Professional Documents
Culture Documents
+Ty2 TIME
(1) INTRODUCTION T k
Vs
T, - T A T
The application of f.m. techniques to radar ranging has been NSTANTANEOUS
BEAT-NOTE
discussed by Keep,2 Tucker,3 Kay4 and others. The advantages FREQUENCY-,-
and disadvantages compared with pulse radars have been set out
Written contributions on papers published without being read at meetings are
invited for consideration with a view to publication.
Mr. Hymans and Mr. Lait are at the Royal Military College of Science. Fig. 1.Production of the beat note.
[365]
366 HYMANS AND LAIT: ANALYSIS OF A FREQUENCY-MODULATED CONTINUOUS-WAVE RANGING SYSTEM
This beat note is the equivalent of the video-frequency pulse (4) PRODUCTION OF THE BEAT NOTE
in a pulse radar and contains all the information. Since the Let the reference or ground wave, which passes direct from
beat note is a repetitive waveform with repetition time Ts its transmitter to receiver, induce an oscillatory voltage in the aerial
Fourier transform must consist of a spectrum of lines spaced P ^ s i n ^ , where <f>g is the expression in eqn. (6), with phase
at intervals u)s = 2TT\TS. arbitrarily zero at t = 0. The echo from a stationary reflecting
In the analysis that follows two approximations are made for target will induce a voltage Ve sin <f>e, where <f>e is a function
simplicity. They are as follows: similar to <f>g but delayed in time by T, given by
(a) The frequency increases linearly with time and has instan-
taneous flyback, although this is not, of course, possible in practice. Range of target 2R
With the type of radar considered, however, the assumption of = 2X (7)
Velocity of propagation
negligible flyback time is thought to be a reasonable approach to
the truth.
(b) The exact phase of both signals is calculated, taking into In most cases we may assume that Vg > Ve. These two
account the fact that the echo is merely the direct signal delayed by oscillations are 'mixed', 'heterodyned', or made to 'beat' together,
a time T. The medium through which the signal travels is assumed in some non-linear device. The resulting signal or beat note
to be non-dispersive. will contain a d.c. (zero-frequency) term and a product term
GVeVg sin <j>e sin <f>g, where G is a numerical constant, and other
(3) MATHEMATICAL FORMULATION OF THE SIGNAL higher-order products.
In general, only the lowest-order product will have a significant
The instantaneous frequency, <oh is given by the following set amplitude, apart from the large d.c. term.
of expressions: The product may be expanded as a difference, namely
Wi = OJ0 + 2a/, where -\TS < t < + $TS
[CQi ( ^ - fa) - COS (<j>g
= o)0 + 2a(t - Ts), where TS < t < + f Ts
The phase-sum term is an oscillation at radio frequency and is
removed by filtering. The difference term is the desired oscilla-
= CDO + 2<x(t - nTs), where i(2n - l)Ts < t < i(2n + 1)7; tion at video frequency, and contains all the range information.
We are thus interested in the function \GVeVg cos (<f>g <f>e).
(1)
It will be seen that <f>g <f>e has two forms in the nth interval.
For convenience in handling the expressions, the sweep rate They are
has been taken as 2a rather than a, and the origin of time is set
atTthe centre of one sweep cycle. (i) During time - 7 ; < tn < - \TS + T,
It has been found useful to make the substitution
when cf>e = a)0(tn_l - T) + a(tn_l - T) 2 + ( - l)co0Ts
t n = t - nTs (2) and <f> = oj t + atl + nu) T g 0 n 0 s
so that the expressions can be generalized to give, in the nth $8-<l>e = oJ0r ~ <TS - r ) 2 + 2 a ( r - Ts)t,, . (8)
interval,
(ii) During time -$TS + r<tn< + %TS,
= a)0 (3)
2
when <f>e = <o0(tn - T) + oc(tn - T) + na>0Ts
Instantaneous frequency is not a physically measurable
variable; only voltages and currents can be so regarded. The a n d <f>g is as in (i),
voltage in the transmitter is proportional to the sine or cosine
of the phase angle <f>h where <f>g <f>e = OJQT O.T2 + 2<XTtn . . . . (9)
r'
+ constant (4) (5) FREQUENCY ANALYSIS OF THE BEAT SIGNAL
The oscillation \GVeVg cos ((f>g <j>e) is the beat note between
Thus we must obtain this integral of the instantaneous fre- the echo and the reference wave. This oscillation is not a pure
quency by substitution from eqn. (3) into eqn. (4). The constant sinusoid and can be analysed into separate harmonic components
of integration can be evaluated by arbitrarily putting <f>t = 0 at by the use of the Fourier transform. If tu is the general variable
t = 0 (there is no loss of generality here) and then stipulating in the transform, we define the Fourier transform as
that phase shall be a continuous function. Such a restriction is + 00
necessarily imposed by the continuity of currents and voltages
in real circuits.
