60 | MEASURE www.ncsli.org

REVIEW PAPERS

1. IntroductionTime and frequency measurementsoccupy a special place, and possess acertain mystique, in the world of metrol-

ogy. The unit of time interval, the second(s), and its reciprocal unit of frequency,the hertz (Hz), can each be measuredwith more resolution and less uncer-tainty than any other physical quantity.NIST and a handful of other nationalmetrology laboratories can currentlyrealize the second to uncertainties meas-ured in parts in 1016 [1], and NIST has

experimental standards already in placethat promise uncertainties at least one ortwo orders of magnitude smaller. [2]These uncertainties represent the pinna-cle of the metrology world, and have a“gee whiz” quality that attracts mediaattention and captures the public’s imag-ination. These tiny uncertainties are alsoof interest to scientists and design engi-

Legal and Technical MeasurementRequirements for Time and Frequency1

Michael A. Lombardi

Abstract: This paper discusses various technologies and applications that rely on precise time and frequency, andexplores their legal and technical requirements for measurement uncertainty. The technologies and applications dis-cussed include financial markets, the wired and wireless telephone networks, radio and television broadcast stations,the electrical power grid, and radionavigation systems. Also discussed are the legal and technical requirements for“everyday” metrology situations, including wristwatches, commercial timing devices, and radar devices used by lawenforcement officers.

Michael A. Lombardi

Time and Frequency DivisionNational Institute of Standardsand Technology325 BroadwayBoulder, CO 80305-3328 USAEmail: lombardi@nist.gov

1 This paper is a contribution of the United States government and is not subject to copyright.The illustrations of commercial products and services are provided only as examples of thetechnology discussed, and this neither constitutes nor implies endorsement by NIST.

MEASURE | 61Vol. 1 No. 3 • September 2006

REVIEW PAPERS

neers, because history has shown that astime and frequency uncertainties getsmaller, new technologies are enabledand new products become possible.

For metrologists, however, it can bedifficult to place the tiny uncertainties ofstate-of-the-art time and frequency meas-urements into their proper context. Mostmetrology work is performed in supportof “real world” systems that require theirmeasuring instruments and standards tobe within a specified tolerance in orderfor the system to perform as designed.Thus, metrologists are concerned withquestions such as: What type of fre-quency uncertainty is required so that apolice officer knows that a measurementof vehicle speed is valid? How closedoes a radio station’s carrier frequencyneed to be controlled so that it does notinterfere with another station? What fre-quency tolerance does a telephonenetwork need in order to avoid droppingcalls? These questions are answered bylooking at both the legal and technicalrequirements of time and frequencymetrology, the topics of this paper. Thesetopics are covered by first looking at therequirements for “everyday” metrology,and then examining the requirements foradvanced applications.

2. Requirements for “Everyday” Metrology

In “everyday” life, we check our wrist-watches for the correct time, pay for timeon parking meters and other commercialtiming devices, play and listen to musicalinstruments, and drive our cars at a safe

speed that is at or below the postedspeed limit. The modest time and fre-quency requirements of these activitiesare described in this section.

2.1 WristwatchesWristwatches are unique devices, theonly metrological instruments that weactually wear. Most wristwatches containa tiny quartz oscillator that runs at anominal frequency of 32 768 Hz. Thereare no legally required uncertainties forwristwatches, but at least one majormanufacturer specifies their watches asaccurate to within 15 s per month, orabout 0.5 s per day, a specification thatseems to be typical for the quartz watchindustry. This translates to an allowablefrequency uncertainty of about 0.2 Hz, ora dimensionless uncertainty near 6 # 10-6.

2.2 Commercial Timing Equipmentand Field Standard Stopwatch

Commercial timing equipment includesdevices such as parking meters, taxicabmeters, and coin operated timers used inlaundries and car washes. NIST Hand-book 44 [3], which is used by all 50states as the legal basis for regulatingcommercial weighing and measuringdevices, uses the terms overregistrationand underregistration when defining thelegal requirements of commercial timingdevices. Overregistration means that theconsumer received more time than theypaid for; underregistration means thatthey received less time than they paid for.The laws are intended to protect con-sumers, and underregistration is of much

greater concern. For example, a personwho pays for 10 minutes on a parkingmeter is legally entitled to receive closeto 10 minutes before the meter expires,but no law is broken if the meter runs formore than 10 minutes. Table 1 summa-rizes the legal requirements of commer-cial timing devices. [3]

