Timekeeping. From clocks to a time scale for reliable and highly accurate national, global, or regional timing reference. Sam Stein Chief Scientist

Transcription

1 Timekeeping From clocks to a time scale for reliable and highly accurate national, global, or regional timing reference Sam Stein Chief Scientist October 7, 2013

2 Clocks Definition of Terms and Performance 2

3 For Frequency Generation You need A pendulum (mechanical oscillator) The Earth rotating Electronic Oscillator Atomic resonance a method of generating a repeatable event 3

4 Frequency Frequency = the number of cycles per second Ideal frequency source generates a pure, repeatable sine wave 4

5 Frequency Definitions Offset the frequency error from the ideal (fast or slow) Accuracy refers to the frequency offset of a device without direct calibration Stability the constancy of frequency over a given time interval Short Term Stability the change of frequency over seconds from noise and vibration. Sometimes called jitter Long Term Stability the change of frequency over hours, days, or months. Likely due to aging and temperature Aging change of frequency over time (also called drift) Temperature Stability the change of frequency over temperature 5

6 What is Frequency Stability & Accuracy Courtesy John Vig 6

7 Oscillator Stability Over Time Frequency stability typically improves in the short term, stabilizes, then becomes worse in the long term 7

8 What are the Influences on Oscillator Frequency? Time Short term (noise) Long term (aging) Temperature Static frequency versus temperature Dynamic frequency versus temperature (warm-up) Thermal history (retrace) Acceleration Gravity, vibration, shock Other Power supply variation Humidity 8

9 Taking Frequency Measurements Frequency measurements are performed over a time interval: t 1 to t 2 : Signal 1 = Signal 2 t 1 to t 3 : Signal 2 is more stable than Signal 1 9

10 Short and Long Term Oscillator Stability Stability 1.00E E E E E E E E E E-15 Frequency Stability 1 Sec 10 Sec 100 Sec 1K Sec 10K Sec 100K Sec Week Month Time OCXO Hi-Stab OCXO Rb Cs-High Perf H-Maser GPS Some oscillators perform better short term others long term 10

11 Puttingthe Fundamentals into Perspective What is one part in 10 10? (As in 1 x /day aging) ~1/2 cm out of the circumference of the Earth. ~1/4 second per human lifetime (of ~80 years). Power received on Earth from a GPS satellite, -160 dbw, is as bright as a flashlight in Los Angeles when viewed in New York City, ~5000 km away The second is the most precise SI unit of measure! 11

12 Frequency Sources for Precision Timekeeping Quartz Crystal Oscillators Gas Cell Passive Atomic Frequency Standards Gas Cell Active Atomic Frequency Standards Atomic Beam Frequency Standards Cold Atom Frequency Standards 12

13 Hierarchy of Oscillator Types Oscillator Type* Accuracy** Typical Applications Crystal oscillator (XO) Temperature compensated crystal oscillator (TCXO) Microcomputer compensated crystal oscillator (MCXO) Oven controlled crystal oscillator (OCXO) Small atomic frequency standard (Rb, RbXO) Active Hydrogen Maser High performance atomic standard (Cs) 10-5 to to (with per g option) to Computer timing Frequency control in tactical radios Spread spectrum system clock Navigation system clock & frequency standard, MTI radar C 3 satellite terminals, bistatic, & multistatic radar Timekeeping, radio astronomy Strategic C 3, EW * Sizes range from <5cm 3 for clock oscillators to > 30 liters for Cs standards Costs range from <$5 for clock oscillators to > $50,000 for Cs standards ** Including environmental effects (e.g., -40 o C to +75 o C) and one year of aging 13

14 Crystal Oscillator Types Voltage Tune Output C f f +10 ppm 25 0 C C T Crystal Oscillator (XO) -10 ppm Temperature Sensor Compensation Network or Computer XO C f f +1 ppm -1 ppm C T Temperature Compensated (TCXO) Oven Oven control XO Temperature Sensor Oven Controlled (OCXO) C f f +1 x x C T 14

