Chapter 2

Time Scales, Epochs, and Intervals

Time is an easy concept to envision. Simply put, it is a nonspatial continuum in

which events occur in apparently irreversible succession from the past through
the present to the future. However, the term is also used to describe an interval
separating two points on this continuum, or duration. Or it can refer to the length
of this duration. It can be a number representing a specific point on this
continuum, reckoned in hours, minutes, seconds, or in some other scale of
measurement. It may denote a system by which such intervals are measured or
such numbers are reckoned, as solar time. It may define a season (e.g., harvest
time) or an era in history (e.g., war time).

Time can therefore be a daunting concept to grasp and quantify, just because
there are so many ways in which it is defined, measured, and applied. Even in
scientific usage, the term spans a large repertoire of contextual applications.

The attributes of time that are important in metric prediction are

1. Epoch, or instant.
2. Interval, defined by two instants.
3. Duration, or interval length
4. Scale, or unit of duration measure
5. System, or application context
Because a number of timescales and reference systems apply to metric prediction
generation, conversions among systems of measure are also of great importance.

1
2 Explanatory Supplement to Metric Prediction Generation

2.1 Epochs
An epoch is an arbitrary fixed instant of time or date used as a chronological
reference datum for calendars, celestial reference systems, star catalogs, and
orbital motions. The standard epoch used within the MPG is noon on January 1,
2000, otherwise known as J2000, which is the start of the Julian year 2000. The
Julian Date of this epoch is JD 245,1545.0. Other instants of time within the
MPG are calculated as interval points having duration measured in seconds past
J2000. Instants before J2000 have negative numeric values.

2.2 Calendars
A calendar is a system of organizing units of time (e.g., days) for the purpose of
reckoning time over extended periods. By convention, the day is the smallest
calendrical unit (the measurement of fractions of a day is classified as time
keeping). Actually, “day” here means a day and a night. To avoid ambiguity
whether “day” means “the daylight hours” or “a day and a night,” calendricists
use the word nychthemeron (pl. nychthemera) for the 24-hour period spanning
a day and a night. Generally, “days” as used in this work refers to nychthemera.

According to a 1987 estimate cited in the Explanatory supplement to the

Astronomical Almanac (ESAA), there are about forty calendars used in the world
today. This chapter, however, is limited to only those appear in MPG
computations, with a short history or description of each. For a fuller
description, consult the ESAA ([Seidelmann1992]).

Early calendars commonly counted years from the beginning of the rule of a
King, Emperor, or leader (regnal years). The Romans counted from the start of
the reign of the Emperor or Caesar and reset to one when the next Emperor took
over. Alternatively, they counted from the founding of Rome, and so indicated by
the letters AUC (ab urbe condita).

The Julian calendar was introduced by Julius Caesar, in what we now denote as
45 BC, as a replacement for the more complicated Roman calendar, taking force
in 45 BC (709 AUC). The calendar consisted of 3 years of 365 days followed by
one having 366 days. But years continued to be counted as regnal years.

In about AD 523, the monk Dionysius Exiguus (Denis the Little) devised a way
to implement rules set forth by the Nicean council (the so-called "Alexandrine
Time Scales, Epochs, and Intervals 3

Rules") for calculating Easter. In his calculations, he chose to number the years
since the birth of Christ, rather than the accession of the current monarch. He
(wrongly) fixed Jesus' birth at 25 December 753 AUC, thus making the Christian
era start with AD 1 on 1 January 754 AUC. How he established the year of
Christ's birth is not known, although a considerable number of theories exist. He
proposed this system of counting, but it was not immediately accepted.

When another monk, The Venerable Bede (AD 673-735), wrote his history of the
early centuries of Anglo-Saxon England, he promoted the system of Dionysius.
Its use spread during the Middle Ages until it became a de facto standard. Bede
himself seems to have instituted the "BC" and "AD" year naming convention. In
academic historical and archaeological circles, particularly in the United States,
the AD period is sometimes referred to as the Common Era (CE) and the BC
period as Before the Common Era (BCE). The AD-BC convention will be used
in this work.

While it is increasingly common to place AD after a date, by analogy to the use

of BC, formal English usage adheres to the traditional practice of placing the
abbreviation before the year, as in Latin (e.g., 100 BC, but AD 100).

The Julian calendar based on the AD 1 epoch was in common use until the AD
1500s, when countries started changing to the Gregorian calendar. Some
countries (for example, Greece and Russia) used it into this century, and the
Russian Orthodox Church still uses it, as do some other Orthodox churches.
However, the mean year in the Julian calendar was a little too long, causing the
Vernal equinox to slowly drift.

The Gregorian calendar is a modification of the Julian calendar that was first
proposed by the Neapolitan doctor Aloysius Lilius, and then authorized by Pope
Gregory XIII, for whom it was named, on February 24, 1582. The papal bull was
signed in AD 1581, but, for unknown reasons, was not printed until 1 March in
1582. It was devised to correct the equinoctial slippage and to bring calendar
dates back into alignment with equinoctial phenomena and to correct the method
by which leap years were calculated. The standard civil calendar in most
countries today is the Gregorian calendar.

Dates that occur prior to the adoption of a calendar system may still be reckoned
according to that system by prolepsis, or the anachronistic representation of
something as existing before its proper or historical time. Thus, it is possible to
4 Explanatory Supplement to Metric Prediction Generation

determine the actual year that was 46 BC, even though no calendar contained that
year.

2.3 Julian Date

Astronomers now commonly designate calendar dates by Julian Date (JD),
which is the interval of time in days and fraction of day since the epoch 4713 BC
January 1, Greenwich noon, according to the Julian proleptic calendar. The
convention was devised by John F. Herschel (son of astronomer William
Herschel) in 1849. He chose he beginning year as one in which the number in
each of three subordinate calendar cycles 1 was unity, and because it predated all
historical dates at that time.

Astronomers adopted Julian Dates in the late 19th century, but established the
meridian of Greenwich as the datum instead of Herschel’s use of Alexandria,
since the former had been made the Prime Meridian by international conference
in 1884.

Formulas for conversion between Gregorian dates and their corresponding Julian
Dates (for JD > 0) appear in the ESAA (12.92). Computer programs for these
conversions also appear in Numerical Recipes (1.1).

The modified Julian Date (MJD) is defined as the Julian Date minus 2400000.5.
Thus J2000 is MJD 51544.5.

