Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://www.atnf.csiro.au/iau-comm31/pdf/2009_IAUGA_JD6/JD6_Petit.pdf Äàòà èçìåíåíèÿ: Fri Aug 6 03:07:24 2010 Äàòà èíäåêñèðîâàíèÿ: Tue Oct 2 18:05:40 2012 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: http www.badastronomy.com bad tv foxapollo.html

Atomic time scales TAI and TT(BIPM): present performances and prospects

GÈrard Petit Bureau International des Poids et Mesures 92312 SÕvres Cedex, France gpetit@bipm.org

IAU'2009 JD6


RÈsumÈ
· · · · · Time scales: EAL-TAI-TT(BIPM) Already achieving sub-10-15 TT(BIPM): accuracy, consistency of primary frequency standards Towards 10-16 and below? Contributions of frequency standards to TAI

IAU'2009 JD6


EAL and TAI
· TAI calculation is done each month (i.e. in "real time") · The BIPM computes a free atomic scale, EAL, from some 350 atomic clocks worldwide, aiming at optimal 1-month stability.
­ An average of N identical clocks should be N more stable than each one. ­ Clocks in different laboratories have to be compared by time transfer techniques: presently GPS and Two-way; the goal is that time transfer noise is negligible. ­ f(EAL) is stable but it can have any value (with respect to the SI second).

· Each month, primary frequency standards (PFS) are used to estimate f(EAL).
­ PFS are expected to represent the SI second.

· The frequency of TAI is then steered so that f(TAI) is close to the SI second.
f(EAL) - f(TAI)
71.5 71 70.5
14

70 69.5 69 68.5 68 67.5 1999 2001 2003 Ye a r 2005 2007 2009

10-

IAU'2009 JD6


TT(BIPMxx)
· As TAI is computed in real time and is not updated if an error is discovered, it is not optimal. · Therefore the BIPM computes a post-processed time scale TT(BIPM) · Each new version TT(BIPMxx) updates and replaces the previous one. · TT(BIPMxx) calculation
­ Post-processed using all available PFS data, as of year 20xx. ­ Complete re-processing starting 1993 (possibly with change of algorithm). ­ f(EAL) is estimated each month using available PFS. Monthly estimates are smoothed and integrated to obtain TT(BIPMxx). ­ Last realization: TT(BIPM08), released in January 2009.

· Significant and time-varying frequency difference between TAI and TT(BIPM) integrates to (up to) 100 ns/yr, and even more before 2003: TAI should NOT been used as a long-term reference e.g. for pulsar analysis.

IAU'2009 JD6


Achieving sub-10-15 stability/accuracy
· Time scale
­ Stability of ensemble time scale assessed by statistical analysis ­ Accuracy depends on PFS performance

· Time transfer
­ Performance assessed by comparison of independent techniques ­ Also by the ability to compare ultra-stable clocks

· Frequency standards
­ Numerous Cs fountains claim to achieve this level ­ Other transitions also available, some have been named "secondary representations of the second"

IAU'2009 JD6


Time scales: achieving 10-15 and below
· EAL: < 4.10 @ 1 month since 2003, from the stability of participating clocks. · TT(BIPM): < 1.10-15 @ any averaging since 2003, from statistical treatment of PFS uncertainty. · TAI: In between. Close to EAL @ 1 month, < 2.10-15 @ years.
-16
Compute d 1-month insta bility of EAL 0. 0. 0. 0. 0. 0. 0. 0. 0. 1 9 8 7 6 5 4 3 2 1 0 1999

10-

15

2001

2003

Date

2005

2007

2009

Uncertainty in f(TT(BIPM08))
3 2.5 2
-15

10

1.5 1 0.5 0 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 Ye ar

IAU'2009 JD6


Time transfer: achieving 10-15 and below
· · · · · Present best: GPS CP and Two Way TW-GPS-CP for four links (Bauch et al. 2006) show both techniques cross 1.10-15 @ 1 day. TW needs 24 pts/day and same transponder to achieve this. Performance of GPS CP is about independent on the distance => PPP to be used in TAI GPS-code only, as well as TW with two transponders are slightly less stable 1.10-15 @ 2-3 day
U SNO-PTB 02/2006-05/2006
TW (X) PP P TW (Ku)

1.E-13

M o d y (T a u )

1.E-14

1.E-15

1.E-16 0 .0 1

0 .1

1
Ta u / d

10

10 0

IAU'2009 JD6


Accuracy of frequency standards: achieving 10-15 and below

Primary Standard IT-CSF1 NICT-CSF1 NIST-F1 NMIJ-F1 PTB-CS1 PTB-CS2 PTB-CSF1 SYRTE-FO1 SYRTE-FO2 SYRTE-FOM SYRTE-JPO

Type /selection Fountain Fountain Fountain Fountain Beam /Mag. Beam /Mag. Fountain Fountain Fountain Fountain Beam /Opt.