15 Crystal Oscillator Portfolio: Price vs.performance overview Price 9250 Low-g OCXO 9940 VCXO 9960 TCXO 9638 OCXO 8200LN Rb 9700 OCXO 9600QT OCXO 9500B USO Tactical Military UAV/Avionics Satcoms Master Osc Applications Low Portable/Battery High Flight/Mission Critical Space Ref Standards Performance 15

16 Gas Cell Atomic Clocks Rb: 6.8 GHz Passive Buffer gas H Maser: 1.4 GHz Active Wall coated cell The resonant frequency of atoms does not age the apparatus to interrogate or confine atoms is affected by the environment and ages in time 16

17 Rubidium Frequency Standard Basics Magnetic Shield Lamp Oven Filter Oven Cavity Oven Lamp Filter Absorption RF Excitation Lamp Exciter Rb-87 Lamp Rb-85 Rb-87 Photo- Detector Signal Out Coil Cell Cell C-Field Coil C-Field Current (3) Oven Temperature Sensors and Heaters Physics Package µw Interrogation RF Chain Servo Modulation Discriminator Signal Frequency Lock Loop Servo Amplifier Control Voltage Crystal Oscillator O/P Amp O/P 17

18 Rubidium Gas Cell Frequency Standards Most widely used type of atomic clock Smallest, lightest, lowest power Least complex, least expensive, longest life Excellent performance, stability & reliability Device of choice when better stability is needed compared to crystal oscillator Lower aging, lower temperature sensitivity Faster warm-up, excellent retrace Used as an inexpensive holdover technology 18

19 SA.31M Laser Pumped Rb& Chip Scale Atomic Clock (CSAC) Rb Miniature Atomic Clock (MAC) Small form factor: 51mm x 51mm x 18mm (H) Lower power: 25 o C Stability 1s <3E-11; 100s <8E-12 Aging: <3E-10/month Temp Stability: <1E-10 ( 10 o C to +75 o C) CSAC (Chip Scale Atomic Clock) Volume: <17 cc Weight : 35g Very Low power: <120 mw Stability 1s <2E-10; 100s <2E-11 Aging: <3E-10/month Temp Stability: <5E-10 (0 to +75 o C) 19

20 Active Hydrogen Maser Block Diagram Teflon coated storage bulb Microwave cavity Microwave output State selector Hydrogen atoms 20

21 Active Hydrogen Masers Excellent frequency stability up to 1 month 40X superior to high performance cesium Mature technology with good operating lifetime and reliability Design of choice when the ultimate frequency stability is required Applications: National time scale and Radio Astronomy applications MHM 2010 MASER: Microwave Amplification by Stimulated Emission of Radiation 21

22 H Maser Applications Metrology Where? International timekeeping laboratories Why a Maser? Provides superior frequency stability out to one month Stability is the key attribute in a timescale application and today s primary standards research Time scale reference clock steered to the cesium ensemble or primary standards Radio Astronomy Where? VLBI Very long baseline interferometry VLBA Very large baseline arrays Why a Maser? Offers frequency stability for multiple VLBA stations to operate coherently Perfect Clock : Maser (short term) + Cesium (long term) 22

23 Cesium Beam Tube Fundamentals 9192 MHz Detector F=3 + F=4 S N F=3 F=4 F=3 + F=4 S N F=3 F=4 "A" Magnet "B" Magnet Magnetically-Selected CBT Advantages: Unperturbed flow of atoms Disadvantage: Finite life, relatively short interrogation time 23

24 Cesium Technology Applications Cesium Technology is considered the most comprehensive holdover option against GNSS vulnerabilities Exhibit no frequency drift Maintains 5x10-13 accuracy over the life of the instrument Critical for long-term autonomous operation No on-going calibration required More expensive than Rubidium and OCXO Consumes more power and space Typical applications Fixed wireline communications infrastructure Under sea (Submarine) Satellite ground stations Metrology and Time Keeping 24

