Positional accuracy

[corey_dyck]

I use a DL1 for autocross in an A-Stock Honda S2000. The DL1 is mounted just behind the passenger seat in the trunk area, the antenna is mounted on a flat space ahead of the trunk lid. All runs were taken with the top down so there is minimal GPS blockage.

My concern is a large amount of inaccuracy of position. As you can see in the screenshot below, the runs do not overlay very well at all. Is there something that can be done to improve the accuracy or is this an accepted amount of error?

A friend has an identical unit in a Pontiac Solstice, he has similar or worse errors on his runs. We were hoping to use this to see if taking different lines pays off in terms of time, but the positional error is such that you can have a 20' error at any time! Note that there are no trees or other cover anywhere near this (or our primary) site. This error has presented itself on multiple days from sunny to snowing.

Any thoughts?

Screenshot:
http://members.shaw.ca/ckdyck/axvids/PCA_event.jpg

Data log files:
http://members.shaw.ca/ckdyck/axvids/RUN___03.RUN
http://members.shaw.ca/ckdyck/axvids/RUN___04.RUN

Onboard video of the 4th run:
http://www.youtube.com/watch?v=zIMJAUSrb_0&fmt=18

A second question is in regards to graphing parameters. As shown in the screenshot, the steering angle trace covers less than 50% of the white space. Is there any way to make the max/min of the graph match the max/min of the reading? The blank white space effectively cuts the visible resolution in half, making subtle changes invisible. I've tried changing the min/max values in the variable setup with no change in the white space.

Thanks in advance for any help/advice you can offer!

[Support]

To get the graphs with the scales you want, use a standard XY graph. If you open two graphs, you can use one graph in interactive mode, and use the other graph to highlight the data you wish to view. You can then set the graph up with whatever max and min values you want.

It is very difficult to see how much error is on the plots without knowing how closely the two data sets were repeated. Do you have a feel for how accurately the two runs should be overlayed?

Martin

[corey_dyck]

To get the graphs with the scales you want, use a standard XY graph. If you open two graphs, you can use one graph in interactive mode, and use the other graph to highlight the data you wish to view. You can then set the graph up with whatever max and min values you want.

This is a little disappointing to hear, the Quick Graphs are such a nice area to check general trends quickly. I feel this really limits it's usefulness without adding any value. Consider this my request to remove as much of the dead space in Quick Graph above and below the graphed values in future software packages!

It is very difficult to see how much error is on the plots without knowing how closely the two data sets were repeated. Do you have a feel for how accurately the two runs should be overlayed?

The most troubling area is just to the right of "Sector5" in the screenshot. This is a relatively narrow gate that can be seen at the 0:21 and again at 0:41 in the video. To take the path suggested by the GPS data would put me outside this gate.

Would you consider this error reasonable or do I need to do something differently?

[Dan in Saint Louis]

Could it be that

1) Consumer-grade GPS is limited to 20 - 30 feet accuracy anyway, and

2) A car that leans in the turns distorts the lateral g reading, so the fine-tuning of the GPS data that can be accomplished with help from the accelerometers is also distorted?

I think the track maps are to give a race engineer a general picture of where the car was when a certain event took place, not to thread autocross gates.

[Support]

For large sections of your run the two lines are almost exactly on top of each other, showing a spread of a coupe of feet or less. In other areas they do deviate from each other by as much as 30 feet. If you colour your track map based on GPS positional accuracy, you will see that through this section on Run3 the positional accuracy is not very good, hence the difference in positions. So the majority of the time the difference is probably a few feet, on the sections where the is more of a difference one run is showing as having a poor positional accuracy.

As for why the GPS accuracy dropped off just there, I don't know, where is your antenna positioned?

Martin

[corey_dyck]

Thanks Martin. The antenna is just ahead of my trunk lid, it can be seen in this picture: http://solosnapshots.com/gallery/main.p ... alNumber=1

Any thoughts regarding the white space? I just processed yesterday's data and got this:
http://i26.tinypic.com/2rz9ogx.jpg
The time slip is almost useless without exporting and re-graphing in Excel. I really have to admit ignorance, maybe there's something I'm missing.

[Dan in Saint Louis]

Any thoughts regarding the white space?

I guess you mean the rather compressed vertical axis? I'd guess that is trade-off for using "quick graphs."

Once you have built a customized X-Y graph, it is easy to re-use -- and with a couple of mouse clicks you can adjust the scale of any variable. Just think of "quick" and "customized" as being choices you get to make.

[corey_dyck]

[quote="Dan in Saint Louis"Just think of "quick" and "customized" as being choices you get to make.[/quote]
Well put, thanks!

[Dan in Saint Louis]

Corey, what I have done is to define -- once -- the graphs that I want; then named and saved them. Now each time I open "Analysis" I am asked if I want to restore my last screen layout. "Yes" gets me right back to the same graphs, and I can open any RUN file to view!

If I fiddle with the scales etc until they are no longer useful, the "good" ones are saved under the name I chose and thus never lost.