APPLICATIONS OF PLASMA DENSITY MEASUREMENTS TO SPACECRAFT RADIO TRACKING by Thomas Marshall Eubanks B.S., Massachusetts Institute of Technology (1977) SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS OF THE DEGREE OF MASTER OF SCIENCE at the MASSACHUSETTS INSTITUTE OF TECHNOLOGY June 1980 \c, Massachusetts Institute of Technology Signature of Author Department of Earth and Planetary Sciences May 26, 1980 Certified by Irwin I. Shapiro Thesis Supervisor Accepted by Chairman, Departmental U)ndgrer MASSACHUSETTS INSTITUTE OF TECHNOLO Y JUN 19 1980 LIBRARIES Graduate Committee -2- APPLICATIONS OF PLASMA DENSITY MEASUREMENTS TO SPACECRAFT RADIO TRACKING by Thomas Marshall Eubanks Submitted to the Department of Earth and Planetary Sciences on May 26, 1980 in partial fulfillment of the requirements for the degree of Master of Science. ABSTRACT S-band plasma delays are estimated as part of a test of the general relativistic time delay effect conducted during the Viking Mission to Mars. The processing of radio tracking data taken with the Viking Orbiters and Landers is discussed. The statistical properties of Viking Orbiter dual-frequency It was condelay and Doppler measurements are described. cluded that the plasma delay can be adequately modeled as a random walk. The implications of this model on estimation of The results of Viking Lander plasma delays are discussed. use of the random walk model for Viking lander plasma delay correction are compared with the results from other plasma models, and it is concluded that this model is sufficient for estimation of Viking Lander plasma delays. Thesis Supervisor: Title: Dr. Irwin I. Shapiro Professor of Physics and Geophysics -3- TABLE OF I. INTRODUCTION II. THE A. B. C. D. E. CONTENTS OBSERVABLES Introduction The Propagation of Electromagnetic Waves in a Plasma of the Group and 1. The Definition Phase Velocity 2. Group and Phase Velocity in the Coronal Plasma Group Delays and Doppler Shift caused by the Solar Wind 1. The Thin Screen Model of the Solar Plasma The Measurement Apparatus Terrestrial Propagation Effects 9 10 11 12 17 18 24 27 III. COMPUTER DATA PROCESSING A. Introduction B. Data Collection C. Data Editing and Calibration 1. Range Calibrations 2. The SX Bias 3. The PRA Demod D. Data Editing 32 32 32 35 35 39 40 41 IV. THE STATISTICAL NATURE OF THE PLASMA DELAY A. Introduction B. The Plasma Autocorrelation The Autocorrelation of the SX Delay 1. 2. The Autocorrelation of the SX Doppler C. Estimation and Smoothing of a Random Walk Process 46 46 46 48 51 V. LANDER PLASMA CORRECTIONS A. Introduction B. Computer Processing of Plasma Corrections C. Lander Residuals and the Plasma Corrections Experimental Tests of Our Conclusions D. 60 60 60 CONCLUSIONS 73 VI. 64 70 -4- Chapter I Introduction advances in many fields have In the past twenty years, made possible an enormous increase in the accuracy of meas- urements of positions and velocities of objects in the inner solar system. An important factor in this progress has been the placement of the program of probes throughout the solar system during interplanetary exploration initiated by the United States in the early 1960's. Radio tracking using spacecraft transponders makes possible accurate measurements of delays and Doppler shifts of signals propagating between a ground station and the spacecraft. With probes in interplanetary space, the solar system can be used as a vast laboratory for gravity research, cluding research on general properties of the planets. relativity dynamical the and in- The solar system, not being under the control of the experimenter, is poorly designed for such experiments. One major complication to the interpretation of present day radio tracking data is the effect of the inter- planetary medium on propagating radio waves. of the total plasma spacecraft given delay from downlink another problem addressed delay in for delay observations from only spacecraft this The estimation measurements is thesis. the of major the one plasma experimental This work was done in connection with a test of general relativity conducted during the Viking mission to Mars (reference 1). -5- At present, the solar system is used for the most definAs was first shown by I. itive tests of general relativity. I. the mass of the sun causes Shapiro in 1964 (reference 21), an -increase in the radio propagation delay over that expected from Euclidean delay occurs geometry. at superior directly between the close to the The sun. maximum conjunction, earth and Mars and For Viking, about 250 microseconds excess when the the the delay relativistic sun moves raypath passes at that time (Usec) while the corresponding is total Unfortunately, the round trip delay is about 2500 seconds. plasma effect is also at a maximum at superior conjunction, with the greatest measured plasma delay being on the order of 100 Usec at a radio wavelength of 12 cm. mates of the delay are spacecraft and plasma thus Accurate esti- vital the to general relativity experiment. The Viking under good conditions, possible, round trip radio propagation ground equipment to measure time with an the make it Earth-Mars uncertainty of about 10 nanoseconds (nsec) and the carrier frequency Doppler shift with an uncertainty of propagation medium makes it on the order impossible to of 1 mHz. infer The the vacuum range and line-of-sight velocity to that accuracy, and currently constitutes the largest source of error in the interpretation of interplanetary radio tracking data. The propagation effects from the medium between Earth and Mars are dominated, at radio frequencies, by the the -6- contribution from the interplanetary plasma in the solar co- atmosphere terrestrial the parts of the propagation medium are major The other rona. and dual-fre- The ionosphere. quency data includes a contribution from the ionosphere but not from the from contribution non-dispersive the Earth's neutral atmosphere. solar The plasma is a highly is proportional to the excess plasma delay or Doppler shift inverse square of plasma density. the carrier The dispersive medium. frequency, and to the local The solar wind is very complex, with density It is impos- fluctuations at times of the order of the mean. sible to adequately estimate the solar plasma delays purely from a time terrestrial solar wind difficult averaged density atmosphere from first than modeling as model, propagation principles the weather is delays. would on with the model the done To probably Earth. It be more is thus necessary to consider statistical models of the plasma delay and delay rate, similar in spirit to the models discussed in references 13 and 14. This thesis is concerned with Viking Orbiter and Lander radio tracking data taken between July 20, 1976 and September 3, 1977. Viking Lander (VL) 1 and Viking Orbiter 1 were (VO) launched as a single spacecraft on 20 August 1975 and were inserted into. Martian orbit on 19 June 1976. VL2 and V02 were launched together on 9 September 1975 and were inserted into Martian orbit on 7 August 1976. VLl landed on the -7- surface 1976. able of Mars on 20 July As of January 1980, V02 of communicating The Landers are at S-band by VL2 on 3 September is inactive and VL2 is incap- directly take part in these radio ing followed and Earth with is tracking experiments. equipped for interplanetary radio (12 cm) only. The Orbiters have, track- in addition, a coherent dual-frequency downlink, at S-band and X-band cm). Differenced measurements can to unable dual-frequency provide (S-band minus X-band or estimates of the time delay (2.3 SX) and Doppler shift contributions from the interplanetary plasma. On November earth and through Mars 25, 1976 and again (together with superior conjunction. on January the Viking The 21, 1979, spacecraft) plasma and the passed relativistic effects are at a maximum near superior conjunction, while the signal-to-noise ratio tion and solar is radio lowest there (due to plasma attenuaDespite interference). this, it was possible to track the Viking spacecraft to within 2 or 3 days before and after superior conjunction. The SpaceCraft designation (S/C) numbers are used as an alternate for the Viking spacecraft: VLl = S/C VO1 = S/C 27 26 VL2 = S/C 29 V02 = S/C 30 The Orbiters are. subject to unmodeled accelerations such as gas leaks from the attitude control system, which compli -8- cate the interpretation The Orbiter Orbiter, of Orbiter SX data are not since the range or range and affected by the Doppler data. motion of the Doppler shift to the spacecraft The cancels out in the differencing. Landers, which cannot make dual-frequency measurements, are fixed on the surface of Mars, which is nearly free from stochastic accelerations. is thus necessary to plasma estimate Lander delay measurements for both the corrections uplink for It the and downlink from Orbiter downlink dual-frequency measurements. Chapter differenced II will observables discuss the that can and how the measurements are made. computer processing basic observables, be constructed the from them, Chapter III discusses the required before the dual-frequency ter data can be used to obtain plasma delays. Orbi- Chapter IV gives the results of a statistical study of the plasma data, and Chapter V describes the application of these results to Lander range measurements. -9- Chapter II The Observables A. Introduction Radio tracking of interplanetary spacecraft provides two observables, the round trip propagation (group) delay and the Doppler shift of the carrier frequency. The group delay and Doppler shift are a function of the group and phase velocities, respectively, in the propagation In a tenuous plasma, signal path. medium along the such as the solar corona, the phase and group velocities are displaced by opposite and nearly equal amounts c, the from velocity of light. phase velocity is greater than the group velocity, which The is the velocity of energy and information transfer in the medium (reference 6). In a tenuous plasma, the group and phase velocities can be approximated by aN v phase = c(l + e) 2f (2.1a) and aN v group = c(l - 2f2e) (2.b) where c is the velocity of light in a vacuum, f is the carrier frequency (Hz), a is a constant equal to 8.1'10 , and N is the electron density in electrons cm-3 e -10- On Earth, good vacuum, usec to tion. model solar but the the total The coronal and corona range plasma delay density fluctuations difficulty is Orbiters ranging taken have the code near especially ability frequency measurements using the Viking to at of the Orbiters are the a very 100 conjunchard mean severe for in the to some Viking sensitive to delay range conjunction. The Viking a received S-band and X-band. retransmit both superior notoriously exceeding superior coherently plasma delays, and considered near is relativity experiment, which is most measurements be would coronal plasma can contribute up to S-band shows This cases. the S-band group delay and Dual- Doppler shift used to estimate Viking Lander improve estimates of the true range to the Landers. In this chapter I will first discuss the nature of the observables, both the delay and Doppler shift and the differenced dual-frequency between second the part observables, observables of this and the chapter, I including the connection propagation medium. will discuss the In the techniques and equipment used to make observations with the Orbiters and the Landers. sing required The to next chapter concerns handle available from the Viking mass of computer proces- dual-frequency data experiment. The Propagation of Electromagnetic Waves in a Plasma B. In of the the an a vacuum, there electromagnetic is only one signal, c. In velocity a of propagation neutral medium, such 'as the Earth's somewhat atmosphere, less than c. In the a velocity of tenuous propagation plasma, there is are (at least) two velocities of propagation, the group and the phase The group velocity is velocity. of wave packets or of velocity of the modulation of the carrier; delay is the round-trip propagation time. is the velocity wave crest is of propagation indistinguishable ured modulo only the change in the The val. of 21r), the concepts total propagation The phase velocity wave crests. (the phase phase delay the group Since can only be cannot be group and phase meas- measured, phase delay over a measurement of each inter- and velocities the expressions derived for them are approximations, which become less valid strength as the plasma increases. In density Section or field I these B.2 external magnetic will show that approximations are precise enough for the Viking radio propagation experiments. (Section 1 is adapted from references 6 and 7.) 