A brief history of the Time Department from 1923 to 1948

The following account is an extract from a lecture, (which had a seperate section devoted to the Chronometer Department), that was given by the Astronomer Royal, Sir Harold Spencer Jones, on 27 January 1949. Click here to read the text of the lecture in its entirety.



The provision of a time service is a normal function of a national observatory and is closely related to the work of the meridian department. The right ascension of a star is the sidereal time of meridian transit of the star. In positional astronomy a selected number of the brighter stars in the equatorial belt, suitably distributed round the sky, are selected as ‘clock stars’. The positions and proper motions of the clock stars, as used at Greenwich, have been derived from the long series of meridian observations and are progressively refined as the observations continue. Observations extending from 6 to 12 hr. serve to control periodic errors in the right ascensions of the clock stars. Using these adopted right ascensions, the observations of the clock stars determine the errors of the standard sidereal clock night by night. The right ascension of any other star is derived by correcting the observed sidereal time of its transit for the clock error, interpolated for the time of transit.

Before 1927 the time determined at the Royal Observatory was based upon observations with the Airy transit circle. In 1926 a world programme of longitude determinations, in which a large number of observatories in all parts of the world participated, was undertaken under the auspices of the International Astronomical Union. For this special programme, a small reversible transit circle was used at Greenwich, the telescope being reversed near the middle of each transit, thereby eliminating the correction for collimation error. For such a programme it is necessary to adopt a common system of star places, in order to ensure that the derived longitudes are as free as possible from systematic errors in star positions determined with different instruments; in this particular programme the star places in Eichelberger’s Fundamental Catalogue were used. It was found that the clock errors derived from these observations had a much smoother run than the clock errors derived concurrently from observations with the Airy transit circle. The latter are affected by obscure instrumental errors, [...]. It was found, moreover, that when the smoother clock errors provided by the small transit observations were used for the reduction of the Airy transit circle observations, the derived right ascensions of the stars became discordant; but that when the more irregular errors given by the transit circle observations were used, the derived right ascensions became accordant. The instrumental peculiarities of the transit circle are evidently involved.

Since 1927, therefore, the time determinations have been based entirely upon observations with small transit instruments, which are reversed in the middle of each transit. By resolution of the International Astronomical Union the revised Auwers fundamental system, known as the FK 3, is used for the system of star places. The apparent places of the stars are taken from the annual volume Apparent Places of Fundamental Stars, published by the Nautical Almanac Office.

Prior to 1923 the standard clocks employed in the Royal Observatory were regulator clocks with Graham dead-beat escapements. A Cottingham clock, fitted with a Riefler escapement, installed after the first World War, was expected to give a higher standard of performance, but failed to come up to expectations. The development of the Shortt free-pendulum clock introduced a new standard of precision in time-keeping. In this type of clock a master pendulum, mounted in an airtight case, exhausted to a pressure of about 1 in. of mercury, and placed in a constant-temperature room, synchronizes a slave clock of commercial type. The master pendulum swings freely, except when given a small impulse once each half-minute, and is relieved of the work of moving a train of wheels to show the time on dials and of sending out signals. The first clock of this type, Shortt no. 3, was installed at the Observatory in November 1924 and soon showed its superiority over other types of pendulum clock. Other clocks of this type were therefore installed and used both as sidereal and as mean time standards in the Observatory.

The introduction of the new standards resulted in an important change in the system of time employed. The transit of the first point of Aries or the vernal equinox defines the beginning of the sidereal day, 0 hr. sidereal time. But the precessional motion of the true equinox is not uniform, being affected by irregularities, due to solar and lunar perturbations, which are known as nutation. In consequence the sidereal day varies slightly in length. If we imagine a point moving uniformly along the equator, with a motion equal to the mean motion of the true equinox, and so that its extreme distances from the true equinox on both sides are equal, we may term this point the mean equinox. The sidereal time determined by observation is apparent sidereal time. A mean sidereal time can be defined by reference to the mean equinox, in which all days are of equal length; it is obtained by subtracting the nutation from the apparent sidereal time. Apparent sidereal time was good enough before the introduction of the Shortt free-pendulum clocks; their superior precision made it necessary to introduce the concept of mean sidereal time, which has been universally adopted. The detailed study of the performance of the Shortt clocks at Greenwich, which proved their superiority over other types of pendulum clocks, stimulated their introduction into observatories in many parts of the world.