Type B std. Uncertainty (0.5 to 0.7)x10 (0.8 to 1.5)x10 0.3x10 4x10 8x10
-15 -15 -15

Operation Discontinuous Discontinuous Discontinuous Discontinuous Continuous Continuous Discontinuous

Comparison with H maser UTC(NICT) H maser H maser TAI TAI H maser H maser H maser H maser H maser

Number/typical duration of comp. 6 / 10 to 20 d 2 / 10-15 d 5 / 15 to 25 d 7 / 15 to 25 d 10 / 30 d 12 / 30 d 2 / 15 d 8 / 10 to 30 d 9 / 10 to 30 d 6 / 10 to 30 d 12 / 10 to 30 d

-15 -15 -15 -15 -15 -15 -15

12x10

0.9x10

(0.4 to 0.6)x10 (0.4 to 0.6)x10 (0.7 to 0.9)x10 6x10
-15

Discontinuous Discontinuous Discontinuous Discontinuous

Primary standards reported to the BIPM in 2008 (8 fountains and 3 beams)

IAU'2009 JD6


Accuracy of frequency standards: achieving 10-15 and below
Type B standard uncertainty (x1016) Fountain Physical origin 2nd order Zeeman Blackbody Radiation Cold Collisions + cavity pulling First Doppler Synchronous phase fluctuations Microwave Leaks, spectral purity Background gas collisions Microwave recoil Ramsey & Rabi pulling Second order Doppler Red shift Total uncertainty FO1 Uncertainty 0.3 0.6 2.4 <3.2 <0.6 <0.7 <0.3 <1.4 <1 <0.1 1 4.7 0.5 <1 <1.4 <1 <0.1 1 4.8 FO2 Uncertainty 0.2 0.6 2.9 3

SYRTE FO1/FO2

NIST-F1

IAU'2009 JD6


TT(BIPM): the latest realization TT(BIPM08)
· Post-processed in January 2009 using all primary frequency standards data until December 2008. · Frequency accuracy over the period: decreases from 2.5x10-15 in 1999 to <1x10-15 since 2004, <0.5x10-15 in 2008.

Uncertainty in f(TT(BIPM08))
3 2.5 2
-15

10

1.5 1 0.5 0 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 IAY'20ar JD6 U e 09


Contributions to TT(BIPM)
· TT(BIPM) performances due to increasing number of Cs fountains · A rough statistical estimate would put TT(BIPM) accuracy in the low 10-16, but time transfer and instability of EAL limit this.
Founta in e va lua tions

Number of evaluat ions
45 40 35 30 25 20 15 10 5 0 1999 2000 2001

Unc ert aint y / 10-16

Unc ert aint y of 1-y ear average

2002

2003

2004

2005

2006

2007

2008

Y '20 IAUe ar09 JD6


Fountains vs. TT over 2004-2008
· Results globally coherent, but 2 slightly too large · Situation similar (even slightly worse) in the most recent period
f(PFS)-f(TT) (fountains)

0.8

125 fountain evaluations. 2 = 1.30 (1.19 without 3 "ou

tliers")

0.6

0.4

NIST-F1 PTBCSF1 FO2

0.2 10
-14

FOM IT-CSF1

0.0

NPLCSF1 NMIJ-F1 FO1

-0.2

NICT-CsF1

-0.4

-0.6 2004

2005

2006

2007

2008

2009

IAU'2009 JD6 Years


Estimation of systematic effects in Cs fountains
· · Because the accuracy of TAI/TT(BIPM) depends on that of the PFS, we want to check for the existence of systematic effects in Cs fountains. When one Cs fountain has enough evaluations, we compare its results with an ensemble average of the other PFS: Tp(Rest_of_the_world) ­ Computation equivalent to TT(BIPM) but using all primary standards except the one under study This estimation has been done (until end 2006) for four fountains with largest number of evaluations (13 to 19). Results show ­ Good self coherence of the four studied PFS. ­ But some systematic differences: none of the four studied PFS agree with the "rest of the world" within the uncertainties. This may be due to · Systematic effects: Estimated accuracy of PFS may be (slightly) optimistic. · Uncertainty in frequency transfer: may also be (slightly) optimistic.

· ·

IAU'2009 JD6


Aiming at 10-16 and beyond
· Ensemble time scale ­ May be limited by the clocks available · Time transfer ­ Will depend on technology developments. ­ Always improved by longer averaging · Frequency standards ­ This is already achieved both for the stability and for the capacity to evaluate systematic effects. ­ Practical application will depend on the achievable continuous operation time (i.e. possible averaging time). ­ See e.g. Proc. of 7th Symposium on frequency standards and metrology, 2008

IAU'2009 JD6


Ensemble time scale: aiming at 10-16 and beyond
· With present technologies, a 10-fold increase in clock number would be needed to reach 1.10-16 . ­ Back of the envelope calculation (for the present situation): 100-200 clocks, each with 5.10-15 stability @ 1 month provide 3-4.10-16 for the ensemble time scale. · Reaching 1.10-16 and beyond on the long term (1 month and above) will depend on the availability of new clock technologies.