Commercial timing devices are oftenchecked with field standard stopwatchessince they can not be easily moved to acalibration laboratory. Most modernstopwatches are controlled by quartzoscillators, and they typically meet orexceed the performance of a quartzwristwatch (as discussed above). Stop-watches are sometimes calibrated using auniversal counter and a signal generator(see Fig. 1), or with a device designed tomeasure the frequency of their time baseoscillator. However, most stopwatch cal-ibrations are still made by manuallystarting and stopping the device undertest while listening to audio timingsignals from NIST radio station WWV ora similar source. For this type of calibra-tion, the longer the time interval meas-ured, the less impact human reactiontime will have on the overall measure-ment uncertainty. [4] To avoid unreason-ably long calibration times, the legallyrequired measurement uncertainty istypically 0.01 % or 0.02 % (1 or 2 partsin 104). NIST Handbook 44 [3] specifies15 s for a 24 hour interval, or 0.017 %.Some states and municipalities havetheir own laws that list similar require-ments. For example, the state of Pennsyl-vania code [5] states that an electronic

Table 1. Legal requirements of commercial timing devices.

CommercialTiming Device

Overregistration Underregistration

Requirement Uncertainty Requirement Uncertainty

Parking Meter None NA 10 s per minute5 minutes per half hour7 minutes per hour

11.7 %to 16.7 %

Time clocks and time recorders

3 s per hour, not to exceed 1 minute per day

0.07 % to 0.08 %

3 s per hour, not to exceed1 minute per day

0.07 % to 0.08 %

Taximeters 3 s per minute 5 % 6 s per minute 10 %

Other Timing Devices 5 s for any interval of 1 minute or more

NA 6 s per minute 10 %

62 | MEASURE www.ncsli.org

REVIEW PAPERS

stopwatch shall comply with the follow-ing standards:(i) The common crystal frequency shall

be 32 768 Hz with a measured fre-quency within plus or minus 3 Hz,or approximately .01% of the stan-dard frequency.

(ii) The stopwatch shall be accurate tothe equivalent of plus or minus9 seconds per 24-hour period.

2.3 Musical PitchThe pitch of a musical tone is a functionof the speed at which air has been set inmotion. The speed is measured as thenumber of complete vibrations – back-wards and forwards – made by a particleof air in one second. When pitch is pro-duced by a vibrating column of air, thepitch of the same length of pipe varieswith temperature: for a 1 °F difference,pitch will vary by 0.001 Hz. [6]

The international standard for musicalpitch was first recognized in 1939, andreaffirmed by the International Organi-zation for Standardization in 1955 and1975. [6, 7] It defined internationalstandard pitch as a system where Aabove “middle” C (known as A4) istuned to 440 Hz. A 440 Hz tone is broad-cast by NIST radio stations WWV andWWVH for use as a musical reference. [8]

The ability of the human ear to dis-

criminate between differences in pitchdepends upon many factors, includingthe sound volume, the duration of thetone, the suddenness of the frequencychange, and the musical training of thelistener. However, the just noticeable dif-ference in pitch is often defined as5 cents, where 1 cent is 1/100 of theratio between two adjacent tones on apiano’s keyboard. Since there are 12tones in a piano’s octave, the ratio for afrequency change of 1 cent is the 1200throot of 2. Therefore, raising a musicalpitch by 1 cent requires multiplying bythe 1200th root of 2, or 1.00057779. Bydoing this five times starting at 440 Hz,we can determine that 5 cents high isabout 441.3 Hz, or high in frequency byabout 0.3 %. [8] Some studies have shownthat trained musicians can distinguishpitch to within 2 or 3 cents, or to within0.1 % or better. Thus, frequency errorsof 0.1 % or larger can change the waythat music sounds for some listeners.

2.4 Law Enforcement Law enforcement officers use radardevices to check vehicle speed. Thesedevices are normally calibrated by point-ing them at tuning forks whose oscilla-tions simulate vehicle speed. Forexample, a radar device might be cali-brated by checking it with a tuning fork

labeled 30 mph (miles per hour) to testthe low range, and another fork labeled90 mph to test the high range. Thenominal frequency of the tuning forkvaries depends upon the radar devicebeing used; a K-band tuning fork labeled30 mph will oscillate at a higher fre-quency than an X-band fork with thesame label.

