PROCEEDINGS OF THE IRE

A Swept-Frequency Interferometer for the Studyof High-Intensity Solar Radiation

at Meter Wavelengths*J. P. WILDt AND K. V. SHERIDANt

Summary-The paper describes an interferometer which hasbeen constructed for studying the various spectral types of solaractivity at meter wavelengths. The instrument is a "swept-frequency"interferometer and is capable of measuring the one-dimensionalposition and angular size of a transient source on the sun's disk, andof determining the polarization and intensity of the received radia-tion. All these characteristics are determined as a function offrequency in the range 40-70 mc. Independent position measure-ments are made at one-second intervals, and source-size measure-ments at two-second intervals. Polarization measurements are madeas an alternative at chosen times. The two antennas of the maininterferometer are spaced 1 km apart on an east-west line. Thetwo antennas of the polarization system are erected on a commonaxis maintained in the direction of the sun.

Some preliminary observations have given an indication of thespectrum of source size for activity of spectral Types I and III. Inboth cases the angular size is found to decrease from about 10minutes of arc at 45 mc to about 6 minutes of arc at 65 mc.

I. INTRODUCTION

r HE sun's electromagnetic radiation in the meter-wavelength part of the radio spectrum is domi-nated by a spasmodic variable component which

originates in active areas above certain sunspots. It con-sists of short-lived bursts of radiation lasting from a fewseconds to several minutes, sometimes superimposed ona variable background continuum of longer duration.The majority of bursts conform to one of a relativelysmall number of spectral types of which three have beendescribed in some detail.1-4 Hypotheses have been ad-vanced to interpret these spectra in terms of physicalprocesses taking place in the solar corona. For example,in one of the commonest types of burst (spectral TypeIII) two frequency bands are emitted, one the secondharmonic of the other, and these drift rapidly in thedirection of decreasing frequency, sweeping throughtens or hundreds of megacycles in a period of a fewseconds. Reasons have been given for supposing that

*Original manuscript received by the IRE, October 18, 1957.t Radiophysics Laboratory, CSIRO, Sydney, Australia.1 J. P. Wild and L. L. McCready, "Observations of the spectrum

of high-intensity solar radiation at metre wavelengths. Part I-The apparatus and spectral types of solar burst observed," Aust. J.Sci. Res., vol. A3, pp. 387-398; September, 1950.

2 J. P. Wild, "Observations of the spectrum of high-intensitysolar radiation at metre wavelengths. Part II-Outbursts (typeII). Part III-Isolated bursts (type III). Part IV-Enhancedradiation (type I)," Aust. J. Sci. Res., vol. A3, pp. 399-408;September, 1950, vol. A3, pp. 541-557; December, 1950, vol. A4, pp.36-50; March, 1951.

3 J. P. Wild, J. D. Murray, and W. C. Rowe, "Harmonics in thespectra of solar radio disturbances," Aust. J. Phys., vol. 7, pp. 439-459; September, 1954.

4 J. P. Wild, J. A. Roberts, and J. D. Murray, "Radio evidenceof the ejection of very fast particles from the sun," Nature, vol. 173,pp. 532-534; March, 1954.

the physical process responsible for such a burst is theexcitation of plasma oscillations in the solar corona by adisturbance (e.g., a corpuscular cloud of ionized matter)traveling outwards through the corona exciting levelsof continuously decreasing plasma frequency. The de-duced velocity of an average disturbance is some 70,000km and a possible connection with the cosmic rays ofsolar origin has been suggested.Much the same interpretation has been given for the

longer-lived bursts of spectral Type II, though here thededuced velocity is typically 500 km of the order ofthe velocity of solar corpuscles responsible for aurorasand magnetic storms on the earth. Also there are long-period storms of Type I bursts individually narrow inbandwidth, short in lifetime, and circular in polariza-tion, which, together with other rarer spectral types, allremain without satisfactory explanation. In order tomake further progress-to test the hypothesis on whichthe existing interpretations are based, and to explorethe emissions of unknown origin-further informationis required to supplement the spectral data.

In particular, we wish to know the position, angularsize, and polarization of the radiating sources. Our pres-ent knowledge of the positions and motions of thesources on the sun's disk is very limited. In 1950, Payne-Scott and Little5 made valuable directional observa-tions of these sources with a swept-phase interferometeroperating at a single frequency (97 mc). But to obtaina proper understanding of the origin of the bursts itseems essential to study the positions over a range offrequencies, since different frequencies are thought tooriginate from different levels in the corona. Also ourknowledge of the angular size of the short-lived sourcesis almost nonexistent. Previous measurements of angu-lar size,5-8 which probably all refer to Type I storms,indicate that at frequencies between 97 and 200 mc thesize is not more than a few minutes of arc. A quantita-

5 R. Payne-Scott and A. G. Little, "The position and move-ment on the solar disc of sources of radiation at a frequency of 97mc. Part II-Noise storms. Part III Outbursts," Aust. J. Sci.Res., vol. A4, pp. 508-525; December, 1951, vol. A5, pp. 32-49;March, 1952.

6L. L. McCready, J. L. Pawsey, and R. Payne-Scott, "Solarradiation at radio frequencies and its relation to sunspots," Proc.Roy. Soc., vol. A190, pp. 357-375; August, 1947.

7M. Ryle and D. D. Vonberg, "An investigation of radio-fre-quency radiation from the sun," Proc. Roy. Soc., vol. A193, pp. 98-120; April, 1948.

8 Y. Avignon, E. J. Blum, A. Boischot, R. Charvin, M. Ginat,and P. Simon, "Observation des orages radio6lectriques polairesavec le grand interferometre de Nancay," Compt. Rend. Acad. Sci.,vol. 244, pp. 1460-1463; March, 1957.