1. The Definition of the Group and Phase Velocity Assume u(x,t that at time to wave packet can be described ) which has Fourier transform of A(k)- 1 /u(x,t )e -ik.x 3 - -dx -I Here k is tude of the wave vector, k, and the in units of cm wavelength, solution to the Helmholtz wave is by X, is equation 1 , k 2ir/k. for a (2.2) is the The magnigeneral traveling wave -12- u(x,t) where w is number k. the - A(k)e /2w angular is equal a tenuous vphase < c. c, plasma, vphase with to ck, but, the wave- in general, w Sw(k) k(2.4) c > and, (2.4) in a neutral medium, Although the phase velocity can be greater information is propagated less than c, (2.3) The phase velocity is defined as Vphase In k frequency associated In a vacuum, w is a function of k. - O than the group velocity, which at is and the postulates of special relativity are not violated. If compact the wavenumber and centered A(k) distribution some about value of some k 0 , then w(k) can dw (k w(k) = w(k ) + dw - (k - + higher order k (2.5) terms terms, the integral in Equation 2.3 can be performed to give - w(k )) u( ,t) - e -- 0 be k 0 it(k is expansion in k: expressed as a Taylor series Ignoring the higher order signal u(x dw d-w dke k 0 t,to) 0 (2.6) -13- A comparison with the original signal u(x,t ) shows that, to within a phase factor, the wave packet travels undisturbed at the group velocity, which is thus defined to be dw Vgroup If the higher order terms tant, then the wave d (2.7) in the expansion of w(k) are impor- packet will change shape as it travels and the group velocity may lose much of its meaning. with 2. Group and Phase Velocity in the Coronal Plasma The solar about 4% corona consists ionized helium of largely by weight ionized and hydrogen, essentially no (about 10 This plasma is so tenuous component. -3 electrons cm 3 at 1 A.U.) that interactions between particles neutral in the gas can be Although the free particles. the is 10 6K, it of order it can be treated as a sea of and ignored, wind has solar not temperatures relativistic medium, a on since thermal velocities are on the order of only 4000 km sec-1 for electrons ics can corona. at that be used The temperature. to treat following Thus radio classical electrodynam- propagation discussion is adapted in from the solar reference 7, page 210 and following. Let r describe the position coordinate system centered upon the position of p. Under the of is particle, p, in a instantaneous equilibrium influence wave, the equation of motion of p a of a propagating radio -14- m dL -eB dt 2 m where and tively, B is of direction E is e the are the propagation (2.8) -eE mass particle's magnetic external transverse the x dL dt c- field field of respec- charge, parallel to constant), the and be a the propagating wave. to (assumed electric and the transverse magnetic field of (Notice that we are ignoring interactions.) the radio wave, as well as particle Equation 2.8 can be solved to yield e(mw(w r = wb where ejB)/(mc) is the 1 w)E (2.9) frequency of particle in the external magnetic field, frequency. precession of the called the cyclotron The displacement of all charged particles in the plasma gives rise to a net dipole moment and thus to a macroscopic dielectric constant of 2 w = 1- w(w P + (2.10) wb ) where w = (Tne2) 1/2 = 5.64'10 4 /n radians sec is called density refers the plasma (particles cm to the (2.11) (for electrons) n frequency, and -3 ). (The ± in is the particle Equation 2.10 two senses of circular polarization.) number and 2.11 -15- kc w - the Using expanding given definitions in a power series in Vphas e k phase v 1/2- 2 - 1/2 2 wb w w w + .) 2 w and 2.7 2 w2 (12--+3-b...) 2 w 4 1/4-- + 4 w 2 w and 2.4 Equation (_- + wb w c(l group + (2.12) in wp and wb, we get 2 w c+ c(1 w the wavenumber k is E, and relation between w, The ... (2.13a) w 4 - 1/8 w 4 + ... (2.13b) w and dw d - = w + order first The in terms the distortion that is a measure and expansion of w(k), term in w rotation. Faraday to I will a propagating wave packet. in second-order the correspond wb T is the next term in the of (2.13c) cw dk and dominates, that show T, the distortion term, is negligible. Clearly the higher order the radians solar the sec-1 and radii Under wind, solar trons/cm3 (R ), and these the At increases. density plasma terms will be more important as Ak is which the equal is conditions, 2"10 7 radians sec -1 . w Thus, Thus, 4 *10 - as deep as average average to S-band, = 210 7 is density field is radians considering considering 2.3109 2r radians cm - At is 105 about 1 sec 5 data probes the Viking plasma magnetic 4 w -1 the electrons and elecGauss. wb only, the electrons only, = -16- 2 w P - 1. 510-6 (2.14a) 1.2"'10 (2.14b) w and TAk = 10 Only the first term in w p -9 << v group - c (2.14c) need be retained. To first order, therefore, the group and phase velocity differ by equal but opposite amounts from c. group velocity, the velocity of information tenuous plasma, is less than c, as expected. Note that the transport in a For the rest of this thesis, we will assume that the group and phase velocity in the solar wind are given to sufficient accuracy by Equation 2.1a and Equation 2.1b. In this part of this chapter, we have used CGS units. From now on we will use natural units in which c = 1. w/27 If f = is the radio carrier frequency, then Vphas e = 1 + 4.03038'10 72 (2.15a) v = 1 - 4.03038'10 7 ne (2.15b) and group f2 -3 with n in electrons cm-. e C. Group Delays and Doppler Shifts Caused by the Solar Wind If will be R is the true range delay, the measured delay, r, -17- R m R = ds S0 Vgroup - 0 2 ds(1 + 2w (2.16) ) Le t (e ) 7 R P = 2m ff/ O n ds = 0.03038 R 107 e e ds (2.17) O P is proportional to the integrated columnar content (the4R ne ds) thus t -= R + P/f 2 (2.18) R is independent of f, and therefore P can be estimated if m is measured for two different frequencies. The derivation of the Doppler phase delay proceeds much the same fashion as that of the group delay. in The total phase delay cannot be measured, only the relative delay from the start of the observation session or pass. Since it is not possible to continuously monitor the phase delay from the start of the mission, the total phase delay is unknown. If r p is the phase delay at the ground station, then the Doppler shift measured, Dm, is m From dual-frequency ferred. dr dR dt adt dP 1 w 2 dt(2.19) Doppler measurements, It is possible in theory to use ments to find the initial value for P, dP/dt can be in- SX range measure- and to use the more -18- accurate Doppler This process is provides most the to measurements called Range accurate estimate Integrated estimates the cycle slip problem, which the change Doppler of P. is discussed (or in P. RID) and Unfortunately, in Chapter III. D, has prevented use of the Doppler measurements, and the potential accuracy of RID measurements has yet to be achieved. delay 1. The Thin Screen Model of the Solar Plasma The Viking over the corrected for Orbiters downlink plasma can be used only. delays The for to measure measured both the the range plasma must uplink and be the downlink, and thus the uplink plasma delay must be calculated from downlink = Pdown) the to model assumes that same as find plasma delay measurements. a consider the the The approximation some model of the (or Pup uplink plasma delay, (measured) downlink delay, better static to solar Pdown" P In ,upis order Pup' it is necessary corona and of the to measure- ment. Experimental based term and studies (reference 20) using both ground- in situ spacecraft measurements show that the average coronal electron number density can be long modeled by n where r is model of the (r) = in 1.55 10 solar 8 + 3.0 10 r r radii. If we 6 electrons cm use Equation local plasma density, it is clear contribution to the integrated -3 2.21 (2.20) as a that the major plasma density will come from -19- near the thin-screen point, the point on the raypath closest Figure 1 shows to the sun. the observation geometry. The vacuum delay is R. = R i P 2 are P 1 and R + up down called (A. + B ) + (C. + D.) = i 1 1 (2.21) 1 P2 ' is thin-screen points. the the thin-screen point for the downlink matched to the ith uplink. Each thin-screen point has a location in time as well as in. space, and the ith thin-screen time is the time at which the signal passes and the through P1 and P 2 ' are matched Pi . separation spatial between in time, points thin-screen is The thin-screen model assigns all of Pup to P1 and ignored. Pdown to P2 and ignores contributions from other parts of the This model is most realistic near superior conjunc- raypath. tion, when the Far from raypath nearly grazes the conjunction, superior irrelevant at since such the times the limb of model thin-screen plasma sun. the and delay is delay rate are small and changing slowly. In assume our that implementation P1 is of coincident the thin-screen P2'' with which model, we amounts to ignoring the distance between P 1 and P 2 ' and any asymmetry in the plasma contributions away from the thin-screen point (which would contribute. at different times to Pup and Pdown ) . Ri is an estimate conjunction, we of can the use delay at Ri to time ti . approximate Near the superior thin-screen -20- delay Atts. Earth and If Re (= is the distance between 1 A.U.) the Mars-Sun is (= 1.5 A.U.) the sun at t. 1 and Rm the distance at ti,*then 1 C. 1 2Ri (R 2 + R 2 - R 2 ) i Assume that Ci = Bi; if Atts is m (2.22) e the thin-screen delay then 1 in seconds and 998 tts is r is measured in A.U., where is twice the conversion factor. The spatial separation between the Orbiters and Landers is small, about 5 104 km at most, and the spatial separation The velocity between the thin screen points is even smaller. of the solar wind is about 400 km sec - 1 1 A.U., at which implies that the spatial separation between the thin-screen points introduce With plasma a timing delay rate these timing errors would 35 nsec, which of errors on the order of 1 psec/hour introduce is not negligible. (a 2 minutes. value), large a plasma delay error We found of from numerical studies that our plasma estimates were remarkably insensitive to thin-screen errors on the order of an hour or less. It was decided to ignore the spatial separation between measurements and desired corrections for the present. In ponder. two-way ranging, the spacecraft acts as a trans- It receives the uplink ranging signal, and amplifies -21- and rebroadcasts talk between the the craft, frequency. to The the and Rup is and by turn-around carrier by b k (equal (equal ratios to and is P P up -~f+ f - down 2 (kf) (2.24) P P up down r x = Rup +R down +u 2 + (bf) 2 (2.25) up down and Rdown are the true uplink and downlink range, frequency independent effects have been ignored. and an total r are measured. Thus, Sx - estimate of delay Pdown tion. down f2 - only, which can be found 2 2 b as 1 1 ( ) (2.26) (2.26) k indicated if earlier. The S-band correction is 1 = f the x ), (rs - be equal to SXCOR In these +R R the is multiplied range measured at S-band = R cross- is Let SXdelay I multiplies downlink, Using to avoid receiver on board the space- coherently S-band X-band. s and at X-band In order received frequency for for Earth. transmitter and Equation 2.19, the where to transponder 240/221) 880/221) it back static model, 2.25 becomes which (P up + assumes 1 Pwn -7 down b that P (2.27) = P down Equa- -22- Pdown f2 (1 + SXCOR f with = 2.3544 ) standard convention any range is If reception at the Earth. P2 from P refers i) measurement with to down (ti ik the S-band ti, then, time-tag signal influence (which will t 2 SXdelay (ti) SXCOR(t of the time is measurement on the ground at time t + Di) If associated tag Pdown (t) denotes the plasma delay time at time the that Doppler measurement or (2.28) SXran b The contribution 1 -- - the then 2 Di) ( -b 2 22 correction for Equations from (2.29) ) a range 2.26 and 2.28 SXCOR(ti) - f 2 (P up (t i - (B. + C. i 1 + D)) 1 (2.30) + an Given estimate, P f+ 2down estimate SX(t), of (t. - D.)) 1 the 1 time,& ~, thin-screen and an of the SXdelay then the estimate of SXCOR, SXCOR is SXCOR(ti) 2 = k k2 A2 2(SX(t i) - b for an S-band delay correction. for b and k, A A + b SX(t i -b&s )) (2.31a) Using the numerical values we get A A SXCOR(t ) = 1.080357 (SX(t ) + 1.179337 SX(t.-ts)) (2.31b) -23- the approximations used in Equation It might be wondered if 2.18 are calculated I compared t justified. by Equation 2.32 with delays determined from an accurate ephemeris. comparison is shown in I. Table The maximum timing (a large value), error With a plasma near superior conjunction is about 30 seconds. delay rate of 1 psec/hour This this timing error would cause a 9 nsec error in the plasma delay correction. TABLE I t (seconds) s Approximation Date MM/DD/YY Accurate Julian Day Error 11/04/76 11/14/76 2443087 2443102 1541 1507 34 1522 1501 21 11/27/76 2443110 1511 1497 14 12/09/76 2443122 1496 1488 8 From an ephemeris, the spatial separation between P 1 and P2 ' can be calculated explicitly. ally take P2' account of the It is planned to eventu- spatial separation between P 1 and by multiplying the matched downlink delay by r(P 1) 2.4 (r(P2')) where r(Pi) is the Sun-P D. i distance. The Measurement Apparatus Range Deep Space and Doppler Tracking are measured Network (DSN). at the stations of the The DSN maintains three -24- tracking station complexes, spaced so that any interplanetary spacecraft is visible from at least one of them at any time. One station at each complex is responsible for most of the DSN 14 at Goldstone, Califor- dual-frequency radio tracking: nia, DSN 43 near Canberra, Australia, and DSN 63 near Madrid, Spain. Each of these mounted, paraboloidal is fully a steerable, The in diameter. 