Troubleshooting ADXL345 Data Inconsistencies Under High-Frequency Vibrations: BW_RATE Tuning & Interface Selection

[Jane Will]

I am working on a project that involves real-time motion tracking using the ADXL345 3-Axis Digital Accelerometer with an embedded system. While configuring it in I2C mode, I encountered unexpected data inconsistencies when the device is subjected to high-frequency vibrations. The raw output values fluctuate beyond expected noise levels, even after implementing filtering techniques like low-pass filtering and moving average smoothing.


Upon further investigation, I suspect it might be due to aliasing effects or improper bandwidth selection in the BW_RATE register configuration. However, even after setting the output data rate (ODR) according to the Nyquist theorem, the issue persists.


Additionally, I have considered SPI mode for better stability, but I am unsure if this will make a significant difference in high-frequency conditions.


Key Questions:

1. How can I precisely tune the BW_RATE register settings to minimize aliasing effects in high-vibration environments?
2. Would switching to SPI mode improve stability, or is the issue fundamentally related to sensor noise characteristics?
3. Are there advanced signal processing techniques (beyond standard filtering) that can be implemented at the firmware level to enhance data reliability?

References:

• ADXL345 3-Axis Digital Accelerometer Configuration & Applications
• ADXL345: Measurements Differ Between SPI 3-Wire, I2C, and SPI 4-Wire
• ADXL345 Sampling Rate Off in SPI Mode

Would love to hear insights from those who have worked with MEMS accelerometers in high-frequency motion applications!

 

[gruntguru]

Don't know this device specifically but I think the following information would be helpful.

• ". . . . subjected to high-frequency vibrations" - Is this a pure sine wave of known, fixed frequency?
• Bandwidth (data rate) to 3,200 Hz not what I would call high-frequency. Is your test frequency above 1,600 Hz?
• Applied to one axis only?
• If so, are you also seeing crosstalk?

 

[GregLocock]

Could you post a spectrum? I'd guess you set AA to at least 1.2*the highest frequency of interest. I think (from reading elsewhere) that this sets the sampling frequency on this chip or vice versa, which is a bit odd, normally I control them separately. Typically we sample at a very high rate and then decimate the resulting signal. This lets you actually see the unaliased HF noise, which can be useful. If you can't get enough attenuation from the AA filter you might need to consider a compliant base for the accelerometer.

Here's a bunch of links to signal processing books if you want some light reading

Introduction to Signal Analysis for Acoustics

www.eng-tips.com