IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. IM-19, No. 4, NOVEMBER 1970

-110 dB/Hzz _

-130dB/Hz'

-140 dB/Hz -_-

102

..I __ __L= I= =====

f (Hz)

Fig. 14. Itesidual noise of 645A synthesizer at 49.9 MHz.

XVI. CONCLUSIONS

AIL accurate method for the measurement of phasenoise has been explained and the theoretical relationshipsbetween the electrical parameters of the equipment in-

volved and the measures of phase noise we desire toestimate have been developed. Practical calibrationtechniques that can be used to increase the accuracy

of the measurement have been developed. Several typesof performance tests which can be made on different typesof equipment have been explained, and some examplesof the actual performance of the test technique for actualmeasurements have been given. The techniques derivedhere should be of direct use to other researchers in thisfield who wish to make more accurate measurements ofphase-noise performance in various types of equipments.

REFERENCES[1] V. Van Duzer, "Short term stability measuirements," Proc. 1964

IEEE-NASA Symp. on Definition and Measurement of Short-Term Frequency Stability. Washington, D. C.: U. S. Govern-ment Printing Office, NASA SP-80, 1963, pp. 269-272.

[2] D. Halford, NBS Seminar on Frequency and Time Stability,lecture notes, 1969 (unpublished).

[3] D. J. Glaze, "Improvements in atomic cesium beam frequencystandards at the National Bureau of Standards," IEEE Trans.Instrumentation and Measurement, vol. IM-19, pp. 156-160,August 1970.

[4] L. S. Cutler and C. L. Searle, "Some aspects of the theory andmeasurement of frequency fluctuations in frequency standards,"Proc. IEEE, vol. 54, pp. 136-154, February 1966.

[5] D. W. Allan, "Statistics of atomic frequiency standards,"Proc. IEEE, vol. 54, pp. 221-230, February 1966.

[6] M. Schwaritz, Injormation Transmission, Modulation, and Noise.New York: McGraw-Hill, 1959, pp. 429-432.

[7] D. G. Meyer, "An ultra low noise direct frequency synthesizer,"Proc. 24th Ann. Frequency Control Symp. (unpublished).

Comparison of National Time Standards

by Simple Overflight

JEAN BESSON

Abstract-The Office National d'Etudes et de RecherchesAerospatiales has devised a method of time synchronization by whatis described as "simple overflight," relying on a high-precision air-borne timepiece, without the need to go close to the clocks undercomparison. The method has repeatedly been tested, and the resultsshow the accuracy of synchronization to be better than 50 ns. A com-

parative experiment between the observatories of Paris, France,Hailsham, England,^Ottawa, Canada, and Washington, D. C. will becarried out in September 1970.

INTRODUCTION

N IMPROVED accuracy of synchronization be-tween clocks at widely separated stations is an

essential requirement in a great manv differentfields. As a matter of fact, the use of high-precision clocks

Manuscript received May 26, 1970.The author is with the Office National d'Etudes et de Recherches

Aerospatiales, 92 Chatillon, France.

based on atomic frequency standards has by now becomeestablished practice, with the effect of boosting the per-

formance of various systems of trajectography [1], locali-zation [2], navigation [3], geodesy [4] [5], and time distri-bution [6]-[9].A wide range of time distribution methods, keyed to the

requisite accuracyr levels, have been evolved for specificapplications.The main time-synchronizing techniques currently in

use include

1)2)

3)4)5)

reception of time signals,reception of Loran C radio navigational transmis-sions,reception of TV broadcasts,transfer of atomic standards,the use of signals delivered from satellites such as

Transit.