64 meters antenna azimuthally 26-meter diameter antenna at DSN 12, on Goldstone, California, has been used for SX measurements since mid-1978. Since about 1970, range delay at the DSN has been measured by ranging (reference technique wave or sine wave transmission, and correlated square wave to which machines 2-5). estimate period. modulates transmitted the To a and ranging, the the received delay a carrier range delay modulo resolve ranging sequential In sequential sequence the use square before signals the sine ambiguity, are or the period of the modulation sequence is doubled and the correlation repeated, yielding the range modulo the longer period. This process is repeated until the codelength is larger than the a priori range uncertainty. Each sequence with a given period is said to determine a ranging code component. The synthesizer frequency used, fs, (, 22 MHz) is multi- plied by 96 to yield the transmitted carrier frequency, fc, of 2.1 GHz. For the Planetary Ranging Assembly (PRA), which uses square wave modulation only, the transmitter range coder output has a period of -25- 64-2 tn= 3f n 2 n+ll f= (2.32) where n, a positive integer, is called the order of the code Note that the transmitter code period is a func- component. The MU-2 machine, which can tion of the carrier frequency. use sine wave with operate essentially n It can be seen from a graph of the correlator equal to zero. outputs can modulation, 3), (Figure the that must correlation be done in quadrature to completely resolve the range code phase. Two types of ranging machines, the PRA (also called the PLanetary OPerational ranging machine, or PLOP) are now used to make dual-frequency The ,MU-2 measurements. (only one exists) and is an and the MU-2, range. and Doppler experimental shift ranging machine is the only machine capable of making unambiguous SX range measurements. The PRA machine is the standard DSN ranging machine, and is used at every DSN staFrom the start of the mission tion. 1977, the MU-2 machine was at DSN 14. until mid 1979, (1975) until April 15, From November 15, 1978 this machine was at DSN 43. While the MU-2 was at DSN 14 it was used exclusively with square wave modulation, but after it was moved to DSN 43 it was used exclu(A. Zygielbaum, sively with sine wave modulation. private communication). The bandwidth of the spacecraft transponder is about 3.5 MHz, which, lengths together (Equation 2.34), with the choice limits the of smallest code component code component -26- to a period and (or length) tions (references shift is 1, which using the shows a a The Phase Locked of diagram the (PLL) Loop the in a phase as frequency carrier phase of signal transmitted block 5). cycles counts an about 10 ns or less under good condi- 4, by measured has and level, sub-nsec measured The modulation. wave sine experimental scatter of ceiver for square wave modulation, is discretized at the delay range for that half to of 2 psec S-band Doppler signal received Figure reference. ranging re- the in 2 system (after range compo- reference 3). The PRA machine can (S-band or one nents of other frequency code be ponent. Thus be resolving the PRA respect SX the of to the be by other ambiguity are The X-band code) code com- (2 usec) shortest S-X the measure code. range means. modulo range This SX range (see Equation 2.34). resolved must The machine 2 usec ambiguity chapter, the all received (usually chosen PRA the is about which X-band) correlated with can only ti, correlate The discussed details the in of next in Section C.3. spacecraft to the will, ground in general, A change station. be in in motion with during range a range measurement causes a change in the phase of the returning modulation. frequency shift To is increase used to integration times, the produce a corresponding the phase of the range correlation template. carrier change in -27- E. Terrestrial Propagation Effects The Earth's the to contributes ionosphere, of for the each plasma of delay tracking every 60 this model stant and the resulting use by is ionosphere and of data the at a rotation from thin to shell be a constant each in sampled mapped are the Satellite continuously, assumed density improve Technology integrated available are to used Faraday is direction the be Applications an orbit thickness spherical to the model. In shell height of cona above Earth. be the Let E(t) an to seconds spacecraft's The correction. geosynchronous can stations, also rotation Faraday which delay, ionospheric DSN station plasma delay. measured the measurements a dispersive medium, being of estimate the spacecraft elevation angle and D delay, ionosphere zenith craft ionospheric delay estimate (t) be the space- then is D (t) ION E(t) =2 2 (rl/r (1 - where r1 is the of radius the 2 r2 - r1 is equal station, a tion of fifth time entire pass. order rather E, is (2.33) and is r2 mean the then (r 1 /r 2 )2 is 0.89866, For each pass at each ground polynomial, than 1/2 E) If r 1 is assumed to be 6378 to 350 km, which is the value used at JPL. cos Earth, radius of the ionospheric shell. km and ) considered fitted to ION These polynomials, prepared under to be E(t) a func- over the the direction -28- of H. Dr. Royden at JPL, will be available made to MIT at some time in the future. Each DSN station uses one several in measured was data could concluded delays by Dr. from observing Thus the Faraday times) be and (the from assumed zenith delay delay, about the bias to delays. that, satellites sufficient to for only use that last With interrupted change in due to the delay satellites geosynchronous sometimes for equipment failures. contain measurements delay ionosphere comparison there is estimates on be the one cannot measurements infefinitely, rotation 2 nsec It mapped a many consistant An attempt was made to estimate the bias unknown bias. consistency ATS addi- between comparison rotation session. only are could be This mapped are (H. Royden, private communication). Faraday sessions and from the plasma delay, only the measure the total months, improve is it satellites simultaneously. different polynomials, Unfortunately, observing to the local ionospheric delay the Royden satellite per station an and used several cases, directions be obtained production over some from one station, visible tional In delay. ionospheric satellite to measure at same a still S-band. at be positive ionosonde data, with the must bias order There each of in the is no station, the but at from all it must ionospheric nighttime zenith reason to expect which means that there probably would be systematic errors on the order of 2-5 nsec between polynomials from different stations. Note that -29- the zenith delay causes nonconstant error in the a bias in delay mapped to the spacecraft. general, the be Thus the bias will not, in the when comparing same delay ionospheric estimate from one station at different times. The ionosphere data can be used to improve the interpretation Lander of delay First, measurements. and Lander Orbiter tracking data are often taken at different stations, and these stations do not share the same ionospheric contribution. polynomials The to contribution the can plasma be used from the thus delay replace to the station Orbiter ionosphere with that from the Lander station ionosphere. Second the thin-screen model does not properly model the The uplink ionospheric delay contribution. is contribution before the made at the matched downlink polynomials can be used to uplink the replace some time contribution. The time, send ionospheric ionospheric receive iono- time spheric contribution in the matched observable by the appropriate send time ionospheric delay. These corrections will be most important far from superior conjunction, when the solar plasma delay is small, and the ionosphere contributes a large fraction of the plasma delay. The zenith ionosphere delay can be modeled fied diurnal sine wave with a peak S-band by a recti- zenith delay of about 10 nsec at local noon, and a fairly constant night-time value of about 2 nsec. At an elevation angle of 100, the S-band ionospheric delay is therefore approximately 30 nsec. -30- Only differences in the ionosphere between the stations cause however, errors, and the delay ionospheric could be the dominant cause of error only for lander observations at low elevation angles. The terrestrial atmosphere does not contribute to the SX neutral atmosphere delay is indepen- since the observables, The atmospheric delay must therefore be dent of frequency. The atmopshere can also be treated estimated by other means. by the slab model in 2.33) which r 2, the slab This im- the radius of the earth. to rl, is equal radius, (Equation mediately gives a cosecant law mapping between the spacecraft atmopsheric delay and the typically about 7 averages from monthly This station. ground nsec at of zenith delay. is frequencies, radio the zenith delay, The estimated pressure and humidity correction is calculated in at the PEP and stored in CAL(1) (see Appendix III). Data from both the Orbiters and the Landers are necesfor sary the general Landers are are the delay and the used to determine the motion of Mars. To use system dynamics. used to estimate the plasma Orbiter plasma delay measurements to estimate delays many In this experiment, studies of solar Orbiters for and experiment relativity Lander plasma requires an extrapolation across space and an inter- polation in time. The temporal separation between Lander and measurements is not a negligible source of error. Orbiter SX A scheme -31- for temporal extrapolations tion of the statistical application of range measurements. processing and required devised after an investiga- properties of the plasma delay. investigation is described the was plasma The delay Chapter 5 discusses interpolations to Lander next chapter discusses the computer before Doppler measurements. in Chapter 4. This use can be made of the SX delay -32- Chapter III Computer Data Processing A. Introduction Extensive computer processing is required before use can be made of the data collected at the tracking stations of the Deep Space Network Chapter II. editing, and (DSN). Computer The observables wre described in processing reformatting at JPL involves merging, the raw data The tapes. results are then copied and mailed to MIT, where the computer processing The completed. is processing done at MIT includes applying calibrations, merging, editing and reformatting the data. This Orbiter chapter radio is concerned tracking data. with First, the processing I discuss of the the trans- ferral of data from JPL to MIT and the problems of obtaining a complete data set. Second, I discuss the nature and type of the various range calibrations and the method of removing PRA SX range ambiguities. Finally, I describe the algorithms used in data editing. B. Data Collection Range and Doppler data were recorded Tracking Tapes (PTT), which also contain at JPL on Project engineering from both the spacecraft and the ground station. At JPL the PTT are read and processed by the JPL Orbit Data Editor program. data (ODE) The resulting ODE tapes, called ODFILES at JPL, are edited and used for orbit determination at JPL. Originally, -33- SX data I used of ODE the to modified that more 20% about tapes. Therefore, process the they would contain MIT good (the PTT, not data the missing data. turn contained missing data The but resolved, satisfactorily been This did not two. other available on the problem has never hope the JPL printout) and ODE tapes, the the in MIT the three data sets available at Each of out to be the case. were to sent tapes PTT programs computer our MIT copies the from available than- were data tapes mailed ODE the used at JPL contained edited SX data print-out of A to MIT. copies of from obtained at least part of the problem seemed to be the use of different editing The algorithms. data PTTs the ODE data not on the be can As sources. three all from consisted experiment this in used data in seen Table III, of the SX superseded by about 20% contributed of data used. April, In the tion must Data types metry, be tion are (MDA). blocked Each a from telemetry from the and range and is of station. formatted data block the sent by to variety ground data, Doppler variety ground A ATDF's. from transmitted include itself the for flow stream the data Figure (ATDF). File Data Tracking Archive was format data PTT the 1978, the the describes informato Metric JPL. tele- spacecraft at star of station sources, Data 4 and engineering the ground sta- Data Assembler switch ler, which creates packets from the data.blocks. control- -34- is switching Packet in packet switching there intermittently, and pre-allocate data rates. This overhead, since cost some of (data information data is no need to flexibility is obtained at the packet each time, source, send sources many since used needed etc.) contain must to the reconstruct the data stream at JPL. The packets are and JPL, are the and for indicates, ODFILE's program ODFILE's The are at used of production storage from JPL are the of The for much in ephemerides, orbit name data. The PTT) (ODE). Editor determination same the two the as from the the NDP, about for tracking (or Data Orbit kept designed, are ATDF the the at The output from (IDR) tapes ATDF called JPL are Record archival created (NDL). Logs (NDP) is a program which separates the Data JPL. at Data Network reconstructs the data. Intermediate weeks the onto written Network Data Processor packets over satellite data links to transmitted manner by a The and the as the OBSLIB data type is used at MIT. The processing up to tion of stream, and human times on The ATDF are editing. is designed intervention. the the NDL resembles a data communica- NDL in case to transmit data with a minimum NDP The of is run several reading errors. available data, without sometimes suspected supposed to contain all tape The ATDF tapes have never been checked appropriate ODFILE data data problem is solved. tapes to ensure that against the this missing -35- tapes The data from JPL in a onto use for reformatted OBServation at Upon receipt of LIBrary chosen to store Orbiter described in Appendix in time order. MIT, was data used to later modified as a this In processing lar, the to range Orbiter range is SX CALibration way The the the from program ATDF format used at MIT to read by written (SXCAL), tapes data on OBSLIB it was to possible into use the ODFILES. Lander of Orbiter Lander data format which can be and X-band dual-frequency data from the similar is (then with the Radio Science Group at S-band format. very format SXFILE, and X-band ODEPEP, program, program, Robert Goldstein convert and MacNeil to process by Dr. by Dr. The The SXFILE containing the data. into the OBSLIB modified SXFILE tapes. is called the S-band remove I the data placed SX data are placed upon the