time transfer – algorithms for the galileo programme · time transfer – algorithms for the...
Post on 03-May-2018
225 Views
Preview:
TRANSCRIPT
Time Transfer – Algorithms for the Galileo Programme
John Davis 3rd June 2009
Monday, 22 March 2010
2
WHY DO WE NEED A STABLE REFERENCE TIMESCALE FOR GALILEO?
• Galileo will be used for time and frequency dissemination, requires traceability to UTC
• Stable timescale required for synchronising clocks on satellites, and obtaining optimal performance from Galileo’s navigation services.
Monday, 22 March 2010
3
REQUIREMENTS OF GALILEO REFERENCE TIMESCALE, GST(MC)
• Physical realisation of a timescale plus back- up GST(MC).
• TAI – GST(MC) < 50 ns (2σ)• TAI – GST(MC) uncertainty < 26 ns (2σ)• TAI – GST(MC) Normalised frequency
uncertainty < 5.4E-14 (2σ) (τ
= 1day) • GST(MC) normalised frequency stability <
4.5E-15 (τ
= 1day)
Monday, 22 March 2010
4
COMPUTATION AND AVAILABILITY OF UTC
• Coordinated Universal Time computed once per month by BIPM
• Based on an ensemble of mainly active hydrogen masers and high performance caesium clocks
• National standards laboratories maintain UTC(k) physical realisations of UTC,
• UTC – UTC(k) offsets computed up to 50 days in arrears
Monday, 22 March 2010
5
WHY DO WE NEED TIMESCALE ALGORITHMS?
• GST(MC) stability and reliability requirements cannot be achieved using a single clock
• Requirement to steer GST(MC) timescale to UTC / TAI
• Need to obtain best estimate of timescale offsets e.g. TAI – GST(MC) using satellite time transfer methods
Monday, 22 March 2010
6
TIME AND FREQUENCY RESPONSIBILITY WITHIN THE GALILEO REFERENCE
TIMESCALE
• Precise Time Facility (PTF) provides a stable physical timescale for navigation applications.
• Time Service Provider (TSP) provides the traceablilty to UTC / TAI, via several European UTC(k) laboratories.
Monday, 22 March 2010
7
CLOCK AND TIME TRANSFER NOISE MODELS
• Clock noise is modelled using a linear combination of Integrated Markov Noise Processes
• Close to FFM in the case of Active Hydrogen Masers • Close to WFM in the case of Caesium clocks
• Time transfer noise is modelled as linear combination of Markov Noise Processes
• Part way in characteristics between WPM and FPM, stationary in long term
• Similar model for TWSTFT, GPSCV and internal Galileo measurements
Monday, 22 March 2010
8
NOISE MODELS FOR CAESIUM CLOCKS AND ACTIVE HYDROGEN MASERS
Monday, 22 March 2010
9
NOISE MODELS FOR TWSTFT AND GPS CV MEASUREMENT NOISE
5 5.0005 5.001 5.0015 5.002
x 104
-1
0
1
2
3
4
5
6
7
8
9x 10
-9 Tim e Tra ns fe r Nois e S im ula tion
MJ D
Tim
e O
ffset
GP S CV TWS TFT
Monday, 22 March 2010
10
TWSTFT MODEL, ADEV, MDEV, HDEV CHARACTERISTICS
1 0 2
1 0 3
1 04
1 0 5
1 0 6
1 0 7
1 0 - 1 7
1 0 - 1 6
1 0 - 1 5
1 0 - 1 4
1 0 - 1 3
1 0 - 1 2 A D E V , H D E V M D E V T W S T F T
ta u
AD
EV /
MD
EV /
HD
EV
A D E V s im u la t io nH D E V s im u la t io nM D E V s im u la t io nA D E V e x p e c ta t io nH D E V e x p e c ta t io n M D E V e x p e c ta t io n
Monday, 22 March 2010
11
GPS CV MODEL, ADEV, MDEV, HDEV CHARACTERISTICS
1 0 2
1 03
1 04
1 0 5
1 06
1 071 0
- 1 6
1 0- 1 5
1 0- 1 4
1 0- 1 3
1 0- 1 2 A D E V , H D E V M D E V G P S C V
ta u
AD
EV /
MD
EV /
HD
EV
A D E V s im u la t io nH D E V s im u la t io nM D E V s im u la t io nA D E V e x p e c ta t io nH D E V e x p e c t a t io nM D E V e x p e c t a t io n
Monday, 22 March 2010
12
ENSEMBLE ALGORITHMS
• Aim is to produce a “composite” free running timescale from an ensemble of individual clocks
• Realised via estimates of (CClk – CI )• Early algorithms based on weighted mean
approach• Later algorithms based on use of Kalman
Filter, these are used in Galileo
Monday, 22 March 2010
13
ENSEMBLE ALGORITHM• TSP ensemble algorithm will include clocks from
several UTC(k) laboratories.
