Condition monitoring, minimizing machine downtime
Vibration detection and analysis, suitable for MEMS accelerometers
ADXL1002, capable of detecting vibrations from DC to 11kHz


Today, monitoring the condition of machinery and technical systems utilizing devices such as gears, motors, and generators is a critical task. Preventive maintenance is becoming increasingly important to minimize the risk of downtime, not only in the industrial sector but in all fields that use machinery. This is achieved through the detection and analysis of vibration patterns in machinery.

For example, vibrations generated by a gearbox are typically detected at multiple shaft speeds in the frequency domain. If inconsistent patterns are detected at different frequencies, this may be a signal indicating wear, imbalance, or loose component connections.

Micro-Electro Mechanical Systems (MEMS)-based accelerometers are often used for frequency measurement. Compared to piezoelectric sensors, MEMS sensors have high resolution and excellent drift and sensitivity characteristics. In addition, not only is the Signal-to-Noise Ratio (SNR) excellent, but it can also detect low-frequency vibrations at levels close to direct current (DC).

By utilizing Analog Devices (ADI) ’s ‘ ADXL1002 ’ low-noise, high-frequency MEMS accelerometer, it is possible to design highly linear and low-noise broadband vibration measurement solutions. These solutions are suitable for any application requiring a high dynamic range of up to ±50g and a frequency response from DC to 11kHz, such as bearing analysis or engine monitoring.

The figure above is an example of a circuit. The analog output signal of the ADXL1002 is passed through a 2-pole RC filter to the SAR (Successive Approximation Register) ADC ' AD4000 '. The AD4000 converts analog signals into digital values to enable additional signal processing tasks.

The ADXL1002 is a high-frequency single-axis MEMS accelerometer that provides an output signal passband extending beyond the sensor's resonant frequency range. Consequently, it can detect frequencies exceeding the 3dB bandwidth. The ADXL1002's output amplifier supports a signal bandwidth of 70kHz. It can directly drive capacitive loads up to 100pF. For loads greater than 100pF, a series resistor of 8kΩ or greater must be used.

An external filter at the output of the ADXL1002 is required to remove aliasing noise from the ADXL1002's output amplifier or other internal noise components. For example, such noise can occur due to the coupling of the internal 200kHz clock signal. Therefore, the filter bandwidth must be implemented accordingly.

With the conditions in the figure above (R1=16kΩ, C1=300pF, R2=32kΩ, C2=300pF), attenuation of approximately 84dB can be achieved at 200kHz. Additionally, the ADC sampling rate must be selected to be higher than the amplifier bandwidth, for example, 32kHz.

The designer must select the power supply voltage of the ADXL1002 for the ADC to be suitable for the corresponding reference, because the output amplifier is proportional to the power supply voltage. In this case, the power supply voltage tolerance and the voltage temperature coefficient (often connected to an external regulator) between the accelerometer and the ADC interact to offset the implicit errors associated with the power supply and reference voltages.


Frequency Response and Mechanical Considerations for Mounting of the ADXL1002
The frequency response of the accelerometer is the most important characteristic of the system, and the gain begins to increase from a frequency of about 2 to 3 kHz. At the resonant frequency (11kHz), the gain reaches a peak value of about 12dB (about 4 times) at the output voltage.

The ADXL1002 has an output (OR pin) to indicate measurement range overshoot (overrange). If severe overrange occurs, the built-in monitoring function generates a warning.

There are a few things to keep in mind when properly mounting the ADXL1002. To prevent measurement errors caused by vibrations of the circuit board itself, the sensor must be mounted on a solid point on the board. This causes the circuit board vibrations to exceed the mechanical sensor's resonance frequency, so they are effectively not detected by the accelerometer.

Fixing the sensor at multiple points and using a thicker board can also reduce the impact of system resonance on sensor performance.


This article is a summary of the article titled 'How to Build a MEMS-Based Solution for Vibration Detection in Condition Monitoring' written by Thomas Brand, a Field Application Engineer (FAE) at Analog Devices (ADI).