25 Frequency and Time Relationship 25

26 Time is Derived from Frequency Every clock ever made is an oscillator + a counter The 1 pulse-per-second (PPS) is the epochor definition of the on-time marker of a clock Further counting of the 1PPS is used to keep track of seconds, minutes, hours, days, and years OSC Counter 1PPS FAST / EARLY SLOW / LATE 1PPS Clock 1PPS rising edge is typically on-time: HH:MM:SS

27 OCXO Accumulated Time Error Clock operating from an OCXO with an offset & aging of and a temperature error of 1 degree C Microseconds Accumulated Time Error Aging Error(uS) Temp Err.(uS) Offset Err.(uS) Total Err.(uS) Hours Oscillator errors accumulate impacting clock performance 27

28 Cesium Time Error Due to Offset Cesium Time Error (Offset 2E-12) Microseconds Time error = Freq offset * seconds/day or x 86400s = ~0.2µs / day Offset Err.(uS) Days Cesium atomic oscillators do not age 28

29 Time Scale Basics 29

30 Concept of Time Identification when: An event occurs Duration of an event Interval between events Three concepts that relate to time are: Date: A reference point represented by the exact time-ofday and, often day and year, that indicates when an event occurred Interval: The duration or elapsed time between two instants or subsequent events Synchronization: Refers to two clocks set to the same time or two events happening at the same instant of time 30

31 What is Time Scale and Types Time scale is an agreed upon measuring system for counting time Time Scales Types Astronomical: UT1 Universal Time (polar corrections) Earth rotation angle Atomic: TAI International Atomic Time Commercial & Primary clocks Atomic -with periodic adjustments: UTC UTC(Lab) Coordinated Universal Time UTC per contributing laboratory Time scales are based on agreements between humans 31

32 What is UTC Time Scale USNO (US) NIST (US) NPL (UK) PTB (Germany) CRL (Japan) Atomic Clock Atomic Clock Atomic Clock Atomic Clock Atomic Clock International Earth Rotation Service (IERS) ENSEMBLE AVERAGE Add leap seconds to correct for Earth s rotation BIPM* International Atomic Time (TAI) BIPM* Coordinated Universal Time UTC BIPM Circular T Report *BIPM Bureau of Weights and Measures National Metrology Institute (NMI) UTC(k) Radio, Telephone, Network, Satellite Dissemination of UTC(k) 32

33 Timing Keeping by Clock Types BIPM International Atomic Time Component Clocks by Weight 12% 1% 5071A Cesium Maser Others H-Maser(s) 87% Cesium Based on BIPM Annual Report % of world time keeping powered by Cesium and Masers 33

34 What is Precise Time-Scale System 5MHz UTC(k) Cesium (5071A) BIPM 5MHz Correction H-Maser(s) Precise Time-Scale System The frequency stability of Symmetricom sensures the uniformity of the world s time 34

35 How Does a Time Scale Work? Step 1 Measure the time differences between all the clocks and the reference clock Step 2 Estimate the time, frequency, and aging differences between all the clocks and the reference clock Step 3 Apply the time scale algorithm to calculate the corrections to the reference clock s time, frequency, and aging needed to render it equal the time scale Step 4 Connect a synthesizer to the reference clock and steer the output of the synthesizer to approximate the time scale using corrections from step 3 35

36 Symmetricom Time Scale Hardware Non-redundant Time Scale with 5 Cesium Clocks Database Computer Real-time Clock Measurement System Cesium Clocks Time Servers UPS Charger Batteries 36

37 Baseline Precise Time-Scale System Consists of: Equipment rack One 5071A high performance cesium standard 8 channel measurement hardware with database server Real-time clock with chassis mainframe and modules UTC recovery, steering, common view, monitor and control software LCD monitor Keyboard Battery backup unit System integration and packaging On-site installation and training 37

