Out project adopts adxl355 for vibration detection. The problem comes up when sensors are running for a while，but most of time they can work well. If data of the sensor goes wrong , what we should do is to cut down the power and power up to make it normal.
The Platform we use is STM32F429 with SPI interface and 2.8M clocks speed. The Program is based on HAL lib. The operation is to read 11 bytes continuously from the register 0x06.
The setting of sensor is ODR in 1Khz. And HPF is turned off, rang is+-8G, other settings are in default.
The testing environment is that 7 sensors are fixed on the robot arm, and the arm will run periodically. It is normal for the sensor to run for a period of time after power on, but there may be data abnormalities after a long -time running. Such problems would not happen in the sensors always, but it may appear from time to time.
Abnormal data is very regular, it appears every 32 normal data. That data looks like it adds a large amount of positive and negative value to the normal data. Even the arm totally stop, the phenomenon is still here. And this situation only appears on the X axis, however the other axes is normal.
Here is the images.
When this problem occurs, we use a logic analyzer to capture the data in the SPI and find that the data on the MISO behaves as described. We suspect that there is a problem with the sensor, and such problems often happen.
Did you find the problem? Also we are using adxl355 for our vibration measurements. We use many adxl355 sensors on different locations. Sometimes, we get very strange temperature data.
For example, outside temperature is nearly 25 degrees celcius. Sensor get stable data for 1 hour and suddenly it gives around -20 degrees celcius. For nearly 15 minutes temperature data is -20, 0, -13, -18 ..... Then it goes back to normal with around 25 degrees.
When we see the wrong temperature data, we also investigate the vibration data. Vibration is also a bit shifted. We know that there is no vibration change but values are a bit shifted.
We are still investigating but we could not find anything.
Hi vvZhong, faaltech,
Sorry for having overlooked into this question. May I know the estimated magnitude and frequency of vibration in your applications?
Subjecting ADXL355 in a high vibration environment could result into stiction as the anti-stiction material wears out.
Thanks for the fast reply. For our case, vibration is really low. For the problem that we have seen above, our product was in our lab and it was on the table. Even after integration, it did not see any shock or heavy vibration. Below is temperature and velocity (calculated from accelerometer data of adxl355).
As you can see, at the beginning everything is perfect. But suddenly temperature gives bad results. Also, vibration rises. But we are sure that there is no vibration or any change around the sensor. Everthing is still. This is a test case.
We are investigating whether it is a power, EMI/EMC, etc. There are same sensors on the table connected to same power source. We did not see that error on others.
Thanks in advance
I'm also experiencing abnormalities that are specific to the X axis at ODR 1 kHz. There appear to be some erroneous sign reversals.
ADXL355 vibrated with a 0.9 g RMS (1.3 g peak) sine wave: X axis, ODR 1 kHz, range 8 g, SPI read from FIFO:
Same as above, zoomed in with reversals marked in red:
The Y axis and Z axis do not have this problem even when rotated to be oriented in the same direction of vibration. ODR 4 kHz does not have this problem either. Changing the HPF and lowering the SPI clock speed from 10 MHz to 1 MHz made no difference.
Perhaps there is a problem with the digital decimation filter that is incorrectly flipping a bit.
I use the analog ADXL354 and i do not see that problem you are observing. Will be e digitalization problem.
I encounter other problems with noise, but maybe i pretend to much from this sensor.