1

ANALYSIS OF TERMINAL SEPARATION STANDARDS AND RADARPERFORMANCE

Dr. Steven D. Thompson, MIT Lincoln Laboratory, Lexington, MA

Dr. Steven R. Bussolari, MIT Lincoln Laboratory, Lexington, MA

AbstractThe United States is in the process of

commissioning new consolidated Terminal RadarApproach Control (TRACON) facilities to serve largecomplex terminal areas. In addition to receivingsurveillance data from radars formerly serving eachindividual TRACON, the consolidated TRACONfacilities will receive surveillance data from all availablesensors in the area including long-range en route radars.The new facilities will have the ability to use mosaicdisplays to combine surveillance data from multiplesurveillance sensors.

Current terminal area separation requirements allowthree-mile separation between aircraft provided bothaircraft are within 40 miles of a single terminal radar. Atissue is whether three-mile separation can be maintainedthroughout consolidated TRACON airspace using amosaic display.

This paper reviews the error characteristics of long-range and short-range sliding-window Air TrafficControl Radar Beacon System (ATCRBS) andMonopulse Secondary Surveillance Radar (MSSR)surveillance sensors. Errors in the separation distancebetween targets displayed to a controller were analyzedfor both single sensor and mosaic displays. Monte Carlosimulations were run to compute the errors in displayedseparation as a function of range from the sensor.

It was found that MSSR sensors offer anapproximately three-fold increase in azimuth accuracyover sliding-window ATCRBS and provides equivalentseparation performance at a range of over 100 miles insingle sensor mode. It was also found that MSSRsensors in a mosaic display offered separationperformance equivalent to a single sliding-windowATCRBS sensor when each aircraft is within 40 miles ofits respective sensor. Sliding-window ATCRBS short-range sensors in mosaic display mode offer acceptableperformance when each aircraft is within 28 miles of itsrespective sensor.

This work is sponsored by the Federal Aviation Administration underAir Force Contract No. F19628-00-C-0002. Opinions, interpretations,conclusions, and recommendations are those of the authors and are notnecessarily endorsed by the U.S. Government.

IntroductionThe United States is in the process of

commissioning new consolidated Terminal RadarApproach Controls (TRACON) facilities to serve largecomplex terminal areas. Southern California TRACONis now operational, Northern California TRACON andthe Potomac TRACON are in the transition stage. Otherconsolidated TRACONs are planned in the future.Consolidated TRACONs replace the individualTRACONs serving the area, each formerly relying on asingle terminal radar to serve a single major airport andsurrounding satellite airports. Consolidated TRACONsoffer operational advantages in complex airspaceincluding increased surveillance coverage and thecapability to optimize airspace and operations within thearea covered by the facility. This promises moreefficient control of aircraft during transition to and fromairports.

In addition to receiving surveillance data fromradars formerly serving the individual TRACONs, theconsolidated TRACONs will receive surveillance datafrom all available sensors in the area including long-range en route radars.

The new facilities will have the capability ofmosaic display of multiple surveillance sensors. Inmosaic display mode, the airspace is divided into equallysized square or rectangular “sort boxes.” Each sort boxrepresents a geographic area and is assigned a preferredsensor, a supplemental sensor, and tertiary sensors (ifavailable). The preferred sensor is used to providesurveillance data as long as a satisfactory track of thetarget is maintained. In mosaic mode, different sensorsmay track aircraft being separated by a controller.

Problem Statement and ApproachAn Individual TRACON controlling traffic in it’s

respective terminal area separates traffic by threenautical miles1 for aircraft less than 1000 feet apart inaltitude. At issue is whether or not the current three-mile_________________________________________1 In this paper, “miles” refers to nautical miles.

2

terminal separation standard can be applied throughoutall of the airspace in a consolidated TRACON. Theability to use three-mile separation throughout theconsolidated TRACON airspace will enable much moreflexibility in airspace design and utilization. For thisreason, the FAA Operational Evolution Plan hasidentified the expansion of the use of the terminalseparation standard as one of its key goals.

Radar separation standards are contained in FAAOrder 7110.65N [1]. The order allows three-mileseparation between aircraft as long as both aircraft areless than 40 miles from and tracked by the same sensorantenna, otherwise the traffic must be separated by fivemiles. A separation of three miles is not permitted with amosaic display; five-mile separation is required. Theorder makes no distinction in separation requirementsbased on the performance of the radar, and appliesequally to short and long-range radars. In order todetermine if FAA Order 7110.65N could be modified topermit three-mile separation in the mosaic environmentof a consolidated TRACON, an analysis was undertakenof the surveillance errors associated with single-sensorand mosaic displays. The objective of this analysis wasto determined the conditions under which a mosaicdisplay would be equivalent to a single sensor display forthe purposes of traffic separation.

