So the electronics deck on my RC car is hard mounted to the chassis. When the robot vibrates so do all the electronics including the IMU. This points tons of noise into all the channels of the IMU. I have talked to a few of my mechanical friends and they all have come back with various suggestions including floating the IMU in a bottle of glycerin. Like that would work on an RC car!
I think I have a few ideas on my own. First I think I need to stiffen the electronics deck. I could see it flexing like crazy under the vibration. Then I need to modify the hard mount system with perhaps a rubber band mount. And lastly I am going to need some kind of viscous damper. Here is where I am stuck. Perhaps I can find a VERY light duty air filled shock absorber.
So here is another job for the todo list.
Some thoughts that, I hope, will be of help. Perhaps you’ve consider all this already, if so, nevermind 🙂
Noise only matters if one cannot retrieve the desired signal out of it. What’s the frequency of the noise versus frequency of the normal signal? Is there a frequency range for low pass filtering that attenuates noise but not the signal? What do the quadcopter guys do about vibration?
One good question to ask is how much noise is too much? How much does gyro noise impact the final heading estimation, for example? Considering that integration acts as a filter. And considering the steering resolution of the vehicle, etc.
Put another way, will the reduction in noise result in improvement in heading/position estimation that is worth the effort and time?
It’s not hard to try to go too far in solving problems, cutting into the time needed to get the basics working.
Noise can be attenuated (possibly with the signal as well), its source can be reduced (I use wall mounting putty to balance tires), and its frequency spectrum can be changed (going slower or faster, stiffening mounts).
After talking with various mechanical gurus, I have been contemplating many of these questions and issues myself. It is hard to give up the desire for an ultra clean signal even if a lesser on will give the same results.
So I have a new set of tires on order. These should be in round and much of the vibration should go away. Next I plan to stiffen the electronics deck as I noticed it oscillating significantly. After that, I plan on just trying a few simple rubber grommet mounting systems to isolate the worst of the noise. After that I will fall back on the pass band filters on the gyro and accelerometer parts.
I guess this is not like at work where someone else has figured out all the main and sub harmonics of the system and designed an isolation system to leave me with a near perfect signal. I will just have to deal with some crud. Oh well.
http://technicaladventure.blogspot.com/p/hefnycopter2-gyro-noise-filtering.html
This is a study of filtering out Gyro filter from Quadcopter by analyzing real data.