The Royal Observatory is responsible for the distribution of time to the public. The first steps in the distribution outside the Observatory became possible with the development of telegraphic communications. In 1852 an electric clock was installed at the Observatory, which transmitted a signal each day that caused a time-ball, on the offices of the Electric Telegraph Company in the Strand, to drop. In 1865 signals were sent hourly to the Electric and International Telegraph Company’s office, whence they were distributed over the railway network of the country. After the telegraph system was taken over by the Post Office, in 1870, a complete system for the hourly distribution of time through the Post Office was gradually developed, which made Greenwich time widely available. A further step in the widespread dissemination of accurate time followed naturally upon the development of broadcasting. Two of the Dent regulator clocks were modified to run as synchronized clocks under the control of one of the mean time free pendulums, and were provided with a system of contacts to enable time signals to be transmitted automatically every quarter of an hour to the British Broadcasting Corporation. The signals were in the form of six dots at intervals of a second, the last coming exactly at the hour, the quarter, or the half hour. These signals are the familiar B.B.C. ‘six pips’ Greenwich time signal, which, since 5 February 1924, have been transmitted on all B.B.C. wave-lengths several times daily.

A further service, designed to be of value for navigation, was commenced on 19 December 1927. From that date radio time signals have been sent out on a frequency of 16 kc./sec. twice daily, at 10 and 18 hr. G.M.T., from the Observatory, via the Rugby wireless station. These signals, which last for 5 min., are of the so-called vernier type, spaced 61 to the minute, enabling the error of a chronometer to be accurately determined by observing the instants of coincidence between the signals and the ticks of the chronometer. Corrections to the times at which the signals were emitted are published by the Observatory at approximately monthly intervals, for use where higher precision is required, as, for instance, in survey operations. For the distribution of these signals a special ‘diminished seconds’ slave clock was installed, whose pendulum swings 61 times in a minute, and which is kept in synchronization by the mean-time master pendulum. The service has more recently been extended by simultaneous transmission of the time signals on several short wave-lengths. The introduction in 1936 by the Post Office of the ‘speaking clock’, designed by the Post Office engineers and constructed at the Post Office Research Station, which is automatically controlled by hourly time signals from the Observatory, has made accurate time continuously available.

It is necessary that the Observatory should keep abreast of developments in precision horology so that the time service can meet all demands for precision that are made upon it. After the development of the quartz crystal clock, it soon became evident that a new standard of precision in time-keeping had been reached. It was therefore decided to instal[l] a clock of this type at the Observatory in order that some direct experience could be gained of its performance in comparison with the performance of the free-pendulum clocks. A clock, using a Dye-Essen ring crystal, was constructed under the supervision of the National Physical Laboratory and installed in 1939. The quartz vibrator was adjusted to have a frequency of 100 kc. per sidereal second; demultiplier circuits provided an output with a frequency of 500 cycles per sidereal second, which was used to drive a phonic motor. Although the performance of this clock was not altogether satisfactory, experience showed that it was free from the small erratic changes of rate to which the pendulum clocks were liable and which made prediction uncertain; it was found also that it would be more convenient, when additional quartz clocks were installed, to employ a fundamental frequency of 100 kc. per mean solar second instead of per mean sidereal second.

The outbreak of war stopped for a time developments which had been planned for an improved time service. A skeleton time service was at once installed at the Abinger Magnetic Station as a safeguard against the possibility of the service from Greenwich being put out of action. Towards the end of 1940 when the frequent air raids made night observations impossible at Greenwich, the time service was moved in its entirety from Greenwich to Abinger. Shortly afterwards a second time service was installed at the Royal Observatory, Edinburgh, with the co-operation of the Astronomer Royal for Scotland. For the remainder of the war, the two time stations, at Abinger and Edinburgh, were in use; the two stations were connected by direct land line with teleprinter communication, enabling clocks at either station to be recorded at the other station.