='7"'=227

103

IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, NOVEMBER 1970

Thus the oldest time distribution method, using atransported timepiece, rates at present as the mostaccurate, provided it employs a perfected version of theatomic clock.The Office National d'Etudes et de Recherches Ae'ro-

spatiales (ONERA) has made use of this method on morethan one occasion [14], in particular, in a series of geodesicexperiments it carried out with French satellites of thetype series D, at the request of the Direction des Re-cherches et Moyens d'Essais (DRME), and in conjunctionwith the Service Technique des Constructions et ArmesNavales (STCAN.) On this occasion the synchronizationto within 50 ,us of three test stations situated around theMediterranean was ensured over a period of 3 months.For certain locations (mountain posts, sea marks or

beacons, ships, aircraft, etc.) synchronization by trans-ported precision clocks can be difficult if not impossibleto achieve. Again, there is an unavoidable time lapsebetween measurements taken of each clock in turn, evenwith highly advanced means of transport, such as aircraftor helicopters in use. This introduces a source of errorthrough a shift in the time scale between timings. Syn-chronizing accuracy can only be improved by cuttingdown the time between comparisons.Time synchronization by simple overflight [15], [16], as

described in the following, also rests on the use of a high-precision timepiece conveved from place to place, butdispenses with the need to move close to the clocks to becompared. In this method, the aircraft carrying thetiming clock has a transmitter aboard for sending outthe time scale to be synchronized, at the station which isequipped with a receiver. The aircraft merely has to passnear the station concerned, and transmit its time scale.As a result, even relatively inaccessible stations comewithin reach, and by saving the repeated transshipmentof equipment, the intervals between the timing of theclocks are minimized.

THEORETICAL CONSIDERATIONS-NOTATIONThe main components of clocks used for precise time

synchronization include 1) an atomic frequency standardand 2) a time-scale generator-with-divider. The frequencystandard supplies 100 kHz, 1 MHz, or 5 MHz sinusoidalvoltages, while the time-scale generator divider producesa sequence of electric pulses representing the unit of timeand termed "time scale."Atomic clocks are compared by correlating them on a

comparative basis. The two parameters in general usefor rating the comparative characteristics of two clocksare the inaccuracy measured between two frequencystandards and the deviation between two time scales.For two frequency standards having nearly equal

repetition periods, the instantaneous inaccuracy E atinstant t can be defined as:

E(t) = (T2 - T1)/T2 or T, (1)

where T, and T2 are the repetition periods at instant t.

In practice, E(t) is unamenable to measurement, aninaccuracy E0 measured over a 0 period of time beingconsidered instead. A mean value Eo of the inaccuracytaken over a period of time 0 is obtained by making anumber of successive measurements. A measure of thismean inaccuracy El is provided by its standard deviationo(EO), which obviously depends on the measurement time0.The shift in the time scales is of course bound up with

the inaccuracy.Let two elementary time scales be considered, one made

up of pulses of T, repetition period, the other of pulsesof a closely similar T2 period, the two assumed to coincideat instant t = 0. At the end of a period, the two scalesshow a (T2 - Tj) deviation. Now the value E(t) =(T2 - T,)/T, expresses the shift per unit time. Theshift over time t can therefore be written:

t

D)(t) = E(t) dt. (2)

Introducing the mean inaccuracy E9 and taking accountof the shift over time t as starting from time t = 0, atwhich the initial deviation is D0, we have [14] the scaleshift

D(t) = Do +E0t. (3)This expression is used to analyze the deviation betweentwo atomic-clock time scales.

Let Hs denote the time scale obtained for a clock X,and HV that for a clock Y (Fig. 1). The two time scalesHz and Hy are as a rule compared by the use of a chronom-eter that measures the difference between two pulsesof the same serial number in the two time scales.By convention, the shift or deviation between the two

time scales H, and Hy at instant t is written:

D(t) = (H. - Hy)t. (4)The quantity (H. - H), has a time dimension and isalgebraic if based on the following conventions:(HZ- H,), > 0 when HX is in advance of H,; (H. - H,),< 0 when HZ lags behind H,.