SXFILE II. PEP. to SX data Another well. convert ODE tapes MIT), (OBSLIB) is data A program, SXDIFF, written by or Paul MacNeil PTT and to write an was tracking Lander MIT. convenient format information within is ODFILE tapes, the the PTT, ATDF or the not use by the MIT data processing programs. for of are range S-band data calibrations calibrations, and range is processing. have the the conceptually In same particuformat as computer subroutines that were written to apply the Orbiter calibrations have been adopted to S-band data Dr. apply for Goldstein. the programs Lander range calibrations. The Lander this experiment were calibrated and edited by He put them onto OBSLIB tapes with the use ODEPEP and SXCAL. of -36- C. Data Editing and Calibration in processing I wrote a number of programs to assist SX the data, and (MFE) program Edit the important of most SX LISTSX data: pair a and of PLOTSX. The range calibrations and as SX and LISTSX from data tapes. The the data, inspect to used are plot as well data, SXFILE and actually MFE program input Fix and list deletes bad several programs PLOTSX Merge to programs applies the merging the which were the especially in connection with data editing. There are These data. four of no require calibrations 2 the and PRA demod) (called the of removal data do but require SX the BIAS ambiguity Doppler The editing. the RANge range PRA Psec SX range raw use, in order of application, resolution (SXBIAS), are, four operations (RANCAL) CALibration of in the processing stages data data edit- in their order extensive ing. These' operations although of use, In particular, it this is will be order often discussed below is not always strictly observed. necessary to the iterate various steps, especially the data editing. Range Calibrations 1. Range geometry measurements delays that contain must be and instrumental estimated and removed. RANCAL provide an estimate of this excess range. is the Craft sum (S/C), of three and the the station- DSS, components, called Z-Correction (Z-Corr) delays. The Each RANCAL the Space- The DSS -37- ranging at hardware by the the and DSS The is delay delay the delays three Delay antenna Device The surface. by a device used on only are made mostly at SX measurements the inside after (for the called mounted transponder a on 26 meter the test translator, I will discuss only the diameter antennas. since is is ZDD delay the test translator or which (ZDD), the of is measured before and It stations) diameter meter 64 the all that the in spacecraft, of estimate Note ranging pass by a device called Zero the on board an estimate an ground station electronics. each delay the to round-trip delays. refer the is delay geometry. site of estimates station Z-Corr The respectively. caused are delays spacecraft and 64 meter three the stations. The the of end test translator which couples the (Figure end appropriate S- and the By shows intermediate are scatter due typically contain being definite about outliers, typically many tens of it can with the or nsec. bad mixer receiver front with signal noise local The S- and The DSS the thus and standard DSS the both the independently. measurement 5 a is simulate and ratios, to front translator transmitted measured occasional equipment changes, tion test frequency, turn-around X-band a the mixing the at transponder transmitter klystron and the calibrations X-band delay 6). a The system. ranging as acts due to devia- measurements measurements, which standard deviations away from the are -38- to delete any point greater chose away the from to edit the It was thus necessary local mean. We 5 standard deviations than Figure mean. local RANCAL's. shows 5 a of plot RANCAL's with deletions. The spacecraft delay is an estimate of the from telemetered spacecraft turnaround time and is calculated Before launch, temperature and signal strength measurements. the delay spacecraft was spacecraft temperatures construct a was delay ments by a table During delay. calculated these table calibration internal versus at measured and from were measurements of spacecraft mission, the internal variety of a used to temperature the spacecraft measure- temperature telemetered lookup and transponder interpolation. range delay measured The Z-Corr calibration converts the by the electronics to the delay that would have been measured if ranging the location had machine at been the station reference The Z-Corr includes the propagation (see figure 6). delay from the antenna aperture plane to the test translator, as as well r , the delay between antenna the aperture plane and the site reference location, both of which are calculated from station geometry. the waveguide which is itself calculated, which station mission. between The Z-Corr also includes the delay in the and is measured. feedhorn the The delay and the test translator in the test translator Z-Corr delay estimate is constant at the 0.1 nsec level (The Z-Corr for the the path length to the ZDD.) 26 meter for any throughout the stations must include -39- in this used The RANCAL's the direction of Tom Komarec at bration pass each for called the the cali- RANCAL cards, with These cards band. under prepared The value of JPL. is punched onto cards, one card per are experiment are mailed to MIT where they are edited and stored on disk. Dr. Goldstein package, USeR and others CALibrations at MIT (USRCAL) , wrote which a subroutine finds the RANCAL for a particular pass, performs data conversions, and returns the calibration frequency. in seconds of delay scaled to the appropriate is USRCAL data, dual-frequency called and the once SX for each is calibration the differenced S-band and X-band calibrations. exists for some From used. spacecraft, set the ground the selects another RANCAL of RANCAL's pass, all and station RANCAL from the band, pass value for the is If no RANCAL and selected appropriate the in the from formed algorithm selection closest for band time to the target pass. 2. The SX Bias Immediately after within a few million was negative. team at JPL average estimate km of It was that systematic errors ured during launch, this spacecraft Earth, the measured the concluded by the spacecraft negative bias was in the ranging system. caused The SX was the bias from in the all SX SX range range is SX delay by unmodeled range measthis measurements. called still navigation the early cruise phase was averaged, and was subtracted of when the the The SXBIAS. -40- The estimated added to still unknown. SXBIAS is tributions of ignored inspire this followed in currently is estimating the The plasma delay con- ionosphere averaging, (these values are SXBIAS the of confidence. terrestrial the in cause procedure The SXBIAS does not were The SX delay). Table II given in so and plasmasphere that there is an tional bias on the order of -10 nsec of SX range still addiin the SX data. TABLE II STATION SXBIAS nsec DSN 14 before JD 2443300 20 DSN 14 after JD 2443300 26 DSN 43 26 DSN 63 26 3. The PRA Demod The PRA ranging machine can measure the SX range only to within modulo tl, ponent the period of the shortest PRA range com- (about 2 usec - see Chapter II.D and Equation 2.29). An appropriate integer multiple of tl, determined from nearby MU-2 data, must be added to each called "demodding" the PRA DATA. PRA SX range, a process Demodding is reliable if the total SX delay can be estimated with an error much less than tl from nearby MU-2 SX measurements. -41- The Further time. of difficulty conjunction, is demodding conjunction, the 20 than to rapid Within delay plasma before or is delay unnecessary. depends demod days 30 SX total the PRA 2 upon strongly superior after and less than 1 isec to days of superior 3 fluctuations make demodding unreliable. PRA cards, demod and input the demod. until values to are determined onto MFE program, which actually applies the PRA SX range data are adjusted by multiples of slope and they match nearby MU-2 SX in both the demod typed manually, value is not clear from the data, tl If level. PRA datum is the deleted. Lander delay superior residuals have a scatter the If conjunction. pends on PRA SX range data, used in a which to test delay provides an demod values used calibration plasma Lander range PRA demod. is in lander nsec near de- correction residual An immediately independent check upon can be error of t detectable, the validity of range calibrations PRA (of course, we in just those data). are especially interested D. the of the the validity Lander Lander 100 of Data Editing Approximately 2 104 SX delay measurements and 5 105 SX Doppler shift measurements are available for this experiment. The sheer amount of data to be processed made data editing an important part of this experiment. edited semi-automatically. The SX delay data The SX Doppler data would were require -42- automatic or subset of the interactive data were data to editing be used. if more The than Doppler a small cycle slip problem considerably complicates automatic Doppler editing. SX first delay editing program data editing pass, was used immediately to done was delete after all iteratively. the SX PRA demod, delay points In the the MFE in a data segment for which SXMIN eSXrang e SXMAX range did not data hold, in order segment SXMIN and cessed near typically delete covered SXMAX were chosen, data, results to for to lie the a segment data. Each of data. look at the unpro- month's from a first just outside data obviously bad of the worth range of processed. being reasonable For example, superior conjunction, SXMIN = -100 nsec and SXMAX = +100 Psec were used. After the first pass of data editing, the data were reviewed and edited manually, with the help of the LISTSX and PLOTSX programs. to be Delete cards deleted) were then (which specify a prepared for input to data editing span of data the MFE pro- gram, which actually deleted the data. Table 4070 III SX delay gives a of summary measurements (or 19%) delay measurements were deleted. (or 16%) came from the delete rest were manual deletions. the on of the total process. 21924 Of the 4070 deletions, SXRMIN and SXRMAX and SX 636 the -43- Table III SX Delay Data Sources JD 2442950 - 2443434 01n PTT ODE ODE b ut no t Total Total Total Deletions Good Points Data Data PTT Data VO1 9083 9905 1199 10282 VO 2 9391 10560 2251 11642 4-- Total 18474 20465 3450 21924 4070 Spacecraft 17854 JPL print227 out data 18081 Grand Total Two categories of bad SX delay data can be guished, isolated bad data and groups of bad data. distinWithin a bad data group, the SX delay estimate are typically scattered between SXMIN and SXMAX, with no apparent correlation between adjacent measurements. All observations within a bad data group were deleted, not just the points away from the local mean (Figure 8). Bad data groups typically, but not always, occur. at the beginning or end of a pass. It is known that the ranging system collected "data" at times when the spacecraft was not above the local horizon. At least some of the -44- bad data groups can be attributed to this cause. Other bad data groups are caused by equipment problems at the tracking station (A. Zygielbaum, personal communication). An isolated bad datum (Figure 9) is one bad point in the The cause of such bad data is midst of a good data sequence. unclear. For isolated bad data, the local standard deviation was estimated and the if it was more than 5 point deleted standard deviations from the local mean. Isolated bad data are hard to catch by eye, and this stage of the data editing might be possible to automate It was iterated several times. the deletion of isolated bad data in the future. The SX Doppler data this for this experiment. included in the craft, which of improve a multiple of phase counting. multiplication up to Lander 1 cycle of noise is Uplink phase X-band at the signal-to-noise reduces to The Doppler data suffer from cycle errors are phase phase in the Doppler used Data editing problems have prevented plasma delay estimates. slips, which be could of the the space- signal re- At times of rapid spacecraft accelera- ceived on the ground. tion (such as near periapse) the X-band Doppler phase changes too fast for the Doppler cycle counter at the ground station to maintain lock. For reliably SX Doppler detect conjunction, cycle slips. the editing cycle slips scatter in it in is necessary to the data. be Near able to superior the Doppler data masks possible Figures 10 and 11 illustrate the problem. Two -45- integrated overlapping Doppler sequences overlapping min hr 40 data, which diverge due to cycle slips at about 23 spacecraft are compared. Figure 10 shows from different the It would be impossible to reliably detect these cycle slips without overlapping Doppler data. Figure 11 shows the over- lapping data after the removal of the cycle slips. After processing, the Orbiter SX data must be applied to Lander range data. Chapter IV describes the statistical nature of the plasma delay, and Chapter V the application of the plasma data to the Lander observable. -46- Chapter IV The Statistical Nature of the Plasma Delay A. Introduction As was shown in Chapter II, the solar plasma contributes to the measurements of the delay and Doppler shift made with the Viking spacecraft. To ments, is to it necessary make full estimate use and of these remove measure- the plasma delay contribution to the measured delay to the Lander. The plasma density wild fluctuations ties to model in the solar corona is subject to (reference 8) and is beyond our capabili- adequately from first principles. In such circumstances it is natural to consider statistical models of the plasma delay. It was decided separation between measurements, and to ignore the spatial to treat the SX plasma measurements as a function of time