The approach taken in this analysis is to examinethe error characteristics of the various types of beaconsensors in the FAA inventory and to analyze theirperformance with regard to providing accurate separationmeasurements to controllers. The metric forperformance is the error in the aircraft separationdisplayed to a controller. The error in measured ordisplayed separation is the difference between the trueseparation (actual distance between aircraft in the air)and the separation displayed to a controller on a radardisplay. The required surveillance performance isderived using this metric for current legacy sensors. Thisrequirement is used to determine the range for providingacceptable separation performance for newer moreadvanced sensor systems in both single sensor andmosaic display modes.

Surveillance AnalysisThe surveillance analysis begins with a summary of

the error characteristics of the secondary sensorscurrently used by the FAA to provide separation. Theadditional errors in displayed separation introduced bymosaic display are then discussed. This is followed by adescription of the Monte Carlo simulation that wasperformed to measure the errors in separation displayedto the controller for each of the sensor cases analyzed.

Error Characteristics of FAA SensorsSecondary (beacon) radar error characteristics

include errors in estimating range and errors inestimating azimuth to the target.

Range measurement errors are due primarily toerrors in measuring the interval between the time a pulseis sent from the radar to the time a reply is received fromthe aircraft’s transponder. This includes errors in theaccuracy with which the sensor can measure the timeinterval and variations in the allowed turn around time ofthe transponder. Range errors due to timing arerelatively small (< 200 feet) and do not increase withrange. Refraction effects are only significant at verylong range and were not included in this analysis.Propagation anomalies, such as atmospheric ducting,were also not included in this analysis. Likewise, errorsintroduced by aircraft not equipped to report altitudewere not considered.

Azimuth measurement errors are primarily due toerrors in estimating the target position within thebeamwidth of the transmitted pulse. Typical 3-dBbeamwidths are approximately 2 1

2°, which results in a

beamwidth of 1.75 miles at a range of 40 miles.Azimuth measurement errors are less than the beamwidthand depend on the technique used to estimate the target’sposition within the beamwidth. There are two azimuthmeasurement techniques used by sensors in the FAAinventory.

The older generation of Air Traffic Control RadarBeacon System (ATCRBS) radars use a “slidingwindow” technique to estimate azimuth. This technique(illustrated in Figure 1) requires detection of replies inthe leading and trailing edges of the beam where thesignal is weakest. The azimuth of the target is estimatedas the center of the leading and trailing edges. FAABeacon Interrogator BI-4 and BI-5 sensors use thesliding window technique.

Figure 1 “Sliding window” azimuth measurementtechnique

3

Hereafter, this paper will refer to sliding-windowATCRBS sensors as simply ATCRBS sensors.

Newer Monopulse Secondary Surveillance Radar(MSSR) sensors use two beam patterns for interrogationand require only a single reply. This technique(illustrated in Figure 2) offers an approximately three-fold improvement in azimuth measurement accuracyover the sliding window technique. FAA Mode S andBI-6 sensors use the monopulse technique for measuringazimuth.

Figure 2. Monopulse azimuth measurement technique

A detailed description of these two azimuthestimation techniques is given by Orlando [2]. Figures 1and 2 were taken from that article.

Additional errors include residual registration errorscaused by location and azimuth biases not removed byalgorithms designed to align multiple sensors.

The position estimates are disseminated using theFAA Common Digitizer (CD) format. In this analysisthe format resolution was not modeled as an additionalerror source but the position estimates were rounded tothe allowed CD formats.

The scan time of the antenna determines the lengthof time between target updates. While the positionestimates are not affected, the motion of the targetsbetween their respective updates results in errors indisplayed separation.

The radar source errors used in the analysis arepresented in Table 1. The values of the errors used arebased on radar specifications and field data fromARCON [3] for radars in Southern California TRACON,and from MIT Lincoln Laboratory [4] for radars in thenortheast region.

Table 1 Radar Source Errors Used in Simulations

Sensor Error Sources

MSSR1 ATCRBS sliding windowShort Range Long Range Short Range Long Range

LocationBias

200 ft. (0.033 nmi.) Uniform in any directions = 115 ft. (0.019 nmi.)Registration

Errors AzimuthBias

±0.3° Uniforms = 0.173°

RadarBias

±30 ft. (0.005 nmi) Uniforms = 17 ft. (0.003 nmi.)

Range ErrorsRadarJitter

25 feet rmss = 25 ft. (0.004 nmi.)

