Radio Astronomy Receivers

31. E. TIURI, MEmER, IEEE

Summary-A general surveyof the principles of radio astronomy T A +

receivers is presented. System noise temperature, the sensitivity of

different receiver types, and the calibration of receivers are studied.
A total-power receiver is analyzed as a basic radio telescope receiver
and the results are used toobtain the performance of other receiver
typessuch as the Dickereceiver,Graham'sreceiver,correlation
receiver, and phase-switching receiver.

A NTENNA-RECEIVER systems used in radio as-

tronomy are called radio telescopes or radiom-
eters.
Radio
telescopesmeasure
radiatedbyastronomicalobjects.Inmost
R F power
cases the
Fig. 1-Total power receiver.

will probably not be used in the future because of the

signal power is broad-band noise, the statistical prop- increased activity and interference in the radio spec-
erties of which do not differ from the noise originating trum.
in the receiver itself orfrom thebackground noise In most receivers the selectivity is determined by the
coupled tothe receiver bytheantenna.The signal- I F amplifier and, hence, the bandwidth of the I F ampli-
power level in radio astronomy receivers is commonly fier is normally given as the bandwidth of the radio tele-
quite low, of the order of to w. This means scopereceiver(alsoin the case of atwo-channelre-
t h a t high sensitivity is one of the first requirements of ceiver). The main part of the predetector amplification
the receivers. In the following a general survey of the (of the order of 80 db) is furnished by the I F amplifier.
principles of radio telescope receivers is presented. Spe- The detector used in radio astronomical receivers is of
cial attention is directed to the system noise tempera- the square law type. The receiver calibration is inde-
tureconcept,sensitivityandsomeotherimportant pendent of the detector level and there is a linear rela-
questions. tionshipbetweenthe signal input power and
the
receiver output indication. The receiver sensitivity for
OF RADIOTELESCOPE
PRINCIPLES RECEIVERS small signals when using square law detector is also as
Radio telescope receivers have construction basically good as for any other detectorlaw [l], [2]. The detector
similar to the receivers used in other branches of radio output is adcvoltagedirectlyproportional tothe
science and engineering, especially in radar systems and receiver input noise power, and a disturbing low fre-
in scatter communication systems. The most common quency fluctuation or noise voltage. The amplitude of
type is thesuperheterodyne receiver.Fig. 1 showsa the fluctuation is smoothed by a low-pass filter amplifier
block diagram of the basic radio telescope receiver, that and an integrator. A typical signal integration time in
is, a total power receiver or direct radiometer. Although radio astronomical receivers is in the order of ten sec-
the total power receiver is not used very much nowa- onds and, hence, is usually not affected by the low-pass
days, i t contains the same parts as the more complex amplifier. Asimple RC circuitiscommonly used as
receivers and many characteristics of the latter can be integrator.
inferredfrom those of the corresponding total power The integrator output is fed to a suitable recorder or to
receiver. Hence, the total power receiver can be consid- an analog-to-digital converterfor further processing [SI.
ered as the basic radio telescope receiver. In total power receivers the dc-output voltage resulting
The R F amplifier is in most cases a maser, a para- fromtheantennabackground noise and from the
metric amplifier (oftena cooled type),a low-noise receiver noise is usually compensated by an equal bat-
travelling wave tube a ortunnel-diode amplifier. Triodes tery voltage - 17, (dc compensation) in the first stage
are used in the H F range. The R F amplifier is followed of the low-pass amplifier. Hence, a dc-output voltage is
by a low-noise crystal mixer which served as the first obtainedonlywhenasignalispresent.Thesuitable
stage in earlier receivers. I t was normally used without output level is in the order of one volt which requires a
any preselectivity so that the noise on the image fre- post-detector gain of several tens of decibels because the
quency as well as the noise on the signal frequency en- signalpoweris normally only a small fraction of the
tering the receiver resulted in a more sensitive system total antenna and receiver noise power.
by almost 3 db. Although the bandwidths of the RF The high R F and dc gain used in the total power re-
amplifiers may be sufficient such two-channel receivers ceiver requires very stable amplifiers. The receiver can
not distinguish between a signal power variation and a
gain variation. In practice the gain stability, instead of
Manuscript received May 4, 1964.
The authoris with the Instituteof Technology, Helsinki, Finland. the low frequency fluctuations resulting from receiver
930
 Tiuri: Radio Astronomy Receivers 931

noise, determines how small a signal can be detectedby

the total power receiver. The limits set by the receiver
system noise and by the gain instability will be consid-
ered in more detail in the following sections.