only. In this chapter, I will discuss a statistical study of the SX plasma measurements. I will define the autocorrela- tion of a random sequence, and will describe the results of autocorrelations of the SX data. Given the available data, I conclude that the appropriate model of the plasma delay is a then discuss the implications of that conclusion for plasma delay estimation. In the next chapter random walk. I will I will discuss results obtained from applying these conclusions to the estimation of plasma delay corrections from the Lander S-band delay measurements. -47- B. The Plasma Autocorrelation A random process, xt, called weakly is (or wide sense) stationary (reference 19, pp 55-56) if it has a time invariant probability density function, and if - Expectation (xtxt+r) = <xtxt+ > c(T) exists and is a function of r only. (4.1) In this case, the func- tion c(r) is called the autocovariance function of xt , and p( is function the autocorrelation Let process, x(t) be a starting at time t . + At, ... , t zero mean, (ACF) (4.2) of x(t). stationary, weakly At for random N observations (Thus, the observation times are t , to An unbiased (N-1)At.) + c(O) intervals at equal sampled ) minimum variance) estimator for c(nAt), (but not necessarily e(nAt) is: N-n c N-n t +iot t +(i+n)At i=0 This estimator processes; is indeed, not generally the unbiased autocovariance for nonstationary as defined above may not even exist for nonstationary random processes. The epochs of the plasma delay and delay-rate measure- ments are not in general evenly spaced and thus Equation 4.3 cannot be used directly. The standard modification of Equa- tion 4.3 for unevenly spaced data is -48- ); n > 0 S(nAt) n > N(n) all pairs x t * x t * w(n,t 22-t ); with tlt 1 2 2 (4.4a) (4 with (4.4b) 2 -tl) w(nt N(n) = all pairs with t -t and In = 2 -tl) w(n,t Equations really mean 0 t 2 - tl 1 (n-1/2)At 0 (n + 1/2)At (4.4a) all and pairs which tt tstart -Ct t 1 < (n - 1/2)At (4.4b), t is due program, to Parzen, see "all pairs with tfinish times at wheret tfinish reference I 24. the Plasma Autocorrelation wrote Program tl t2 t tl, 2 for tis start is to Equation 4.3 This extension called the integration time. (4.4c) t2 - t -2 1 observations of t2 - t2-t 1 < (n+1/2)At (PAP), a computer to imple- ment this algorithm. 1. The Autocorrelation of the SX Delay The plasma delay is weakly not stationary, as neither its mean nor its variance are constant with respect to time. (See Figure superior 12 for SX plasma delay measurements taken near conjunction.) The long-term non-stationary over trend time in spans the plasma much longer delay makes it than a day. This non-stationarity must be removed before PAP can give meaningful autocorrelation estimates for lags on the -49- a of order of tenth in nonstationarity a We primary The more. SX delay the raypath geometry. or day measurements the of changing that this nonstationarity hypothesized could be removed by normalizing is cause the integrated plasma density with a geometrical correction factor, obtained by integrating a static plasma density model A simple, commonly over each ray path. used model of the plasma density is (reference 20): p(r) r is the radius, R0 and r0 Here Numerous trons The cm distance have best the from (4.5) in electrons cm Sun, in units of the fit shown the that coronal P 2.4 and P P 7.5 plasma density at raypath given the over density integrated n solar P215 R O. is generally taken to be 1 A.U. or studies -3 - 3 - = elec5RO 0 r Chapter in II is H(z) ar = (r)dx (4.6) Earth where It is function of nistic tive since useful a the to assume that the measured SX delay stationary random process, function function impact parameter. raypath t is the of such SX' variance as H(t). and of H(Z), the SX If SX' the not be a a determi- and SX delay was could delay SX', is an addi- stationary, fluctuations increases -50- We hypothesized that the SX delay near superior conjunction. was a multiplicative function of SX' and H(t), or that SX' = SX/H(a) (4.7) is the assumed stationary process. From an analysis of the Viking Orbiter SX delay measurements I found that (A) SX varied from 10 nsec to 50 usec over of observations, the region or by over 3 orders of magnitude. The local variance was roughly proportional to the mean. (B) SX' varied the with from 0.2 to local variances SX' with time. 1.6 over varied being over the same period, roughly less than constant 1 order of magnitude. (C) There is still conjunction, a but long-term the closer to stationarity data. As an SX' some Figure 13 is than before do near data normalized illustration, 2 days trend the come unnormalized superior conjunction. and Figure 14 is a plot of the corresponding SX' vs time. the average of much compare SX and I will a plot of SX vs time superior the SX ranges shown Let SX be in Figure 13, and let ASX be the peak-to-peak variation in Figure 13, then ASX/SX = 0.14. In an analogous way we -51- constructed The ASX' still trend local of account the With obtained. simply in as lag, a trend (or white) of (Figure to inade- the take into Equation partly removed, show that ACF due in the (see ACF decreases the noise as 0.05. SX delay could random walk the and to to not r, = be might does be detrended 15). Figure linearly with the first differ- looks like uncorre- SX delay measurements ences of 16). The Autocorrelation of the SX Doppler Continuity implies that the become correlated as t]i+ 1 ured more often than the higher or terms appear random walk lated SX' long-term behaves increasing 2. the ASX'/SX' in which order They SX range a model, that autocorrelations of the long arc For visible higher 2.21. found fluctuations plasma quacy and frequency plasma measurements should t must first difference of xt . The Doppler shift range, and thus provides a The fluctuations. also be more accurate plot of SX delay SX is measprobe shift Doppler than the of SX delay measurements. Figure 17 the delay and and that the With respond to a and integrated Doppler Doppler sampling this in mind, ACF's of the agree rate ACF's range of 20 on the 20 times is the Doppler integrated 16 days after superior conjunction. vs time some range. is Doppler differences) that Notice nsec that level of (which were the cor- obtained -52- (Figure 18). Doppler The average sequences data and used the in root mean square of Figure 18 shown are the in SX Table IV, the average of each SX Doppler data sequence is less than the corresponding rms in all thesis that the SX Doppler cases, which supports the hypo- are a zero mean random process. TABLE IV Study Statistics of SX Doppler Data Used in Autocorrelation t , Jul an Date Spacecraft t , Jul an Date RMS of SX Doppler Data, Hz Average of Doppler Data, Hz Number of Observations VO2 2443095.353 2443095.595 349 -0.000491 0.1397 VO2 2443101.080 2443101.347 383 -0.2927 0.3516 VO2 2443117.685 2443117.800 166 0.0209 0.10587 V01 2443124.839 2443124.9577 177 -0.00135 0.19388 VOl 2443160.885 2443161.317 606 0.00281 0.00937 Since the epochs estimate provided Equation 4.3, known. mates In of the by to where the included values N is (This 4.4 ACF 4.3 is known number estimate pp. 34-36). with the plots of Lines the a white to be of of are noise a at is due the Doppler esti- process equal approximately observations 5a ACF. to in better of a, ACF that much deviation, of the to identical is standard the 14, equispaced, properties of the are data statistical the Equation estimate. reference whose this Equation particular, provided 1//N, by of included see Bartlett, level have Several in of been the -53- sample ACF's at correlations cant ACF's, an with 19b, be best An AR(n) and d, c, modeled a as process suggest of nature the with decay the that order be modeled difference equation of the show signifisample increasing in possible damped oscillations low can The exponential e, and 19a and and c) 19a level. this apparently lag in Figures Figures Figures (in particular Doppler data SX by an nth process. (AR) Autoregressive order may stochastic form n = x zero is 8 where III). chapter i-k mean white noise be expanded x, can sequence. , xt. the If Bt B the zero mean, in terms is a of 14, a linear x mean random multiplied by zero are 4.8 if the expected value is then taken and normalized equations resulting (reference sequence Equation of sides 1,Aand (4.8) i . are both for m = + B t k=l combination of past Since K i by dividing by c(O), the so-called Yule Walker or Normal Equations result k lkm-k kilkPm-k p= m These equations Estimates for Equations the ignore Pi. best solved be the model estimates of the necessarily can m = to parameters, The $.i estimate possible Yule Walker equations are errors 4 give $j obtained k in ^, in terms of can be obtained in this manner fi, since of (4.9) 1, n but the solutions Yule are pm. from not Walker to the frequently used as a guide in model -54- identification. value with From the rapid decay of the SX Doppler ACF only lag, increasing order low AR are models suitable. For an AR(1) process, Equation 4.8 reduces to xt = 1xt + ti (4.10a) 0 < 1 for stationarity the ACF decays exponentially 1€ Pk (4.10b) kI and the Yule Walker Equations become the trivial (4.10c) P1 = 01 An AR(1) process can be interpreted samples of as a first order Markov process dx dt - x(t) + 8(t) y (4.11) Under this interpretation, Equation 4.10 becomes -At/y x t= (4.12) + Bt. xt and y can be identified with the correlation or Since At is known, y can be estimated by an AR(1) process. y = -At/ln$ In testing a suspected a white noise data results is shown AR(1) 1' of in 1 a1 , the Table that, the the (4.13) 1 null hypothesis estimate of the is stan is just 1//N. analysis V. = -At/lnp process, Under process. dard deviation of The "l/e" time of At of the selected least two sequences SX Doppler have sta- -55- tistically significant values for The Correlation times *1. are all very short though, on the order of a minute or less. imply that The short correlation times of the SX Doppler and the -Doppler random to minutes. measurement Lander nearest SX practical modeled as versus 21b) delay adequately typically the time modeled measured of to shows that are further measurement. every the number the nearest 82.4% than The of a four Lander SX of all 58 minutes delay (usable) invervals of longer than a few minutes. As integral of a the SX white Doppler from are therefore, SX as data matter, the is be histogram measurements uncorrelated over a The (Figure delay can rate SX delay observations delay the The walk. eight delay delay noise data process, can or as be a random walk: Table V for SX Doppler Estimates Parameter Data Fract N 3095.353 3101.080 3117.685 3124.837 3160.885 349 383 166 171 606 The integrated data, $ Time JD-2440000 estimation Doppler however, multiples of of to contain one cycle $ 1 1 min. .054 .051 .078 .076 .041 .599 .286 .199 .210 .310 SX delays extend cycle of could range slips Doppler be improved sequences. (errors that phase 1.95 0.80 0.62 0.64 0.85 11.2 5.60 2.56 2.74 7.63 at the The are by using Doppler integral ground sta- -56- tion), especially when the Doppler phase counter stressed (i.e., when the Doppler rate is high). about 106 SX Doppler observations, Doppler is highly As there are validation is a critical problem, one that I have chosen not to tackle so far (see Chapter III.D). C. Estimation and Smoothing of a Random Walk Process A random walk can be modeled by a stochastic differen- tial equation: dx = w(t) (4.14a) where w(t) is a white noise or Wiener process. correlated random back" term linear in X(t) tially has variable an An exponen"feed additional on the right hand side.The equiva- lent discrete version of Equation 4.10a is x + Bt. = xti 1 1-i (4.14b) 1 where the t i are assumed equally spaced and St. is a zero i mean white noise sequence, assumed to have a symmetric uni2 2 modal distribution and a variance of a2 and a variance of a . unimodal symmetrical distribution. In that case x The 1 first difference of a random walk is Vx t x t x = B ti-1 t A random walk can be written as a sum (4.11) -57- k-1 2 + = xt x (4.12) t =0 The Maximum Likelihood Estimate (MLE) of a random variable with a symmetric unimodal distribution is the conditional mean of the variable, which is also the Minimum Mean Square Error 157). xt, (MMSE) estimate (reference 19, p. 149, pp.156- Under these assumptions, the MLE of the present state, given knowledge of some past state, x t < > = <x Ixt i k- + i-k , > = x (4.13) i-, i-k is i-k this is called the forward state estimate. The mean square error is 2 2 f = <(Xt x i- x 1 where Atf is (ti - tik) i-k 2 > = At fa = 2 f (4.14) and is called the forward interpola- tion time. In a similar fashion the MLE of xt. given knowledge of , some future state, x t is i+m <xt > = x 1 i+m (4.15) 1+m -58- This is called the backward state estimate and the expecta- tion of the mean square error is 2 where Atb is ti+ - m t i and 2 2 (4.16) = Atb ob is called the backward interpola- tion time. The optimal linear (MMSE) smoother is the weighted average of the two estimates of x. (reference 15, chapter 5): xxt Ix = t Using 1 Equations 1 1 + 1 -+2 2 f ob 4.14 X t Ixt i-k 1 and a a f 4.16, i+m 1 2 the (4.17) 2 b optimal linear smoother for a random walk process can be simplified to obtained pxt t. 1i where a p is Atb/(Atf + Atb). i-k (4.18) (l-p)xt + i+m The optimal random walk is thus just a linear linear smoother for interpolation between the nearest data before and after the smoothed point. Given based a model the upon corrections. cation of of the model, In plasma it plasma is possible the next chapter, delay behavior, to and an to plasma calculate I will discuss measurements estimator the the appli- Lander range -59- corrections, and the results of that application. smoother derived