RSS
Want to fly your plane or drive your car using GPS signals, but finding that your receiver just isn't accurate enough to make things work? Well, MAKE subscriber Bruce Mueller writes in to point us at an impressive solution: an open-source real time kinematic GPS receiver. Researchers Tomoji Takasu and Akio Yasuda of Tokyo University developed the RTKLIB library to perform the RTK-GPS calculations, and then ported the whole thing to run on a low-cost beagle board and commodity GPS receiver. Want to try it out? Full source code, circuit layouts and instructions are provided on their site.
So, how does it work? A GPS receiver normally works by measuring the delay between an internally generated signal and one received by a satellite. This specially crafted signal makes it possible for the GPS receiver to find and latch onto the satellites signal, however it's wavelength limits the accuracy of the receiver. The real time kinematic system gets around this limitation by measuring the phase delay in the carrier signal. Because this signal has a much sorter wavelength, it is possible to make a system that is accurate to the centimeter.
I use RTK GPS every single day for geophysical surveys, and this system simply can't work properly as described.
As the wikipedia article on RTK points out, you can't guarantee phase alignment with a single receiver. Anybody actually using RTK in the field has a base station (a tripod-mounted GPS receiver and a radio transmitter) set up on a known (to great precision) point.
In order to position our base station in the first place, we either need to locate it in RTK mode, and thus position ourselves based on a pre-existing known point, or we set up our base station and let it find its position over the course of hours.
If it were simple enough to have a single device that could determine GPS positioning with this much accuracy, there would be all sorts of actual products utilizing it. Unfortunately, it's simply impossible to use RTK in a situation like a car or plane where you're moving over large distances, or in any situation where you don't have time to set up a transmitter on a known point.
It's a neat little project, sure, but to actually use it you're going to need a lot more hardware and setup time.
Reply to this comment
This isn't a single-receiver solution - the paper makes it clear that there is a fixed second station on the other side of the HDSPA network link.
Reply to this comment
As the web page makes clear, two receivers are used. The paper linked at the top of the page shows the second receiver as a geodetic-quality (presumably dual-frequency) receiver, but another copy of the low-cost rover would be fine for differential measurements not requiring a link to the global WGS/ITRF coordinate system.
It's worth mentioning that many GPS reference stations publish their data freely, so an amateur can indeed get few-centimeter accuracy to ITRF with a single low-cost L1-only receiver provided a reference station is available within 10 km or so. A good fraction of the continental USA meets this requirement. Once the L2C and L5 signals arrive, this distance will be vastly increased.
It's great to see projects that make carrier-phase GPS more widely available.
Reply to this comment
how well/badly would it perform in urban tunnels. hopefully someone will bring a user-perspective test soon. joe
Reply to this comment
As GPS satellites work in the negative SNR domain, phase tracking the carrier is tricky at best. You tell me how wll your PLL works in this domain ;-)
But even if you could - you might achieve greater precision in your output, but not accuracy - which is limited by the Satellite Ephemeris data accuracy (Satellite position), ionospheric and tropospheric signal delays, GPS receiver antenna phase delays, etc.
Now... when you add a ground based signal - the game completely changes. You CAN track this signal's phase very accurately, and then apply corrective biases. I am still not sure how this is then different than Differential GPS?
Reply to this comment
While SNR is below 0 dB over the full signal bandwidth, this is not what a PLL sees---it sees the signal after the spread-spectrum code has been wiped off, and with a bandwidth of perhaps 10 Hz rather than several MHz. The SNR in this 10 Hz bandwidth is at least +30 dB for unobstructed satellites (at L1 with no squaring loss). PLLs can track this easily with typical RMS errors of a few degrees or less. Any of the textbooks (e.g. Kaplan and Hegarty or Parkinson and Spilker) has pretty graphs of all the tradeoffs.
Satellite ephemeris and clock-bias data can be downloaded freely from the IGS, and they are much better than the broadcast values from the GPS satellites themselves. The other errors you cite are still operative, though, and limit system accuracy to a few centimeters globally and a few millimeters locally. (Also multipath is a limiting factor.)
Reply to this comment