Design Freescale MEMS Acceleration Sensors

The acceleration sensors are getting popular in air-bag, GPS navigator, HDD, game console and mobile phone. The latest acceleration sensors adopt the MEMS (MicroElectroMechanical System) technology, offers high reliability and sensitivity in a small size. I am involved in an impact triggered automatic camera with a MEMS sensor. The product is simple. It is an event recorder as a black box for after market. The camera module keeps caching the image/video into a DRAM buffer, until the sensor triggers it to save the buffer into the external SD card. By this means, the whole process of the traffic accident can be recorded as legal evidence. I selected ADI DSP for my camera module, and I am searching the suitable MEMS sensor for my project. Because the DSP has I2C/SPI and ADC, it is flexible to interface any kinds of acceleration sensors with analog output or digital output in any axis.

 

Freescale offers a very informative 1080-page reference manual for its sensors, which you may have a look at. I was quite surprised when I first realized the size of this manual. Don't be nerves. The document includes three sensor families from Freescale, including acceleration, pressure and electric field sensors. The acceleration sensor chapter covers overview, selector, data sheets, application notes and packages information. 

Since Freescale has many acceleration sensors in low/medium/high-g, it defines the numbering system for MEMS Acceleration (MMA) sensors in automotive and consumer classes. The automotive devices have MMA12XX/22XX/23XX/32XX/62XX. And the consumer devices have MMA745X/736X/734X/733X/726X/627X/628X/626X/623X. The designer should read the manual carefully and pick the correct one for his design according to the acceleration range, sensing axis, sensitivity, roll-off frequency, supply voltage and optional interface.

Acceleration Range

This parameter confuses me for a while. My product is similar to an air-bag system. When the high-g impact takes place, the air-bag controller should trigger the air-bag immediately. However, my camera is intended to record the small rub and impact for all kinds of traffic accidence. That means the sensors for my product should be triggered by much lower acceleration value. That means my design should take balance between sensor dynamic range and sensitivity. Finally I decided to pick a medium-g MMA for the initial design. Additionally, I add a push button to trigger the recorder.

Sensing Axis

I have only one camera lens in this project, so one axis accelerometer seems enough. However, if the product is upgraded to detect the other accidents like side impact, roll-over, hitting animal/human, then it should use two and three axis sensors with more lenses. Because all of these MMA sensors are SMD, the designer must mount the sensors and let them work in a correct axis. In another word, the designer should select the sensor with correct axis. Freescale offers one axis MMA in SO package for X and Z axis. X axis MMA works in the horizontal direction of the PCB, while Z axis MMA works in the vertical direction. Since my camera is installed vertically, so I pick the Z axis to sense the vertical direction of the PCB, i.e. the impact direction along with the car body.

Sensitivity

The sensitivity of a sensor is related to its analog output voltage and acceleration range. The formula is:

Sensitivity = output voltage range / acceleration range; 

If we pick a 5V sensor with 40g (-40g to +40g), and its analog output is about 4V. The result sensitivity is 4000 / (40 * 2) = 50 mV/g. You can refer to the resulting sensitivity of MMA3201D as cross check. 

Additionally, the multiple-axis MMA has different sensitivity for each axis. The system designer should consider the mounting and sensitivity for each axis as well.

Roll-off Frequency

Each MMA sensor has an internal filter to eliminate shakes in unnecessary frequency range to preserve pulse shape integrity. That means the designer should investigate the background of the product. Since the internal filter of most of the MMA sensors are fixed, you can consider to add extra filters with circuit, mechanical shockproof or digital filter implemented in software. 

Supply Voltage

My system controller works on 3.3V, while most of the medium-g MMA works on 5V. So I have to solve the sampling issue in the design.

Interface

Most of the analog sensors offer output voltage represents the real acceleration value in the MMA sensor. The system controller can measure the voltage with ADC. If the connection track between the MMA and the system controller is very long, the designer should add a proper filter between two parts. Of course, the RF link is even better if your MMA mounting location is too far away from the system controller. You can check out the related AN1611 for detail information. Some parts have embedded the ADC in the package and allow the system controller to read out the value by the popular SPI bus. The digital interface is much easier, but its price is higher.

Reference Designs from Freescale

Freescale offers many reference designs for its MMA family, including the big reference manual and other application notes. Among these documents, AN1611 describes all necessary design details about the impact measurement with a toy car model, including schematics, algorithm flowchart, and assembly code for the project. The other one, AN3380 is much in detail, which introduces Freescale MMA7206Q low-g sensor for brake detection, and uses MM908E625, a microcontroller with power control module for dedicated algorithm and LED control. AN3380 analysis the application in detail, and introduce the solutions for mechanical anti-shake and digital filter implemented by software. 

All of these application notes imply us a fact, that the success factor of an MMA sensor project is actually the system design, especially in software algorithm. Even if we select the suitable sensor, interface the sensor correctly, we still need investigate its application environment, implement suitable algorithm to eliminate the unexpected noises and accidental influence and then detect the correct pattern for the desired events.

Reference