SYSTEM OF THE RECEIVER

NOISETEMPERATURE
T h e noise power originating from a radio telescope
receiver is given by
P.\-R = kTRB (1)

where K isBoltzmann'sconstant (1.38 X joules/"K),

B the bandwidth of the receiver, and TR is the noise
temperature of the receiverreferred to the telescope
antenna terminals as given by

where Fig. 2-Antenna sky noise temperature TAwith beam pointing at the
galactic pole andgalacticcenterandduetotheatmosphere
T p = noise temperature of the antenna feed line (all (dashed) as a function of the frequency. Noise temperatures TR
of some typical receivers are also shown.
transmission line components included)
G F = p o \ ~ e rgain of the antenna feed line
defined for one signal channel [ 3 ] . T h e noise tempera-

(<1, attenuation L p GP' >

=- ture of the radio telescope receiver in the case of two
channels is
T ~ ~ = n o i temperature
se of t h e R F amplifier
GRF=available power gain of the R F amplifier
T.~I
= noise temperature of the mixer IF-amplifier "A 2-channel noise figure" FR2 = $ FR cannot be used in
combination. (5) for calculating the effective 2-channel receiver noise
temperature."
If the R F gain is sufficiently high, TR is determined T h e radio telescope antenna delivers a disturbing
by the two first terms. The noise temperature of the noise power to the receiver which depends on the back-
antenna feed line is ground noise from the sky, the atmospheric noise, the
side-lobe noise, the noise from the losses of the antenna,
T F = (LP- ~ ) T F P (3)
etc. This powerisexpressedin terms of the effective
where T F p is the physical temperature of the line. If antenna noise temperature TA in which the signal noise
the receiver noise temperaturewithouttransmission power A T is not included.
line losses is TR' then the excess noise temperature due Fig. 2 shows the antenna sky noise temperature and
to the losseswill be the noise temperatures of some typical receivers as a
function of frequency.
+
T R - TR' = ( L p - l ) ( T p p TR'). (4) The system noise temperature TS,Vof the radio tele-
If, for example, the line is a t room temperature and the scope receiver is
noise temperature TR' of the receiver itself is 150"K,
each 0.1 db attenuation will add about 10'K to the re-
ceiver noise temperature. Hence, in low noise receivers The total disturbingnoise power referred to thereceiver
i t is necessary t o keeptransmissionlinesshortand input is then
low loss.
Earlier it hasbeen common t o use the (average) noise
figure FR instead of the noise temperature TR.The two and the signal-noise power
are related by Ps = KATB. (9)
T R = ( F R - 1)To (5) SENSITIVITY O F THE TOTAL P O W E R RECEIVER
where To = 290°K. As stated before in the case of super- The sensitivity of the radio telescope receiver is de-
heterodyne receivers, when there is no R F selectivity, fined as the signal which will give a dc-output voltage
the image and signal bands are equally useful in radio equal to the effective value of the output fluctuations
astronomy applications except incases where the signal due to the systemnoise. Fig. 3 shows the signal voltage
spectrum is narrow. T h e noise figure of the receiver is vs time and the power density a t 3 places,in the I F
932 IEEE TRANSACTIONS O N AAVTE~YKAAS
AND PROPAGATION December
T14BLE I
HF NOISE BANDWIDTHS
EEF

Type of filter
--
n cascaded single
tuned stages
It= 1 3.14
2 1.96
3 1.76
5 1.62
53 (Gaussian) 1 .so

nz cascaded2n-pole
Butterworth 6lters
V O U T W
nr = 1, ft = 2 1.48
n=3 1.26
n = 00 1.00
tit = 2, I t = 2 1.30
M
BLF Aw=w-wa where wa=angular center frequency.
Bs=3 db bandwidth of a single section.
Fig. &-Voltage waveforms and power spectra at different stages in R = 3 d b bandwidth of the amdifier.
a radio telescope receiver.