160 January

Wild and Sheridan: A Swept-Frequency Interferometer for Solar Radiation

tive knowledge of source size is necessary both to gaugethe scale and spread of the disturbed region and to cal-culate the intrinsic radio brightness of the sources.Polarization observations, notably those of Payne-Scott and Little' and Hatanaka,9 have been more ex-tensive, but apart from some recent observations ofKomesaroff (in preparation), these have been restrictedto single frequencies.The instrument described in this paper is an interfer-

ometer constructed with the intention of making rapidmeasurements of the position and size (both in one di-mension) and polarization of the transient sources overa substantial frequency range. These data are obtained)side by side with records of dynamic spectra using theradio spectrograph at Dapto (40-240 mc). Hence it ishoped to obtain a fairly complete set of observablecharacteristics of the different spectral types of burst.The instrument covers a frequency range of 40-70

mc, the lowest range of the spectrograph. There wereseveral reasons for choosing the lowest range. The firstis dictated by the nature of emitted spectra. Referencehas already been made to the natural emission ofharmonic frequencies, a phenomenon common to burstsof Types II and III. On current evidence it appears thatthe fundamental frequencies rarely exceed 100 mc,while harmonics are received over a much wider range.By choosing a frequency range below 100 mc, therefore,we should be in a position to observe the sources both intheir fundamental and harmonic emissions. Second, itcan be shown that an interferometer of given base lineis capable of discriminating between the 40- and 70-mcplasma levels at the limb of the sun more efficiently thanbetween the levels of a pair of higher frequencies (e.g.,70 and 120 mc) bearing a similar ratio'0-this despitethe sacrifice of angular resolution at the lower frequen-cies. Finally, the technical problems in the building of anaccurate interferometer of given base line decrease withfrequency. A disadvantage of low-frequency operationis the increase in errors of position due to refraction inthe ionized atmospheres of the sun and earth.To obtain a sequence of positions of sources which

exist for no more than 5 or 10 seconds it is necessary tomake measurements at intervals of about one second.Little and Payne-Scott"' fulfilled this requirement at asingle frequency by rapidly sweeping the phase of sig-nals entering through one arm of their interferometer,thus causing the fringes in the antenna pattern to be

9 T. Hatanaka, Polarization of solar radio bursts," in "RadioAstronomy," H. C. van de Hulst, ed., Cambridge University Press,Cambridge, Eng., pp. 358-362; 1957.

"The Faraday effect in the earth's ionosphere with specialreference to polarization measurement of the solar radio emission,"Publ. Astron. Soc. Japan, vol. 8, pp. 73-86; 1956.

10 This merely expresses the fact that the distribution of electrondensity, N, in the solar corona is such that VN (proportional toplasma frequency) decreases outwards more slowly than expo-nentially.

11 A. G. Little and R. Payne-Scott, 'The position and move-ment on the solar disc of sources of radiation at a frequency of 97mc. Part I-Equipment." Aust. J. Sci. Res., vol. A4, pp. 489-507;December, 1951.

WEST EAST

SWEPT SWEEPEXTRA FEEDE FREQUENCY GENERATORLENGTH, E RECEIVER

INTENjSITY 4JL

-m..FREQUENCY

Fig. 1-The principle of the swept-frequency interferometer.

scanned rapidly across the sky. In the present case weneed an interferometer in which both the antennapattern and the frequency are swept. This has beenachieved in a single operation; i.e., sweeping the fre-quency in an asymmetrical two-antenna interferometer.The principle of such a "swept-frequency" interferom-

eter has previously been applied by Wild and Rob-erts'2 to the problem of radio-star scintillations."3

II. THE SWEPT-FREQUENCYRECEIVER

A. The PrincipleThe principle of the swept-frequency interferometer

is illustrated in Fig. 1. Plane waves incident on anasymmetrical two-antenna interferometer at an angle0 with the normal plane of the interferometer travelthrough the two antennas along paths which differ bya sin 0+1, where a is the spacing of the antennas and £the extra feeder length in one side. The interferingwaves are received in a swept-frequency receiver anddisplayed on a cathode-ray tube which traces the pat-tern of intensity vs frequency. It is evident that thispattern will be just the spectrum of the source modu-lated with an interference pattern having maxima atwavelengths for which a sin 0+I= nX and minima forwhich a sin 0+I= (n+ )X, where n is an integer. Whenall frequencies arrive from the same direction, suchmaxima occur at equal intervals, Af, in the frequencyscale, given by

12 J. P. Wild and J. A. Roberts, "The spectrum of radio-star scin-tillations and the nature of irregularities in the ionosphere," J.Atmos. Terr. Phys., vol. 8, pp. 55- 5; February, 1956.

13 The latter instrument was also used for some preliminary solarobservations. See R. L. Loughhead, J. A. Roberts, and M.McCabe, "The association of soiar radio bursts of spectral Type IIIwith chromospheric flares," Aust. J. Phys., in press.

