I am looking for a positioning accuracy that is higher than normal consumer-grade GNSS receiver (mostly GPS/GLONASS hybrids with +- 5m precision). I came across RTK with dual receiver setup (base station + rover) which is supposed to do cm-level precision although it has the "disadvantage" while it needs a base station or another source of correction data that one may not be able to supply.
I am interested in getting the most of a single receiver setup. I read Precise Point Positioning (PPP) is able to achieve sub-meter (or maybe even 20cm-level) accuracy. I think PPP needs raw data from sensor which I can do. I will do readings from a moving vehicle in my application so I can not go with stationary "averaging" over time for getting the accuracy. However, I do not need a real-time solution as I will only collect mobile data and evaluate it later so I can do positioning in post-processing.
U-blox also has a page on PPP where I get the impression their solution is just an ordinary raw data chip with a "better firmware" that can do some additional calculations (quote: "... by stabilizing measurements of the distance between GNSS satellites and the receiver (pseudo-ranges) using carrier phase tracking") so it has better precision than ordinary chip.
I read the Wikipedia article but I do not really understand how PPP works. So how exactly is this PPP correction technically done? Where do the corrections come from?
I read RTKLIB can do PPP too. My location is Czech Republic.