amplifier output, in the detector output, andin the inte-

grator output. The dc-output power due to the system
noise is (see Appendix) where G L F ( f ) is the post detector power gain or power
transfer function. Valuesof B L p for some common types
PDC,,,= C(kTsxBrrp)2 (10) of lorn-pass filters and integrating circuits are given in
and the dcpower due to thesignal noise is Table 11. For example the equivalent LF bandwidth of
the ideal integrator which integrates the signal for T I
PDCS= C(kATBaF)2. (11) seconds is 1/(2rI).
C is a constant and BHFis the noise bandwidth of the The sensitivity of the receiver is obtained by taking
P D C=~ P N which~ ~ yields
receiver HF section prior to the detector asgiven by

[ s,"v)df]'
BEF = (12)
When using an idealintegratorwiththeintegrating
Jo%f)%f time TI the sensitivity is

GCf) is the predetectorpower gain of the receiver. Values

of BHPfordifferentIF-amplifiertypesare givenin
Table I. BHFis close to the 3 db bandwidth of the am-
For otherfiltersandintegratorshavingthe Iow fre-
plifier except for cascaded, synchronously-tuned stages
quency noise bandwidth of BLF an"equivalentinte-
in which case it is about 1.6 times the 3 d b bandwidth.
grating time" canbe defined as
The dc-output voltageof the detector due to the sys-
tem noise is normally compensated with an equal bat- 1
TI=-.
tery voltage - Vo. The actual dc-output power of the (1 7)
~BLP
receiver is then given by (11).
The output fluctuationnoise power is (see Appendix) Theseare alsoshownin Table 11. For example, the
equivalent integration time for a simple RC circuit is
+
PNLP= 2Ck2(Tsw AT)'BAFBLF T I = 2 7 ~ Hence
~ . (16) can be used for all total power re-
E2C(kTsx)'B~pB~p (13) ceivers in connection with (i), (12), (14), and (17).
Eq. (16) can also be obtained by physical reasoning.
provided AT issmallincomparisonwith Tshr. The As indicated in Fig. 3 the predetection noise signal is
quantity B L Pis the equivalent bandwidth of the low- correlated over a period of time of the order of l / B ~ p
pass amplifier and integrator as given by seconds.Hence there are BHFeffectively independent
 1964 Tiuri: Radio dstrolzomy Receicers 933
T.-ZBLE I1 assumption will not holdin practice because the H F
LF NOISE BANDWIDTHSBLPASD EQUIGALEXT gain is close to 100 db, and it is very difficult to con-
INTEGRATIOSTIMESTI
struct a receiverwhich will not have gain variations
I
Type of filter of at least of the order of 0.1 db. Even for such stability
electronicallyregulatedanodeandheatervoltages,
integrator,
Pure
grating time rr
inte-
'
sin2 ( t w q )
(+UT, j II, 1
71

1
stabilization of operating temperature, etc. are required.
As mentioned earlier the receiver is not able to distin-
guish between the increase in signal power and the in-

I B
Ideal low-passfilter 1 __
I
2B crease due to higher gain. For example, if the system
~t RC filters in inde- noise temperatureis 200°K andthepredetectorgain
pendent cascade, increases 0.1 db, the output indicatorwill show a signal
time constant r of about 4°K. iyariations of the noise figure and of the
RC
n= 1 1. j i B 2rRC
bandwidth of the receiver have similareffects. However,
2 1.22B h R C
slow driftduring a relatively long timeinterval (in
1 comparisonwith the signaltime) is notveryserious
a (Gaussian) 1.06B because i t can be readily recognized on an analog out-
2nd order flter (WO put.
=undamped natura In the following discussion only the gain instability
frequency of the fil- is investigated. Output fluctuations due to gain varia-
ter; .t=damping tions are independent of the fluctuations resulting from
constant) system noise. Hence,
-
23 = 3 db bandwidth.

