judah levine, nist, cenam, sept 2012: 1 introduction to time and timekeeping judah levine time and...
TRANSCRIPT
Judah Levine, NIST, CENAM, Sept 2012: 1
Introduction to Time and Introduction to Time and TimekeepingTimekeeping
Judah LevineJudah Levine
Time and Frequency DivisionTime and Frequency Division
NIST/BoulderNIST/[email protected]
Judah Levine, NIST, CENAM, Sept 2012: 2
OutlineOutline
Introductory background Introductory background Requirements of a time service Requirements of a time service
operated by a timing laboratoryoperated by a timing laboratory The error budget for time The error budget for time
disseminationdissemination Description of methods with Description of methods with
examples examples – advantages and limitationsadvantages and limitations
Judah Levine, NIST, CENAM, Sept 2012: 3
Background
Laboratory is generating a local estimate of UTC: UTC(lab)– Time scale methods in future talk
UTC(lab) is steered to UTC with an uncertainty adequate for the time service 100 s for NTP, telephone services,
radio time signals, …
Judah Levine, NIST, CENAM, Sept 2012: 4
Requirements - 1
Integrity– Time signals must be protected so
they are not modified or changed during transmission
– Easy:• Telephone service• Radio broadcast service• Authenticated Internet service
– Hard:• Normal Internet services
Judah Levine, NIST, CENAM, Sept 2012: 5
Requirements - 2
Availability– Service should not have single point
of failure• Multiple sources at different locations
– Minimize Time to Repair– Balanced with Cost– GNSS signals can be jammed or
spoofed• Total reliance on GNSS signals possible
long-term problem
Judah Levine, NIST, CENAM, Sept 2012: 6
Requirements - 3
Accuracy– Service should transmit UTC(lab) only
when operating correctly– Should transmit nothing or error
message when failed• Some services do not use this principle
– Example: implementations of NTP
• Principle can never be 100% reliable
Judah Levine, NIST, CENAM, Sept 2012: 7
Requirements - 4 Technical Traceability
– Each link between user and UTC should be calibrated with delay and uncertainty
• Magnitude consistent with user requirements Legal Traceability
– Traceability can be documented and proven in legal proceedings
• Log files and documents show proper operation and also errors
Users are responsible for traceability with assistance from timing laboratory
Judah Levine, NIST, CENAM, Sept 2012: 8
The Error BudgetThe Error Budget
Internal accuracy of the time sourceInternal accuracy of the time source– Usually not the limiting factorUsually not the limiting factor
The transmission delayThe transmission delay– This is usually the hard partThis is usually the hard part– Uncertainty often limits traceabilityUncertainty often limits traceability
Statistics of the user’s clock and the Statistics of the user’s clock and the measurement processmeasurement process– Is calibration interval consistent with Is calibration interval consistent with
accuracy requirement?accuracy requirement?– Dynamic, adaptive calibration processDynamic, adaptive calibration process
Judah Levine, NIST, CENAM, Sept 2012: 9
Methods of Time Dissemination Methods of Time Dissemination
Simple one-way methodSimple one-way method One-way method with model of One-way method with model of
delaydelay Common-viewCommon-view Partial two-way methodPartial two-way method Full Two-way methodFull Two-way method
Judah Levine, NIST, CENAM, Sept 2012: 10
Simple one-way method - 1Simple one-way method - 1
Ignore network delay completelyIgnore network delay completely– Delay << required accuracyDelay << required accuracy
Simple broadcastsSimple broadcasts– Low-frequency services (WWVB, …)Low-frequency services (WWVB, …)
• 60 kHz, 2 × 50 kW covers most of US60 kHz, 2 × 50 kW covers most of US
– Short-wave services (WWV, …)Short-wave services (WWV, …)• 2.5 MHz, 5 MHz, … delay, coverage variable2.5 MHz, 5 MHz, … delay, coverage variable
– Internet service in broadcast mode Internet service in broadcast mode (NTP)(NTP)• Delay, coverage very variableDelay, coverage very variable
Judah Levine, NIST, CENAM, Sept 2012: 11
Simple one-way method - 2Simple one-way method - 2
Simple receiver and transmitterSimple receiver and transmitter Transmission cost does not Transmission cost does not
depend on number of receiversdepend on number of receivers Receiver is passiveReceiver is passive Timing error < 1 s, often < 20 msTiming error < 1 s, often < 20 ms Traceability possible with Traceability possible with adequateadequate log files log files
Judah Levine, NIST, CENAM, Sept 2012: 12
One-way with delay model
GPS
Receiver
Geometric delay, 65 ms estimated using ephemerisand known position
65 ns, Ionosphere delay from model or L1-L2 dispersion
5 ns,Troposphere delay from T, RH, or multiple satellites
Geophysical effects,earth models, 1ns Calibration,
Multipath, 10ns
Judah Levine, NIST, CENAM, Sept 2012: 13
Common-view methodCommon-view method
Source
Rcvr 1 Rcvr 2
T1= t(1) – (S +)T2= t(2) – (S + )
t= T1-T2= t(1)-t(2)
Path delaysare nearly equaland cancel in thedifference
Source clockcancels too
The time is S
←δ→
Judah Levine, NIST, CENAM, Sept 2012: 14
Common View Sources
GNSS Signals Television Broadcasts
– Synchronization pulse in blank line FM radio signals
– Stereo sub-carrier Phase of mains voltage