The term “Julian date” is also used in many applications to refer to a date format
that combines the year and the number of days since the beginning of the year.
Depending on the usage, the year is represented by either 2 or 4 digits, and the
day of year by 3 digits. For example, January 1, 2007 is represented either as
2007001 or 07001. August 24,1999 is stored as 1999236 or 99236, since August
24 is the 236th day of the year.

This form is not based on the Julian calendar, nor is it the Julian Date discussed
above. To avoid confusion and ambiguity, the term is not further used in this

1
The cycles are indiction cycle, Metonic cycle, and solar cycle. The indiction cycle is a Roman tax
cycle of 15 years declared by Constantine the Great. The Metonic cycle is a particular approximate
common multiple of the tropical year and the synodic month, or 19 tropical years, which differs
from 235 synodic months only by about 2 hours. The solar cycle of 28 days is related to lunation
rate.
Time Scales, Epochs, and Intervals 5

work. Should it be necessary to refer to dates in this format, the term year-doy
number will be applied.

2.4 Time Scales

There are many time scales that are used in generating metric predictions for
deep space missions applications. These may be categorized as civil time, solar
time, proper time, and coordinate time. Each of these is related to the others, and
there are subtle subclassifications within its category. For example, civil time is
designated as Universal Time, or UT; but UT comes in a number of flavors, such
as UTC, UT0, UT1, and UT1R, discussed later in this Chapter. Figure 2-1
enumerates the various timescales covered in this Supplement, along with the
means of conversion among them.

2.5 International Atomic Time (TAI)

Proper time on Earth is reckoned in units of the SI (Système International)
second, defined as the duration of 9,192,631,770 cycles of radiation
corresponding to the transition between two hyperfine levels of the ground state
of cesium 133, as measured on the geoid (mean sea level). This definition,
though precise, is not totally accurate, as some variation among clocks
implementing this standard is inevitable.

International Atomic Time (TAI, from the French Temps Atomique

International), is a practical implementation of the standard that conforms as
closely as technology now permits to the definition of the SI second. It is
calculated as a weighed average of timescales obtained from a number of
individual commercial atomic time-standards and primary frequency standards in
many countries, as directed by the Bureau International des Poids et Measures in
Sèvres, France. Corrections are applied for known effects to maintain accuracy,
and the adjusted timescale is published as TAI.

As will be shown later in this chapter, velocity and gravitational potential at a

given location affect the rate of a clock’s oscillator. This fact causes clocks at the
equator to perform differently than the same clock at a pole, due to the
differences in rotational velocity and gravitational potential. For this reason,
comparisons of times from various locations require that a coordinate reference
frame and a set of comparison standards need to be established.
6 Explanatory Supplement to Metric Prediction Generation

In 1980 the Consultative Committee for the Definition of the Second (CCDS)
proposed to the International Committee of Weights and Measures (CIPM, in
French word order) that TAI be specifically defined as a coordinate timescale at a
geocentric datum line having as its unit the SI second, as obtained on the geoid in
rotation, and that TAI at other locations near the geoid be extended by applying
corrections for relativistic effects.

2.6 Dynamical Time (Ephemeris Time)

Dynamical time represents the independent variable of the equations of motion of
celestial bodies, spacecraft, and light rays in the solar system, measured in SI
seconds relative to an established epoch and inertial coordinate reference frame.
It is the time coordinate of ephemerides of such entities.

Two principal inertial reference frames are used by the MPG: the solar system
barycentric frame and the terrestrial geocentric frame. According to relativity
theory, there must exist mathematical transformations that correspond
phenomena that are observed in the two frames. The timescales of the two
frames, therefore, cannot both be unique. However, they may be chosen in such
a way that the timescales may differ only by periodic variations. This will be
addressed sections to follow.

2.7 Terrestrial Time

The dynamical timescale for apparent geocentric ephemerides was selected by
the IAU to be a unique proper timescale, from which others would be derived,
and termed Terrestrial Dynamic Time, or TDT. They later decided to drop the
“Dynamical” part of the name, to define the timescale as Terrestrial Time, or TT.

They decided that the TAI instant January 1.0, 1977 would be made equal to the
TT instant January 1.0003725, 1977. This introduced a difference of exactly
32.184 s between the two timescales. The unit of the timescale was the SI
second at mean sea level.

They also decided that a related barycentric timescale would be defined such that
no periodic variations between that time scale and TT would exist.
Time Scales, Epochs, and Intervals 7

TT is defined generally to be in step with TAI in order to take advantage of the

direct availability of Coordinated Universal Time, or UTC, which is also based
on the SI second measured on the geoid. But since TAI is a statistical timescale
based on average observed times and TT is an idealized uniform timescale, the
two are not totally consistent. The current definition is

TT  TAI  32.184 s
but may me altered in the future, if deemed appropriate.

The relationship between TT and TAI provides continuity with an existing

“ephemeris time” that was the independent variable in ephemeris generation in
that era. The chosen offset above is equal to the estimated difference between
TAI and ephemeris time at the time TT was introduced.

2.8 Barycentric Coordinate Time

Proper distance in any frame conforms to the equation

ds 2  gij dxi dx j

established in Chapter 1, where ds lies along the geodesic and dx i are

coordinate differentials along the path of travel. Substituting the g ij components
of the metric tensor for the n-body problem given in the previous chapter (also
see Moyer, 1971, 6), using a Cartesian coordinate system, inserting the
coordinate velocity, which is

 dx 2 dy 2 dz 2 
v2   2  2  2 
 dt dt dt 
and retaining only terms having order greater than 1/ c 2 , the following equation
results, relating proper time at some point on Earth to coordinate time in the
barycentric reference frame:

d 2 ds 2  2U   2U v
2

2
 2 2
 1  2   1  2  2
dt c dt  c   c c
8 Explanatory Supplement to Metric Prediction Generation

Here U is the sum of the Newtonian gravitational potentials of the ensemble of

masses at the body whose motion is being addressed. As is usual in relativistic
time transformations, proper time runs slower than coordinate time, as evidenced
by the right side of the equation being less than unity by an amount of order
(2U  v 2 ) / c 2 .
If proper time is being measured by a TT clock on the geoid, then the velocity
above is that of the clock with respect to the barycenter, and the significant
gravitational potential terms are those due to the Sun, Moon, and each of the
planets (including Earth). The barycentric timescale is called Barycentric
Coordinate Time, or TCB. Since TCB runs at a higher rate than TT, over
centuries the two drift farther and farther apart.

