ABSTRACT The Millstone operated Hill incoherent since 1763 to perform densities, and electron have been conducted digital This report Fi -regions). from These to permit which computer changes By operating include the echo autocorrelation and performing of the reflected coherently, below 200 km (i.e., signals. spaced pairs of echo it has also been possible clutter signals Samples presented of the results from timing of pulses can be determined pairs ex- the E- and to the radar of close-spaced function ( stronger) are electron The se measurements modifications appropriately tbe unwanted these programs. has been made to the system in $969 to permit to altitudes changes the radar system long pulses analysis the transmission by processing radar study of F2 -region single and spectrum of the measurements equipment a synoptic by transmitting describes scatter and ion temperatures. sweep integration tensions (Thomson) in the samples. to subtract tbe ionospheric echoes. that can now be obtained from CONTENTS iii Abstract Svbols vi and Abbreviations 4 1. INTRODUCTION 11. OPERATING A. 5 MODES 5 General 5. 33. Mode E 5 C. Mode F D. Mode G E. Mode H F. Model G. Observing H. Mode Modifications io 10 io m. Iv *3 program ii 24 EQUIPMENT Zi A. General B. Frequency C. Receiver D. Data Sampling E. Timing Equipment 1. 2. 3. 4. 5. 6. 7. 8. B. c. 26 29 PROCEDURE Timing 29 Control Computer Computer Gene ral Computer comPuter Computer E F G H I Output @tput -Modes Output –Mode Output –Modes E andl F G and H iv 29 29 36 36 37 37 37 39 39 Operations General computer OperatiOns– MOde compute rOperatiOns– MOde Compute r Operations-Mode Computer Operations –MOde Computer Operations -MOde Real-Time 4. 2. 3. 4. 25 Transmitter Control –Mode E Transmitter Control –Mode F Transmitter Control–Mode G Transmitter Control-Mode H Transmitter Control–Mode 1 Transmitter Modulator Control Receiver Control Data-Sampling ContrOl Real-Time i. 2. 3. 4. 5. 6. ?5 Equipment OPERATING A. 2+ Control 39 4t 44 46 47 47 48 48 48 49 49 V. ANALYSIS 53 A. General 53 B. Power-Profile 1. 2. 3. C. 2: 7. 8. D. E. 58 Analysis General Phase Analysis Magnitude Analysis – Library Search Ion-Composition Problem Collisional Region – h < i i 8 km Temperature Equality Region – 148< h < i30 km Ion-Composition Transition Region – i 30< h < 225 km (J+ Region – 225 km < h Neutral-Atmosphere Profile 1. 2. 3. 4. 53 55 56 Mode E – Routine Analysis Mode E – RASEM Analysis Mode 1 Correlation-Function 1. 2. 3. 4. 53 .inalysis Deductions and Ion-Composition 65 General Tn, [0] Profile Ion-Composition Profile Electron-Density-Profile Debye-Length 5* 58 59 6i 63 63 65 65 Corrections 65 65 66 67 68 Corrections Acknowledgments 69 References 70 APPENDIX A - Systematic Distortions Measurements APPENDIX B – Statistical Errors Measurements in Correlation-Function in Correlation-Function v 79 SYMBOLS I 1 ABBREVIATIONS A* complex A*B convolution JAI magnitude [A] concentration [A] ~ [A] acf autocorrelation c speed of light in free f frequency fo conjugate radar Im (A) imaginary k Boltzmann’s ——-. . density) space above the earth’s part of complex number constant mass electron Re (A) real part of complex S/N signal-to-noise T pulse length Te electron Ti ion temperature Tn neutral T. exospheric density number ratio temperature temperature temperature Tn at h = i20 krn Vd drift velocity Ah height resolution wavelength ion -neutral collision correlation lag vi 1 density. frequency Ne v in A function ion mass radar A x in kilometers mi Tizo number (number (height) electron number of A and B center altitude of complex of complex at altitude h ‘e , AND frequency A surface A Of gas species A INCOHERENT SCATTER DENSITY, MEASUREMENTS TEMPERATURES, OF AND AT E- AND COLLISION MILLSTONE F-REGION FREQUENCY HILL 1. 1NTRODUCTION This report the Millstone (Thomson) menced is intended to complement Hill, West ford, scatter radar system tron and ion temperatures.z listed Lincoln in Table These power the measurements studies directed of F-region incoherent These electron studies density com- and elec - 1963 through i970 have been published as in a number of journal articles, as measurements bythe pulse length Adequate long pulses together The height resolution T according height resolution T to a suitable Te and Ti requires value. with echo- Ah afforded by to can be achieved However, a spectrum on the pulse length that may be used. is approximately using single computer. (i) by reducing and ion temperatures in a digital cT/2 c is the speed of light. constraints during as well have been conducted performed is determined Ah= ) Reports the manner in which vertically of the F2-region. measurements obtained Technical that described 74.5”W) I. F2-region sweep integration where synoptic The results Laboratory onei (42.6oN, is operated’ for studies in 1963 and have provided in earlier an earlier Massachusetts for electron-density the determination analysis of the signals, The overall doppler of the electron and this places broadening of the signals 3Afi (Ref. 9), where (2) is the doppler speed, shift produced k is Boltzmannls of the information constant, on the frequency width of tbe pulse (-l/T) broadening. Thus, As discussed spectrum in Ref.9, this restriction long-pulse this useful range is approximately and ion temperatures. to extend the measurements length (thereby increasing fully 23-cm-wavelength possible to achievea tude range 130< on the transition the F-region, the papers Afi) listed andlor radar height resolution between in Table to recover dOppler operated the molecular drifts length limits adequate For the 68-cm-wavelength 250< vertically to lower altitudes, one may reduce oblique with the beam directed Using this radar, incidence. obliquely, information NO+) in the E-regiOn in the E- and F-regions; the waveUsinga it has heen to cover thealti- has been obtained and atOmic iOns (0+) in these results are published II. .. for directed h< 1000 km depending upon ionospheric of 35 km and extend tbe measurements ions (0~, the range height resolution , . .. ...-—= most that the spectral Of the overall use fulpulse provide make the measurementsat h< 260 km approximately. and on horizontal to one-sixth on the minimum observations radar steerable In order it is necessary (3) of electron In order at its mean thermal pulse length is measurements system, wavelength. of the returns, to about one-fifth acceptable a radar . 1 over which single conditions. i and h is the radar be restricted the minimum Tmin~2/Af. of altitudes by an ion of mass mi approaching . in 1TABLE I PUBLICATIONS CONCERNING THE MILLSTONE HILL UHF (68-cm WAVELENGTH) THOMSON SCATTER RESULTS Year lv!entln Covered March, 1963 July, December, August, April, Fublicotion September July, November February 1963 to January 1964 Ref. 3 Ref. 4 Technical Repoti 374, Lincoln lmbcdc.ry, (22 Jonuary 1965), M. I.T. DDC AD-616607 Apil, 1964 July, November January fhrwgh December Ref. 5 Technical Relwt 430, Lincoln Laboratory, M.I. T. (15 November 1967), DDC AD-668.436 January, 1965 April, Jmwarytkwgh August Ref. 6 December Technical Report 474, Lincoln Labcr.tory, M.I. (8 December 1969), T. DDC AD-707501 1964 January through December Ref. 7; Technical Lincol. Report 481, Laboratory, M. I.T. (19 January 1971 ), DDC AD-725742 1967 January through December Ref. 7; Technical Reporf 482, Lincoln Laboratory, (22 July 1971), M.I. T. DDC AD-735727 1968 January through December Technical Report 499, Lincoln Laboratory, (23 Ja”wy 1973), M. I.T. DDC AD-767251/2 1969 Jwwmythrough December Ref. 8; Technical Report 513, Lincoln Laboratory, M.I. (23 July 1974), T. DDC AD-AO+38505/O 1970 January thrwgh December Technical Report 522, Lincoln Laboratory, M. I.T. (11 I%y ! , 1976) TABLE II PUBLICATIONS CONCERNING THE MILLSTONE HILL L-BAND (23-cm WAVELENGTH) THOMSON SCATTER RESULTS Perbd Covered Apri 1- November October Unfortunately, incidence time. herent scatter limitations spacing resolution ber 1975. Ref. 13 of the L -band radar measurements, system by an interval located adequate ~ and measuring Peru, is routinely resolution between employed and at Arecibo, can be set via Eq. (1) to any desired The spectrum employing by transmitting the correlation This approach at Jicamarca, for oblique- only during the day- on the E- and Fi -regions As outlined in Ref. 9, this may be achieved stations Puerto the pairs of pairs of at the incoRico? Using value subject only to is set independently by the the pulses. technique program is barely and useful data can be gathered has now been implemented at altitudes before. as it etisted hardware This report from at Millstone (addition Hill, describes the Millstone in November of a hard-wired program. permitting measure- about 90 and 600 km with far better its implementation modification to tbe 2-pulse between These Hill 2 -pulse inco- 4969 through Decemin January correlator) modifications height 1976 will be the subject of report. II of this report radar equipment All Z-pulse the various and computer to the data to derive operations are available plots of tbe various to the data o“ a routine basis. 3 modes performed physical on request physical of the 2 -pulse In Sec. 111, the modifications in the 2-pulse timing results modes of operation measurement. needed for operation applied analysis contour not been applied describes for ionospheric equipment procedures for producing has 1969 and temperature and their purpose6 analysis September some changes Section describes 1968- by SIN considerations. A major necessitated 1-pulse November than could be achieved scatter a future 12 between This 2-pulse ments of density herent Ref. the height resolution imposed maximum 11 1969 T separated radar Ref. September taken with the same separation. this scheme, 10 1968- radar. of length echo samples 1968 we have sought to obtain information vertical-incidence pulses M.y Ref. September scatter Therefore, 1964 1965- the sensitivity Thomson P.bl icotion are presented to outside users. parameters to the original are discussed. upon the received results (altitude-time experiments Section signals. IV The in Sec. V. A program contours), exists but it 11. OPERATING A. MODES GENERAL The goal of the 2-pulse ionosphere, where heighi gradients those encountered H, and I– experiments at greater range ments, and other characteristics Modes profiles. experimental and scale heights each designed purposes. TabIe for these modes, in the bottomside smaller compared to probe 111lists while the altitude Figs. i(a) this table and the figures Modes efficiently with E, F, G, a specific ranges, through (c) give are referenced measure- more detailed frequently in sections. E and I are designed Modes frequencies, F, to study electron G, and Hare and composition; tude of the acf yields the phase provides height resolution designed height resolution A number of experiments-designated on the acf measurements; the following larger for this purpose, altitude information for specific become heights. have been devised is to gain improved deeigned hence, comparable tudes where collisions clude a first zero-crossing collision drift information. to extend outtoa.t to the local least smear temperatures, the acf of the scattered frequency, ionospheric scale height. function). of the function, Theacf of the function from scattered-power motions, power. and composition zero- crossing the correlation and minimum they measure Each mode bas been designed thesecond severely hence, to study ionospheric they measure the temperature, plasma density; collision The magni- information while to probe with a measurements (except are atloweralti- This is needed in order which considerable to in- information is derived. All measurements ground-cIutter echoes, antenna sidelobes. digitally below which are reelections Removal those returns enough for ionospheric I believe scatter length variations; tests The following clutter suggest sections neighboring E to yield a complete in practice ionospheric through the by filtering –a time long refractive index. We owing to slight phase path signal E tbrough I in more B. Mode E is designed milliseconds the atmospheric of the total clutter these new Modes entering with but short enough such that the clutter is achieved in which they have been used in experimental I mountains of several due tovarfationof that 99 percent are contaminated and is accomplished completely, the manner MODE F, and I Modes) for a period subtraction discuss (E, is necessary, to decor relate appreciably tbat perfect from of these echoes which are stationary path length does notvary do”ot about i60 km altitude is normally removed. detail, and indicate profile of electron program. E- and F-region density. A fOO-psec from pulse is used, 6 km at the lower than i5-km spacing, C. F MODE altitudes however, Mode F is designed vicinity of the ionospheric in tbe lower imaginary giving a height resolution to 15 km at the higher Ab . 15 km. altitudes The sample (Table III); interval samples varies with less are not independent. to yield profiles E-region, part of that region. part of the correlation of electron and ion temperatures and also a profile of ion-neutral Due to the need for clutter function is omitted collision subtraction, to avoid doubling Te and Ti in the frequency the calculation the integration Uin of the time. All TABLE OPERATING Altitude Sample Height Range’ Interval (km) (km) Resolution (km) Mcde E F MODES 92 to 176 176 to 347 347 to 677 Ill AND THEIR Meawrernent 6 15 9 15 15 15 Tots I power 65 Real part of 106 to 124 35 124to 166 6 167 to 317 15 6to FUNCTIONS Range of step Spacing* Interval (Ps:c) (p~e;) NA NA Clutter Subtraction Yes No No 12 corre Ioti on Yes o to 380 20 O to 380 20 No o to 190 10 No Yes function complex m G 15 to 27 correlot ion function Complex H 215 to 515 I 56 to 140 ‘Slightly different altitudes in Mcde fHeight resolution +Themaximum 30 to 42 30 1.5 3 altitude rrmgesapply E, -2kmin Mcde is.afunction lagmeasured 5A d-km sample interval G, oflag-see isafunction and 6-to Total power for data collected .nd Okm in Mcde Fig. prior to Decembsr NA 1971: NA add 10km Yes to all H. 1. ofaltitude-see 12-km height corre Iat i on function resolution Fig. 1. apply for data taken before 30 Decembr 1970. TIME-BASE CENTER cm.AY. HE,GHT+ (km) 0\20]40[ (.$..) 0.74 0.76 0.78 O.@r” 0,82 I 1!% Iw! 112 115$ $18 . ,. 50j80]1CO] . (w..) 120\ 140 160 180 2C0 220 240 I 260 I 280 L 3(0 320 ?40 350 300 . . . . . . . . . . . . . . . . . . .1. AS ABOVE AS ABOVE AS ABOVE A5 A80Vf AS ABOVE . . . . ., . .1. . . .4. A5 ABOVE ASMO”, 1.C4 ! TM ! ~~~ ; ,%. I.ct l.m. 1,08 1 ...— 1.10 160 ,.,2 ,. ,2. ~~~. .. —1.14 ?U i ~“ ,.,, L As ABOVE ASAOOVf — ASABOVE . Ind?e., a,C..F.,,.,,.” ofr-lw“ .,,,,..1 ....).... a.,.1.,,., ,.!.. s,!, Iv !3!, (,,. )p!,!.”. 4.1.,s, + ,, 8.b;,..,100,,., fro., I.,T >200we (W; d.., wd P,,]. .D.!.$..oIE+ .rc!.k. ”.!fl.. -bos. +S..l.ble1!1 re,odin, .I!!!.d. d-m,.!. !T,,,, .1,,,.4., .dd.d30.,,mb=t,970, ,,,., ).wM.6 !h......!.!..” ,,6,., b.]..!30k.w, ,....s!k., ,hw.b 130k. (*w c.=-,.,. ,+~ ,.,!09,0, T ~lM!J$,C). W~Ib38’h3..1.9..=u,sa ‘1=. (a) Mode F. Fig. i(a-c). Diagrams showing time-base delay t of first for computing correlation for different pulse spacings ?. “:1 I 7 sample of each pair TEzE!_ j ,, I 4 t s..T.bl. 111 W.dm .l!ild. AO!w”. (b) Mode G. Fig. i(a-c). Continued. -lEzL I TIME-CASE I CENTER t A’P ,..) HE, GHT (km) ,,,-, 1,s3 1“”-1 II III lmiTTmTTTl , 5, I 0 I I 1.55215 10 20 . I . 30 ] 40 I . I . 50 I . 60 I m I 80 I 90 I lW I . 110 120 . ..0...0 . . . . . . 1,57 1 1 .tmmw@! maIokm.!?Ime-bwd+, f+ T. Full,., 4 .m.-he delmp,,W, I,*W.6”. r d,.- i,I.t.ld ASABOVE, (c) Mode H. Fig. i (a-c). Continued. 9 . . . plasma-drift information difference between A 40.psec the magnitude pulse is used, interval sample portion, to a scheme of averaging (Fig. 1). from above that covered not expected altitude, results from adjacent adjacent (Table III); in the upper lags due Such averaging 6-km height =egions but a shorter im- are no longer maximum lag applies at the higher is 20 Wsec. T ~, and drift velocity and its filtering giving degrades Vd as a function of height in the peak. Ground-clutter is not attempted. with clutter, echoes (The lowest and in very adverse were Mode G situations of 200 km. ) Ah . 15 km, but averaging of neighboring this to 27 km at some correlation i 5-km intervals function are not completely is measured a3titudes (Fig. i). Therefore, independent. T = O to 380 psec, from time-base lags (Fig. 4). The spacing between but a shorter maximum lag lags is 20 vsec. MODE R to as high an aItitude designed Te, results Over be studied the increase the acf is measured (Fig. 1). MODE adjacent region in ionospheric range of correlation giving with this mode. temperature only from this to 42 km at some 30-km intervals F-region, This mode is also specifically which useful nighttime returns of neighboring correlation are not completely time-base lags (Fig. i). Ac- independent by Mode H, the echo power and decrease in comparison is then reduced spectrum widens owing to 9. in ion mass; this causes the interesting with the Mode F and G values. r . 0 to i9 O psec, The spacing ionospheric at these times. Ah = 30 km, but averaging examined to be reduced in the topside below the F2 peak from F and G, degrades from the altitude and Vd profiles F and G may be omitted pulse is used, as in Modes cordingly, but a shorter maximum Accordingly, lag applies at the higher to *IJ wee. I Mode 1 is designed region covered values using ionosonde A 20-psec extends to yield however, an electron density upward into the lower determinations pulse is used, giving in the ionospheric D-region. to allow normalization The of the density of foE. Ah . 3 km. The sample are very weak, and below the D-region which are thought to be caused by aircraft. the height region profile E-region interval is f. 5 km; adjacent sam- are not independent. Mode I signals observed Ti, as can usefully so that Modes A 200-psec samples, altitudes to yield to include enough of the region are obtained ples, lags to yield Ti, pulse is used, Mode H is designed , of Mode F, the to $2 km at some correlation by Mode F as high as the layer as in Mode F, applie S at the bigher F. are not independent is often seen to be contaminated The correlation ,,~ from O to 380 psec, has been seen out to ranges A 100-~sec samples, ., !. portion along the time base (Fig. i). but results to extend to these ranges, however, this clutter I samples altitude MODE G region . In the lower 3-km samples The spacing between This mode is designed ,1 ‘;! the small part of the acf. Ah . 6 km. neighboring also neglects independent. altitudes E. of these data therefore is 6 km, but Ah is degraded of the acfrs, The acf is measured D. and real giving interval the statistics completely Analysis is 3 km, but adjacent the sample proves is thus lost. that can be studied. io These ‘fledge* weak clutter essentially returns set a lower limit are to The D-region Mode 1 observations have been the subject paper, 14 but are in- of a separate eluded here for completeness. G. OBSERVING PROGRAM The observing program in that any sequence of modes might use a repeated during Such a scheme duration. Operations Besides to-time. of E-mgicm These A few operations were i 5 min. EFFFFGH), termed consisting to observe The Mode I signals Tables W, A typical of 8 and 20 min.. observing are very of E Modes achieved sequence a sequence every hour. satellite. disturbances in these 496?. time- has been modiOccasional One eclipse run in rapid succession ionospheric of 30-min. have been run from the typical completing Respectively. (e.g., has 2-rein. with fine time resolu- E NJOde (RASEM). usually i“ coordination weak and require with Wallops long integration Island rocket times (typically by night). V, and VI provide November programs tides, oPeratiOn respectively, sequence the S/N differences pass of a scientific traveling &Apid Sequence by day and 30 min. tained from times a complete specialized have been made during an overhead have been conducted launches. i on Mode F (e.g., more atmospheric Mode I bas been run only a few times, 4 with integration of maintaining may be used. of 4, 10, 8, a“d 4 min., are made unequal to offset observations, been observed. times times E, F, G, H, and I is flexible such as this have he en run about once a month since November these routine fied to concentrate of integration EH sequence times using Modes with integraticm For observations observations runs), and choice sequence has the merit The integration modes. tion. EFGH the day and a repeated at night. ‘,,; that can be established lists of regular 1969 through December 2-pulse, 4975. RASEM, and Mode I operations Some RASEM results ob- have been presented in Ref. 45. H. MODE MODIFICATIONS The evolution to existing modes of the Z-pulse that affect their dates of implementation, (i) 6 November (Sec. IV-B (2) (3) at Millstone analysis, and the locations 20 December 1970- three additional lower 16 December changes 3971- These modifications modifications, with them are given belOw: was modified XI). Fig. l(a), is written of the data. in the text dealing added to Mode F (Sec. V-A, but many comments Hill has seem three major and quality 1970 – the mamner of noise measurement and Table ment (Table This report program collection, timing E-region and Table modified altitudes were III). the altitudes of measure- HI). to apply specifically have been included to data collected to cover commencing tbe operations prior i6 December to that date. i973, TABLE TWO-PULSE UHF IV OBSERVATIONS PRIOR TO 1976 start End (GMT) (GMT) 1440 1410 Modes F, G only 19-2U Decemlxx 0200 0200 Modes F, G, H only 30 December 1445 2050 Mode, F, G, H Only January 1350 1450 Mocks F, G, H only 30 Ja”IMty 1420 2040 1420 1440 5 March 15C0 2250 Pre-eclip* run 6 March 15C0 2215 Pre-eclipse run 7 March 1350 2400 Eclipse olxervations 9 March 1415 1940 9 Apri I 1415 1940 1 May 1330 2050 2 July 1215 1900 July 1305 0225 August 125S 0140 25 Augtmt 2105 2240 Mcdes E, G, H only 26 Augwt a345 0510 Mcdes E, G, H only 19 Septembsr 1130 232U 6 Novemter 14@l 21C0 30-3 I Decemter 2010 04w Date Comments ~9 28-29 Nc.vembar 1970 — 13-14 5-6 25-26 9-10 February Low tmnsmitter power First Mcde F with correlation fun.tio”s every 3 km (h <127 km) I I *2 —--.., --:.. . . .