What the experiment involved
Gibson says he bought the F5t at an e-waste shop and remembered that some models shipped with a u-blox 6 GNSS receiver. He expected to compare it against the location services on modern tablets. A third-party copy of the F5t user guide supports the general idea. It says the optional mobile broadband module includes a GPS receiver that works with most GPS navigation and mapping applications. That manual doesn't confirm which chipset his unit used. The u-blox 6 identification is his own recollection and observation.
His setup was simple:
- He placed the tablet on a roof and connected to it over Windows Remote Desktop.
- VisualGPSView confirmed that the receiver was producing a standard NMEA data stream.
- He then moved to u-blox's u-center software. Its views cover satellite sky position, signal strength, timing and position plots.
He is also right to separate GPS from GNSS. GPS is the US system. GNSS is the umbrella term that also covers Russia's GLONASS, Europe's Galileo, China's BeiDou, India's NavIC and Japan's QZSS. A given receiver may use only some of them. The author says his old chip didn't listen to every constellation.
Compatibility note: u-center is a Windows tool, but it's still worth checking its version if you try this yourself. Per the research notes, u-blox's newer u-center 2 targets newer receiver generations. An older u-blox 6 module would need the original u-center. That is historical compatibility information, not a guarantee for every current Windows install.
How signals become a position
The explanation Gibson walks through is sound, and GPS.gov backs up its core ideas.
- Satellites broadcast, receivers listen. GPS satellites send timing codes plus orbit and clock-correction data. The receiver never transmits back.
- Delay becomes distance. The receiver lines up its own copy of each code with the incoming one to measure how late the signal arrived. GPS.gov's trilateration lesson frames this as distance equals rate times time. The rate is the speed of light, 299,792,458 meters per second.
- The receiver's clock is wrong. Its cheap clock isn't synced to the satellites' atomic clocks. Every measurement is therefore off by the same unknown amount. Engineers call these skewed measurements pseudoranges, not true distances.
- Four unknowns need four satellites. The unknowns are latitude, longitude, altitude and receiver clock error. Each satellite adds one equation, so a basic 3D fix needs at least four.
The scale of the timing problem is worth stating. One microsecond of error is roughly 300 meters of signal travel, or about 984 feet. A separate encyclopedia entry makes the same point with a bigger number: a clock error of only one-thousandth of a second causes a position error of almost 200 miles. This is why Gibson's clock-bias graph is more than a curiosity. The receiver has to solve for its own clock error continuously.
Note that "four satellites" is a minimum for a solution. It doesn't guarantee a precise one. More satellites add redundancy, and spread across the sky helps even more.
The graph and the scatter plot
Gibson reports that the clock-bias trace sloped toward roughly minus one millisecond and then jumped back toward zero twice. He reads this as the receiver managing its own clock while positions stayed valid. That is his interpretation of an on-screen diagnostic. Without the log or the specific u-blox message, it's best treated as an observation, not a verified account of how the chip works.
His stationary test is more concrete. He reports:
- 690 valid position samples over 11 minutes and 29 seconds
- six to eight satellites in use, with no recording gaps
- about 49 feet of east-west spread and 32 feet north-south
The tablet didn't move, yet the reported position wandered. That is a useful lesson. A valid fix is an estimate, not a fixed point.
These numbers are one run on one roof. There was no surveyed reference point, no comparison device and no repeatable protocol. They are not a measure of the F5t's accuracy. They also don't show that it beats or loses to a modern tablet. Gibson mentions comparing against software-based location sensors, but the write-up includes no side-by-side test.
Why a stationary fix wanders
GPS.gov says that what you receive depends on satellite geometry, signal blockage, atmospheric conditions, and receiver design features/quality. It lists reflections off buildings or walls ("multipath") among the common causes of degraded accuracy. Gibson's list is consistent with that:
- Atmospheric delay slows signals slightly as they pass through the sky.
- Reflections from metal roofs or water make signals take a longer path.
- Poor geometry hurts when satellites are clustered. Satellites at wide angles to each other do better.
A roof is not automatically a good test site. Parapets, vents, metal surfaces and nearby buildings can block or bounce signals. Gibson doesn't describe his roof in enough detail to judge.
If you want to repeat the experiment, the reasonable approach is:
- Pick a spot with a wide, unobstructed view of the sky, away from buildings and dense trees.
- Leave the tablet and its orientation unchanged during the run.
- Keep it dry.
- Log for long enough to build a scatter plot.
- Compare the points against a known reference, if you have one.
Raw fixes versus the map dot
Gibson also contrasts raw receiver output with what Google Maps or Waze show. Those apps use map matching, which snaps a series of positions to the most likely road. That is a general technique, not something he tested here. It explains why the marker on a driving map looks so steady. The scatter plot has no such smoothing layer. It's also a reminder that map errors and GPS errors are separate problems.
Timing is the bigger story
Gibson ends on timing. GPS.gov says that besides longitude, latitude and altitude, GPS provides a fourth dimension: time. It says receivers decode the signals and effectively sync themselves to the satellites' atomic clocks, which lets users determine the time to within 100 billionths of a second. The agency names communications systems, power grids and financial networks among the infrastructure that benefits. That matches his point about cellular networks, grid monitoring and transaction timestamps.
One caveat: serious timing applications use specialized receivers. A decade-old tablet's clock panel doesn't prove it could serve as a timing reference for anything else.
Why Windows readers might care
This is a hobbyist project, not a how-to, but it carries a few practical takeaways:
- Old Windows tablets and rugged laptops with real GNSS hardware can be cheap learning tools. Check the hardware first, since not every configuration includes the receiver.
- Windows' "location sensor" entry in Device Manager doesn't tell you whether dedicated GPS hardware exists. Gibson says many modern tablets rely on software estimates, for example from nearby Wi-Fi networks.
- Seeing raw NMEA data and a sky view is a good way to understand why your phone sometimes puts you in the wrong place.
The tablet itself is a relic. The takeaway is that the dot on a map is a statistical estimate, built from clocks in space and a very imperfect clock in your hand.