F(o>) =\\GV. cos (<f>g - (10)
J-oo
Taking the interval of zero order as an example, we then have,
from eqn. (1), This integral may be reduced to
+-oo
oo r i(2w+i)r,
M2n+1)T,
F(o>) = iGVeVg 21 s(^
cos (<f>g - <f>e)e-Ja'dt . (11)
and >i = J (^o + 2oct0)dt0 co-Nu^nor,
nr.
<f>i = c V + x t
n + ncoQT s . . . . (6) I- 2a(r - Ts)l
An examination of eqn. (6) shows that <f>{ is continuous at the + cos (CUOT - a r 2 + 2artn)s-J<->'dtn\ (12)
ends of each sweep.
HYMANS AND LAIT: ANALYSIS OF A FREQUENCY-MODULATED CONTINUOUS-WAVE RANGING SYSTEM 367
The first integral in the outer bracket gives two terms: For our purpose, it will thus be sufficient to consider the
functions for positive values of co, since the negative values
sin [co - 2OL(TS - r)]-?
contribute nothing new.
F,(co) = I ' exp - j[o)0r - \co(Ts T)]
(6.2) Amplitudes of Lines
[co - 2a(Ts - r)]T- . . . . (13)
Inspection of eqns. (13) and (15) shows that the spectrum
has two maxima on the positive axis
T sin [co + 2<x{Ts - T ) ] - at (i) co = 2ar = coBl (say),
FM expy[w0T -f %o)(Ts - T)]
and at (ii) co = 2a(Ts r) = coB2 (say).
[co + 2oc(Ts - T ) ] I . . . . (14)
Consider for the present the case when coBX and coB2 are so
The second integral in the outer bracket gives two terms: far apart that the contributions of F! and F 3 can be considered
independently. The maxima at coBl and coB2 may or may not
coincide with a particular line, depending on the value of r
sin (co 2ar) I relative to Ts. The zeros of the amplitude functions F, and
F3(o>) = T e\pj(co0T \COT) F 3 occur when
. . . . (15) (co-2ar)(Ts-r)l2= mm \
\ for m = 1, 2, 3, . . .
and and [co - 2OL(TS - T)]T/2 = + mm \ . ' .
s i n ( a , + 2 a r ) ( 'T.-
- 2 J
~ J
but not m = 0
F4(w) = - ^ -= -exp -;(CU O T 2nm
i.e. at co = 2<XT + (20)
. . . . (16)
and at co = 2oc(Ts - T) . . . . (21)
The remaining factor in F(cu) is a delta function, T
r~-(R0+vt) (24)
c
The instantaneous beat-note frequency now consists of a
pair of sliding tones, alternating in a progressively changing
ratio, for the delay time rn and T W+1 at the beginning of the
nth and (n + l)th cycle of the transmitted frequency are related
by the expression
Therefore there are two or three lines between first zeros. (25)
(b) T T,/4. Zeros spaced at j=? - r - ' and during time -iTs + :r<tn <TS
Therefore there are an integral number of cycles. But there are three lines
in the first maximum and one line at the peak of the second (upper maximum
not shown).
(c) T -> T,I2. > * -to- ( 2ar + 2Q- + 2<XT . (26)
Maxima nearly equal in amplitude. Lines overlapping.
{d) i > T,12. The corresponding positive values of the beat notes a>Bl and
coB2 thus both contain Doppler shifts, and these depend as
integral number of harmonics in coBi, the spacing between first expected on the instantaneous angular transmitted frequency;
zeros is 8coJ3 and there are three lines in the first maximum. in fact, cogj and cvB2 are, respectively,
Furthermore the (k 2)th lines on either side now occur at
about the peak of the second maximum, and have amplitudes of
about 21 % of the kth line as shown above. 2a(7; - T) - y [cu, + 2a(Ts - T ) ] | and
Fig. 2(c) shows the case when r^-$Ts. Quite clearly, the
envelopes of the two sets of lines begin to merge, and tofindthe 2ar + (oj,-2aT) (see Fig. 3). It is of interest that the
amplitude of the resulting oscillations we have to consider the
phases of the two contributions, due to the upper and lower Doppler component appears as a decrease in the beat-note
beat notes, respectively. Consideration of the argument leading frequency during the earlier, and as an increase during the later,
to expression (19), and reference to the exact forms for Fj and part of the cycle.
