Technical Article • May/June 2015
Estimating the ShortTerm Stability of InOrbit GNSS ClocksFollowing Launch on GEO/GSO SatellitesAn engineering team proposes a method for determining the time of atomic clocks onboard satellites in geosynchronous and geostationary orbits when precise ephemeris data from ground monitoring stations are unavailable during initial checkout following launch.
Share via: Slashdot Technorati Twitter Facebook Global navigation satellite systems provide position, velocity, and time (PVT) solutions to users whose receivers calculate position based on oneway ranging from satellites. As is wellunderstood, a key step in the positioning process involves a determination of the difference between the time of signal transmission identified in the satellite’s broadcast navigation message and the time of its reception by user equipment. The accuracy of time of transmission depends on a satellite’s onboard clock stability, with the clock’s shortterm stability affecting — among other applications — precise point positioning. Hence, it is very important to monitor the shortterm stability of inorbit satellite clocks and find out the onboard time error compared to the navigation system time. GNSS satellites transmit navigation signals on Lband and/or Sband frequencies. Highly accurate receivers track carrier phase as well as code phase of transmitted signals and use the carrier phase observables to determine an inorbit clock’s shortterm stability. These calculations estimate deterministic errors due to geometrical range and range rate by using precise satellite ephemeris data to estimate the shortterm stability of the satellite clocks. However, the algorithms employed in these estimation techniques are limited during inorbit testing of satellite carried out after the launch when precise ephemeris data may not be available for processing. So, it is very important to characterize the onboard clock for initial operations of navigation satellites once in orbit. This article proposes a mathematical model to remove deterministic errors, without using satellite ephemeris data, to analyze the shortterm stability of GNSS inorbit clocks in the presence of adverse environmental and equipment effects. The proposed technique is useful for geostationary orbit (GEO) or geosynchronous orbit (GSO) navigation satellites where the carrier doppler rate remains constant for short periods of time due to the orbital characteristics.
Description of the Case Study & Algorithm For analytical purposes, we used typical GEO/GSO navigation satellite signalsinspace (SIS) in a case study, comparing the measured results of the inorbit satellite clock’s shortterm stability against a time reference on the ground. We begin by modeling carrier phase observables. In our analysis, we used the L5 band SIS to estimate onboard clock stability by measuring the carrier phase differences between the onboardtransmitted carrier phase and that received at the user equipment, which provides a satellitetoreceiver range measurement in terms of the number of carrier phase cycles. Onboard transmitted carrier phase is generated from highly stable atomic frequency standard, and the receiver measures carrier phase using a reference source, typically a crystal oscillator. These measurements are affected by various parameters such as integer cycles ambiguity, ephemeris errors, satellite clock bias, receiver clock bias, ionosphere effects, troposphere effects, and receiver measurement noise. The carrier phase observable model equation is Equation (1) (see inset photo, above right, for equations) where:
L = measured carrier phase at L5 band in seconds A receiver cannot measure the absolute carrier phase difference between satellite and receiver. It measures carrier phase within 0 to 360 degrees of one cycle as a first measurement and then keeps track of change in carrier phase over a period of time. Hence, it has an ambiguity of integer cycles (as a residue of range) that can be estimated based on precise ephemeris parameters and code phase measurements. Satellite clock errors are deterministic errors, which can be estimated, based on leastsquares estimation method. Receiver clock errors are also deterministic in nature. To estimate onboard clock performance, the receiver clock stability should be at least one order of magnitude better than onboard clock stability so that the effects of receiver clock stability do not introduce uncertainties into the carrier phase measurements. Ionosphere delay is inversely proportional to square of frequency and varies with local time and season. Ionosphere delay adds maximum range errors on carrier phase measurements over a day. Ionosphere delay can be removed using dual frequency measurements. Troposphere delay affects signals at up to 50 kilometers of altitude and contains delays due to wet and dry components, which can be estimated based on available statistical models or local measurements. Multipath also affects carrier phase measurements, but its contribution will be much less compared to other error sources. The carriertonoise density (C/N_{0}) of the link will add to the carrier tracking error on carrier phase measurements.
