Precise T & F Intercomparison via VLF Phase Measurements

8/6/2019 Precise T & F Intercomparison via VLF Phase Measurements

1/12

PRECISE T & F INTERCCMPARISON VIA VLF PHASEMEASUREMENTSA . S e n Gupta , G.K.Goel and B . S .M a th u rTime and F r e q u e nc y SectionNational Physical Laboratory, Hillside Road,N e w Delhi-110012, India.

ABSTRACT

The Indian subcontinent does not fall In thegroundwave range of any LORAN-C transmission.As such, at present the only alternative tech-nique in this region for convenient and routineT & F intercomparison Is via VLF phase measure-ments. At NPL, New Delhi, continuous phaserecording of the 15 kHz transmissions fromGBR (UK) s being made. In addition the pub-lished midday phase data of GBR from severallaboratories-NPL (UK),RGO(UK), PTB (FRG) andUSNO (USA) - are being received regularly. Inthe present paper we discuss T & F intercom-parisbns between the local time scale, UTC(India), at NPL and those at the above mentionedlaboratories, using the VLF phase data. A majorfactor which limits the accuracy of long termcomparison is the seasonal variation in the VLFpropagation delay over long paths. It Is shownthat by taking into account the seasonal propa-gation delay variations in a semiempirical waythe accuracy of T & F comparisons can beconsiderably improved. In fact over a one yearperiod accuracy of few parts in 10^ in frequencyand 1-2 p-sec in time have been obtained.The relative frequency offset difference betweenUTC (India) and UTC (PTB)evaluated in the presentwork as (7.0 0. x10-13 agrees very well withthat obtained via the satellite experimentdes-cribed in a companion paper.

INTRODUCTIONThe National Physical Laboratory (NPL), New Delhi, Indiahas the statutory obligation of maintaining the India

851


2/12

Standards of time and Frequency.- This is being achievedat present with the help of- two commercial cesium clocks:NPL-1 (Oscilloquartz Model 3200) and NPL-2(Hewlett PackardModel 5061A with option 004). A time scale, UTC (India),is being maintained using the NPL-1. One of major problemsbeing faced by us at NPL is that of a regular time andfrequency transfer link between UTC (India) with thoseof the other international timekeeping laboratories. TheIndian subcontinent does not fall in the groundwave rangeof any LORAN-C transmitter thus prohibiting the use ofthis technique.The first serious attempt in having an accu rate link wasmade in May-June, 1979, when a two way satellite timetransfer experiment was performed between NPL and TB,West Germany (Mathur et al, 1980 and references there n).With the help of this experiment UTC (India) was synchroni-sed with UTC' (PTB). From July, 1979 onwards we arecontinuously tracking the phase of 15 kHz transmissionsfrom GBR. (UK). The choice of this station was dictated bythe fact that; (a) the received signal strength at ourlocation is very good so that a precis e phase tracking ispossible and (b) there are several timekeeping laboratorieswhich track this station and publis h their data. We areregularly receiving the GBR phase data from PTB (FRG),USNO (USA), NPL (UK)and RGO (UK). Utilizing the VLF datait has been possible to establish a fairly precise linkbetween UTC (India) and UTC of the abovementioned labora-tories as will be described in the subsequent sections,A confirmation of the accura cy of the VLF links was madepossible with a portable clock trip from USNO in September,1930. This is described briefly in a separate subsection.THE TECHNIQUE .The VLF time transfer technique has been described byBecker et al (1969). this basically consists of recordingthe time difference between the 1 pps of the local timescale and some specified phase (generally the po si ti vezero crossing) of the received VLF signal appearing Justsubsequently. This measurement is referred to as the'phase time 1, or simply 'phase' of the received signalrelative to the local time scale. The r ecor din g is done'at the local noon over the path mid poin t when VLF propa-gatio n conditions are most stable. Subtracting the dailyvalues of phase m easurements recorded at two laboratories