MODE OF OPERATIONGiven two time scales HBl and HB2 at far-apart geo-

graphical locations, synchronization is aimed at as-certaining the difference HB1 - HB2 at a particularinstant. By the method described here, an atomic timingclock HA is conveyed by an aircraft overflying stationsB, and B2 in succession (Fig. 2) with a view to determiningHB1- HA and HB2 - HA*HB - HB2 can be calculatedby correlating these measurements.The novelty of the method lies in the determination

of the deviations HB1 - HA and HB2 - HA throughtransmission to the stations of time scale HA at themoment of overflight, using a transmitter-receiver as-sembly of ONERA design. The schematic diagram shownin Fig. 3 affords a closer insight into the technique of

228

BESSON: COMPARISON OF TIME STANDARDS BY OVERFLIGHT

n-I n nI1

n-4 n nI

ncerrerencFig. 1. Shift between two time scales.

Aircraft

Overflight at instant t1

Aircraft

Clock

HA

Transtmi.si t

overflight at insTtant t2

229

and variations in the reception level that would distortthe results of the time-interval measurement. A suitabledecoder is correspondingly associated with the receiver.The stations accordingly obtain a time scale HA that

lags behind HA. The delay through transmission fallsinto two categories:

1)TER, the time delay produced in the transmitter-receiver system;

2) r,, the time delay that equals the radio-wavepropagation time corresponding to the air-to-groundradial distance.

Thus we have

HA - HA = TER + TIP. (5)

A timing assembly measures the difference HB -HAfrom which is derived:

HB - HA = (HB - HAI) - (TER + TP)

Station 1

A HB2-HA

2BClock B2

Station 2

Fig. 2. Principle of synchronization by overflight.

Fig. 3. Functional diagram.

measuring the difference between the local and the air-borne clock.The time scale of the airborne HA is sent out from the

aircraft and received on the ground by a transmitter-receiver specially designed to transmit and receivesteep-fronted narrow pulses. To this end, the HA clockpulses are encoded (transmitter control) as a protectionfor the time-keeping equipment against spurious signals

(6)A recorder registers the H.B - HI values at the pulserepetition rate of the time scales used. By evaluatingTER and Tr, HB - HA can be obtained from (6),

1) TER is an equipment constant measurable to withinthe accuracy of the timing setup;

2) T, depends on the aircraft position relative to thestation, and is determinable by various methods, twoof which are described in the following.

METHOD 1

Here the aircraft is instructed to pass directly over thestation at a constant altitude. During its passage, theHB- HA values are recorded. The real deviation HB - HAdoes not change during the measurement (approximately1-2 minutes), while HB - HA decreases as a result of thedelay due to propagation, with the progressive advanceof the aircraft towards the station and reaches a minimumwhen the aircraft is directly over the station. At thisinstant the radial distance equals the altitude indicatedby the airborne equipment.There are here three potential sources of error incident

to the determination of propagation time Tr,, for example,

1) vertical error,2) altitude error,3) error due to the discontinuity of recording and to

the possibility of the true minimum value occurringbetween two measurements.

Using a low-flying (100-300 meter altitude) aircraftat a flight speed of about 50 m/s (180 km/h), the experi-ment showed that the angle a, the minimum deviationfrom perpendicularity, should be less than 10 degrees.At the rate of 10 timings per second, and with a 3 percentaltitude error, the propagation time inaccuracy falls below30 ns.

AT, < 30 ns. (7)

IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, NOVEMBER 1970

METHOD 2

This method dispenses with the determination of air-to-ground propagation time by a two-way communicationbetween aircraft and ground station. The time scale ofthe airborne clock is transmitted to the station, and thatof the local clock to the aircraft. Two timings done con-currently in the air and on the ground eliminate propaga-tion time 'r.

In concrete terms, note the following.

1) HA and HB denote the two time scales to be cor-related.

2) TER (AB) and rER (BA) are, respectively, the air-to-ground and ground-to-air transmitter-receiver systemdelays.