To meet legal requirements that varyfrom state to state, tuning forks must beperiodically calibrated, often with a fre-quency counter or an oscilloscope. A fre-quency uncertainty of 0.1 % (1# 10-3) issufficient for tuning fork calibrations.Although this seems like a coarserequirement, a frequency uncertainty of0.1% translates directly to a speeduncertainty (for example, 0.03 mph at 30mph, 0.09 mph at 90 mph) for either X-band or K-band radar devices. This isinsignificant when you consider thatspeeding tickets are seldom issued unlessa motorist exceeds the posted speed limitby at least several miles per hour. [9]

3. Requirements for Financial Markets

To protect investors from securities fraudand to ensure that financial transactionsoccur in an orderly fashion that can beaudited if necessary, financial marketsoften require all recorded events to betime tagged to the nearest second. Forexample, after an August 1996 settle-ment with the Securities Exchange Com-mission (SEC) involving stock marketfraud related to the improper executionof trades, the National Association ofSecurities Dealers (NASD) needed a wayto perform surveillance of the NASDAQmarket center. As a result, the NASDdeveloped an integrated audit trail oforder, quote, and trade information forNASDAQ equity securities known asOATS (Order Audit Trail System).

OATS introduced many new rules forNASD members, including requiring allmembers to synchronize their computersystem and mechanical clocks everybusiness day before the market opens toensure that recorded order event timestamps are accurate. To maintain clocksynchronization, clocks should bechecked against the standard clock andresynchronized, if necessary, at predeter-mined intervals throughout the day, so

Figure 1. A stopwatch calibration that employs the totalize function of a universal counter(courtesy of Sandia National Laboratories).

MEASURE | 63Vol. 1 No. 3 • September 2006

REVIEW PAPERS

that the time kept by all clocks canalways be trusted. NIST time was chosenas the official time reference forNASDAQ transactions.

NASD OATS Rule 6953, Synchroniza-tion of Member Business Clocks, appliesto all member firms that record order,transaction, or related data to synchro-nize all business clocks. In addition tospecifying NIST time as the reference, itrequires firms to keep a copy of theirclock synchronization procedures on-site. One part of the requirements [10]reads as follows:

All computer system clocks and mechan-ical time stamping devices must be syn-chronized to within three seconds of theNational Institute of Standards andTechnology (NIST) atomic clock. Anytime provider may be used for synchro-nization, however, all clocks and timestamping devices must remain accuratewithin a three-second tolerance of theNIST clock. This tolerance includes all ofthe following: • The difference between the NIST stan-

dard and a time provider’s clock: • Transmission delay from the source;

and • The amount of drift of the member

firm’s clock. For example, if the time provider’s clockis accurate to within one second of theNIST standard, the maximum allowabledrift for any computer system or mechan-ical clock is two seconds.

Prior to the development of OATS, bro-kerage houses often used clocks and timestamp devices that recorded time indecimal minutes with a resolution of 0.1minutes (6 s). The new OATS require-

ments forced the removal of these clocks.Fig. 2 shows an OATS compliant clockthat synchronizes to NIST time via theInternet. Clocks such as this one are syn-chronized to the nearest second, but upto 3 seconds of clock drift are allowedbetween synchronizations.

4. Requirements for BroadcastingUnlike time metrology, which has originsthat date back thousands of years, fre-quency metrology was not generally dis-cussed until about 1920, when commercialradio stations began to appear. Radiopioneers such as Marconi, Tesla, andothers were not aware of the exact fre-quencies (or even the general part of thespectrum) that they were using. How-ever, when the number of radio broad-casters began to proliferate, keepingstations near their assigned frequenciesbecame a major problem, creating aninstant demand for frequency measure-ment procedures and for frequency stan-dards. [11] Today, with stable quartz andatomic oscillators readily available, keep-ing broadcasters “on frequency” is rela-tively easy, but all broadcasters must provideevidence that they follow the FederalCommunications Commission (FCC)regulations as described in Section 4.1.Figure 3 shows a mobile calibration vanthat makes on-site visits to transmittersites to check their frequency.

4.1 FCC Requirements for Radioand Television Broadcasting

The FCC specifies the allowable carrierfrequency departure tolerances for AMand FM radio stations, television sta-tions, and international broadcast sta-tions. [12] These tolerances are specifiedas a fixed frequency across the broadcastband of ±20 Hz for AM radio, ±2000 Hzfor FM radio, and ±1000 Hz for theaudio and video television carriers, andas a dimensionless tolerance of0.0015 % for international shortwavebroadcasters. The allowable tolerancesare converted to dimensionless uncer-tainties and summarized in Table 2.

4.2 Frequency Requirements forColor Television Subcarriers

For historical design reasons, the chromi-nance subcarrier frequency on analogcolor televisions is 63/88 multiplied by5 MHz, or about 3.58 MHz. To ensureadequate picture quality for televisionviewers, federal regulations specify thatthe frequency of this subcarrier mustremain within ±10 Hz of its nominalvalue, and the rate of frequency driftmust not exceed 0.1 Hz per second. [13]This corresponds to an allowable toler-ance of ±0.044 Hz for the 15 734.264 Hzhorizontal scanning frequency, a dimen-sionless frequency uncertainty near 3#10-6.

