SparkFun Forums 

Where electronics enthusiasts find answers.

Tips and questions relating to the GPS modules from SFE
By russnagel1
#135460
Looking at the information on the webpage (sparkfun) regarding the accuracy of some of the modules (EM-408) indicates an accuracy of 10 meters. I suspect the actual accuracy is much better. Can I get a few comments on accuracy from actual experience and what model you are using? I have read the turorial.

I need to buy a unit for my bot as soon as possible.

Thanks
Russ
By fll-freak
#135492
Although the quality of receiver plays a part, other factors are often much more important in the accuracy of the solution. These include:

"Constellation/Geometry" This is the geometry of the satellite vehicles (SV) in the sky. All the SVs in one quadrant in none in others will give you an awful solution. The orbits of the SVs are designed to minimize this problem, but sometimes the geometry is poor.

"Number of SVs". The more the merrier. Three is bare minimum for a 2D solution. Four for a 3D solution, and five SVs if you want to be able to begin to have an idea of what your errors are. With that extra fifth SV, the problem is over constrained and you can solve the equations multiple ways. This gives you an idea if one range equation should be discarded because it is "out of family".

"Poor horizon". If you live in an urban area where the horizon is obscured by trees and buildings, you will not see SVs that are low on the horizon. These are the ones (if they are more than 10 degrees up) that give you the best horizontal positioning error.

"Multipath" If your receiver is around items that will reflect the GPS signal (buildings, cars, ...) the timing signal from the SVs will arrive a multiple times (think echo). This can play havoc with a solution. If you have a military GPS that uses multiple bands (L1,L2) you can avoid multipath to a large extent.

I have a GPS antenna that is bolted to the roof and has not moved in 4 years. The variability or accuracy of the solution changes on a day by day hour by hour basis. Could be as much as 10 meters off on rainy days with a poor constellation. Could be as little as a meter off on good days. Some days it is perfect (but only due to probability)!
By russnagel1
#135510
Thanks Skye, I had not considered all the variables. My primary usage at this point, is for a robot that will be outdoors, on sunny days with a clear view of the sky. A couple of meters accuracy would be great.

Russ

P.S. I'll check out your blog
By fll-freak
#135553
I will add one more factor and that is the antenna. Not all antennas are created the same. Cheap "patch" antennas often do not have a very good reception pattern. The will not see SVs very well bellow 30 deg of the horizon. A larger and more expensive helicoil antenna will pick up low SVs better. Pretty much the more you pay the better your antenna performance. Sometime it can make a difference, other times it is not worth it.

And one more factor. Running your GPS near electronics that have harmonics and the GPS frequencies will also play havoc with a solution. I have a GPS that reports it is being jammed whenever a certain piece of test equipment is turned on and closer than a few feet from the receiver. Not the antenna that is 60 feet away on the roof, but the receiver. If you have such a condition, a Faraday shield might help. Just remember the antenna has to be on the outside of the cage!
By mbavaro77
#135759
Hi Russnagel,

If your bot needs a couple of meters accuracy most modern GPS receivers will be able to achieve that, provided that you have clear sky visibility (low multipath).
I would suggest to try a newer model than a SiRFStar III: MTK 3329, uBlox 6, CSR GSD4e, Skytraq Venus 638LP, etc..

For sub-meter accuracy you need to look at DGPS and RTK so the choice narrows quite a bit (even though you may still find low-cost ones).

Cheers,
Michele