noise contributions per second producing the detector

where AG is the effective value of the receiver HF-power-
output. These are averaged in the post detection inte-
gain variation, and Go is the average predetector power
grator over the integration time 7 1 . During this time
gain. I t is assumed that the dc voltage resulting from
there are BHPTI independent contributions and there-
the system noise is compensated in the first stageof the
fore the rms deviation is l/\/BHFrI times the average
low-pass amplifier and, hence, post detector gain varia-
(dc) output.
tions will have no effect on the sensitivity.
Eq. (16) shows that the sensitivity of the radio tele-
In practice, AGi/Go is the determining term in (18) so
scopecanbeincreased by increasing thepredetector
that the actual sensitivityof the total power receiver is
bandwidth and the equivalent integration time. Forex-
considerably less than the theoretical value to be ex-
ample when T ~ s = 200"K, B H F =10 RIc and the inte-
pected from a consideration of the system noise. There-
grator is a simple RC network with the time constant 20
fore, total power operation is not suitable for high sensi-
seconds, ATmin=O.O1"K.
tivity receivers. However, the total power mode is use-
An increase in the H F bandwidth is limited by the
ful, for example, in checking the stabilityof a radio tele-
usable bandwidth of the amplifiers and by interfering
scope receiver.
signals. I t is also not desirable to increase the integrat-
ingtimebeyondcertainlimits.Forexample,source DICKE RECEIVER
structure may be lost if T I is more than the time re-
quired for a point source to traverse the antenna beam Dicke [7] was the first to introduce the use of the
width between half power points. Too long an integrat- modulationprincipleforeliminatingfluctuations in-
ing time may also cause a significant displacement in duced by receiver instabilities. Fig. 4 shows the block
the position of the source [4], [SI. diagram of the most common radio telescope receiver,
In practice, the minimum detectable signal is taken the Dicke receiver. The inputof the receiver is switched
to be several times greater than that given by (16) in between theantennaandthe referenceload with a
order to increase the probability that the deflection is repetitionfrequency f , T f . If the noisepowerfrom the
not a random fluctuation. The probability of exceeding referenceload kTcBeF is equal to the antenna noise
an amplitude of four times the effective value is only power kTdBHFthen the signal power kATBHp is ampli-
3 X 10-5 which roughly means that this event may occur tude modulated a t t h emodulating frequencyf,t~ because
once every 3.3 X 10%r seconds. For example when T I = 20
the signal can enter the receiver during half periods of
seconds this interval is about 7 days. the switching frequency only. If f 2 1 f is high enough in
comparison with the gain instability frequencies, theni t
is possible to detect thesignal without disturbingeffects
EFFECTS
OF RECEIVER INSTABILITY
due to gain fluctuations by using a multiplier or phase
Intheanalysis of sensitivityabove, i t has been detector as shown in Fig.4. The multiplier switches the
assumed that the gain of the receiver is constant. This receiver detector output in synchronism with the an-
-..

934 IEEE TRAAi-SACTIOhTS O N ANTENNAS

PROPAGATION
AND December
power system. Eq. (20) assumes square wave modula-
tion and square wave multiplication.It is advantageous
to introduce the post detection gain needed before the
multiplier state in which case a video amplifier can be
used. The bandwidth of the videoamplifier must be
from f M to about ten times f M in order to include all of
the important harmonics of the square-wave signal.
Fig. G D i c k e receiver (band-pass filter optional). Several amplifier stagescanbesaved if the video
amplifier is made narrow band ffM plus or minus a few
cycles), as for example, by using a narrow band-pass
filter as shown by the dottedlines in Fig. 4. This means
some reduction in the sensitivity because only the first
3 harmonic of the square-wave signal is used in the multi-
plier. The amplitude of the first harmonic is 4/n times
I
the amplitude of the square wave and itseffective value
SWITCH.
GEN.
is 4/4/2n. Hence, the expression AT,, in (20) must be
multiplied by t/%/4. Because thereduction in the
Fig. S--hlultiplier and integrator stageof a Dicke receiver. sensitivity is only a little more than 10 per cent, this
modification of the Dicke receiver is usually preferred.
tenna switch in opposite phase to the integrator (Fig. In this case the waveform from the switching generator
5). Theoutputvoltagefromtheintegrator will be to the mdtiplier can also be sinusoidal. If the input
directly proportional to the signal temperature AT and modulation is also sinusoidal insteadof square wave the
there is no dc output if TC= T Aand no signal is present. sensitivity is further reduced by the factor4/n.
The modulationfrequency is in practice 10-1000 Some workers have proposed unsymmetrical modu-
cycles per second (preferably not a multiple of power lating waveforms, for example, in order to balance the
line frequency). The waveform and phase shift circuits noise powers in bothpositions of the input switch. These
of the switching voltage must be stable. Switches are will causesomereduction in sensitivitybecausethe
commonly semiconductor diode or ferrite switches hav- modulation of the signal is not so effective as in the
ing an attenuation of about 0.2 d b in "on" position. symmetrical case [4].
When the receiver is connected to the antenna, the Sensitivity (20) assumesabalanced receiver or TC
dc-output voltage is = TA. There are several different methods to achieve
this. One possibility is to use a second antenna which is
+
V A = CIGK(TA T R ) B R P , directed toward a cold region of the sky, such as near
the galactic pole, so that its noise temperature is less
and when i t is connected to the comparison load i t is
than that of the telescope antenna. The balance is then
+
V c = ClGK(Tc T R ) B H F . obtained by feeding additional noise from the adjust-
able noise source to the reference Ioad through a direc-
The output voltage from the multiplier is hence tional coupler so that Tc equals T A . In hydrogen line
measurementsthe referenceload canbethe signal
V = C~G(VA - Vc) = C~CZGK(TA - TC)BHF
antenna when the receiving frequency during the refer-
or as given in terms of the noise temperatures; ence period is situated outside the hydrogen line spec-
trum. The input switch is not required when switching
AG
AT = ( T A- Tc) - * (19) is made by side tuning the local oscillator.
Go
GAIN-MODULATED DICICERECEIVER
If Tc= TA, outputvoltagefluctuationsduetogain
A useful method for obtaining balanced operation in
instability disappear and receiver sensitivity is deter-
Dickereceiversis gainmodulation. Fig. 6 showsthe
mined by the system noise alone.
block diagram of thegain-modulation receiver. Two
In the Dicke receiver the signal is connected to the
stable passive attenuators are switched alternately in
receiver only half of the time (in the case of square wave
synchronism with the input switch into the I F amplifier.
modulation).Hence,thesensitivity of theDicke re-
T h e H F gain of the receiver with the attenuator is GA
ceiver is
when the signal antenna is connected and Gc when the
comparison load is connected. If

orone half of the theoretical sensitivity of the total then the receiver is balanced.
 1964 Tiuri: Radio Bstlaonomy Receivers 935

m
L- - - - - _ _ _-
.
Fig. 6-Dicke receiver using gain modulation for balancing.

b NULL-BALANCING
RECEIVER
DICKE 1
V,",
All the Dicke receiver types investigated so far suffer Fig. 7-Kull-balancing Dicke receiverof Machin, Ryle, and\:onberg.
from gain instability when the signal is present, espe-
cially when the signal is relatively strong. Instabilities
will distort the shapeof the signal curve and reduce the
accuracy of the results. A receiver where the balanced
condition is realized all the time is the null-balancing
Dickereceiver of Machin, Ryle and Vonberg [SI. Its
block diagram is shown in Fig. 7. The comparison load
has an adjustable noise source. The output noise power
of thesource is controlled bythereceiver-integrator
output so that the output is always zero. The actual
output signal of the radio telescope receiver is then the
controlling signal. On VHF and UHF frequencies the The multiplier output contains only a correlation signal
comparisonsourcecanbe a noise diode whose anode or a signal proportional to the noise power coming from
current is directly proportional to the noise power and the antenna which is the same for both receivers. T h e
can be used as a receiver output signal. A t microwave noise powersfrom thetwo receivers areuncorrelated
frequencies the comparison source could be a fixed out- and, hence,will not produce a dc output.
put noise generator (discharge tube, cold load etc.) in Thesensitivity of thecorrelation receiver is (see
connection with a current-controlled attenuator. Appendix)
The null-balancing method can also be combined with
the gain-modulation method byusing a servo-controlled
current-dependent attenuator in the comparison attenu-
ator at the intermediate frequency.
GRAHAM'S RECEIVER
where T is the systemnoise temperature (T= 4 T A TR). +
If all the antenna noise is signal noise then the sensi-
In the normal Dicke receiver the signal power is ob- tivity of the correlation receiver is the same as that of
served only half of the time. This may beconsidered Graham's receiver. If the signal noise is a small fraction
I wasteful when large telescope antennasare used t o col- of the antenna noise and receiver noise is small in com-
lect the signal power from the source. Full efficiency in parison with the antenna noise then the sensitivity is
observing time is achieved by switching the telescope the same as that of the total powerreceiver. In this
,. antenna between two receivers. Both receivers can be latter case the gain fluctuations will in practice deter-
of the Dicke type and the outputsof both can be added mine the sensitivity of the correlation receiver. Hence,
following the observation thereby increasing the sensi- the correlation receiver is useful only when TA is small.
I tivity
by 42. In practice it maybe difficult to couple two receivers in
The combining operation can also be made electroni- parallel. Care must be taken so that the input circuit
cally during the actual observation. One method is pre- noise from one receiver does not enter the otherreceiver
.. sented by Graham [SI and is shown in Fig. 8. Due to the since this will cause anextracorrelatedsignal.The
adding of two independent observations the sensitivity coupling can be arranged, for example,by using two
is 4 2 times better than in the case of one receiver. circulators.
Thecorrelation principle has also been applied by
CORRELATIOK RECEIVER Ryle in the phase-switching radiometer used in inter-
Two identical radio telescopereceivers can becon- ferometer applications, Fig. 10 [12]. Two identical re-
nected as shown in Fig. 9 resulting in a correlation re- ceivers are connected to two antennas of the interferom-
ceiver [lo], [ l l ]. Both receivers are coupled in parallel eter. The intermediate frequency signal of one receiver
to the antenna and the IF-output signals are multiplied. goes through a phase reversing switch which is operated
.- 1
~.