2.9 Barycentric Dynamic Time

Earlier it was mentioned that coordinate times could be made to differ from TT
by only periodic terms. Such a timescale would not then drift apart from Earth
clocks in the way that TCB does. It is defined by scaling spacetime coordinates
by a constant factor   1  L greater than unity chosen in such a way as to
remove the secular drift over time. The scale factor  does not affect the
equations of motion for bodies or light. However, it does alter the relationship
between the rate of an atomic clock that records proper time on Earth (which is
fixed at TT rate) and the rate of coordinate time in the barycentric frame.

The coordinate timescale defined in this fashion is termed Barycentric Dynamic

Time, or TDB, and is the independent variable in all ephemerides used by the
MPG. It is synonymous in this work with the term ephemeris time, except where
otherwise specifically noted. It differs from TT by at most by 1.7 ms.

TCB   TDB  (1  L ) TDB

The value required to render no long-term secular variation between TT and TDB
has been determined (see the Appendix to this chapter) to be
L  1.550505 108 .
The MPG uses two conversion routines provided by NAIF to translate between
TAI and TDB. These are HPTA2E (High Precision Time, Atomic to Ephemeris)
and HPTE2A (High Precision Time, Ephemeris to Atomic). It also uses two
routines that translate between UTC and TDB; these are HPTU2E (High
Time Scales, Epochs, and Intervals 9

Precision Time, UTC to Ephemeris) and HPTE2U (High Precision Time,

Ephemeris to UTC). Their usage is documented in the code commentary.

The time translation formula used in these utility functions was derived by JPL’s
Ted Moyer (see Moyer81). Moyer’s derivation of the TDB-TAI relationship is
lengthy and detailed, as it also analyzes the magnitudes of all elements of the
approximation, including terms which were then omitted from the final result as
being inconsequential, insofar as its use in the ODP was concerned. An
abbreviated version of that method appears in the Appendix to this chapter for
readers who may wish to have insight into the relativistic theory of the
transformation. Those with further interest may consult the reference, which is
now archived in electronic form. Those with less may skip the Appendix
altogether.

2.10 The TDB-TAI Equation

Moyer’s 1981 paper [Moyer81] derives two forms of the TDB-TAI formula.
One is a numeric expression involving the clock location and properties of
various mass centers (e.g., gravitational constant, velocity, and longitude of
Jupiter, Saturn, and the Earth-Moon barycenter). This formula appears in the
ESAA (2.222). The other is the vector form shown below.

The vector form of the solution gives TDB  TAI with considerably less
computation and slightly greater accuracy than does the expression given in the
ESAA.

The vector form of solution, used in the NAIF high-precision time routines, is

2 1 1
2  SB
TDB  TAI  TA  r  r SB   2  rBE  r B   2  rEA  r E  
c c c
1 1  J 
 2  rSB  r S   2    rSJ  r SJ  
c c  S   J 
1   Sa 
    rS Sa  r S Sa 
c2 
 S   Sa 
10 Explanatory Supplement to Metric Prediction Generation

in which TA  32.184 s , rx y denotes a vector from point x to point y , over-

dot denotes differentiation with respect to TDB, b  G M b is the gravitational
constant of body b , and A, E, B, S, J, and Sa refer, respectively, to the positions
of the atomic clock, the geocenter, the Earth-Moon barycenter, the Sun, Jupiter,
and Saturn.

This equation is the same for all Earth-fixed clocks, whether on the geoid or not,
as no restrictions are made as to clock location. However, evaluation of the
equation does require the clock’s location relative to the geocenter, and is
therefore does not yield the same result for all DSSs. This is a natural result,
whose first-order results are predicted by special relativity.

2.11 Evaluation of TDB Given TAI

Because the TDB  TAI formula is implicitly based on TDB , then given a
TDB value, the difference may be computed and TAI found via the relationship
TAI  TDB  (TDB  TAI )
However, when TAI is given, TDB must be found by iteration. The estimation
of TDBn by

TDBn  (TDB  TAI ) n 1  1  TAI

fortunately converges very rapidly. The first two terms of the TDB  TAI
formula are the most significant, and provide the means for a good first estimate
of TDB .

The convergence rate is illustrated by considering the difference between two

succeeding estimates,

TDBn  TDBn 1  (TDB  TAI ) n 1  (TDB  TAI ) n  2

d 
 (TDBn 1  TDBn  2 )
d TDB
The derivative factor has a value whose maximum magnitude is much smaller
than unity (on the order of 109 ). Each succeeding estimate is therefore much
closer to the correct value than the preceding one.

This iterative method is used in the NAIF HPTA2E utility.

Time Scales, Epochs, and Intervals 11

2.12 Universal Time (UT)

Prior to 1925, mean solar time was reckoned from noon in astronomical practice,
and designated Greenwich Mean Time, or GMT. Beginning in 1925, a 12h
discontinuity was introduced in order that mean solar time was reckoned from
midnight, rather than noon. The Nautical Almanac used GMT for the new
reckoning, whereas, the American Almanac used Greenwich Civil Time, or
GCT, for it. This confusion in terminology was finally removed by dropping
both designations and instituting Universal Time, or UT.

However, in the United Kingdom, GMT is sometimes still used. For civil
timekeeping, it means UTC, and for navigation, it refers to UT1. Thus, GMT has
two meanings that can differ by as much as 0.9s, and is not used in DSN
subsystems.

UT is a measure of time that closely approximates the mean diurnal motion of the
Sun. UTC, which is really an atomic time scale for civil timekeeping, is
discussed separately a little late in this Chapter. Several other forms of UT exist
and are used within the MPG, principally for determination of Earth’s attitude at
a given UTC instant.

The timescale designated as UT0 is determined directly from observations of the

diurnal motions of the stars; it is slightly dependent on the place of observation.
The timescale UT1 is obtained by correcting UT0 for the shift in longitude of an
observing station caused by polar motion. Whenever the designation UT is used
in this document, UT1 is implied. UT1 contains 41 short-period terms between 5
and 35 days that are caused by long-period solid Earth tides, designated as UT1.
When these are removed from UT1, the result is designated as regularized
universal time, or UTR.

UT1 is defined in such a way that it can be directly related to mean sidereal time
through a mathematical formula, appearing later in this Chapter. It thus does not
refer to the motion of Earth, nor is it precisely related to the Sun’s hour angle.