F 3 , shows that there will be present in the receiver (video- The difference-frequency component, GVeVg cos ((j>g <f>e)
frequency section) two oscillations both of frequency ka>s, one in Section 5, will also, in general, have a non-repetitive waveform,
from the lower beat-note envelope with phase ((JO0T ^ka)sr) and its Fourier transform F(co) must therefore properly be
and the other from the upper beat-note envelope with phase interpreted as a continuous spectrum. The detailed analysis
ika)s(Ts T) U)0T. Thus, if the two frequencies are super- is beyond the scope of the present work. In most practical
imposed and the resultant amplitude is measured, one has to take applications, however, it is adequate to think in terms of a slowly-
into account the difference of phase 2CO0T %kcusTs. Since cos = changing line spectrum of the type described in Section 6. Even
HYMANS AND LAIT: ANALYSIS OF A FREQUENCY-MODULATED CONTINUOUS-WAVE RANGING SYSTEM 369
INSTANTANEOUS DOPPLER
FREQUENCY SHIFT
amplifiers have been omitted. The beat-note frequencies and
SIGNAL
FREQUENCY
the output from the variable-frequency selective amplifier are
TRANSMITTED passed into a balanced modulator, which produces the diflFerence
frequencies. Only one of these will be at exactly lOOkc/s, and
this can be selected by the use of a phase-sensitive detector whose
switching frequency is the unmodulated intermediate frequency.
With an integration time as long as desired, one has the equiva-
lent of a narrow-band filter, tuned to exactly lOOkc/s as suggested
by Gnanalingam.1 It will be noted that great stability of the
INSTANTANEOUS
BEAT - NOTE \
intermediate frequency is not required, since it enters the system
FREQUENCY only as a 'carrier' of information, and is subtracted out again
at the phase-sensitive detector. It is suggested that the inter-
mediate frequency should be lOOkc/s, since that is a standard
value for this type of work and techniques are therefore well
known. Moreover, for the projected system, it falls well outside
the possible range of beat-note frequencies so that second-channel
interference is minimized and does not lead to confusion between,
Fig. 3.Production of the beat note when the echo comes from a say, echoes from targets at ranges 3R, 5R and higher-order
moving target. modulation products arising from the beat note of a target at
range R. This was a difficulty on which Gnanalingam remarked
for a target receding at Mach 1, rn+l rn is of the order of in Section 5.3 of his paper.
only 2 x 10~6Ts, so that T will increase from 0 to Ts (thereby
completing one cycle of spectral repetition) in a time of approxi-
mately 5 X 10 5 r s ; and at lower line-of-sight velocities the (9) RANGE AMBIGUITY
periodic time for T will be proportionately greater.
An f.m. c.w. radar ranging system of the type discussed
suffers from a difficulty not encountered with pulse-type radars.
(8) SCHEMATIC FOR A COHERENT SYSTEM Reference to Fig. 5 shows that for every range RA there is a
As shown above, the scheme proposed by Gnanalingam will complementary range R, such that the upper and lower beat
not produce the desired result of a single-line beat note except notes are identical. Thus the echo from range A has beat
in the trivial case of zero range. It is, however, possible to frequencies 4/^a/c and 2aTs 4-R^a/c, and echoes from range
produce a coherent system in which each line of the beat-note B have beat frequencies 4i?fla/c and 2<xTs 4RB<x/c. Ambiguity
spectrum is examined separately with an arbitrarily long time of in range measurement will occur when
integration equivalent to an ideally narrow pass-band filter.
Fig. 4 shows a block diagram of an apparatus for doing this. (27)
The swept frequency is generated by a sawtooth voltage which
is itself triggered by a train of clock pulses at intervals Ts, thus The obvious way of resolving this problem is to have available
ensuring that the duration of one sweep is accurately denned. different sweep rates 2a, a method which is equivalent to a
The same train of pulses is used to amplitude-modulate an choice of values of the pulse-repetition frequency in a pulse
externally generated intermediate frequency (for convenience this radar. Another method of attack would be to use what may be
has been taken as lOOkc/s), thus producing side frequencies called 'channel switching' to distinguish it from the first method,
spaced at intervals of \)TS cycles per second. Each side which is 'range switching'.
frequency can be selected in turn by a variable-frequency Fig. 5 has been drawn to show the special case when RA = RB.