936 IEEE
TRANSACTIOLVS OiV A N T E X N A S AND PR.OPAGATION December

NOISE
T wT
F l
NO1 SE

ISECT. I V,,,

Fig. 9-Correlation receiver. (a) (b)

Fig. 11-Calibration circuits.

be a t room temperature). Values of LD and L must be

known accurately and they must be stable in order t o
obtain an accurate calibration. If, for example, T G is
I 10,000'K and the required calibration signal is 1°K f5
per cent then the total attenuation LDLmust be 405 0.2
Fig. 10-Phase-switching receiver.
db.
The arrangement shown in Fig. 1l(b) is suitable for
an overall calibration of a radio telescope. However, the
at frequency f-w. If signals VI and Vz are uncorrelated attenuation between auxiliaryantennaand telescope
the switching will have no effect on the square law de- antenna must be known accurately.
tector output. When VI and V, contain correlated com-
ponents the detector outputis different for VI+ Vzthan CONCLUSION
for V I - Vz. This means that the detector output varies
at frequency f,v due to the correlated signal. Assuming Table I11 shows the comparison of the sensitivities
that the desired signal is the only correlated signal it is of different radio telescope receiver types. Sensitivities
clear that the sensitivityof the phase switching receiver are given with reference to the total powerreceiver
is the sameas the sensitivityof the simple Dicke receiver which is theoretically the most sensitive type ( a = 1)
using a similar low frequency section [4]. when gain variations are neglected. With refrigerated
radiofrequencyamplifiers and carefullydesignedan-
CBLIBRkTION O F RADIOTELESCOPE RECEIVERS tennas the system noise temperature of microwave radio
telescope receivers can be lowered to a few tens of de-
Radio telescope receivers must be calibrated in order grees. I t is to be expected that the sensitivity in many
to measure the signal power. When using a square Iaw cases will then be limited by external reasons such as
detector the calibration is independent of the detector
level but will, of course, depend on the receiver gain.
TABLE I11
Therefore, i t is common practice to perform the calibra-
SENSITIVITIES OF DIFFERENT RADIO TELESCOPE RECEIVER TYPES
tion before and after the observation period or use a
calibration signal switched on automatically a t regular
time intervals. The possibility of using AGC for achiev-
ing a more stable calibration has been investigated by
Tmi,= u
Tsx
-V'BHFTI
-
Colvin [13]. Total power receiver
(Fig. 1). I 1
~ ~