AzimuthError

AzimuthJitter s = 0.068° (0.8 ACP)3 s = 0.230° (2.6 ACP)3

Range1/64 nmi.s = 27 ft.

(0.005 nmi.)

1/16 nmi.s = 110 ft.

(0.018 nmi.)

1/64 nmi.s = 27 ft.

(0.005 nmi.)

1/16 nmi.s = 110 ft.

(0.018 nmi.)

DataDisseminationQuantizationCD format Azimuth 360°/4096

s = 0.025°Scan Time

UncorrelatedError2

4-5 sec.s = 219 ft.(.036 nmi.)

10-12 sec.s = 536 ft.(.088 nmi.)

4-5 sec.s = 219 ft.(.036 nmi.)

10-12 sec.s = 536 ft.(.088 nmi.)

1Note: MSSR can deal with both Mode S and ATCRBS transponders in a monopulse fashion.2Note: Aircraft assumed three miles apart in trail at 200 knots.3Note: ACP=Azimuth Change Pulse (1/4096 of a scan)

Transponder Error Sources

Mode S ATCRBSRange Error ± 125 ft. (0.021 nmi.)

s = 72 ft. (0.012 nmi.)± 250 ft. (0.041 nmi.)s = 144 ft. (0.024 nmi.)

4

The errors for individual radars are in goodagreement with the errors for radars reported in a studyconducted by Lockheed Martin and included as anAppendix in the ARCON report.

Additional Errors in Displayed SeparationIntroduced by Mosaic Display

The target position accuracy of an individual radaris unaffected by whether it is used as a single sensor or ina mosaic display. However, two factors add to the errorsin the separation between targets displayed to acontroller in mosaic mode compared to single sensormode.

First, for a single sensor tracking two or moreaircraft, the bias errors in range, azimuth, and sensorlocation affect the position estimates of all targets in thesame way. That is, the errors are correlated. Thus eventhough the position of the targets may be in error, theerrors are equally applied to all targets and have verylittle affect on the separation distance displayed to acontroller on a radar display. However, in mosaic mode,when two different radars are used to track the targets,those bias errors are, in general, uncorrelated and willresult in additional errors in the separation betweentargets displayed to the controller.

Secondly, when a controller is separating twoaircraft using the estimated positions on a display, thetargets are updated at different times. This introduces anerror in the displayed separation because of the motion ofone aircraft relative to the other between updates. With asingle sensor, for two target aircraft relatively near eachother, the time between updates can be explicitlycomputed and is generally small. However, in the caseof mosaic display, the target updates are asynchronousand, in general, the time difference between targetupdates is larger for mosaic display. This in turn resultsin larger errors in displayed separation for the mosaiccase.

Statistical Analysis of Errors in SeparationDisplayed to a Controller

The procedure followed in this analysis was todetermine the distribution of errors in measuredseparation observed, on average, for two aircraft thatwere three nautical miles apart in-trail traveling at 200knots. Monte-Carlo simulations using the sensor errorcharacteristics described above were run for singlesensor and mosaic display for long-range and short-rangeATCRBS and MSSR sensors at various ranges(1,000,000 trials each).

All of the characteristic radar errors wereindependently re-sampled for each trial. Separationmeasurement errors are highly dependent on range andthe relative geometry of the aircraft and the radars. Theprocedure used for the single sensor case was torandomly orient two in-trail aircraft that are separated bythree miles about an angle j relative to the radar, withthe midpoint of their locations at the specified rangefrom the radar. This is illustrated in Figure 3.

Mosaic

RandomOrientation

Single Sensor

Range toMidpoint

j

3 nmi

j

Rangeto Midpoint

RangeTo Midpoint

RadarSort BoxBoundary

Figure 3. Relative geometries used for the single andmosaic (two-radar) measurements of errors in

separation measurements

In the mosaic case, with two radars, each trackingthe aircraft nearest the sensor, the midpoint of theseparation of the aircraft was kept at a constant range forboth sensors as illustrated in Figure 3. Note that becauseof the random orientation of the aircraft about j ,theresults for the mosaic two-radar case are valid for anyrelative angle between the aircraft and radars.

The surveillance errors were sampled according tothe error distributions described in Table 1 to produceestimated positions for each aircraft. This procedure isillustrated in Figure 4. The estimated positions werethen reported in the CD format. The individual positionreports for each aircraft were used to compute estimatedseparation distances. The time between target updateswas computed explicitly for the single sensor case and

5

randomly sampled according to the relative radar updaterates for the mosaic display case. The additional error indisplayed separation caused by movement of the aircraftbetween updates was added to compute the separationdistance displayed to the controller.