1958 161

PROCEEDINGS OF THE IRE

cAf = asinO+l ((0) < 7) (1)

where c is the free-space velocity of light. Thus measure-

ment of the frequencies of maximum or minimum en-

ables the position angle, 0, to be measured at those fre-quencies. Provided the instantaneous position of thesource is approximately independent of frequency, none

of the ambiguities in position encountered with single-frequency interferometers arise since, for a given vlaueof 1, Af is a unique function of 0. (The value of Af can beadjusted to any desired value by varying the length 1.)Also by measuring the maximum-to-minimum ratio ofthe intensity of the interference pattern we can obtaina useful measure of the angular size of the source if thelatter is no wider than about one fringe spacing and no

narrower than about one tenth of a fringe spacing.While the simple interferometer shown in Fig. 1 is

suitable for the rough measurement of position, diffi-culties arise when measurements of high precision are

required (e.g., a few hundredths of a fringe spacing). Inthe first place, the finite bandwidth of solar emissionscan cause a considerable distortion of the sinusoidalpattern and so complicate the precise location of thetrue interference maxima and minima. Second, in thecase of sources of relatively large angular width, theinterference pattern is superimposed on a high-incoher-ent background with consequent loss in the visibility ofthe pattern. We have overcome these problems bymeasuring positions with an interferometer in which thesignals entering the two antennas are multiplied to-gether rather than added. This technique is familiar inradio-star interferometry where it is desirable to elim-inate the galactic background.'4 Fig. 2(b) shows thekind of pattern obtained when a partially resolvedsource is viewed with a multiplying interferometer; thismay be contrasted with the adding pattern shown inFig. 2(a). With multiplication the receiver amplifica-tion may be increased as desired, and the phase of thepattern can be accurately measured using the points atwhich the pattern crosses the zero line.

However, the adding interferometer is still requiredfor source-size measurements. Our instrument, there-fore, has been made to combine both adding and multi-plying facilities, and is automatically switched betweenthe two conditions during alternate ('-second) sweepintervals. Where possible, the two systems make use ofthe same equipment. In describing them, however, it isconvenient to regard them as two independent instru-ments.So far we have restricted the discussion to problems

of measuring position and source size. We shall continueto do so until Section IV-C, where it will be shown thatpolarization can be measured by replacing the spacedantenna system with a pair of crossed antennas.

14 For example, see M. Ryle, 'A new radio interferometer andits application to the observation of weak radio stars,n Proc. Roy.Soc., vol. A211, pp. 351-375; March, 1952.

(a)

Frequency

(b)Fig. 2-Characteristic patterns of intensity (vertical) vs frequency

obtained with (a) an adding interferometer and (b) a multiplyinginterferometer.

B. The Adding Interferometer

A block diagram of the adding interferometer isshown in Fig. 3. For source-size measurements the sig-nals are received in two identical broad-band, wide-beam antennas spaced 1 km apart on an east-westline. Connection to the receiver is made by long feederlines, one of which includes a special section capable ofbeing varied from 0 to 480 m in steps of 15 m. The sig-nals are combined in a simple junction, detected by a40-70 mc swept-frequency receiver3 and displayed ona cathode-ray tube provided with a horizontal sweepsynchronized to the receiver sweep. The receiver isswept at 2 revolutions per second, the operative periodbeing 8 second. Thus traces of the form shown in Fig.2(a) are generated twice per second, and these arephotographically recorded. Facilities are provided forinjecting standard power levels fronm a noise generatorand frequency markers from a harmonic crystal gen-erator.The function of the variable line is to provide control

of the frequency spacing between maxima in the inter-ference pattern [Af of (1) ]. If this spacing is allowed tobecome too large, the interference pattern is liable to be-come confused with the envelope of the natural spec-trum; also we would sacrifice the number of independentmeasurements across the frequency range. If the spac-ing becomes too small the interference pattern wouldbecome blurred by the finite bandwidth of the receiver(chosen to be fairly large, 0.5 mc, to minimize noisefluctuations) and the reading accuracy would decline.In practice the most suitable spacing is found to beAfr-5 mc, so that we need to arrange for the differencein the total path length from the source to the receiverthrough the two antennas to be approximately 60 m[see (1) 1. Now as the earth rotates, the path length be-tween the sun and the two antennas changes by up to+ 1 km in 12 hours. The present equipment has beendesigned to operate during the interval noon ± 2 hours,during which the path-length difference varies by about

162 January

1958 Wild a

F1 - --1 Km.

W. MELJAL Ar--G_

EXTRA LENGTH60m. I

ind Sheridan: A Swept-Frequency Interferometer for Solar Radiation

__E.1 t AERIAL

IbLEIll1 CLOCK-REGULATED

IEXI CONTROL UNIT

GENERATOR

HAMOI L- HoNc 2s CRYSTAL|

FREOUENCY MARKERS

40-70 Mdis ,

SWEPT-FREOUENCYRECEIVER ______TUNING

L.F BANDWIDTH MOTOR__ 1025

I | ~~~~D.C.AM UFIER

EliFig. 3-The adding interferometer.

± 480 m. This path length is compensated by the vari-able line which is automatically inserted in the appro-priate feeder line and adjusted to the correct length(see Section III-B) .For polarization measurements the two spaced an-

tennias are replaced by two antennas having a commonprincipal axis but orthogonal planes of polarization. Anadditional 60 in of feeder length is included in one arm.The reason for this will be discussed in Section IV-C.

C. The Multiplying InterferometerA block diagram of the multiplying interferometer is

shown in Fig. 4. The antenna and feeder arrangementsare the same as those of the adding interferometer butthe signals in the two arms are amplified and convertedto the intermediate frequency (30 mc) before beingcombined. It is necessary to preserve the relative phaseof the two signals up to the common mixing point andso the two converters are supplied by a common localoscillator. The frequency is swept by tuning, in syn-chronism, each radio-frequency amplifier and mixer,and the common local oscillator. The two 30-mc signalsare then multiplied by phase switching and phase de-tection; the principle of this method is contained in (6)of Section IV and the technical arrangement is essen-tially similar to that used by Mills, Little, Sheridan,and Slee.'5 A high switching rate of 5 kc is used becauseit is required to make several switchings during the time

B. Y. Mills, A. G. Little, K. V. Sheridan, and 0. B. Slee, "Ahigh resolution radio telescope for use at 3.5 m," this issue, p. 67.

