[aplusbi]

He's probably using a GPS watch which is very convenient but error prone. They'll frequently get misreads (where they briefly think you are miles away from where you actually are) and accumulate small amounts of error over time (the granularity isn't great, so sometimes it will continue to increment distance even if you are standing still).

Really reliable GPS trackers are big and heavy, and for someone attempting speed records just not worth the weight. And even with reliable GPS tracker there will still be doubters who will claim the data was faked.

[logicallee]

How does error-prone GPS work on a technical level? Based on a description of GPS I would think it's kind of an "all or nothing" thing (no signal, or exact calculated position)

https://en.wikipedia.org/wiki/Global_Positioning_System

I just don't get how you can get "low-fi GPS". What happens exactly? Either you receive the signal or you don't, and any bounces off the ground or something will just add a couple of meters.

[jrockway]

https://en.wikipedia.org/wiki/Error_analysis_for_the_Global_...

The largest source of error is ionospheric delay which changes throughout the day. This can be corrected with additional ground measurements (DGPS, WAAS) or a second receiver on a different frequency. GPS watches do WAAS but nothing else.

(As an aside, I have been unable to get my hands on any modern multi-frequency GPS receivers. If someone has a link to one that I can buy with a credit card, I would love it! But that rules out the industrial ones that are "call for a quote" because if you have to ask, you can't afford it ;)

Other sources of error are inconveniently positioned satellites ("dilution of precision"), multipath reception, and the intrinsic quantized nature of the code being transmitted by the satellite. (This last one can be mitigated by tracking the phase of the carrier signal independently from the navigation code, however. Not sure if GPS watches do this or not, but I suspect they do. A comment below implies they don't have the power budget for the extra CPU cycles, though, which might be true.)

An ideal model of GPS makes it seem like there are four unknowns, but in reality there are many more unknowns. You don't know exactly where the satellites are (ephemeris errors). You don't know if the clock on the GPS satellite is right. You don't know the speed of light through the atmosphere on the exact path between you and the satellite. You don't know the exact time at the receiver's location (internal oscillator error, etc.). You don't know that the signal from the satellite took a straight line route. You don't necessarily have an optimal view of the sky to be able to hedge against the other errors (HDOP/VDOP/TDOP). As a result, you don't get a pinpoint location in the real world.

(One last thing, it turns out there is a reason why GPS receivers give you separate speed and position outputs. Speed can be determined independently from position by monitoring the doppler shifts on the received signals. Many receivers do this, which is why your speed can be correct but your position wrong.)

[mojoB]

> As an aside, I have been unable to get my hands on any modern multi-frequency GPS receivers. If someone has a link to one that I can buy with a credit card, I would love it!

Try SwiftNav or Tersus. I have not used either of them so I can't comment on the quality, but $1-2k for an RTK setup is pretty great. They are both more 'startup' than traditional suppliers.

uBlox F9 looks good but isn't quite available yet. They have a solid line of single frequency receivers so I'm looking forward to seeing pricing.

I have briefly tested some Unicorecomm from China, and they work well, but they miss your online store requirement, and the documentation is nonexistent.

[moron4hire]

GPS satellites mostly just send out a timing signal. The receiving device then measures the differences between timing signals from the satellites it can see. This calculation is not simple, and lower power devices don't have the processing budget (or power budget, which is practically the same thing on a battery powered device) or don't have implemented as many error correction methods. For example, the signal echoing off of rock faces (or more often, buildings, for a more familiar environment), can make it look like one satellite is more than one satellite, which will obviously not give you the right location.

[jhayward]

> any bounces off the ground or something will just add a couple of meters.

I have raw data captures showing a multipath signal being more than 1/2 mile off from a hiker in steep mountain terrain. It's not as simple as you think.