- TABLE IV (Continued) Start End (GMT) (GMT) 6 January 1430 2100 January 0105 0200 8-9 February 2240 0400 23-24 Febru.wy 1625 I 650 25-26 March 1525 1540 13- /4 April 2135 2050 10-11 May 2015 2015 June 21@l 2100 Date Comments 1971 — 29-30 3-4 July 2005 0900 27 July 1410 2010 28 Ju[y 0045 1415 26-27 I ,, 30-31 August 2045 2000 21-22 September 1610 1625 Mode F with 20-Psec Modes E, G, Only pulse H only Mode E gced Adder+ sudden cammencemenl Adalert magstorm 23-24 Octo!xm 1135 1120 16-17 November I&Xl 1640 21-22 December 1535 1755 27 December 1415 2135 Modes E, F only (tidal studies) 28 Decemker 1310 2115 Modes E, F only (tidal studies) 29 Decemter 1330 I 900 Modes E, F only (tidal studies) 30 December 1330 2045 Modes E, F only (tidal studies) TABLE stat IV (Continued) End Comments (GMT) (GMT) January 1710 1820 February 22C0 0500 1603 1700 March 16CiI 1640 3 Apri I 1300 2100 EFFFFGH (tidal studies) 4 April 1300 2330 EFFFFGH (tidal studies) 5 April 1505 2330 EFFFFGH (tidal studies) 6 April 1250 2100 EFFFFGH (tidal studies) 7 April I 500 2320 EFFFFGH (tidal studies) 16 May 1025 2400 EFFFFGH (m.agstmn alert) 17 May I 020 2345 EFFFFGH (rn.agstorm alert) 18 May 1010 2350 EFFFFGH (magstorm alert) DcIte ~ 11-12 7-8 1-2 March 28-i9 .~ I 20-21 June 1515 1600 24-25 July 1810 2205 17-18 Augmt 1135 1615 20-21 September 1000 1340 21-22 September 1430 2330 EFF FGH (tidal studies) 23 September 0955 2400 EFF FGH (tidal studies) 24 September I 000 2300 EFFFGH (tidal studies) October 1340 1550 21-22 November 1415 2205 1>13 Decem&r 1155 1200 13 Decernbsr 1215 2345 EFFFFGH (tidal studies) 14 December I 2C0 2305 EFFFFGH (tidal studies) 24-25 TABLE IV (Continued) start End (GMT) (GMT) January 1720 2145 March 2225 2235 April 1640 1940 8-9 MOy 1730 2320 10 May 1715 1955 AEROS Date C.mnments 1973 24-25 6-7 13-14 .: and AEROS transit transit 5-6 June 1835 2100 ISIS II transit 18-19 June 1450 2200 Magsiorm a Iert 20 June 1145 2020 21 June 1150 2230 EFFFFGH (tidal st.dies) 22 June 1125 2230 EFFFFGH (tidal studies) (tidal studies) 23 June 1210 I91O EFFFFGH 10 July 1805 1925 AEROS transit AEROS 30-31 July 2005 2155 28-29 August 1530 2210 24-25 September I 835 2315 26 September 1200 2230 EFFFFGH (tidal studies) 27 September 1150 2200 EFFFFGH (tidal studies) 28 September 1150 2210 EFFFF GH (tidal studies) 25 October 12C0 2040 6 November 1915 2145 8 November 1955 2200 10 November 1915 2145 23 November 1225 2135 6 DecemEer 1135 1300 Decembsr 2030 2150 19 Decemker 1235 2145 20 December 1235 2130 21 December 1540 2120 17-18 I # Magstorm alert i5 EFFFFGH transit (tidal studies) TABLE IV (Continued) start End (GMT) (GMT) 3 January 1215 2235 4 Janmry 1225 2248 5 January 1205 2230 16 January 1115 2115 14 February 1545 1745 15 Febrvary 1540 1755 22 February 1135 2144 4 March 1225 1350 18 March 1110 2205 EFFFFGH (tidal studies) 19 March 1115 2240 EFFFFGH (tidal studies) 20 March 1105 2300 EFFFFGH (tidal studies) 19 Apri I 0220 0450 30 Apri I 1055 1915 EFFFFGH (tidal studies) 17 May 1450 1750 28 hby 1015 2340 EFFFFGH (tidal studies) 2 June I 045 2355 EFFFFGH (tidal studies) 18 June 0940 2240 EFFFFGH (tidal studies) 19 June loa5 0015 EFFFFGH (tidal studies) 1545 1715 0315 0?20 Modes E, Honly 1 August 1025 2350 EFFFFGH (tidal studies) 10 August 1130 2325 EFFFFGH (tidal studies) 11 AuguN I 050 2245 EFFFFGH (tidal st”dies) 12 August 1040 0C05 EFFFFGH (tid~l studies) 13 August 1050 0015 EFFFFGH(tid.al studies) 14 August 1105 CC05 EFFFFGH (tidal studies) 15 August 111o 2330 EFFFFGH (tidal studies) 17- I B September I 040 2105 EFFFFGH (tidal st”dies) 30-31 October 1455 1535 12-13 Noveinber I21O 2215 EFFFFGH (tidal studies) 10-11 Decembsr 1250 2150 EFFFFGH (tidal studies) 13-14 December 1205 2125 EFFFFGH (tidal st”dies) Date Comments 1974 — ,,..:, 29-30 June 2 July EFFFF GH (tidal studies) EFFFFGH (tidal studies) (unable to read data tape) TABLE IV (Continued) Date start End (GMT) (GMT) Comments 1975 — 15-16 January 1330 1400 12-13 Fetxuary 1150 2305 21-22 March 1725 2040 15-16 April 1440 2150 2-4 May 2145 0230 7 May 2155 2400 May 2320 2345 10 June 1055 1I June 12 June EFFFFGH (tidal studies) 2240 EFFFFGH (tidal studies) 1035 2305 EFFFFGH (tidal studies) I 030 2245 EFFFFGH (tidal studies) 2130 2000 11 August 1125 2250 EFFFFGH (tida I studies) 12 August 1103 1605 EFFFFGH (tidal studies) 2250 EFFFFGH (tidal studies) 23-24 ‘:/ 16-18 hly 13 August 1945 1215 1355 13 October 1110 2205 EFFFFGH (tidal studies) 14 October 1535 2130 EFFFFGH (tidal studies) 15 Octoter 1040 2140 EFFFFGH (tidal studies) 16 October 1210 1910 EFFFFGH (tidal studies) 17 October I 040 2210 EFFFFGH (tidal studies) November 22Ca 2400 13 December 1210 2310 EFFFFGH (tidalstudies) 16 December I21O 2120 EFF FFGH (tidal 24-25 21-22 Septembsr studies) TABLE RASEM Stati OPERATIONS V PRIOR TO 1976 End (GMT) (GMT) 13 March 1400 20Q0 interrupted 1448-1520; 10 April 1420 22C0 Interrupted 1430-1442 Date Comments ~ 8-9 May 1545 C4Ds Interrupted 0000-CO02 15-16 May 1555 0330 Interrupted 1758-1 806; 1700-1B20 2052-21 and 0000-0014 22-23 &y 170Q 0400 Interrupted 0000-0241 28-29 May 1540 0320 Interrupted ODOO-0002 1345 2400 1110 2330 8-9 August 26 September 1571 — 20-21 May 1730 0340 Interrupted OLWO-0LY38 IC-11 June 2010 2020 interrupted 0000-WCG; 23C0 0400 30 Novemter 1 December i - 0041-0044 10; .. . TABLE MCIDE Date VI I OPERATIONS start End (GMT) (GMT) PRIOR TO 1976 Comment ~ 27-28 )November 2110 2130 Operated with separate recovery 31 Decemt& 1600 channel 21CQ 1971 — 23-24 April: 1545 .0040. 1972 20 January .1730 1800 Coordinated rocket with Wal 10ps Island launches. 1715 ~~~1900 15 knuary 1605 1810 Ccdinated with Wallops rocket “iaunches 16 January 1645 2000. Cocmdincded with W.al 10ps Island 3 I January.. Ccbrdlnated with Wallops. Island ~~~ rocket icwnches” 1973 — rocket 1974 15 January 4 1355 1850 launches Island III. A. EQUIPMENT GENE RAL The system with a brief parameters description Fig. 2, directs of the 68-cm of the apparatus; a pencil there bills. is a virtual In normal to ranges absence operations, of the order ter echoes radar ing regions. Figure come from of natural screening out to 200 km range. display of 8 nmi (14.8 km). Nearly a,re observed The most distant i 50 km being the White Mountains. Smaller short pulse: appear tbe clutter side of the antenna. a fence would need to be over out on tbe basis include of cost. observed has been placed solutions Additional (Fig. 2). and are prob- along each of the three legs a distance i.e., the taper to have decreased so that more such choke round its edge. despite the increase and strengthening however. the intensity to come via re - these, absorbing netting ?./4 from the inside energy is directed over the aperture face. toward has been increased by adding to the feed horn an extension The choke appears with the result that the effective One final step that may have contributed been a reworking Unfortunately, are believed to reduce of approximately of the illumination This last change was accomplished a quarter-wave is the construction have been attempted, returns In an effort the size of tbe feed horn has been increased of the reflector, returns skirt to the antenna feed horn to reduce from the horizon. feed support legs of the feed horn, appear of a Large conical directly t O to 20 dB. ficiency those at 100 ft high and seve r-al hundred feet long and thus far has been ruled A number of less -expensive the addition off the tripod incorporated in range and to the south are not permanent for combatting from de- by aircraft. of the northern the center with with the main to be extended fence over a 60- sector In addition, clut - of the offend- in New Hampshire, that has heen considered flections out conditions, on this display One scheme of echces distant and searching in this manner, of an anti-clutter These from and directions The range markers with mountains returns reflections i“ the antenna sidelobes the ranges returns shown in the radio star During very’ adverse obtained of 0.2-. together Since the antenna is on a By transmitting all can be identified allows that would prevent of returns in Ref.1 The antenna, purposes. we have been able to locate antenna at an elevation here. this slight offset beam for calibration 3 shows a PPI the north. ably caused this is not repeated strong ground returns antenna, beam of the steerable note intervals have been presented of $00 km, and again near i 50 km. can be detected the steerable system beam 20 south of the zenith; Cygnus to pass through the radar hill, radar collecting to have improved that the ef- area of the antenna does not in beamwidth.l to improving of the surface the radiation that has reduced pattern of tbe antenna has the rms surface tolerance to about an inch (h/24). B. FREQUENCY CONTROL The modifications point where, makes it possible the ground the radar required citer, described as a rule, to consider returns be obtained from local the echoes a single oscillator coherent to phase-sensitive the intensity the receiver to be coherent. so that phase coherence that all the receiver nals must be applied saturate removing are expected equipment above have reduced they no longer from modifications of the ground returns frequencies, oscillator 24 Also, linear as have been made to will be preserved. of a simple This in frequency, as well as those generated as shown in Fig. 4. in place to the gain setting. the data by filtering Accordingly, rectifiers of the ground returns at its normal This has in tbe ex- the received rectifier Sig- employed Fig. 2. Millstone Hill Field Station, To left is 220-ft vertically pointing UHF antenna. Tripod supports horn feed at focus. Building to right contains transmitting and receiving equipment. Larger, fully steerable antenna operates at L-band and has been employed for ionosphere studies at Ohlio,,e incidence. Fig. 3. PPI display of ground-clutter Range marks are 14.8 km apart. returns 22 observable from Millstone. I 18-9-6603 I TRAN&TTER FROM TR SWITCH 440 -MHz AMPLIFIER 440-MHz PREAMPLIFIER d 440 d 410 41 O-MHZ AMPLIFIER A I II AWL!%I I I I 30 w%%. I-1 II 1 I 1 .n 1 28.2-MHZ SYNTHESIZER 4!0 X8 MULTIPLY 4 28.2 A L 1 2-MHz AMPLIFIER 2-MHZ AMPLIFIER 2-MHz AMPLIFIER (I n 52.5 51.25-MHz SYNTHESIZER % 1 2 1 LOW PAss LOW PASS 1 TO sAMPE LOW PASS LOW PASS 1 1 -AND-HOLD I CIRCUITS Fig. 4. Block diagram showing frequency control system employed for measurements described in this report. 23 for radar FROM LOW- PASS FILTERS 1 SAMP\;i:~}-HLID 1 I __ SANpLE-ANO-HOLO CIRCUIT 118-9-660+] __ ~PLE+M()-~u 1 __ ~A~pL~+,/&H~~ CIRCUIT CIRCUIT Am 10- BIT REGISTER 1O-BIT REGISTER 20-/JsecCLCCK PULSES 1-4 I t T “ DATA ACKNOWLEDGE” PULSES ( 1.75-psec spacing] 10-~sec % CLOCK PULSES (HI) I ] CONTROLLER “ TAKE 0 OATA” PULSE “INTERRU=T AND OIRECT MEMORY ACCESS “ DATA REQUEST ENAELE” PULSE “INTERRUPT 44” 77 f XOS 9300 COMPUTER Fig. 5. Block diagram employed to measm.e computer. I showing data-sampling and data-transfer equipment signal voltqes (Fig. 4) ~d c.mmnu”icate these to 24 40” hitherto. With these changes it is possible turns and see that this does indeed vary The new frequency frequency rapidly C. adjacent of a single to receive been modified spacing 30-MHz These switch S2. moved. the transmitter the radar frequency to to be performed dur- tuned to the correct by ~ltering frequencies, frequencies the setting rectifie M, in turn, channels This is of particular drive low-pass filters the role played in their importance Frequency by the common rectifiers ope r- having a time constant rectifiers can be reversed by the two phase detectors gain m- incomplete in attempting has as shown in Fig. 4. The of the 28.2 -MHz a pair of phase -sensitive of the phase -sensitive due to differences tbe receiver in each channel, phase separation to determine of by the ionospheric can be re drift ve- accurately. In order to minimize the time over which the noise in the receiver a wide receiver bandwidth. Thus, tbe overall set at 50 kH-z and the post-detector bandwidth at f 00 kHz, of the order represents of 45 kHz. This value bandwidth as narrow as possible and the need to limit the distortion width (see Appendix In order to improve second has been changed. sion was ac cmnplished by applying and suppression In this way, DATA tbe recover-y additional in the receiver) of finite band- modifications the time interval operating receiver time to approximately employed Finally, vacuum tubes, pentodes A solid-state diode gates between to -iO 2 psec. A tube TR switch by one employing switch bad a cleanup time of ZOO to was about iOO Fsec. time is reduced the trans - As originally 200 psec to tmm off the transmitter. to short- s“pp~essor-base by inserting have between condition. to nomnal in a few microseconds. pulses is, carried the gating and suppres- in various receiver various IF is now em- amplifier stages. psec. out in the rna.rmer indicated aPPearing at the outputs of the low-pass circuits. 0“ eratine by a. receiver to keep the SAMPLING Data sampling sampled abow gas-discharge The original is achieved i.e., the turn-off tube in the older The time constant for recovery ployed, time, req”itwd the original The gas-discharge 100 km, to its normal to reduce b“t the new switch is restored of the receiver stages. recovery modulator has since been rebuilt diodes. 300 psec, below of the receiver change has been to replace solid-state we rail bandwidth the desire the mnvanted noise power data produced at altitudes the receiver the UHF transmitter The modulator an effective between it is nec - bandwidth has been A). mitte r pulse and the restoration designed, of the spectral is correlated, pm-detector yielding a compromise (and thus minimize to make measurements been necessary command for an interval over from filters of i 00 nsec, a range of * 5 V assigns in Fig, 5. (Fig. 4) are applied a clock pulse (normally and this voltage any applied 25 i_ permits two experiments m adjacent channel drives By commutating effects es sary to employ i of switching must be kept to a value that can be accepted Each receiver The sine and cosine throwing ‘1 permits transmitted may be varied but the separation any systematic pulses two IF channels IF amplifier. 40 ps’?c. locity separately the frequencies ated in quadrature. D. re - sweep of the radar. to provide between synthesizer, ,. the capability This provision and, in effect, distant clutter with time. (Fig, 4) provides switch Si. pulses the phase of the more RECEIVER In order .1 only slowly system via the solid-state be changed between ing tbe course control to monitor every to four 20 psec), is then stored. voltage The video signal voltages ‘sample-and-hold the applied A digital voltages voltmeter a sign and one of 512 levels. m are 0p- The digital words representing to the computer format register i.e., Transfer namely, are required (see Sees, IV-A-4 a“d -5). instead of every TIMING equipment etc., operations. modes the computer In order of the data, channel is employed digital at a time samples time is now 7 psec and is carried is trans- out eve~ B timer version equipment provided i O psec pulse length This timing generator in Ref. i (A timer), for the four preset of the B timer allowed operating a variety made in Mode was which provided Modes E, F, of pulse lengths, I. Subsequently, unit. was constructed from by push buttons, the desired system. that described for measurements into a single is provided selects employed to cent ml the transmitter to the radar it from Tbe combined to synchronize a synchrcmizing commercially available but can be transferred operating mode by loading the data-acceptance pulse ( ‘,Interrupt on each sweep of the radar data pulse n) is required the remainder logic cards. Selec- to the computer. an output register coupler 100 K, allowing In some of the modes In this with an ap- modes and spacings, operation of tbe gates. the required modulator any one modulator gates gene rater Accordingly, pulses synchronizing the noise level yields and connect pulse can hold the excite r ON. in Fig. 6. 26 ( ,,Inte r- pulse. terminals via a in the receiver supe rheterodyne (diode) ON and OFF ON and OFF to employ these through hy RF As shown in Fig. 4, the system. (Fig. 4) and tO the 51.25-MHz separate coincident pulse as many as four separate out using solid-state it is necessary channel for units. which is an inverse require pulse (Wake- 40’1 synchronization equipment, op- equipment This pulse is transmitted at the receiver (temperature) levels interface the alte mate on each of two frequencies). by an exciter These the timing inserted pulse raises This The gating is carried Associates. proper pdse at the 2- and 30-MHz (not shown in Fig. 4). A third with the radar synchronizing from ‘of pulse spacings. hy the new timing are generated samples of the next ,tlnterrupt measurement to amplifiers by Sanders and -5). ccmve rsion to absolute provided (2-pulse RF pulses Gating is applied factured IV-A-4 a noise -calibration intensity at the computer H and I, a second to accept to the end of a cycle at the end of each sweep. signal are required transmitter In Modes 200 ~sec in advance E and I include operations 40 ‘t) is tram mitted to the digital the computer (see Sec.. the computer to the data interface Modes time base. to alert of the sweep rupt 4481) alerts erate rate, word. erations, as well multiplica- clock to the presence in time is iO.5 wsec. receiver was const rutted experimental and was initially The new timing tion between pulses the computer position for immediate out at the computer only one pair of the available the computer to distinguish A second, all these units were coaxial by the computer only a single I“ this case, and to synchronize for single -pulse case, in a 10-bit word to first is carried to alert one at a time The function of the EQUIPMENT and spacing propriate (H, I), and transferred register. 20 psec. known as the B timer spacings, position so that the total transfer The data transfer New timing -gene rater G, and H. first required In order modes into the computer, in registers and a format into tbe memory (3.5 #see) In two of the operating ferred are stored (switch) place it in the location of the samples one word per 1.75 psec. two cycles E. voltages multiplexer is to move the sign bit from a 24-bit word, tion, the sampled via a digital switches triggers. pulses signal leading the trigger manu- For to im- pulses to gen- 13 OR** gates in such a manner This arrangement is indicated some that schematically I 1 +5 INVERTER Fig. 6. Block diagram showing manner in which exciler ON and OFF triggers are mixed to control Changes timing reverses reverse of the exciter equipment the frequenciess arriving at St; by the receiver to be averaged but is retained of the switch which it is under the cOntrOl Of the cOmPute r and is used ‘0 appear channels in the pulse train. and thereby out in the sweep integration E, F, G, H, and I use this capability useful purpose, under control Not shown in Fig. 4 is a ‘commutate” in which the frequencies tates the role played Modes are made during the train of pulses via switch S, in Fig. 4. the order the two channels frequency separate transmitter RF pulses generated. for simplicity (E -Mode of timer 27 allows any gain difference performed frequency operation). This effectively commubetween by the computer. commutation accomplishs All no , OPERATING IV. A. EQUIPMENT 1. Transmitter Mode iOO-~sec frequency. transmitted from the second pulses (Fig. 2. for 20 sweeps. Figure Frequency from from the first tbe first vs altitude. pulse are summed between The roles in which the pulses of the computer, out. and causes together irrespective and/or differences the transmitter. in the performs b“t is retained on receipt a and then returns This commutation by the computer TWO one m each radar the order for all other modes) the relationship” and Fig. 7(a). pulse for 20 sweeps will be averaged .is pe=formed VII of the echo power 20 s-weeps by chazging in the gains of the two channels 7(a) shows Control for simplicity and: sync pulses of the ?!Interrupt 4411 timing pair), pair is made at a fixed interpulse operating at an inter’pulse shown in Fig. 8(a). period period These appear four pulses displaced by counting the number time-base through the logic from * O sweeps. of sweeps, since generator (BI genera$or (BZ timer), summing the first network in time-base timer) two each successive of sweeps Thus. and resets the maximum the two timers the sweep. A pulse counter the B2 counter spacing were as shown irr length, by 20 #see per sweep. counts the number with the B1 [ Fig. 8(a)] every .(wide-spaced the last two pulses. of the series for each pair is made to increase of pair as shown in Fig. 7(b). Owing to the difference exciter. 20 psec further T < 380@ec by a time-base generates are passed 200< the close-spaced are com$rdked A pair of the time base – one pair for de - of 8.0 O msec ....A second on the output of the Bi timer is set each sweep to surround of. 8.02 msec, the transmitter 3n this way, the spacing into coincidence are shown in .Tahle VII and Fig.7 (b). the other for delays that are trazzsmitted operating two pulses characteristics (close-spaced The wide-spaced to control - Mode F on each of two frequencies firs.$ two pulses pulses Returns commutation is transmitted operating iII Table of the time base, 5). lays O < r < i 80 ~sec Fig. 8(a) shown This is under control at the two frequencies Transmitter second estimates to handle the returns Mode F transmitter The yield so that any differences operation. in Mode E. pair). separately are cm each sweep function in the E Mode (it is essential of timer pulses characteristics transmitted are Fevers ed every channel transmitted no useful are at the two frequencies. receiver of frequency, - Mode E timing pulses channels given pulse Control These of the receiver powers COXTROL E transmitter adjacent are TIMING PROCEDURE between forced into coincidence. At the instant the Bi and B2 timers first of the B2 pulses coincide (T = O). are forced 200 wsec. Thus, on the first sweep the second sweep one obtains the correlation and 380 ~sec. values The B2 timer of T are obtained A complication subtraction coherent procedure from into coincidence, The wide-spaced the correlation is measured at 20 and 220 psec, is then desynchronized the second pair of pulses of the Bi and is then separated at T .0 and T = 200 psec. and this continues to the Bi timer. by In this way. COI until T = 180 all possible in i O sweeps. is introduced is basically sweep-to-sweep. by the need to remove a simple For digital this reason, 29 filter clutter echoes that removes it is necessary (Sec. 111-A). echoes to repeat The clutter that appear each sweep to be at least mm Mm”,,,,% 1 TM.. VI, C“A14CTERI s!., MA, , F G “ F+. tam. [.mkd] (p-) ,Ca do ICC X0 I.,WF.JS. *CM (“mid) h,..) * 8 h...,., rmw+,.d ,W ~,,. W,” Med.. :.