for volume production and takes less than an hour, which is insignificant compared to the overall production time: contrast this with the highly anisotropic DRIE process used by other MEMS manufacturers to create cavities and define structures in CMOS, requiring more elaborate equipment and costly process steps.

The holes are then hermetically sealed using a standard thin-film deposition applied to the whole wafer in a single step. Finally the chips are singulated and packaged as required.

Since only standard, high volume CMOS processes are used, NanoEMS structures may be monolithically integrated with other nanoEMS sensors and active circuitry as required enabling smart sensors to be created. The MEMS and electronics are formed simultaneously using a single, standard mask set.

A generic CMOS motion sensor cell
NanoEMS resolves the previously mentioned complexities by creating generic motion sensor cells that may be adapted to create all of the motion sensor types.  This not only provides the key to sensor integration but also enables exciting new concepts, like smart reconfigurable sensors.

In motion sensing, MEMS devices measure mechanical deflections caused by applied forces. The forces to be measured are those that are applied to the sensor by:

1. Linear acceleration, a  (Newton: Force = mass x a)
2. Angular velocity, ? (Coriolis: Force = 2 x mass x velocity x ?)
3. Earth’s magnetic field, BE(Lorentz Force = BE x conductor length x conductor current)
4. Pressure, P (Force = P x Area)

The ‘released’ mechanical structures and the associated capacitive sensing structures that form a NanoEMSmotion sensor cell are designed in principle to detect and measure the presence of an applied force. The mechanical and electrical design criteria determine the range (expected maximum strength of the force) and resolution of the measurement of the force, given appropriate constraints: -
? the bandwidth (BW) required of the measurement system (how fast the force is changing)
? the Signal to Noise Ratio (SNR) required to provide the appropriate range and resolution
? the sensor current (I) required for the measurement

So, in principle, each of the quantities of interest (a, ?, BE, P) translate to a force applied to the sensor device, and each can therefore be measured using the same ‘generic’ MEMS structure. The specific design constraints for the sensing structures are adjusted according to the characteristics of the force applied by the quantity of interest (a, ?, BE, P).

The generic motion sensor cell concept not only cracks the ‘integration roadblock’ of different processes for implementing the four different sensor types, but also allows all three axes of the accelerometer, gyroscope and compass to be deployed in a single plane, i.e. the plane of the chip. Indeed, the generic nature of the motion cell design ensures a rapid time to market for different sensor variants.

Multiple sensors on a single chip
Using the NanoEMS generic sensor cell concept, each of the key issues limiting the cost-effectiveness of co-packaging all the motion sensors are systematically resolved.  Monolithic, integrated NanoIMUs become an exciting and realistic prospect, bringing the complete set of motion sensors together on the same chip and in a package size similar to today’s discrete motion sensors.

Low-cost manufacturing: NanoEMS enables all motion sensors and even other MEMS devices to be manufactured in standard CMOS. Manufacturing test time is reduced.

Reduced process complexity: All four motion sensor types can be manufactured in the same CMOS process with no special materials. By using the Lorentz force compass technique, no additional magneto-resistive material deposition or magneto-concentrator device is required, further reducing manufacturing cost and increasing cross-axis performance by virtue of the sensor alignment being defined by the very tight x-y alignment accuracy of conductors on a CMOS wafer.

Low-cost packaging: Since the sensors are created and sealed inside standard CMOS chips, they can be packaged in either standard, plastic moulded packaging, or, because the device is monolithic, a flip-ship style CSP (chip-scale package) for minimal PCB area.

2 Fusion intelligence takes individual data streams provided by the 2, 3 or 4 types of motion sensors, and processes them into simple, standard motion vectors

Low-complexity packaging - even where a vacuum is required: The NanoEMS cavity inside the CMOS chip is sealed in a vacuum, sufficient to accommodate resonant devices requiring internal vacuum to provide the desired Q factor of resonance. This elegant solution precludes the need for specialist packaging and getters to form and maintain the vacuum: with the evacuated cavity sealed inside the silicon, the chip can once again be encapsulated in standard plastic molded packaging or flip-chipped.

The single-chip motion sensor
With all four motion sensors capable of being manufactured in the standard CMOS flow, NanoEMS uniquely enables them to be placed on the same CMOS chip. Indeed together with the sensors, the electronics can be consolidated, optimised, and deployed on the same chip: namely signal conditioning, digital conversion and ‘intelligent’ digital signal processing.

This prospect of a single-chip motion sensor delivers a host of benefits:
? size: 10-axis motion sensor in the same tiny form factor as today’s discrete sensors
? cost: all standard processes - overall manufacturing costs can be reduced by at least two thirds
? performance/features: co-design of electronics and MEMS enables performance optimisation, reduced power and smart new detection features

Single-chip to super-smart
The CMOS integration and simple packaging capabilities enabled by NanoEMS technology truly transform the roadmap for motion sensing devices, shifting the focus from today’s discrete single and dual-function sensors implemented as multi-chip modules, to:

?       single-chip, multi-function, motion sensors
? single-chip, 9 or 10-axis NanoIMUs
? smart, single-chip nanoIMUs with on-chip sensor fusion intelligence
? and onwards to ‘super-smart’ reconfigurable sensors.

Fusion intelligence in its simplest form takes the individual data streams provided by the 2, 3 or 4 types of motion sensors, and processes them into simple, standard motion vectors that can be directly used by the application without preprocessing.