Figure 2. A OATS compliant clock used totime stamp financial transactions (courtesyof the Widmer Time Recorder Company).

Figure 3. A mobile calibration van that tests whether or not a transmitter is within toler-ances specified by the FCC (courtesy of dbK Communications, Inc.).

64 | MEASURE www.ncsli.org

REVIEW PAPERS

5. Requirements for ElectricPower Distribution

The electric power system in NorthAmerica consists of many subsystemsthat interconnect into several massivegrids that span the continent. The systemdelivers the 60 Hz AC frequency to manymillions of customers by matching powergeneration levels to transmission capabil-ity and load patterns. The entire powersystem relies on time synchronization,and synchronization problems can leadto catastrophic failures. For example, themassive August 2003 blackout in theeastern regions of the United States andCanada was at least partially caused bysynchronization failures. [14]

The timing requirements of the powerindustry vary (Table 3), because differentparts of the system were designed at dif-ferent times, and the entire system hasevolved over many years. The olderparts of the system have less stringenttiming requirements because they weredesigned using technologies that pre-dated the Global Positioning System(GPS). The newer parts of the system

rely on the ability of GPS to provideprecise time synchronization over a largegeographic area.

Since electrical energy must be used asit is generated, generation must be con-stantly balanced with load, and the alter-nating current produced by a generatormust be kept in approximate phase withevery other generator. Generationcontrol requires time synchronization ofabout 10 ms. Synchronization to about 1ms is required by event and faultrecorders that supply information usedto correct problems in the grid andimprove operation. Stability controlschemes prevent unnecessary generatorshutdown, loss of load, and separation ofthe power grid. They require synchro-nization to about 46 µs (±1° phase angleat 60 Hz), and networked controls haverequirements one order of magnitudelower, or to 4.6 µs (±0.1° phase angle at60 Hz). Traveling wave fault locatorsfind faults in the power grid by timingwaveforms that travel down power linesat velocities near the speed of light.Because the high voltage towers are

spaced about 300 meters apart, thetiming requirement is 1 µs, or the periodof a 300 meter wavelength. [15] Newermeasurement techniques, such as syn-chronized phasor measurements, requiretime synchronization to CoordinatedUniversal Time (UTC) to within 1 µs,which corresponds to a phase angle accu-racy of 0.022 ° for a 60 Hz system. Alocal time reference must be applied toeach phasor measurement unit, and GPSis currently the only system that can meetthe requirements of synchrophasor meas-urements. [16] Commercial phasormeasurement units that receive GPSsignals are shown in Fig. 4.

The 60 Hz frequency delivered to con-sumers is sometimes used as the res-onator for low priced electric clocks andtimers that lack quartz oscillators. Thelegally allowable tolerance for the 60 Hzfrequency is only ±0.02 Hz, or 0.033 %[17], but under normal operating condi-tions the actual tolerance is much tighter.

6. Requirements forTelecommunication Systems

Telecommunication networks make useof the stratum hierarchy for synchroniza-tion as defined in the ANSI T1.101 stan-dard. [18] This hierarchy classifiesclocks based on their frequency accuracy,which translates into time accuracy rela-tive to other clocks in the network. Thebest clocks, known as Stratum 1, aredefined as autonomous timing sourcesthat require no input from other clocks,other than perhaps a periodic calibra-tion. Stratum-1 clocks are normallyatomic oscillators or GPS disciplinedoscillators (GPSDOs), and have an accu-racy specification of 1#10-11. Clocks atstrata lower than level 1 require inputand adjustment from another networkTable 3. Time synchronization requirements for the electric power industry.

Table 2. FCC requirements for broadcast carrier frequency departure.

Broadcast ToleranceLow End of Band High End of Band

Carrier Uncertainty Carrier Uncertainty

AM radio ±20 Hz 530 kHz 3.8 # 10-5 1710 kHz 1.2 # 10-5

FM radio ±2000 Hz 88 MHz 2.3 # 10-5 108 MHz 1.9 # 10-5

Television ±1000 Hz 55.25 MHz(channel 2 video)

1.8 # 10-5 805.75 MHz(channel 69 audio)

1.2 # 10-6

International 0.0015 % 3 MHz 1.5 # 10-5 30 MHz 1.5 # 10-5

System Function MeasurementTimeRequirement

Generation Control Generator phase 10 ms

Event Recorders Time tagging of records 1 ms

Stability Controls Phase angle, ±1° 46 µs

Networked Controls Phase angle, ±0.1° 4.6 µs

Traveling wave fault locators 300 meter tower spacing 1 µs

Synchrophasor measurements Phase angle, ±0.022° 1 µs

MEASURE | 65Vol. 1 No. 3 • September 2006

REVIEW PAPERS

clock. The specifications for stratumlevels 1, 2, 3, and 3E are shown inTable 4. The “pull-in range” determineswhat type of input accuracy is requiredto synchronize the clock. For example, a“pull-in-range” of ±4#10-6, means thatthe clock can be synchronized by anotherclock with that level of accuracy.