-%, 20 8 TWOF,,, 0’ w (k., F- ,-.% .“. =.-h fr.q”.m, l.m Pi. d PJlsV -.W -. ~ .O* lr.q..rs, o !. ,8, ,..., ph. 0 !. !80 h-r Cm. P,, d ?!’” 5 cm $+. p. SW.., F- *.-P f-+.-, ,Pc,., ,.M... Fw1’,,1 (,.4 ,W 2C0. 3s0-F.,64 W*., .1 ,“,,. SF4q, ;,,, s?,, (,,4 N. ,,,-d F.,,”, {,s.4 ‘, ‘* I ., ,kmtims.~m. c1... .,dW* Swd %. (wed 4.,”,., .’ 5.,.. 20 . . 20 20 ,0 ,4. 0,0,20,20,W Zm,202, w 0,10, !0,30, 0,)0, 20,!O,,,. “ ,W .. .. 2, 1 20 ,Cw... %q.e.o, C.m.!..m m.,- atSpe h,,.(.,n!.m.i 4,. ) CO,.,),,”, “,,. Coiw,d,.! wilh ,.,,0,,..-, ,.,,., ,.em, ~,,. w.1.w L.q* sm. 220,s., MC., s,.. ,-.,,.. ..- p,,- pi. C.!”ddcm! wi. ~;:, ... C.i..idw! WI!h m no War!d ,.!,. p, k. 6CQ . 5.,.220,,64.,. ,.,”220,,.$ 6.* ,,.% ,,.. w.,,+, Ma”,mu.?SW.icq is. ‘.-*W 320,,,. Cd-id..! ‘“M ,... “cd (,,.,, )dUlwef film sm. w,,. w: P“k, ,-,-,, . . 20 INa m 0!. lm *O 0,0,20,20, 40, Z.w, 2C0, 220, MA Fdr, 2c0* 384W* Fir ., .!, !”* -,- .,,,. ,. WWA, 30 s.,” 220,,.. ,.,.”. ,,m 5,0,,, m ,,.. L4m, ,,.. PRF [0’I, WW”P1 40”) 0 COMMUTATE I FREQUENCIES i“lntemurd 44”) I I —.. — f, I 1 +mmse. (every 2W ~we.P) RF PULSE _~ -Ioopsec ‘ . +7.9 m8ec 0 12 RF PULSE y FREQUENCIES COMMUTATED y=:; INTERRUPT —.— 0 44” — +Ioopwc TX MoDULATOR PULSE E.oop$ec .._~ +7. T8m8ec -220/I*ec I + 380psec .,,oF,w / +500@c ----~ I ! RX SUPPRESSION PULSE NOISE cALIBRATION PULSE \ ~---- ,,.om,en “TAKE +770 msec DATA” PULSE (not .sed ) ———— + 6.4 “WC (a) Fig. 7(a-d). differs from Mode E. Timing diagrams (not to scale) for Modes E, F, H, and 1. Mode G Mode F only in that all four RF pulses are $00 psec in length. 31 PRF(” Interrupt 40”) I r I +8,0 0 mm SET r = 0 [“Interrupt 4 4“) COMMUTATE FREQUENCY ——— I +7.0 nwec [every 20th sweep) f. RF PULSES -1 ‘“w (r= Q, 0,20,20,40,..., ----~ I FREQUENCIES COMMUTATED 180psecl JG51-LJ:::::LJLGJ (, = 2Q0,200,220,220,240,..., TX MODUL4TOR S00psee) PULSE _____~ -220/L1Oc -220psec +7.78 “s,, +380p8eo +7.78 +500Pse0 I r-------~ RX SUPPRESSION PULSE “TAKE _——~ +6.9 msec (b) Mode F. Fig. 7(a-d). Continued. ‘1 I 1 ‘1 ““,{ nwec 32 DATA” PULSE [not used) PRF [“ Interrupt 40”) J~ L._._.. +8.0 msec 0 SET T =O (“ Inierrupt 44” ) _____–_~ +7,8 m$ec (every 20th tl(or f2)RF I } (r= 200+ f2 (or fl) RF PULSE FREQUENCY PULSE 0 sweep] w:i’’’’up’::~ 1’ psec O, 0,10, IO,2O,,I9OPW1 Tx MOOUL ATOR PULSE 600psec -i_________~ +390 psec -Zzopsec -220+0. +778 msec +778 rose, +700psec r-------~ 7 RX SUPPRESSION PULSE “TAKE OP.TA’CPULSE (change RX channel) (c) PRF [“ Interrupt Mode H. 40”1 +5.0 me.. 0 f)(or f2) RF PULSE “} FREOUENCY Commutate EVERY “INTERRUPT PULSE f2(w f,) RF PULSE 40” n 0 +5.Omsec +20pse. ~ -Zoopsec +20 /Lsoe +4.8msec -200 +400/L%ec +4.8msee fisec I RX SUPPRESSION PULSE NOISE CALIBRATION I +4.3msec PULSE I +4. srnsec “ TAKE DATA” PULSE (chmqe RX channel] +3.9m$ec (d) Mode I. Fig. 7 (a-d). Continued. 33 Wm!!.L LRFPULSEION , > RF PULSE 1 oFF PRF l/RF B1 PULSE 2 ONICHANGE - RF PULSE 2 OFF b MODULATOR ~ MODULATOR PULSE OFF FREQUENCY PULSE ON/RX oFF R&ii i RX ON TIMER * NoISE + NOISE PULSE OFF PULSE ON \ 1 ~ i + “INTERRUPT 44” w “INTERRUPT 40” 1 i TO COMPUTER T (PRF 2) B2 — TIMER (a) Original scheme; Block diagrams Fig. 8(a-b). with right spacing (Fig. 6). RF PULSE 3 ON . RF PULSE 3 OFF * RF PULSE 4 ON/CHANGE + RF PULSE 4 oFF spacing incre%ses showing ::1 1 34 after TO RADAR FREQUENCY } every method of generating sweep. RF pulses lo-p,,, -i - CLOCK PULSES m RF PULSE 1 ON RF PULSE 1 OFF PRF 1/RF PULSE 2 ON/CH&NGE FREQUENC RF PULSE 2 OFF 01 TO . ‘ RADAR MODULATOR PUL2E ON/RX OFF TIMER MODULATOR PULSE OFF RX ON NOISE PULSE ON NOISE PULSE OFF [Recycle) [ j RESET COUNTER RESET ,?: ,~ “TAKE DATA” TO COMPUTER ‘“INTERRUPT 44” “ INTERRUPT 40” 1 GATE -F! 4 OR L -J ,. I 9-% — PRF 1) GATE COMMUTATE — OR ‘IMER -RF E= PULSE 30N] 4=F’E” ‘T”’’’” RF F’”LSE 4 ON/CHANGE FREQUENCY I (b) Final scheme; spacing increases Fig. 8(a-b). after Continued. 35 every second sweep. once in order to get an estimate 20,40, 40,... psec for the close-spaced for the wide-spaced pair [Fig. these two on alternate scheme sweeps only on every through all spacings. spacings. (Figs. pair of pulses [ Fig. 8(b)]. second radar Commutation 5, 7, and 8). from In Modes being measured and receiver 3. Transmitter gains, of the frequencies Control Mode G transmitter [Fig. cm one frequency pulse transmissions simplicity spacing utilize of timer were From Figs. is the reason the longer Control timing than the pulse length (2OO ~sec), where frequency between characteristics maximum the first spacing the two pulses etc., frequency pair of pulses for and and to best simply obtained is transmitted. This edge of the second pulse rather This also explains commutation employed, since always as outlined the superflu- in Sec. 7 below. VII and Fig. 7(c). only a single the maximum overlap, is transmitted the ionospheric i.e., pair of pulses spacing a single on successive Owing to (i 90 ~sec) is is less pulse of length time-base sweeps as echoes are taken in the other modes. are expected Accordingly, to extend out to the transmitter on each sweep of the time base so that a pulse of a given length is trans - may then be obtained by taking these from on one frequency channels (A/D) converters nels is initiated is retained and the pulse lengths of different at the leading Actually, first between COntrOl is is needed in Mode G, the but this repetition are shown in Table echoes digital subtraction it can be seen that Mode E is very mitted over pulses Transmitter of Fig. 7(c). the noise samples is altered VII. one conceived, and phase-detector Due to the long H Mode pulse lengths, delays This transmitter that the length of all four pulses (as in the H and I Modes). 200, 200, 210, 210, 220, 220, . . . ~sec, diagram except As no clutter so that ody per sweep of the time base. shown in the timing the close-spaced lags on the other. – Mode H pulse length and reduced transmitted of 44” pulse in the results. as for the F Mode, VII, the B2 timer commutation Mode H transmitter to to step available. edge of the first Transmitter of an ‘tInterrupt using different are shown in Table why the Mode E sync pulse is located ous E Mode frequency with respect is now required prevents to be measured Mode G was the first 7(a) and (b) and Table than at the leading 4. 40 to i 00 psec. then chosen to avoid overlap Mode G by disabling commutation in all respects need not be repeated power between out at the end of each cycle on receipt and the wide-spaced disparity characteristics control operation. the transmitter from timing from is carried function for the B2 timer – Mode G to Mode F transmitter 7(b)] is increased are O, 0, 20, generator is made to retard hy the computer with an inevitable by arranging reset-pulse and a total of 20 sweeps F a“d G, this frequency always tie Thus, the B2 timer sweep, would cause the two parts of the correlation powers employed and 200, 200, 220, 220, 240, 240, . . . psec of Fig. 8(a) is accomplished As in Mode E, this is performed identical the actual spacings a pulse at 8.oO and 8.02 rnsec and commutating the B1 timer lags Thus, 7(b)]. This change to the basic to provide of the clutter. and repeated is accomplished on the second frequency. by controlling (Fig. 5) is transmitted by the radar timer the alternate by means of the ‘Itake-datat! free channel. which pair of words to the computer. 36 Samples receiver from the analog-to- The command pulse. of ionospheric This change- In effect, to change chanthe radar frequency is switched pulse of a given back and forth in advance frequency of the sync pulse so that the delay between and the correspo”di”g “ois,e samples is approximately a t 5 instead of 7 msec. 5. Transmitter Mode Control I transmitter timing surements were adjustment of all transmitter single initially number wing timing in order the problem can be considered allows the original completely samples of noise frequency of tbe B timer, to minimize the delay required suppression. steps. In order the interpulse between mea- wbicb allowed In this mode, only a the start of the to compensate period These for the reduced is reduced to 5 mSec. This in tbe H Mode that there is no place on the time base that free of echoes. Accordingly, the radar frequency is switched by the ‘rtake -data!( pulse [ Fig. 7(d)] to tbe other channel. to be gathered (instead version VII and Fig. 7(d). in the 10-psec made per unit time, encountered a point along the time base defined on a given are shown in Table parameters and the end of the receiver of measurements introduces characteristics conducted pulse is transmitted transmission - Mode I apprwxinmtely of 4 msec). 9 msec after a pdse In Mode L the “Interrupt at This has been transmitted 44 ‘p pulse is not used by the computer. Transmitter b. Modulator The beam current penultimate amplifier) TabIe In Modes VII). length (6OO ~sec) in the klystron is controlled E, F, is also used to initiate to 220 ysec wasteful, 7. Receiver Table above, table. from the RF drive modulator VII). of the receiver. In the RASEM is made to reduce the timing is suppressed The receive= Provision a fixed used (r = the long (600 -psec) pulse length is modulator pulse is it to 320 psec. sequence are especially pulse, insofar as it affects at the time the modulator is restored to operation is made for manual control returns The noise -calibration in the H and I Modes accepting pulse waveform In Mode 1, the modulator mode, pulse (Fig. 7, pulse is maintained the sync pulse by 22o psec and strong, employed after of the turn-on causing samples of noise, and marks the receivers. pulse is turned on. time; receiver the position This is true given in the this is useful when the near- saturation. only in the E and I Modes, which in these modes As noted the sync pulse at delays is placed at the end of The IItake -data!! pulse is em- the time base and turned off at the same time as the receiver. ployed via a modulator the widest pulse precedes (and in the klystron Control the receiver ground clutter of the transmitter long enough to encompass suppression VIII summarizes for all modes. separately The “modulator-on” (Table and provision final amplifiers G, and H, the transmitter that is just 380 psec for the G Mode), reduced Control on the time base where are taken from the computer the (alternate) starts channel con- taining noise only. The sine and cosine modes receipt except F. of every 80 sweeps channels Modes For second of the phase detectors Interrupt 44 pulse (i. e., for Mode H), and hence proceeds E Mode commutation performs are commutated E, G, and H this commutation every is carried 40 sweeps (via S2 Of Fig. 4, in all out (by the computer) for Modes at half the rate of the frequency no useful purpose 37 but is implemented because E and G, every commutation. of the way in on TABLE Vlll RECEIVER TIMING CHARACTERISTICS M&e Pammeter I nterpu Ise Pericd (n.aminal) RX ON “ . F G H I 8 8 8 8 5 500 500 600 700 40Q 220 220 220 220 200 NA NA NA t4A NA NA (N.:) 6.9 (msec) (after sync) (Pse.) RX OFF (before sync) (Psec) Noise E Puke ON 7.0 4.3 (after sync) (msec) Noise Pu Ike OFF “Take-Data” (after (r4) (u) Every 40 sweeps NA Every 40 sweeps Commutation _.A.. ...,.=.., At Rx OFF Pulse sync) (rime.) Phase Detector At Rx OFF --- ... . . Every 80 sweep 3.9 Every 8 sweeps which it is obtained mutated every by modifying 8 sweeps. Mode G timing In this instance, only. In Mode I, the phase detectors the computer counts Interrupt are com- 40 pulses t.o control commutation. 8. Data-Sampling Control The !Isample -and-hold!! voltages continuously frequency is increased computer and format words @se, register made every In Modes controller 5). sends These the sampled 20 psec, These at i. 75-Wee cha””el output register (not shown in Fig. 5) used to commutate controller puter... .ln all modes, The computer In Modes trati”of delays. over in Table which the data samples pulses whi5h opens S3 whe”n th”e”cOrrect pulses. so-that the radar ace sent to the comvia S2..in Fig. 5. a ‘rdata - numb~r has been re- pulse time base is sampled over obtaimed,.. together via the same and transmitting ‘take-data’! into governs... and ttie phase detectors. the train of samples werds is i 0.5 psec. are transferred is accepted of the computer the frequencies for the thb digital of the following the “data-request” 40 pulse is used to initiate H and I this pulse also arranges. ‘!data-~-questt! and transfer E, F, and G a request pair. ~f words the length of the train. by counting request -enable 89pulse to tha: controller, ceived. to define of multiplexer In Modes and this is under control (Fig. 5) serves an Interrupt determines On receipt intervals, and only two words of which which receiver The interface the sample are also sent to the to take data. during the course intervals, Choice is being s.ampletl; of the signal In the H and I Modes, for the computer are transferred 7 ‘psec. take samples clock pulses into the computer. are at f O-psec during the following voltmeters ,,d?.ta-acknowle dge!l pulSeS to the digital are spaced voltages and four words H and I the requests the computer digital E, F, and G. as requests the comPuter (Fig. representing and associated to iOO kHz (via S1 in Fig. 5). via the interface a t!da~ .reque~tu circuits at a 50 -kHz rate in Modes to initiate a second two regions. The with the sample. spacings,. are listed ~: IX. i B. REAL -’T2ME COMPUTER OPERATIONS of operation, the computers ) 1. Gene ral In all modes almost Accordingly, continuously. been separated. in one section The data samples of the memory, during the previous are being written ‘,’. ,! .,,. .! ~ area. period tectors l~(t)l and stored ina Modes of time (de finedin program different is operating section. period to take data samples data samples Table have IX) are stored on the samples In other words, the previous gathered while data samples set is being processed is timed by the computer E, F, G, H. and 1 uses twO frequencies to a separate, a pair of low-pass and a pair of 10-bit is to resolve signal sin a, where and the arbitrary physical filters, storage the instantaneous cos a and l~(t)l the received interval the length of the integration corresponding phase detectors, converters, given requests and processing in another through count- 40 pulses. Each Of the operating frequency fora to answer of storing while the processing into one area of the memory, III each mode, ing Interrupt required the functions a pair of sample-and-hold registers receiver (Figs. cOmPlex t is the time-base phase-detector in channel i as =(t) and in channel 2 as ~(t), ith transmitter sweep is Of transmissi0n8 data channel containing 4 and 5). reference. the information 39 circuits, each a pair of a pair of A/D The action of the phase de- VOltage ~(t) delay ad an amplifier, into orthogonal components a is the phase difference Denoting presented the samples he~een obtained tO the cOmPuter On the TABLE IX DATA-SAMPLING ARRANGEMENTS Mde Parameter Delay E .af First Data Sample Accepted COmputer (msec) 0.56 G 0.56 H 0.56 I - 0.55 0.35 by Spacing of Succeeding Samples (psec) 20 Last Dcit.a Sample Accepted F 20 7.560 20 7.460 10 10 7.460 4.95 2.95 6.90 3.9 by the Computer for This Train OF Samples (msec) Delay to First Data NA NA NA NA NA NA NA NA NA 7.50 4.59 I T 4 Sample in Seccmd Wi ndc.w (rnse.) Spacing of Succeeding 10 10 Samples (psec) Last Data $ample Accepted by the Computer for This Tmi. of Samples (msec) Number of Sweeps Over Which Data 2 2 Are Gathered Before Being Processed w All delays are measured with respect to the Interrupt 40 40 pulse. Xi(t)i where = ];(t)il yi (t)i = I;(t)il sin @i x2(t)i = l~(t)il cos ei y2(t)i = l~(t)il sin ei (4) @i and e i are the phase differences and 2, respectively. the designed are used, The choice purposes and Table from the phase-detector of when to take samples of the individual X lists uses to which the samples modes. are put. The sections that follow are being received. The altitudes quoted below for the various (see Sec. II-H). cated in Table 2. Prior Computer III). modes describe them depends gathered the computer to modify upon in T#hle fx is conducted apply to data taken starting it is necessary 1 and the operations i 6 Decem- these a~titudes as indi- III. Operations Mode E is employed (Table to that date, on channels and how to process over which data processing data samples ber i97i references Not all the data samples the delay intervals out while carried ] Cos @i – Mode E to. determine To determine the echo power the echo power profile over as a function of delay, the altitudes the computer 92 to 677 km squares the in- put samples ‘1 , 51 (t)i = X:(t)i + Y:(t)i = :(t)i :*(t)i = I=(t)il 2 Sz(t)i = X:(t)i + Y;(t)i = 7(t)i T*(t)i = .\F(t)il 2 The measured unnormalized Pm(t) power vs delay Pm(t) . ~ si(t)i + ~ i ] where –im 100 -psec delay in reality in the discussion In general, is then taken as the summation (6) wec)i for the time separation The frequencies alternates are actually accordingly between commutated between channels. every the pulses transmit- 20 sweeps, and the This complication has been in this section. the total power wanted noise power, (5) i the shift in delay by i 00 psec allows ted at the two frequencies. ignored 52(t . Pm(t) measured and unwanted clutter includes echo power. wanted ionospheric Denoting echo power, un - these by the subscripts s, n, and c, respectively, Pm(t) The signal interval of less samples = P~(t) + Pn(t) are correlated than i msec over correlated over an interval the receiver (45 kHz). It follows Ground-clutter intervals of the order des crfbed O< T < b~~f, that the signal echoes, (7) . for tbe measurements are likewise sweep-to-sweep. + Pc(t) where and noise of O < T < in this report. samples i.e., The noise beff is the effective over correlated an samples bandwidfi are not correlated cm the other hand, should be well 41 Af~*, Of from over an TABLE X COMPUTER REAL-TJME SAMPLES Mode Delay Parameter E F t.a First Echo 0.68 0.74 G H 1.18 1,55 i 0.35 Sample (msec) Sample Spacing Delay (psec) 40 to End of This 20 NA NA NA 1.24 0.86 NA NA NA 1.30 NA NA NA NA NA NA NA NA 2.38 NA NA NA NA 2.48 0.90 NA NA NA Train (msec) Delay to First lnter- mediate (msec) Echo Sample Intermediate Sample 60 Spacing (psec) Delay to End of This Train “i :1 Delay (msec) tO Next mediate (msec) l“ter- Echo Sample Sample Spacing Delay (psec) to End of This I 00 40 I 00 200 10 4.58 1.14 2.18 3.55 0.91 6.50 6.96 6.?6 6.95 4.00 Train (msec) First Noise Noise S.ampie(msec) Sample Spacing 40 20 20 10 10 (psec) Last Noise Sample (msec) First Culibratim 6.82 7.14 7.18 7.09 4.14 7.20 NA NA NA 4.40 Pulse Sample (msec) Calibration Pulse SOmple Sp.aciW Last Noise N@g 40 Pulse (msec) Alldeloys NA NA NA NA NA NA arem~s.red 7.40 with respect tothel.tem"pt 4Op"lse. changes in these delays apply f.ard.at.a take” prim tO.]6 resulting in small shifts i.. Ititude -see Table Ill. .,j 1 42 I ! —. 10 (psec) 4.49 Slight December 1971, - interval as short as 8 msec, puting the amount of power p=(t) = 2 ~ =(t)i so that an estimate that is correlated ~*(t)i+f + 2 ~ i This expression imaginary of 2 allows clutter is purely real T(t if the clutter is not actually can be obtained by com- sweez-tO-SWeep - foo +sec)i The product ~(t)i is per fectl;. calculated, for the fact that the clutter i odd. from Pc(t) v*(t –100 wsec)i+i (8) . i part of Pc(t) only for of the echo power is calculated =*(t)i+i correlated from but is assumed sweep-to-sweep. to be zero. or.i? once fOr everY can be expanded in terms Here, The the factor Pair Of swee Ps, i.e., of signal, nois$, and contributions Ii(t)i:J* (t)i H = [X, (t)i+ I“(t)i+ TC(t)il . [;: (t)i+i (t)i+l+ ;:(t)i+i1 + ;; = Y$(t)i:: (t)i+i+ ;n(t)iT:(t)i* + ;C(t)i;: (t)i+i+ cross The terms and hence representing cross have an expected are uncm-r-elated from products sweep-to-sweep, xG,(t)iq echoes i (10) . are essentially ‘ identical, i.e., clutter subtraction scheme proposed (MTI) radar here represents The advantages to any desired the signals so that there value, (b) precise by analog means, is no possibility described Pm(t) of the digital –P=(t) = ~ :(t)i approach timing of the receiver [T*(t)i period of the Sampling and delayed gain differing out more simply implementation reflections to store are (a) tbe freedOm and (c) the current above can be carried a digital method of c~celing sweep. The operation Thus, (ii) which uses a delay line equal in length to the interpuke identically C=~c(t)i+i. i moving-target-indicator period =C(t)i . the previous thao delaying phase angles and noise samples out to zero ary objects, interpulse have arbitrary the signal also average -0 $= standard since i hand, the clutter The simple (9) and clutter Similarly, these terms x=(t)iIi”(t)i+i zIuc(t)il of tie noise, . xIin(t)i:: (t)i+l (t)i+i= i On the other of signal, mean value of zero. products from signals station from tO adjust the is eaSier to aCbieVe samPles are treated fOr the twO. as –ti*(t)i+.il 1 + ~ F(t –100 wec)i [T*(t – 100 w=c)i -~*(t –~oo wec)i+il (i2) i but this has not been done. Since the clutter 92 to i 76 km, it is not clem- that a. significant subtraction is performed amcmnt of computer 43 only over the height range time would be saved. The clutter signal subtraction spectrum scheme described above effectively imposes a weighting of the applied of the form (i3) H(f) = i – COS(2fif/fr) where f is the frequency frequency (425 Hz). displacement Thus, from the admitted the radar power will If the echo power was distributed in frequency this filtering echo power would be to remove Although at 40-psec vals, data samples arrive intervals, as listed 1.30< in Table X. (- 3Afi) at the computer for the second pulse transmitted ● mfr, m is any integer. to f=, the action of as a function of frequency. every IX), the range 0.68< and 2.48< [via This occurs than fr. 2.0 Wsec (Table over intervals, is estimated where comparable is much larger are formed t < 7..38 at 60-psec The clutter an interval selectively fo, and f= is the pulse repetition at f. be zero over in Mode I, but for Mode E the echo bandwidth nal power frequency t <4.58 Eq. (8)] only over products of the sig- t < 1.24 msec at 100-psec the first only inter- of these intervals. To estimate 6.50< level t <6.82 the average msec spaced in the receiver, For calibration An important covered taining data from the result error echoes the first up and down i 5 km. maximum lag applies Description profile This causes assume for the moment the wide -spaced the first pulse transmitted, t <7.40 msec the same operations was unfortunately was, in fact, no problems profile operations throughout delay which should be applied applied an average in regions features. to the other channel, of the desired of linear This error the 1969-i 975 to the channel conwith measurement power-profile was not discovered gradient, in time to program. to determine the echo acf over altitudes operations the range O < T < 380 psec [a shorter – see Fig. 1 (a)] for altitudes is facilitated that the close-spaced pair in channel 2 (v channel). acf can be calculated 7.20< are taken – Mode F at higher of computer six samples computer power Operations A further in the real-time in the analysis Mode F is employed from The i 00-psec attempt Computer of t. on the time base [Eqs. (6) and (8)]. near sharper any correction of these nine is taken as the mean noise pulse Pn+cal (t) in the range pulse transmitted that the measured at delays fOO @see earlier but could be important 3. An average for nine samples of temperature. existed by this report. is computed to be independent and noise -calibration in terms out but for samples period shifted 40 psec apart. the channel containing are carried the power and this is assumed of the sum of the noise for power noise ~n(t), by initially ignoring pulse pair is sampled In this case, i 06< h < i 66 km (Table III). channel commutation, i.e., in channel i (u channel) and the total unnormalized measured as Pm(t, 7) = ~ [Xi(t)i xi (t + T)i + Y, (t)i Y4(t + T)il i Pm(t, T) = ~ [Xz(t– 100 psec)i x2(t + T – 100 w=)i i + y2(t – 100 wsec)i TIIUS, the acfvs computed that the vertical in Mode F are assumed drift velocity ZOO< T <380 y2(t + T – ioo wsec)il for the altitudes to be wholly covered 44 real. Vsec . (44) This amounts to assuming by this mode is negligibly small. Pm(t, T) includes contributions due to signal, clutter. Pm(t, 7) = P~(t, 7) + Pn(t, T) + Pc(t, T) The clutter component is computed and noise . via (see Sec. 2 above (i5) for discussion of clutter subtraction technique) Pc(t, T) = 2 ~ [Xi(t)i Xi(t +T)i+i Pc(t,~) . 