The apparent diurnal motion of the Sun involves both the nonuniform rotation of
Earth and the motion of Earth in its orbit around the Sun. UT1 was not based on
the hour angle of the Sun because such a system of time measurements could not
be related precisely to sidereal time, and could not be determined by observations
star transits and other such celestial measurements.
12 Explanatory Supplement to Metric Prediction Generation

UT1 is counted from 0 hours at midnight, with unit of duration the mean solar
day, defined to be as uniform as possible despite variations in the rotation of
Earth. It is continuous (no leap seconds), but has a somewhat variable rate due
Earth’s non-uniform rotational period.

The International Earth Rotation Service (IERS) tabulates the difference between
UT1 and UTC, UT 1  UT 1  UTC , as IERS Bulletin B, which is available via
the Internet. However, the MPG receives this data from the Kalman Earth
Orientation Filter (KEO) interface, along with other Earth Orientation Parameters
(EOP). UT1 is then calculated as

UT 1  UTC  UT 1

2.13 Coordinated Universal Time (UTC)

Coordinated Universal Time (UTC) was established on January 1, 1972 at 0h as a
measure of time that now serves as the basis for almost all civil timekeeping. It is
commonly accessed within the United States via broadcasts by the NIST radio
station WWV and other services.

As discussed above, it is related to Universal Time (UT), which is mean solar

time, but UTC is really an atomic time. UTC uses the SI second as it
fundamental unit and is adjusted infrequently in order to maintain the transit of
the prime meridian at a time near noon. Since Earth’s rotation is not uniform with
respect to atomic time, a leap second is added or subtracted as necessary to
prevent the error between UTC and mean solar time from exceeding 0.9s. Leap
seconds are typically added at the end of December or June, but they can also
appear (added or subtracted) at other designated times throughout the year.

UTC  TAI  ( number of leap seconds )

Leap second adjustments affect the number of seconds per day and thus the
number of seconds per year. The number of leap seconds incorporated into the
January 1, 1972 initial UTC epoch was 10. Subsequently, the UTC timescale has
marched backward relative to the TAI timescale exactly one second on each of a
number of scheduled occasions.

Precisely when leap seconds occur, and in which direction the correction is made,
is currently the responsibility of the IERS, which publishes periodic bulletins
(Bulletin C) available via the Internet. As specified in CCIR Report 517, a leap
Time Scales, Epochs, and Intervals 13

second is inserted following second 23:59:59 on the last day of June or

December and becomes second 23:59:60 of that day. A leap second would be
deleted by omitting second 23:59:59 on one of these days, although this has not
yet happened, as of this writing.

On July 4, 2005, the EIRS announced the introduction of the 33 rd leap second on
December 31, 2005. The 32nd had been introduced on December 31, 1998.

The MPG converts UTC into TDB, and vice versa, through use of the NAIF
special routines HPTU2E and HPTE2U, which require the presence of a file
containing a time-tagged list of leap seconds, or, in SPICE parlance, a leap
seconds kernel.

Earth is divided in to standard-time zones, and local times differ from UTC by an
integral number of hours according to the particular time zone. Parts of Canada
and Australia differ by integer-plus-half hours.

UTC times are referenced to the Zero meridian (Greenwich, England), which is
often designated by a “Z” affixed to the time format, as 12:45:03Z. Z was thus
designated as the international time zone for the prime meridian. It is sometimes
thus referred to phonetically as “Zulu” time. The U.S. local time zones are
Eastern ["R", "Romeo]; Central ["S", "Sierra"]; Mountain ["T", "Tango"]; Pacific
["U", "Uniform"]; Alaska ["V", "Victor"], and Hawaii ["W", "Whiskey"].

2.14 Sidereal Time (ST)

True sidereal time measures the Greenwich hour angle of the true equinox of
date, measured westward from the true prime meridian of date (0=) about the true
pole of date to the true vernal equinox of date. As such, then, sidereal time is a
direct measure of Earth’s diurnal rotation. The MPG ordinarily uses the
Precision Earth Model (PEM) for determining DSS locations and SPICE utilities,
such as SPKEZ, for spacecraft positions. However, it does require sidereal time
to calculate the directions of radio sources. For this reason, a discussion is
included here.

The period of time between two consecutive upper meridian transits of the
equinox is a sidereal day. Since the rotation of Earth is subject to irregular
forces, sidereal time is irregular with respect to atomic time. The practical
determination of sidereal time from observations of radio sources and other
14 Explanatory Supplement to Metric Prediction Generation

means requires inclusion of Earth’s precession, nutation, and polar motion

effects.

2.15 Greenwich Mean Sidereal Time

.Mean sidereal time is referenced to the mean equinox of date, which is perturbed
only by precession. Greenwich Mean Sidereal Time (GMST) is directly related
to UT1 by a numerical formula. Since January 1, 1984, Greenwich Mean
Sidereal Time (GMST), measured in seconds, at 0h UT1 on day d , measured in
days past JD2451545.0, is given by the following formula appearing in the
ESAA,

GMST0s  24110.54841  236.5553679087 d 

 6.9789  10-11 d 2  1.27  10-19 d 3
Note in this expression that Julian Dates begin at noon, so a given d at 0h has a
fractional part of 0.5. Since the sidereal year is a full day longer than a solar year,
each sidereal day is about 86400 / 365.25  236.55s longer than a UT day.
Translated into units of degrees, this formula is

GMSToo  100.46061838  0.985647366286 d 

2.9079 1013 d 2  5.3 1022 d 3
That is, the Greenwich hour angle of the equinox advances about one degree per
day. The conversion of this expression into a general expression for mean
sidereal time at any UT1 instant was done by Moyer (see Moyer2000, Eq. 5-
173). His result, with d now including days plus fractional days past J2000, is

GMST s  67310.54841  24.06570982442 d

6.9789 1011 d 2  1.27  1019 d 3
Expressed in angular measure, this is

GMST   280.460618375  360.9856473663 d 

2.9079  1013 d 2  5.3  10 22 d 3
Time Scales, Epochs, and Intervals 15

2.16 Greenwich Apparent Sidereal Time

Apparent sidereal time is referenced to the true equinox (the intersection of the
true equator of date with the ecliptic of date), and thus includes the effects of
both precession and nutation. Greenwich Apparent Sidereal Time (GAST) is
related to GMST via the equation of the equinoxes (before 1960, this effect was
called “nutation in right ascension”). Nutation describes the motion of the true
pole relative to the mean pole and may be resolved into components in longitude
and obliquity. The longitude component, between the mean equinox and the true
equinox of date, is denoted  , and the angle between the mean ecliptic of date
and the true equator of date is the true obliquity, and denoted  . Both of these
quantities are available via IERS bulletins, and to the MPG via the KEO Filter
interface.