– Within building or small area Loran signals (no longer in US) Source is used passively at no cost
Judah Levine, NIST, CENAM, Sept 2012: 15
Common View Limitations
Paths to receivers have very different un-modeled delays– Calibration of local equipment– Atmospheric delay
Receivers too far apart to see the physical transmitter
Judah Levine, NIST, CENAM, Sept 2012: 16
All in view melting pot
S1 S2 S3 S4 S5 S6 S7 S8
Com ref
Rcvr 2Rcvr 1
2-C1-C 3-C 4-C 5-C 6-C 7-C 8-C
1=(S1+S2+S3+S4)/4 2=(S5+S6+S7+S8)/4
T=1-2
Judah Levine, NIST, CENAM, Sept 2012: 17
Partial two-way method
Delay is stable and is white pm– Measure only occasionally– Unique to PTP/1588– Useful only in special cases
• Problems in wide-area networks• False-tickers and the trust problem
Judah Levine, NIST, CENAM, Sept 2012: 18
Full Two-wayFull Two-way
Measure round-trip delay on every Measure round-trip delay on every calibrationcalibration– Delay is not stable and not white pm Delay is not stable and not white pm
over longer periods over longer periods – Transmission delay is one-half of Transmission delay is one-half of
measured valuemeasured value• Delay is symmetric on the averageDelay is symmetric on the average
Telephone system using ACTSTelephone system using ACTS Internet using full NTP Internet using full NTP
Judah Levine, NIST, CENAM, Sept 2012: 19
Real-world limitationsReal-world limitations
Inbound and outbound delays are Inbound and outbound delays are not equalnot equal– Realized as a two-way physical circuit Realized as a two-way physical circuit
with some one-way componentswith some one-way components• Physical component dispersionPhysical component dispersion
– Realized with a reversible one-way Realized with a reversible one-way physical circuitphysical circuit• Time dispersionTime dispersion
– Realized using a packet networkRealized using a packet network• Asymmetric queuing and routing delaysAsymmetric queuing and routing delays
Judah Levine, NIST, CENAM, Sept 2012: 20
Effect of Asymmetry - 1Effect of Asymmetry - 1
Method assumes one-way delay is one-Method assumes one-way delay is one-half of round-trip value. Time error is half of round-trip value. Time error is given bygiven by
fractionoutboundk
delaytripround
k
)5.0(
0≤ k ≤ 1
Judah Levine, NIST, CENAM, Sept 2012: 21
Effect of asymmetry - 2Effect of asymmetry - 2
0 Round-trip delay→
k=1, =/2
k=0, = -/2
Smaller delay has smaller asymmetry error
Judah Levine, NIST, CENAM, Sept 2012: 22
NTP Service model
Operate servers at many locations– Minimizes delay error for all users– No single point of failure– How are remote servers synchronized?
• Time link to source of UTC(lab) Performance limited by delay jitter
and asymmetry– Few percent of round-trip measurement
• Accuracy < 50 ms, often < 10 ms, maybe ~ 1ms
Judah Levine, NIST, CENAM, Sept 2012: 23
Asymmetry – the bottom lineAsymmetry – the bottom line
Static asymmetry generally cannot Static asymmetry generally cannot be detected or removedbe detected or removed– Limits accuracy of any protocolLimits accuracy of any protocol– Multiply-connected networks Multiply-connected networks
sometimes help in detecting sometimes help in detecting asymmetryasymmetry• Apparent time difference over different Apparent time difference over different
pathspaths
Judah Levine, NIST, CENAM, Sept 2012: 24
Summary - 1Summary - 1
One-way methods are simple and are One-way methods are simple and are good enough for many applicationsgood enough for many applications– Path delay can be ignoredPath delay can be ignored– Path delay can be modeled adequatelyPath delay can be modeled adequately
Common-view depends on equality of Common-view depends on equality of delays along two one-way pathsdelays along two one-way paths– Requires data exchange between stationsRequires data exchange between stations
Neither method can attenuate local Neither method can attenuate local effectseffects
Judah Levine, NIST, CENAM, Sept 2012: 25
Summary - 2Summary - 2 Two-way depends on equality of delay in Two-way depends on equality of delay in
opposite direction along a single pathopposite direction along a single path Limited by the symmetry of the link delay Limited by the symmetry of the link delay
between the transmitter and the receiverbetween the transmitter and the receiver– Magnitude of the delay not importantMagnitude of the delay not important– Message format not importantMessage format not important
Error in time data proportional to Error in time data proportional to asymmetry and delayasymmetry and delay– Shorter paths will always have smaller Shorter paths will always have smaller
errorserrors
Judah Levine, NIST, CENAM, Sept 2012: 26
For more informationFor more information
List of publications of the NIST time List of publications of the NIST time and frequency division are in the and frequency division are in the publications menu of our web page:publications menu of our web page:
tf.boulder.nist.govtf.boulder.nist.gov Many of these publications are on-line Many of these publications are on-line ““Time and Frequency Measurement” Time and Frequency Measurement”
by C. Hackman and D. B. Sullivan, by C. Hackman and D. B. Sullivan, published by the American Association published by the American Association of Physics Teachers, 1996.of Physics Teachers, 1996.