View Single Post
  #17  
Old 12-16-2002, 10:01 AM
BlueJeeper BlueJeeper is offline
Registered User
 
Join Date: Jul 2002
Posts: 524
Hi Frank,

I have a BS in Computer Science from the U of I. As well as having taken a few EE courses and labs, I had a specialization in applied physics, so those things help me stumble through this stuff a little better. That's really my only connection other than having a few friends on staff. I'm still here because my wife and I stuck around to work at a couple of the tech companies here. Are you connected to the industry in any way?

I really like your ideas about the display. I had not really thought about representing the information in other ways besides visual images of a Jeep or by raw numerical data. I can see now from your work that we might be able to come up with additional displays that would represent the information differently, which also might be more or equally as intuitive.

One thing I can think of right away, is that it might be useful to let the user define tolerances on their calculated COG. Fuel, human bodies, and other things that might shift will certainly change the COG a bit. Additionally, when off-camber, theoretically the orientation of unsprung mass as well as the reactive forces of the suspension on the frame may very well alter the COG or change rollover tendencies quite a bit?? Perhaps this might be estimated computationally based on the orientation of the axles to gravity and the frame? Just thinking out loud. The thing that sort of struck me when I saw your COG graphic is that in practice it might be useful to have a "red zone"; an x, y, and z tolerance component to the calculated COG that might be displayed (and trigger a light, piezo buzzer, etc.) so that the driver could know when their tolerance is exiting the gravitational track. Something simple, but something that might also prove quite useful in practice? I'm not sure.

If you would like to work on this then great! The more the merrier. It seems you want to work on the display end or with the COG algorithms?? I say go for it! I'm all for coming up with *many* different ways to represent the data so don't be bound by the fact that I mentioned a Palm device as a way to represent data. I would say as long as you could come up with some stuff that computes its results based on several inputs--angle of inclination of the body and frame (in 2 dimensions... front to back and side-to-side) with respect to the earth, and angle of inclination of the axles with respect to the earth--then we can probably somehow meet in the middle and go from there. Any of your input will definitely be massively useful!!

As for the accellerometers themselves; man, are these cool little devices. I have worked on the prototype some more in the past week in my copious spare time. I now understand them quite a bit better. The digital sampling period is set by a resistor and I need to go to Radio Shack this week to get some more resistors, because I currently have the resolution of the device to the point where there are about 250 datapoints per 90 degrees of tilt and I want better than that. With different resistors I can get thousands of points of resolution for every 90 degrees of tilt. One thing that introduces error into the system is acceleration in the direction of the axis being measured. This won't be much of a problem for the axle measurements, but it will be for the pitch measurement of the body and frame. With greater resolution and more frequent sampling I should be able to "average out" most of the error from the system; at least, hopefully, enough for the purposes of this experiment.

What I have found some people do to eliminate this error altogether is integrate a digital gyro to be able to subtract out this extra component of accelleration that we don't really want getting into our data. However, integrating a gyro with an accellerometer means adding lots of math to the solution (called "Quaternion") and I would like to stay away from that for now. Even with a gyro there eventually is a small amount of drift in the calculation so it is not perfect. Nonetheless, I'm pretty sure that just using the solid-state accellerometers should be good enough for our purposes with a little bit of averaging.

Rick
Reply With Quote