Analysis Time Period for Estimation of Constant Doppler Rate To solve this problem, we used GEO/GSO satellites’ carrier phase data for periods during which the doppler rate is constant so that higher order effects would not be present and effects due to relative motion on carrier phase measurement could be estimated using a leastsquares method. This approach is valid only for navigation satellites in GEO/GSO orbit and not applicable for satellites in middle Earth orbits (MEO). We used longerduration doubledifferenced carrier phase measurements to determine the constant doppler rate in order to estimate clock stability. This process was carried out for two successive days, and approximately 350 seconds (analysis time) of data from each day were used to estimate onboard clock shortterm stability using the SIS. Estimation of Deterministic Errors. As discussed in the section describing our modeling of carrier phase observables, deterministic errors arise from the satellite and receiver clocks as well as those related to relative motion. Our analysis used the stable ovencontrolled crystal oscillator (OCXO) from a phase noise measuring instrument as reference source for the payload test receiver (PTR). The OCXO’s shortterm stability is one order of magnitude better than the onboard clock in the observed navigation satellite. So, the effect due to receiver clock stability is nullified in the carrier phase measurements, which were used to estimate and remove the combined deterministic effects of onboard clock errors, receiver clock errors and errors related to relative motion before estimation of clock stability. The combined deterministic error (C_{e}) model is given in equation (2): C_{e} = Δt_{r}  Δt_{s} + d_{0}t + d_{1} t^{2} (2) where, Δt_{s} and Δt_{r} are the satellite clock bias and receiver clock bias, respectively. d_{0} (sec/sec) and d_{1} (sec/sec^{2}) are normalized doppler and doppler rate, respectively. Frequency Stability. The Allan deviation is the most important time domain measure of frequency stability. Similar to the standard deviation, it is a measure of the fractional frequency error and has the advantage of converging for most types of random clock noise. In this analysis, we used an overlapped Allan deviation mathematical tool to estimate frequency stability. The result is usually expressed as the square root of the Allan variance, using two samples of fractional frequency errors to estimate the stability of frequency. The overlapped Allan variance in terms of phase data is given as follows: Equation (3) where, x_{i} is the i_{th} sample of the N phase data spaced by the measurement averaging time τ = mτ_{0} (seconds). m is an averaging factor, and τ_{0} is the basic measurement interval. The relationship between fractional frequency values and phase data is y_{i} = (x_{i}_{+1}–x_{i})/2. Because the Allan variance is twosample variance, any bias and drift components in the phase data will cancel out and thus not affect frequency stability estimation. However, the drift rate on phase data will affect the frequency stability estimation, as discussed in the articles by D. W. Allan and M. Y. Shin et alia listed in the Additional Resources section near the end of this article. Hence, before estimation of the frequency stability of an SIS signal, all possible deterministic errors must be removed from the carrier phase observables.
Test Setup A choke ring antenna was used to receive the L5 band signals. Benefit of using choke ring antenna is stable phase center and higher multipath rejection capabilities. Thus effects due to receiver measurement error, receiver clock stability and multipath are negligible on carrier phase measurements. Figure 1 shows the test setup for estimating the clock stability.
Results As mentioned earlier, our analysis used at least 350 seconds of data for the estimation of the shortterm stability of atomic clock, which will average out random errors. Hence, any change in ionosphere delay or troposphere delay up to 15 millimeters per second will not affect the estimation of shortterm stability of the inorbit clock. Integer cycles ambiguity is also a constant term, which will not affect estimation of clock stability. As discussed earlier, any multipath error and receiver measurement error are also ignored. Now, only deterministic errors due to onboard atomic clock, receiver clock, doppler, and doppler rate need to be estimated and removed from carrier phase observables as per the model proposed earlier. In analyzing the results, navigation data is not used to estimate the shortterm stability of clock. L5 band carrier phase data of a typical GEO/GSO navigation satellite has been collected through the PTR on day of year (DOY) 36 and 37 of 2014, which are shown in Figure 2 and Figure 3, respectively. The carrier phase was measured at a onesecond rate. Double differencing of carrier phase data, which results in nothing but the drift rate, is also shown in Figures 4 and 5. Our analysis selected data from the interval 56005950 seconds from the DOY36 data set and the interval 4050 4400 seconds from DOY37 data set for estimation of clock stability, as no higher order errors are present in carrier phase observables during those intervals and the doppler rate is also constant. Figure 6 and Figure 7 shows the fractional frequency error after removing deterministic errors from carrier phase observables as per equation (2) using leastsquares estimation method. Figure 8 shows the estimated stability of the onboard clock using oneway carrier phase measurement compared with performance of the time reference on the ground.
Conclusions
Acknowledgments
Additional Resources ManufacturersThe phase noise measuring tool was the PN9100A Frequency Synthesizer Module from Aeroflex Inc. (now a Cobham plc company), Plainview, New York USA. The MultiFrequency Choke Ring MC antenna was from Septentrio Satellite Navigation, Belgium. The payload test receiver was a demonstration unit of the NavXNTR Navigation Test Receiver from IFEN GmbH, Poing, Germany.Copyright © 2017 Gibbons Media & Research LLC, all rights reserved. 