3) r,, stands for the propagation time over the givenradial distance. Let it be assumed that the two time scalesdo not deviate markedly (not more than by a few milli-seconds) so that the aircraft position may be taken asfixed for the two-way measurement of two pulses of thesame serial number.The locally received time scale is HA, and HB- HA is

measured. From (6), we have

HB - HA = (HB - HO) - (7rER(AB) + rP)* (8)

Correspondingly, the time scale received in the aircraftis HB1. The measurement gives:

HA - HB = (HA - HB) - (rER(BA) + r,). (9)

If (8) and (9) represent two concurrently performedmeasurements, the elimination of T, between the twoequations leads to:

HB - HA =

(HB - HA) - (HA- HB)2

,rER(BA) - TER(AB)

2 (10)

The only limitation on accuracy is here set by the meas-

urement of the deviations and by the definition of theequipment constants.On the other hand, the aircraft overflies the station

without any special precautions as to altitude or whetherit flies directly over the station.

EQUIPMENTThe functional diagram in Fig. 3 displays the various

equipment units required for the aircraft and the groundstations in the experiment. For the two-way communica-tion, the station is equipped with a transmitter assembly,and the aircraft carries receiving and measuring devices.The frequency standards are as a rule rubidium-vapouror cesium-beam generators. The time scales used are

provided by a pulse sequence of 10-Hz repetition fre-quency, thus admitting of 10 deviation measurementsper second.The transmitter controls were evolved by ONERA to

meet the specific requirements of trajectory studies andairborne time signalling. They are designed to furnish atime scale capable of transmission by a definite code andin perfect keeping with the master clock.

This system provides a 10-Hz rate time scale with thefollowing pulse characteristics:

1) amplitude, 5 volts;2) rise time, < 10 ns;3) half-amplitude duration, 300 ns.

Via a parallel channel, the 10-Hz pulses are encoded fortransmitter control; every 1/10 second the single pulsegives way to a 10-pulse train of 1-,s duration each,spaced 100 As apart. The fluctuations in the pulse-leadingedges of the time scales are of the order of 3 ns between0°C and 50°C.The time-scale transmission is amplitude modulated

on a 2.2- or 2.3-GHz carrier having a 10-MHz passband,and the transmitter peak power is 50 watts.The receiver and associated decoder help to convert

the 10-pulse word into a single pulse with the followingcharacteristics:

1) amplitude, 3.5 volts;2) duration, 200 ns;3) rise time, < 10 ns.4) variation of leading edge with reception level, < 10

ns/30 dB.

The equipment at ground stations dealing with devia-tion timing and aboard the aircraft in case of two-waycommunication is composed of the following items: 1) achronometer of 10-ns resolution and 2) a printer operatingat the rate of 5 or 10 measurements per second and re-cording the deviations timed by the chronometer togetherwith the corresponding clock time as given by the clockin terms of 0.1 second, seconds, and minutes.

RESULTS

The effectiveness of this method of time synchroniza-tion is well illustrated by the results of a comparisonbetween the times of the observatories of Paris [BureauInternational de l'Heure (BIH)] and of Braunschweig[Physikalisch Technische Bundesanstalt (PTB)] made onJune 23-24, 1969.

This undertaking, organized under DRME sponsorshipand named Synchronisation Franco-Allemande-Franco-German Synchronization (Operation Synfral) was basedon the principle described, using two-way communicationat the Paris observatory and a single-channel (low-flyingaircraft) at the Braunschweig PTB.The airborne clock is described by HA (Horloge Avion),

and the clocks of the Paris and Braunschweig observa-tories are described by HSP (Horloge Synfral Paris) andHSB (Horloge Synfral Braunschweig), respectively. TheParis observatory was overflown at about 14.00 hoursuniversal time (UT), the Braunschweig PTB at about16.00 hours UT on June 23.

230

BESSON: COMPARISON OF TIME STANDARDS BY OVERFLIGHT

For the case of one-way communication, the calculationof the HB - HA deviation values obtained bv one-waytimings included:

1) determination of the HR - HAI minimum as a func-tion of time;

2) correction for minimum propagation time;3) computation of HR - HA and assessment of the

errors incurred.