Monte Carlo simulations of 1,000,000 trials wererun for single-sensor and mosaic display for short-rangeand long-range ATCRBS and MSSR sensors at variousranges. Each simulation of 1,000,000 trials generated adistribution of separations displayed to the controller.This is the probability density function of displayedseparation illustrated in Figure 4. The mean of thedistribution was three miles, the actual separation, foreach case. The shape of the distribution depended on theerror characteristics for the particular sensor beingsimulated. The standard deviation, s , for eachdistribution was computed.

Figure 4. Procedure for generating errordistributions for displayed separation

Simulations were run for the eight combinationsrepresented by single-sensor and mosaic display, short-range and long-range sensors, and ATCRBS and MSSRsensors. All eight cases were run at various ranges.

As an illustration, results for single-sensorATCRBS, single-sensor MSSR, and mosaic displayMSSR short-range sensors are compared at four differentranges in Figure 5. Note that the shapes of thedistributions as well as their standard deviations aredifferent and expand with range. The single-sensorMSSR has the lowest average error and the single-sensorATCRBS and mosaic MSSR have similar errorsalthough the distribution shapes are different.

It is important to note that these distributions arenot Gaussian. The assumption that they are Gaussianand the use of the Gaussian standard deviation s to drawconclusions regarding the percentages of trials in the tailsof the distribution would lead to serious errors. Anexample computation illustrates the point. In a Gaussiandistribution, 0.26% of the trials will fall outside the 3slimits in the tails of the distribution. This is not the casefor the distributions shown in Figure 5. For the mosaicshort-range MSSR distribution at 30 miles 0.32% of thetrials were outside the 3s limit. For the ATCRBS short-range single-sensor distribution, 0.94% of the trials wereoutside that distribution’s 3s limit. For the MSSR short-range single-sensor distribution it was 0.44%.

Figure 5. Probability distributions of displayedseparation errors as a function of range for ATCRBS,

MSSR, and MSSR mosaic display

6

ResultsThe results of the Monte Carlo simulation runs are

presented as plots of displayed separation error (3s)versus range for single sensor and mosaic displayATCRBS and MSSR sensors. The results for short-range radars are presented in Figure 6 and the results forlong-range radars are presented in Figure 7. The lineindicating currently acceptable performance (i.e.,consistent with FAA Order 7110.65N) is derived fromthe separation performance of a single short-rangeATCRBS sensor at a range of 40 miles.

It is apparent from Figure 6 and 7 that MSSRperformance is significantly better than ATCRBS,especially at long range. This is because of the three-fold increase in azimuth accuracy for MSSR.

For single sensor ATCRBS and MSSR sensors theseparation performance of the long-range sensors isalmost equivalent to the performance of the short-rangesensors. This is because the slower rotation rate of theantenna does not contribute significant error whenaircraft are relatively near each other and are undersurveillance by the same sensor. At long ranges, thetime difference between target updates does not go upsignificantly at the slower rotation rates because theazimuth angle between the targets is so small and themotion of the aircraft between updates is small. Atcloser ranges, this error does increase for long-rangeradars relative to short-range radars because the relativeazimuth angle of the targets is greater, but theperformance is better than the currently acceptable levelat close range. The only other difference for long-rangeradars is the lower resolution in the CD reporting formatfor range and this does not appear to make much of acontribution to the separation performance.

Comparing Figure 6 and Figure 7, the separationperformance of long-range mosaic ATCRBS and MSSRsensors is significantly degraded over short-range mosaicsensors. This is because in the mosaic mode the targetupdates are asynchronous and the slower rotation rates ofthe long-range sensors now make large errorcontributions compared to the single-sensor surveillancebecause the aircraft can move a significant distancebetween updates.

Figure 6 shows that the short-range mosaic MSSRperformance is better than the single-sensor ATCRBSperformance at ranges greater than about 32 miles. Thisis because at longer ranges the increased azimuthaccuracy of MSSR begins to dominate over the errors indisplayed separation introduced by mosaic display. Atcloser ranges, mosaic MSSR performance is not as goodas single sensor ATCRBS but is still better than thecurrently acceptable performance of single sensorATCBRS at 40 miles.