2 ~ [x2(t–400 psec)i +Yi(t)i Yi(t + 7)iti) + T –400 ~sec)i+l 0< T<i80psec i xz(t i + Yz(t – ~oo w=)i That is, the first factor sample on one sweep of 2 in Eq. (16) allows sweeps, i.e., is multiplied by the second for the fact that the operation sample T< 380 psec . (i6) on the next sweep. is only carried out for every The pair of only for odd i. The noise acf is measured Only the channel ficient 200< Yz(t + T – 100 w=-)iAl number containing of samples i O independent complete and 5 per sweep using samples the close-spaced is taken to form pair samples several acf ,s are computed for 100 G r < i 80 psec, echo acf measurements at the end of the sweep (iO time-base entire per sweep giving at t . 6.96 msec. is used fOr this measurement. noise acfts per sweep. when the pulse 75 noise starting spacing acf measurements A suf- To be specific, is 0< r < 80 psec for each profile of sweeps). m Pn(~n)i . ~ [xi (t + jA’rn)i xi(t + jATn + Tn)i j.o + Yi(t Pn(Tn) = 1/75 ~ + jA7n)i Yi (t + 3ATn + Tn)il (17) Pn(T”)i i with m . 9 for O < T G 80 +sec and m = 4 for iOO < r < i 80 vsec, ber. Equation (17) applies and hence there real. accuracy sible the first at these lags, to achieve others, t and T employed to the above equation amount of physical sample constructed Psec, p“(r) Pn(T) -0, to be of the second 30 December ● transmissions does not lend itself to height at these lags. before the values to increase While more An alternate this is carried i 970) for T > i 00 #sec. to the wanted one. of delay tabulated it is POS- often than the to this. In the F Mode, 3 km with respect sample may be obtained fn order are made near them. spaciog at Millstone information of the function. multiplications the desired are for altitudes The location FOr T >100 (i 7) assumes must be changed to O, 0, and i O minimum and first by repeating . Equation sweep num- XI). are shown in Fig. i (a) where of any pair. products. which the greatest with respect computed further crossing h > f30 km (and all altitudes tional correlations sample 1970 (Table additional equipment iS to average at altitudes zero this simply the timing approach in forming 4, 9, and 75 applying to 6 November The lags of the acf from are those near for O < Tn < 100 psec and ATn ‘ 20 Wec is no purpose The numbers for data prior and i representing out The addiThe values are for the first is obtained by adding the vahe of T to TABLE XI NUMBER N OF NOISE-CORRELATION PER SWEEP IN FUNCTIONS MODES F. G, AND FORMED H Close-Spaced Pair Pulse Spacing O to 80 F I G ; H ; Note: the corresponding (6 km before ; December ber i970 in order Figure i (a) ignores 120 t.a 180 12 (1) 4(1) I (1) o to 40 50 to 90 15 (1) 10 (1) i 06< group, to data taken h < i 24 km, Since the roles of the frequency for T .380 Computer to reverse Operations Mode G is employed maximum lag applies are obtained and thence is carried every’ 3 km 6 km over 30 Decem- For the higher to save some of the first every out starting of the measurements. psec fn order computer time. pulse for T ~ 200 #sec. In used to compute the full, in the imaginary part. information part and magnitude in this mode, Aside these basic differences it is evident 20 sweeps. contaminated correlation, differences, echoes. 221). Accordingly, computer drift information time is contained only for O < ~ < i 80 psec as the drift small, Also, The difference as no clutter between subtraction the real is needed independently. F and G Mode computer being those of timing, as given in Fig. i and Tables psec. b < 3i 7 km (Table the available to obtain the plasma part is calculated T <380 167< by unwanted clutter and, instead, as tbe real part becomes fm. 200< the range O < T < 380 psec [a shorter – see Fig, i (b)] for altitudes a,cf in order each sweep is processed from the remaining calculated clutter This imaginary is ignored 20 sweeps, then in channel 2 for 20 sweeps; of u and v in Eqs. (14) through (i 7) every the echo acf over altitudes complex content decreases every – Mode G is not normally is made to subtract are commutated in channel i for 20 sweeps, the roles to determim at higher This range of altitudes no attempt channels pair will appear it is necessary 4. height averaging, IX through XI. ,,I 46 , ‘1 functions no beigbt averaging the shift of i 00 psec in the location that the close-spaced :1 .] 350 (40) 5 (1) the full height resolution is not computed 104 (40) llx)t0190 1 (a) shows that correlation 197 O) between . this is taken into account. practice, ) in 150 (40) (1) 60t011Yl Figure to preserve the correlation > o to 40 For the lower group, hence, lo(l) delay. i 30 s h < i 66 km. Total 20 Sweeps 100 to 180 Numbers i“ parentheses .pply before 6 November 1970. — T (ysec) Mcde operaticms are largely and number the same, of noise acfis ,., The section) full acf is calculated (once agafn ignoring channel commutation as in the previous as Fm(t, T) = ~ ;(t)i li*(t + T)i (i 8) 0< T< 180 psec i Pm(t, 7) = ~ Re [~(t i. Commutation effect of reversing designed the roles F, P“(T) analysis to be symmetric A discusses Computer the effect Operations lag applies (i9) every frequency . (i9) channels has the 20 sweeps. function as the receiver bandpass was This has been found to be not completely to cause an important of ignoring between ‘r< 380 #see systematic tbe imaginary distortion in Im [ ~m(t, r)]. part of the noise acf. upon the data function to determine Computer III). major differences Thus, the equation = ~ and measurement Pc(t) .2 profile III). complex cor- G and H Mode com- height averaging, IX through XI. to retain z over lines arise the range 55< h <140 of E Mode processing, due to the transmission of signal-plus -noise processing in the computer for Mode 1 corresponding 51 (t)i + of alternate and noise-only of sweeps km but several on op- in groups a pair of sweeps of four on each to Eq. (6) for Mode E becomes (20) s2(t)i i even to Eq. (8), ~ [:(t)i psec. being those of timing, This necessitates purposes i odd and, corresponding 5i5 km (Table and the full, O < r < 190 along the general sweep. subtraction Pm(t) the echo power follows sweeps on a given lags in Fig. i and Tables Most of these differences on alternate 2i 5< h< at these altitudes, all the differences as given to determine exist. for O < T 4 i90 psec [a shorter for altitudes - Mode 1 1 Mode processing at a time for clutter frequent y. the same, Operations frequencies Eq. (i8)l acfts calculated Mode I is employed (Table frequencies [i. e., the range - see Fig. l(c)] need not be considered are largely of noise the echo aci over altitudes is measured puter operations and number – Mode H at higher As in Mode G, clutter posite and real frequency. 200< psec)i pulse pairs to be a purely about its center exists Mode H is employed maximum 6. + 7 –iOO results. 5. relation %t of u and v in Eqs. (i8) is assumed Enough asymmetry Appendix psec). of the C1OSe- and wide-spaced As in Mode true. -100 :* we have (t)i+2 + v(t)i+i ~“ (tIi+31 (i - i )/4 an integer (21) . i AS with Mode E, only the real altitudes calculated. Clutter is subtracted at all in the I Mode. A separate version to a frequency outside Tbe integration rately part of Pc (t) is acma~y from of the Mode I program the receiver was allowed the others. passband to proceed In this way, was developed during four as normal, the recovery 47 except in which tbe transmitter consecutive sweeps that these sweeps of the receiver was switched out of every were stored could be monitored, eight. sepa- albeit at a cost of half the integration time. the receiver recovered; had essentially not attempt c. to monitor REAL-TIME 1. the receiver COMPUTER At the termination are first pose of providinga computer 2. and P “+Cal(t) of the program, these operations that does power P,(t) = Pm(t) ismultipliedbyhz, Error integration-time temperature process, Pn+cal(t) P~(t) a modest form. Also. of the prfmary the results. These has determined Pm(t), data are saved on magnetic only. P“(t), tape. The mean receiver noise level t. as (22) . (base 10) is then taken, (Appendix a height profile B). values and the result of tbe logarithm are calculated Signal temperature temperature from T~(t), T”arecalculated is arbitrarily of relative scat- signal-to-noise and noise-plus-clutter from the relationships =PS(t)/Tn [P”(t) T “R(t) over c(t) and noise (system) T~(t) X. of these power considerations Tn& (t), for the pur- performs t, and the mean noise plus calibration-pulse value of 5 to yield estimates The”, in detail. the computer is then calculated –Fn-P andacf is employed meaningful presentation tonthe line printer Pn(t) over the logarithm to a maxinmrn power. in more t given in Table by averaging signal are in binary. the computer visual power format 1 are now made foroutp” is estimated normalized a real-time Eand the integrated to an immediately discuss by averaging period, themselves the information of the integration ~“ is estimated Ionospheric tered version to. BCD and binary copy of these resdts, provides at delays calculations A mixed display Output –Modes At the terminaticm Several printed t.a reduce sections Computer integration tape. is in BCD and the results oscilloscope The following P~(t) was reverted of 60 km OUTPUT onto magnetic real-time amount of processing Fn+ca, it was found that by an altitude the original recovery, of each experimental written which the data heading level hence, General results Pc(t), Using this program, = +Pc(t)l T n (23) Altitude h is calculated as h=i50(t where T is the radar -~) pulse length and D is the delay introduced t, T, and D are measured of header information standard-deviatiori (24) km in milliseconds Tn. and lists error for this equation. TS, and scatter of h, TnW, bounds (relative logarithms) 1 48 vs t. by the receiver The line printer power (-0.03 listing msec); consists plus its upper and lower l%eos.illoacope displays Pm(t) 225 150 75 I o~ > POWER [ relative (a) and Pc(t) 3. Several power T) correlation t and T listed are then calculated for instrumental The line printer and P=(t, 7)/P~(t, directly by Eq. (25). ] printed. This latter measured 4. Ife[~m(t. HI and X. listing by averaging of a The best-fit for only. Pn(t. r) over in Appendix A. information VS T. (described April t. The signal theoretical Altitude tape. noise power these tables, in Appendix t used. h is computed plus tables of P~(t, ~)/P~(t, [For Figure the values of unity, h of P,(t, O)/~n(0) is (S/N). 9(c) is an example of a function is also shown (see Sec. V). G and H process, (Mode the computer G), ,These data are saved on magnetic The mean receiver 49 and tape. Pn(t, has determined T) at delays Several noise pewer of A) and not as given to a maximum 1974 by a list vs O) vs T for each T > 2oo wsec only. The mean receiver (25) of header smearing starting of the integration r)] Pm(t, ‘r), Pn(t. 7), data are saved on magnetic T) as discussed consists P~(t, 7)/P~(t, for output on the line printer ,_ has determined These a list of h2P~(t, O) vs h, normalized Output - Modes At the termination Tables 9(a) is an example the computer III and X. – Pc(t, smearing list was replaced Mode F acf. T < ~90 psec, of 5. as ‘P”(T) for instrumental Finally, displays Computer Mode I. to a maximum Figure O) vs b and T, and a list of Pn(T)/Pn(0) P~ are those corrected The oscilloscope vs h. I.q) Fig. 9(b) is a Mode I example. process, in Tables is estimated function ~n(r) p~(t, 7) = Pm(t, 7) via Eq. (24). normalized are now made for output on the line printer functions and corrected 10garithm) by Mode E; of the integration at delays calculations correlation obtained power, J 5 Output - Mode F At the termination and P=(t, of scattered pOwer (relative profile Computer (b) Logarithm and scattered scattered-power ‘POWER (relative 109) Mode E. Fig. 9(a-b). ( 4 ! . 50 2 ~m(t, t and T) fOr T given in calculations are now made correlation function ~n(r) O) 1 1 1 I 1 * 1 18 SEPTEM8ER 1974 1735-1747 UT 121 km ALTITUDE Ti=404?16K 0.1 ‘. T, = TI ASSUMED ,1, = 0 ASSUMED THEORETICAL 0.: FIT -a r (F*..) (c) Mode F. , * 18 SEPTEMBER 1S74 1748-1756 UT 197 km ALTITUDE T,=967f21K T.= 14691 18 SEPTEMBER 197, 1756-1800 UT 395 km ALTITUDE T,=1135*36K 1.0 T;= 1803* 40 K . . -65+ 13m/, ec 0,s 19K ● 54% O“COMPOSITION ASSUMED MAGNITUDE x PHASE ~ 0,6 g ● THEORETICAL MAGNITUDE g x PHASE FIT :0.4 5 q LINEAR g FIT 50. ? x $ . 8 / 0 xx .“ LINEAR / -0.2 “02L--...J m - -0.4 b 0 ,60 2+0 320 4oo 0 , 40 so 120 T (psecl r (pee) (d) Fig. 9(c-e). (e) Mode G. Measured correlation Mode H. function and fitted theoretical 50. FIT function. 240 . is estimated power _,_....,,. . _, by averaging correlation Pn(t, r) over t. The magnitude function ~~ (t, ~) are calculated M(t, T) = l~~(t, ,)1 = I ({w [~m(t, Im [~~(t, T)l}z + (I~~m(t, T)l)2)i’2 (26) –~n(T)l T)] I @(t, 7) . arctan M(t, T) and phase @(t, T) of the signal as (27) Re [~, (t, r)] choosing @ to lie in the quadrant *tr/2. Im [~~(t, velocity ~)] is assumed at each altitude to be zero for T >200 ~sec (Mode G). @ is then interpreted as a via (28) where dq/dr center of the remaining is taken to be the slope +6 points z of the line passing through O(t) = O, T = O. and the weighted TC)(T) Mz(r) &+=r (29) x r2M2(r) 7 This equation calculation approximates the uncertain of q@ (Appendix M is corrected y in $ (u ~ ) by i/M to avoid the complicated rigOrOus in Appendix h is cal- B). for instrumental smearing as discussed of header information A. Altitude culated via Eq. (24). The line printer fm [~~(t, ~)]/M(t, listing M used are those corrected ized to a maximum consists O) vs h and T, plus a list of ~n(T)/~n(0) for instrumental of unity) vs h are printed. liSt Of M(t, O)/~n(0) VS h (s/N). acfts. Fitted theoretical functions Figures effects. plus tables vi r, where, Finally, This latter lists of M(t, ‘r)/M(t, O) and again, of of v and hZM(t, O) (normal- list was replaced 9(d) and (e) are examples the values in April of measured 1974 by a Mode G and H are also shown (see Sec. V). 54 L ——. ,.. - . . . ...-. ”.— ,-.. ..=-. v. A. GENERAL The real-time and clutter to analyze sity Ne, V& magnetic power these [Nz], conce rni”g atomic munity. oxygen tape, [0], the neutral and the neutral and a library The following alyze F Modes (See, II-H). analysis fourth, realized containing 14 heights, by averaging that as the first degree 7 heights HI), these ventional to the 2-pulse altitudes created information. (secOnd, fOurth, to an- were are created heights com- data. was written mea- in the [second, It should also be F Mode are at 3-km intervals altitudes are saved whereas and sixth) have only noise estimates, not additional measurements. Mode E - first ANAI,YSIS Routine Analysis calctdates the POwer PrOfile, h vit112q. (24) and P~(h)via via a radar eq”aticm which, vertically directed p:tirabOlic antenna, 0.76 PtAI,cT P8=———— i6n 1 I same three independent and that due to independent Mode E isusedtomeasum related additional nitrogen to the scientific program These altitudes. d rift these a munbe.- 4970 only *I heights missing den- vertical of this analysik applied Tbe analysis to 30 December in the i4-height from for distribution procedures tbe three the two neighboring of independence, POWER-PROFILE program but Prior F MOdes, and sixth in Fig. t (a)] do not yield The results is used such as electron the abundance of molecular Tn. the analysis noise, program ions [0+] /Ne, also calculates is available for completeness. For the 1i -height program echo-signal i. describe here atomic namely, temperature of such results se ctiom Ah = 6 km at best (Table a small atmosphere, of signal, A separate of the ionosphere percentage The program Vin. consists delays. quantities Te and Ti, frequency One note will be entered sured the physical and ion temperature collision experiment of time base and correlation data to determine and ion-neutral on magnetic outPut frOm a 2-Pulse -t:kpe as a function electron of parameters B. ANALYSIS frOmwhich Ne(h) is deduced. P~ and electron Eq. (22). in the case of a moncmtatic radar system The analysis density Ne are employing a con- maY be written9 N u > (30) 11* where j Pt . peak tr:msrnitter A . effectivv As discussed power, antenna aperture, L . waveguide and other losses, o . scattering cross in Ref.9, provide, d Te/Ti section and per election. & 3.0, the scattering cross section per electron may be written a. “e (i+azl (f +Te/Ti+02) (31) (32) 53 De = 69~e m In Eq. (3i ), Oe is the radar cross 28 2 scattering, is u= = i O m (Ref. 9). From these expressions, (33) . section for a free electron, which, in the case of back- it may be seen that Ne (34) P~ = (4 + a2) For typical a 2< IVe and Te, (i + Te/Ti De< 0.03 and may be ignored, This approximation variation Equation Mode E, ignoring generally reasonable smoothing Te/Ti yields The constant profile constant from values simultaneous is in megahertz. vertical-incidence ionosonde; faulty local readings. ‘7 through them. terpolation at the midpoint Statistical Te/Ti From uncertainty applied in determining iza.ticm should be applied. frequency to derive foF2 between of the Te/Ti the absolute to determine measurements of the F2 -region from Ottawa, at-e made using an on-site Wallops Island, when available and a curve are read off at half-hour values this (36) have been considered values in the power is used to normalize measurements, the still Billerica to correct is drawn by hand intervals. Linear in- tbe E Mode Ne profile This is resolved profile over the eight consecutive vertex of the parabola larger plus the use of discrete-height the exact point on the umormalized the peak and the ummrmalized interpolation Application time. to them, marks (the smooth- via are plotted vs time, these half-hour uncertainties frequency measurements this smooth curve, un- a 2- or 3-point m-’ (Massachusetts) of its integration corrections critical All these readings in time between Linear It has been the practice of the critical X 1040 similar (Mas sacbusetts ), and Maynard - basically, is used. is profile. may be established These varia- to remove .wmmtbing is applied program of Ne mea~urement measurements = L24(foF2)2 profile profile data are edited a smooth Ne profile. demsity profile. NmaxFZ for Te/Ti As the power the temperature of the analysis data analysis. F, G, and H following in Eq. (35) must also be determined from the unnormalized through foF2 density profile Modes < 4) and some profile to the heights an ummrmalized which the peal+ density where Te/Ti of proportionality N ~axF2 I 0.5< the power from must then be made for the 2 -pulse mode measurements. profile, has been used) to try to maintain correction ‘! between than the Te/Ti (requiring and allowance to correct has changed during development smoothed density altitudes, data are obtained the time difference values ing algorithm tiese temperature much smoother (35) but Eq. (35) is used in the present (35) shows that it is necessary These Thus at our wavelength 500 km down at higher of a (Ref. i), tion to find Ne (h). below about 500 km. yielding h< breaks “ i cm for heights W* altitude + a2) heights the desired Ne value by fitting a parabola ““certainties measurement Ne profile to which foF2 normal- to the logi o Ne (unnOrmalized) fo~ which the sum of these logarithms point of normalization, corresponding 54 in the cause some specifying to the density maximizes. The the height hmax of Nmax F2. Occasiomlly, bad points (due to satellite occasional instances At the higher ground clutter, of improperly normalized Mode E altitudes, quite scattered. profiles the signal could be more accurately this region. A straight to decrease with altitude decrease, represented line is fitted instead height of measurement. (decreasing scale Mode H altitude) The lowest altitude