6.1 Requirements for Telephones (land lines)

The North American T1 standard fortelecommunications consists of a digitaldata stream clocked at a frequency of1.544 MHz. This data stream is dividedinto 24 voice channels, each with 64 kHzof bandwidth. Each voice channel issampled 8000 times per second, or onceevery 125 µs. When a telephone connec-tion is established between two voicechannels originating from differentclocks, the time error needs to be lessthan one half of the sample period, or62.5 µs. Half the period is used to indi-cate the worst case, which exists whentwo clocks of the same stratum are

moving in opposite directions. If the timeerror exceeds 62.5 µs, a cycle slip occursresulting in loss of data, noise on the line,or in some cases, a dropped call. The useof Stratum-1 clocks throughout anetwork guarantees that cycle slips occuronly once every 72.3 days (62.5 µsdivided by 0.864 µs of time offset per

day). In contrast, Stratum-3 clocks couldproduce cycle slips as often as every169 s (Table 4), an unacceptable condi-tion. Thus if resources allow, the use ofStratum-1 clocks is certainly desirablefor network providers.

Table 4. Stratum timing requirements for clocks in telecommunication networks.

Figure 4. Phasor Measurement Systems receive time signals from the GPS satellites (courtesy of ABB, Inc.).

Stratum Levels Stratum-1 Stratum-2 Stratum-3E Stratum-3

Frequency accuracy, adjustment range 1 # 10-11 1.6 # 10-8 1 # 10-6 4.6 # 10-6

Frequency stability NA 1 # 10-10 1 # 10-8 3.7 # 10-7

Pull-in range NA 1.6 # 10-8 4.6 # 10-6 4.6 # 10-6

Time offset per day dueto frequency instability

0.864 µs 8.64 µs 864 µs 32 ms

Interval between cycle slips 72.3 days 7.2 days 104 minutes 169 s

66 | MEASURE www.ncsli.org

REVIEW PAPERS

6.2 Requirements for Mobile Telephones

Mobile telephone networks depend uponprecise time and frequency. Code divi-sion multiple access (CDMA) networkshave the most stringent requirements.

CDMA networks normally comply withthe TIA/EIA IS-95 standard [19] thatdefines base station time using GPS timeas a benchmark. Thus, nearly all CDMAbase stations contain GPSDOs (morethan 100,000 CDMA base stations are

equipped with GPS in North America).The time requirement is ±10 µs, even ifGPS is unavailable for up to 8 hours.During normal operation, base stationsare synchronized to within 1 µs. The fre-quency requirement is 5#10-8 for thetransmitter carrier frequency, but thecarrier is normally derived from the sameGPSDO as the time, and is usually muchbetter than the specification. Figure 5shows a cellular telephone tower con-taining a large variety of antennas.Several small GPS antennas near thebase of the tower are used to obtainCDMA time references (one antenna isshown in the inset).

Although not yet as popular as CDMAin the United States, the Global Systemfor Mobile Communications (GSM) isthe most popular standard for mobilephones in the world, currently used byover a billion people in more than 200countries. GSM is a time division multi-ple access (TDMA) technology thatworks by dividing a radio frequency intotime slots and then allocating slots tomultiple calls. Unlike CDMA, GSM hasno time synchronization requirementthat requires GPS performance, but theuncertainty requirement for the fre-quency source is 5#10-8, generallyrequiring a rubidium or a high qualityquartz oscillator to be installed at eachbase station. [20] Unlike CDMA sub-scribers, GSM subscribers won’t neces-sarily have the correct time-of-daydisplayed on their phones. The basestation clock is sometimes (but notalways) synchronized to the centraloffice master clock system.

6.3 Requirements for Wireless Networks

Although they operate at much higherfrequencies than those of the radio andtelevision stations discussed earlier, wire-less networks based on the IEEE802.11b and 802.11g have a similaracceptable tolerance for carrier fre-quency departure of ±2.5#10-5. Thespecifications call for the transmit fre-quency and the data clock to be derivedfrom the same reference oscillator. [21]

Figure 5. Cellular telephone towers contain a myriad of antennas, often including GPSantennas used to obtain a CDMA time reference.