38 Precise Time-Scale System Key Functions Generates timing outputs synced to UTC using a GNSS receiver Option to incorporate multiple clocks to optimize availability Measurement system to provide clock time differences Distributes the real-time clock: 5MHz, 1PPS and IRIG-B Uses a time-scale algorithm to combine the clocks into one output Archives data for review Performs GNSS common view time transfer using BIPM procedure Automatically prepares BIPM reports 38

39 Precise Time-Scale Key System Specifications Frequency Accuracy: +1x10-14 for 10 day average after 60 days of continuous operation Frequency Holdover: +1x10-13 for 30 days over the full temp range Phase Noise Offset Freq (Hz) 5 MHz (dbc/hz) 10 MHz (dbc/hz) KHz KHz, 100 KHz Short-Term Stability (ADEV) Ƭ (s) σ Y (Ƭ) 1 5x x x k 2.7x k 8.5x k 2.7x

40 Precise Time-Scale System Features and Benefits Features State-of-the-art UTC generation Support for participaton with BIPM User friendly GUI and Database for archiving Performance Frequency accuracy +1x10-14 Time accuracy of +10 ns RMS to UTC Customer Benefits Less than 1 year to full UTC participation Performance on par with the best national laboratories due to time scale algorithm from Symmetricom Reduced complexity and ease of integration 40

41 Precise Time-Scale User Interface & Database Web-Based User Interface Configure MMS Manage Clocks Retrieve and manipulate data Add user accounts Database (includes Manager) 10 years of data storage capacity TCP/IP interface to MMS w/listener process Every second data from MMS is sent to listener for data processing and storing Web Interface for database and measurement systems RAID 5 Disk Array 4 on-line diskesstriped for 360 GB actual storage 1 auto-failover hot spare Clock Steering GUI with estimated offset from UTC Front Panel of Database w/raid 5 disk array 41

42 Precise Time-Scale System Sample Performance Performance vs. UTC UTC - UTC(lab) ns MJD 2 August 2007 through 11 June 2008 RMS = 5 ns Mean = 2.5 ns 42

43 Performance of KAS-2 Time Scale Analyzed using the method of the N-cornered hat 5 High Performance and 1 Standard Performance Cesium Clocks 3 High Performance Cesium Clocks And 3 Hydrogen Masers 43

44 Effect of Additional Cesium Clocks Precise Time-Scale can achieve better performance with additional clocks: Availability Stability Cesium maintains specified accuracy for life of instrument 5071A Cesium clock Long term autonomous operation Quartz or Rubidium not suitable since they exhibit frequency drift 44

45 Effect of Maser Clock to the System Adding an active Hydrogen Maser to the system improves the stability of the time scale Symmetricom s Maser has a reputation for its excellent short and long term stability, reliability and long life MHM-2010 active hydrogen Maser 45

46 Precise Time-Scale System Service Offering Standard Services System design, integration and testing System packaging and documentation On-site installation and training Rack, monitor, keyboard and cables Optional Services Site survey and verification Customer-witnessed factory acceptance testing Extended warranty and in-country support System spares program Software and systems support 24x7 BIPM contribution and timekeeping consulting 46

47 Time Scale Technical Details Overview What is a time scale? Why is this an interesting problem? Types of time scales Historical perspective Symmetricom turn-key time scale Performance 47

48 What is a Time Scale? Why do We Need Them? A time scale is a method of computing corrections for each member of an ensemble (group) of clocks so that any of them may be steered to produce the time and frequency of a more nearly perfect clock Why are they needed? Extend beyond the life of one clock More uniform than any single clock More robust than any single clock Improved availability An example of a time scale is International Atomic Time (TAI) Applications for time scales Metrology can benefit from post processing Synchronization requires real time output Science highest performance in real time and non-real time 48

49 Why is This an Interesting Problem? Computing a time scale is more than just an estimation problem The inputs to the calculation are the time differences between the clocks There are an infinite number of solutions for the times of the clocks that are consistent with the measurements There are two components to the problem How much of a measured time difference between two clocks is attributable to each clock? We can estimate this How much of the time change of a clock is unobserved because the clocks move together? We have no way to address this question 49