in others, of a parabola able results. casionally, Efforts or the parabola No profile vertex is below is applied from the real data profile and not replaced this high-altitude f~r the data just above highest to the smoothed fit, but by the fit. smoothing scheme to yield unreason- successful, md, oc - results. to RASEM jwt described except for m rtai” and normalization. data, run-to-run by a weighted i+n .2 Mode E analysis correction, to reduce P8 (b)i is replaced ~~(h)i the maximum fit is substituted and 546 km (i. e., appears fit also takes into account the data up to 50 km below to the routine temperature smoothing of the data i“ Analysis in time is applied power smoothing is identical of data smoothing, smoothing signal profile Mode E - RASEM analysis versions are retained echo causes increase if the slope of the parabola made to cull out these bad points have not been very a distorted RASEM facets a satellite exponen- high-altitude fit to the logarithm for which the parabola smooth transition data in this 50-km interval Occasionally, 2. The parabola height to insure the original so weak that the data are it has been felt that the trw is set at i50 km above the F2 peak in some program this lowest becomes by a parabola height) and occur. sometimes ionosphere N , profile is expected to decrease nearly h, with its scale height increasing only slowly as temperatures and mean ion mass and the pull of gravity densities etc. ) cause this method to go astray, As the topside with increasing tially echoes, but becawe fluctuations. time average the integration time is short, For the ith run in time, the ~~ (b)i WjP9 (h)i . j=i-~+n z j.i-n w. J Wj.2(n–lj–il)+l i+n 2 This provides a triangular been employed. possible (37) Wj=(”+p+nz j.i-~ weighting This smoothing weighted average over process 2n + 1 points. truncates that can be formed Values n = O, 1, 2, 5, and 10 have the data series is for somewhat, since the first i . n + i and the last is for i = N – (n + 1) for a total of N rons. As no temperature perature data are gathered dw-ing a ftASE M expe rinmnt, correction is modeled: (a) Te/Ti = 1 (b) Te/Ti increases linearly to 2 (c) Te/Ti decreases linearly to i.2 h< the density-profile 930km i30< 300< 55 h < 300 km h < 2000km , tem- As in the regular ization to ionosonde RASEM Mode E analysis, readings. measurements to determine are made much more I-”n-to-run. Thus, to maintain fit determines the ionosonde the point for profile readings frequently Nmax F2 [Eq. (36)] for each profile, This should serve dar. a parabola However, (e g., are taken every every 2 min.). while Rather than attemPt we make use of the inherent log, ~ NmaxF2 normal- half-hour stability - log, ~ p~ (hmax ) approximately of the ra- constant (38) r . logi ~ Nmax F2 – logl ~ p~ (hmax) is formed for all the available As the radar line is fitted electron sensitivity profile values a magnetic to make the program 3. the run (i.e., with time, every storm a least-mean-square the time variation i+ (Te/Ti) I with the program of this parameter F(t). The (h) (h ) Inax I These (39) + ‘(:) program. when the FI -Iaye r has been pronounced It is then possible r and hmax. -st raight . using + 10gf O I + (Te/Ti) in summer). half-hour). events for tbe FI peak to be mis- are sufficiently infrequent that at- cope with them have not been made. Mode I The Mode I results Mode E results, density profile to an Ne profile that it is assumed to the E-region usually and magnitude are converted except made with respect contains critical frequency making the average of power and tbe normalization in place of the F-region. peak, Accordingly, as the value in much the same mariner as the routine that Te . Ti at all altitudes no well-defined of normalization. 125 and 135 km is employed N slowly t is then obtained have been encountered following throughout b and t can be provided” by the processing taken for the F2 both in obtaining tempts to drift = 10gi OP~(h) plots of “log, o Ne vs Difficulties (usually of NmaxF2 of r to yield at any time logiONe(h) Contour values is likely to the available density from the difference it difficult of the values corresponding is The E-region to decide the location of 10gl o p~ (h) between to E -region maximum density E. max The Ne profile clutter the center wise, results subtraction frequency subtraction high collision carried an appreciable in the notch of the filter the echo amplitude frequency. becomes fraction function correlated That is, for the influence H(f) [Eq. (43)]. over the echo power of the of the echo power intervals lies at Stated other- as long as i O msec PI remaining after clutter becomes PI . P~[i where i“ the case of Mode I, be corrected This is because and, hence, at low altitudes owing to the very must, scheme. p (TC) is the correlation out. (40) –p(TC)] D. T. Farley over (private tbe interval (1 O msec) communication) autocorrelation function in the collision-dominated of the spectrum equation of Dougherty and Farley at which the clutter has pointed case, subtraction out that the expression based on taking the Fourier (Ref. i8 – see Sec. C IR1ow), is for the transform becomes - T/* P(, ) . e Tn = (4i ) (42 ) r(2rAfi/A) 56 9 = vin/(2nAfi/X) Adopting k = 0.68 m, mi . 3i amu, yields P(TC) . exp[-i820 In order to use this result, K(h) = i – for model U.S. (43) . atmospheric Standard Ti/vin] we have computed (45) conditions. i962 Adopting one computes be employed from in Table and [N2] (h) from and the simple the CI RA i965 expression (Ref. i 9) and Zi of Wand and Perkins, Vin and [Nz] x iO-i5[N2] K(h) as listed by Banks differs Ti(h) (Ref. 20) models hased on the work of Banks, 22 relating .0.9 the function p(Tc)h Atmosphere v in (44) . (46 ) see-i XII and plotted that of Dougherty and in Eq, (44) is half that given in Fig. 10. Farley by The definition a factor of 2 so that of Vin employed the value of vin to by Eq. (46). TABLE X11 COMPUTATION OF THE EFFECT OF CLUTTER ON THE MODE I ELECTRON DENSITY Height (km) N(N2) vi” (m-3) (s..-l) Temperature (K) K(h) Source 120 381 4.01 x 10’7 1.80 X 102 I .0 Ref.19 110 263 1.62 x 1018 7.23 X 102 I .0 Ref.19 100 213 8.18X 1018 3.68 X 103 I .0 ReF.19 90 186 4.96 X 1019 2.23 X 104 1.0 Ref.19 80 186 2.96 X 1020 1.33 x 105 0.92 Ref.19 70 219 I .41x 1021 6.35 X 105 0.47 Ref.20 60 243 4.95 10’1 2.22 x 0.18 Ref.20 The final electron loglo density Ne(h) profile x is therefore = loglo[P~(h)l obtained ~~ (h) is the aforementioned average 106 from the expression – 10glo[~~(h)l + 10gl ~ Nmax E – logio where SUBTRACTION PROFILE value over (47 ) [K(h)] the interval i 25< h < 135 km. 57 —.. ..——... ..—.,-.---.-.,,- -- -. 100 I I I I I I I I ~ 90 — ~ :. 80 - + $ y 70 - 60- I 50 I 0.2 o I I 0.4 I I 0.6 I 0.8 I I 1.0 K (h] Fig. i O. Altitude variation of fraction of echo power remaining subtraction operation has been conducted in computer. As discussed in Ref. 34, the results obtained using the Mode I program doubt at altitudes below about 75 to 80 km .cmdmay be in error difficulty to be the presence appears which may be clutter returns tudes causes the Ne profile as illustrated in Fig. 9(b). first altitude creased. from this is usually of the filter Each measured the real [Eqs. (25) through mental effects analyzed appears the expected plotting larger to be increasing than a factor at these low alti- cw-ve program below about 80 km, the plot is halted at the as the altitude to 80 km so that the correction applied is de- to the results of 2. ANALYSIS (27)]. independently Phase form. are calculated (Appendix M by i/aO inEq. estimate phase estimate The statistical for purposes of the datauncertainties of weighting program forinstru- see Sec. lV-C), in these and subsequent first and -4 but is real parts, temperature is negative is greater (statistical in error to calculate [Eqs. (28) and (2,9)] to determine aveloc - instead of the approximation o . i/M for weighting, i.e., + ‘$ Lags are eliminated from this fitting procedure if the data fluctuation than half a radian are systematically failure as in Sec. IV-C-4 of o (29). trum, i.e., at this point in the analysis printout program in Sees. IV-C-3 Analysis using now the true estimate variance The analysis B). The phase @ of an acf is analyzed replacing of all other data. and phase of an acf as described M(T) is not corrected in its uncorrected and phases calculations 2. acf is analyzed part or magnitude (as was done to obtain a real-time directly magnitudes, ity, from as yet unidentified, observed Gene ral calculates error close [Eq. (47)] is rarely CORffELATION-FUNCTION i. progressively in the 1 Mode profile The cause of the mechanism, power clutter are open to serious even higher. scattering The additional below 80 km at which the density In practice, for the effect c. of an additional aircraft. to depart Thus, after due to failure due to near-zero (unreasonable signal-to-noise) velocity). The velocity to account for the asymmetry the imaginary part of the noise acf. 58 or the estimates of the receiver The effect spec- of this neglect is examined in Appendix discussed. Ffgu=es 3. A, where the velocity 9(d) and (e) show examples Magnitude Analysis The magnitude - Library part of the acf is not measured) acf magnitudes. has been felt that additional Pin) library general The library independent is used at lower A coarse-grid library search for closer by the fine search. scrutiny Table XIII gives given values The Te/Ti of normalized in Table collision are direct Farleyi8 causes Debye error plasma); causes occurs other errors made in order These functions a very in Te/Ti rn by the coefficient parabolic or (Tf slight Tbe of abcmt the minimum for each search parameters are tai - for the coarse The “in grid values The area under investigation. and are stored $ grid values also are acf!s formulation stores -i. for O; 4 percent assumption correlation delay was grid point, rn as de- 240 per acf in steps of 0.05. values; the functions are calibrated To obtain the lags needed smearing smeared by the equipment (Appendix between such that the comparison is addressed 9(c) through The for all ion species. and increases acfls method The lengths are assumed in Iag simply for direct by dividing comparison with the in lag is used. is mathematically tYPe Of normalization a simple parameters. for 0+ plasma, ionospheric The library cases assumption This zero-electron-mass would suffice and Tbe zero-electron- in all determined i87 points per acf for each Te/Ti were For each comparison error at zero These for the limiting 24 has developed as a function of normalized for reasonable of r in Eq. (42). interpolation field This zero -Debye -length (-2 percent rn in each case starts lags are needed. of a magnetic of the acf (see Sec. E below). than 1 percent. are stored acf magnitude is stored. 23 and by Dougherty given by Farley Moorcroft systematic with the data, Figures frOm the data. space to find the proper in Eq. (43). ratio. determination To acccmnt for the instrumental normalized XIV. Vin)] a theoretical analysis are not more are sufficient if greater smeared it and the estimation centered grid and step sizes + as defined without the presence after The library for each $ grid point; of these a libra~ altitudes. and number of grid points wed underestimated. that one library theoretical fined in Eq. (42). to be zero search of the physical from the equations plasma of this error mass assumption greatest .deduced at higher regression tbe entire are given in Table length and electron-to-ion-mass T= to be systematically for correction data, Te/Ti) calculations for a monoionic of small library library limits of variation frequency with variables); XIV. At each grid point [(TV functions Te/Ti) covers step sizes, grid values comparison B. The grid and ranges psec in Mode G as (two independent by a fine-grid search the Ti g=id limits, fine searches. in terms The coarse through could not be reliably a (Tp is followed grid. to the expected parameters in Appendix point in the coarse lored is analyzed the 2-parameter is presented F and for T >200 is 3-dimensional altitudes, method used fm. performing by this technique chirp is also determinations. M (set equal to the real part in Mode of theoretical errors of velocity frequency Search the imaginary A (Td bias due to transmitter of the data, by the computer before in the same manner and a theoretical closest in Appendix (e) show examples acf, comparison in which the data A). a measured yields each library acf, fit in a weighted, the data are multiplicatively least-squares sense. B. of theoretical data. 59 acf magnitudes fitted to measured This TABLE X111 Ti GRID Fine Gridt Altitudek M&e Came Grid (8 grid ~ints) Lower Limit 106 lW 12 78 522 150 450 12 522 572 I 50 450 12 12 78 128 200 500 128 572 12 18 128 92 572 758 200 200 500 500 284 650 124 130 18 92 758 200 18 192 858 300 650 750 136 18 18 192 858 8CL9 30+3 250 70Q 858 3C0 750 350 400 450 800 850 900 121 142 Limit Lower Limit Upper Limit 750 18 18 18 242 292 18 342 9m 958 1038 342 1(XJ8 1058 450 5C0 900 950 18 392 492 492 420 1158 1158 600 600 1050 18 30 1530 600 160 164 167 18 18 182 197 212 227 Upper 142 192 148 154 G (13 grid ~ints) Step Size 115 118 I (K) (km) 112 F LIMITS 1Q50 242 30 420 1530 600 1350 1350 257 30 420 1530 60G 1350 272 287 302 30 30 420 42o 1530 1530 6043 6CiI 36 384 600 317 36 384 1716 1716 1350 1500 600 15C0 215 36 384 1716 245 36 384 1716 600 6W 1500 275 36 384 1716 6C0 1500 305 384 1716 600 1500 335 36 36 1716 600 15(XI 365 395 425 36 36 36 384 384 384 384 1716 1716 600 600 1500 1500 455 36 384 1716 1716 60il 600 1500 150U 1350 I ,! H ●See I Table tF1.e-grid 485 36 384 1716 600 15CXI 515 36 384 1716 600 1500 Ill regarding altitude search uses 13grid toextendkyond 150Q variations. Wi.tscenterd o"came~rid-s~rch grid limits). 60 optim"m fit(ktnotall~wd I TABLE XIV Te/Ti AND NORMALIZED COLLISION FREQUENCY $ GRID g 0.75 0.87 ~ 1.12 1.25 1.37 1.50 1.87 2.00 ~ 2.25 2.37 ~ 2.62 2.75 3.oiI 3.12 =5 3.37 3.50 ~ POINTS I .62 =5 + 0.00 — 0.05 0.10 ~ 0,20 0,25 ~ 0.35 0.40 0.45 0.50 0.55 Q&T 0.65 0.70 @ 1.00 1.20 1.40 — 1.60 1.80 2.20 — 3.00 & Ccarse-grid-se-srch Fine-grid points are under] ined. search uses nine cc..sec.tive search optimum Fit (but not allowed Different eters altitude expected problems Table ranges are analyzed to be influential in incoherent XV is a summary in different in different scatter regions. data analysis, of the analysis grid poi”fs centered to extend manners Following these process ob.o.t cowse-grid- beyond grid I imits). altitude in different according to the physical a discussion regions regions param- of ion-composition are discussed to be referred in detail. to in the next few sections. 4. Ion-Composition The major are 0+, Problem ions likely ’to be encountered NO+, o;, H+, md occasionally depends upon Ti and mi cmly as a ratio, scatter in ttie altitude sporadic-E range of these metallic im identification ions. 2-pulse is not possible from acf (e. g., 02+ at iOOO K and 0+ at 500 K would be indistinguishable). could be a problem in certain day and NO+ and O; outnumbered species circumstances, hy night, by tbe metallic expected, e.g., or in a sporadic-E ions. and tbe major However, near 200 km where layer in general. a given incoherent we have mostly 0+ by ions are well- one knows the identities of the relative description Such identification in which the ambient task is the determination experiments As plasma-wave of the major composition if a mix- ture is present. When the plasma contains tions f~om each species ratio is a factor so closely proached entire this profile (e. g., of very seen in the incohercmt shape that essentially from a given measured Hill data, of data simultanecmsly type of analysis different In previous spectrwn mass scatter 0+ - 02+ - NO+ or Fe+ - O; the monoionic detemnination with Millstone profile can be easily of 2 or less approximates composition two ion species (e. g., 0+, H+), contribu24 spectrmn. If the mass – NO+), error-free (m- acf).25 however, data are required This quality data is diwwssed in Sec. D below. for is not ap- papers,~ i >i 2 we have shown bow znalysis can be used to deduce this composition. to tbe 2-pulse the spectrum Application of an of TABLE XV SUMMARY OF ASSUMED IN Height MEASURED PARAMETERS ANALYSIS Range (km) 106to AND THE DATA Assume 118 Te=Ti=T Measure Ti, n V. m ~=1) I l18t.a 130’ Te=Ti=T Ti n ~=o v ;“ =“0 130to 166t Ti=T v in 166t to 225 n from Eq.(48) Te, p from Eq.(48) T=, P, Vd =0 Ti=T n vi” =0 ~=1 225 to 515 T=, Ti, vi “ =0 92 to <700 *See text for discussion of special tThisaltit”de actwl N an.alpes at these altitudes. break corresponds tOthech.ngefrom height wosa little higher forsome 62 e eorlydat~ Fto GMcde; (see Table the ill). Vd At the lowest in the daytime, similar altitudes except covered and unknown ~ priori, uncertain within a +3-percent generally the more to errors of +2/-5 percent, Above mass spectrometer i.e., possibly NO+ and O; region conditions rocket determination at even higher of ion composition too low for useful ions takes place creating (giving of usable very low Te/Ti (> 40 percent) nighttime temperature Collisional In the lower cases of midlatitude range measurements night) - a sufficient number of error that extremely in unusual ions to be more important Region of transition mixtures ratios exponentially may well exist in analysis, from is to atomic Both the rate of as an ion descends transition in tbe volume could cause significant assuming density molecular of the F2-layer. The ion-composition this occurrence as tbe electron from the is, in fact, so responsible distortion a fixed composition), for of the but this for these data. effects in the acf’s during the day, but may comprise of the ion content near 500 km at night. Its neglect may cause overestimates. - h <148 km by the influence the shape of the incoherent scatter spectrum is significantly affected 9 At the ~i~~~tone Hill radar wavelength these effects of ion-neutral collisions. are consistently seen below sional effects from fore 6. here precludes in the li8- to i30-km reliable Equality is significantly in the previous region, is a significant Te-Ti also, - ii8 determinations. higher than Tr i30 km, colli- Analysis for h < i i 8 km there- An ion mass of 30.5 amu is assumed. determined no consistent, there is no evidence find Ti with Te ,= Ti and Vin . 0, assuming Above 130 km there is no evidence from measured data. < h < i30 km section, difference below Ti and vin. of Ti and vin values Region as high as i30 km, though the small- in the region and seeks to determine t 1 shows an example Temperature consistent, In addition, Hill data that Te Ti = Te, As was mentioned that there ’138 km and may be significant are undetectable. Millstone assumes Figure I i ion E-region, ness of the effects I is uncertainty. from available It is possible the region of 0+ decrease returns; portion 5. data. on the sharp underside H+ does not seem to cause significant serious further shows that the altituffe 200 km at night is moot, sharp 0+ gradient. has not been addressed a sizable to ca~~ ) may cause molecular At night, region and the lifetime a very altitude spectrum problem however, error. This summary scatter below sharp that a wide range of composition the lowest systematic day) to 260 km (summer Millstone acf measurement. in a narrow diffusion F 2 peak, NO+, is are always altitudes. The question ambipolar composition data in this region within this range to cause a 3- to 100-percent from incoherent over are so of a mean mass of 30.5 may lead of 0+ ions may appear but that in its extreme times masses the NO+ - O; from but assumption a larger molecular a mean mass of 31 amu. experiments. i 30 km (winter (e. g., trough passage As determined and NO+ ions predominate to 0+ ions has been studied by Oliver is variable, from O; These be distinguished. proportion ions may exist at certain temperature cannot range if one assumes data from the range extends minor layers. abundant of these species, from of this transition in sporadic-E the temperatures i 30 km, a sufficient This transition experiments, occasionally (3 O and 32 amu) that they variable by the 2-pulse below 130 km. from reliable Millstone This altitude an ion mass of 30.5 amu. I 63 .in information is obtained 27 Hill data9 or theory range is analyzed to --EEL c-w -+’- . T; OR “in X T; FROM (T;) FIT AsSUMING ❑ 11 FROM (Ti, T#l) FIT ASSUMING Vin = O A 1. (T;, Tfl;) FIT ASSUMING Vin = 0 FROM FROM (Ti,”ln) FIT ASSUMING Te = 1:, Te = T, v;n = Q — ~ — + & -’0- + —, -+- I I ~ .500 TEMPERATURE 0 [K) I 500 I I 1500 I I I 3500 COLLISION FREQUENCY (see-l) ++ k++{ ,A*f~;, , Fig. 43. Lower F Mode results on 48 September T, T. I 2500 1974, 4735-1747 UT. T. “+ 2= — ++ *+ ,00 0 500 1000 1500 TEMPERATURE 2000 [K) 259Q 1 50 PERCENT 0+ Fig. 12. Profile analysis for data on 18 September i974, 1731-1800 UT. Thermosphere results yield: T- . i052 * 17 K, s = 0.0262 * 0.0041 km-i; T420 = 408 * i3 K; and [0]400 = (2.46 + 0.99) x i08 cm-3. 64 I ~500 1< Some additional assumptions formed calculations 7. Te . Ti and Vin = O, i.e., i t shows results Ion-Composition Te and Ti differ this region, species vi”) Transition ionosphere. and a (Tp proportions Tbe temperature this estimate to test the adequacy Te/Ti) search 8. 0+ Region of the has also heen per- -225 Because search obtained is thus performed. from analysis are systematically is obtained. 