GPS Antenna

MEASURE | 67Vol. 1 No. 3 • September 2006

7. Requirements for Calibration Laboratories

Calibration laboratories with an accred-ited capability in frequency usually main-tain either a rubidium, cesium, or aGPSDO as their primary frequency stan-dard. This frequency standard is used tocalibrate time base oscillators in testequipment such as counters and signalgenerators. The test equipment is gener-ally calibrated in accordance with manu-facturer's specifications, which typicallyrange from a few parts in 106 for lowpriced devices with non-temperaturecontrolled quartz oscillators to parts in1011 for devices with rubidium timebases. Therefore, a frequency standardwith an uncertainty of 1#10-12 allows alaboratory to calibrate nearly any pieceof commercial test equipment and stillmaintain a test uncertainty ratio thatexceeds 10:1. For these reasons, calibra-tion laboratories seldom have a fre-quency uncertainty requirement of lessthan 1#10-12. Laboratories that requiremonthly certification of their primaryfrequency standard can subscribe to theNIST Frequency Measurement andAnalysis Service (Fig. 6), and continu-ously measure their standard with anuncertainty of 2#10-13 at an averagingtime of one day. [22] Laboratories thatdo not need certification can often meeta 1#10-12 uncertainty requirement by using aGPSDO and a frequency measurementsystem with sufficient resolution.

7.1 Requirements for Voltage Measurements

The uncertainty in voltage measurementin a Josephson voltage standard (JVS) is

proportional to the uncertainty in fre-quency measurement. Typical high leveldirect comparisons of JVS systems at 10V are performed at uncertainties of a fewparts in 1011. Therefore, each laboratoryinvolved in a JVS comparison requires afrequency standard with an uncertaintyof 1#10-11 or less at an averaging time ofless than 10 minutes to ensure propervoltage measurement results. [23] Thisfrequency requirement is generally metby using either a cesium oscillator or aGPSDO. Figure 7 shows the NIST JVSsystem with a GPSDO located at the topof the equipment rack.

7.2 Requirements forLength Measurements

Since 1983, the meter has been definedas “the length of the path traveled bylight in a vacuum during a time intervalof 1 / 299 792 458 of a second.” Thus,the definition of length is dependentupon the prior definition of time interval,

and time and length metrology have aclose relationship. Until recently, the bestphysical realizations of the meter haduncertainties several orders of magnitudelarger than the uncertainty of the second,due to the techniques used to derive themeter. [24] However, the optical fre-quency standards [2] now being devel-oped at national metrology institutes canalso serve as laser wavelength standardsfor length metrology. As a result, theuncertainties of the best physical realiza-tions of the second and the meter willprobably track very closely in futureyears. [25]

7.3 Requirements for Flow Measurements

Flow metrology normally involves col-lecting a measured amount of gas orliquid in a tank or enclosure over a meas-ured time interval, which is known as thecollection time. Thus, uncertainties inthe measurement of the collection time

Figure 6. The measurement system sup-plied to subscribers to the NIST FrequencyMeasurement and Analysis Service. Itmakes frequency measurements traceableto the NIST standard by using GPS as atransfer standard.

Figure 7. Josephson voltage standards require a high performance frequency reference(courtesy of Yi-hua Tang, NIST).

REVIEW PAPERS

68 | MEASURE www.ncsli.org

REVIEW PAPERS

can contribute uncertainty to the flowmeasurement. However, they are gener-ally insignificant if they can be held, forexample, to a few tenths of a second overa 100 s interval. Nearly any commercialtime interval counters can exceed thisrequirement by at least two orders ofmagnitude, but most collection timeuncertainty is introduced by delay varia-tions in the signals used to start and stopthe counter. These delay variations needto be measured against a time interval ref-erence, and included in the uncertaintyanalysis of a flow measurement. [26]

8. Requirements forRadionavigation

Radionavigation systems, such as theground-based LORAN-C system and thesatellite based GPS system, have verydemanding time and frequency require-ments. The precise positioning uncer-tainty of these systems is entirelydependent upon precise time kept byatomic oscillators. In the case of GPS thesatellites carry on-board atomic oscilla-tors that receive clock corrections fromearth-based control stations just onceduring each orbit, or about every 12hours. The maximum acceptable contri-bution from the satellite clocks to thepositioning uncertainty is generallyassumed to be about 1 m. Since lighttravels at about 3#10-8 m/s, the 1 mrequirement is equivalent to about a3.3 ns ranging error. This means that thesatellite clocks have to be stable enoughto keep time (without the benefit of cor-rections) to within about 3.3 ns for about12 hours. That translates to a frequencystability specification near 6#10-14,which was the specified technicalrequirement during a recent GPS spaceclock procurement. [27]

