time stability of the k-band catalog sources · allan variance to determine noise floor (1/3)...

Time stability of the K-band catalog sources

K. Le Bail (1), A. de Witt (2), C. S. Jacobs (3), D. Gordon (1)(1) NVI, Inc., Greenbelt, MD, U.-S.

(2) SARAO/HartRAO, Krugersdorp, South Africa(3) Jet Propulsion Laboratory, California Inst. Technology/NASA, Pasadena, CA, U.-S.

24th Meeting of the European VLBI Group for Geodesy and Astrometry (EVGA)March 17-19 2019, Las Palmas, Gran Canaria, Spain

Acknowledgements:Copyright ©2019. All Rights Reserved.U.S. Government sponsorship acknowledged for work done at JPL-Caltech under a contract with NASA.The VLBA is managed by NRAO, funded by the National Science Foundation, and operated under cooperative agreement by Associated Universities.The authors gratefully acknowledge use of the VLBA under the USNO’s time allocation.This work supports USNO’s ongoing research into the celestial reference frame and geodesy.HartRAO is a facility of the National Research Foundation (NRF) of South Africa.The Hobart telescope is operated by the University of Tasmania and this research has been supported by AuScope Ltd., funded under the National Collaborative ResearchInfrastructure Strategy (NCRIS).

Outline

• K-band in ICRF3• Brief review on how to use the Allan variance to

determine noise floor of source time series• Noise floor of the latest GSFC K-band time series

solution and comparison with the latest GSFC S/X time series solution

2/17

K-band

• K-band in ICRF3See poster P306: A. de Witt et al., “The K-band (24 GHz) Celestial Reference Frame”.

3/17

K solutionJan 2019

S/X solutionJan 2019

Source # 906 4775

Session # 65 6271

Observation period

05/2002 –11/2018

08/1979 –01/2019

Source #(>10 sessions)

354 788

• This paper: Study of the latest GSFC time series solutions for K and S/X.

−80 −60 −40 −20 0 20 40 60 800

20

40

60

80

100

Declination (deg)

Sour

ce n

oise

(µas

)

ICRF3 decimation test − Right Ascension

S/X noise floorK noise floor

−80 −60 −40 −20 0 20 40 60 800

20

40

60

80

100

Declination (deg)

Sour

ce n

oise

(µas

)

ICRF3 decimation test − Declination


310 common sources

K-band and S/X-bandPosition time series

• Different ways to study these time series:• Standard deviation => static quantity;• Noise floor using the Allan variance => takes into account the time variable.

4/17

1980 1985 1990 1995 2000 2005 2010 2015 2020−1

−0.5

0

0.5

1

RAcosDEC (mas) 0552+398

S/X Standard dev: 1.512 K Standard dev: 0.139

S/XK

1980 1985 1990 1995 2000 2005 2010 2015 2020−1

−0.5

0

0.5

1

DEC (mas) 0552+398


S/XK

1980 1985 1990 1995 2000 2005 2010 2015 2020−1

−0.5

0

0.5

1

RAcosDEC (mas) 3C120


S/XK

1980 1985 1990 1995 2000 2005 2010 2015 2020−1

−0.5

0

0.5

1

DEC (mas) 3C120


S/XK

Allan variance to determine noise floor (1/3)

• The Allan variance is a statistical tool that gives level and type of noise of time series.

• If !" "#$,& are the measurements and ' the sampling time, the Allan variance is:

() ' = $) !"+$ − -!" )

• The type of noise is determined by the slope of the curve

log10(Allan variance)=f(log10(sampling time))

• Cons: it has to be applied to regularly spaced time series.

-1 White noise

0 Flicker noise

+1 Random walk

5/17


6/17


Real data: sources not observed regularly => difficulties in statistical determination due to gaps in between observations, number of observations,… We need to prepare the data, to preprocess the time series.• Step 1: Keep sources with 10 or more observations.• Step 2: Averaging (yearly, monthly and weekly) and interpolation.• Step 3: Allan variance processing for each source, each coordinate,

each averaged time series.• Step 4: Noise floor determination for each source and each

coordinate. We look at the noise type determined by the slope of the Allan variance curve:• If white noise or flicker noise, the noise floor is the lowest Allan variance.• If random walk, the noise floor is undefined.

7/17

Comparison of global solutions310 common sources

The K noise floors tend to be smaller than the S/X noise floors.