1“ znd atomic-ion of a measured under the assumption too low, Section acf of a puce 0+ but can be corD-2 below explains to Te and Ti are then applied. significant in this region, amounts of mOleCUlar beyond 26o km at night, and a (Ti, Te/Ti ) library ions may on occasion there is the possibility The H+ abmdance search exist up to 250 km of underestimating is also assumed true at night and can cause the temperature increasing Figure km library profile to be important in these height intervals. is not always data. km < h during the day and possibly Ti and Te/Ti so obtained on measm.ed amounts of both molecular- range are analyzed and bow corrections Only 0+ ions are assumed is performed. Te/Ti) cannot be reliably values performed h <225 of significant of the ion-composition derived fs analyses -430< and a (Ti, all acf, s in this altitude when m estimate the error Region in this region, and, as their rected from these various the ion content may consist (Sec. 4 above), This a (Tr on data in this region. Figure how have been made for this region to be negligible. to be overestimated above 400 km, with altitude. i 2 includes some sample results from this height region from an analysis of mea- sured data. D. NEUTRAL-ATMOSPHERE i. DEDUCTIONS Althcmgh the incoherent scatter can be made concerning properties are equal. as N2isthe equal below “in, concerning of data and invoking 2. estimate Tn, Ba”er”~ [0] &z’ measures be assumed additional ionospheric below km, i30 Also, equal whenever atmosphere PROFILE time-continuity cases, is essentially Tiand a number of ionic and neutral proportionalto Tnaretbought to be Te is found to be eqwal to Tr constraints. constraints, bas been obtained can then be used to estimate properties, can be made by considering profile-continuity information In some atmosphere. in this region. the neutral appropriate by invoking with these data. ) In this way, Tn profile constitutmt 200 km and can reliably may be obtainable technique the neutral which is measurable predominant Other deductions profile ION-COMPOSITION General deductions [N2], AND (Further in fm-matim but this has not been done on the [0] and Tn profiles. the ion composition an entire This in the F1-regiom Profile have shown how one may determine where the ion composition t ion. Their analysis Tn(h) Tn from incoherent scatter data in regions is known and Te > Ti > Tn through use of the ion heat-balance eq”a 29 profile shape above h . 420 km assumes Tn to follow the Bates .T~-(Ta -Ti20)exp [-s(h-i 65 ZO)] and [0] to be in diffusive equilibrium altitude (usually 400 km) a“d Tn(h) then define of Bauer Et ~, specified the parameters above s, and [0]400 T-, and using a two-step ‘Odel ‘alue ‘0= Ti 20 from the data above 300 km [0] ~oo value, determined For analysis of Millstone Hill Z-pulse for simultaneowly. data, than parameters mmli”ear 20 mode130 and the remaining the fit (e. g., at night), retaining than 75-percent free parameters cases, is used. Data this lower-altitude is that all four unknowns are solved additional of an initial uncertainty, remit calculations 1s too great procedure set to their are or (mmcon- it is fixed using the Jaccbia If no E-region both TIZO and [0]400 are initially asSUming a determined the just -determined of the above analysis are needed in tbe fitting are sought. were scheme 225 km. only a one -step procedure, iterations reference analysis from the data, and [0]400 In some held fixed if the uncertainty If [0]400 has more vergence). Ti2w A second difference least-squares T- Then, a variation of O at some In the original determined First, from the data above While this requires made with certain more were to 435 km are used and help to establish range has been set at i 20 to i 30 km. abundance [0] (h). altitude with s = 0.024 km-i. and s were li6 The procedure. T- from 225 km. i97 i data are available Jacchia for model values before solving to find T ~ and ~ then, mtomatically, the data are refitted with s additionally held fixed -i to find ody T-. at O.02 km Only the results of the last fit made are preserved on magnetic tape, although all fitting results The need for nmltiple the reliability consistent tests and reascmable have produced Figure 3. fitting of the results. are not thought to yield are output onto the line printer. ccmti”gencies values of this profile discussed in that region. [Eq. (48)] is assumed fitting procedure, in the previous For is so nearly analysis complete, lations are given in Table assuming an arbitrary deduce assuming p . i, XVI. Shown here is A(p), A similar relation these tabulated accm-acy due to the slight is lost by choosing (perhaps values a central and F2-region this Tn profile meaestimate temperatures ambiguity can be solved. li, i2 Because the mapping relations ion composition. of ion temperature B(p) is shown for the Te/Ti points is used in 2-pulse differing data, E- As the ion and neutral simple the ratio (02+/NO+) ratio (see Sec. C-4 above). obtain Table Tn profile. temperature Moreover, of Te/TV member one would determination. data analysis. 5 percent] for the cor34 These re one would p . [0+ ]/Ne to the ion temperature uncertain A and B for 2-pulse to relative molecular-im for although Hill investigations. to derive temperatures from unambiguous -composition Millstone ion composition is derived also. of A(0) is somewhat that little uncertainty there is, in E-region in fact, b“t each one is tightly of each family. Linear The exact value mean a family packed such This has been done to XVI. The Z-pulse of p . 4, yielding data i“ the i 30- to 225 -km region profiles of Ti(p . i) and Te/Ti(p 66 I Hill ion-temperature it is possible ion and electron deduce of such relations sinmltaneously, showing the deduced prevents of Millstone the Fi -region of relative between secticm of the Pi -region has been used in some previous mass concerning in determining The [0 ]400 and s parameters when determined to $; valid through the F4 -region are equal in this region, interpolation questions Profile The Tn(h) estimate respondence a“d Ti ~ o. useful results data as tbe .“kmmvn io” cmnpcxiticm ambiguity raises good [0 ]400 re suits when s is fixed in the analysis. i 2 is an example This approach analysis is indeed often unsuccessful of anything b“t T- consistently Ion-Composition surement for profile The fitting procedure are initially . 1). analyzed under the assumption Once the independent Ti estimate TABLE XVI VALUES OF A(p) COMPOSITION Ti(model) simply B(p) RELATING” RELATIVE 1.937 1.000 0. I 1.744 1.065 0.2 1.570 1.123 0.3 1.427 1.165 0.4 1.308 1.188 0.5 1.218 1.192 0.6 1.146 1.177 0.7 1.W2 1.146 0.8 1.052 1.106 0.9 1.021 1.056 1.0 I. 000 I. 000 in this region (see previous fidence interval in p, there may be large onto the p-axis model (A here denotes systematic Ti estimate. Once from the p- to the B-axis On occasion, the physical and the temperatures corresponding Due to the possibility the Ft-region, O+). of large re.wits with the exception perhaps even in the worst-case situation. To date, procedure In addition yields limit (composition, of Te/Ti Te/Ti The . i)] con- to the statistical necessary mmpositim in deriving of Te/Ti . B(p) (Te/Ti) values limit by (p . i). outside of is assumed are adopted. in the independent Ti ,estimate Te) are also subject to sizable which is less than 20-percent F4-region . 1). . 1) i ATi(p the O or iOO-percent uncei-ta.i”ties these . Ti(model)/Ti(p one finds the final estimate to this composition the other F+-region temat ic errors, y). then estimating In such cases, systematic to A(p) due to the assumptions p is found, estimation (O - iOO-percent one finds the ion composition tbe Ti(model)/[Ti(p Ti uncertain errors to find B(p), this composition range section), XVI corresponding found by mapping ION TEMPERATURES 0.0 in p is similarly the independent FI-REGION ELECTRON B(p) errors mapping RELATIVE AND A(p) by finding the value of p in Table error ION P has been formed statistical I AND AND results in sys - uncertain have not been extensively used. Figure 4. i2 shows an example Electron-Density-Profile As explained P~(h). of the Te and composition i“ Sec. B-i The fim.1 estimate evaluation in this region above, (Te/Ti)(h) of (Te/Ti)(h) Sec. B-i above. Sec. 3 above], may result A large is needed before in the region profile the final estimate measured uncertainty in ion composition is grossly Ne(h) can be Calculated data. (Sec. 3 above), determination. from (i30 to 225 km) and thw Ne(h) Once this final esti- of Ne(h) is made “ia the method explained and thus to Ne(h) determination. if ion composition from of unknown ion composition is determined must also await ion-composition )(h) is made, deduced Corrections cannot be made until the io”-imposition mate of (Te/Ti values is not so detrimental A maximum in error. 67 to (Te/Ti)(h) of iO- to 20-percent i“ (see em-cm in Ne(h) E. DEBYE-LENGTH All small CORRECTIONS acf analysis employs a library De bye length (Sec. C-3 above). results appreciably when Ne is very mate is also low due to the Te/Ti above ). Moorcroft24 iterative manner, of theoretical This assumptim low, calculated is never for the limiting quite true, in which case the Te estimate correction employed has shown how one can correct but this correction acf!s 68 but cmly affects the is low and the Ne esti- in its calculation from P~ (Sec. B-i these Te and Ne values has not been applied case of to the Millstone in a simple Hill 2-pulse but data. ACKNOWLEDGMENTS In addition to the authors of this report, several people have heen instrumental in the development of the ‘two-pulse 81program at Millstone Hill. Wallace A. Reid has heen the engineer in charge of development, modification, and maintenance of the radar equipment. Leonard B. Hanson has been the engineer in charge of the development, modification, and maintenance of the radar-computer interface and maintenance of the computer. Glenn W. Armistead was responsible for much of the real-time computer program development. Richard A. Brockelman, Tor Hagfors, Robert A. Power, John K. Upham, Arnold D. Kaminsky, Alice R. Freeman, John M. HoIt, Ralph F. Julian, and Gary A. Bullock provided programming and data analysis support. 69 REFERENCES i. J. V. Evans, R. F. Julian, and W .A. Reid, ‘tIncoherent Scatter Measurements of F-Region Density, Temperatures, and Vertical Velocity at Millstone Hill, at Technical Report 477, Lincoln Laboratory, M. I. T,. (6 February i970), DDC AD-706863. 2. J. V. Evans and M. Loewenthal, 3. J.V. Evans, Planet. Planet. Space Sci. i3, Space Sci. iZ, 1031 (1965), 4. J. Geophys. 5. Planet. Space Sci. ~5, 3387 (1967). 6. Planet. Space Sci. i8, i225 (1970), 7. Planet. Space Sci. 2J, 763 (i973), 8. Planet. Space Sci. 23, i46i 9. Proc. io. Res. 7Q, i$75 (1965), J. Geophys. ii. J.V. Evans and L.P. 12. L.P. Cox and J.V. Evans, Res. ~, DDC AD-616607. DDC AD-6243 *O. DDC AD-7i6056. DDC AD-772 i37/6. (1975), IEEE ~7, 496 (1969), 915 (1964). DDC AD-A024177/8. DDC AD-694i92. 3343 (1967), DDC AD-6589%2. Cox, J. Geophys. Res. ~5, 159 (i970), Evans, J. Geophys. Res. DDC AD-703492. 75, 6271 (1970), DDC AD-7229 ii. 13. J.V. 14. G.W. Armistead, DDC AD-742616. i5. J. V. Evans, Editor, IfThe Millstone FY 1970,. Technical Note 1970-20, ber 1970), DDC AD-717156. 16. J.V. Evans, R.A. Brockelman, R. F. Julian, Radio Sci. ~, 27 (i970), DDC AD-70463i. i7. J. V. Evans, ‘Millstone Hill Thomson scatter Results for 1967, II Technical Report 482. Lincoln Laboratory, M.I.T. (22 July 1971), DDC AD-735727. J. Geophys. Res. 77, 2341 (1972), J. V. Eva”s, i8. J. P. Dougherty 19. COSPAR 1965). and D.T. 20. U.S. Standard At mosp~ D. C., 1962). 21. R.H. 22. P.M. International Farley, (U. S. Government Perkios, Banks, Planet. Space Sci. ~, J. Geophys. D.T. Farley, D.R. Moorcroft, 25. H. Carru, 26. W.L. Oliver, J. Atmos. 27. P.M. Eanks, Proc. 28. P. Bauer, R. Benoit, Terr. IEEE ~, Proc. W .A. Reid, Res. md L.A. Carpenter, 6S, 5473 (1963), (North-Holland, Amsterdam, Printing Washington, Res. ~, Office, 6370 (i968). 4091 (i966). ~, Res. i53 (4972), 1105 (!966). ~, and M. Petit, P. Waldtetiel, Bates, J. Geophys. Res. J. Geophys. Sci. ~, Hill Propagation Study: Progress in Lincoln Laboratory, M.1.T. (2 Decem- Atmosp&e Wand a“d F.W. 24. Radio J. Geophys. Reference 23. DDC AD-752948. and W .A. Reid, 955 (i964). LaNature, Phys. No. 3330, 422 (1962). 37, 1065 (i975). 258 (i969). and D. Alcayde, Roy. Sot. ~, 4825 (1970). D.R. 30. L. G. Jacchia, ‘Revised Static Models of the Thermosphere and Exosphere with Empirical Temperature Profiles, n Smithsonian Astrophysical Observatory Special Report 332 (1971). 31. P. Waldteufel, Res. &, ~, Res. 29. J. Geophys. (London) J. Geophys. 451 (1959). 6995 (i971). 70 .—.. —. APPENDIX SYSTEMATIC The techniques cause a number DISTORTIONS employed at Millstone of systematic into two categories. incoherent scatter transmitter calibrations, and errors in the analysis undistorted theoretical distortions 1. scatter scatter trum. by D.T. function of the impulse noise passed where t receiver passband and is of great in comparison measurements. De flnition may improper for these distortions function between Calculation and evaluation of of these function (or, ~IP(T)Pw(T)l expected fortbe is that the ionospheric of the transmitter plasma waveform, at correlation density incoherent fluctuations, and the bandpass time delay spec~, P(T) pw(~)be the auto- and pn(~) be the autocorrelation function equivalently, the autocorrelation result general inco- autOCOrrelatiOn by the receiver measured waveform, then Farley!s gating upon ionospheric have been considered is then multiplied of the ionospheric GATING and transmitter function signal of the receiver TRANSMITTER remit hytheautocorrelation through the receiver), ~ > denotes His basic backscattered <Pm(T)> distortions to any monostatic, by inaccuracies, consists is possible. AND measurements of the gated transmitter response Allowance and the distorted bandpass be the autocorrelation be the autocorrelation correlation receiver Farley. multiplied If we let pm(T) introduced Such analysis comparison PASSBAND correlation-function of the resulting These intrinsic function of this appendix. of finite-width is first spectrum functions before RECEIVER experiment function by the hardware. correlation correlation those due to use of a finite-width of the measurements. are the objects The effects herent introduced the ionospheric in the measurements. we have those distortions is thus required FINITE-WIDTH to appear MEASUREMENTS there are those distortions -namely, Second, importance the distortion First, experiment gating. Hill to measure distortions be grouped A IN CORRELATION-FUNCTION function of white in the time domain can be written * P“(T) value, and * denotes convolution. pm(r) and Pn(r) are known func - tions for a given experiment. Analysis of measured functions pm(~) at Millstone Hill uses a library of theoretical functions Pt(T) cOvering the range Of fUnCtiOns over which P(T) may reasonably be expected to vary. TO be able to compare the measurements pm(~) with the theory pt(~), we must follow one of twO approaches: (a) Correction approach then compare (b) Convolution Pt(T) P(T) andpt(r) approach -apply Eq. (A-l)fOr P(T) in terms of pm(r), and directly. the convolution effects of Eq. (A-i) to as Ptc(7) =[Pt(T) tD. T. Farley, –solve Radio Sci.4, Pw(d] * Pn(T) 935 (i969). (A-2) and then compare Pm(T) and Ptc( r) directly. etone Hill 2 -pulse data analysis; A. Correction by T. Hagfors A necessary Implementation we attempt to solve at Millstone Hill; assumption to solve Eq. (A-i) a brief Eq. P(T-x)= P(T) in which case we may write Pm(r) = is discussed P(T) +~ Eq. [PW(T) : for p(i). description below. of Eq. (A-i) was originally is given here. is well con- approximated by as (-x)+*+2 (A-1) This problem of his approach for P(T) is that <pm(T)> (A-i) The next step is to expand P(7 -x) Pm(r). of these methods of Mill- Approach In this approach, sidered Each method has been used for some facet + (A-3) . . . as Pn(T)] + Pw(d *, {PW(T) ~ * [-TPn(T)]} ~pn(T) +... II ~z c? p( ,) + ----# I (A-4) or Pm(r) = P(T) Fk(,) = PW(7) Fe(,) ‘k + ~ only three terms forms Eq. of the expansion used in the Millstone terms for p(~). of Eq. estimate p isderived from by the derivatives ference IIOiSy ptc(~) between data. ~gnitudes An iterative estimate calculated Fn(r), F, ( T). -. and F,(T) for pn(T) and the actual transmitter of p(r) procedure is calculated wave- is then used to solve ignoring the derivative of P(T) is estinratedas (T) .~ dp. ~(T) d~ measured noisy data is undesirable, functions HagfOrs waveform Hill experiments. Tbe first-iteration pi(T)= * Pi(r) being retained. Gaussian (A-5) The ith-iteration from (A-5) 0,1,2 1 using an appropriate (A-5) F2(T) k= P“(,) # [ analytically 2 + & Fi(T) of aparabcda d2p. 4(T) + FO(r) the derivatives fit to five plots of pt(~) and pi(~). noisy. req”iredin points inthe 5 iterations F’2(T) —. As the calculation Eq. (A-6) neighborhood were A somewhat (A-6) FO(r) d? data and is therefore for pm(~) show that i. The procedure Fi(T) —_ required larger does not break down upon addition of derivatives are replaced of ~. toyield Tests with noiseless animperceptible number of iterations of random in practice dif - is required noise to P(T); for noise Of UP to P(0)/5 have been tested. 72 .—..—.————- This tortions correction approach is used in Millstone data cm the real-time to the finite-width Hill 2-pulse computer receiver data analysis printout; it is not applied netic tape, nor Is it used in the final data analysis tives noisy data necessitate ed by Eq. from problem can be entirely apprOach B. we perform this calculation. The first of these functions, leaving ET represents convolution Eq. time) espbciall y as the by use of the convolution TWO methods the two methods (A-2) ‘: have bee”. used to are small. directly through numerical function normalized integration to unity at zero using lag, i.e., (A-7) Pn(x) of the waveform W(T), which, in general, quite time consuming, hy the simple may vary with . . These and to speed up the calculation approximation (A-8) than the sum of T (the width of the transmitted Calculations have shown this approximation <00, and 200 psec for Modes most of the convolutions performed (A-2). = Pt(d all must still be performed were PW(T)(X)] been replaced such that ~ is greater allows Eq. correlation are intrinsically Ptc(T) Tn [Pt(X) in Practice, 30 psec and T .40, of Eq. between calculates tbe energy half -width of pn( T)]. tbe calculation the resulting = Er calculation (A ‘7) has. at lags reliable, of computer on magof deriva- = f,(7)/fo(o) f~(x) Eq. data stored it is felt that the calculation is not sufficiently at the expense Discrepancies methcds Pte(T) where (A-6) to the real-time since gating dis - look at the corrected Approach In this approach, sampled (albeit and transmitter to tbe problem. Convolution perform avoided passband to obtain an initial to be avoided F, pulse) to be quite good. G, and H, respectively, in Modes T < In practice, out through +00 ~sec for Mode F and i50 psec for Mode G. With ~n m we see that ~ > T + F and G, for which O< for Mode H, for which O< r < 190 psec. and ~“ [the 380 ~sec, but the convolutions Approximation of (A-7) by Eq. (A-8) for these longer lags produced errors of about +i K in ion temperature Ti -’t and +iO sec m collision freq”e”cy Vin for Mode F, and O K in Ti and +10 K in electron temperature Te for Mode G. These The actual implementation tunately, did not greatly Ptc(T) i.e., a different = errors of this convolution factor and +5 K in T= for Mode H. These each lag instead In October was applied for each lag. however, errors computations time. Sec. A above [Eqs. (A-3) a small error was erroneously The errors have also been ignored was that a different normalization constant actually which, for- calculated as resulting were about in the data analysis. factor needed; K in Ti The main had to be calculated this essentially for doubled the required. i 976, a second method for performing to save computation contained (A-7) fm. Mode F, +6 K in Ti and –5 K in Te for Mode G, and –iO of tbe one normalization number of convolution calculation fr(T)/fr(0) normalizing -i m .in of this error, in the data analysis. Equation affect’ the final results. O K in Ti and +2 sec effect have been ignored the convolution This method uses the results through (A-5)] to write Eq. (A-2) 73 of fIagfors as of Eq. (A-2) 1 calculations was instituted as described in dpt(,) Ptc(T) = Whereas Hagfors Pt(T) originally of pn(T) and PW(T), I?O(T) approach the measured tions; (A-9) w. data before is presently with the measured data. functions instead of noisy, data with theoretical used to convolve The functions 2 -pulse passed Adoption j I of Eq. from (A-9) Fi[~), to approximate A-1. Of course, distortion Receiver (A-9) in Sec. Iahove design. number are discussed for the Millstone Hill used for data Of a parabola tbe rmmber of computations to and 4 mdtiplicatioms Ptc(0) = per lag – t). function of finite is valid receiver bandwidth for any receiver between and and pulse the hardware implemen discrepancies for the 2-pulse experiments miscalibrations, and errors, and have been of instances. Those cases - for whicb important here. to be symmetric Receiver of receiver tilt i“ the from through the receiver. This bias can be mea- data passed but the fear of receiver Sucha asymmetries, an average data taking produced discovery nwnber 74 in the next para- (see Sec. I above) output from profile A-%. de- the receiver of ionospheric of such measurements pn in Table variatio”witb in data interpretation, spectrum of the noise each time an altitude the adopted f”nction bad changed of bandwidth could be important function a bias to dubious wortb. of the iommpheric of a large asymmetries with time is presented data tobeof variation Smearing experiments that receiver variation rift The correlation is measwed; b“t important of data of applying been made. during the 2-pdse that there was a small a fixed six years This tilt has tbe effect characteristics bandwidth. andtobave until some of the spectrum. hese2 -pulsed to the discovery frequency was not evaluated It was tben”oticed oconsidert receiver about its center asymmetry since the amount of instrumental tion functions unmodified as the derivatives discrepancies These pendsupon days of 2-pulse original smooth. -theoretical are presently has decreased for the effects ina calculated time has also recently is measured and (A-9) to b additions correlation for in the data analysis, (evidence In addition (A-2) impm’tance results at the center drifts hasledw.t Eq. inaccuracies, been collected. sured and allowed with time comparison Hagfors’ of the original, functions are calculated there are always was designed spectrum all ionospheric from func - of ,. in hardware of some 45 kHz. had already receiver for direct to Inaccuracies The receiver bandwidth of Eqs. (A-5) than O. 00i (relative experimental Hill consist found to be of significant A. functions correlation method has over comparison These and i4 multiplications, presented tation and the proposed measurement was in an attempt ionospheric are now calculated and F2(T) by Eq. gating upon the ionospheric employed. at Millstone Of actual measurements