Fig. 4-The multiplying interferometer.

(-10-2 to 10-3 seconds) taken for the frequency to beswept through one receiver bandwidth. The output ofthe phase-sensitive detector is passed through a low-pass filter to the Y plates of the cathode-ray tube.The factors influencing the accuracy of position

measurement will be discussed in Section IV, but two ofthe factors require instrumentation and should be men-tioned now. The first is the provision of accurate sharpfrequency markers. These are inserted automatically,at intervals of two seconds, by injecting harmonics ofa 2.5-mc crystal oscillator into the input terminals ofone of the rf amplifiers; at the same time the inter-mediate frequencies are diverted from the main IFamplifier (bandwidth 0.5 mc) into one of bandwidth one

163

PROCEEDINGS OF THE IRE

Fig. 5-Two typical records of Type III bursts obtained on July 3,1957. As indicated by the clock, time advances vertically down-wards. Patterns of the multiplying interferometer are displayedat one-second intervals, and those from the adding interfer-ometer at two-second intervals.

kc, whence they are detected and displayed as a ver-

tical deflection on the cathode-ray tube. The appearance

of the marker pips is shown in Fig. 5.The second calibration facility is the provision of a

means of determining the relative shift in the phase ofsignals passing through the two receiver channels; therelative phase shift must be known as a function of fre-quency and determined at sufficiently short intervalsof time. This is done by replacing the antennas andfeeders with a pair of artificial phase-coherent sources ofnoise which generate an artificial swept-frequency in-terference pattern. Signals from a broad-band diodenoise generator are split into two channels and attenu-

ated in identical networks to prevent interaction be-tween them. One is fed to receiver A through a cable oflength s, the other to receiver B through a length s+b.Thus an interference pattern is set up on the displaycorresponding precisely to that from a celestial source

for which the path difference through space and throughthe antenna feeders is b. In practice the length b ischosen to be 60 m giving fringes spaced at the desirableinterval of 5 mc.

D. The Combined Display

A 12-inch cathode-ray tube is used to display all therequired information, viz., multiplying and adding inter-ferometer patterns, and narrow-band frequency mark-ers which are superimposed on the former. The tube isphotographed with a 35-mm cine camera whose film isadvanced by 1/10 inch between alternate scans (i.e.,at one-second intervals). The final record is illustratedin Fig. 5 and Fig. 6. The record also includes clockimages at 2-second intervals and a bunch of illuminatednumbers to indicate the setting of the variable line.

Fig. 6-Two sections of the record taken on June 20, 1957, during aType I storm which was characterized by a high backgroundcontinuum. (Note: During the 'polarization' section, no addingpattern was recorded in this case.)

The combined display is produced automatically byusing magnetically operated switches to control a cycleof operations which covers 4 scans (2 seconds). Thescans in each sequence are shared as follows: 1) multi-plying interferometer, 2) adding interferometer, 3)multiplying interferometer, and 4) frequency markers.During 2) the trace is deflected to the right and thehorizontal sweep reduced in amplitude.

III. THE ANTENNA FEEDER SYSTEMAND ITS CALIBRATION

A. Antennas and Feeders

The spaced antennas of the interferometer are single-wire rhombics each mounted on a wooden cross in thevertical north-south plane and pivoted at the center toallow for two different declination settings (summerand winter); no azimuth movement is necessary be-cause the beamwidth of the antennas is wide enough toaccommodate the sun's daily motion during the operat-ing hours (noon±2 hours). The antennas are designedfor uniformity of effective area in the frequency range40-70 mc (see Wild and McCready'). Each side is oflength 3.64 m, and the semiangle at the side corners is270, giving an effective area of about 9 m2. Each antennahas a characteristic impedance of about 700 ohms and isconnected directly to 700-ohm open-wire transmissionlines."' The length of each feeder is about 800 m, havinga measured attenuation of about 4 db at 50 mc (seeSection III-C). Near the receiving station the feedersare quasi-exponentially tapered to transform the char-

w These consist of a pair of copper wires of diameter 0.064 inchspaced 11 inches by thin perspex spacers at intervals of 30 feet andsupported 10-15 feet above the ground on wooden poles.

January164

Wild and Sheridan: A Swept-Frequency Interferometer for Solar Radiation

acteristic impedance to 300 ohms to facilitate connectionto the receivers.

Polarization measurements are made with a differentpair of rhombic antennas which are mounted on a com-mon axis and with their planes perpendicular. Eachantenna has the same dimensions as the interferometerantennas, but the pair is equatorially mounted anddriven to follow the sun. This allows polarization ob-servations to be made throughout the day.

B. The 0-480-Meter Variable LineAs explained in Section II-B, the function of the

variable line is to maintain the frequency spacing be-tween fringes in the interference pattern at a value ofthe order of 5 mc for different positions of the sun in thesky. While it would probably be feasible to develop asuitable compact nondispersive artificial line, we havepreferred to minimize problems of matching, dispersion,and loss by the use of long lengths of the normal 700-ohm open-wire line.The variable line consists of 5 loops of the open-wire

line of length 15, 30, 60, 120, and 240 m, respectively.Each of these may be either switched into the antennafeeder line or by-passed with a direct connection. Aplan view of two of the loop sections is shown in Fig. 7.The switches were designed to avoid appreciable re-flections in the radio-frequency line and are operated bysolenoids.-7

Similar switches are used in a system to transfer thevariable line from the eastern feeder to the westernfeeder at noon each day.The six banks of solenoids operating the five loops

and the "noon transfer" system are controlled by aclock-regulated unit which arranges for the total lengthof the eastern feeder to be decreased in 15-m steps be-fore noon, and then the total length of the westernfeeder to be similarly increased after noon. By thismeans the quantity a sin 0+1 in (1) is maintained in thevicinity of 60 m and hence the frequency interval Af inthe vicinity of 5 mc.