50 Types of Time Scales Batch processing vs. Bayesian (recursive) Batch algorithms save measurements over time and compute the next set of time scale values using all measurements at once The Bayesian approach is to save the state at the time of the last computation and update it using measurements at the next time Real-time vs. after the fact (filter vs. smoother) Smoothers compute the time scale at a point in time based on prior and subsequent data Smoothing produces better estimates than filtering Model dependent vs. model independent Clocks are weighted based on their performance at one averaging time vs. multiple averaging times (single vs. multiresolution) Cookbook recipes vs. designed approaches 50

51 Historical View of Time Scales Earliest time scales (circa 1968) used the basic time scale equation Sum of the random shocks for the clock phases is zero Detailed performance of the algorithm depended on the method of frequency estimation In 1990, Stein developed an algorithm where the sum of the random shocks for every clock state is zero (KAS-2) Time scale performance can be optimized over each averaging time for which a different noise process is dominant For the first time the state estimation problem was separated from the time scale computation problem In 2004, Davis, Stacey, and Greenhallused Markov noise processes to model flicker noise in clocks in a time scale In 2011, Senior and Percival developed a model independent multiscale time scale algorithm using the discrete wavelet transform 51

52 Clock Model Traditional model Noise in a clock can most generally be describe as S y (f) = h α f α r r r x ( t ) x ( t ) s ( t ) i k 1 i k i k + = Φ + δ δ δ δ Φ = δ / / 2 White phase noise White frequency noise Random walk frequency noise Random run frequency noise Addition of Markov noise per Stacey et. al. For each noise process N i= 1 a ˆ ε ( t) = 0 i i So there are 4 weights that determine Performance over 4 averaging times 4 Markov processes 52

53 Time Scale Computation Kalman filter operates on the clock time differences to estimate the noise differences between each clock and the reference Observed time differences are used to estimate the random shocks of the time, frequency, and frequency aging KAS-2 Time scale algorithm separates the noise difference estimates into individual clock noise estimates Uses the assumption that the sum of the random shocks over all the clocks is zero (true in the limit of an infinite number of clocks) References U. S. Patents 5,155,695 and 5,315,566 S. R. Stein Advances in Time Scale Algorithms, 23 rd Annual PTTI Meeting, S. R. Stein, Time Scales Demystified, Proceedings of the 57 th Annual Frequency Control Symposium,

54 Why Add Markov Noise? Maser Allan Deviation White phase noise Random walk frequency noise plus aging Markov noise processes can be used to model the transition between white or random walk phase noise and random walk frequency noise (flicker region) 54

55 Sum of 4 Markov Processes Models Flicker FM Over 3 Decades of Averaging Time From Davis, Greenhall, and Stacey (PTTI 2004) 55

56 Comparison of Model Dependent Time Scale with a Wavelet Multi-Resolution Time Scale Issue the multi-resolution analysis does not have a built in forecasting method, which is desirable for real-time applications and steering 56

57 Technical Take Aways Use of Kalman filter provides minimum squared error estimates that have optimum transient response Bayesian approach is just as effective as those that require large memory of past data and extensive recomputation Allows compact highly automated time-scale system with low power computer and small memory All noise types of atomic clocks are well modeled making future time prediction highly optimal Weighting by noise type provides equal advantage (multiresolution analysis) as the wavelet approach but provide the added capability to forecast clock times 57

58 Precise Time-Scale System Key Take Away s Affordable time scale system for international markets Base system includes RTC/Measurement HW, SW and one Cesium clock System integration, Installation and Training included Optional support services available Option to add additional clocks Can use existing 5071A and Symmetricom Maser in the lab Key customer benefits Lower total cost of ownership High availability Affordable price point with options to grow Proven support services Reduced complexity and ease of operation Ability to participate in the BIPM time scale quickly 58

59 Samuel R Stein sstein@symmetricom.com 4775 Walnut St Boulder, CO Tel: Fax: