RADIO TELESCOPES OF LARGE RESOLVING

POWER
Nobel Lecture, December 12, 1974

by M A R T I N R Y L E
University of Cambridge, Cavendish Laboratory, Cambridge, England

I think that the event which, more than anything else, led me to the search
for ways of making more powerful radio telescopes, was the recognition, in
1952, that the intense source in the constellation of Cygnus was a distant
galaxy - 1000 million light years away. This discovery showed that some
galaxies were capable of producing radio emission about a million times more
intense than that from our own Galaxy or the Andromeda nebula, and the
mechanisms responsible were quite unknown. It seemed quite likely that some
of the weaker sources already detected with the small radio telescopes then
available might be similar in character; if so they would be at distances
comparable with the limits of observation of the largest optical telescopes. It
was therefore possible that more powerful radio telescopes might eventually
provide the best way of distinguishing between different cosmological models.
It was not until 1958 (1) that it could be shown with some certainty that most
of the sources were indeed powerful extragalactic objects, but the possibilities
were so exciting even in 1952 that my colleagues and I set about the task of
designing instruments capable of extending the observations to weaker and
weaker sources, and of exploring their internal structure.
The early observations were severely limited both by the poor angular re-
solution and by the limited sensitivity. It was usually impossible to obtain any
information about the structure of a source, and adjacent sources could often
not be properly separated, whilst attempts to relate the radio sources to
optically visible objects were often prevented by the poor positional accuracy.
The use of interferometers allowed better positions to be obtained, and some-
times made it possible to derive simple models for the source structure. Few
of the sources were found to have an angular size greater than 2-3 minutes
of arc.
The problem of making detailed maps of such sources arises simply from
the fact that the wavelengths used are some million times greater than optical
wavelengths - so that even to obtain a radio picture with the same resolution
as that of the unaided human eye ( ~ 1’ arc) we would need a telescope
having a diameter of about 1 km operating at a wavelength of 50 cm. At the
same time the instrument will be effective only if the surface accuracy is good
enough to make a proper image, corresponding to errors of < l/20 or a few
cm; the engineering problems of building such an instrument are clearly
enormous.
With the development, around 1960, of masers and parametric amplifiers
capable of providing receiving systems of good sensitivity at wavelengths of
a few cm, it became possible to build telescopes of diameter 10-100 m with
 Physics 1974

(a) (b) (c)

Fig. 1. The use of (a) a paraboloid, (b) an array of dipoles or (c) the sequential
sampling of the wavefront by small aerial elements to achieve a high resolving power
by combining the signals from a large part of the incident wavefront.

sufficient sensitivity and with angular resolutions of ~ 1’ arc; even with

such instruments the engineering problems of constructing a rigid enough
surface are considerable, and it is likely to be difficult to build a conventional
paraboloid capable of angular resolutions much better than 1’ arc.
I would like now to describe an entirely different approach to the problem
in which small aerial elements are moved to occupy successively the whole of
a much larger aperture plane. The development and use of “aperture
synthesis” systems has occupied much of our team in Cambridge over the
past 20 years.
The principle of the method is extremely simple. In all methods used to
obtain a large resolving power, that is to distinguish the wavefront from a
particular direction and ignore those from adjacent directions, we arrange to
combine the field measured over as large an area as possible of the wavefront.
In a paraboloid we do this by providing a suitably shaped reflecting surface,
so that the fields incident on different parts of the sampled wavefront are
combined at the focus (Fig. 1 (a)); the voltage produced in the receiving
dipole represents the sum of these fields. We can achieve the same result if
we use an array of dipoles connected together through equal lengths of cable
(Fig. 1 (b)).
Suppose now that only a small part of the wavefront is sampled, but that
different parts are sampled in turn (Fig. 1 (c)); could we combine these
signals to produce the same effect ? Since in general, we do not know the
phase of the incident field at different times this would not normally be pos-
sible but if we continue to measure one of the samples while we measure the
others we can use the signal from this one as a phase reference to correct the
values measured in other parts of the wavefront. In this way, by using two
small aerial elements, we can again add the fields over the wavefront-the
area of which is now determined by the range of relative positions taken by
the two aerial elements.
It might be thought that this method would be extremely slow, for if we
are to sample an area of side D using elements of side d, it is necessary to
2 Da
observe with - 3 different relative positions of the two aerial elements. In
d
 M. Ryle 189