IAU'2009 JD6


Limits to long-term stability of EAL
Compute d 1-month insta bi li ty of EAL

· Has decreased from about 69.10-16 in 1999-2000 to about 4.10-16 in 2003. · But more or less constant since 2003. Total number of clocks still increasing, but total number of good continuous clocks only slightly increasing. · Some marginal improvements still possible. · But gaining e.g. one order of magnitude does not seem possible without new clocks.

0. 0. 0. 0. 0. 0. 0. 0. 0.

1 9 8 7 6 5 4 3 2 1 0 1999

Change of weighting strategy

10-

15

2001

2003

Da te

2005

2007

2009

Num be r of cl ocks use d (good: 11-m onth cont. w good HP c loc k s good H mas ers

ave=

0.1w

max)

All good

Tot al

3 2 2 2 1 1

2 8 4 0 6 2 8 4

IAU'2009 JD6

0 0 0 0 0 0 0 0 0 1 9 9 9 20 0 0 2 00 1 2 00 2 2 0 0 3 2 0 0 4 20 0 5 2 00 6 2 00 7 2 0 0 8 2 0 0 9 Date


Limits to the long-term stability of EAL
f(EAL) - f(TT(BIPM08))

· f(EAL) is compared to TT(BIPM): Some systematic frequency trends persist for many years · Systematic drift is only partly due to H-masers. · Long-term (1 year) stability of EAL is presently limited · TAI not too affected if the number of PFS evaluations is large enough, but life would be easier without EAL drift.

72.5 71.5 70.5
-14

10

69.5 68.5 67.5 66.5 1 99 3

19 95

1 997

19 99

2 001 Y e ar

2 003

200 5

20 07

200 9

IAU'2009 JD6


New clock technology
· LITS 199Hg+ (Burt et al. 2008) : 9-month uninterrupted operation at JPL. ­ 5x10-14/1/2 (estimated from maser comparison) ­ floor of less than 2x10-16 (from comparison to Cs fountain) ­ drift < 2.7x10-17/day (from comparison to TT(BIPM07) · Rb fountain (Ekstrom et al. 2008) ­ 1.5x10-13/1/2 ­ floor at about 3x10-16 · Your favorite clock...

IAU'2009 JD6


Frequency transfer: limits to the current techniques
·The current techniques may still be (slightly) improved ·GPS/GNSS: ­More systems, more satellites, more frequencies, different codes ­Best performance presently is with GPS dual frequency phase and code (such as PPP) ·TW: ­More bandwidth, more time (both very expensive), use phase ·Comparisons of techniques in ftp://tai.bipm.org/TimeLink/LkC/yymm ­Typical values of RMS of (TW-PPP) over one month: 0.2 to 0.6 ns ­All techniques stable to 100-300 ps up to 10-day averaging ·TW and PPP can provide 10-16 frequency accuracy, but only averaging > 10-15-20 days. ·But TW and PPP cannot presently reach that level at 1-day averaging: New techniques needed

IAU'2009 JD6


ftp://tai.bipm.org/TimeLink/LkC/0906
USNO-PTB: TW(X) - PPP

IAU'2009 JD6


Frequency transfer: aiming at 10-16 and beyond · ACES Microwave link (from Salomon et al 2007) · T2L2 (from Guillemot et al. 2006) · Optical fiber (from Lopez et al. 2007)

IAU'2009 JD6


Contributions of frequency standards to TAI
· CCTF 3 (2004) recommends that TAI scale unit be conform to its definition to within 3 .

· This is generally not achieved except end 2006-early 2007

f(TAI)-f(TT(BIPM))
0.2

0

-0.2

-0.4
-14

10

-0.6

-0.8

-1

-1.2

-1.4 19 99

00 IAU'220009 J2006 D1

20 02

2 003

2 00 4
Year

20 05

2 00 6

20 07

20 08

2 00 9


Contributions of frequency standards to TAI
· Evaluations of PFS always needed. · Presently (2009), nearly 4 fountain evaluations per month. Quite good in regard to the number of available fountains. · New FS encouraged ­ PFS ­ "Secondary representations of the second" are also expected to provide evaluations, in order for us to get experience with their use.

IAU'2009 JD6


Conclusions
· Sub-10
­ ­ ­ ­
-15

level is proven for all components of time scale formation:

Ensemble time scale stability Time transfer Primary frequency standards More studies to check for systematic effects
-16

· How to reach 1x10

(and beyond)?

­ Very stable clocks already exist. Better reliability and wider availability are needed for time scale formation. ­ Present time transfer techniques may reach this level for long averaging time (10 days). New techniques needed to reach it at short averaging time, and go beyond.

· In the mean time, more (P)FS data needed (more regularly) · Start to study alternative algorithms for
­ minimizing long term instability, drift, in EAL ­ TAI steering ­ TT(BIPM) computation.

IAU'2009 JD6


IAU'2009 JD6


Stability of EAL and TAI vs. TT(BIPM)
· TT(BIPM) is correlated to EAL at 1-few month averaging. The 1-month stability of EAL is estimated at 3-4x10-16 over the past years. · EAL: the long-term behavior indicates that a drift is present. · TAI: the long term stability has been much improved due to better steering (because better PFS are available).

IAU'2009 JD6