geez
[Top] [All Lists]

Re: Calibration and flipping data?

To: geez@autox.team.net
Subject: Re: Calibration and flipping data?
From: Byron Short <bshort@AFSinc.com>
Date: Wed, 08 Mar 2000 19:49:37 -0700
It's pretty hard to get 1.16 accelerating, so yeah,
something's wacky.  Take a look at your calibration string. 
If you calibrated gas and brake backwards you'll have
numbers around 2300 or so in the last two positions.  

Here's how the calibration string works:

1B37A12B129B108

The first 1 is for calibration method #1, the only one that
exists so far.

After the 1 there are four sets of calibration
alpha-numerics, one for each direction.  The four sets in
our above example are B37, A12, B129, and B108.  The sets
are for left, right, gas, and brake in that order.  The A or
B stands for Above or Below, meaning that the 1 G level is
Above or Below the nominal values.  The nominal values
happen to be 1250 in each direction, but in real life
G-Cubes vary and are usually in the range of about 1050-1350
counts per G.  
In other words, the G-Cube measures G's at a native
resolution of between 1/1050 to 1/1350.  Anyway, the idea in
the numbers is that if your cube gets only 1150 counts per g
when you lay it on it's left side in calibration, then your
first set of numbers will be B100, or 100 counts Below the
nominal value of 1250.  

So because of all of this, it's easy to see if you
calibrated backwards just by looking at your calibration
string.  If you calibrated backwards it might read something
like 

1B37A12B2242B2284

The big 2000-2600 type numbers tell us that the readings
were off of the expected level by essentially 2g's, since it
expected +1g, and got -1g, for instance.

Okay, that was a long explanation, and probably more than
anyone wanted to know.

So I'll answer your other question much faster.  You asked
if you were to calibrate with the G-Cube against a vertical
surface, or flat on a desk, where the built in 1degree tilt
would effect it.  It doesn't make much difference.  The
reading difference between 90 degrees and 89 degrees is
determined by taking the sine of 89 and subtracting it from
the sine of 90.  The sine of 90 is of course 1.  The sine of
89 is 0.99985.  So whether you use the bottom or the side
for your flat side, the difference between the two will be
about 0.00015g.  We can't measure that small of a change.

--Byron


"Darren P. Madams" wrote:
> 
> I think I may have figured out what our problem was on sunday.
> 
> First: when you calibrate, are you supposed to put the cube down on each
> edge (including the front and the back)?  They're all sloped inward.  Or
> are you supposed to put the bottom up against a wall or something and make
> it a perfect 90 degree angle?
> 
> The problem we have (and I think what caused our autostart to goof) was
> that we were getting crazy high acceleration numbers (peak 1.16g) and no
> deceleration (peak .33g).  Not likely in an MR-2, methinks... even it it
> was Brian Priebe! :)  Peak Left was 1.27g and peak right was 1.17...
> sustained was 1.12 and .99 respetively so almost believable.
> 
> I'm thinking that maybe I reversed the top and bottom part of the
> calibration... so accel geez are coming in as decel?
> 
> The course started out with a not too hard acceleration into a tight right
> left 90 degree kink where the lights were.  It seems to me that the data
> starts being collected right at the braking point for that turn which would
> support that theory.
> 
> Any way to flip the data around to see if that was indeed the problem?
> 
>         --Darren

<Prev in Thread] Current Thread [Next in Thread>