practice, however, the method is not significantly slower than the use of the
large equivalent instrument for although a large number of observations must
be made, the results may be combined in a computer using additional phase
differences, which correspond to many different wave directions (as in a
phased array or dipoles), so that with the one set of observations an area of
sky may be mapped which is limited only by the diffraction pattern of the
D2
small elements themselves; there are in fact some -2 different directions
d
which can be scanned in this way, and which would have had to be explored
sequentially by a conventional instrument, so that the total observing time of
the two methods is nearly the same.
It can also be seen that the sensitivity of the system is much better than
would be associated with the small elements, for the signal from a particular
point in the sky is contributing to that point on the map for the whole ob-
serving period; the resulting signal-to-noise is in fact equivalent to the use of
2 2 D2
an instrument having a collecting area 2d √ d2

~ 3Dd, a figure which may be much greater than that of the elements them-
selves, and although it is not as great as if the full instrument of area D2 had
been built, it may exceed that of any instrument which can be built.
Unlike a paraboloid or array, in which both the sensitivity and resolving
power are fixed as soon as the wavelength is decided, the value of d may be
chosen so that the sensitivity, for any particular wavelength and type of ob-
servation, is matched to the resolution.
The method of aperture synthesis avoids the severe structural problems of
building very large and accurate paraboloids or arrays, and allows both high
resolving power and large effective collecting area to be obtained with a
minimum of engineering structure and therefore cost. Provision must be made
for the relative movement of the small elements, and their relative positions
and electrical connecting paths must be known with an accuracy equal to the
surface accuracy of the equivalent instrument (≤ λ/20). Automatic comput-
ing is needed to carry out the Fourier inversion involved in combining the
observations to provide a map of the sky.
Historically, the forerunners fo this type of instrument were realized in the
early days when observations in both Australia and England with aerial
elements having a range of separations were used to determine the distribu-
tion of radio brightness across the solar disc. In the earliest observations the
Sun was assumed to show spherical symmetry, and no measurements of phase
were necessary so that a precise knowledge of the relative positions of the
elements, and of the electrical path lengths to the receiver were unnecessary.
A similar technique was used to establish the profile of radio brightness across
the plane of the Galaxy (2).
The first synthesis instrument capable of mapping an abritrary distribution
of sources was built at Cambridge in 1954 by John Blythe (3); it consisted of
a long thin element covering, in effect, a whole row of Fig. 1 (c) used in con-
190 Physics 1974

Fig. 2. (a) The arrangement used in the instrument built in 1954 by J. H. Bythe.
(b) The equivalent instrument using two small elements.

junction with a smaller element moved to 38 different positions along a

perpendicular line (Fig. 2 (a) ) to synthesise a square instrument giving a reso-
lution of 2O.2. This instrument provided the first detailed maps of the galactic
emission at a long radio wavelength (7.9 m).
Larger instruments using this same configuration were built at Cambridge
during the succeeding years, including an instrument of high sensitivity and
45’ arc resolution also at λ = 7.9 m (4) and a second operating at λ = 1.7 m
with 25’ arc resolution which was used by Paul Scott and others to locate
nearly 5000 sources in the northern sky (5, 6).
These instruments used a very cheap form of construction; for λ> 1 m an
efficient reflecting surface may be provided by thin (~ 1 mm diameter)
wires 5-10 cm apart. In the case of the λ = 1.7 m instrument, wires stretched
across simple parabolic frames of welded steel tube provided a cylindrical
paraboloid 450 m long and 20 m wide (Fig. 3) at a cost of about £2 per m2.

Fig. 3. Photograph of the east-west arm of the λ = 1.7 m instrument built in 1957
with which nearly 5000 sources were located.
 M. Ryle 191

la) (b)
Fig. 4. (a) Two aerial elements mounted near the North Pole observing throughout the
day are equivalent to one ring of a much larger instrument.
(b) The elements may be used at other latitudes if arranged on an east-west line and
used to track the chosen point for 12 h.

With the need for still greater resolving power, we realized that physically
larger systems operating at metre wavelengths would no longer prove suc-
cessful, because of the limitation imposed by irregularities of electron density
in the ionosphere. But at shorter wavelengths where these are unimportant it
becomes difficult to make efficient reflectors by using stretched wires, both
because of their deflection by the wind, and because with the closer spacing
needed there is difficulty with them twisting together. For operating wave-
lengths of < 50 cm a much more rigid supporting structure must be used,
and the engineering costs of building a long element become very great.
The obvious solution is to use the system illustrated in Fig. 1 (c), in which
the engineering structure is confined to two small elements-where much
higher costs per m2 are acceptable. The method for altering the relative posi-
tions of the two elements presents some practical problems; suppose that the
elements are mounted on two railway tracks at right angles (Fig. 2 (b) ), so
that for each position of A on the N-S track B is moved to every position
D
along the E-W track. For values of - ~ 50, there are then 5000 different
d
arrangements, and if B is moved each day, the observations will take 5000
days and although a map will then be available for the whole strip of sky,
the period is too long for a graduate student’s thesis!
Alternatively B could be moved rapidly - so that several positions could be
fitted into the time during which the area of sky remains in the beam of the
small elements. This will reduce the total time of the observations, at the
expense of observing only parts of the strip of sky. We can clearly extend this
period, and so allow more relative positions of A and B each day, if we arrange
for the elements to track the chosen point in the sky for an extended period.
As soon as we do this, we realize that the rotation of the earth is itself
providing us with a relative motion of A and B as seen from the source,
without our having to move them on the surface of the earth at all. Suppose,
for example, we have our two elements mounted near the North Pole and we
use them to observe an area of sky centred on the Celestial Pole; in this case
we do not even have to arrange for them to track. Over a 24 h period, one will
have traced out a circular path about the other (Fig. 4 (a) ), and the signals
192 Physics 1974

recorded during this time can be combined to provide the same response as
that of the equivalent ring aerial; by simply altering the separation along a
line on say 50 successive days a complete aperture can then be synthesized.
Miss Ann Neville and I set up an experimental system in 1960-61 to test
the method and develop the computing; we used it to map a region 8’ in
diameter round the North Celestial Pole at a wavelength of 1.7 m (7). We
connected up different 14 m sections of the long cylindrical paraboloid (Fig.
3) with some other small aerials to simulate the use of two 14 m diameter
elements at different spacings. The effective diameter of the synthesized in-
strument was 1 km and it provided an angular resolution of 4’.5 arc.
As well as showing that the method really worked, it provided some inter-
esting astronomical results - in particular by allowing the detection of sources
some 8 times weaker than had been observed before; even though the area
of sky covered was only some 50 square degrees the results were useful in our
cosmological investigations.
In practice only 12h observations are needed because of the symmetry of the
system and observations need not be made from the North Pole or limited to
the Celestial Pole, provided that the elements are situated on an East-West
axis, and each is able to track the required region of sky for 12h (Fig. 4 (b) ) .
At low declinations the synthesized instrument becomes elliptical with the
north-south aperture reduced by sin δ. The engineering simplicity of moving
the elements along a line, and the consequent great saving in the area of
land needed are, however, such great advantages that we eventually built
three large instruments in Cambridge with equivalent instrumental diameters
of 0.8, 1.6 and nearly 5 km.
These instruments are known as the Half-Mile, the One-Mile and (because
its construction coincided with the early negotiations for the entry of Britain
into the European Community), the 5 km Telescopes! The One-Mile tele-
scope was the first to be built, and this started observations in 1964.
It is interesting that as early as 1954 we had discussed the possibility of
building a high resolution instrument on exactly these principles, and I have
recently found two entries in an old note-book: -

“8.6.1954 Possible research student and other projects.

. . . 3(f). North Polar Survey on 81.5 MC / S . Effective gain area ~ 25 X
1500=37,500 sq. ft. Effective resolving power area N 106 sq. ft.”
(The entry included a diagram of the proposed aerial element)

"29.6.1954
Do 3(f) in all directions where 180’ rotation available? above about 20’
might be possible by directing aerials in successively different directions-
i.e. observation not on meridian.”
A third entry on 22.7.1954 discusses the east-west rail track to be used
for the latter programme with two 30 ft aerials mounted on it, the arrange-
ment of cabling needed to compensate for the different path lengths to the
two aerials when observing off the meridian, and the selection of directions
of observation “to give uniform cover of Fourier terms”.
 M. Kyle 193

Why then, with its obvious simplicity and economy, did we not build this
instrument in 1954? The answer is that at this time there were no computers
with sufficient speed and storage capacity to do the Fourier inversion of the
data. EDSAC I, which was the first stored-programme computer, was built
by Dr. M. V. Wilkes of the Cambridge University Mathematical Laboratory,
and came into operation in 1949. It was used for reducing John Blythe’s ob-
servations and took some 15h of computing to do the 38-point transform for
every 4 m of the 24h scan of the sky. It would not have been practicable to use
it for the 2-dimensional inversion needed for the earth-rotation synthesis. By
1958 the completion of the much faster EDSAC II, and the development by
Dr. David Wheeler of the Mathematical Laboratory of the fast fourier trans-
form (incidentally some six years before these methods came into general
use) made possible the efficient reduction of the 7.9 m and 1.7 m surveys,
and also enabled the trials of the 1.7 m earth-rotation synthesis to be made in
1961; even with EDSAC II, however, the reduction for the small area of sky
covered in the latter survey took the whole night.
During the early stages in the design of the One-Mile telescope in 1961, I
discussed with Maurice Wilkes the considerably greater problems of reducing
the data from this instrument, but by then the replacement of EDSAC II was
planned and the new TITAN computer, which came into operation in 1963,
was easily capable of dealing with the output of the One-Mile telescope. The
development of aperture synthesis has therefore been very closely linked to the
development of more and more powerful computers, and it is interesting to
speculate how our work in Cambridge would have proceeded if, for example,
computer development had been five years behind its actual course.
The two programmes in my 1954 note-book subsequently formed the basis
of two Ph.D. theses in 1964 and 1965.
Now I return to the design of the large instruments whose layout is shown
in Fig. 5. The One-Mile telescope consists of three 18 m dishes, two fixed at
0.8 km spacing, the third mounted on a 0.8 km rail-track (Fig. 6); this ar-

Fig. 5. Sketch map showing the arrangement of One-Mile, Half-Mile and 5 km

telescopes.
 Physics 1974

Fig. 6. The One-Mile telescope, showing the west, railmounted, dish in the foreground,
with the two fixed dishes behind.

rangement was cheaper than building the longer rail track and it also provided
two spacings at a time. It was designed for two main programmes: (a) The
detection of much fainter and therefore more distant sources (see Fig. 7) in
order to explore the early history of the Universe, and so try and distinguish
between different cosmological models, and (b) To make radio maps of in-
dividual sources in an attempt to understand the physical mechanisms within
them; most of the sources studied have been powerful extragalactic objects,
but the remnants of supernova explosions are perhaps physically as important.
 M. Ryle 195

Fig. 7. Map obtained with the One-Mile telescope showing sources about 100 times
fainter than had been observed before.

The problem of the physics of radio galaxies and quasars and the cosmo-
logical problem are strangely linked; we appear to be living in an evolving
Universe, so that very distant sources which, due to the signal travel time, we
observe as they were when the Universe was younger, may be systematically
different from a sample of nearby sources. But the intrinsically most powerful
sources are so rare that there are no nearby ones, whilst the weak sources can-
not be detected at great distances. If we are to understand how the Universe
is evolving, we may first have to solve the physical problem of the individual
source - so that we can infer the differences in its evolution at earlier cosmo-
logical epochs.
The Half-Mile telescope was built later by John Shakeshaft and John
Baldwin. It was actually built very cheaply because as can be seen from Fig. 5,
it made use of the same rail track, and we were able to get the four 9 m
dishes at scrap-metal prices from a discontinued radio link service, and only
the mounts had to be built. It has been used mainly with a radio spectrometer
covering the 21 cm wavelength band of neutral hydrogen to map the distri-
bution of density and velocity of the hydrogen in a number of nearby galaxies,
196 Physics 1974

Fig. 8. The 5 km telescope, with the movable dishes in the foreground.