9. Requirements for Remote Comparisons of the World’s Best Clocks

The current primary time and frequencystandard for the United States is thecesium fountain NIST-F1, with uncer-tainties that have dropped below 1#10-15

[1]. To determine that a clock is accurateto within 1#10-15 relative to anotherclock, the time transfer technique used tocompare the clocks needs to reach uncer-tainties lower than 1#10-15 in a reason-

ably short interval. NIST-F1 is routinelycompared to the world’s best clocksusing time transfer techniques thatinvolve either common-view measure-ments of the GPS satellites, or two-waytime transfer comparisons that require

the transmission and reception of signalsthrough geostationary satellites. Cur-rently, both the carrier-phase GPS andthe two-way time transfer techniques canreach uncertainties of about 2#10-15 atone day, reaching parts in 1016 after

Table 5. Summary of legal and technical time and frequency requirements.

Application or DeviceRequired Uncertainty

Time Frequency

Wristwatches 0.5 s per day 6 # 10-6

Parking Meters 7 minutes per hour 11.7%

Time Clocks and Recorders 1 minute per day 7 # 10-4

Taximeters 6 s per minute 10%

Field Standard Stopwatches 9 s per day 1 # 10-4

Musical Pitch NA 1 # 10-3

Tuning forks used for radar calibration NA 1 # 10-3

Stock Market time stamp3 s absoluteaccuracy

NA

AM Radio Carrier frequency NA 1.2 # 10-5

FM Radio Carrier frequency NA 1.9 # 10-5

TV Carrier Frequency NA 1.2 # 10-6

Shortwave Carrier Frequency NA 1.5 # 10-5

Color TV subcarrier NA 3 # 10-6

Electric Power Generation 10 ms NA

Electric Power Event Recorders 1 ms NA

Electric Power Stability Controls 46 µs NA

Electric Power Network Controls 4.6 µs NA

Electric Power Fault Locators 1 µs NA

Electric Power Synchrophasors 1 µs NA

Telecommunications, Stratum-1 clock NA 1 # 10-11

Telecommunications, Stratum-2 clock NA 1.6 # 10-8

Telecommunications, Stratum-3E clock NA 1 # 10-6

Telecommunications, Stratum-3 clock NA 4.6 # 10-6

Mobile Telephones, CDMA 10 µs 5 # 10-8

Mobile Telephones, GSM NA 5 # 10-8

Wireless Networks, 802.11g NA 2.5 # 10-5

Frequency Calibration Laboratories NA 1 # 10-12

Josephson Array Voltage Standard NA 1 # 10-11

GPS Space Clocks NA 6 # 10-14

State-of-the-art time transfer < 1 ns parts in 1016

MEASURE | 69Vol. 1 No. 3 • September 2006

REVIEW PAPERS

about 10 days of averaging. [28] Thereare practical limits to the length of thesecomparisons, because it often not possi-ble to continuously run NIST-F1 andcomparable standards for more than 30to 60 days. Although these time transferrequirements might seem staggeringlyhigh, keep in mind that the uncertaintiesof the world’s best clocks will continue toget smaller [2] and time transfer require-ments will become even more stringentin the coming years.

10. Summary and ConclusionAs we have seen, the world of time andfrequency metrology is extensive, sup-porting applications that range from theeveryday to the state-of-the-art. It haslegal and technical uncertainty require-ments that cover an astounding 15orders of magnitude, from the parts perhundred (percent) uncertainties requiredby coin operated timers, to the parts in1016 uncertainties required for remotecomparisons of the world’s best clocks.Table 5 summarizes the requirements forthe applications discussed in this paper(listed in the order that they appear inthe text).

11. References[1] T.P. Heavner, S.R. Jefferts, E.A. Donley,

J.H. Shirley, and T.E. Parker, “NIST-F1:recent improvements and accuracy eval-uations,” Metrologia, vol. 42, pp. 411-422, September 2005.

[2] S.A. Diddams, J.C. Bergquist, S.R. Jef-ferts, and C.W. Oates, “Standards ofTime and Frequency at the Outset of the21st Century,” Science, vol. 306, pp.1318-1324, November 19, 2004.

[3] T. Butcher, L. Crown, R. Suitor, J.Williams, editors, “Specifications, Toler-ances, and Other Technical Require-ments for Weighing and MeasuringDevices,” National Institute of Stan-dards and Technology Handbook 44,329 pages, December 2003.

[4] J.C. Gust, R.M. Graham, and M.A. Lom-bardi, “Stopwatch and Timer Calibra-tions,” National Institute of Standardsand Technology Special Publication 960-12, 60 pages, May 2004.