The H, - HA minimum was determined on the strengthof measurements as exemplified in Fig. 4.To improve on the accuracy of the graph, the applica-

tion of mathematical smoothing offers an advantage.One method used is the fitting by least squares of a second-degree polynomial representing HB - HA"

Propagation time r, was corrected for by drawing oninformation obtained from the Centre d'Essais en Vol(Flight Testing Center), the relevant data comprisingthe flight altitude and flight path relative to the groundreceiving aerial.

Referring to the experiment

(6)HB - HA = (HB - HA) - (TE R + Tr),

the mean-square error can be defined as:

T(HR - HA) = VoS0(HB HA) + Of(TER) + 0f(7P) (11)

O(HR, -HA) is known after smoothing; cT(r,) < 30 ns asseen in (7); 0-(TER) < 10 ns.

In two-way communication. propagation time no longeraffects the measurements, and the difference HR - HA isdirectly obtained through relation (10), Fig. 5.HR - HA is constant in theory over the actual period

of an overflight (some seconds). All that remain to becalculated, therefore, are 1) the mean value of the devia-tion during an overflight and 2) the mean-square error.

In the light of relation (10), the error in HB - HAdepends on the error in the half difference of the groundand air measurement values and on the error in the meas-urement of equipment constants:

a(Hs- HA) = V/o_2(measured values) + o-2(equipment)

where o- (measured values) is given by smoothing ando- (equipment) < 10 ns.For each flight over the observatories, the resulting

curves were similar to those shown in the figures. Takingthe mean value of the different overflights as the endresult for the deviations, the relevant findings were asfollows.

1) for the Braunschweig PTB in onte-way communica-tion note the following.

16.10 hours UT, June 23, 1969:

HSB - HA = 2335.89 ,us oJ(HSRB - HA) < 25 ns. (12)

08.10 hours UT, June 24, 1969:

H(HSR - HA) < 30 ns. (13)

HsB_-i1

(,us)

3230,000 _ _

3233,50¢

0 20 40 s0 80 100

Fig. 4. Mleasurement of deviations in one-way communication.

fgs,0o0o1o0.too

0oo .

600

191,500.

400.

300.oo0.too

tIfl, 000.

-_ 1JA) overflight

(P5)overflight 2 on June 24 th

overflight on June 23 rd

lOans

SO '2 A(6 20

Fig. 5. Measurement of deviations in two-way communication.

2) For the Paris observatory in two-way communica-tion note the following.

14.00 hours UT, June 23, 1969:

HSP- HA= 1998.47 ps o-(H - HA) < 20ns. (14)10.15 hours UT, June 24, 1969:

HSP-HA = 1998.68,US (HspSP- HA) < 20 ns. (15)

The final synchronization result as given here allowsfor the fact that the deviations HSP-HA and HSB- HAwere measured at different instants. Bearing in mind thedrift of the clocks between the overflights, the deviationHSP-HSB was determined at a given instant, for exam-ple:

16.10 hour.s UT, Junic 23, 1969:

H,S - HA = 1998.49 Us ±t 20 ns

HSR - HA = 233.5.89 /is ± 25 nis

(16)

(17)

PLANNED EXPERIMENT

An experiment to be conducted in September 1970will be aimed at correlating the time scales of severalnational observatories to within an accuracy of some tensof nanoseconds. The observatories concerned will includethose of England (Royal Greenwich Observatory, Hail-

we_. n-,

- , 'go.. .. V - ..