and forms part of a programme concerning the formation and evolution of

galaxies.
The 5 km telescope was completed in 1971, and because it represents a
rather more advanced design I will describe it in more detail. It was designed
solely for the purpose of mapping individual sources, and besides its larger
overall size, the individual dishes are more accurate to allow operation at
wavelengths as short as 2 cm. As a result the angular resolution is ~ 1” arc,
a figure comparable with the resolution attained by large optical telescopes
on good mountain sites. It is at present being used on a wavelength of 6 cm,
where the resolution is 2” arc.
In order to improve the speed of observation, four fixed and four movable
elements mounted on a rail-track are used, as shown in Fig. 5; this arrange-
ment provides 16 spacings simultaneously, and a single 12 h observation
produces a 2” arc main response with circular grating responses separated by
42” arc. Sources of smaller extent than 42” arc can therefore be mapped with
a single 12h observation; more extensive fields of view require further observa-
tions with intermediate positions of the movable elements to suppress the
grating responses.
For operation at these short wavelengths the positioning of the elements,
and the electrical cable connections, must be stable and measured with an
accuracy better than 1 mm. Conventional surveying methods allowed each
element to be located to ± 10 mm, and the final alignment had to be based
entirely on radio observations; the distance between the two outer fixed
elements (on which the scale of declination is based) was found in this way
 M. Ryle 197

to be 3430828.7±0.25 mm, and no changes outside this error have been

found over a 2-year period. The combination of azimuth and longitude, on
which the measurement of right-ascension depends, was established by ob-
serving the bright fundamental star Algol, which is a weak and variable radio
source.
The telescope is controlled by an on-line computer which continually up-
dates the position of the selected map-centre for precession, aberration etc.,
and uses this to compute the path differences (corrected for atmospheric
refraction) to each pair of elements; these values are then used to control
electrical delays in the signals from each element before they are combined
in the receivers. The outputs of the receivers are sampled by the computer and
stored on a magnetic disc, so that at the completion of the observation they
may be combined to form a map of the area observed. The map is then
drawn on a curve-plotter controlled by the computer.
This instrument has been used in a wide range of astronomical programmes
from the study of ionized hydrogen clouds in our Galaxy to distant quasars.
Following the accurate calibration survey it became evident that as an
astrometric instrument - to establish a coordinate system across the sky - its
measuring accuracy was comparable with the best optical methods, whilst
overcoming some of the difficulties in optical work such as the measurement
of large angles. Bruce Elsmore is involved in a collaborative programme with
optical observers to relate the positions of quasars - (some of which are com-
pact sources at both optical and radio wavelengths) - as measured by radio
means, to those derived from the fundamental stars, in order to determine any
large-scale non-uniformities which may exist in the present astrometric
systems. He also showed how this type of instrument may be used for the
direct measurement of astronomical time-without the need for collaborative
observations at different longitudes to correct for polar motion-again with
an accuracy comparable with optical methods (~ 5 mS in a 12 h observa-
tion) .
Another programme is concerned with a study of the birth of stars; when
a cloud of gas condenses to form a star, the dust which it contains provides
such an effective screen that newly-formed stars, with their surrounding
regions of ionized hydrogen, can never be seen optically; only after this dust
cloud has dispersed does the star appear. The dust introduces no appreciable
absorption at radio wavelengths, so that radio observations allow these regions
to be studied at the earliest stages.
NGC 7538 is an example of such a region, and the upper part of Fig. 9
shows the radio emission as mapped with the One-Mile telescope. The large
diffuse component corresponds almost exactly with the optical nebulosity, and
represents the cloud of gas ionized by one or more O-stars formed about a
million years ago, with the dust sufficiently dispersed to allow the light to be
seen. The compact lower component corresponds to gas ionized by much
younger stars, which are still embedded in dust too dense for any optical emis-
sion to escape, and it is invisible on the photograph. When this southern
component was mapped with the higher resolution of the 5 km telescope, the
198 Physics 1974

Fig. 9. The ionized hydrogen cloud NGC 7533. The upper radio map shows the large
cloud associated with the optical emission, and another, compact, component to the
south. This compact component is shown with greater resolution below.

lower map was obtained, showing that there is an ionized cloud some 10” arc
in diameter, probably produced by the radiation from a star of spectral type
O8, and an even more compact cloud to the south of this, produced by a still
younger star, only a few thousand years old. The dust surrounding these two
compact regions is heated by the young stars they contain, and both have been
detected by their infra-red emission (8).
But the most extensive programme has been the mapping of extragalactic
sources - the radio galaxies and quasars; galaxies which, during a brief frac-
tion of their lives, produce some 10 60 ergs of energy, equivalent to the total
annihilation of the matter in about a million suns, by a mechanism which is
not understood.
 M. Ryle 199

Fig. 10. The powerful radio galaxy in the constellation of Cygnus mapped with the
5 km telescope. The compact outer components are exceedingly bright - (31 and 41
contours). The central component - which corresponds to the nucleus of the optical
galaxy is very weak and is drawn with contours spaced at l/5 the interval.