Plans were proceeded with for the installation of quartz-controlled frequency standards. The first group of three clocks was installed in 1943. These clocks, which were constructed at the Post Office Research Station, were of the Post Office Group IV type, in which a GT-cut plate of quartz is maintained in oscillation at 100 kc. per mean time second, by a bridge drive circuit, which reduces to a minimum the effect of variation in the supply voltages. The crystals are mounted in thermostatically controlled ovens, and the temperature range permitted by the modified Turner circuit employed ensures that variations in frequency from this cause are small. Regenerative frequency dividers are used to provide an output at 1000 c./sec., which can operate phonic motors, provided with contacts from which signals can be taken off.

The installation has since been considerably extended and the time service is now based upon six groups of three quartz crystal clocks, four of the groups being at Abinger, where the main time-service station has remained, and two at Greenwich, enabling a time service of much higher precision to be provided than was possible with pendulum clocks. The quartz clocks have the further advantage that relative errors and rates can be obtained much more readily and with much greater accuracy than with pendulum clocks. For measuring time and frequency differences, decimal counter chronometers are used, embodying scale-of-ten counter circuits. The 100 kc./sec. output from one of the primary standards is fed into the counting unit. A seconds impulse from one clock can be applied to start the count and a seconds impulse from another clock to stop it. A single reading of the time difference between the two clocks is obtained, being shown on five dials reading successively in units of tenths, hundredths, thousandths, ten-thousandths, and hundred-thousandths of a second. To provide a check, the roles of the two clocks can be interchanged.

Alternatively, by throwing over a switch, a series of successive readings may be obtained, when the results are added in the counter. This procedure is of value when the intervals are nominally constant but are subject to small erratic variations, as when comparing a clock with a time signal; ten consecutive readings can then be obtained with advantage. For frequency comparisons, the nominal 100 kc./sec. outputs from two clocks are fed into a comparator, where the levels are adjusted to a standard value. The adjusted outputs are then combined and the resultant beats are shown on a meter. By means of a trigger circuit, a pulse is sent to the decimal counter when the beat-frequency voltage passes through zero. The intervals between the beats are thus accurately timed, enabling frequency differences to be determined with an accuracy of at least one part in 1010. The quartz clocks are provided in addition with automatic beat counters; by automatic counting of the number of beats between each pair of oscillators in a 24 hr. period, the change in the time difference between each pair of clocks in the course of a day is recorded in units of 10-5 sec. The intercomparisons between each pair of clocks, in each group of three, provides an automatic check against incorrect action of any of the beat counters.

An additional advantage of the quartz crystal clocks is that though they are rated approximately to mean time they can serve all purposes. Separate sidereal time clocks and diminished seconds transmitters for the rhythmic time signals are no longer required. The phonic motors, driven by the 1000 c./sec. output from one of the clocks, can be adapted to these additional requirements. In the phonic motors used at the Royal Observatory, the rotor consists of a laminated iron ring with 100 teeth cut on its inside surface. The six-pole stator within the rotor has corresponding teeth cut on its pole faces. The motor runs synchronously at 10 revolutions per second, driving by gearing a commutator wheel at one revolution per second. A contact spring, bearing on this wheel, closes an electrical circuit for one-tenth of each second. In order to obtain sidereal seconds, a special gearing is used. The gearing ratio employed 119/114 x 317/330, which is about four parts in 109 smaller than the correct ratio. This small error in the gear ratio is immaterial, for when the rate of the clock in mean-time milliseconds has been determined, the rate of the sidereal impulses, in sidereal-time milliseconds, can be at once inferred.