estimators. The linear in this chapter will be compared to other -60- Chapter V Lander A. Plasma Introduction In the previous smoother for required to dent the of the Chapter, SX delay calculate estimator was Lander possible used, delay The compare to I chapter, of part plasma I solutions optimal of the corrections and was it processing is indepen- to decided write a In this way, several estimators. estimators linear effects from their on residuals. first Lander Most the plasma this chapter, processing necessary to convert into derived data. Lander program which could realize it Corrections will the using In the Lander will describe the Orbiter SX range measurements corrections. report I the results delay second of data, part multiple and of this parameter the finally be- havior of the postfit Lander residuals is explored. B. Computer Processing of Plasma Corrections As observations, Library of applying a means wrote I Tapes) which a program, reads to the plasma corrections an UPOLT (UPdate OBServation Lander Observation LIBrary (OBSLIB) tape and writes a new OBSLIB tape with the needed corrections stored on converting SX corrections, lators as functions: UPOLT was designed it. range and they to were measurements serve as a found. to provide into Lander flexibility in plasma framework for better UPOLT performs the delay interpofollowing -61- A. Read B. For in an OBSLIB tape and an SXFILE. each Lander for C. Various subsets of One for the downlink and the SX data are selected: one two measurement, range uplink. the linear fits are made to each the se- uplink and of lected sets of SX data. D. The Lander downlink correction is correction, average of the the each sum of of the which is a weighted linear fits to the data. Each of these functions is described more fully below. A. Part of the buffer containing As Then the OBSLIB tape SX delay. ured is the buffer needed, observation and the meas- time of each the used to fill a SXFILE is read and input is scanned sequentially. filled with later SX information or the SXFILE tape can be rewound and the buffer refilled. last is option observations The desired. as the OBSLIB tape does not contain in time order. time necessarily needed This tag the for time any for The measured which the observation is not plasma delay estimate Lander delay is gation delay over both uplink and downlink only. delay Lander sum of the downlink. The the is propa- SX data are Thus, each Lander range correction is the sum of two SX interpolations, one corresponding to the uplink and one corresponding to the downlink part of the signal. using the the entire so-called plasma "thin-screen" effect occurs model, at the which point We are assumes where the that ray -62- path passes closest to the sun The 4). screen thin is (Chapter II C and probably a good reference near assumption superior conjunction, where the plasma density increases as a high power of the Far to the sun. distance from superior conjunction, it may not be a good assumption, but there the effect of timing (Only near su- is much decreased. errors perior conjunction does the plasma show a rapid time varia- tion.) Both the SX delay time tag and the Lander delay time tag are the time of reception or, more precisely, the time of the As discussed in Chapter II, start of the range correlation. time for a Lander delay calibration the downlink thin-screen is just the Lander delay reception time, and the uplink thin screen time is the reception time minus the round-trip propagation time between Mars and the thin screen point. be possible to given that in plasma timing particular, detected delay errors an after residuals. thin-screen delay approximation improve the Chapter on II. However, estimates the error order of run AP-43C, It would numerical are of remarkably many several causes tens hundred show experiments of insensitive to minutes. In seconds negligible in changes Atts, in the It is doubtful that much would be gained by using more accurate thin-screen times. B. For each of the two thin-screen times, the follow- ing SX data were selected. -63- 1. two The SX nearest (before and points after the thin-screen time) were found. All 2. SX points within certain a called AT, SPAN, of those two points were used. C. A from the linear was fit of made Each selected data. squares least fit a to three linear straight fit is to line data a the sets chosen (non-weighted) selected data. The data sets fit to are: selected All 1. SX epoch of the epoch of the the before data Lander observation. All 2. selected SX the after data Lander observation. 3. The D. tion of All the selected plasma SX data. delay is estimate by means separate estimates three obtained by combina- of a weighted average SX/w S= /w i=l In Equation 5.1, SX and A SX is the i is the value of interpolated (5.1) i=l SX the SX from the value. The weighting used was W2 c i (a 2 + alt i S2 cii. n -t 1) ts ith fit, i = 1,2) i = 3 scheme -64- where w i is the weight given to the ith fit, a i is the formal regression, from the linear error SX the time of the tn is datum that was nearest in time to tts and in the ith data set (see above), screen time and the c i for Equation 2.23). c is the thin observation (see In the standard parameter set, cl = 1, c 2 = (seconds of range) 1, c3 = 0.1, and a = 0.25 x 10 i, range Lander current the tts and a are constants. /day 2 3 The value of SPAN in the standard parameter set was 1 hour Our working hypothesis was that SPAN should be 12 minutes. set to the correlation time of the plasma measurements. One hour was the initial guess of this correlation time and SPAN equal to 0.05 day, or 1.2 hours, was the value chosen for the <1 minute would be turns out, As it standard parameter set. a more appropriate choice of AT. Given SX for each of the two thin times, screen the Lander plasma correction at S-band (SXCOR) is given by SXCOR(t) = A SX(t) b2 A + b2 SX(t-) ) k 2/ (k 2 2) (5.3) where k is the X-band turnaround ratio, and b is the S-band turnaround ratio C. (see Equation 2.31a). Lander Residuals and the Plasma Corrections Lander residuals provide an independent test of our model of the plasma correction, which is not usually available in time series analysis. Residuals are defined by -65- r(t) - = O(t) (5.4) C(t,?) where O(t) = the measured roundtrip range to the lander. C(t,2) = the computed value of the observation given the parameter set a. y = the best available estimate of the true parameter set vector y. = the time of reception of the lander range measure- t ment. The residuals Program (PEP), are computed which is by the Planetary described elsewhere Ephemerides (reference 18). PEP can calculate the theoretical observable and the partial derivatives of the observable to selected parameters from a parameterized model of motions of bodies in the solar system. PEP can use the resulting tials, value. to do a residuals, least-squares The partials, together with the par- solution once calculated, for the parameter can be stored on the OBSLIB tape with the observables, thus reducing the cost of each parameter solution. In the work reported in this chapter, PEP was used to fit a model of the motion of the earth and Mars to the Lander delay data, and the sensitivity residuals was investigated. earth and of the resulting postfit The model of the motion of the the location of the ground stations were obtained -66- the be the of model used parameters orbits and teh of at and six On seconds. PEP, ables with additional some postfit axis runs of rotation Jupiter period of Mars, of Landers on Mars unit in light- relativity the the parameter for. are residuals predicted from and the partial derivatives of to the parameters. respect than cheaper earth, the the The for each Mercury, of and the value of the astronomical adjustments meter the 26-parameter observations. coordinates of the the "RELDEL" was also solved In mass the a conditions initial direction of the to rotation phase of a given epoch, (three each) fit Mars, and earth the is system were: the Moon, the rotation all solar parameters) UPOLT the from distinguished (these parameters must solution parameter a standard In of 17. in reference rotation of Mars is described model The observations. radio and optical previous from reintegrating the This equations is the para- the observconsiderably of motion using the new parameter values. estimate An provided lander by the range are of the UPOLT plasma Lander The program. updated with the SXCOR, correction, residual new plasma and is computed corrections as is described in Appendix III. We are driven to consider the lander residuals by two facts: A. PEP uses a Weighted Least Squares y. The optimum weight for Estimator to find each observation is the -67- error associated with the observation. Appropriate weights can best be determined by a consideration of the residual scatter. B. independent check on the The residuals provide an validity of the Given plasma model. any plasma model, we can derive the sensitivity of the plasma estimate to external conditions. By actually finding the sensitivity of the residual scatter to the same external conditions, a check of the validity of the plasma model can be made. I wrote a program, the HistOGram (HOG) program, to 1. Group the residuals by external conditions. 2. Find the rms scatter within each group. 3. Plot the rms spread of the residuals vs the external conditions. Even a cursory glance at a plot of lander range residuals versus time shows an increase in the (Figure 19). near superior conjunction residual scatter Thus the sensitivity of the Lander delay residuals to the temporal separation between the measurement time and the time of superior conjunction was investigated. Let 6Tsc = IT - Tscl (5.6) T = time of range measurement, in Julian days T sc = time of superior conjunction, T sc = (244)3108. -68- HOG was run with all residuals for which nAt < 6Tsc were grouped together where are Figure shown error in 20, which was used The results in deriving 16T were chosen to make the root mean square weighted least residuals be squares slight changes sigmas, the solution - estimate parameter in They were used 1. in postfit parameter rms of the PEP, equal now on, we will consider the (residual/a). An attempt was made to find the residuals weighted possible that dependence upon the the time from the range 20. These (rms) of the in a new weighted and estimates. weighted were (5.8) to caused With only the new .9727 for one (Appendix 1i, Run AP-41R). From thought formal The formal errors used were This corresponds to the heavy black line on Figure sigmas a 200 days I sc < 200 days sc60-0.2"-T sc )nsec; 16T ssc (20 nsec; ) = (5.7) At is the bin size. (a) for each observation. o(6T (n + l)At < point), or lander upon upon the fit, fit (which is the external weighted residual the sensitivity of It conditions. residuals might range observation to the the number statistical a measure of of SX range error how well the associated a (i.e., nearest points SX data was show interpolation time of the estimate or the on weighted used with agree SX in the with -69- the curve these fit three Figures to them). possibilities, 21, 22, and no and and 23. Note relationship apparent the other was run to the results that parameters. the investigate each between shown the residuals rms of This the can in There residuals weighted be of are residuals were used. 1.0, as weighted now center about is HOG explained by the random walk model as follows. For the a random walk, is From that (Figure this data far measurement SX point contains associated with from noise has rms reality, each range of P20 it nsec improved plasma delay interpolaLander plasma the estimate of the In this case, is not delay an of Near superior conjunction this measure- 19 and 20). error. all conjunction, superior ment noise is probably swamped by the tion In extrapolation. random measurement noise measurement. seems for available information there the nearest Far by including more SX points. from superior conjunction, this condition might not be valid. The with autocorrelation increasing lag of (Figure SX sitive to the error, which tion time). observations interpolation is dominated One and might by terms lander .points very times slowly of the extrapolation will be insen(or time expect over Thus, 17). order of one tenth of a day, the decreases range the containing that was to if on the the statistical the gap order interpolabetween of days that some relationship could then be detected. easily test this by deliberately deleting SX data. SX several We could -70- D. Experimental Tests of Our Conclusions We have hypothesized that the random walk is a suitable Under the random walk hypo- model for the SX.plasma delays. thesis the total information content of a sequence of plasma measurement is contained in the last measurement. of the information available for in two the plasma values thin screen time. Thus, all interpolation is contained before and after immediately the From this, two further hypotheses can be derived: 1. Any linear plasma delay from extrapolated contribution will have estimator larger a mean includes that a plasma delay rates square error than the optimal linear smoother and the more it relies upon the extrapolated the slopes, larger will be the mean square error. 2. The plasma delay tive to the interpolation number of SX be should data points insensiused (see Equation 4.23). Now, if C 3 is set to o estimate will include slopes. the plasma delay a contribution from the extrapolated We hypothesized plasma interpolator. RMS 106), (in O -C - error practice, that this should be an unreliable We found that 1.2 using C 1 =1, C2 =1, C 3 =0.1 (AP-48F) 5.3 using C =1, C3 =10 C2 =1, with SPAN = 0.05 day. 6 (5.9) (AP-48E) -71- This large If increase in the rms error supports our hypothesis. SPAN is measurements, set to less then only two than interval between measurements, one before and one after, will be selected for use If, in addition, = c 10 5 C the time in the plasma delay estimate. = 10 6 , C3 = 1 or 1, C 3 = 106 is used, then the estimate lies line through the optimal linear 4.18. 