852


3/12

gives the difference between their time scales. .Thisdifference is ambiguous to an additive constant whichinvolves the difference between the propagation delaysover the two paths. While this ambiguity is immaterialfor frequency comparison between the two time scales it hasto be eliminated if a time comparison is desired.Eliminationof the additive constant can be achieved by an initialcalibration with 'the help of either a portable clock or twow ay satellite experiment,A major factor which degrades precision and accuracy of theVLF time transfer is the non constancy of the propagationdelay. There is a day to day jitter in the propagationdelay because of the normal variability of the ionosphericD-region. This is m o r e pronounced dur in g winter thanduring s u m m e r , as will be shown inthe next section (seealso Belrose,1958). The jitter evidently reduces theprecision of the link. 'Sudden ionospheric disturbances(SID) due to solar flares may occasionally cause pa th anom-alies, but these generally last for short durations andcan be isolated by careful insoection of the data (Keder,1971).In addit ion, over long paths there are also systematicseasonal variations in the propagation delay w h i c h cans o m e times be as large as 15 fi sec (lilima et al, 1953;S w a n s o n & Kugel, 1972). If these are not taken intoaccoun t then they may in trod uce significant inaccuracy inthe time and frequency comparisons. The seasonal variationin the propagation delay occurs main.ly due to variationsin the D-region ionization. This consists of two parts,namely (a) seasonal variation of the Solar Zenith angleat the mid path noon and (b) variations in the mesosphericneutral atmospheric constituents, which are to some extentrelated to meteorological p h e n o m e n a (Belrose 1963). Of theabove, only the first part ,(a),can be modeled with definite-ness. It has been shown by Swanson & Kugel (1972) to beof the fo rm M(l-cosX ), where, CosXis the average of thecosine of the solar zenith angle */ over the path at themidday. M is an empirical constant dependent on thepropagation path and to some extent on the solar activity.TH E DATA(a) KPL (India} data : The m i d d a y phase observations ofthe 16 kHz GBR signal relative to UTC(India) recorded dailybetween 03-30 to 09-30 UT are shown in Fig.l(a) for the

853


4/12

period 10th July, 1979 to 10th July, 1930. Twice duringthe one year time the NPL-1 underwent discontinuities inoperation for short periods. Once during 23-25 August,1979 and again during 30 August - 7 September,1980. Inboth cases it was pos sible to restart UTC (India) by syn-chronizing NPL-1 with NPL-2 and giving suitable time cor-rections (of 1.3 and 5.5 ysec respectively). The timecorrections wer e determined from the intercomparison databetween these two clocks which was taken periodicallythroughout,We ob serve the following features in Fig.l(a). The day today variability is muc h smaller in summer than in winter.The rms jitter in summer is less than 1 psec while inwinter it is about 2.5 /isec. There is a gradual drift inthe phase data arising due to a relative frequency offsetbetween UTC (India) and the transmitted frequency ofGBR.This drift is modulated by a seasonal variation in thepropagation delay. As mentioned in the earlier section,the Solar Zenith angle related seasonal varia tion of theform M (1-cosx ) can be modelled and eliminated. To deter-mine cost, we follow lijima et al (1969). We divide thepropagation path (6700km) into 10 equal segments and deter-mine cosX, at the midpoint of each. The average of thesevalues gives cesfl, over the path. To determine the valueof M we take help of the observed diurnal variation ofphase delay. It has been shown by Swanson and Kugel (1972)that around midpath noon, the temporal variation of phasedelay is also of the form M (1-cosX ). We thus makecomparisons between the temporal variation of the observedphase delay and that of the calculated factor M (1-cos/C )by varying M. The value of M which gi ves best agreementbetween the forms of the two variations is selected.Utilising the whole years data the best value of M=12.5+lwas obtained. In Figs*2 and 3 we have shown as illustra-tions the observed diurnal phase for groups of 15 days inApril and June arid the calculated M(L.cosX ); The goodagreement between the two is evident. Adopting the abovevalue of M the calculated seasonal variation of M(l-cos )at the midpath noon for the full year is shown In Fig.l(b).This shows an annual variation with a summer to winterphase retardation of 7/usee. On subtracting out thisseasonal variation from the observed data in Fig.l(a) weget the resultant variations as shown in Fig.4(a). Thephas e drift now appears more regular but there stillper-sist some short period variations during the periodSept ember, 19 79 to February, 1980 with peak amplitudes of2-3 /usec.