INACCURACIES The development graph) (A-9) In general, in Table ACCW.aCy is better 11. HARDWARE scheme ! data. required 6 additions savings. transmitter I the theoretical representations is highly desirable. are tabulated The derivatives functions that the present by Eq. analytical use of Eq. (A-9) with theoretical through three points i“ the neighborhood necessary a large measured Pn(~) and FO(~), experiments analysis. required calculations original comparison The only advantage method is that the derivatives (A-9) F,,,, has used sampled Hag forsl of pn( T) and pw( r) for this calculation. deconvolve + - FO( T), Fi ( T), and F2 ( T) from calculated the present F,(T) + ~ from correlatbe earlier Mm-e recent TABLE A-1 Pn AND THE FUNCTIONS Fo, F,, AND F2 FOR MODES F, G, AND H — ,. .,. J-fKme r. Mcde G Delay (yse.) P n o 1 .M 10 0.4s 20 0.0s 30 0.01 40 O.oc Fo — .00( F1 O.OQ( Mode h . F2 — Fo F1 F2 — Fo F1 30.6 .003 0.00( 38.7 1 .Ooc O.ooc .193 -2.623 50.2 .067 .158 -1.680 48.5 .068 -0.920 46.0 .003 -0.077 31.8 .C6a -0.920 46.0 30.6 .06a -0.920 46.0 .068 -0.920 46.o ,001 –0.030 39.2 -0.89C 45.6 F2 41:0 1.028 -0.330 43.2 1.032 -0.430 44.3 1.033 -0.4-45 44.5 50 60 70 80 .000 O.ow 90 100 110 120 130 140 ,000 O. COO 38.7 150 160 170 180 190 200 — — ‘o is dimensi.an less; F1 has dimensions of microseconds; 4 column of pericds means that the hnction node. continues 75 F2 has dimensions of @see) unchanged thrwgh the maximum delay for the measurements have shown that the receiver value changfng from lyzed with tbe Pn of Table A-i tinue to analyze deduced from before has narrowed this time variation all data with this same several in time, years been ana- it was decided The effects in the adopted the pn ( iO -ysec) had already was discovered, Pn for consistency. the data due to these inaccuracies are probably B. spectrum Since much data from 0.48 to O. 54. to con- upon the temperatures p“ have not been evaluated, but small. FulSe -Shape Distortions The pulse waveforms tangular and flat. employed Rise power capacitor-bank time. In addition, discharge significant enter into the calculation for calculation in the 2-pulse and fall times during pdse are, are designed of course, trammission pulse timing errors of the correlation of the effects experiments in the equipment causes the pdse have been identified. function shapes to droop with pw of Eq. upon measurement rec- and the znain - All these factors of the pulse waveform of instrmnenta.1 smearing to be perfectly finite, directly (A-i) needed of the ionospheric corre- lation function. The droop in the pulse waveforms transmission schemes. well described as dP/P 1s the longest cant. , shape, correction which sbmld transmission accounted for this effect scheme for. should be multiplied P is power should rigorously this distortion for power variation factor is small, pulse pairs waveform it was decided the slight distortion In making this correction, cm different is that tbe measured all droop is Since 0.4 msec drooping however, vs time and ignore effect. for the droop correction by the correction affects is small but signifi- be made by using the true, Since the correction be a second-order discharge has show” that this power and t is time in milliseconds. experiments, using close - and wide-spaced Tbe result capacitor-bank with tbe equipment in the 2-pulse the true pw of Eq. (A-i). make the simple form . dt/15, where pulse transmitted Correction to calculate d“e to main-power Experimentation to i“ v..ave - the real pulse- frequencies nmst be autocorrelation function f i90 psec f . i + 7/45000 T< f = i + (, –200)/i5000 r > 200 ~sec for lag T in microseconds. The finite rise slopes and fall times of the leading cause of the shortness short compared cially apparent. portance. effect “’ and trailing of the pulse-generating of the F-Mode with tbe 40 -psec Were of leading-edge rcm”ding pulse. F-Mode a21 F-Mode Tbe 20- and 40-psec edges of the transmitted equipmcmt, seem The rise and fall times pulse Ie”gth; pulses a rounding of the same length, lag waveforms, is relatively however, different pulses, to be important due to the finite ody for Mode F be - cannot be considered of tbe leading this rounding very edge is espe - would be of small use 60- and 80-psec for these lags compared pulses, im- and the with other lag waveforms. Investigation mode which were of the leading-edge pulse rounding nmch more tba” the pulse edge-rounding. pulse length was discovered since pw of Eq. (A-i) seriow to be 39 psec instead. is easily modified sistently. Measurements a 59-~sec, pulse is used for the 20-psec gap are used instead of waveforms of two 39-psec for Mode F uncovered This difference, to account for it were for other lags, lag, pulses bmvever, separated 76 by a i-psec errors the basic in itself, this basic and that two 36-psec timing First, in this 40 -psec is not detrimental pulse length used con- have show” that a 58 -psec, pulses separated not by a 6-psec gap for the 40-psec lag. Furthermore, this 6-psec to the shortness crepancies gap appears of the pulses, in the F-Mode in the measured the waveforms to be more pulse waveforms autocorrelation functions, was not successful and even correction in sufficiently complexities, uncertainties, and discrepancies and distortions of the transmitted ignored Transmitter c. Although single from electron F-Mode all the frequencies These removing for the major observed encountered timing errors discrepancies. i“ accounting waveforms, diseffects and correcting the 20- and 40 -psec standard, that expected is open to direct for lags have been in the transmitte~ the RF phase difference due to transmitter klystrcm experimental approximation, chirp. and receiver between This arises a.mplifiep changes are derived a pair of transmitter because the transit during the course i“vesti.gat ion, or may be calculated the voltage Vt applied from to the klystron from varies energy It is klystron in reality, and C is the capacity Equation this is not precisely (A-i O) assumes true. where time of the first during the pu18e as shift between condenser achieve that stores ccmstamt during the pulse, in the beam are accelerated Vt, and thereby a“d mass of an electron, (2)”2 If d is the distance This principles. between the though, the a speed Ve given in . Vte e and Me axe the charge ‘e’ of the high-voltage that It remains The electrons cathode and ring anode by the potential $ NfeV~ pulses (A-iO) current, for the pdse. a of the pulse. dVt — = – It/c dt where in Due to the data. employed beam in the transmitter To a first in shape and, due slanted. Chirp stable frequency differs effect in analyzing than rectangular to be significantly have berm seen to produce “quite detrimental these errors completely triangular the pulse edges appear respectively. It follows that ~ between the input and o“tp”t cavities in the klystrcm, there will be a phase the two of ~=_?#d radian =-’”:d(+s’z where f is the radar frequemy. The change in Vt during tbe pulse produces a rate M $$ = 2rfd r Me ‘ -Tfd It z> r d~t i ~e$~dt c “ t I I 77 L a change in @ at This, in turn, leads to a frequency = -~ fd r M e ~ev3 t shift in the transmitted $ (A-ii) c For the case of the Millstone Hill UHF transmitter C = 24 HF so that Eq. (A-ii) reduces In practice, & is typically shift which, unless -8 m/see. To compensate from G and H. Modes corrected, (employing Eimax x626 klystrons), d . i 08 cm, to of the order of –23 Hz. will he interpreted for this, pulse of a correction 78 This corresponds as a plasma drift must be applied to an apparent away from the radar to any drift velocities doppler of obtained APPENDIX STATISTICAL I. EXPRESSIONS FOR The correlation (sine and.osine ERRORS UNNORMALIZED functions channels). during the same sweep, our purposes, CORRELATION are estimated In practice, from digital separated MEASUREMENTS FUNCTIONS samples p“lsetransmission and the returns this maybe B IN CORRELATION-FUNCTION into two pairs taken as being equivalent oforthogonal may bemadeon receiver of orthogonal to two separate sweeps outputs two frequencies outputs. For and one orthogonal receiver. Samples from sweep i at time t (meamn’edf romthec noted by Yi(t) for the sine output and Ui(t) for the cosine to distinguish samples taken in the noise region in the signal-plus-noise scatter over signal, region. the computer fn calculating forms certain ommencementof output. of the sweep Primes are de- are used (i. e., Y~,U~) (near the end) from theautocorrelation products the sweep) samples taken function of the incoherent of the samples in real time and averages a number of sweeps. In general, is calculated the unnormalized according K AR(T) real part of the autocorrelation function AR(T) for time lag r to the expression . & s. s M-t ~ ~ l=i j.O [Yi(t +jc) Yi(t +7 +jc) +Ui(t +jc) Ui(t +7 +jc)] Kn —— Zin ,2 [y;(t) 1.1 K~/2 - & E i=i unnormalized K AI(T) = & s ,Z 1.1 U;(t + T)] M-t ~ [yzi.i(t j=O +U2i-i(t The corresponding y~(t + T) + u;(t) +j=) Y2i(t +~ + K) + jc) U2i(t +T + jc)] imaginary . (B-i) part of the autocorrelaticm function AI(T) is given hy M-% Z j.O [Yi(t + j<) Ui(t +7 + jc)-Yi(t +T + j.) Ui(t + jc)l (B-2) 79 fn Eqs. (B-i) and (B-2), plus-noise region, and cosine samples the separation separated terms noise, in Eqs. (B-1) and possibly from tion of the transmitted subscripts ensemble average of r. There products from from noise contribution. It is assumed, of sine U M # 1, ●. incoherent scatter and airplanes. signal, The purpose Each pair of noise-region sweeps however, M pairs and third terms. distant hills be taken in different <Y~Y~ > = (Y:> or sweeps that noise samples <Y~> for i =# j. samp- where the separa- The brackets with-different <> denote an terms are only value. in Eqa. (B-i) in F- Mode calculations, used fn forming term i.e., or expected The third terms present pulses is T. are uncorrelated, the first contain contributions returns the system in this term may not necessarily are taken in the signal- taken in the noise region. the next is taken to be a constant, and (B-2) also clutter of the second term is to subtract les for which samples pairs by T are used to calculate of one pair of samples Tbe first system KS is the number of sweeps Kn is the number of sample and (B-2) effect a clutter but are included in these terms here for generality. is taken from are only half as many products subtraction. consecutive used in calculating These Each pair of samples sweeps with a pulse spacing term three as compared with one. 11. EVALUATION In order assumption variance to evaluate istaken tohavea T)> .–<ui(t)Yi(t+ samples andp+~) of terms in Eqs. (B-i) +T)], normal are defined +7)) variables Yi(t and (B-2), with zero [Yi(t), Ui(t +T)], distribution. the basic mean and [Ui(t), The correlation by .UZPR(T) T)> .U%l(T) Xi(t) of the receiver . (B- 3) output are written .Yi(t)--Iui(t) x;(t+ (B-4) correlation T)> represent function is given by .2 U2[PR(T)+IP1(T)] tbe real and imaginary output in the signal-plus-noise an odd function of r, SAMPLES are normal bivariate for these distributions The complex of the receiver Adopting +T)] <Yi(t)ui(t+ <xi(t) pR(T) and Ui(t) T)> . <ui(t)ui(t xi(t) I . ~. and variance Yi(t) <Yi(t)Yi(t+ notation, RECEWER each pair ofvariables[Yi(t), Yi(t pR(.r) andpl(r) In complex Thus, INVOLVING the expectation Furthermore, and [Ui(t), coefficients where PRODUCTS is made that the samples 92. Ui(t +T)], OF and wehavep region. R(0) . 1, PI(0) .2 U2P(T) . (B-5) parts of thecomplex correk.tion function PR(T) iS an eVen fUnCtiOn Of r, PI(T) is . 0. the notation Zi. we may summarize Yi(t +7) the following , Vi. ui(t properties +7) of samples 80 (B-6) in the signal-plus-noise region <Yi> = <Zi> <Y;> More = (2:) = (vi) = <u:> <Yi.zi> = <uiVi> <Yivi> = –<uizi> complicated be evaluated = <Ui> relationships = o = <v:> = 02 = U20R(T) = UZP1(T) involving (B-7) . the expectation value of products of four samples may using the theorem, <X,X2X3X4> which applies = <X1X2> when Xi, <X3X4> +< X1X3> Xz, X3, and X4 are jointly <Xzx’t> normal +<xix4> (B-8) <X2X3> random variables with zero mean. For example. I <Yiziuivi> <Yfz;> In the noise denoted by u; region, = u4[i and pn for relationships to-sweep. term AR(T) In evaluating no correlation and AI(T), clutter function of the receiver output are (B-ii) the primed samples subtracts the variant es of AR(T) from sweep-to-sweep, as in Eq. (B-7) by substituting o: <Yj(t2)> = o <Ui(tl) uj(t2)> = <Ui(ti)> <Uj(tz)> = o e of the clutter term but it does affect OF variances manner TBE We may now proceed and AI(T) ‘ii’ their in an empirical UNNORMALIZED the assumpticm does not affect of the variant = ‘l’t + ‘<) ‘ Cij ‘ ‘ijzij + ‘fjvij Eij . YijVij – ZijUij c~ = Y~z: + u;v~, is, values The effect of of non- A. CORRELATION and (B-2). ‘ii is made that there the expected and so must tie retained. in Appendix with the computation , sweep- (B-i2) defined ‘!(L +]’) from . COMPLEX in Eqs. (B-i) that exists for i ~ j = <Yi(t,)> the present any correlation and AI(T), i.e., Yj(t2)> is evaluated AR(T) and (B-2) <Yi(ti) 111. VARIANCES functiom correlation Defining v; = U;(t + 7) between in Eqs. (B-i) Under this assumption, zero , (B- iO) u and p. The clutter in fact, + 7) . and complex = ,onR(7 ) + rPd(T). z; = Y;(t we obtain similar + 2P:(T)] the variance and P*(T) (B-9) – P:(T)] = 04 [I@) FUNCTIONS es of the unmmmalized Adopting correlation the notation Zij = Yi(t +. + j.) “iJ > = “J ‘2i,j ‘zi, +’ +j<) = ‘2i-4,jz2i.j j = ‘2i.4, + ‘2i. jvzi,> E: = Y;V~- Z~U: 8i %,jvZi,j – ‘Z1, ju*l. j,j (B-i3) we may write, (B-14) The complex correlation to as the measured related correlation to the correlation Appendix function after function system noise i.e., pm(r), function of the incoherent and clutter <AR(T)> scatter subtraction - I<AI(T)> signal will tie referred = U~Pm(T). Pm(T) (see Sec. VII below, A). Taking expected values <AR(T)> The variance of AR(T), in Eqs. (B-i4) and (B-45), = C72PR(T) – fJ:PnR(7) <AI(T)> = U2P1(T) – the variant U:Pnl(T) = U:PmR(T) = e of AI(T), U:PmI(T) (B-16) . and tie covariance of AR(T) and AI(T) are given by u:(T)= <A:(T)>– <AR(T)>2 q%) To calculate and then subtract three variables of Cij variables, i (sweep 0:(7) = +T) OR script is and also o:, <42(7)> we square (B-48) <A1(T)>2 AI(T)> – <AR(T)> = <AR(T) <AR> 2. Cif - the right-hand is correlated. is different. Also, I + * terms take expected disappear because there is no correlation values. no pair of the between any pair or any pair of D variables for which the sub21, j With these simplifications, u:(r) may be written M-i <C;> s (B-i9) . side of Eq. (B-i 4) for AR, 21, j or any pair of Cl variables, . & <-41(T)> A great many of the resulting Cl, and D number) (B-17) – <Cij>2 + ~ ~ (M -k) [< CijCi,j+k> - <Cij> I k=i [<C;2> <Ci,j+k>l - <C; >2] n M-1 I <D~i j> _ <D2i,j>2 x [< D2i,jD2i,j+k> + & – 82 z k=l (M–k) <D2i,j> <D2i,j+k>] I ‘ (B-2o) The expressions in terms in Eq. (B-20) involving of the autocorrelation rmise region (PW PI, the variables functions . 2U%R(T) , <c:> <Cijci,j+k> = 2U4[2P:(T)+p:(kc) XPR(T <c;) = 20:pnR(T) <D2i,j> = O <Dzi jD~i,j+k> The general , <D;i = 2U4 [p~(k~) result for u;(T) iv+ ; Ph, = 2U4 [i + 3P:(T) +P1(T , (pRn, + pf(kc) -k) C;, and D may readilybe evaluated 21,3 of the receiver outnut in the sienal-&m-. and va~iances and U2) and in the “noise region <Cij> Cij, <C;2> + OR(T +k~) o:). These results are -P:(7)] + ke) P1(T–k~)] = 20: j> = 204 and [1 + 3P:R(T) – P:I(T)] , + p~(kf)] (B-21) . is then i ~ (M – k) [Sp;(kc) + Sp12(ke) +PR(T + kc) PR(7 –k~) k=l – PI(7 + Aim ‘n + — ~4 where S . 1 if no clutter used. The summation k,) PI(T – k.)] +P;R(T)–P;l(T)] —--= [i Kn (B-22) I s“bt~ action is used in calculating term must be omitted and S . 3 if a clutter AR(T), if M . 1, i.e., if only one product term is formed is per sweep. The calculation variance of the variance of AR(T). tuted for Cij, The of AI(T) follows U12(T) is given by Eq. (B-20) along similar if the variables lines to the calculation Eij, ., and C;, respectively. 21,J general reswlt for ai2(r) is q%) D = & * F 21,j’ of the and E; are substi- s + P:(r) - +) M-1 + ~ ~ (M-k) [S&(kd +Sp:(kc )-j@ +kc) OR(T -k, ) k= 1 + PI(T + kc) PI(T - k-)] (B-23) 83 i An equation for the covariance of AR(T) AI(T) may be derived in a similar maoner as Eq. (B-ZO), M-$ OR@ . & <CijEij> - <Cij> <Eij> + ~ ~ s (M -k) k=i X [ <CijEi,j+k> + EijCi,j+k> _ <Cij> – <Eij> ++[ c;E;) <Ei,j~> I <Ci,jw> - <C:> 1 <E\>] n M-i + * F ,?i,J> -<D2ij> , ‘D2i,j <F2ij>, +4 I x I< D2i,jF2i,jw> + ‘F2i,j The terms D 21,j+k to evaluate <CijEij> = 8U4P1(T) PR(T) and the result <F2i,jtk> <D2i,j+k>] . (B-24) this are = <EijCi,j+k> x [PI(7 ‘D2i,j (M-k) I + [OR(7 + k~) + PR(T – k~)l = a4 (4pR(T) PI(T) <C;E;> – <D2i,j> > – <F2i,j> required <CijEi,j+k> ~ k=i +@) = + kc) + PI(7 - k~)}) PnR(T) F.. >=(F 21,j+k ~i jD2i,j+k> , of substituting these terms = O (B-25) into Eq. (B-24) is M-i 4 OR *(T) = & 2PR(T) @ + ~ s [ (M-k) k=i x [PI(T ‘n ‘~ ~ +k~) –k~)l +P1(T 4 2MK~ _ PnR(T) Pn1(7) Kn 84 . I [P# + kc) +PR(T - k)] Iv. MAGNITUDE OF NOISE The last terms tem noise terms. in Eqs. (B-22), subtraction, The values SUBTRACTION This term TERM (B-2 3), and (B-26) is of order (for u:, u;, K -4 compared of M and the ratio Kn/ KS f~r Modes and OR ~) reSUR from with order F, the SYs- (MK~)- 1 for the other G, and H are given in Table B-i. TAB LE B-l VALUES OF K#, AND K~K~ Mcde F I * Numbers that U2 & a; at least 52(20) I 1,3,0,5 75 (20) I l,3,0r5 in parentheses apply 6 November mOin text. is always 10,3 I I H 1970. to dalu taken prior tO See also Fig. and using the values in Table an order-of-magnitude MODES M 75 (20)” G Noting M FOR THE 2-PULSE smaller I and Table Xl in B-i, it is seen that the noise than the other terms subtraction term and may he safely neglected. v. CORRELATION OF Theunnormalized correlation lated using signal-plus are still from correlated ESTIMATES -noise estimates LAGS at different different the noise subtraction lags [i. e., AR(’ri), AR(T2)] (and bence independent) terms in Eqs. (B-1) sweeps, and (B-2) arecalcuhuttbey use samples for all lags. Thecovarianceof AR(7i), AR(T2) is Kn ~RR(Tf,72) DIFFERENT sa.mples from because the same sweeps AT =-& which may he evaluated [( -Cc’(’))(:ic+ x n to Kn i= t C;(T,)Z C;(T,) i. i ) (B-27) 4 ‘n ‘RR(T4JTZ) = ~ [P~#~-T~) ‘pnR(Ti) 85 PnR(T2)-Pn1(T4) Pn1(T2)] . (B-28) Similarly, the covarianceof AR(ri), A1(T2) is 4 OR1(T,, T2) = $ [Pn1(T2– Ti) +PnR(T,)Pn1(T2)l +PnR(T2)PnL(Ti) (B-29) . n The covarianceof the estimates at different lags is of order K~i, -i (MK~) . In the noise subtraction method used, Knis always of order larger vI. than MK~ (Table B-i), and the covariance VARfANCE IN THE AMPLITUDE CORRELATION FUNCTION Theunnormalized subtraction amplitude [Eqs. (B-1), B(T) We wish to derive =[A~(, xz)> ‘J:= <Ax, #x2)> i/2 correlation without serious COMPLEX function after noise and clutter (B-30) for evaluating of B(r) and its variance the expectation and variance u~(r) for later of a general 2 u ,, and X2. u;, Xi, xz are $=f(<xf>, (B-3i) <X2>) –<f(x, ent applications, mations are, and u Thepartial 2(:)2‘2” #xz)>z=ri ,3X i$z(~) (%) (B-32) are adequate The expectation <B(T)> u:(T) neglects +P&(T)l~’z . <B(T) >-2 [<AR(T)>2 U12, ‘d X2 =<X2>. Since KS is large, in the pres - these approxi- (B-33) . +P:l(T)l-~ 2 ‘R, I aregiven u:(T) + <Ar(T)>2 u~(T) + 2<AR(T)> OR ~(7)] [P: +2 PmR(T)Pm1(T) u;, K~2. Of xi of B(r) =[P;R(7) where of the order at xi = <xi>, +<A1(T)>21i/2 x <AI(T)> i terms of Xi and X2 and the covariance are evaluated of B(T) is then =[<AR(T)>2 The variance the variances coefficients for our purposes. =U:[P:R(T) I respectively, ~~$ferential Eq. (B-31) use. function ‘“:(%Y where error. . . of the expectation X2) of two randOm variables <f> = <f(x*, be neglected UNNORMALIZED B(T) of thecomplex ) +Af(?)] expressions THE are is an estimate Approximate f(x,, (B-2)] OF may always while the variances an order-of-magnitude R(T) a:(, )+ P:l(T) u:(T) (B-34) UR,l(T)I in Eqs. (B-22), 86 (B-23), and (B-26). VII. RELATION BETWEEN MEASURED CORRELATION FUNCTIONS Theunnormalized tractions complex AND correlation INCOHERENT SCATTER function ohtained after system noise and clutter sub- denoted byu~pm(’r). U:Pm(T) = (AR(T)> o:Pm(r) = +I<A1(T)> (B-35) . Using Eq. (B-16), a* P(T) As both the signal-plus-noise are Hermitian, statistics –m:Pn(T) correlation in the signal-plus-noise maybe (Appendix A) written region in terms where {[ Pw(T)P~(T)] W. is the transmitter the transmitter PW(T) V(T) = For the experiments the measured correlation correlation function p5(r) func - exp[luor] * Pn(T) (B-37) The weighting function pw(r) is theautocorrelaticmof * V*(–T) (B-38) ’xv*(–T)]T=o . being considered, length are transmitted It is desirable scatter function pn(r) noise has the same V(t) [V(T) of the basic as in the noise region, correlation thesystem *Pn(T)}7=~ frequency. waveform andthenoise Assuming of the incoherent P~(T)l [PW(T) ‘m(T)= function p(r) i.e., pm (T) =P%(–T). so also is pm(r), tion pm(r) (B-36) . to express thevarianceum This basic (r . O). thevarianceu2 m forother 2. M only estimated single-pulse variance pulse transmissions when single is denoted in terms pulses byu~. of u; 2 2 ‘m ‘ ‘rob (B- 39) where [v,(t) * V:(-t)}=o (B-40) aT= [Vo(t)*V;(-t)]t=o Vo(t) being the transmitted waveform form when lag T is measured. when lag zero is measured, and VT(t) the transmitted a7 is just the ratio of the total transmitted energy wave- for the two wav ef 0 rms. It is adequate convolution. Defining for our error More calculation complicated a new weighting we may now express procedures Eq. (B-38) areused bytakingp~(r) in practice outside the (Appendix). function P!W(T) by PW(T) PL(T) to simplify deconvolution = [P w (T)* * Pn(T) (B-4i) Pn(T)]T=o the incoherent scatter correlation function intemns of the estimates AR(T) and AI(7), <AR(T)> +I<A1(T)> (B-42) P~(T) = &(T) 87 —... with 8(T) Because erable (B-43) . our aim is to estimate to use PS(?) and covariances rather the incoherent scatter place of the signal-plus-noise in for the variances quantity o; C17P’W(T) = of AR(T) and AI(T). than U2 which is not directly correlation correlation function P*(T), It is also preferable The required est~ated. it is pref- function p (?) in the expressions to use tbe estimated substitutions are (B-44) o:Pm(T) p(r) . 