ION GNSS 2012
Session C4: GNSS Ground Based Augmentation Systems (GBAS)

Title: Analysis of GPS Pseudorange Natural Biases using a Software Receiver
Author(s): S. Gunawardena and F. van Graas, Ohio University
Date/Time: Thursday, September 20, 2012, 2:35 p.m.
Room: 206 (NCC)

High integrity safety-of-life GNSS applications such as civil aviation use a differential architecture to remove common mode GPS system errors such as orbit errors, clock errors, and errors due to the ionosphere to yield centimeter-level relative positioning accuracy. For short baseline ground-based augmentation systems (GBAS) and space-based systems (SBAS) differential corrections to each satellite in view are computed using sets of pseudorange measurements from one or more reference receivers. These corrections are then broadcast to airborne users via a wireless datalink. Airborne systems remove the common mode errors by effectively applying these corrections to the pseudoranges measured by their GPS sensor(s). Errors not removed by this technique include multipath, errors due to the troposphere, and a type of error known as pseudorange natural biases - the subject of this paper. Extensive statistical analysis and modeling is typically used to assign bounds for each of these non-common-mode errors according to system integrity requirements. For example, a typical error bound for pseudorange natural biases for GBAS is on the order of 0.1 m. Practical validation of this error bound is important for both current and future SBAS and GBAS.

Conceptually, when two dissimilar GNSS receivers in a zero baseline configuration produce pseudorange measurements for the same satellite in view, and these two measurements are corrected for receiver clock biases and other deterministic effects (such as relative front-end delays), any non-zero-mean residual is the pseudorange natural bias for a given GNSS signal between these two particular receivers. Natural biases can occur due to distortions in the signal broadcast by the satellite and the deformation of each receiver´s correlation function as a result. For example, excessive ringing when the BPSK signal transitions from a -1 chip to a +1 chip and vice-versa can cause oscillations to be superimposed on the otherwise-triangular correlation function. Depending on a particular receiver´s front-end bandwidth and correlation point placement, these deformations can give rise to natural bias errors between different types of receivers. Noteworthy is the fact that a receiver´s natural bias error-producing mechanism is similar for multipath errors. Hence, any experiment designed to measure natural biases must take care to ensure that observables are not corrupted by multipath.

Direct measurement of pseudorange natural biases as a function of receiver parameters is complicated by a variety of factors: 1) the variation of each receiver´s group delay as a function of time, 2) inability to change a receiver´s bandwidth and limited ability to change its correlator spacing, and 3) modern receivers´ use of sophisticated multipath mitigation techniques which attempt to ´correct for signal deformations´ thus clouding the desired observables.

A wideband software GNSS receiver is a near-ideal tool for characterizing natural biases in a true ´apples-to-apples´ sense since pre and post-correlation digital signal processing steps (that emulate various receiver configurations and are completely characterizable) can be applied to the same set of wideband IF data samples. For example, the effect of correlator spacing can be studied in isolation while holding all other parameters constant (such as bandwidth, type of discriminator, tracking loop bandwidths, etc.).

This paper summarizes the results of an extensive analysis of pseudorange natural biases that was performed using Ohio University´s Transform-Domain Instrumentation GNSS Receiver (TRIGR) platform. The paper reports C/A code natural biases for the current GPS constellation as a function of satellite elevation angle and other significant receiver parameters. Also included is the ~18 ns-resolution correlation function output of the receiver and its correspondence to the observed natural biases.

Previous Abstract Return to Session C4 Next Abstract