854


5/12

(b) PTB datat The midd ay phase observations of GBR relativeto UTC(PTB) is shown in Fig.4(b). This path is only 790 kmlong and the solar zenith angle related seasonal variationin the phase delay is not expected to be significant. It isnot apparent in the data also. During the period fromSeptember'79 to February'SO, the nature of variations in thePTB data are very muc h similar to NPL (India) data in Fig.4(a). It is notable in this connection that GBR-PTB path isalmost overlapping with 1/8 of the GBR-NPL (India) path. Thisindicates the possibility that the shorter period variationsmentioned abov e are arising from this portion of the path.(c) RGO data: The midda y phase variation of GBR relativeto UTC (RGO) in shown in Fig.4(c). In this .case also thepath length being very short (180km) seasonal variationsare almost absent.(d) USNO data j The midday phase variations of GBR Relativeto UTC (USNO) are shown in Fig.5(a). The propagation pat hin this case is long (5800 km) and so the solar zenith anglerelated seasonal variation is significant. As we do not havethe diurnal phase variation for this station it is notpossible to determine M as was done earlier for NPL data.Inabse nce of anything better we have just adopte d the samevalue of Msl2.5 and calculated the variation of midday valuesof M(L-cosX,). This is shown in Fig 5(b). Subtractingvariations in Fig 5(b) from those in Fig 5(a) we get theresultant as shown in Fig 5(fl)) .In this case also duringSeptember 1979 to February 1930, we observe anomalous varia-tions of similar nature as in PTB and NPL (India) data. Butthese variations are of larger magnitude (~12 j i sec peakto peak)(e) NPL (UK) data : The midday phase observation of GBRrelative to UTC (BPL,UK) showed occasional jumps of 5,10and 15 jusec over the whole year. These were most probablydue to some equipment malfunction. Thus, the data from thisstation have not been used.RESULTS AND DISCUSSIONFor intercomparison between UTC (India) and UTC of theother laboratories, the individual phase data in Figs 4(b),4(c) and 5(c) have to be subtracted from Fig 4(a). InFig 6 we have shown t (a) UTC (India)-UTC (PTB), (b) UTC(India)r UTC(RGO) and (c) UTC (India)-UTC(USNO). It isclear that of the three links the one with PTB has the

855


6/12

minl tnum jitter and anomalies due to propagation variations.In Fig.6 we have drawn regression lines through the pointswhich represent the overall drift rates. The slope ofthese straight lines give the relative frequency offset,S,between UTC (India) and the other UTCscales. Followingresults are obtained.

sIndia,PTB = ( 7.0 0.1 ) x 1013s!ndla,RGO = ( 7.3 + 0.2 > x 10"13SIndia,USNO = ( 6.9 + 0.2 ) x 1013

T he value of slndia,PTB obtained above is in fairly goodagreement with a value of (7.1 + 0.5) x 10- obtained viathe satellite experiment (Mathur et al, 1980) duringMay-June, 1979. This indicates that the UTC (India) has beenmaintaining a fairly constant rate. The relative offsetwith the other two laboratories are also consistent (withinthe uncertainties) with LOHAN-C links connecting these.LORAN-C link results have been calculated as follows :

SPTB,USNO = 0.4 x 10-13SPTB,RGO = 0.0 x 10-13

T o get an idea of the precision of the frequency transferestimates we have made some computations of the variance,, on the India-PTB link which is the best of the threelinks. We get the following values :

^y ( -c - 1 day ) = 2 x 10-11


7/12

on 29 June'79 U TC (India)-UTC(PTB) = (2.7 + 0.1) /usec. TheVLF observations were started on 10 July onwards (as theG B R transmission was off for a month before this). Thus inFig 6(a) we assign a value of 2.7psec to the ordinatewhere the regression line crosses the abscissa on 29 June'79. Since th e difference between U T C (PTB) a n d U T C (RGO)&nd UTC(USNO) are. known from the LORAN-C links it ispossible to calibrate the other two graphs also*P O R T A B L E CLOCK TRIPA portable clock trip from the USNO was m a d e during 18-21,September,1980. This made it possible to evaluate theaccuracy of the time transfer accuracy of the VLFlink. InFig. 7 we have shown the results of only one link U TC (India)-UTC(PTB) extended upto the end of September130. There aretw o crosses in Fig. 7 on 29 June '79 representing the satel-lite experiment time transfer and on 20 September '80representing the portable clock result. The discrepancybetween results of the portable clock and the regressionline predicted by the VLFdata is only 1.5 jisec which isquite small.C O N C L U S I O NIn this paper we have discussed a VLF time and frequencyintercomparison link between UTC (India) and UTC of PTB,R G O and USNO. It is clear that the India-PTB link is thebest of the three. It has been shewn that by taking dueaccount of the seasonal variations over long paths it ispossible to achieve frequency inter comparison to a fewparts in 1014 over one year period. Ti m e transfer can beachieved to an accuracy of 1-2 psec.W e realise, however, that this accuracy level is still notacceptable to BIH for inclusion of U T C (India) in theinternational UTCcoordinated by them. In fact a similarproblem m u s t be faced by many other remote timekeepinglaboratories w h i c h are not covered by LORAN-C groundwavesor regular portable clock trips.O u J ? future plans in improving our links with the inter-national time keeping communi ty are : (a) reception of moreVLF stations possibly O M E G A ( J A P A N , Liberia, L a Reunion),(b) having regular portable clock trips,(c) using the NNSSsatellite signals and (d) using some geostationary satellitelinks on a regular basis.