The rhythmic time signals consist of a long dash at the minute, followed by a series of dots spaced at intervals of 1/61 min. These signals are derived from a contact drum, which is driven through a 60/61 gear from a phonic motor. In order that the signals may be sent out at the desired instants, a phasing adjustment is provided for the signal transmitter. There are day-to-day variations in the time lag introduced by the land-line joining the Observatory to the Rugby wireless station. Some test signals are transmitted a few minutes before the time signals themselves; these test signals are received and recorded at the Observatory and by comparing the times with those of the outgoing signals, the land-line lag is deduced. An adjustment to correct for the lag is made by rotating the phonic motor stator, thus advancing or retarding the phase of the rotor. A rotation of the stator housing by 360o advances or retards the contact time by one-tenth of a second.

The most recent phonic motor equipment installed at the Observatory provides a complete contact assembly for the control of all the time signals sent out, including not only the rhythmic signals but also the B.B.C. ‘six pips’ time signals and the hourly signals for the control of the Post Office speaking clock.

The use of quartz clocks has resulted in a much improved precision in the time service. Their freedom from small erratic changes of rate makes accurate short-term prediction possible. This is of importance for the control of precision frequency standards, which can be checked against a 24 hr. time interval of high precision. The Rugby 10 hr. time signals are transmitted so as to give an accuracy in the 24 hr. interval between the signals on consecutive days which does not normally exceed 1 msec. The difficulties of accurate long-term prediction, which is needed to carryover periods during which no time determinations can be obtained, are much greater. The quartz crystal oscillators are subject to an ageing effect, which causes a drift in frequency, more rapid at first and gradually decreasing, though never, as far as present experience is a guide, completely disappearing. Time determinations extending over some months are needed in order to derive the frequency drift with the accuracy needed for prediction. But a complication is introduced by the motion of the Earth’s poles, which causes small displacements of the meridian; the result is that a perfect clock, compared with absolutely accurate time determinations, would appear to have a small variable error, which at Greenwich can amount to about ± 25 msec. As the polar motion has two principal components, with periods of 12 and 14 months, an incorrect determination of frequency drift is inevitable unless the effects of polar motion can be allowed for. The motion of the pole along the meridian can be determined by observing the changes of latitude which result from it; the motion in the perpendicular direction can be determined only from observations of latitude variation at another observatory, differing by about 90° in longitude. The determinations of latitude variation at Washington, longitude 77° W, are communicated regularly to Greenwich and are used to correct for the effects of polar motion and thereby to derive more accurate values of the frequency drifts of the clocks. It is noticeable that the application of these corrections has appreciably smoothed the apparent errors of the clocks.

The introduction of quartz crystal clocks has demanded an improvement in the precision of the time determinations. The period of several months, which is required for a satisfactory determination of frequency drift with the relatively large errors inherent in the time determinations with the small transit instruments, could be much reduced if an appreciable reduction in the errors of observation were achieved. Much consideration has therefore been given in recent years to the design of new instruments for time determinations.

A photographic zenith tube, which is expected to reduce the probable error of a time determination to a few milliseconds, is now in an advanced stage of construction. This instrument is based on Airy’s design of the reflex zenith tube at Greenwich, with modifications to adapt it for photographic observation due to F. E. Ross and incorporated in his photographic zenith tube, now in Washington. The essential principle is the employment of a zenith telescope, whose tube contains a mercury horizon to reflect the light and to bring it to a focus in the second Gaussian point of the objective, thereby making the observations practically independent of any error of level. The important modification introduced by Ross was the inversion of the objective, placing the flint component uppermost, whereby, with an appropriate separation between the two components, the second Gaussian point is brought a few millimetres below the lower face of the crown component. The observations are made photographically, the photographic plate being mounted in the Gaussian plane. The instrument was designed originally for the measurement of the variation of latitude; the upper portion of the instrument, which carries the objective and the photographic plate, is in the form of a rotary, which can be turned through exactly 180o. If two exposures are made on a star, the rotary being turned through 180o between them, the separation between the two images in the direction of the meridian is twice the zenith distance of the star. In practice, exposures of finite length are given, the plate carriage being travelled along during each exposure with the speed of motion of the star image. If the two exposures are accurately timed and are approximately symmetrical about the instant of meridian transit, the time of transit can be inferred from the small relative displacement of the images in the direction perpendicular to the meridian.