high-Q-factor tuned amplifier, so that one has available a set In the general case, the two tones overlap for a fraction
of frequencies separated by the same intervals as the beat-note / 4RA\
spectrum. I1 - ) of each cycle. Thus, there is in theory the
V cTsJ
The receiver is shown in Fig. 4 only in outline. All linear possibility of confusion during virtually the whole cycle, but for
TRANSMITTER
VARIABLE-
FREQUENCY
TRIGGER PULSES OSCILLATOR
AT INTERVALS Ts
SAWTOOTH PHASE-
BALANCED
VOLTAGE SENSITIVE
MODULATOR
GENERATOR DETECTOR
INTERMEDIATE
FREQUENCY FREQUENCY
MODULATOR SELECTIVE
(lOO kc/s) AlyPLIFIER
TRANSMITTED SIGNAL values of RA small and tending to zero one must take into account
ECHO A ^-ECHO B the difference in amplitudes between the two signals. Even if
only the normal radar range equation is applicable this confusion
is not likely to arise for small values of RA, since the corre-
sponding value of RB is at extreme range. The ratio of the two
(a)
powers received will be
2<xT,
i
41
c ^= 4 =fe) 4 >
Hence, if, for example, RA = %RB, the power ratio is ^
(6)
i f
so that the ambiguous signal is approximately 19dB down on
the signal from range A. For smaller values of RA this ratio
will be even greater. In a long-range equipment such as an
t
2T. 3T,
ionosphere sounder, atmospheric attenuation will also increase
2TS
signal difference by a considerable factor. For two comple-
mentary signals at nearly half range (RA = icTs), however, the
amplitudes will be more nearly comparable. As r-+$Ts,
(C)
VIDEO-FREQUENCY
BEAT NOTES
PULSES FROM
TRANSMITTER(^)
INTERMEDIATE FREQUENCY
( f a - 2f, +
r
i
continuously in frequency, the lines are located at frequencies
which are integral multiples of 2TTITS. Thus wm, the peak of
the envelope, will, in general, occur between two lines, and the
Fig. 7.Gating waveforms for channel switching. problem to be solved is one of interpolation.
(a) Beat-note voltage; echo from range RA-
(b) Beat-note voltage; echo from complementary range RB.
By increasing the bandwidth of the filter so that it passes
(c) Lower-frequency gating waveform. two adjacent lines instead of a single line, the accuracy of inter-
(</) Upper-frequency gating waveform.
polation may be improved by scanning the lines first singly and
passed to two separate narrow-band filters followed by phase- then in pairs. Nevertheless, a fixed-magnitude 'reading error'
sensitive detectors. The channel switch is controlled by the of the order of TTITS will always be present in a>Bl, and there-
setting of the variable-frequency selective amplifier. Thus, if fore in the derived range, in addition to the percentage error
the range selection control is set to receive signals from a range arising from any uncertainty in a.
RA (lower beat-note a}m = 4<xRA/c), the switch will change over
during each sweep cycle at time / = $TS and back at time (10.2) Range Discrimination
/ = Ts + 2RA/c. Reference to Fig. 7 shows that the In certain applications, a radar system deals primarily with
unwanted signal power from the ambiguous range will be extended targets, and range discrimination will be important only
multiplied by the factor (1 -^). in so far as it is essential (as in an airborne altimeter) to assess
\ cTsJ accurately the range of the boundary of an extended target.
Thus in the worst case when RA ~ \cTs and the two interfering Within the target itself, the echo range R and the corresponding
signals are of comparable magnitude, the unwanted one is delay time T will be continuous, so that the spectrum will cover
almost completely eliminated. more or less uniformly a band of frequencies whose width
The realization of the channel switch should not prove difficult, depends on the difference between the maximum and minimum
since small errors in the switching instant will cause only a small echo ranges for the target.
deviation from the ideal case, and the degradation of the signal In other applications, the system may be called upon to dis-
will not be greatly increased. criminate between echoes from discrete objects at adjacent ranges,
and it is the purpose of this Section to extend Gnanalingam's
(10) RANGE ACCURACY AND DISCRIMINATION treatment of this topic.
