Relative versus absolute lateral acceleration?

FormulaRedline
Posts: 11
Joined: Thu Jan 03, 2008 5:05 am

Relative versus absolute lateral acceleration?

Postby FormulaRedline » Fri May 02, 2008 2:01 pm

Hopefully I'm not completely missing a built-in variable somewhere, but is there a way to calculate absolute lateral acceleration?

To clarify, the accelerometers on board the DL-1 move with the car and so always measure acceleration relative to the orientation of the car. Therefore, when you go around a banked turn, your measured lateral acceleration will be the same as when you are on a flat turn. The lateral acceleration in respect to the earth goes up, but that component is lost when the accelerometers bank. Can this acceleration easily be calculated in the software? Can I do it manually from speed and position data?

Thanks!

tristancliffe
Posts: 220
Joined: Thu Mar 15, 2007 3:41 pm
Location: Norwich, UK
Contact:

Postby tristancliffe » Fri May 02, 2008 4:48 pm

If you had some form of inclinometer measuring bank angle then it would be relatively easy. But the DL1 doesn't know where the g-force is coming from...

Perhaps you can derrive this from the GPS position and speed? Turning is an acceleration, which is a rate of change of velocity. And you know the longitudinal accleration (relative to the car)... I'm just thinking out loud, but if you can derrive acceleration purely from GPS position, and remove the long. component, then you'd be pretty close to absolute lateral accel. Of course, it doesn't work if your on a gradient longitudinally, but then you might be able to factor in GPS altitude to correct it.

Not something I'd want to attempt in a weekend, but I wish you luck!
Monoposto 2000 - Reynard 883 Toyota

FormulaRedline
Posts: 11
Joined: Thu Jan 03, 2008 5:05 am

Postby FormulaRedline » Fri May 02, 2008 5:19 pm

tristancliffe wrote:Perhaps you can derrive this from the GPS position and speed?


Yes you can. GPS based data loggers without accelerometers already do this. (like MaxQData) I'm wondering if the DL-1 software has these equations built in or not. I shouldn't need to subtract out anything, because I really am looking for the absolute earth relative accelerations. This should be directly calculable from the position and time.

faraday
Posts: 267
Joined: Sun Feb 25, 2007 1:18 am

Postby faraday » Fri May 02, 2008 5:36 pm

You are missing GPS lat and long accels, but you're not missing much when the logger is a DL1.
I presume the channels are calculated from the native low update GPS solution (Speed and change of heading, perhaps).

It is tempting to say you could calculate more accurately using 100Hz speed, but then you are using the accelerometers...
Body motion and drift are issues as well. Perhaps it would be best to measure the banking, measure body roll from a front on photo, do some vector revision...

Nevertheless, I must be missing something. What is the big deal with lateral acceleration?
It's the result of speed, which is a much cleaner measure and what we are trying to achieve, after all.
If you want to estimate forces in components or deflections, you'll need the angle of banking anyway.
You will need corner radius, which is a standard variable, but I think that is calculated from speed (GPS) and acceleration (that pesky accelerometer again).
You can make the software do a lot automatically, but if you are confidant that the position solution has created a valid 2-D vehicle path, more accuracy is likely from a large plot of the path through the corner to which you fit instantaneous radii with draughting tools.

What accuracy is required?

As the number of R-T products has grown and Analysis has been "unified", more variables have been incorporated. It might not be top of priorities, but it would be nice to see a current list of variable descriptions, and some notes on their accuracy and/or utility from different devices.
Perhaps such a list exists, but I can't find it.
:)

FormulaRedline
Posts: 11
Joined: Thu Jan 03, 2008 5:05 am

Postby FormulaRedline » Fri May 02, 2008 6:00 pm

faraday wrote:What accuracy is required?


Very little. The vehicle relative accelerations are much more important as far as squeezing out extra time at the track, and that's being read at 100Hz by some pretty precise accelerometers. I'm simply interested in seeing approximately what affect the banking has on the ability of the car to turn.

Are you telling me the corner radius variable is derived from the accelerometers, so it would not actually be correct on a cambered corner? If it was correct, using that combined with the speed would seem to be the easiest way to get what I'm looking for.

faraday
Posts: 267
Joined: Sun Feb 25, 2007 1:18 am

Postby faraday » Sat May 03, 2008 2:22 am

I honestly don't know in this case, but certainly in conventional DAQ systems.