The advantages of this type of instrument for time determination are considerable. The observations being photographic, personal equations are eliminated. Errors of level do not affect the observations; there is no collimation correction to trouble about; observations in the zenith are independent of azimuth error. As the instrument is fixed, the various sources of error to which a moving’ instrument is liable cannot occur. A longer focal length can be used than is possible with a moving instrument, with the advantage of a correspondingly greater scale. Observations are restricted to the zenith, where atmospheric transparency is highest and refraction effects are at a minimum.

The instrument which has been designed at Greenwich differs in a number of important respects from the Washington instrument:

(i) It has a larger aperture (10 in.) and longer focal length (135 in.).

(ii) A plain ball-bearing is used for constraint of the rotary, in place of conical bearing, in order to reduce friction and to facilitate construction.

(iii) An autocollimation method is used as a criterion of the angle of reversal of the rotary.

(iv) As a fixed axis of rotation is not required for (iii), a definite constraint in the horizontal plane is not needed. The two working orientations are each defined by a pair of stops instead of by a single stop.

(v) Adjustments to the objective are provided for squaring-on and for coincidence of the nodal plane and photographic plate.

(vi) Automatic reversal is accomplished by means of a system of wires which exert a pure torque on the rotary and therefore no tilting torque on the tube. The system is such that unidirectional rotation of the driving shaft is converted into reciprocating rotation of the rotary.

(vii) The plate carriage is annular and the plate-holder mount is circular so that symmetry of diffraction pattern is secured. The carriage constraints are external to the aperture.

(viii) Relative motion of the carriage and rotary is made to approximate to pure translation by means of a compensating system of flexed rods, which constrain the carriage in the horizontal plane to which the motion is restricted by means of three balls that roll between horizontal planes.

(ix) Uniformity of rate in the relative translation of carriage and rotary is obtained by a specially designed system comprising a differential roller and metallic tapes.

(x) The time scale is produced photographically by means of a clock-controlled lamp giving flashes of very short duration. An independent chronograph is not required.

(xi) The height of the mercury surface is accurately adjustable and, as criterion of adjustment for constancy of scale value, an optical null method has been introduced for use in conjunction with a suspended silica rod.

Some consideration has also been given to the design of a new type of transit instrument, designated as the Horizontal Transit Instrument. The essential feature is that the telescope system remains fixed (though adjustable) with its axis horizontal and in an east-west direction. The light from a star of any declination, near the position of meridian transit, is directed along the optical axis by a subsidiary optical system of constant deviation, which can be rotated about an east-west axis and can be set to the appropriate declination. The effect on time determination of its positional errors (whether due to maladjustment of the axis of rotation or to pivotal errors) is reduced to the second order. Level and azimuth errors of the telescopic system have the same effect on the observed time of transit as they do with the ordinary transit instrument; but since the telescopic system is not deliberately subjected to gross mechanical disturbances and suffers from no pivotal errors, these level and azimuth errors should be far more stable than in the reversible instrument. The collimation error is dealt with by duplication of the telescopic system and reversal of the subsidiary system, so that the essential advantage of the reversible instrument is not sacrificed. Observation is made at the common focal plane of the duplex telescopic system, from the two sides successively. The fixity of the telescopic system avoids errors due to flexure, and permits of the use of a focal length considerably greater than can profitably be used in the ordinary reversible instrument. The level is determined with reference to two mercury surfaces, one at each end of the instrument, by means of an autocollimation method.

Instead of following the star image with a movable micrometer wire, a variable-deviation system is used by which the light in the telescopic portion of the instrument is kept always axial as the direction of the incident starlight rotates. In this way the tolerances of certain essential adjustments are greatly increased. Further, this variable-deviation system acts also as a micrometer and as the means by which signals are sent to the chronograph. An additional advantage of this axial method is that the fiducial line that bisects the star image is not required to move in order to follow the star’s image or to be linked to the signalling system as at present. Thus no mechanical errors are introduced at this point. An optical method is contemplated for defining the position of the variable-deviation system in such a way that in its performance as a micrometer or signal emitter the system will be effectively free from the effects of mechanical errors. A thorough examination of the theoretical aspects of the design has been completed.