C. Calibration of the Antenna Feeder SystemOne of the major technical problems encountered in

making accurate position measurements with the inter-ferometer is the determination of the electrical lengthsof the antenna feeder system including the variable line.Since the switches in the variable line inevitably in-troduce some discontinuities, it is necessary to investi-gate carefully the consequent variations in electricallength with frequency. To measure apparent positionswith an accuracy of ± 1' arc (±3 X 10-4 radians), it isdesirable to measure all relative electrical lengths tobetter than one part in 3 X 103, or + 30 cm in our case.

17 Each switch consists of a thin strip of phosphor bronze, oneend of which is held in a plastic clamp; the other end contains smallrhodium-plated contacts which bear against similar contacts inplastic mounts. The arm is linked through a long rigid plastic arm toits solenoid.

RECEIVER

AERIAL THROUGH FURTHERLOOPS OF LENGTH60,120 & 240 METRES

CONTROCUNIT

Fig. 7-Diagrammatic plan view of two sections of the variable line,in which only one of the two wires of each radio frequency line isseen. The complete line can be varied from 0 to 480 m in steps of15 m.

Also, it will be shown that the measurement of sourcesize requires that the relative attenuation of the pairof feeders be known as a function pf frequency.The lines were calibrated by recording, as a function

of frequency, the impedance at the receiver terminalsof each line with the antenna disconnected from the line.Let us assume initially that the line is uniform and hasa characteristic impedance Z0. The impedance, ZR(f),at the receiver terminals is then

ZR(f) = Zo coth {aL + aT + i (IL + IT)4, (2)

where aL and IL denote the attenuation and electricallength of the line being measured, and where the ter-mination of the line (nominally an open circuit) is repre-sented by a short line of length IT and attenuation aT.In the general case, the attenuation and length param-eters vary slowly with frequency.

In practice the attenuation per wavelength is ex-tremely small, and so it follows from (2) that as thefrequency is increased, ZR(f) oscillates between extremevalues

(ZR)max = Zo coth (aL + aT)and

(ZR)min = Zo tan h(ajL + aT),

and these occur at frequenciesnc

IL + IT

and

(n +j)cfn+ll2 = -

2

IL + IT

(3)

(4)

(5)

respectively, where n assumes integral values.The phase and amplitude of the pattern of ZR(f)

1958 165

PROCEEDINGS OF THE IRE

Fig. 8-A small section of an actual record of impedance at the re-ceiver terminals (vertical vs frequency using the system describedin the text for measuring electrical length and attenuation.

[see (2)] is measured by connecting the line across theterminals of the 40-70 mc swept-frequency receiver inparallel with a noise diode which acts as a current sourceof broad-band noise. The receiver includes a narrow-band (6-kc) IF amplifier, and the detected output isdisptayed on a pen recorder. As the receiver frequencyis mechanically swept slowly across its range, an oscil-latory pattern in phase with Zn(f) is traced on the re-corder."8 A typical pattern is shown in Fig. 8. The phaseis accurately read using frequency markers generatedfrom harmonics of a crystal oscillator, and the electricallength IL+lr may be determined at many frequenciesacross the range using (5). Also the attenuation aL+aTis determined by measuring the ratio (ZR)inx/(ZR).inwith the aid of a series of calibration runs in which thenetwork is replaced by several resistors of known value;it is calculated from (3) and (4), which give

/ (ZR)matanh(aL + aT) = Al -

V(ZR)mi.The measurements are then repeated for the line be-longing to the other arm of the interferometer, which isterminated by an identical terminating impedance. Letthe attenuation and length of the second line be aL' andIL', respectively. As before, we determine IL'+lT andaL+aT as functions of frequency, and so obtain therequired relative length l-lIL' and relative attenuationaL-a Ll.A sample set of length measurements (IL+IT) is

shown in Fig. 9. It is seen that the derived lengths atdifferent frequencies are scattered about a mean valuewith a standard deviation of about 10 cm. This scatterincludes the effects of the nonuniformity of the line(e.g., those due to switches in the variable line). Work-ing within the required limits of + 30 cm, we areclearly justified in regarding the line as a uniform, non-dispersive line whose length is given by the average atdifferent frequencies.A sample set of attenuation measurements is shown in

Fig. 10. It is found that, within limits of about ± 2 db(sufficient for source-size determinations), the measuredattenuations can be adequately represented by a linearfunction of frequency.

is A small additional phase shift may occur in the receivingsystem. This effect tends to be cancelled when we come to measurediferences in length between a pair of line networks.

a

I

Z1

_J

id

-.

AA AA

127940 45 so 55 60 65 70

FREQUENCY (Mci)Fig. 9-A typical measurement of electrical length (IL+lr) as a

function of frequency. (This example refers to the west feeder withthe variable line set to 450 m.)

KC

4j

0

3zQ

zI.-

7

6

West feeder with vwiablelne set at 465 metres. A

AAW vA

A~~~

0

West feeder without/vriable line

49

,_So 60

FREQUENCY (Mc/s.)70

S

40

Fig. 10-Two sample measurements of attenuation(aT+aL) as a function of frequency.

IV. THEORY OF MEASUREMENTS

A. The Measurement of Positions on the Sun's DiskWe now discuss the method of locating sources on the

sun's disk with an E-W swept-frequency multiplyinginterferometer. The first step is to use the records tocalculate the angle t between the incoming rays andthe normal plane of the interferometer (see Fig. 1) tak-ing into account phase shifts in the receiver. The secon(dis to relate 0 with positions on the sun's disk.