• Time Offset, Normalised Frequency Offset, Linear Frequency Drift and Integrated Markov Process state vector components for each clock
• Active hydrogen masers assumed to possess linear frequency drift, caesium clocks assumed to be free of linear frequency drift
Monday, 22 March 2010
14
ENSEMBLE ALGORITHM
• Long term, ensemble timescale will be free of linear frequency drift
• Close to optimal stability at all averaging times
• Includes linear frequency drift in clock models
• May include white measurement noise
Monday, 22 March 2010
15
COMBINING CLOCKS WITH DIFFERENT STABILITY CHARACTERISTICS
2 2 . 5 3 3 . 5 4 4 . 5 5 -1 5 . 8
-1 5 . 6
-1 5 . 4
-1 5 . 2
-1 5
-1 4 . 8
-1 4 . 6
-1 4 . 4
-1 4 . 2
L o g 1 0 ( τ )
Log 10
(σy)
C lo c k 1 (F F M ) C lo c k 2 (W F M ) C lo c k 3 (W F M ) C o m p o s ite S im p le C o m p o s ite O p t im a l C o m p o s ite
Monday, 22 March 2010
16
ENSEMBLE ALGORITHM PERFORMANCE, 2 HYDROGEN MASERS + 4 CAESIUM CLOCKS
Monday, 22 March 2010
17
WHITENESS OF RESIDUALS
5.0008 5.0008 5.0009 5.0009 5.001
x 104
-6
-4
-2
0
2
4
6 x 10
-12 Clock 2 - Clock1, Res idua ls ,
MJD
Res
idua
ls
Monday, 22 March 2010
18
REMAINING ISSUES
• Treatment of non-white measurement noise
• Difference in noise levels between and within laboratories.
• Biases in time transfer measurements
Monday, 22 March 2010
19
TIME TRANSFER COMBINING ALGORITHMS
• Use range of satellite time transfer measurements to estimate offset between two distant timescales
• UTC(j) – UTC(k) links
• Offset between two separate PTF reference timescales
• GST(MC) – GPS_Time offset
Monday, 22 March 2010
20
ALGORITHM PROPERTIES
• Kalman filter based• Multiple input data sets in single algorithm:
TWSTFT, GPSCV, internal Galileo products • Use of both TWSTFT and GPSCV calibration
within algorithm• Model calibration offset as long relaxation time
Markov Process • Timescale and time transfer models already
described
Monday, 22 March 2010
21
ALGORITHM PERFORMANCE
• Separately weight short term noise and calibration uncertainties of TWSTFT and GPSCV data
• Maintain TWSTFT calibration uncertainty on loss of TWSTFT data
• Good performance in case of missing data
Monday, 22 March 2010
22
ESTIMATION PERFORMANCE OF TIME TRANSFER COMBINING ALGORITHM
5 5 .0 0 0 5 5 .0 0 1 5 .0 0 1 5 5 .0 0 2
x 1 04
-0 .5
0
0 .5
1 .0
1 .5
2 x 1 0
-9 T im e O ffs e t E s tim a tio n E rro r
M J D
Tim
e O
ffset
T im e O ffs e t E rro r T im e O ffs e t E rro r E xp e c ta tio n
Monday, 22 March 2010
23
STEERING AND PREDICTION ALGORITHMS
• Based on time transfer combining algorithms
• Kalman filter based timescale predictor added to deal with delays in data set availability
• Steering algorithm added, to either minimise time offset or frequency offset
Monday, 22 March 2010
24
PERFORMANCE OF A STEERING ALGORITHM
Monday, 22 March 2010
25
FIRST RESULTS FROM TSP
Steered and unsteered timescales
-8.00E-08
-6.00E-08
-4.00E-08
-2.00E-08
0.00E+00
2.00E-08
54400 54500 54600 54700 54800 54900
MJD
Tim
e O
ffset
steered unsteered
Monday, 22 March 2010
26
CONCLUSIONS
• NPL has developed a range of state of the art timescale algorithms
• Currently being developed for use in Galileo reference timescale.
• First TSP results within specification achieved
top related