The difference between the true and mean right ascensions of a position on the
true equator is the difference between apparent and mean sidereal time,

GAST  GMST  EE
EE   cos   CT
where EE denotes the equation of the equinoxes and CT includes the
“complementary terms” that were added to the “textbook” form of the equation
of the equinoxes by IAU Resolution C7, Recommendation 3 (1994). The new
formulation takes into account cross-terms between the various precession and
nutation quantities, amounting to about 3 milliarcsec (83 microdeg). The
transition from the old to the new model officially took place on February 27,
1997. These terms were added to compensate for irregularities in the UT1
timescale traceable to side effects of nutation and precession. By convention, the
complementary terms were included in the equation of the equinoxes, rather than
as apart of mean sidereal time.

2.17 Approximate Equation of the Equinox

The U. S. Naval Observatory (USNO) publishes approximate formulas for
sidereal time, which may be accessed via the internet. Their 1978 formulas for
GMST and EE, based on an epoch of 00:00:00 January 1, 1900, were
incorporated into the NSS MP. An updated set is available that uses the J2000
16 Explanatory Supplement to Metric Prediction Generation

epoch. The expression for GMST in this set agrees with the one given above; the
approximation for EE is

EE     cos 
   0.004785sin   0.00036sin 2 L
  125.04  0.052954d
L  280.47  0.98565d
   23.4393  0.0000004d
In this formulation,  is the longitude of the ascending node of the Moon’s orbit
and L is the mean longitude of the Sun. All expressions are given in degrees.

The cited maximum error resulting from the use of these formulas over the
period 2000-2100 is 0.432 seconds, with an rms error of 0.15 seconds. At
Earth’s diurnal rate, the maximum error amounts to about 1.8 mdeg. Since this
exceeds the MPG pointing error specification, it is unusable.

2.18 Local Apparent Sidereal Time

Local Apparent Sidereal Time (LAST) is the apparent right ascension of the local
meridian. It can be obtained from the GAST by adding the meridian’s east
longitude.

2.19 Solar Time

Although solar time is not used in the MPG, it nonetheless has historical
importance that mandates its inclusion in this Explanatory Supplement.
Apparent Solar Time is the oldest timescale of all, being the measure of time
defined by the actual diurnal motion of the Sun. It has been known since
antiquity that the Sun’s motion is not uniform, a fact that became well established
after the invention of clocks, which tended to measure uniform intervals of time.
Apparent Solar Time was the argument in The National Almanac and other
national ephemerides until the early nineteenth century. Mean Solar Time was
defined by the motion of an imaginary fiducial body (the fictitious mean Sun)
that moved uniformly in the equatorial plane at a rate virtually equal to the mean
rate of the true Sun’s motion in the ecliptic. As clocks improved and came into
extensive use at sea in the late eighteenth and early nineteenth centuries, apparent
Time Scales, Epochs, and Intervals 17

time was eventually superseded in civil use by Mean Solar Time. In the mid-
nineteenth century, it became the argument in the national ephemerides.

At first, astronomers began counting time in hours past noon, so the Sun’s
meridian distance was indeed the time of day. When the time of day began to be
reckoned at midnight, it was necessary to add 12 hours to time to determine the
location of the fictitious mean Sun at that time.

The difference between apparent and mean solar times is called the equation of
time. The principles for determining the equation of time extend back at least to
the time of Ptolemy. At first, the equation of time was a way of determining
mean solar time from apparent solar time. As clocks improved, it became a way
of calculating apparent solar time from clock time, which kept, to the accuracy
available at the time, mean solar time. These clocks were regulated by
observations of sidereal time.

The underlying concept of mean solar time was that the rotation of Earth is
uniform. However, in the first half of the twentieth century it became obvious
that this assumption could no longer be deemed acceptable. To replace it, two
new timescales came into being. Ephemeris Time (ET) was introduced to satisfy
the desire for a uniform measure of time that would be the independent variable
in the mathematical computation of ephemerides, and Universal Time (UT) came
to designate a measure of Earth’s rotation, as discussed in previous Sections of
this Supplement.

The irregularities in apparent solar time are principally due to two effects. First,
the motion of Earth in the ecliptic plane is not uniform, but is almost elliptical.
The difference between the true anomaly and mean anomaly of Earth’s orbit is an
angle through which Earth must turn in order for the Sun to transit the local
meridian. The time required for Earth to rotate by this angle is therefore one
component of the equation of time.

The second effect is due to the inclination of Earth’s pole to the ecliptic, which
makes the Sun, as viewed at Earth, traverse above and below the equatorial
plane. The projection of the Sun’s path on the equatorial plane thus causes
further distortions of the apparent motion.

Both effects are annual phenomena, and tend to produce a characteristic cyclic
variation over time. A formula for the equation of time appears in the ESAA
18 Explanatory Supplement to Metric Prediction Generation

(9.311). However, the algorithm given below agrees with this formula within a
few seconds, and is more intuitively understood.

The procedure is to compute the true anomaly in the ecliptic, project this
onto the equatorial plane, and subtract the mean motion in the equatorial
plane, as follows: Calculate Earth’s mean anomaly and obliquity, in
radians (reduced, as necessary, to be in 0  2 ), using
M  6.2400407681  0.0172019698861 d
  0.409092959363  6.36892077921109 d
where d is the number of UT days past the J2000 epoch. Next, invert Kepler’s
equation2 to find the eccentric anomaly E

M  E  e sin E
Then use Earth’s orbital eccentricity e  0.01671 to compute the current
longitude (true anomaly)

  tan 1  1  e 2 sin E /(cos E  e) 

Use the two-argument arc-tangent function, if available, to retain proper quadrant
information.