31 E . ._.

January166

I

I

4

Wild and Sheridan: A Swept-Frequency Interferometer for Solar Radiation

1) A Determination of the Angle 0: The phase of therecorded interference pattern is determined by the dif-ference in phase of the two signals reaching the mixingpoint in the channel mixer (see Fig. 4). At this point letus represent the voltage waveform of the signal passingthrough channel A as

EA = | EA sin 2wrfot,where fo is the intermediate frequency. Then that of thesignal through channel B will be of the form

/ 7~~~~~r?r\EB =EB I sin 2rfot +cs +9L+R + -+ )

where the O's denote phase terms introduced by thedifference in path length through the two channels inspace (bs), in the antenna feeder lines (OL) and in thereceiver circuits (OR). The term "7r/2 + 7r/2" is includedto take account of the two positions of the phase switch.Denoting the two switched values of EB by EB+ andEB-, the output of the phase-sensitive detector is pro-portional to the time average

(EA + EB )2 - (EA + EB+)2= 41 EAI I EB I sin 2rfot sin (27rfot+ s + XL + 'OR)= 21 EA IIEBI cos (Os + FL + OR). (6)The phase terms k. and FL are given by

cs =27r(a/X) sin 0

FL = 27r(l/X)

where I is the length by which the western antennafeeder line exceeds the eastern, and a and 0 are definedin Fig. 1. Similarly, we write

OR = 27r{x(X)/X},where x(X) is the difference in path length introducedby the two receivers. Substituting in (6) we obtain theswept-frequency interference pattern

21 EA I I EB cos- { a sin 0 + I + x(X)}.x

The wavelengths at which the interference patterncrosses the zero line occur at X,, given by

asin 0 + I + x(Xk) = 2(n + 1)X., (7)

where n takes integral values, and is even or odd accord-ing to the sign of the slope of the pattern as it crosses thezero line.To determine the function x(X) at a number of wave-

lengths, we replace the antenna system by the artificialsource (noise generator) (see Fig. 4). This is connectedto receivers by a pair of cables differing by a knownlength, b. We then obtain a second interference patternwhose zero points are given, by analogy with (7), by

b + X(X)m) - 4(m + 1)X?m,X

where m takes integral values.

The electrical length of I in (7) is measured by themethod previously described and so sin 0 is calculated.Providing that neighboring wavelengths X., X.+, arrivefrom approximately the same direction, no ambiguityin choosing n will arise.

2) Conversion of 0 to Solar Disk Coordinates: Theposition of a source in the celestial sphere may be speci-fied by the usual astronomical coordinates-hour angle,H, and declination, S. These are related to the measuredquantity.

S = sin 0

by the well-known expression for an E-W interfer-ometer'9

S = sin H cos 5. (8)Let the position of a source on the sun be referred to thecenter (So; Ho, bo) of the disk in the form (Ho+AH,So+AS). Then, with sufficient accuracy, the small dis-placements AH and AS cause the value of S to become

S = So + AS,

where

aS\ /as\AS =1 AH + -JA6.

aH 0 as i (9)

Thus for a given value of AS there is a family of possiblepositions of the source which lie on a straight line whoseequation, in the (AH, AS) plane, is (9). Let us specifythis line in terms of its (angular) distance, r, from thecenter of the sun and its inclination, i,, with the H axis(see Fig. 11)-, i.e., in the form

r = cos AH - sin ,6AS. (10)

Comparing (9) and (10) we obtain

tan*it = - --1 / as !

and

AS

Vkaa ) j+ WaH )

(11)

(12)

Substituting for S in (11) and (12), using (8), we obtainthe values of r and ,t in terms of the measured quantityS=sin 0, and the sun's position (available from theNautical Almanac):

sin 0 - sin Ho cos Sor7=

V\/ (cos 2Ho cos 26o + 1)i1 = tan-' (tan Ho tan So).

(13)

(14)

19 The treatment given here is for a horizontal E-W interferometer.The actual interferometer axis lies in a surveyed E-W vertical planebut is tilted a small angle a, the W antenna being higher than the E.In this case (8) becomes S= cos a sinHcos 5+sin a(cos H cosO cos 6+ sin 8 sin 4t), where aI is the latitude of the observing station.

1671958

PROCEEDINGS OF THE IRE

s

Fig. 11-Aspect of the sun as seen at noon fromthe southern hemisphere.

(a)B. The Measurement of Source SizeThe method of measuring the angular size of a source

with an adding swept-frequency interferometer is closelyanalogous to the monochromatic method originallyused in optical astronomy by Michelson and in radioastronomy by McCready, Pawsey, and Payne-Scott:'we simply measure the ratio of the maximum-to-mini-mum power in the interference pattern and hence cal-culate the half-power width of the source assuming aplausible shape of the source. The half-power widthdoes not depend critically on the assumed shape if thelatter is a smooth, single-humped curve. Here we shallassume the shape to be one period of a sine wave, suchthat the line-integrated brightness, B(0'), at an angle 0'measured along the scanning direction is given by

B(')=A (1+ cos-) .. ... (15)

B(0') = 0 .............'.......... < A

where A is the half-power angular width of the source.The ratio Pmax,/Pmin of the maximum-to-minimum

power in the interference pattern at any desired fre-quency20 is obtained from the records by drawing theupper and lower envelopes to the fringe pattern [Fig.12(b)]. These are calibrated using special calibrationrecords in which a series of six power levels, each sepa-rated by a 2:1 interval, is generated with a diode noisegenerator. The source-size calculation also depends onthe relative attenuation, a, of the feeder lines which ismeasured for different settings of the variable line bythe technique previously described.The dependence of source size upon Pmax/Pmin and a

is derived below. (Previous treatments describing theanalogous monochromatic method have neglected theeffects of attenuation which are important here.