8/17

Standard deviationNoise floor determined by the Allan variance

500 1000 1500 20000

20

40

60

80

100

120

140

Source standard deviation (µas)

Sour

ce n

umbe

r

RAcos(DEC) − K−band

500 1000 1500 20000

20

40

60

80

100

120

140DEC − K−band


Sour

ce n

umbe

r

500 1000 1500 20000

20

40

60

80

100

120

140

Source standard deviation (µas)So

urce

num

ber

RAcos(DEC) − S/X−band

500 1000 1500 20000

20

40

60

80

100

120

140DEC − S/X−band


Sour

ce n

umbe

r

100 200 300 400 5000

10

20

30

40

50

60RAcos(DEC) − K−band

Source noise floor (µas)

Sour

ce n

umbe

r

100 200 300 400 5000

10

20

30

40

50

60DEC − K−band


Sour

ce n

umbe

r

100 200 300 400 5000

10

20

30

40

50

60RAcos(DEC) − S/X−band


Sour

ce n

umbe

r

100 200 300 400 5000

10

20

30

40

50



Sour

ce n

umbe

r

Comparison of global solutions310 common sources

Noise floor comparison in 10o declination bands

9/17

−80 −60 −40 −20 0 20 40 60 800

100

200

300

400

500

Declination (deg)

Sour

ce n

oise

floo

r (µa

s) RAcos(DEC)


−80 −60 −40 −20 0 20 40 60 800

100

200

300

400

500

Declination (deg)

Sour

ce n

oise

floo

r (µa

s) DEC


Comparison of partial solutions87 common sources from Nov 2016 to Nov 2018

• Focus on same period of observation: November 2016 to November 2018.

10/17

1980 1985 1990 1995 2000 2005 2010 2015

S/X sessionsK sessions

K solution Jan 2019 S/X solution Jan 2019 Common sources

Source # 901 4728 877Session # 44 391Source # (>10 sessions) 254 431 87

Comparison of partial solutions87 common sources from Nov 2016 to Nov 2018

11/17

2016.5 2017 2017.5 2018 2018.5 2019 2019.5−1

−0.5

0

0.5

1



S/XK

2016.5 2017 2017.5 2018 2018.5 2019 2019.5−1

−0.5

0

0.5

1

DEC (mas) 0552+398


S/XK

2016.5 2017 2017.5 2018 2018.5 2019 2019.5−1

−0.5

0

0.5

1



S/XK

2016.5 2017 2017.5 2018 2018.5 2019 2019.5−1

−0.5

0

0.5

1

DEC (mas) 3C120


S/XK

Comparison of partial solutions87 common sources on period Nov 2016 – Nov 2018

The S/X noise floors tend to be smaller than the K noise floors.

12/17


500 1000 1500 20000

5

10

15

20

25

30

35

40


Sour

ce n

umbe

r

RAcos(DEC) − K−band

500 1000 1500 20000

5

10

15

20

25

30

35

40DEC − K−band


Sour

ce n

umbe

r

500 1000 1500 20000

5

10

15

20

25

30

35

40


Sour

ce n

umbe

r


500 1000 1500 20000

5

10

15

20

25

30

35



Sour

ce n

umbe

r

100 200 300 400 5000

5

10

15



Sour

ce n

umbe

r

100 200 300 400 5000

5

10

15

20DEC − K−band


Sour

ce n

umbe

r

100 200 300 400 5000

5

10

15



Sour

ce n

umbe

r

100 200 300 400 5000

5

10

15



Sour

ce n

umbe

r

Comparison of partial solutions87 common sources on period Nov 2016 – Nov 2018


13/17

−80 −60 −40 −20 0 20 40 60 800

100

200

300

400

500

Declination (deg)

Sour

ce n

oise

floo

r (µa

s) RAcos(DEC)


−80 −60 −40 −20 0 20 40 60 800

100

200

300

400

500

Declination (deg)

Sour

ce n

oise

floo

r (µa

s) DEC


Comparison of partial solutions31 common sources in VLBA sessions UD sessions for K-band, UF-UG sessions for S/X

• UD sessions: 24-hour VLBA sessions at K-band.

• UF001 and UG002 sessions: 24-hour VLBA sessions at S/X band. Goals: improving the precision of ICRF3, ICRF3 maintenance, and future updates of the ICRF at radio frequencies. Approximately 3300 of the weakest ICRF3 sources will be re-observed during these sessions.

14/17

K solution Jan 2019

S/X solution Jan 2019

Source # 819 4145

Session # 27 40

Source #(>10 sessions)

149 132

2016.5 2017 2017.5 2018 2018.5 2019 2019.5−1

−0.5

0

0.5

1



S/XK

2016.5 2017 2017.5 2018 2018.5 2019 2019.5−1

−0.5

0

0.5

1

DEC (mas) 0552+398


S/XK

31 common sources

Comparison of partial solutionsUD vs. UF-UG sessions (31 common sources)

The K noise floors tend to be smaller than the S/X noise floors.