Fig. 11 shows two possible calibration and checking Dicke receiver (Fig. 4),square wave modulation,
arrangements. In Fig. ll(a)the noise generator is square wave multiplication.
assumed to havea noise temperature T Gwhen in opera- Dickereceiver(Fig.
tion and room temperature otherwise. The excess noise
temperature T X Ein the receiver caused by the operation
of the noise generator is
plication).
Dicke
~
4) squarewavemodulation,
narrowbandvideoamplifier(sinewavemulti-

receiver (Fig.4) sine wave modulation, nar-

A
-=2.22
IJI
row band video amplifier (sine wave multipli- 242=2.83
cation).
Graham's receiver(Fig.8) square wave modulation, d 2 = 1.41
square wave multiplication.
Correlation receiver (Fig. 9) (only signal noise is
assumed to be correlated).
Phase switchingreceiver(Fig. 10) (onlysignal
Lo is the coupling loss of the directional coupler, and L noise is assumed to be correlated), square wave
switching and multiplication.
the attenuation of the attenuator (which is assumed to
 1964 Tiuri: Radio Asfronomy Receivers 937

variations and fluctuationsin atmospheric transmission, (Fig. 3 ) . The dc voltage due to the noise temperature
noise power pickup from the sun and from strong radio T S N is balanced by a voltage Yo. T h e difference V , is a
starsthroughminor lobes, andbyinterference from measure of the signal noise temperature AT. Its mean
radio transmitters and other receiving systems. In the valueis
future, the simple two channel radiometer, which has E[V,] = + k * T J W / A ( j w ) 12df ( 2 8)
been extensively used in the past, can be expected t o -m
disappear in favor of systems using radiofrequency
selectivity. where E is the ensemble averaging operator.
The low-passfilter eliminatesmost of the random
APPENDIX fluctuations. The square of the standard deviation will
In this Appendix the equations for the sensitivity of then be [from ( 2 7 ) ]
total power andcorrelationradiometersarederived.
TheFouriertransformsinvolvingthe power density
spectrum @ ( j w ) can be expressed by thefollowing pair of
(Wiener-Khintchine)relations;

@(jw) = J-Ie-jwr$(;ldr = (two-sided) spectral density ( 2 3 )

Because thepost-detectionbandwidth is verymuch
smaller than that of predetection (Fig. 3 ) , w may be set
=
2T =
J-~ej~r~(jw)do S-: ejwTa(jw>df
equal to zero in the expressioninside the brackets of
( 2 9 ) . Dividing UOOCTby the average of V O U T which
( ~ ) is
= autocorrelation
function (24) equal to H ( 0 ) E I V e ]we
, obtain the relative deviation

where w=27rf. For a more complete discussion see Mid- Finally expressing the integrals in terms of noise band-
dleton [6]. widths ( TsL,>>AT )

-4. Total Power Radiometer (Fig. 1 )

The two-sided spectral density of VIS is
@ r x ( j U ) = i k ( Ts.v + A T ). where B H Fis the predetection noise bandwidth ( = I F
( 2 5 ) noise bandwidth)asgivenby
After R F a n d Iamplifiers
F the spectral densityis ob-
tained from (25) bymultiplying i t withthe power
I I
amplification ratio A ('jw) z. ( A ( j w ) = predetection volt-
age transfer function).
= +
$ K ( T ~ & -A T ) j rl(yw) p. (26)

Afterthesquare law detector we obtain,assuming VrF and BLF isthepostdetection noise bandwidthas
to be Gaussian [ ~ D E T ( T )= 2 f # I F 2 ( T ) + ~ I F 2 ( 9 ) ] , given by
1 P -

+ 2?iS(w)[ J =arF(jW)df]2
--oo
If low-pass filtering is done by a pure integrator with
integration (27)
time 71, BLF=1 / ( 2 ~ 1 ) .Equivalentintegra-
tion time may now be defined as rr= I / ( ~ B L F ) .Conse-
where 6 ( w ) is the unit impulse. The latter term of the quently the relation ( 3 1 ) can be expressed in a new form
expression ( 2 7 ) representsthedccomponentdueto
rectified
signal and system noise while former
the con- UOUT* = -
Ts.v 1
omponents
frequency
sists
doubleof low
andfrequency (34)
A T m i n V' BRFU
- .--

938 IEEE TRANSACTIONS O N ANTENNAS AND PROPAGATION

Now (38) can be expressed in terms of transfer fun&

i
tions and noise spectral densities. T o simplify the ex-
pressions the channels are assumed to be identical. Thus,