Let the attenuation, expressed as a power ratio, alongthe two feeder lines be kA and kB, respectively [Fig.

20 In this section all quantities relating to the source [i.e., B(4/),A, A, Pm., Pmin, S and v] are functions of frequency. So also arethe instrumental quantities kA, kB and a.

(b)Fig. 12-Defining the symbols required for

source-size measurements.

12(a)]. The power received from a point source of fluxdensity S at position angle 00 is then proportional to thetime average

[(kAS)1/2 COS 1 2irft + 7r(f/c) (a sin 00 + 1)}+ (kBS)'12 cos {2rft- r(f/c)(a sin 0o + 1)} ]2

(kA + kB)S + (kAkB) "2S COS {2r(f/c)(a sin 0o + 1)JSimilarly if the point source is replaced by a source ofsmall but finite angular size having a line-integratedbrightness B(0') at frequency f, the received power isproportional to

00

2 (kA + kB) f B(o')d0'_x

+ (kAkB) 1/2 B(0') cos [27r(f/c) J a sin (Go + 0') + I} ]de'_00

(kA + kB) f B(0')dO'_00

+ (kAkB)1 12 f B(0') cos 127r(f/c) (0' cos Oo

+ a sin Oo+ l) } d&',

where B(0') is taken as nonzero only for small values of0'. Asf is varied, the above expression defines the swept-frequency interference pattern. It is simple to show that:

168 January

Wild and Sheridan: A Swept-Frequency Interferometer for Solar Radiation

1) The spacing between neighboring maxima is

a sin 0o + IAf=

provided that B(0') does not change appreciably withfrequency in the interval Af, and

2) The maximum and minimum envelopes of theinterference pattern due to a symmetrical source are pro-portional to

00

2(kA + kB) f B(0')d0'_00

± (kA + kB) 1/2f B(0') cos i 2ra(f/c)O' cos 0o } dO'.-00

Denoting these envelopes by Pmax, and Pmin, we notethat the visibility, {, of the fringes, defined by

Pmax - Pmin

Pmax - Pmillis glven, by

2(kAkB) 112

kA+ kBrxf_B(0') cos { 2T0'(a/X) cos 00 } dO'

_00~~~~0B(01)d01

Thus t is just the ratio of the moduli of the Fourier com-ponents of the source distribution at angular "fre-quencies" (a/X) cos Oo and 0, with a factor of proportion-ality determined by the attenuation of the lines. If therelative attenuation is ai nepers (a = Ilog. (kA/kB)), thisfactor becomes

2ea-= sech a.

e2a+ 1

It remains to substitute the assumed source function(15), whence we obtain

Pma.x/Pmin- 1 sech a sin 27rvPmax/Pmin + 1 2rvw(1 - 4V2)

where v =A(a/X) cos Oo.A plot of Pmax/Pmin vs v for different values of a is

given in Fig. 13. The conversion of v to the source sizeA for the interferometer a-t Dapto is also given in thisfigure. It is seen that numerical values of source sizecan be assessed in the approximate range 1 to 20 min-utes of arc. Outside this range upper and lower limitscan be derived.

C. The Measurement of PolarizationPolarization measurements are made by replacing the

pair of spaced antennas of the interferometer by a pairof antennas mounted on a common axis, but with

V. a AG COS e

CEQ.a

V.A,gA cose

(a. cos 1-02 Km)

Fig. 13-The chart used for calculating sources size from the maxi-mum-to-minimum power ratio (Pma=/Pmin) in the interferencepattern, and the relative attenuation of the feeders in decibles.The upper part of this figure is quite general, while the lower partcontains parameters peculiar to the interferometer at Dapto.

mutually perpendicular planes of polarization. Theantennas are connected to the adding and multiplyingreceiving systems through feeders differing by a knownlength b( = 60 m) as shown in Fig. 4.The signals generated in the two antennas by polar-

ized radiation incident upon them will in general con-tain phase-coherent components. These components

1958 169

PROCEEDINGS OF TIIE IRE

will interfere with one another at the mixing point and,as the frequency is swept, will generate an interferencepattern. With the multiplying system, this pattern is[cf. (6)]

2orfbCOS t~ + q5P + ORJ

where q5p is the difference in phase between the two re-ceived polarization components of the radiation, andqSg is the relative phase change introduced in the receiv-ers. The latter term is determined in the calibrationprocess and so the polarization phase, Op, is determined(at 2.5-mc frequency intervals) from the frequencies ofthe null points in the interference pattern.

Past observations have shown that polarized radia-tion from the sun is mainly circular. In this case q5 =± wr/2, the sign depending on the sense of rotation.Circular polarization can therefore be directly recog-nized and its sense determined from measurements ofq5p; also the degree of circular polarization is given bythe depth of modulation of the pattern of the addinginterferometer.When the phase Pp indicates noncircular polarization

the problem is more complicated. Linear polarizationgives 'p = 0 or ir/2, but the depth of modulation de-pends both on the angle between the planes of polariza-tion of the antenna and the radiation, and on the degreeof polarization. Intermediate values of 'p correspondto elliptical polarization. In general, however, it issimple to show2" that the complete polarization char-acteristics (i.e., the percentage, eccentricity, orienta-tion, and sense-of-rotation of elliptical polarization)can be determined from measurements of the following:

1) The phase, 'p, of the polarization (given accuratelyby the multiplying interferometer with crossed antennasystem).

2) The maximum and minimum power in the polari-zation interference pattern (given by the adding inter-ferometer with crossed antenna system).

3) The total power received in one polarization (givenby the adding interferometer with the spaced antennasystem).