15/17


500 1000 1500 20000

5

10

15


Sour

ce n

umbe

r

RAcos(DEC) − K

500 1000 1500 20000

5

10

15DEC − K−band


Sour

ce n

umbe

r

500 1000 1500 20000

5

10

15


Sour

ce n

umbe

r


500 1000 1500 20000

5

10



Sour

ce n

umbe

r

100 200 300 400 5000

5

10



Sour

ce n

umbe

r

100 200 300 400 5000

5

10

15DEC − K−band


Sour

ce n

umbe

r

100 200 300 400 5000

5

10



Sour

ce n

umbe

r

100 200 300 400 5000

5

10



Sour

ce n

umbe

r

Comparison of partial solutionsUD vs. UF-UG sessions (31 common sources)


16/17

−80 −60 −40 −20 0 20 40 60 800

100

200

300

400

500

Declination (deg)

Sour

ce n

oise

floo

r (µa

s) RAcos(DEC)


−80 −60 −40 −20 0 20 40 60 800

100

200

300

400

500

Declination (deg)

Sour

ce n

oise

floo

r (µa

s) DEC


Conclusion

• K-band time stability• The K-band observations have reached a level of stability equivalent to the

level of current S/X observations. They benefited from the history of the S/X observations.

• The strength of the S/X data set is in its broad diversity of baselines and sessions.

• Comparing the VLBA sessions UD (K-band) and UF-UG (S/X-band), it seems to show the K-band statistic stability is better than the S/X-band statistic stability.

• We need more K-band observations to continue monitoring and comparing the stability of the frame realized by the K-band observations.

17/17

• Thanks to the VLBA, K-band observations have increased greatly in the past two years, prompting many studies. At the EVGA2019:• Benedikt Soja: Ionospheric calibration for K-band celestial reference frames.• Hana Krásná: Earth orientation parameters estimated from K-band VLBA

measurements.• Aletha de Witt: The K-band (24 GHz) Celestial Reference Frame.

And more…

• Structure?

18/17

1980 1985 1990 1995 2000 2005 2010 2015 2020−1

−0.5

0

0.5

1



S/XK

1980 1985 1990 1995 2000 2005 2010 2015 2020−1

−0.5

0

0.5

1

DEC (mas) 3C418


S/XK

2016.5 2017 2017.5 2018 2018.5 2019 2019.5−1

−0.5

0

0.5

1



S/XK

2016.5 2017 2017.5 2018 2018.5 2019 2019.5−1

−0.5

0

0.5

1

DEC (mas) 3C418


S/XK

Noise floor comparison on individual sources

19/17

Standard deviation

0 500 1000 1500 20000

500

1000

1500

2000

RAcos(DEC)

S/X source standard deviation (µas)

K so

urce

sta

ndar

d de

viat

ion

(µas

)

0 500 1000 1500 20000

500

1000

1500

2000

DEC


K so

urce

sta

ndar

d de

viat

ion

(µas

)

0 100 200 300 400 5000

100

200

300

400

500

RAcos(DEC)

S/X source noise floor (µas)

K so

urce

noi

se fl

oor (µa

s)

0 100 200 300 400 5000

100

200

300

400

500

DEC


K so

urce

noi

se fl

oor (µa

s)

0 100 200 300 400 5000

100

200

300

400

500

RAcos(DEC)


K so

urce

noi

se fl

oor (µa

s)

0 100 200 300 400 5000

100

200

300

400

500

DEC


K so

urce

noi

se fl

oor (µa

s)

0 500 1000 1500 20000

500

1000

1500

2000

RAcos(DEC)


K so

urce

sta

ndar

d de

viat

ion

(µas

)

0 500 1000 1500 20000

500

1000

1500

2000

DEC


K so

urce

sta

ndar

d de

viat

ion

(µas

)

0 500 1000 1500 20000

500

1000

1500

2000

RAcos(DEC)


K so

urce

sta

ndar

d de

viat

ion

(µas

)

0 500 1000 1500 20000

500

1000

1500

2000

DEC


K so

urce

sta

ndar

d de

viat

ion

(µas

)

Noise floor determined by the Allan variance

0 100 200 300 400 5000

100

200

300

400

500

RAcos(DEC)


K so

urce

noi

se fl

oor (µa

s)

0 100 200 300 400 5000

100

200

300

400

500

DEC


K so

urce

noi

se fl

oor (µa

s)

(157

/142

sour

ces)

Entir

e pe

riod

(310

sour

ces)

(74

sour

ces)

2-yr

per

iod

(87

sour

ces)

(24/

28 so

urce

s)VL

BA se

ssio

ns(3

1 so

urce

s)

time stability of the k-band catalog sources · allan variance to determine noise floor (1/3)...

Documents