As well as combining the datastreams, internal corrections can be made to compensate the inherent weaknesses of each sensor: for instance, compass readings can compensate the gyroscope’s inherent drift; gyro and accelerometer can be used to validate the compass reading against magnetic interference, and the accelerometer provides tilt-correction for the compass: accurate headings otherwise require the compass to be held ‘flat’ whereas phones are usually tilted towards the user.

With NanoEMS, the processing electronics required to implement the fusion intelligence (the ‘fusion ASIC’) can be co-designed and co-integrated with the sensor cells to create the smallest, most flexible smart sensor solution with lowest manufacturing cost. Clever codesign techniques allow the electronics to mitigate the effects of both temperature and time related changes in the MEMS structures, defined by the mechanical properties of the interconnect metal used to build them. This leads to the very beneficial side effect of autocalibration, significantly reducing and even eliminating the need to factory-calibrate each axis of each sensor: a significant fab-time cost adder for traditional approaches.

Even further flexibility and performance optimisation is offered by the concept of ‘super-smart’ reconfigurable sensors, made possible by the NanoEMS integration and generic sensor cell capabilities.

The future
Beyond the realisation of single-chip sensors, single-chip IMUs and single-chip intelligent IMUs, NanoEMS technology today enables the visionary concept of reconfigurable sensors, the benefits being:
? optimum performance (accuracy and bandwidth) at the lowest power consumption
? improved reliability
? lowest cost

The reconfigurable sensor concept leverages the unique properties of NanoEMS technology:
? smaller NanoEMS sensor devices: at 100-300µm, sensor cells are typically 5-10 times smaller on a side than traditional MEMS equivalents;
? Monolithic integration of MEMS sensors with electronics;
? Dynamically scaleable performance: the MEMS structures used, and the co-design of MEMS with electronics allows dynamic adjustment of performance parameters to minimise power
? MEMS ‘for free’: the next generation of NanoEMS sensors will use fewer metal layers to implement the mechanical structures, so MEMS may be deployed ‘above’ the electronics, so occupying no incremental real-estate.

Small MEMS devices, coupled with deploying the MEMS above the electronics allow placing many MEMS cells on the chip, introducing the idea of redundancy.

Firstly, redundancy improves yield (and hence, cost) and long-term reliability: any defective MEMS cells may be dynamically switched out, with another, on-chip MEMS deployed in its place automatically.

Secondly, redundancy may be used to realise reconfigurability. Imagine a ‘fusion ASIC’ with sufficient intelligence to control an array of NanoEMS sensor cells deployed above it. Depending on the type of motion currently experienced or expected by the application, the ‘fusion ASIC’ may choose to deploy one or any number of these sensors, dynamically switching in or switching out the sensor cells as required, thus providing the application with the optimum desired motion-tracking performance at minimum power at each instant in time. In this sense, there is no need to drive a sensor at ‘rated power’ at all times to cover the few situations where maximum sensitivity or range are required. The performance and hence the power is dynamically tuned. In this scenario the following strategies are possible:

? Any combination of accelerometer, compass, gyro or pressure sensor may be activated dynamically as needed, the others powered down to save current drain
? 1, 2 or 3 axes of each of the motion sensors may be deployed dynamically as required by the application
? deploying 2 or more similar axes of one type of sensor allows signal averaging, which effectively reduces noise and improves accuracy
? performance (bandwidth, resolution or accuracy) may be dynamically adjusted to need by selecting different MEMS types and changing the active current in the MEMS
?   additional compass MEMS may be deployed to track ferrous materials and other magnetic sources (like loudspeakers) in the environment which would otherwise corrupt the heading reading. Similarly a compass device can be activated from time to time to compensate the long term drift of a gyroscope.

So ‘super-smart’ reconfigurability is achieved by deploying a ‘fusion ASIC’ and an array of tiny NanoEMS motion sensor cells on the same chip. According to the instantaneous sensing requirements of the application, the ASIC selects which sensor cells are to be used and configures their performance using the strategies described above: type of motion detection, range, resolution, tracking accuracy, noise, and current drain. The size, cost, scalability, flexibility, reliability and miserly current-drain promised by reconfigurability will result in a wealth of applications from exciting, new, motion-enabled, consumer applications including motion-enabled accessories, toys, games and sports/health equipment to low-energy sensor nodes in advanced remote sensor networks.

In Conclusion
Baolab’s nanoEMS concept has already demonstrated the lowest-cost single-chip 3D electronic compass device, but offers considerably more by virtue of its unique abilities to shrink the size of the mechanical elements of motion sensors, realise all motion sensors in standard CMOS using generic motion cells and integrate combinations of motion sensors with control and intelligence all on the same chip.

NanoEMS unlocks the barriers preventing integration of different sensor MEMS with the electronics, enabling an exciting motion sensor roadmap from lowest-cost single-chip discrete sensors, to multi-sensor nanoIMUsoffering up to 10 axes of motion sensing in the same tiny form factor as today’s discrete sensors, to smart nanoIMUs with integrated sensor fusion intelligence, and ultimately to super-smart reconfigurable sensor modules.

NanoEMS is the radical change in MEMS manufacturing long sought-after to enable the ubiquitous proliferation of motion sensing innovation.

Figure 1 NanoEMS technology
Figure 2 Fusion intelligence takes individual data streams provided by the 2, 3 or 4 types of motion sensors, and processes them into simple, standard motion vectors

Contact Details and Archive...

Print this page | E-mail this page