[5] State of Pennsylvania Code, 67 §105.71(2), (2005).

[6] Lynn Cavanagh, “A brief history of theestablishment of international standard

pitch a = 440 Hz,” WAM: Webzine aboutAudio and Music, 4 pages, 2000.

[7] International Organization for Standard-ization, “Acoustics – Standard tuningfrequency (Standard musical pitch),”ISO 16, 1975.

[8] M.A. Lombardi, “NIST Time and Fre-quency Services,” National Institute ofStandards and Technology Special Publi-cation 432, 80 pages, January 2002.

[9] U. S. Department of Transportation,“Speed-Measuring Device PerformanceSpecifications: Down the Road RadarModule,” DOT HS 809 812, 72 pages,June 2004.

[10] NASD, “OATS Reporting TechnicalSpecifications,” 281 pages, September12, 2005.

[11] J.H. Dellinger, “Reducing the Guess-work in Tuning,” Radio Broadcast, vol.3, pp. 241-245, December 1923.

[12] Code of Federal Regulations 47 §73.1545, (2004).

[13] Code of Federal Regulations 47 §73.682, (2004).

[14] U.S. – Canada Power System OutageTask Force, “Final report on the August14, 2003 blackout in the United Statesand Canada: Causes and Recommenda-tions” April 2004. Available at:www.nerc.com/~filez/blackout.html

[15] K.E. Martin, “Precise Timing in ElectricPower Systems,” Proceedings of the1993 IEEE International FrequencyControl Symposium, pp. 15-22, June1993.

[16] Power System Relaying Committee ofthe IEEE Power Engineering Society,“IEEE Standard for Synchrophasors forPower Systems,” IEEE Standard 1344-1995(R2001), 36 pages, December1995, reaffirmed March 2001.

[17] North American Electric ReliabilityCouncil, “Generation, Control, and Per-formance,” NERC Operating Manual,Policy 1, Version 2, October 2002.

[18] American National Standard forTelecommunications, “SynchronizationInterface Standards for Digital Net-works,” ANSI T1.101, 1999.

[19] “Mobile Station-Base Station Compati-bility Standard for Wideband SpreadSpectrum Cellular Systems,” TIA/EIAStandard 95-B, Arlington, VA: Telecom-munications Industry Association,March 1999.

[20] European Telecommunications Stan-

dards Institute (ETSI), "GSM: Digitalcellular telecommunication system(Phase 2+); Radio subsystem synchro-nization (GSM 05.10 version 8.4.0),”ETSI TS 100 912, 1999.

[21] LAN/MAN Standards Committee of theIEEE Computer Society, “IEEE Standardfor Information technology – Telecom-munications and information exchangebetween systems – Local and metropol-itan area networks – Specific require-ments Part 11: Wireless LAN MediumAccess Control (MAC) and PhysicalLayer (PHY) specifications – Amend-ment 4: Further Higher Data RateExtension in the 2.4 GHz Band,” IEEEStandard 802.11g, 2003.

[22] M.A. Lombardi, “Remote frequency cal-ibrations: The NIST frequency measure-ment and analysis service,” NationalInstitute of Standards and TechnologySpecial Publication 250-29, 90 pages,June 2004.

[23] Y. Tang, M.A. Lombardi, D.A. Howe,“Frequency uncertainty analysis forJosephson voltage standard”, Proceed-ings of the 2004 IEEE Conference onPrecision Electromagnetic Measure-ments, pp. 338-339, June 2004.

[24] B.W. Petley, “Time and Frequency inFundamental Metrology,” Proceedings ofthe IEEE, vol. 79, no. 7, pp. 1070-1076,July 1991.

[25] J. Helmcke, “Realization of the metre byfrequency-stabilized lasers,” Measure-ment Science and Technology, vol. 14,pp. 1187-1199, July 2003.

[26] J.D. Wright, A.N. Johnson, M.R.Moldover, and G.M. Kline, “GasFlowmeter Calibrations with the 34 Land 677 L PVTt Standards,” NationalInstitute of Standards and TechnologySpecial Publication 250-63, 72 pages,January 2004.

[27] T. Dass, G. Freed, J. Petzinger, J. Rajan,T.J. Lynch, and J. Vaccaro, “GPS Clocksin Space: Current Performance andPlans for the Future,” Proceedings of the2002 Precise Time and Time IntervalMeeting, pp. 175-192, December 2002.

[28] T.E. Parker. S.R. Jefferts, T.P. Heavner,and E.A. Donley, “Operation of theNIST-F1 cesium fountain primary fre-quency standard with a maser ensemble,including the impact of frequency trans-fer noise,” Metrologia, vol. 42, pp. 423-430, September 2005.