2 data points. smoother for This C1 = 1, C 2 upon the straight corresponds a random walk derived = to the in Equation Using these parameter values, we found that 1.2 using SPAN C1 C C - RMS 1.0 using = 0.05 day = 1.0 = 1.0 = 0.i SPAN = C = C2 = 1 minute 1.0 (AP-48F) (5.10) (AP-48C) 1.8 C3 = 10 Run AP-48F is a standard parameter set run, which uses about 20 SX data points for each Lander plasma delay calibration, and for which the plasma delay estimate is a weighted sum of three 4.22 linear and to fits 4.24. Run the selected AP-48C SX data, implements the as in Equation optimal linear smoother of Chapter IV, which uses only two SX measurements per more plasma delay estimate. complicated smoother This does implemented as well as the much in the first run. Under the random walk hypothesis, the extra data used in the first run contains no extra information on the desired delay estimates, and can only degrade the plasma delay estimate. -72- Thus performance supports the the actual smoother random walk hypothesis. As an using smoother linear optimal As can does be not plasma test, additional seen the smoother in is thin screen model. using I, the compared screen Appendix model the use (Run AP-48C) model static of optimal the thin linear with (Run the AP-48G). screen model in the probably dominating any benefits from the the improve estimate thin I postfit residuals. The error -73- Chapter VI Conclusions I have tried to describe the processing In this thesis of plasma delay corrections for the Viking General Relativity Experiment, as well as a statistical study of dual-frequency measurements of plasma delays IV, I hypothesized the that and delay rates. plasma delay In Chapter can be,. for the purposes of Lander plasma corrections, adequately modeled by a random walk, and in Chapter V, I tested this hypothesis on actual Lander plasma corrections. Much work remains to be done with the Viking data. in engaged currently processing Viking Lander I am Orbiter and data covering the period from JD 2443420 to JD 2443804, which is not as straight forward as it might seem. of the data span exposes Lander motions used data is Lander required data. to inadequacies in PEP. break Another in the model Use of S-band Orbiter degeneracies four The extension hundred by exposed days of of the Doppler the Lander new data, including a second superior conjunction, remains to be processed. Other data, such as lunar laser ranging data, should be included in the solutions. possibly a program Doppler data. "bug", in There is currently a problem, the interpretation of Lander It is intended to commit a major effort in the near future to this problem. Some thought must also be given to plasma corrections for the Lander S-band Doppler data. -74- The plasma General Relativity corrections possible to near improve experiment superior Lander is most sensitive conjunction. plasma delay superior conjunction by hand validation of It to might corrections be near the appropriate SX Doppler data. Unfortunately, tional phenomenon do to discriminate is generally sion of needed before the tests the three an tracking would (the down in be of Optical tests have theories. in a preciwill magnitude direct increasing required the the amount are It be impact precision dim. At Lander MU-2 machines would need to be placed at station required RANCALs of in the Chapter would be general POINTS experiment). measurement of III). the program, With it supported soon. relativity Interferometers Considerable complexes. the American space experiment tests precision needed improvement of gravita- round trip dual-frequency tracking of a plasma corrections. tions the of orders chances of by level system required to dramatically improve the accuracy of the would be ment experimental solar gravitational an that several The of the various by test delay the very least all not have between theorists. time tests acknowledged these upon the of current in are likely earth improve- range calibra- current slow- is unlikely that such Future high prcision to orbit be conducted (such as the from MIT -75Appendix I: Results of Parameter Solutions using Corrections* Run ID Root Mean Square Residual/ Errors Number of Lander Delay Observations Used Plasma Comments AP-51 .489114 327 Standard parameter set plasma delay extrapolation AP-37D .350473 330 Corrected SX Bias AP-42A .347055 421 Uses old formal errors, uses points recovered by correcting range code errors AP-42R .346925 421 AP-42A to Identical except that the RELDEL parameter was included in the fit AP-41C .991057 330 Uses new scheme AP-41R .97272 330 AP-41C to Identical except that the RELDEL parameter was included in the fit AP-43C 1.00984 410 AP-42A to Identical except that the new formal errors are used and some lander points were deleted *The Standard Parameter in the above, and consists Set (see equation 5.2) of SPAN = 0.05 day C, = 1.0 C2 = 1.0 C 3 = 0.1 for lack formal of error is the default -76screen (see 0.1 day AP-48B 1.02357 410 thin ERROR in corrected times page 6a) SPAN = AP-48C 1.00984 410 Optimal 1 near smoother 1 = SPAN 10 , = C3 minute AP-48E 5.30030 410 Average of extrapolated 10 slopes with C SPAN = 0.05 da AP-48F 1.20653 410 Standard parameter set plasma delay interpolation AP-48G 1.00984 410 Optimal linear smoother plasma static using model -77The SXFILE Format Appendix II: The SXFILE and Doppler data. the is format chosen to store the SX range Each SXFILE consists of an 80 byte header, used as an 80 character identifier, followed by an indefinite number of Range Each data block observation, mation. or SX data Doppler blocks in time order. is 96 bytes long and contains the epoch of the observable itself, and various other infor- Each SX range data block is followed by another 96 byte data block which contains the original S-band and X-band delay observables, the calibration values applied to the SX observable, the ranging code from the ranging correlators. lengths used, and the output -78Information Name of Variable TIME Storage Inside Each SXFILE Block Type Purpose R*8 Epoch of Observation TIME = Julian Date - 2440000 KIND 1*2 KIND NMOD 1"*2 Number of modifications to this SX datum SCNUM 1*2 S/C number IQUAL 1*2 "Quality" of observationrelates to the nature of its calibrations FREQ R*8 Receiver synthesizer - Hz - SX R*8 S-X range g(- 1 for range 3 for Doppler frequency observable S-3/11X Doppler RUNID GENDTE R*8 R*8 MODDTE R*8 CLOCT R*8 Range Doppler - NCOMP 1*4 Range Used as 8 character tifiers - Doppler - run Maximum Code COUNT TIME iden- Length number of ranging Components not used IEE 1*4 Receiving station number DSNAME 1"*8 8 character name NDOPTP 1*2 Doppler ground mode = 0 for range receiving station = 1 for one way OoppieP= 2 for two way = 3 for three way " = 4 for three way coherent 14 trailing zero bytes -79If data is range Name of Variable RANGE (2) (KIND=1), a second Tvpe ~ Purpose R*8 S range record follows seconds X range SHORTC(2) R*8 S shortest codelength length X seconds RANCAL(3) R*8 S RANCAL values for this X observation S-X IVOLTS(2) 1*4 IVOLTX(2) 1*4 24 trailing zero bytes In phase highest frequency Quadrature code phase - S-band In phase highest frequency Quadrature code phase - X-band -80- Appendix Library Tape and the Plasma The Observation III: Correction OBServation LIBrary (OBSLIB) tapes are used to store delay, Doppler, and other measurements in a form suitable for use by PEP. Information of records. beginning The Type of OBSLIB tape an on OBSLIB Type OBSLIB are grouped series, each III record and Type the IV tape. into series, occurs at the A For includes the information time the residual respect to various IV time of on each order, although the data etc. data one for within a series themselves can be in and the other records tag of carried in series partials parameters, propagation the scaler variable are used CAL and of number corrections the Type IV information observable observable with corrections. The the plasma vectors. each 5, this observable, the (O-C), on is in Chapter the PREDICT runs itself, Type in series are numbered, of variety and stored the identification. record. plasma occur only at beginning follow series must be in time order, is used for types which contain data with the that, Data on 4 station, observable type, measurements record. kept Measurements Note any order. is I and Type II records the same observed body, ground One tapes are Three stored vectors in the and in storing the corrections: vectors to be placed on the NCAL The length of the CAL(I) Stores the actual correction tape a -81- SCAL(I) Stores an error estimate of the actual correction ICAL(I) Stores the rank of the correction; which correction to use The types of corrections and their used in deciding positional assign- ments within the new vectors are as follows: I CORRECTION NAME OBS. RANK TYPE STATIC NEUTRAL ATMOSPHERE STATIC NEUTRAL ATMOSPHERE TMDLY DOP ACTIVE TERRESTRIAL NEUTRAL ATMOSPHERE TMDLY DOP PASSIVE TERRESTRIAL IONOSPHERE TMDLY DOP ACTIVE TERRESTRIAL IONOSPHERE TMDLY DOP 1 1 TMDLY DOP STATIC SX FROM RNGNS CALIBRATIONS FROM DOPNS 20 20 TMDLY DOP 13 14 time) EXTRAPOLATED (in PLASMA CORRECTIONS (SX) 17 17 TMDLY DOP 15 9 10 SOLAR PLASMA CALCULATED IN MEDIA 11 12 INTEGRATED TMDLY 16 RANGE CAL. VIA S-X DOPPLER Not used 17 RANCAL RANGE CALIBRATIONS TMDLY 19 EXTRATERRESTRIAL ATMOSPHERE TMDLY 21 SX CORRECTIONS FROM UPOLT TMDLY The total elements propagation correc tion in the CAL vector. is a sum C A L C U L A T I N P E P over appropriate This total correction is stored in a variable named SUMCOR. The plasma correction, for the work reported thesis, is provided by the UPOLT program. In on in this general, the -82- OBSLIB tape already on case, that new plasma input it, to UPOLT generally the old in will have either a plasma CAL(15) or plasma correction must be correction, the residual and calibration CAL(21). replaced by the theoretical observ- able must be changed, and NCAL also may need to be changed reflect the increased size of the If UPOLT SXCORne w (see correction, is from the to CAL vector. the new plasma correction, calculated Equation In 4.24), and CAL vector, SXCORol then is d the the old update in plasma equations are Computedne w = Computedol d + SXCORne Residual = Residual - SUMCOR The new new new plasma CAL vector, = SUMCOR correction which is old old SXCOR + SXCOR is stored on new placed w new - in - SXCORol + SXCOR SXCOR the old old 21th the new OBSLIB d slot tape. of the -83- Appendix IV: Conversion between Julian Date and Civil Date Comments Julian Date Civil Date 2442779 1 January 1976 2442931 1 June 1976 2442980 20 July 1976 2443025 3 September 2443108 25 November 1976 2443145 1 January 1977 2443296 1 June 1977 2443420 3 September VL1 1976 1977 Lands on Mars VL2 Lands on Mars Superior Conjunction End of Data Span used in this Thesis 2443510 1 January 1978 2443667 1 June 1978 2443875 1 January 1979 2443895 21 January 2444026 1 June 1979 1979 Superior Conjunction -84- References 1. I., I. Shapiro, Goldstein, J. Sygielbaum, W. Viking F. Relativity Research, Vol. 2. Brenkle, D. P. Goldstein, M., R. 82, Experiment," # 28, pp. "Ranging with Martin, W. L., pp. 72-81, 4329-4334, 1977. Sequential and T. II, J. Jackson, and Sons, 8. J. Jokipii, R., planetary 42-36, pp. Pizzo, in "A V., the Vol. II, 35-40, Space Deep Network 1975. Chapter 7, John 1975. 6, Chapter John Wiley 1973. "Turbulence and Scintillations Plasma," 83, pp. Wind," of in the Inter- Astronomy and pp 1-28, 1973. Three-Dimensional Solar Review Annual Astrophysics Volume 11, 9. JPL Astrophysical Concepts, M., Harwit, Rang- "Terminology of Ranging Measure- Classical Electrodynamics, D., Wiley and Sons, 7. pp. 46-49, 1968. Sequential Acquisition Calibrations," Progress Report 6. JPL JPL Viking R013, 1976. Otoshi, DSS and ments Components," "system Performance of the Dual Channel MU-II Martin, L. L., Komarek, T., Geophysical 1969. Sequential Ranging," 5. Coded I. "The JPL Space programs Summary 37-57, Vol. ing System," 4. Binary "A of Journal Space programs Summary 37-52, Vol. 3. Jr., H. Michael, Cuddihy, and W. B. Komarek, A. Cain, T. L. R. MacNeil, E. P. Reasenberg, D. R. Model Journal 5563-5572, 1978. of of Corolating Geophysical Streams Research, -85- 10. van de C., H. Hulst, Corona," Astron. Soc. Bull. of Density Electron "The Nether. Vol. the XI, Solar # 410, pp. 135-150, 1950. 11. L., G. Tyler, P. J. "The Zygielbaum, Geophys. Res., 12. Kraus, 13. Nash, and A. Engineering Box, G. E. Vol. 82, # 28, S. K. Jordan, Corona Experiment," J. 1977. 5, McGraw-Hill, "Statistical Proc. Perspective," I. and Geodesy IEEE, Vol. and P., M. G. An Time Series Analysis- Jenleins, Applied Optimal Estimation, editor, A., Gelb, - 66, # 5, pp. Forecasting and Control, revised ed, Holden Day, 15. 1966. 1978. 532-550, 14. Solar Viking A. Komarek, A. Radio Astronomy, Chapter J. D., R. T. Brenkle, 1976. Press, MIT 1974. 16. Eubanks, M., T. tions, 17. and and R. D., R. Reasenberg, Vol. W. 84, "Determination E., M. Ash, Memo on Correc- Plasma 2 February 1977. Geophys. Res., 18. Goldstein, R. The King, # Bll, Rotation of Mars, 10, Oct. 1979. satellite Earth of J. Orbits," Lincoln laboratory Technical Note 1972-5, 1972. 19. Jazwinski, A. Stochastic Process and Filtering Theory, H., Academic Press, 20. Tyler, G. L., Zygielbaum, P. J. The Geophys. Res., 21. Shapiro, I. I., Rev. Lett., 1970. Brenkle, Viking T. Solar A. Corona Vol. 82, 4335-4340, Fourth Test 13, 789-791, of I. Experiment, J. and 1977. General 1964. A. Komarek, Relativity, Phys. -86- 22. Martin, Ranging Sequential Binary Layland, W. and J. L., W. with Sine Waves, JPL Deep Space Network Progress Report 42-31, pp. 23. 