Finally if theminimumobservedchangein noise The mean value VouT is then

temperature AT,, is defined as the value with which
QOUT* = 1, EVOUT= tk(TA + 4 T ) S-: I A ( j w ) I2df.
The relative deviation will be (if AT<<TA)
-
2T
B. Correlation Radiometer (Fig.9 )
In the correlation radiometer the input signal is di-
videdandputintoseparatechannels.Inputspectral The threshold sensitivity is then [see ( 3 3 3
density of, e.g., channel 1, is then
+
a ~ ~ ~ c=j wa )k ( T ~ , ~ T fA4 4 T ) = +K(T + +AT) (36)
where T R is~ the noise temperature of thechannel 1
while TA is the noise temperature of the antenna and T ACKNOWLEDGMENT
is the system noise temperature of channel 1. Theauthor wishes tothankSeppoHalmewhohas '
The outputsof the two channels are then multiplied. compiled the Appendix andhascalculatedTables I
ForGaussian processes VrIp, and V I P 2 the correlation and 11.
function of the product is
REFERENCES .
d'DET(7) = d'IF1(T)'#'IF2(T) + d'IF1z2(0) D. G. Lampard, "The minimum detectable change in the mean
noise-inputpower to aradioreceiver," Proc. IEE, vol.101,
+ d'rFl2(~)dJrFla(-~) (37) pp. 118-128; Fe;ruary, 1954.
P. D. Strum, Considerationsinhigh-sensitivitymicrowave
4

radiometry," PIOC.IRE, vol. 46,pp. 43-53; January, 1958.

where # I I ~ ~ ~ is
( T the
) cross-correlation function of VIFl "IRE standardson receivers: definitions of terms," PROC.IRE,
and V 1 F 2 . In terms of spectral densities vol.40, pp.1681-1685; December, 1952.
S. R. O'Donnell, "A Comparison of Radiometers," M.Sc. thesis,

@DET(jU) = 2irs(w) [ s-:@rF12(j'")df]~

The Ohio State Univ., Columbus, Ohio; 1963.
J. C. McLaughlin, "A Data Digitizing and Processing System for
a Radio Telescope,"MSc. thesis, The Ohio State Univ., Colum-
bus,Ohio; 1962. c
1 " D. Middleton, "An Introduction To Statistical Communication
@IFl(jW - ju)@1Ip2(ju)du Theory," McGraw-Hill Book Co., Inc., New York, N. Y.; 1960.
+ GS, R. H. Dicke, "The measurement of thermal radiation at micro-
wave frequencies," Rev. Sci. Instr., vol. 17; pp. 268-275; July,
1 " 1946.
+
)

@ ~ I p I 2 ( j U > ~ r p l l l ( j w ;u)du. (38) K. E. Machin, M. Ryle, and D. D. Vonberg, "The design of an

+ GS-, equipment for measuring small radio-frequency noise powers,"
Proc. IEE, vol. 99"; pp. 127-134; May, 1952.
The bar over (P means the complex conjugate. %FI2Cjw)
M. H. Graham, Radiometer circuits," PROC.IRE, vol. 46, -
p. 1966; 1958.
is the cross-spectral density of VrF1 and V 1 p 2 . The first S. J. Goldstein, Jr., "A comparison of two radiometer circuits,"
PROC.IRE, vol. 43, pp. 1663-1666; November, 1955. (See also
term in (38) represents the dc component, i.e., the de- rt 11.)
L--,~,

tectedantennatemperature. I t canbeshown that D. G. Tucker, M . H. Graham, and S. J. Goldstein, Jr., "A com-
parison of two radiometer circuits," PROC. IRE, vol. 45, pp. 365-
A1(ju)A2Cju),
@rFlr(ju) = @ . T N ~ ~ ( ~ W ) because the noise 366; March, 1957.
components due toreceivers are uncorrelated with each M. Ryle, "A new radio interferometer and its application to the
observation of weak radio stars," Proc. Roy. SOG.(London) A ,
other and the antenna noise which has spectral density VOI. 211, pp. 351-375; 1952.
@m12(jw) = i k ( T ~ A T ) . + (39)
R. S. Colvin, "A Study of Radio-astronomy Receivers," Ph.D.
dissertation, Stanford University, Stanford, Calif; 1961.