Fig. 10 shows the new radio map of the source in the constellation of
Cygnus - the first powerful radio galaxy to be recognized. The distribution of
polarized emission from the north component is shown in Fig. 11, giving in-

Fig. 11. The polarization of the emission from the north component of the Cygnus
source which shows the magnetic field to be turbulent on a scale ~ 104 light-years.
Fig. 12. Maps of six extragalactic radio sources.

formation on the magnetic field. Maps of a number of other sources made

with the 5 km telescope are shown in Fig. 12.
In most cases the radio emission originates mainly in two huge regions
disposed far outside the associated galaxy - although weak emission may also
be detectable from a very compact central source coincident with the nucleus
of the galaxy. In some cases much more extensive components or a bridge
linking the components occur.
The finer detail provided by the 5 km telescope has already enabled some
important conclusions to be drawn; the energy is probably being produced
more or less continuously over a period of 107--10 8 years in a very compact
nucleus and not, as was originally thought, in some single explosive event.
The source of this energy may be associated with the gravitational collapse
of large numbers of stars, in the manner which Tony Hewish describes in his
lecture, or by material falling into a much more massive collapsed object at
the nucleus of the galaxy. The mechanism for transmitting this energy to the
compact heads of the main components (e.g. Fig. 10) is not understood, but
 M. Ryle 201

may involve a narrow beam of low frequency electromagnetic waves or

relativistic particles (9, 10). The interaction of this beam with the surrounding
intergalactic medium might then accelerate the electrons responsible for the
radio emission from the compact heads, and their subsequent diffusion into
the region behind the heads can probably explain the general shape of the
extensive components.
While much remains unanswered, the present conclusions were only reached
when detailed maps became available; the physical processes relating the
nucleus, the compact heads, and the extensive tails or bridges can clearly only
be investigated when the relationship between these structural components is
known.
What can we expect in the future? In 1954, the first aperture synthesis
telescope provided maps with a resolution of 2O.2; today we have maps with
a resolution of 2” arc. Can we foresee a continuing development with radio
pictures having much better resolution than the optical ones? The technical
problems of increasing the aperture or decreasing the operating wavelength
are severe, but they do not appear to be as serious as the limitations imposed
by the earth’s atmosphere; in optical observations atmospheric turbulence on a
scale of ~ 10 cm in the lower atmosphere introduces irregularities in the
incident wavefront which normally limits the resolution to ~ 1” arc. At
radio wavelengths the contribution of these small-scale irregularities is not
important, but there are also irregularities of refractive index on a much
larger scale in the troposphere. Two distinct types have been found in a series
of observations with the One-Mile and 5 km telescopes; neither can be
attributed to variations of air density, and both are probably due to non-
uniformity in the partial pressure of water-vapour, which makes an important
contribution to the refractive index at radio wavelengths. One class has a
typical scale size of ~ 0.7 km and is attributed to turbulence in the tropo-
sphere due to solar heating of the ground in the same way that fair-weather
cumulus clouds develop. These irregularities, however, are often detected in
clear air conditions without the formation of cumulus clouds; they only occur
during day-time and are more severe during summer months. The second
class, - which shows only slight diurnal or annual variation, has a much larger
scale size, typically 10-20 km, and there may be still larger scales which have
not yet been recognized. The origin of these disturbances is not known, and
it is therefore not possible to predict how they might depend on geographical
position.
Under very good conditions - representing about 1% of the total time, the
atmospheric irregularities are extremely small and correspond to a distortion
of the incident wavefront by < 0.2 mm over 5 km; under these conditions,
operation at a wavelength of 4 mm or less would be possible and should
provide maps with a resolution better than 0” ._2 arc. These excellent observ-
ing conditions have only been encountered during periods of widespread
winter fog when the atmosphere is extremely stable, a result which illustrates
the differing requirements in seeking good sites for optical and radio ob-
servatories!
202 Physics 1974