R = velocity squared / lat. acceleration
The computational loads get greater from alternative means. Using conventional symbols in plane curvilinear motion, w (omega) is rotational velocity [radians/second], we also know:
R = v/w
w can be found from change of heading.

R-T Analysis gives us two headings as included variables: Standard Var - heading... GPS Var - GPS heading.
R-T's corner radius (my R) is quite nicely smoothed, so it doesn't come from an simple calculation using native GPS outputs. In the affordable low frequency hardware used, I believe anything from GPS other than position and velocity will be jerky, low resolution rubbish. If we were only tracking someone on a pushbike, this wouldn't matter, but...
IMHO, systems without accelerometers are conning their users with kinematic variables smoothed out in software. Just because a trace on the computer screen appears smooth and sufficiently detailed doesn't mean the underlying measurement system has the accuracy, resolution or precision (however you wish to describe quality) to create it.
:roll:
If we are going to make changes to try to improve performance, it's important to use measurements in which we have the most confidence.

FormulaRedline
Posts: 11
Joined: Thu Jan 03, 2008 5:05 am

Postby FormulaRedline » Sat May 03, 2008 3:59 am

Thanks for the help, but you seem to be missing the point of the question. I completely agree that this is not going to be accuracte data used to tune the car in the way of the actual accelerometer data. Hence why I have a DL-1 and not a lesser system.

I think I'll just take simple answer as "No, the R-T software will not calculate this variable for you."

I'll take a look at using corner radius and speed now that I'm home and have access to the data.

EDIT: Ok, here is what I came up with. It's obvious when you look at the data that it's no where near as accuracte as the accelerometer data. The greater the radius (i.e. straight aways), the less accurate it seems to be, especially during the transitions (when corner radius is infinite). But it gets the job done for now:

VAR_0015 * VAR_0015 / 5 / VAR_0052 / 9.80665 * VAR_0004 / ABS(VAR_0004)

Where VAR_0015 is speed in mph, VAR_0052 is corner radius in m, 5 is the conversion from mph^2 to (m/s)^2, 9.8 converts m/s^2 to G's, and VAR_0004 is lateral G's from the accelerometers (since corner radius is always posotive, dividing lateral g's by it's absolute value transfers the sign from the accelerometers without transfering the value).

faraday
Posts: 267
Joined: Sun Feb 25, 2007 1:18 am

Postby faraday » Sat May 03, 2008 5:18 am

Not missing the point, just adding commentary for anyone who wishes to read it. I understand your interest and ability to seperate fact from fiction.
At the end of the day, however, after we have tried to analyse performance more deeply, we might have developed a new technique or insight. Then it would be good to know: "Is this insight scientifically or statistically valid?"
We are working or playing in a very noisy environment were measurement is not black or white.
I applaud your use of "5" and not 5.0039 :)
There are many young engineers who would not. :(

I will be more blunt:

There are too many variables in R-T Analysis that are poorly defined and explained. It is not possible to give concise advice or make confident decisions with them while that state remains.
If I am correct in my assumption that correcting this is a low priority, it will not get done unless more users request that it be done. :!:

PS Are you sick of US/Imperial units :?:

FormulaRedline
Posts: 11
Joined: Thu Jan 03, 2008 5:05 am

Postby FormulaRedline » Sat May 03, 2008 6:56 am

faraday wrote:There are too many variables in R-T Analysis that are poorly defined and explained. It is not possible to give concise advice or make confident decisions with them while that state remains.


I agree. I have no idea how the corner radius is calculated...which means I don't really know if my equation above is even a useful one. It seems to roughly follow the trends of the measured lateral G's, with some extra acceleration through the banking. This is what I was looking for, I was just surprised it was not already built in!

I generally find it easier to work in standard units. I hate making conversions between feet/yards/miles.

Support

Postby Support » Thu May 08, 2008 8:42 am

Just open the variable manager, go to the GPS variables, and enable GPS lat accel. This is the lateral acceleration as calculated from the GPS data.

Point taken about the variables not being described very well. There is some information here:

http://www.race-technology.com/wiki/ind ... scriptions

But it is not complete.

I will also see if we can get a description of each variable put in to the variable manager. Without detailed knowledge of how the variable manager works I don't know how easy it is, but I will ask.

Regards,

Martin


Return to “General software support”

Who is online

Users browsing this forum: No registered users and 6 guests