Using lis3dx to detect the acceleration threshold in a car accident

Kefei Yao · ‎2018-03-22

Posted on March 23, 2018 at 07:53

The purpose is to detect the car accident while ignore the normal acceleration variance during car driving.

I know there is a high pass filter in the lis3dx sensor, and the cut-off frequency can be changed by setting the following register (with actual cut-off frequency related to ODR setting) but don't know how to choose the best value. I guess there is experience value for the ODR and HPCF settings in car application. Anyone have relevant experience and would like to share it with me?

Legacy member · ‎2018-03-23

Posted on March 23, 2018 at 08:39

Hello,

I don't think anyone will write any typical values for a car accident here. There are so many variants! This is why car manufacturers and safety systems manufacturers do millions of hours of testing.

Are you sure you need the high pass filter? If you eliminate the gravity, you loose the sense of the gravity direction and that is not good. What if the car is rolled over a roof? Is that an accident?

Also, if a fender bender an accident for you? Most of the fenders are made from plastic and they soften the first part of the impact. That is why there are usually other sensors embedded, because time is critical in order to fire the airbags and seatbelts, etc.

However, if you are building just some hobby project, it might be sufficient to just experimentally set value and time threshold.

Here is some reading:

https://www.wired.com/2011/04/crashing-into-wall/

David

Kefei Yao · ‎2018-03-23

Posted on March 23, 2018 at 12:04

Thank you very much for your input and reading material. They are extremely helpful.

The real-time quality of the accident detection in my application is not that critical so no question about that. It's a product that help mark the video taken by the camera in case of car accident. Yes, a car rolling over a roof is definitely a car accident but it still can be detected by the lis3dx sensor even if the high pass filter is applied, can't it?

The reason I'm asking about the high pass filter setting is because the high pass filter can help attenuate the acceleration signals generated in normal car driving which is rather gentle in most cases. Also high pass filter can help remove the dc components.

I also find with the same threshold and duration setting, the lower HPCF setting can cause the interrupt trigger more easily in case of normal car driving. I deduce that the lower HPCF attenuate the signal generated in normal car driving less and cause the interrupt trigger easily.

? ?_3 · ‎2018-04-09

Posted on April 09, 2018 at 08:52

Excuse me, did you use the LIS3DH for impact detection? I tapped my table lightly with my device and the acceleration readings were over 8g.

Kefei Yao · ‎2018-04-09

Posted on April 09, 2018 at 09:06

What part are you using? 3DH? If so, the output of 3DH is left aligned in 2's complementary and it needs to be right shifted by 4 bit.

Kefei Yao · ‎2018-04-09

Posted on April 09, 2018 at 09:07

Check AN3308 from ST official for more details.

Legacy member · ‎2018-04-09

Posted on April 09, 2018 at 10:00

Also, this part of the

http://www.st.com/resource/en/datasheet/lis3dh.pdf

states resolution per LSB (''digit'') for each range:

David

? ?_3 · ‎2018-04-09

Posted on April 09, 2018 at 10:01

Thank you very much for taking the time to reply. The conversion of the values is in accordance with the manual, and they are correct. When flat, the X and Y axes tend to be 0, and the Z axis is about 1g. Then I use the device to gently knock down the table when I do the impact test. Some axes can detect 10g or so. It's a bit unintelligible. 10g acceleration shouldn't be very big. I'll tap it so that it can produce such a large acceleration. It's a little puzzled, so I'd like to ask you about it. Have you tried this kind of crash test?

Kefei Yao · ‎2018-04-09

Posted on April 10, 2018 at 03:43

I have not do this kind of test. Is the test designed by you or ST official? Anyway, 10g acceleration is not normal and there must be something wrong. Probably you can open a topic describing the issue in detail. The ST people will try to help you. They are quite reactive.