Polarization observations are made at selected timesby operating an antenna switch (S2, shown in Fig. 4)which replaces the spaced-antenna system by thecrossed-antenna system. A sample record in Fig. 6shows strong circular polarization in storm radiation(spectral Type I).

V. PRELIMINARY RESULTS

Trial observations with the interferometer were madedaily between June 6 and July 6, 1957, and during thisperiod numerous bursts and storms of solar radio emis-sion were recorded. Analysis of these records is incom-

21 J. L. Pawsey and R. N. Bracewell, 'Radio Astronomy,"

Clarendon Press, Oxford, Eng., pp. 59-62; 1955.

plete at the time of writing. The preliminary resultsgiven below are primarily measurements of the angularsize of Type I and Type III sources recorded on two se-lected days.

A. Observations of a Type I Storm-June 20, 1957, 11:30to 12:30 E.A.S.T.2

Parts of the record are reproduced in Fig. 6, and theanalysis is summarized in Fig. 14. The storm was char-acterized by an unusually intense background con-tinuum in which the flux density attained a value of1.5 X 10-19 W m-2 (c/s)-v. The radiation was stronglypolarized (right-handed circular).At 12:08 the source was located on the line AA' in

Fig. 14(a).23 It was not possible to decide which of twolarge sunspot groups, one in the northern and one in thesouthern hemisphere, was associated with the source.However a further measurement made at 13:46 whenthe scanning angle had changed by 112° favored thenorthern group.

Measurements of source size at 55 mc at intervals ofabout one minute are shown in Fig. 14(b). The frequencydependence of source size, averaged over the one-hourperiod, is shown in Fig. 14(c).The mean size is seen to decrease from about 10' arc

at 45 mc to 6' arc at 65 mc.The maximum instantaneous brightness temperature

in this period was estimated to be 2.3 X 10100 K.

B. Observations of Type III Bursts-July 3, 1957, 10:15to 12:30 E.A.S.T.

In this period 26 individual bursts, or small groups ofbursts, were analyzed (Fig. 15). Records of two are re-produced in Fig. 5. Fig. 15(a) shows the location of aburst at 10:39 and indicates an origin above a largespot near a flare in progress.

Detailed source-size measurements indicated that atany one frequency the size remained approximatelyconstant in the course of the rise and fall of an individ-ual burst. Fig. 15(b) shows the distribution of sourcesize at 55 mc, for the 26 bursts, and Fig. 15(c) the aver-age size as a function of frequency. There appears to beno marked difference between the angular size of TypeI and Type III emissions.The maximum instantaneous brightness temper-

ature in this period was estimated to be 10110 K.The above results show that both Type I and Type

III sources cover a remarkably large part of the sun'sdisk frequently exceeding one tenth of the optical disk ifcircular symmetry is assumed. The result that the sizeincreases with decreasing frequency suggests that theemitting region in the solar corona could be funnel-shaped, the diameter increasing with height. It should

22 Eastern Australian Standard Time is 10 hours in advance ofUniversal Time.

28 The sunspot pictures and flare observation in Figs. 14 and 15were kindly made available by the Division of Physics, CSIRO,through the courtesy of Dr. R. G. Giovanelli.

170 January

Wild and Sheridan: A Swept-Frequency Interferometer for Solar Radiation

20 JUNE 1957 TYPE I STORM14r

12

I ;It 0 10

r 1 4~~~~~6In4

A' IW

I E E

.ClRCL; ,,, <

/ b

12h 07m40s CtPOLARIZATION DR.H. CIRCULAR Q 2

(a)

(l30

55 Mc/s.12r

ub,io

2 80w

a2

I)INtn4

1D02

1200 1230

LOCAL STANDARD TIME (150°EAST)(b)

45 su Ss

FREQUENCY(c)

Fig. 14-Observations of a Type I storm. (a) A pair of instantaneous position measurements, averaged overfrequency, (b) the variation with time of the apparent source size at 55 mc, (c) the average spectrum

of source size over the period in (b).

3 JULY 1957: TYPE m BURSTS12r

V)0 48.L00' 2-z

Of

55 Mc/s

0 1 2 3 4 5 6 7 8 9 D 11

SOURCE SIZE AG (minutes of arc)(b)

12

u

-lo0

4.'

w

-9

4-

0.- ..

K

,- -40 45 SO 55 60

FRECQUENCY (Mc/s)(c)

Fig. 15-Observations of Type III bursts. (a) An instantaneous position measurement, averaged over frequency,(b) histogram showing the distribution of source size in the 26 bursts (55 mc), (c) the average

spectrum of source size during the bursts referred to in (b).

be pointed out, however, that the size we measure doesnot necessarily correspond to that of the actual emittingregion since refraction and scattering in the solar at-mosphere could disguise its true form. Further measure-

ments of sources at different positions on the sun's diskshould help to resolve this question.

VI. ACKNOWLEDGMENTThe authors wish to thank Dr. J. A. Roberts and

M. M. Komesaroff who participated in the initial phasesof this work. Much of the variable line assembly and its

control unit was constructed by Mr. Komesaroff. Theauthors also are indebted to G. H. Trent for valuableassistance in construction of equipment, taking of ob-servations, and analysis of results, and to J. Joisce whoassisted in the observing program. Considerable helpwas derived from H. Rabben, Fraunhoffer Institute,Freiburg-im-Breisgau, during an extended visit toDapto field station. Finally, thanks are due to Dr. J. L.Pawsey, Director, CSIRO Radiophysics Laboratory'sradio. astronomy group, for his interest in this work,and Dr. Roberts for helpful criticisms of the manuscript.

60

(Mc/s.)65 70

Ew

(a)65 70

u1-ArN A e t 1

I

1958 171

Wl

-T -

A40