Rotate the ellipse in the ecliptic clockwise about the z axis by the adjusted true
anomaly,    , using a value for perihelion elongation from winter solstice of
  0.223796326795 ,
x1  cos(  )
y1  sin(  )
z1  0
Rotate the ecliptic plane counterclockwise about the y axis by the obliquity

x 2  cos  x1  sin  z1
y 2  y1
z 2   sin  x1  cos  z1
2
Iteration of Kepler’s equation converges rapidly, with E0  M , Ei  M  e sin Ei 1 , until
4
| Ei 1  Ei |  10 . This provides subsecond accuracy of conversion.
Time Scales, Epochs, and Intervals 19

Finally, rotate the equatorial plane counterclockwise about the z axis by the
adjusted mean anomaly,

x3  cos( M  ) x 2  sin( M  ) y 2
y3   sin( M  ) x 2  cos( M  ) y 2
z3  z 2
The equation of time, in minutes, and declination of the Sun, in degrees, are

720
EOT   tan 1 ( y3 / x3)

180 1
 sin ( z 3)

The leading fractions are conversion factors for the cited units.

The resulting behavior plotted over time is shown in Figure 2-2. A plot of the
equation of time versus declination, known as an analemma, is shown in Figure
2-3.
20 Explanatory Supplement to Metric Prediction Generation

2.20 Time Representations and Standard Formats

Epochs and intervals have different representations to humans than they have in
machine form. As mentioned earlier, SPICE and the MPG interpret times
internally in the form of seconds past the J2000 epoch. Human readable contexts
are usually expressed as strings of characters.

However, the variety of ways people have developed for expressing times in
string formats is enormous. It is unlikely that any single software package can
accommodate all of the custom time formats that have arisen in various
computing contexts. However, SPICE and NAIF utilities correctly interpret most
time formats used throughout the planetary science community. It supports ISO,
UNIX, VMS, and MS-DOS formats, with epochs in both AD and BC eras, and
with time zone specifications. These utilities transform from string to machine
form and from machine form into strings. Consult the SPICE required reading
document TIME.REQ for complete details on formats and translations.

Some of the more frequently human-readable formats are listed below. The ISO
formats specify strict forms with required terms and exact punctuation. The
SPICE and NAIF routines are much more general. In ISO formats a “T” is
required to indicate the beginning of a time specification.

Year YYYY

Year and Month YYYY-MM

Complete Date YYYY-MM-DD

Complete Date and Time YYYY-MM-DDThh:mm:ss.ffTZD

Year and Day of Year YYYY-DOY

Year and Day of Year, plus YYYY-DOY DDThh:mm:ss.ffTZD

Time

Times are generally assumed to be expressed in UTC, with a special UTC

designator ("Z"). Local times in ISO formats require offset designators.
Time Scales, Epochs, and Intervals 21

In this table, the representations are

YYYY four-digit year

MM two-digit month (01 is January, etc.)

hh two-digit hour (00 through 23). No “am/pm” in ISO.

mm two-digit minute (00 through 59)

ss two-digit seconds (00 through 59)

f one or more digits representing decimal fractions of a second

TZD time zone designator (e.g., Z) or offset

(+hh:mm or –hh:mm)

SPICE utilities recognize a wider range of more user-friendly formats that are not
as strict as the ISO forms. For example, they also recognize Julian Dates, “am”
or “pm”, and U.S. local time zone designations (e.g., PDT), as well as timescale
designations (e.g., UTC, TAI, TDB, TDT).

2.21 Spacecraft Atomic Time

Given the spacecraft identifier and the ephemeris time of interest, the NAIF-
supplied function HPTDAT produces the derivative of Spacecraft Atomic Time
(TAS3) with respect to ephemeris time in the form Dt = t&(t ) - 1 , where t is a
TDB time and t (t ) is its equivalent atomic time aboard the given spacecraft.

At one stage of its development, the MPG computed spacecraft atomic time in
order to compute one-way light time values. It therefore generated a polynomial
profile that approximated D t (t ) = t (t ) - t within a prespecified error value.
This data type was deemed unneeded for a number of reasons, and was later
omitted from the MPG design.

The reasons for not using the estimated TAS were that (1) one-way light time
data products are used differentially, so that TAS disappears in the differences,
and (2) the TAS profile contained a constant of integration that could not be
3
The word order here matches that of TDB and TAI, even though it does not represent a French
word order. “Spacecraft” in French would become “Astronef” or some other suitable term.
22 Explanatory Supplement to Metric Prediction Generation

evaluated. Nevertheless, the means that were used in the earlier design are
included here for historical purposes.

The MPG produced the TAS polynomial profile by segmenting the time interval
of interest (normally a pass) into intervals (t i , t i + 1 ) , sampling Dt at five optimally
spaced points over each interval, and integrating the results to form a profile of
polynomials of degree 5. Each polynomial in the profile was expressed in the
standard MPG Everett form. The samples values extracted were

y = ( t i , Dt (t i ), Dt (t i + 33h / 181), Dt (t i + h / 2), Dt (t i + 148h / 181), Dt (t i + 1 ))T

Here h is the interval length and t i is time at the beginning of the ith interval. The
method of generating the polynomial from sample values is treated in the chapter
on Interpolation in this Supplement.

In this formulation, the value t i was recognized as a constant of integration, to

be evaluated. Given this value at the beginning of an interval, the subsequent
endpoint value t i + 1 will be determined by the interval length and linear
combinations of sample values. To make the profile continuous at interval
boundaries, the beginning value of the next interval becomes the final value of
the previous one throughout the profile.

Only the t 0 = t (t 0 ) of the initial interval thus remains unspecified. This initial
offset between spacecraft atomic time and ephemeris time is unknown, and was
therefore set to zero in the earlier MPG development. Since users of predictions
involving TAS were generally insensitive to the true difference, this choice is of
no known consequence.
Time Scales, Epochs, and Intervals 23

Appendix A
Derivation of TDB-TAI

This Appendix presents an overview of the method found in Moyer’s [Moyer81]

derivation of the relationship between TDB and TAI. Computation of the scaling
factor L is mentioned in the reference, but not directly computed. The
approximate computation of the scaling factor that appears below is due to the
author.

Computing the Scaling Factor

If the scaled time (TDB) is denoted by T , where

t  T
then the relationship between proper time and TDB is

d 2 ds 2  2U   2U v 
2

2
 2 2
 2 1  2   1  2  2
dT c dT  c   c c 
in which v is the clock’s velocity relative to the solar system barycenter,
translated into the barycentric frame.