8 5 7


8/12

ACKNOWLEDGEMENTSW e are highly thankful to Prof.G.Becker of PTB, W estGermany w h o w a s with us at the begtning of the VL Fexperiment and gave us very useful suggestions. W ethankfully acknowledge the receipt of the VLF phase datafrom PTB, H G O , N P L an d USNO. Thanks are also due to theother members of time and frequency group for their helpand cooperation.REFERENCES1. Becker, G., Fisher, B, Kramer , G..M ethods and results

of international VLF time comparison (In Germany Proc.Colloque International de Chronometric, Pris, 1959,2. Mathur, B.S., Banerjee, P., Sood, P.C., Mithlesh Saxena,Kumar, Nand and Srari, A.K., Precise T & F Intercompari-son between NPL, India and PTB, Federal Republic ofGermany via Satellite Symphonie-1, XII Annual PTTIA p p l i c . A Plan. Meeting, Dec.2-4,1980.3, Belrose. -T.S., Low and Very L ow Frequency Radio wavepropagation, A G A R D Lecture series XXIX,1963.4, Reder, F.N. , VLF Propagation Observed D u ri n g Low andHi gh Solar Activity, Progress,in Radio Science :1966-1969, Vol. 2, USRI-Belglum, 1971.5. Swanson, E.R. and Kugel, C.P., CMEGA VLF Timing, NELCTech. Rep. No. 1740 (Revision.!), June 1972.6. lijima, S., Torao, M. and Fujiwara, K., Phase Varia-tions of VLF wave GBR and GBZ as received. Annalsof Tokyo Astron. Observatory Vol. XI, No.I, 1963..

858


9/12

' I I I

I 10 MS (o)UTC(lndio)-GBR(b)M(i-cosx)

J A S O N D J F M A M J J1979 1980

Fig.l (a} Daily values of ( r B T * phase relative to I7TC(India)(b) Calculated seasonal varia tion factor M(l-cbs%)for GBR-NPL path.

OO I 4

Fig. 2 Diurnal variation of GBR phase for 15 daysin ApriLl.SQ compared with the var ia ti on ofM (1-cosX) (doYt'ed line) around midda y.

859


10/12

PhOM Wfltioo ofPropogotion Path G B R N .

I 0 1 h - Z 5 l h Jura 1980

00 2 4 6 6 K> 12 14 16 18 20 ZZ 24

Fig. 3 Same as in Fig.2 for 15 days in June130,

IOn$] I ~\-~ T" J I

; ( a ) U T C (India)-GBR( c o s X variation ebmnoted)

( b ) U T C ( P T B l - G B R. . .(c)UTC(RGO)-GBR

J A S O N D J F M A M J J1-9 79 1980

Fig.4 (a) G3R phase relative t o - U T C ( I n d i a ) correctedfor solar zenith angle variations, (b) G3ii phaserelative to U T C ( P T B ) . (c) GBR phase relative toU TC ( R G O ) .

860


11/12

I I ITlOMs (o)UTC(USNO)-GBR

(b)M(l-cos%)

'''';v";t;."'--(T)UTC(USNO)-GBR;- "". -'(cos% voriotioneliminoted)

J A S O N D J F M A M J J1979 I98O

Fig.5 (a) GBR phase relative to I . r r c ( U 5 N Q ) . (b)Calculatedseasonal varia tion fac tor Md-cosX,)for GBR-USE0path, (c) GSR phase relative to UTC(17SNO) correctedfor solar zenith angle variations.

J A S O N D J F M M J J

'ig. 5 (a) UTC (India) - UTC (PTB).(b) UTC (India).UTC (EGO), (c) FTC (India) - UTC (USNO).

Documents

Precise T & F Intercomparison via VLF Phase Measurements