SHORT RANGE

Mosaic ATCRBS

Mosaic MSSR

Single Sensor ATCRBS

Single Sensor MSSR

11010090807060504030201000

0.2

0.4

0.6

0.8

1.0

1.2

Range to Center of Two Aircraft Separated by Three Miles (nmi)

3s

Dis

pla

ye

d S

ep

ara

tio

n E

rro

r (n

mi)

Current Acceptable Performance

Figure 6. Displayed separation error (3s) as afunction of range for single sensor and mosaic short-

range ATCRBS and MSSR radars

LONG RANGE

Mosaic ATCRBS

Mosaic MSSR

Single SensorATCRBS

Single Sensor MSSR

11010090807060504030201000

0.2

0.4

0.6

0.8

1.0

1.2

Range to Center of Two Aircraft Separated by Three Miles (nmi)

3s

Dis

pla

ye

d S

ep

ara

tio

n E

rro

r (n

mi)

Current Acceptable Performance

Figure 7. Displayed separation error (3s) as afunction of range for single sensor and mosaic long-

range ATCRBS and MSSR radars

7

ConclusionsBased on the results of the analysis presented above

and the currently acceptable separation performance ofsingle sensor ATCRBS sensors at a range of 40 miles,the following conclusions are drawn:

• MSSR sensors (both short-range and long-range) offer acceptable separationperformance for three-mile separation outto a range of at least 100 miles for a singlesensor.

• MSSR short-range sensors in mosaicdisplay mode offer acceptable separationperformance when each aircraft is within40 miles of its respective sensor.

• ATCRBS short-range sensors in mosaicdisplay mode offer acceptable separationperformance when each aircraft is within28 miles of its respective sensor

• Long-range ATCRBS and MSSR sensorupdate rates are not sufficient to supportthree-mile separation with mosaic displayat any range

In summary, a technique has been developed toderive surveillance requirements for three-mileseparation in the extended terminal environmentusing existing standards and sensors as a baseline.The analysis reported here supports the significantexpansion of the use of three-mile separationstandards with existing MSSR sensors and datadissemination formats in both single sensor andmosaic environments. The choice of single sensorvs. mosaic for a particular facility will depend uponthe desired airspace design. Further expansion maybe possible using fusion trackers that reduce themosaic asynchronous update errors, but additionalanalysis is required to estimate the potentialbenefits of this technique.

References[1] Federal Aviation Administration, Air Traffic Control,Order 7110.65N, February 21, 2002, Department ofTransportation, Washington D.C. Available on theInternet at http://www.faa.gov/atpubs

[2] Orlando, Vincent A., The Mode S Beacon RadarSystem, The Lincoln Laboratory Journal, Fall 1989,Volume 2, Number 3, Lexington MA.

[3] Mosaic Display Target Accuracy Study, Final Report,for the Federal Aviation Administration, Under ContractDTFA01-97-C-00046, ARCON Corporation, Waltham,MA. December 13,1999

[4] Gertz, J.L., and Kaminsky, Fusion TrackerEvaluation, Project Report, ATC-302, prepared for theFederal Aviation Administration, MIT LincolnLaboratory, February 15,2002.

Key WordsTerminal Separation

Terminal Surveillance

Consolidated TRACON

Separation Standards

Radar Performance

BiographiesDr. Steven D. Thompson

Steven Thompson is a member of the TechnicalStaff of the Air Traffic Control Systems Group at MITLincoln Laboratory. He has worked in air traffic controlfor the past fifteen years, and on surveillanceperformance requirements for the last several years.Prior to working at MIT Lincoln Laboratory, Dr.Thompson was at Bell Laboratories in Holmdel N.J. andArgonne National Laboratory in Chicago, Illinois. Heholds a Bachelor’s Degree in Aerospace Engineering anda Master’s and Ph.D. in Nuclear Engineering, all fromGeorgia Tech. He also holds an MBA from theUniversity of Chicago. He is an active pilot with anAirline Transport Pilot’s Certificate with a type rating inthe Citation jet. He also holds a Certified FlightInstructor’s Certificate with single and multi-engineairplane, instrument, and glider ratings.

Dr. Steven R. Bussolari

Steven R. Bussolari is Leader of the Air TrafficControl Systems Group at MIT Lincoln Laboratory. Hereceived a B.S. in Mechanical Engineering from UnionCollege and S.M. and Ph.D. in Mechanical Engineeringfrom MIT. Previously, he was on the faculty in theDepartment of Aeronautics and Astronautics in the Man-Vehicle Laboratory at MIT, where he was PrincipalInvestigator on NASA research grants in the area ofaerospace human factors as applied to flight simulation,cockpit automation, and human performance inweightlessness. He has served as a member of the LifeSciences and Flight Simulation Technical Committees ofthe American Institute of Aeronautics and Astronautics.Dr. Bussolari currently holds a commercial pilot's licensewith single and multi-engineer airplane, instrument andglider ratings.