If it is desired that no periodic terms in this equation appear,  can be set to the
value
1/ 2
 2U   2U  v 2 
   1  2    1  2  2 
 c   c c 

in which the operation removes short-term variations. A one-term Taylor

expansion of the right hand side above in 1/ c 2 yields the evaluation
24 Explanatory Supplement to Metric Prediction Generation

U  v2 / 2
 1 
c2
It is convenient to define L as the small departure from unity in this factor,

U  v2 / 2
L
c2
It remains only to evaluate the secular parts of U and v 2 .

Approximating the Scaling Factor

An approximation method is useful to illustrate the method actually used.
Suppose that the clock path is represented as a circular orbit at distance a about
a given center, and let the path of a given gravitational influence be represented
also by a circular orbit at distance b about the same center. Let d  max(a, b)
and   min(a, b) / d . Then the distance between the gravity source and clock
may be computed using vector algebra, which will result in an algebraic function
involving trigonometric variations over time. Using an algebraic computational
tool, such as Mathematica, the inverse distance function may be expanded in a
Taylor series in  to determine the constant and periodic terms. The secular
portion of the Newtonian potential for that gravity source is then found to be

   2 9 4 
Ui  1    
d 4 64 
where   G M is the gravitational constant of the source. This expression may
be applied to the Sun ( a  AU , b  0 ), Moon ( a  rEarth , b  dist Moon ), Earth (
a  rEarth , b  0 ), and each of the planets ( a  AU , b  dist planet ), to provide the
approximate contributions of these bodies to the scale factor.

Similarly, the velocity may be represented a constant vector whose magnitude is

the mean geocentric velocity about the barycenter added to a rotating vector
whose magnitude is the mean diurnal clock velocity. The result is
2 2
 2 AU   2 rEarth 
v 
2
  
 365.25 days   1 day 
Time Scales, Epochs, and Intervals 25

The composite result is that L  1.55059  108 . The value published in the
ESAA is L  1.550505 108 , and that used in the NSS MP was
L  1.5505204 108 . The approximation has omitted eccentricities in orbits
and the effects of considering the Earth-Moon motion about their barycenter,
which are included in the actual computation. Nonetheless, the agreement is
pretty good.

The rate difference between TDB and TCB thus turns out to be about 48.93
s/century. The rate difference between TDB computations using the ESAA rate
and the one approximated above is only about 3 ms/century, and using the ESAA
and NSS MP rates is only about 0.5 ms/century.

Computing the Periodic Part

In order to compute TDB, it is first necessary to determine the differential
equation relating the rates of the two timescales. This may be done by square-
rooting the metric equation. Then the right-hand side may be expanded in a
Taylor series in 1/ c 2 , with terms of order higher than 1/ c 2 discarded. The
result is

d U  v2 / 2
 1 L
dT c2
It then remains only to integrate this differential equation. In practice, TDB ( T
above) is determined from TT (  above) by means of appropriate mathematical
approximations. Secular terms disappear due to the presence of the L term. The
desired form of solution is the time difference

U  v 2 / 2 
T
T    T0     dT
T0 
c2 
where  Q  Q  L retains only the periodic parts of Q .

In 1981, JPL’s Ted Moyer (Moyer 81) published two forms of solution. One had
appeared in a JPL Technical Report several years earlier, in 1971. This is
essentially the one that appears in the ESAA, and is the one that is used by the
NSS MP. The 1981 article also developed a vector form of the solution, which is
the one used in the MPG. Moyer’s article cites others who have derived similar
formulas. In fact, several formulas appear have greater accuracies (errors as low
26 Explanatory Supplement to Metric Prediction Generation

as 1 ns, but nominally –131 to 64 ns) than the Moyer forms, but they all contain
considerably more (127 to 1637) terms. The Moyer solutions are sufficiently
accurate (within about 4 s) for MPG applications, and the vector form is the one
used in the MPG and ODP.

Moyer’s derivation of the TDB-TT formula is lengthy and detailed, as it also

analyzes the magnitudes of all elements of the approximation, including terms
which were then omitted from the final result as being inconsequential, insofar as
its use in the ODP was concerned. An abbreviated version of that method will be
presented here for readers who may wish to have insight into the relativistic
theory of the transformation. Those with further interest may consult the
reference, which is now archived in electronic form. Those with less may skip
the remainder of this subsection.

Since only periodic terms remain in the solution, only these will be retained in
the analysis as it progresses. The constant terms, of course, accumulate into the
L parameter, and provide its more accurate determination.
The following notation will be used to designate vector positions and velocities:
r denotes the position of the entity  with respect to the solar system
barycenter, and r denotes its derivative with respect to coordinate time
(velocity); r denotes the vector r  r and r is its time derivative. Entities
of interest are the Sun (S), Moon (M), Earth (E), the Earth-Moon barycenter (B),
Jupiter (J), Saturn (Sa), and the atomic clock (A) that is recording TT. The other
planets are indexed by their numerical order outward from the Sun.

The integrand contains the quantity

U A  v A2 / 2
which is the sum of the potential and kinetic energies per unit mass at the clock
location. The first step is to express each of these terms in geocentric terms,

U A  (U A U E )  U E  U E  U E  rEA  
v A2 / 2  r A  r A / 2  (r EA  r E )  (r EA  r E ) / 2  vEA
2
/ 2  r EA  r A  vE2 / 2
The potential U E is evaluated at the geocenter, and thus excludes Earth’s
gravitational effects.
Time Scales, Epochs, and Intervals 27

2
In the second of these equations, vEA is the square of the geocentric velocity of
the clock, which is constant if the effects of solid earth tides, polar motion, and
nutation are ignored. It therefore may be dropped from further consideration. In
the first equation, the remaining terms in the Taylor expansion of the potential
are omitted as inconsequential. Further, Newton’s law can be used to estimate
the potential gradient term within the required order of 1/ c 2 terms

U E  rE
so that

d (r E  rEA )
U E  rEA  rE  rEA   r E  r EA
dT
Notice that the negative term in this equation cancels a term in the earlier
velocity equation. The differential time equation time periodic terms are now

d (T   ) d  U E  vE2 / 2 d (r E  rEA )

  
dT dT c2 dT
Moreover, the derivative term on the right-hand side can be integrated directly,
leaving only the first term to be further manipulated.

The same kind of action can be applied to the variation about the Earth-Moon
barycenter, in which U E  (U E  U B )  U B and rE  rEB  rB . The U B term is
evaluated at the Earth-Moon barycenter, and excludes terms due to Earth and
Moon. The result is

d  U B  vB2 / 2 d (r E  rEA ) d (r B  rBE )

  
dT c2 dT dT
Again, only the first term on the right requires further manipulation.