2 the basic -pulse noise-to- x = 0:/0: Some additional because the lag signal (B-45) is x, . (B-46) .. simplifications may be made in relation T at which AR(T) the pulse separation. ratio + W*(7) a~+x +.; ‘m where @(T) p~m + u: Pn(T) = and AI(T) 2n this case, [Eqs. (B-i) and assuming to F, and (B-2)1 perfect G, and H Mode experiments are rectangular measured pulses is the same as and no transmitter chirp, (B-47) PW(T) = i/ar and a7 . i + T/TB for T < 7B . 2 where Tn is the basic pulse length. calculations, is P’W(T) s (B-48) for 7> TB A further Thus, PW(T). simplification, which should be adequate !3(7) ~ 1 and Eq. (B-45) for error becomes P~(T) + XP*(T) p(T) VIII. SUMMARY The results AI(T) rized. [Eqs. (B-1) = OF RESULTS FOR for the expectation and (B-2)], The noise subtracted <AR(T)> (B-49) a~+x UNNORM.ALIZED and variance of tbe u~Ormalized and the unnormalized term is neglected FUNCTIONS amplitude and p(r) cOrrelatiOn B(’r) [Eq. (B-34)] functiOns AR(T). are now summa- is taken as unity. (B- 50) = U;P,R(T) M-i U;(T) = ~ 0; S(CIT + X)2 + p2(T) – q2(7) + ~ s I + Sq2(k@ ~ (M– k) [Spz(kc) k=i + p(T + kc) p(T –kc) - q(7 +kc) (B-5i) q(7 –k<)] 88 -..-..-—-—— .m.=— <AI(T)> O:(T) = (B-52) @:P,I(T) M-i ~ = ~MK~ S(a7 + X)2 + q2(T) + P2(7) + ~ ~ ,(M,-k) [:qz(kc) k=l + Sp2(k@ + q(7 + kc) q(’f – kc) – p(T (B-53) kc) p(7 -kc)] + 1, M-i 4 U;,l(T) = ~ ‘b ZP(r) q(7) + ; ~ s .x[q(7 <B(T)> 0:(7) = (M –k) [p(T + kc) + p(T - k~)] k=i +kc) +q(r– (B-54) kc)] I U: [P:R(T) +P:l(T)]’/2 ‘[ P:R(T) = U:(T) +P.&) (B-55) C:(T) + 2P~1(7) P~R(T) mR, r(T)] (B-56) [P&&) )1 +P:l(T where P(T) = P ,R(7) q(r) The various = P&) + X/InR(T) + xPnl(T) quantities appearing PJT) = P~R(T) + IP~l(T) Pn(T) = P“R(T) + IPnl(T) (B- 57) . in these expressions The incoherent The system and pn(T) P:(-T), x scatter correlation noise correlation are both Her-mitian, Pn(T) Noise-to-signal altitude are: function function. i.e., PS(T ) p~(~) . = P:(-T) power at the measurement using the basic pulse length TB [Eq. (B-46)] % s Ratio of total transmitted energy when lag T is measured to the total transmitted ergy when lag O is measured Clutter subtraction constant; clutter subtraction term S . 1 when no clutter (Modes 89 G and H) [see en- Eq. (B”-48)] S = 3 when a is used (Mode subtraction F), is made I TABLE B-2 VALUES OF TB, S, AND C. FOR THE 2-PULSE MODES I F c ‘B (pee) Mode 40 F (s.;) s 0. OB 3 G IN 0,08 1 H 200 0.16 1 TABLE B-3 VALUES OF M AND e VS ALTITUDE AND LAG FCf THE 2-PULSE MODES Center Altitu&* Mcde F (:=; (km) IM (N/A) 130 (lo6)t0 All to 124 (N/A) 0 to 16.5 Oto All BO 40 120 to 380 Al I o to 50 to lootolw 1 3 6ot0100 H (p:.) 1 loot03Bo 130 (104) to 165 G M 1 3 20 5 20 40 1 %3 3 20 5 20 *Numbers in parentheses apply to daiu COI Iected prim to 30 December 1970. note i“ Table Ill OF rnai” text regarding alti~de charges prior to 16 December see also F;g. 1 ond Table Ill ;“ main text, 1 ) , 4 , 90 20 See 1971; M Number of prod.cte for a particular Time G formed sweep lag and altitude shift between a particular in & successive products lag and altitude, for only defined ifM>i Number KS of sweeps for which lag sured KS = t~/C~, where ? is mea- t~ is the integra- tion time in 6 econds and C~ is a constant d Variance (power) of incoherent nal at the measurement altitude scatter sig- for the basic pulse transmission Table .B-2 gives functions of altitude To evaluate TB, S, and c~ for Modes. F; G, and H, while” Table and lag for. Modes the variances know these functions are the required AnR(T)/’AnR(0) exactly,. M“ sfid” ~ as F, G, and H. we need to.have values of P6(T), and x. Pn(r), tbe best that can. be done is to use quantities Thus, values. B-3 gives we mayuse for OnR(~ ), Anl(T)/AtiR(0) AR(’r)/AR(.O) forp~R(7), Sirice we do not whose expectations A1(:)/A~(0) for PiI{T”)”, and ARn(0)/[ AR(0) -. ARn(o)l fOr P~1(7), ~~~ fOr X, where Kn AnR(T).= ~ ~ [Y\(r) n i=+ K AnI(T) = + Y\(t + T) + U;(T) U~(t +.T)] n. [Y\(7) ”U~(t + T) – Y:(t ~ + T) U;(t)] . (B- 58) How ever,. o~. drops n. 1,4. AR(0) could be used for the. remaining sions for the variances cwant.i.ty,u~. of tbe nm-maliz ed mm-relation function out of the exPres - (see next section) and this substi- tution is not required. 1X. NORMALIZATION Final estimates tained by dividing Two methods u: “ measurement, or B(T) AR(Tf) A. of the amplitude the unnormalized of deriving and AR(T2) depending F only tbe real correlation-function estimate 8catter correlation function may be ob- Eqs. (B- 50), (B- 5.3), Or (B-55) incoherent scatter by an estimate Of one method uses the zerO lag correlation on the mode of tbe experiment. as discussed function. Either Tbe covariances AR(T) of ~ Sec. V Of this append~. Lag real part of the measured and may be used to normalize Normalization ESTIMATES of u; are considered; and A1(T2) are ignored, Using Zero the unnormalized of u: In Mode estimates an estimate and of AR(Ti) Normalization 1, of the incoherent the other uses a theoretical has to be normalized, AR(0), estimate OF CORRELATION-FUNCTION the estimates of Real Part of Correlation correlation correlation function AR(T) function at zero lag, is an at all other lags. Function is measured. ad the incoherent scatter is 8R(T) where I ,- +&) . PR(T) = ~“ (B- 59) The mean and variance covariance of AR(T) of 2R(T) are calculated using Eqs. (B-3i) and (B-32), ignoring the and AR(0). <AR(,)> <bR(T)> = — <AR(0)> P~R(T) —+— z“[ = <aR(T)> <AR(0)>Z ‘R = Note that as p~l(T) and u:(O) In Modes and PnI(T) scatter PA(T) correlation (T)] andvarianceof$A(T) correlations Normalization B(r) =qiij {0:(, + Al(?) +P:l(T)] 1 u;(O)} (B-64) . Function that a theoretical to playa O;(T) <B(T)>2 incoherent when searching to find the one which is the best fit to the measured of AR(T), (B-62) . . +[P:R(T) a Theoretical This would be the situation VAJeS using (B- 63) <AR(0)>2 we assume and the amplitude are measured, +P~I(T)l = ~sing Is Measured i]z ‘[P:R(T) . .& Function are PA s“red in the expressions i/2 AR(0) (T) In this case, we may take them as zero Correlation u,? <8A(T)>2—+) available. (B-61) function may be estimated +A; = <6A(T)> B. I . are not estimated. When Complex [.A&) A The .expectation z [Eq. (B-51)]. G and H both real and imaginary of the incoherent <AR(T)> 0;4[+) +P:R(T) 0:(01 Normalization z. o;(r) +0) 0/?(7) for o:(T) (B-60) . scatter correlation through alihraryof function. part in determining function pt(r) theoretical This approach the estimate of o; is functions allows all the mea- rather than just AR(0), Thetheoretical rectly for F-Mode be zero. function pt(r) analysis, For G- and H-Mode the amplitude is for zero since imagina~y analysis, doppler shift and is wholly correlations thecomplete is [pt(T)\ and phase is ~(~). ‘?2 real. are not calculated theoretical function pt(r) isuseddi- and are assumed is /pt[T)/eiO(T), to where ,,,, ,, 1. Normalization The estimate of o; ~ When Only Real Part of Correlation is taken as the value Wi [AR(Ti) - Function 11 a!t which minimizes Is Measured the sum apt(Ti)12 o (B-65) i The sum is over It is convenient all measured values of Ti, of Ob because a~(ri) and the weights Wi are proportional to U~2(Ti). to use 4 ‘h ~,=T 1 ‘R(Ti) which makes VIi independent 0b4. Solving for ‘va;’ (B-66) “ Z ‘ipf(Ti) value of a is given by -0 when : Thus, a term ‘R(7i) ~ . ~ ‘iPt(Ti) The expected contains a is an unbiased estimate Pt(Ti) of al - PSR(Ti) (B-67) . if the theoretical function Pt is close tO the true functiOn as will be assumed. pSR, The variance of a is o: Z ‘fPf(Ti)‘~(7i) - [2 wiP:(Ti)]’ ~: _ 4 ‘b x Since a is an estimate “ PR(Ti) Expressions (B- 68) wiP;(Ti) of o;, = “ the normalized ‘R(ri) ~ for the expectation <Y~(Ti)> =P~~(Ti) A autocorrelation function ~R(Ti) may be defined as (B-69) . and varianCe of ~R(Ti) are (B-70) 0: (Ti) = P:R(Ti) Upo:(,i) t– o:(Ti) ‘R z P&(Tj) (B-71) u;%j) 1 j The method of normalization a small eters correlation (such as ion or electron ize the best -fit theoretical ances). Directly Typically, using a theoretical between the statistical temperature comparing the covariances <6R(Ti) - P~R(Ti)> <&(Tj) <~R(Ti) <YR(Tj) – P~R(ri)> function has an advantage of different or ion-neutral function are more the RHS of Eq. (B-72) Ieast-squares errors collision accurately - P, R(Tj)> - . 3.:(0) P~R(Tj)> is about 20, clearly frequency) determined for the estimates in that it produces This means that the param- lags. which character- (i. e., have Smalley Vari- $R and ~R gives ~ p:Ro;%i) . (B-72) 1 making it more desirable to use &R(r) in fitting than $R(r). Complete 2. Correlation The method used is similar The estimate of o; ~ Function Measured to that just described, is the value which minimizes Wi [B(Ti) - b[pt(7i) except that B(T) is used instead of AR(T). the sum (B-73) 1]2 i with .i=—“: (B-74) u:(Ti) As before, ~ Wi lpt(Ti ) [B(Ti) b= ~ \vipt2(Ti) and <b> =Su: Tbe normalized v PA(Ti) <YA+(Ti)> . amplitude (B-75) ~A(Ti) iS defined as B(ri) = (B-76) ~ = P~A(Ti) = [P~~(Ti) (B-77) + Pjl(Ti)li’2 P:A(Ti) U: ‘A (Ti) = U;4U&i) i [ – u;(Ti) z P~A(Tj) 94 oB2(Tj) 1 (B-78) —- For the same reason as discussed in the .laet section. A PA when deri vmg the best-fit theoretical function. x. ERROR IN PHASE OF CORRELATION The phase of the complex correlation it is desirable to use %A instead of FUNCTION function measured in Modes G and H is $(T)[1 .arc (B-79) tan ~ R The expectation and variance of O(T) are derived using theapproxhnate Eqs. (B-3i) apd (B-32), [.$ 1 .:(,) =-, 1[ -)+-) .p~:~’y(,,] . arc tan <$(.)> [ s s i-l XI. ERROR IN DRIFT The plasma (B-80) 13JT)2“2 P~R(T) (B-8i) “ VELOCITY drift velocityvd is related tothephase $(r) of the ~coherent scatter cOrrelatiOn function by (B-82) where A is the transmitted @(Ti) atanumbe~of lags ~ifromwhich The estimate weighted wavelength. ofd@/dT least-squares vd maybe is the slope sense, i.e., The G- and Ii-Mode experiments yield phase values estimated. S of a straight line fitted to the measured phases in a S is chosen to minimize (B-83) ~ Wi [@(Ti) - sTi]2 i with Wi . U~2(Ti). The solution for S is (B-84) ‘= ‘~wi,: 1 and its vm’iance v: is (B-85) 95 The drift velocity 1 is estimated using (B-86) I and its variance c; is given by (B-87) ! X11. ERRORS IN ACCURACY OF PARAMETERS DETERMINED REGRESS1ON ON A MEASURED CORRELATION FUNCTION Themagnitude i parameters –Ti tion estimates obtained 1 I { eters and vininthe the errors in the library discrete values lower f(x, y,zi) In searching data d(zi) variable (theoretical are squared, according Zi (lagri) incoherent are given correlation correlation of x and y, thedeviations weighted, and summed to yield two physical other altitudes. This secvalues in terms function) through comparison scatter . to the goodness-of-fit the derivations (measured LIBRARY to determine and Ti and Te/Tiat parameters For generality, from for the best estimates functions function is analyzed E-region, regression. of the independent functions library correlation in these physical x and y to be determined retical / of a measured FROM of param- measured with a library at of theo- functions), between the measured and the residual 4 1 6(X, y) = ~ j Wi[f(x, (B-88) y,zi)–ad(zi)J2 i I where the weights Wi and normalization constant a are given in Sec. IX of this appendix, Next, a function of tbe form :1 i ~ ! ~*: ::: q(x, y)= is fitted in a least-mean-squares of x and y are the values reasonable ,? Ax2+By2+Cxy+Dx assumptions, of the coefficients We assume f(x, y,zi) A, +Ey+F sense to the values ~ and ? which produce to estimate it is possible by m amount b(zi) mean. j that the probability 6(x, y). The least-mean-squares a minimum in the function 6(x, y). the statistical uncertainty estimates By making in $ and $ in terms B, and C in Eq. (B-89). that if x and y are given, ,1 (B-89) We also assume a measm’ed which is normally tbat the error of observi”ga point d(zi) distributed is independent given value d(zi) will deviate with standard oftbe fore errors from the true value deviation in adjacent O(zi) and zero points. It follows x and y is +zi) P[d(zi)lx, The joint probability product y] .P[8(zi)lx, of observing of these individual P[d(z)lx, y] aexp a complete probabilities, y]c — Zuz(zi) (autocorrelation) (B-90) functional(z) given x and y is the viz:- exp –~62(zi)/2u2(zi) [ [1 i 96 I . (B-91) ff we make various ing probabilities P [x, y] d(z)], tion d(z). hypotheses of obtaining regarding the resdt which is the probability d(z). the values of x and y, Eq. (B-9i) However, what we really that x, y are the correct values gives the result- want is the function given the observed func- We have P [x, yld(z)] = m P [d(z) lx, y] (B-92) . The .— a priori probability P(x, y) is essentially unknown, and the best that cm be done is to consider all hypotheses as to the true values of x, y to be equally likely. P [d(z)] is a constant, independent of x, y. Thus, P [x, yld(z)] where ~ K e -6(x, y)/2 (B-93) K is a constant. From Eq. (B-93), also maximize we see that the least-squares P [x, y Id(z)]. Thus, estimates ?, ~ are identical ~ and $ which minimize to the maximum-likelihood 6(x, y) eStimateS Of x and y. The eXact form of the function P [x, y Id(z)] tion is made that it is a bivariate A as ~ and Y, the variances as u: is difficult to determine, so the simple assump- normal distribution in the variables x, y. The means are taken 2 and the correlation coefficient as u*y. and o Thus, Y’ p[xyld(”)lm’’pl-2(,:.2 J(+++(+$z”xy(x:;:(y(B-94) XY and, from Eq. (B-93), (X–$)2 6(x, y) = + (i – U:y) u: From the fit [Eq. (B-89)] ‘y’ (,A~#2 _ –c. :, that . OF THE ERROR 2NTRODUCED INTO THE FUNCTIONS BY CLUTTER ECHOES F-MODE lntroducticm Clutter Sees.11-A, practice, subtraction returns III-A, are removed IV-B, through digital and IV-C autocorrelation algorithm. ffitering, andasindicated that points tlrat exhibit alarge the expected as explained by Eqs. (i6) clutter-to-signal it seems appropriate and (25). inthe main text in Ithas ratio often depart function by an am.mmt not explained Accordingly, 97 . (B-95) . (i – U;y) Oxuy dzm EST1MATION CORRELATION A. it follows (xi%)i” ‘y XIII. to 6(x, y), (y –9) t constant (i – U:y) u; - “x’ 0 20X (X-Q) (Y - ?)2 been found, considerably by the statistics in frmn of the clutter to give these points less weight than ._., .,,, F’ ) I 0.12 1 L(r)= +,:, [p(r, t8,p - P [T t p(r) I 0.08 ALL PO INTS, WEIGHTS= llu,~,o, X P (T)th,Or SUPPRESSING 20-, 40- 0 SUPPRESSING 20-, 40-vsec p (T),heo, 1 h VS LAG AUTOCORRELATION p iT),h,Or ● I I I I Psec.LAGS, WEIGHTS = I/w;heO, LAGS, WEIGHTS= 1/02 ● -0,04 t I -O,oe . o . ..lz~ o 40 sO 120 160 200 240 2S0 320 360 400 T (L!S,C) Fig. B-1. theoretical 0.14 Comparison of experimental autocorrelation function and its best-fit function –average difference vs lag; i4 April 1971, c/s <0.5. I -E&L I o A ,m, [T) = 0,12 ● USING 0.10 1 1 . ALL x SUppreSSing 0 SUPPRESS +! ,., 1 1 [p(ri)e,p -p(ril,ha., I I ]z POINTS LAGs, wEIGHTs= 20-,40-wc lNG20-,40-pSeC LAGS, WEIGHTS l/w~haO, =l/C2 0 ● xx 0.0s v 0 ~ a 0.06’ i? f ● 0.04 , G e .- ———+~—8— -.-/” r’*—r—*-— —A 888 0.02 0 ! 1 1 1 40 80 12o 160 I 200 I 240 I 280 I I 320 360 r [ we.) Fig. B-2. theoretical Comparison of experimental autocorrelation function – rms different e vs lag; i 4 April function and its best-fit 197i, c/s <0.5. 98 4 d —- . . ., ,. ,... one would otherwise This section B, assign to them when matching outlines Experimental In order studied proper functions in some detail. Statistical weights in Mode rived self-clutter, in this way, the threshold Figure and a d-c clutter contribution ratio c/s of >0.5 were were examined deviations estimated accordtig between theoretically to Eq. (B- 51). Using theoretical tbe theoretical including taking However, fun~tions de- and experimental those for which c/s exceeded value of 0.5. B-1 shows the mean deviation Only points where that at ‘r . 40 pew pec A). the theoretical ~ec points as a square fit eliminating that major An aver-age is probably pulse assumed, the 20- and 40- @ec less than O.Oi, indicating fit. a“d particularly for this discrepancy the perfect that especially points shows an systematic errors have for the case when c/s < 0.5. The rms difference Arms between the experimental and theoretical tions is shown in Fig. B-2 for the same data as in Fig. B-i. c/s were and experimental wsed in tbe least- sqwares The reason from A revised that is gemxally now heen removed too low. pulse shape departs (see Appendix deviation between c/s c 0.5 were It is evident that the point at T .20 am systematically the actual transmitted small days in f 97 i were below i25 km. were ignored. for all points, to experimental on four sample to altitudes functiom of 120 points was used at each lag. average was limited the random and systematic f“nctiom function of time lag. for T .40 functions. are established. fit of theoretical correlations F autocorrelation all points with a clutter-to-signal autocorrelation for the least- squares F, the obsewed The analysis em-m-s in Mode account of noise, to theoretical weights Study to derive autocorrelation the observations the manner in which the appropriate used, Arms is close to the expected autoc orrelation func - Since only points with relatively value @theOr, at least for lags shorter than 300 psec. The effects autocorrelation c/& of clutter The latter 0.6, 0.6to c/s >2. were functions. parameter 0.8, 0.8 to i, Thus, was quantized linear relationship model will suffice of systematic the fit. Arms is seen to exist between to estimate errors The rms deviation As expected, Arms the additional in the measured 2 ‘c = A:ms as a function of 0.2 to 0.4, 0.4 to deviation with c/s >2 for c/s >2 is apparent at is about 2 per-cent of could be removed by elim- as a function of c/s is shown in Fig. B-4 is close and c/s, statistical O to 0,2, A systematic to Ctheor for small indicating error c/s. that a simple introduced autocom. elation points due to clutter – ot~eor Four days of measurements A roughly empirical by clutter. were O:(T) = 0.0i3 as (B-96) used in deriving the empirical is defined . B-5 shows the data points on each day, fitted to the data gives the empirical and the average relationship over all days. between u: and c/s. The straight line relationship (c/s) . (B-97) 99 . ranges: and theoretical calculated Results The variance Figore in the following the number of samples the effects with Utbeor. the experimental deviation i to 1.5, 1.5 to 2, 2.5 to 3, >3. inating all these points from and is compared by comparing B-3 shows the average For the four days studied, the total points. c. then obtained Figure 0.12 ! 1 I I I -FmmL J 0.08 8* . . rZ[CA) = +,:l[PICA!T )exp- p (r, l,,.., ] i t 20-, 0,04 40-M,.c LAGS . wEIGHTS 0 WEIGHTS= SUPPRESSED = \/o,&O, I/vz [ ● io e -0,04 -0,08 0 ● [ 1 -0.12 0 0,s I I I I 1.0 1.5 2.0 I 28, I I I I I 1.?6 66 43 59 2) NUMBER Fig. B-3. Comparison of experimental best -fit theoretical function – average 0.30 1 1 I 3,0 2.5 ,/s I ,8,1 d 0 1 la I 35 OF SAMPLES autocorrelation function and its cliff erence vs c/s, f 4 April i 97 i. I I 1 1 118-9-665$ t Arm, (C/S I = 0.23 20-, 40- WEIGHTS ~,~, “,,. [pi./., LAGS rllO.O -p (T( ),h,.r 12 SUPPRESSED = l/u:heO, a20 35 !8 21 ~. z 0.15 59 )46 z —— 0.10 43 281 — 951 ~-———— 0.05 0 —— ——————— rm. r ————— —————- ———————— i I I I 1 1 0.9 1,0 1,5 2.0 2.5 3,0 C,A Fig. B-4. Comparison of experimental autocorrelation best-fit theoretical function – rms difference vs c/s, i 00 function and its i4 April i971. 1 1 u: - A:., 20-, 0.0s 40- I I ~ - U,:W “s,, LAGS S“PP~SSED o 25 M4RCH 1971 . 21 W2H4BER !971 A 13 APRIL 1971 . ~, 1$71 DECEMBER 0,04 . 0 ~b” ~,o, 0 / 0,02 x 0 x 0.01 1 0 ,.0 0.5 1.5 2.0 2.5 m ,/s Fig. B-5.. Comparison of experimental best-fit theoretical function - variance Thus, the total error in the autocmn-elation U*(T) = ut:mr(T) where o~eor 1/02 (T) instead The elimination .,.! I I and u: it was decided of l/ut~eor (r). function as a fumtion of lag is given hy + 0:(7) is given by Eq. (B-51) Based on the foregoing, autocorrelation function and its increase vs c/s. Also, of points c/s >2, (+-98) is given by Eq. (B-97). to weight the F-Mode correlation the 20- and 40- psec lags were while probably desirable, omitted function points by from was not implemented. consideration. 3lBLlOGRAPHlC DATA SHEET 3. Recipient’s Accession No. I. Title and Subtitle 5. Report Date 23 February 1979 Incoherent Scatter Measurements of E- and F-Region Density, Temperatures, and Collision Frequency at Millstone Hill 7. Author(s) Joseph E. Salah John V. Evans William L. Oliver Ronald H. Wand ). Performing Organiiation Name and Address Lincoln Laboratory, M. I. T. P.O. Box 73 Lexington, MA 02173 6. 8. Performing Organization Rept. No. Technical Report 531 10. Project/Task/Work Unit No. 11. Contract/Grant No. ATM 75-22193 12. Sponsoring Organization Name and Address National Science Foundation Atmospheric Science Section Washington, DC 20550 13. Type of Report 8r Period Covered Technical Report 14. 15. Supplementary Notes 16. Abstracts The Millstone Hill incoherent (Thomson) scatter radar system has been operated since 1963 to perform a synoptic study of F2-region electron densities, and electron and ion temperatures. These measurements have been conducted by transmitting single long pulses and performing digital sweep integration and spectrum analysis of the reflected signals. This report describes changes made to the system in 1969 to permit extensions of the measurements to altitudes below 200 km (i.e., the E- and Fl-regions). These changes include modifications to the radar timing equipment to permit the transmission of close-spaced pairs of pulses from which the echo autocorrelation function can be determined in the computer by processing appropriately spaced pairs of echo samples. By operating the radar coherently, it has also been possible to subtract the unwanted (stronger) clutter signals from the ionospheric echoes. Samples are presented of the results that can now be obtained from these programs. 17. Key Words and Document Analysis. Millstone radar E-region F-region diurnal variations electron density ionospheric scatter seasonal variations temperature effects spectrum analyzers 17a. Descriptors 7b. Identifiers/Open-Ended Terms 17~. COSATI Field/Group 19.. Security Class (This 18. Availability Statement ‘DRM NT1535 (REV. 10-73) ENDORSEDBYANSI ANDUNESCO. Report j U N % Security Class (This Page UNCLASSIFIED THIS FC :MMAY BE REPRODUCED 21. No. of Pages 108 22. Price VSCOMM-DC 82SS-P7A