30-40. Goldstein, Signal 234: 24. Dunsmair, Series 1979, MacNeil, R. B. Breidenthal, J. P. Brenkle, D. L. P. M. Dec. 15, by Komarek, A. T. Relativity Retardation L219-L221, W., E. Viking Zygielbaum, of P. Kaufman, M. T. Cain, C. J. Shapiro, I. I. D., R. Reasenberg, Experiment: Solar Gravity, A. and I. Verification Ap. J., Vol for Time 1979. Robinson, Asymtotic Theory Containing Missing and Amplitude Modulated Data, to be published. -87- Figure 1) The Observation Geometry j4 . Eh 5',., 0-. c9~; FEe. L~C f 1 E~r3 -88- Figure 2) Simplified Ranging System Block Diagram -89- Figure 3) Idealized Ranging Receiver Correlator Output COyul_ ~r o4;p3 Coo#.stdo. 04-4Ir"ep Co'Y't-~k L"W,e Os".h. C044 1 Recei"e C.&s. reI I Data Flow Stream Figre 4,ATDF r ic, ur e 4 2 e/ a~t Re cc Oopp(et 6 f~?C~SY~.AO, e'4w st46Of P0 &Ipe id. 4 C)(t0-'VL ~ o s ,4 *% M C,4A &tsFc4.(.; / i 4c *tA J3E.td' 0 - DSA; AOCeSSOv'A Akt &J4C ,0cf&Lo5 wet4 - Dss /;.I, 34Iu C0 Nvet F"Ile- ( N 0L) .. i&A. PIOftessoot *e*dSbhZ :r+A ioa:rz Pecord. Atekue TeaOIG;A (x Aro P) ( ArP R. t cA~ -91Figure 5) 111 7 ncals - - - - - - - 4.5 Iec II - - - ----- - -- III II I II I I I I I I I I I I I I I I I I I I I I II I II I II I I I I I I II II II I II I4 II I I II I I II I I II I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I II I I I I I I I II I I I I II I I I I I No0 - - - -- .NCAL I II II I I S 4,o I RA I 4.2.'II Ps I 3.9 2. II II II rlr l I 3.5 1I II II II I I I I I - - ----- I I I r r I I r r I I I I I I I I I I I I - - I I I I I I I I I I I I f r I I I I I r I f I I I I I I I I III I 3. I I I 3.61 I I 3. I 3170 - 3180 3190 3200------ -92- Figure 6) The Geometry of the DSS Correction - sk.6 #e~ttafde P la Fs Al tea.4a. rofC f-0 $A 4& -J --- Re cei Le Q- r S~c~j.r'Cc~ Pe +e~elZy, ,Ce'*o~ Mic -93- Figure 7) SX Delay Data Processing at M.I.T. ATOrF or rT 00ODL PEP X 03J L 3 SXcAL WE - X.ml44 EJLf PLoP OfoA riei 4 SXr, .fs TF TV.pj SX r Lm Cfd i6J*.41€, I -, Oeo,. PAO, - Oelt"., .( 61 FE 3 X F 8.4 Possi;6 repet*{'PLdo SX FtLE en 0 SI 2 S9 0 4I o po OD S I I I 9 9 9 I --------- - 9 9 99 9---I 0 IV c% - 9 II 0 C N l-49 I---------I C 0 -I---999---I--9 ...... 0 o -9 9 0 999 9 9999 99 9 999999 9 99 99 9 9 .-.-----.. 9I--9..-- 0 00o 0 I I 1 IC- I I I On I\0 I 9-I 9 I I 99 9 9 9 9999 99 1 99999999 9 0 99999 999 I---------I-------------------------------------------------------------99999999--9 9 99 9 9 9 9 9 9 9 9 99 9999 99 999 I---------------------------I---------------------------------------------99-9999--9S99 I. | S9 I....--9 0 0 --------------------------------------------0 v 4 0 I9 I---------1------------------------- ----99--1--------1---------1-----------------1------ 9 9 9 I 0 >0 -I 1 Figure 9) Plot of isolated bad datum a a 500 I I I I I I- I a I I I I I I 450 40o I I I I I I I I I I I I I I I a a a I a I Ia II I a II! I I I I I I I I I a a I I I I I I I I I I I I a a a a a-I - I I I I I 350 a rII a a nsec a a I I I I 300 a I I I 250 II II I! 200 I I- '0 ! ! ! '0 '0! I 1- 150 a' I' '0 I I a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a I a a a a 100 I a a a '0 a a a a 50 o ' .0. '0.0 a a9 a a a a a 0 S3389 a a a 3390 Time, JD - 2440000 a I a - a- -96Figure 1o) Overlapping Integrated SX Doppler %fore Deletions 1 - VO-1 '- V0-2 1.0 *ihS3 .i9' 0.0 oo;t I -OIot .1i. ~ b i~i 1 j3t : *j E ?; 11 ,iI I~ T 0 .I "1, L. SX nsec t Alt ~C~ -40 ',. A L_ _ .975 .982 .989 .996, Time, JD - 2443200 '.003 -97Figure 11) Overlapping Integrated SX Doppler After Deletions 1 - VO-1 2 - VO-2 0 - Overlap 0.3 0.0 Al 1 -0.3 27 i I ; .9~ SX '1 9.t nsec -0.6 1 0 S It -0.9 " Ot ii 1 . -1.2 .975 .982 --' .989 Time, JD - 2443200 .996 1.003 -98- SX range vs. time near Superior Conjunction Superior Conjunction = JD 2443108 or Nov. 25, 1976 1 - DSN 14 4 - DSN 43 6 - DSN 63 Figure 12) 20.0 18.0 - I 'W- a a I a I I I a a I a a II I SI - I a I I I t I S I I a II I I I I I .4 I a I I II a a * * a, 14.0 a II I a a a a 12.0 '-- Ia a a 10.0 Ia I I I I a a a a a a aa a I a 8.0 aa a I a - a a a I1 S6.0 4.0 - I I i a a ae a a a a Ia aa a aa a a a , I a a a Ia a a I a a a a a a • -- a a I a a a a a a I a a i a I II a a a a a a a - -r a a a 2.0 '.- . I I I aa a a a a a a a a i , I I aI I aa a a a a a a a I Ia a a a Ia a a I a I a a I a a a a ar ar a a a a a I i I S, a a a I i S SI a a 16.0 I1 I1 I I I II II II II I II II I a I I I ,t .t a a a a a a a a a a a a a a a a a a a a a a a a . a aI a a a a a I I a a a a a a 'I- a a a a a a , a a a a a a "r a a a a a a a a ..C. II II I a a I a a a a t 1I I I I I II II I ( a I a a a I a I a Ir II I1 I II I 1I 1I r a a a a a lr a a a a II II II sIs( JI Jll 4r I 3 . oi0 Jo .m .- Tl*4V --I Time days -- w 31o .-----o T--r II II II II 1I II I I II 'I II 11 II II II I I I I I1I I III I rI I II II I I I I1 I I II 9( r(J9 99 , -,-r-- So f -99- SX Range from V01 DSN 43 on JD 2443110 to JD 2443111 Nov. 27-28, 1976, 2 days after Superior Conjunction Figure 13) ___________ 4A n I U.U _________I______________ II 1I - I 9.6 a 9.2 8.8-d -a a I a I ai a~ Ia I I I 8..- aa a a ± II 7.6-' 7.2 - aa 14 I I aa a aI I a aa a aa a a Iaa a aI a a a 2. 6 e ,I 4P ,4' a a I I I ! I ! I I Ia I ! *a I !I -! I 6.4-- I 8.0" II t a • .o " a. 6.8 - I I I - IaI St a I I aI I a a II I 1t I I I I I I t 7.6a a a a a a I I I I I I I SI SI aI I e I I I , I ! 6.r _ ,4 I JJ 9 93SA4 I I I I 1I II I Ju II r1 II 1 1 I I I I I q.9( II I II II I I1 II I I I I I - 4 7 .97 Time (in days) Data taken at 7 min 30 sec intervals Time = Julian date - 2443110.0 1.0o l.OZIr i.os3 C0 o - In n .I ... C" SoI rI Sp H I I I ........ I -- II......... - - I I......... *- II - I I ........- -1-- -- I I ........ --- t - K ......... -I -1-----I --- I I ........ - - I I .... I ... - 44 I ...... ......... I I- - I J II I I 1---------1 I- - ---------- ,I -I-i I-----.. . I......... i ......... l --------------------------- 1 I.... ... I ... ... I ... ... I........ I.... ... I........ I........ I-------I-1---- -------- I ------ t I I I S---------I------I ----- I -------~cU~--------------I------1.-----oxg E t o ', 0 o -101- Figure 15) SX' = SX/H(L) Autocorrelation t I = 2443058 t2 = 2443158 at - 0.1 day Mean subtracted from SX' sequence Straight line provided for reference only - ---_ -- -------------. -- -- --- , - - - ----- _ _ So.A - r- 0 1"67 * o" -- e -------- -- -. . Lag --- days 4 r -102- Figure 16a) Correlation of First Differences of SX' - SX/H() t i = JD 2443058 t= dt = 0.1 day (the mean was subtracted from the Differenced Sequence) ----------------------------7 ------1.0 0.2 - -I I 0.8 I I II * -0.0 * * Note: I * .I I I I 1.o0 I I I I data usedI in Figure IS. This Iis over the same I I 0.6 JD 2443158 0.4 * I I I I I I I I 0 I I I '0 0 I a 0.0 *Goo I 0 a to * I a -0.2 " II* I I I -0.4 • I I I I I II I I I~~ I I I II I I Lag, days I I I. I I I I I I I I I I I I I I I ..I I ooo I * I 6-o o1 ILg Noe 10 2. 3. days Thi aedtaue isoe nFgr -0.8, -1.0 I I I I I I I I 1.01 -0.6-- -- 0 I *, ~ooa coc aoI I I - -....-- " - - -- 5 -103- Correlation of First Difference of SX' - SX/H(J) Figure 16bi t -= JD 2443350 t2 = JD 2443400 At - 0.01 day 5107 Range Differences The mean was not subtracted from this sequence - -e -- --- 0.6. I I aI 0 0 0. 0 0 .. ' 0. 41 0 a a a o~~ a a : a 0.1 0.2 a . o 11 0 0 0.3 a a a a a a~ a a a a U -0.4.--' 7 o Ia a 0.0 a I ,o 0a 0 o~ a -o)... a 11 a -1 Lag, days -104Figure 17) Plot of Integrated SX Doppler Circled points are Delay .-------------------------. . .---- - T TT . . -- . . T i_' ,4 - 1 -- 7 i -- 7 SX 1Time, 1*3 -- I -- - 7 D- 2443124 _ psec 7 7 7 .84 1, 7 77 7 78 .6877 7 7 7 77 7 7 7 7 7 7 7 788 77 ± - -- - - ,8 -- - - -- Tit - -- - .586e7 - JD- -- - - -- -- -- 4412 - - -- - - -- - - 8 -105- Figure 18a) Autocorrelation of SX Doppler t1 I JD 2443095.353 t2 = JD 2443095.595 At = 1 minute 349 V02 SX Doppler Observations The mean was subtracted from this data sequence ---1.0 -o ---------------------------------I I 0I UI -S0.2 I 1 I aa a a) a a a a a 0.6 M-a a a 4) 0 110 a a o 04 -0.4 a O. ' - - ( II 0 -- o 0.6 0. a I a a a a1 a. ao a a a a1 I I~ a a a a a ra (a ~ -- -- -0.6---Lr( 02a'-0 4)I -0.8riI 00't -I ..... a Ia -- -- in t Ia a an-r ao a a a a aa a aa a Lag ~~ Ia 20( a a04 a a a - - - - -I - - - - 1a aa ao a a a a Ia0 a- , a aa a Ia 30 a a a a --- -- a a a a a a 40 a a -- -- aI Oo:II oaI a 00II o a . 04 a I0 0 a 0 a 08... a a a 3 a 4 ate ~ -106- Autocorrelation of SX Doppler Data tI = JD 2443101.080 Figure 18bI t2 - JD 2443101.347 At = 1 minute 383 SX Doppler Observations - all V02 The mean was subtracted from this data sequence 1.0 lo 0.8 " 0.6 - 0.4 0 0, 0.2 - 0 -4 o II.. 'U a a O0 a a I 0.0 I a O a 0a ,oo Sao ao a o so _ ,, I 01i a a------ a 0 a aa a aJ a a6a I -0.2 - -go% 0 I -0.4 I I I a a a I a a a a -* a -0.6 I I SI -- -0.- -II a a a a -084. i . a Ia a. a aa a a a a a a a1. a a a a i .2 Lag, minutes a a -107- Figure 18 c) Autocorrelation of SX Doppler Data t = JD 2443117.685 t 2 = JD 2443117.800 &t = I minute 166 Observations from VO-2 1.0 - o 0.8 - 0,6 - - 0. .L - 0.2 0 Correlation 0 Coefficient -0.2 ..aa aI aI 0 I0 a 0 a 0 0a ca o a a o a 0 LAG, 0 a a o n0 o I 8 n a minutes co a aC C a a a a 10 20 30 a a -0.6 -0.8 - -1.0a.a.. 0 LAG, minutes 40 -108- Figure 18 d) Autocorrelation of SX Doppler Data tI = JD 2442124.839 t2 = JD 2443124.9577 at = minute 171 ObserLations from V0-1 1.0 o, 08 , , ; . 0. i I I 7I I I I I I , I 7, 7s I I ' I I I 7 -7 7 , a , 7 7 7 a.a a , a i I 0.6 0.2 -0.2cin -- SI a a a'0 T T I T I - ..a 0 (O aI 0. Correl a1 a a SI Coeffi -I_ I a I~s I I SI 000 Sa a a a -0. a a a a a a a 4 0 0"aaOI -0. -o.6 a a a a a a a a a a a a a a a a a a a a a a a a a a a a a 10 a a a I 2. 2 2 2 -L -086 00 a a a a , a a a aa. a 00 l a Go,, ... a a a a a a O aIO 0 O 0 00 a a a a a a a a a S ,a a a a a a a a I Ia a a a a a a a a a a a a a a a a 0 00 0 O OOIa I a a a a 0 a0 a00 I 0 I , 000 I 0 a a a a a a a a a a a a a a a a a a a a a a a 30 aa a a a 40 a I I 2. a a . a a a a a a a aaa a 20 aa a a a 0 02 3 -1.0 ..... LAG, minutes , - " -109- Figure 18 e) Autocorrelation of SX Doppler Data ti = JD 2443160.8851 t2 = JD 2443161.317 at = 1 minute 606 Observations from V0-1 1.0 0.8 - - - - 0.6 0.4 a 1 Correlation Coefficient 0.0 ac * a * -- I| / 0 a 0 £ 0 0 a Y a !- , Oi a - -0.4 -0.6 0a 10 a - - 20 a I.................. - - -o0.8 -1.0 LAG, minutes 30 S -110- Residual Plot from Run AP-48F Standard Parameter Set Plasma Delay Interpolation Figure 19al 400- 300 200 100L 44. 4 I 4; x, ft 0tP a fg UI - ioz E:z ~_ ___ e $ z"- -V% ee * z Y z -100 -200 -300 -400 3000 3100 3200 3300 Time Julian Date 2440000 3400 -111- Figure 19b) Residual Plot from run AP-48C Optimal Linear Smoother with Thin Screen Plasma Model 400 300 200 10 z 1* SA A ia. r - " W z = ~- - -- z & a , i SI . r • z x -100. z U 0 U -2001 0 -300 7 -400. , 2980 3000 . I 3200 3100 Time Julian Date - 2440000 3300 - : .3420 3400 -112- Figure 19c) Residual Plot from run AP-48E Plasma Correction in Average of Extrapolated Slopes 400 - 3001 z z 200 1 0 x 6 0 z 0 4 + S1~ I z Xf a . a~ £ t 00 I A O0 - I II J *Z c I Y z z*C -200 -3001 -400 . 2980 3000 . . . . .. . " 3100 L 3200 3300 Time Julian Date - 2440000 S'3420 3400 -113- Optical Linear Smoother Residual Plot from run AP-48G ) Using Static (Pup = Pdown plasma model. Figure 19d) 400T Too 300oo 2001 100" z z+ i'= 3* z 1 * 0 C iI z Z +z 0 - i I - , - zz 4 a AA z ry I z z~+ I -200 o o oo - -300 -400 2980 ;,3000 3100 3200 Time Julian Date - 2440000 3300 -3420 3400 -114- Figure 20 0) HOG Output Residual Scatter versus 6Tsc 410 Lander Range Residuals Bin Width = 10 days Date of Superior Conjunction: or November 25, 1976 Bin No. 10 0 o JD 2443108 20o - - - - I 6 a a 440 0 0 T a- 0 a IJ i1 o 80 , 1 0 TSC 100 SC, 20 6 Tsc IT - 200 days 2I TSCf, days 300 0 300 -115- Figure 20 b) Histogram of Number of Lander Observations versus Time to Superior Conjunction for Run AP-48C. Bin Width = 10 Days 50 L gg r I Number of Lander Observations in Bin 30 2• 1I .i -rr iI 1 12 r Ii r II 1i ii Il ii ii rr i* rr rr Il rr i ,irI i rr rrr; r i*ri *rixi rrrr Bin r i 1 i iumber i x * i r 111 Izi s s -116- Figure 21 a) HOG Output Residual Scatter vs. Smoother Interpolation Time Interpolation time- IThin Screen time - time of nearest SX observationj 410 Lander Range Residuals Bin Width = 0.01 day (14.4 minutes) Run AP-48F Bin No. S o 1.2 1.0 a0 00 0.4- 7 0 ~.0 a C 7 a 7 7 - 0.1- -- - - 0.2 -.-7 -- - - 0.3 0 0- -- - - -6. Interpolation Time Days - --- - -117- Figure 21 b) Histogram of number of Lander Obervations versus Smoother Interpolation Time for Run AP-48F. Bin Width = 0.01 Day 60 - 50 40 rr rI rr rr rr rr rI r rr rr Et 111 II ) Lander 30 Observations in Bin 20 10 0 I * Il II II II II II II II 1: ~ I Ii .. r rrr II rrrr -r Irri II 11 rr 1111 rr rrnrrrrr r rir rr rrn Irirrrrir rIIrrII nrr ~r rirrrrrr rr rrr rrr*rrr+rxrr rrr rrrr*r rr r rrr ~ a~ll 11~1 rr*- II*l II*1IILII n " o0 Bin Number I= I Sjo ! -118- Figure 22a) HOG Output Residual Scatter vs. Number of SX Range Observations Used in Interpolations 410 Range Residuals Bin Width = 2 SX points Bin No. 0 10 20 30 7 7 Ci "4 2.0 M 7 0 7, C.. I , 16' w 1.21.0 0 0.8 .4 Ci o I , , C,) 44. 0 '0 0.4._: 10' 0 0.0 Cn o 10 0 I 0 20 0 0 40 Total Number of Points Used in SX Interpolation 60 -11ii9- Histogram of number of Lander Observations Figure 22 b) versus number of SX delay observations used in interpolation for Run AP-48F. Bin Width = 2 SX delay observations 100 r 90 I L z L z s ,I 60 "I Lander Observations 50 11 in Bin 1 40 z z r .i r 1 11* t 30 ,,fL ilI 11 III 11 '11I II iI1 1I ;L 10 0 II J '/I I) 11 1I II II1 1 11III 1IfllXI 11 fII11 rt I 1 I ! t I 111 1x, 1 t *,r 1 1 CJI ILllt r Bin Number i