The remaining potential term,

U B  U B, S  U B,i
i

is the sum of the gravitational potential at B due to S and due to each of the
other 8 planets. However, only the contributions due to Jupiter and Saturn were
found to be significant, so only these terms are assumed in what follows.
28 Explanatory Supplement to Metric Prediction Generation

S
U B, S 
rSB
i i   3 
U B,i    i 1  i cos(i )  i3 cos(2i ) 
rBi | rSB  rSi | rS i  4 
The quantity i is the gravitational constant of the planet i , or, when it
references the Earth-Moon barycenter, to the combined gravitational constants of
Earth and Moon. The i parameter is the same as that defined earlier, the
periodic terms of the earlier Taylor expansion are now retained, and i is the
longitudinal angle between i and B as viewed at S . Moyer’s work determined
that only the first two terms in the expansion were found to be needed for the
accuracy required.

 i i
U Bi   rSB cos(i )
rS i rS2i

As for the velocity term, since rB  rS  rSB , vB becomes

2

vB2  vS2  vSB

2
 2r S  r SB  vSB
2
 2r S  r SB
 d r  r 
 vSB
2
 2  S SB  rS  rSB 
 dT 
The approximation assumes that the squared velocity of the Sun about the solar
system barycenter has negligible periodic terms. As earlier, the acceleration term
may be replaced by its Newtonian potential within the accuracy limits imposed,

i
rS   rS i
i rS3i

The sum extends over the Earth-Moon barycenter and retained planets. The dot
product rS  rSB is then

i 
rS  rSB   3
rS i  rSB    2i rSB cos(i )
i rS i i rS i

These terms cancel with corresponding terms in the expansion of U B . The

resulting time differential equation is now
Time Scales, Epochs, and Intervals 29

d  1   S vSB
2
  d r  r d (r E  rEA ) d (r B  rBE ) 
 2   J  Sa  S SB   
dT c  rSB 2 rSJ rSSa dT dT dT 
The remaining steps presume that the Earth-Moon barycenter, Jupiter, and Saturn
are in heliocentric elliptical orbits. The squared-velocity term is given by the vis-
viva equation,

 2 1 
vB2  (  S   E   M )   
 rSB aSB 
in which aSB is the ellipse semimajor axis. The Earth-Moon barycentric terms
are then

 S vSB
2
2   E   M  S   E   M
  S 
rSB 2 rSB 2 aSB
 2 1 
 S   
 rSB 2 aSB 
in which the final terms contain no periodic elements, and may thus be dropped
from consideration.

Also for an elliptic orbit, the inverse of the radial distance is a function of the
eccentric anomaly E (not to be confused with the E used elsewhere to denote
the position of Earth) and the ellipse eccentricity e , given by

1 1 e
  cos( E )
rSB aSB rSB
Only the latter term contains periodic elements, so the first may be omitted.

Further, the derivative of the eccentric anomaly with respect to coordinate time is

1 S   E  M 1 S
E  
rSB a rSB a
The ratio of the left-hand side of this equation to the right-hand side is unity.
Multiplication of the expression for the periodic elements of the Earth-Moon
barycentric terms by this form of unity gives
30 Explanatory Supplement to Metric Prediction Generation

  S vSB
2
 2 S e  aSB 
    rSB  cos( E ) E
 rSB 2  rSB  S 
 2e  S aSB cos( E ) E


d
dT

2e  S aSB sin( E ) 
Another property of elliptic orbits is that

rSB  r SB  e S aSB sin( E )

As a consequence, then, an integrable form is obtained,

  S vSB
2
 d
    2 rSB  r SB 
 rSB 2  dT
A similar treatment of Jupiter and Saturn contributions provides

 i  d  i 
   ri  ri 
 rS i  dT  S  i 

The TDB-TAI Equation

Since all terms in the time differential equation are now integrable, the final
result is at hand. Integration adds a constant value T0 to the integrated periodic
terms, which may be chosen to make T   have an agreed-upon value at a
given epoch. Since TT and TAI are defined so as to differ only by a constant, the
integrated equation can also express the difference TDB  TAI by appropriate
choice of the integration constant.

The 16th General Assembly of the IAU adopted the value TA  32.184 s. The
final expression for the time differential is
Time Scales, Epochs, and Intervals 31

2 1 1
2  SB
TDB  TAI  TA  r  r SB   2  rBE  r B   2  rEA  r E  
c c c
1 1  J 
 2  rSB  r S   2    rSJ  r SJ  
c c  S   J 
1   Sa 
    rS Sa  r S Sa 
c2   S   Sa 
Moreover, this equation is the same for all Earth-fixed clocks, whether on the
geoid or not, as no restrictions were made as to clock location.

This vector form of the solution gives TDB  TAI with considerably less
computation and slightly greater accuracy than does the expression given in the
ESAA, which is a function of time and the topocentric coordinates of the atomic
clock.
32 Explanatory Supplement to Metric Prediction Generation

References
[Moyer1971] Moyer, T. D. (1971), “Mathematical Formulation of the Double
Precision Orbit Determination Program,”, Technical Report 32-1527, Jet
Propulsion Laboratory, California Institute of Technology, Pasadena, CA, May
15, 1971.

[Moyer2000] Moyer, T. D. (2000), Formulation for Observed and Computed

Values of Deep Space Network Data Types for Navigation, JPL Publication 00-7,
Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA,
October 2000.

[Seidelmann1992] Seidelmann, K. P. (ed) (1992), Explanatory Supplement to the

Astronomical Almanac, University Science Books, Mill Valley, CA,1992.

[NAIF1997] TIME.REQ (1997), SPICE Toolkit, Navigation and Ancillary

Information Facility, Jet Propulsion Laboratory, Pasadena, CA, July 1997.

[Taus1998-1] Tausworthe, R. C., “Translation of Polynomials to and from

Everett Form,” Mathematica study Everett-Horner.nb, MPG Archive,
NPP Release Directory, January, 1998.

[Taus1998] Tausworthe, R. C., and Walther, J. Y., “Estimation of Spacecraft

Atomic Time from its Derivative Using the Error Ridge Method,” Mathematica
study TAS-Ridge-Errors.nb, MPG Archive, NPP Release Directory, July,
1998.