For most of the time the atmospheric irregularities are considerably worse,
and although there is insufficient information on scale sizes > 20 km, the use
of instruments much larger than this will introduce difficulties associated with
the curvature of the atmosphere. One might guess that it should be possible
to build instruments which would give a resolution better than 0”.5 arc for
perhaps 50% of the winter months.
To reach a greater resolution new techniques capable of correcting for the
atmospheric effects will be necessary. One simple, though expensive, solution
would be to build a second dish alongside each element, so that observations
of a reference point source close to the area to be mapped, could be made
simultaneously at every spacing; the observed phase errors for this reference
source could then be used to provide a continuous correction for the signals
from the area being mapped.
Such techniques can clearly be extended to the interferometers having
baselines of many thousands of km (VLBI) which have been made possible
by the development of atomic frequency standards. These instruments have
shown the existence of very small components, ~ 0”.00l arc in some sources.
The use of a comparison source for eliminating both atmospheric and in-
strumental phase was first used at Jodrell Bank in the special case of sources
of the OH maser line at λ = 18 cm, where different components within the
primary beam can be distinguished by their frequency; if one is used as a
phase reference the relative positions of the others can be found (11).
For continuum sources a reference outside the primary beam of the in-
strument must, in general, be used and two elements at each location are
needed. This technique has been used in the U.S.A. to reduce both in-
strumental and atmospheric phase variations in measurements of the gravi-
tational deflection of radio waves by the sun (12) ; one pair of elements was
used to observe a source close to the Sun, while the other pair observed a
reference source about 10’ away.
The accuracy of the correction, and hence the shortest wavelength at which
mapping could be achieved, would depend on the angular separation between
the area to be mapped and a reference source sufficiently intense and of suf-
ficiently small angular size. But even if adequate phase stability can be at-
tained in this way, there is a serious practical difficulty in making maps with
resolution ~ 0”.001 arc, due to the inevitable poor sampling of the aperture
plane. Even with 5 or 6 stations distributed across one hemisphere of the
world, and using every possible combination of the signals from them, with
observing periods lasting several hours, the fraction of the aperture plane
which can be filled is still very small, so that the field of view which can be
mapped without ambiguity from secondary responses is unlikely to exceed
~ 0”.02 arc. Whilst there seems little hope of deriving complete maps of most
sources with this resolution, there are certainly some central components
where such a map could provide very important information.
But I think it may also be important for our understanding of the
mechanisms operating in the main components of radio sources, to obtain
complete maps with intermediate resolution; for this work extensions of the
 M. Ryle 203

present synthesis techniques, while retaining good filling of the aperture plane,
are needed.
The last 25 years have seen a remarkable improvement in the performance
of radio telescopes, which has in turn led to a much greater understanding
of the strange sources of “high-energy astrophysics” and of the nature of the
Universe as a whole.
I feel very fortunate to have started my research at a time which allowed
me and my colleagues to play a part in these exciting developments.

R EFERENCES
1. Ryle, M. (1958) Proc. Roy. Soc. A., 248, 289.
2. Scheuer, P. A. G. & Ryle, hi. (1953) Mon. Not. R. astr. Soc., 113, 3.
3. Blythe, J. H. (1957) Mon. Not. R. astr. Soc., 117, 644.
4. Costain, C. H. & Smith, F. G. (1960) Mon. Not. R. astr. Soc., 121, 405.
5. Pilkington, J. D. H. & Scott, P. F. ( 1965) Mem. R. astr. Soc., 69, 183.
6. Gower, J. F. R., Scott, P. F. & Wills, D. (1967) Mem. R. astr. Soc., 71, 49
7. Ryle, M. & Neville, A. C. (1962) Mon. Not. R. astr. Soc., 125, 39.
8. Wynn-Williams, C. G., Becklin, E. E. & Neugebauer, G. (1974) Ap. J., 187, 473.
9. Rees, M. (1971) Nature, 229, 312.
10. Scheuer, P. A. G. (1974) Mon. Not. R. astr. Soc., 166, 513.
11. Cooper, A. J., Davies, R. D. & Booth, R. S. (1971) Mon. Not. R. astr. Soc., 152,
383
12. Counselmann, C. C., Kent, S. M., Knight, C. A., Shapiro, I. I., Clarke, T. A.,
Hinteregger, H. F., Rogers, A. E. E. & Whitney, A. R. (1974) Phys. Rev. Lett..
33, 1621.