231

100ns L,4/1o s

H,,, H_, = 2335-77 As

IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, NOVEMBER 1970

sham), the United States (US Naval Observatory, Wash-ington), Canada (National Research Council, Ottawa),and France (Observatoire de Paris).The experiment will adopt the principle, here described,

of simple overflight of the observatories. At the Wash-ington observatory, it is planned to carry out synchroniza-tion by two-way communication. The master clock willbe carried aboard a DC7 aircraft of the Centre d'Essaisen Vol (CEV), Amor Section, Bretigny.On the outward flight, the itinerary will be Paris-

Hailsham-Ottawa-Washington, with the homeward flightfollowing the reverse route. Under the planned experi-mental conditions of three or four overflights over eachobservatory, a nonstop flight either way, outward orreturn, will take about 18 hours.For easier time-deviation measurement, the use of

time scales in universal time coordinated terms wouldseem both more meaningful and more convenient.

This comparative experiment should ultimately resultin correlating national time to within better than 50ns.

CONCLUSION

The accuracy of time synchronization by simple over-flight has been found to be of the order of some tens ofnanoseconds; errors are traceable partly to the resolvingpower of electronic circuitry (transmitters, timing setup),partly to aircraft position finding.An improvement in the accuracy of the procedure can

be affected by timing to within the nanosecond order ofmagnitude, and by further perfecting time-scale trans-mission and reception facilities. These advances are feasi-ble at short notice. On the other hand, there is littlelikelihood of higher accuracy levels being achieved inaircraft position fixing with the aids available at present.

There is, then, all the more to be gained by the use of thetwo-way communication method, which dispenses withthe need to determine propagation time and lends itselfto synchronizing time within the range of a fewnanoseconds.

REFERENCES[1] R. Moreau, "Nouveaux dispositifs de trajectographie," Onde

Elec., vol. 491, pp. 95-107, February 1968.[2] P. Fombonne, "Navigation et localisation a la mer," Navig.,

pp. 255-279, July 1968.[3] R. H. Waldman, "Les systemes de navigation pour les avions

subsoniques de transport civil dans les annees 1970," Navig.,pp. 303-310, July 1969.

[4] J. Kovalesky, F. Barlier, and I. Diellmacher, "Premiersr6sultats dus aux mesures Doppler sur le satellite frangaisDiapason," Recherche Spatiale, pp. 1-6, August-September1968.

[5] R. Moreau, "Participation de l'ONERA aux operations degeodesie au moyen du satellite Diademe," Aerosp. Res., pp.51-52, July-August 1968.

[6] B. Guinot, "Temps atomique et temps coordonne," ParisObser., p. 1, February 1969.

[7] P. Parcelier, "Comparaison des fr6quences," Rev. GMn. Elec.,vol. 78, pp. 287-290, March 1969.

[8] G. Becker, "Die Neudefinition der Sekunde und das Problemkunftiger Definitionen von Zeitskalen," PTB Mitt., pp. 270-275, April 1968.

[9] G. E. Hudson, D. W. Allan, J. A. Barnes, R. G. Hall, J. D.Lavanceau, and G. M. R. Winkler, "A coordinate frequiencyand time system," Proc. 23rd An. Freq. Cont. Symp. USA COM.,May 6-8, 1969.

[10] J. C. Husson, "Rapport sur les problemes de synchronisation,"CNES, pp. 1-14, 1967.

[11] L. D. Shapiro, "Time synchronization from Loran," IEEESpec., vol. 5, pp. 46-55, August 1968.

[12] P. Parcelier, "Developpement des synchronisations de tempspar la t6levision," Colloq. Internat. Chronom., Paris, pp. 1-6,September 1969.

[13] L. N. Bodily, "Correlating time from Europe to Asia withflying clocks," Hewlett-Packard J., vol. 16, pp. 1-8, April 1968.

[14] J. Besson and J. Cumer, "Synchronisation de bases de mesurespar transport d'une horloge atomique," Assoc. Tech. Marit.Afron., pp. 1-24, April 1967.

[15] J. Besson, "Synchronisation d'horloges atomiques par simplesurvol," Aerosp. Res., pp. 48-52, March-April 1969.

[16] J. Besson and J. Cumer, "Synchronisation pr6cise de bases parsimple survol," Colloq. Internat. Chronom., Paris, pp. 1-24,September 1969.

232