He considered two targets whose echoes were of equal ampli-
(10.1) Range Accuracy tude and assumed that the relative phases of their spectral lines
In a radar system of the type under consideration, the problem were sufficiently random to justify the assumption that the
of measuring range is essentially that of measuring the time delay average response Vs of two components Vsl and Vs2 was given
372 HYMANS AND LAIT: ANALYSIS OF A FREQUENCY-MODULATED CONTINUOUS-WAVE RANGING SYSTEM
by V} = V}x + V}2. Eqn. (15) shows that, for stationary targets, targets but gives little help towards their resolution. The separa-
at ranges having delays T, T + 8T, the corresponding phase of tion must exceed 2TTI(TS T) by a significant factor before the
lines at frequency cu are (w0 io)r and (co0 ia>)(r + ST); positions of the individual targets can be confidently identified,
thus there will be a phase difference (co0 ^CO)BT between the and even then the line structure of the spectra may give only a
interfering components. Beating will thus not be entirely relatively inaccurate estimate of the actual separation between
random, but will be more or less severe according to the rapidity the targets concerned [cf. Figs. 8(c) and (d)]. The probability
and extent of carrier fluctuations; pronounced beating will also of separation does not appear to be markedly altered if the
occur, for a v.h.f. or h.f. carrier, for quite small differences ST amplitude of one target is reduced by half. In the presence of
between the delays of the targets. Even in the idealized case noise, integration times must be increased if the resolution is
when noise is absent, the problem is not readily susceptible of a not to suffer.
mathematical analysis which is at the same time simple and The existence of the term Ts r in each denominator implies
rewarding; instead, typical cases are presented in Fig. 8, which that range discrimination will deteriorate with increasing range;
shows alternative line spectra, with the upper and lower limits of thus a separation in tosl of 2TTI(TS T) is equivalent to a range
beating, for target pairs of equal and unequal amplitude. In separation 8R = itc\2<xTs at T = 0, rising to 2TTCI3UTS at
the left-hand spectrum of each pair, the range of the nearer T = $TS and to TTCIOCTS at T = \TS.
target is such that a line occurs exactly at the peak of the spectral
envelope; the right-hand spectrum of each pair shows an inter-
mediate case. (11) CONCLUSIONS
Figs. S(a)-(d) show, respectively, separations in coBl of The foregoing analysis of a frequency-modulated continuous-
wave radar system shows that, except in the trivial cases of
T = 0 and r = Ts, the beat-note spectrum cannot be reduced
EQUAL TARGETS TARGET AMPLITUDES IN RATIO 2:1 to a single line.
An alternative method has been proposed which does permit
(a) the detection, with adequately long time-constant, of individual
lines in the beat-note spectrum. As a refinement, channel
.i ill ilia switching enables the additional ambiguity which occurs in the
Tt
region of r = 7^ to be minimized, and the use of a suitably
-SEPARATION! 4 ? chosen carrier frequency eliminates the possibility of confusion
between echoes from targets at harmonically-related ranges.
Examination of the beating between nearby target echoes
demonstrates the effect of carrier instability and suggests that,
1 j I ! even under optimum conditions, the separation between the
4 4 4 \ 4 t beat-note tones for two such targets must exceed 1 /(Ts T)
^SEPARATION 42
cycles per second by a significant factor before the positions of
the individual targets become clearly defined.
(C)
(12) ACKNOWLEDGMENT
T t T f The authors gratefully acknowledge the facilities made avail-
^SEPARATION! i able to them at the Royal Military College of Science. They
wish to thank the Dean for permission to publish the paper.
(d)
i
(13) REFERENCES
.ai 4 4
WiLa i (1) GNANALINGAM, S.: 'An Apparatus for the Detection of
4 4 4 t
'-SEPARATION \ Weak Ionospheric Echoes', Proceedings I.E.E., Paper
No. 1670, July, 1954 (101, Part III, p. 243).
Fig. 8.Typical line spectra for adjacent targets, showing upper (2) KEEP, D. N.: 'Frequency-Modulation Radar for Use in the
and lower limits of beating.
Mercantile Marine', ibid., Paper No. 1940 R, November,
The arrowheads below each line spectrum show the precise positions of WJBI for
the targets. 1955 (103 B, p. 519).
(3) TUCKER, D. G.: 'Underwater Echo-Ranging', Journal of the
$2TTI(TS - T), 27r/(7; - r), f 2TT/(7; - r), and |2T7/(7; - T); British Institution of Radio Engineers, 1956, 16, p. 243.
Fig. 8(6) corresponds to Gnanalingam's 'critical separation'. (4) KAY, L.: 'A Comparison between Pulse and Frequency
It is legitimate to conclude from the Figures that there is little Modulation Echo-Ranging Systems', ibid., 1959,19, p. 105.
probability of resolution for separation in coBl up to and (5) WOODWARD, P. M.: 'Probability and Information Theory
including 27r/(Ts T); the presence of beating indicates multiple with applications to Radar' (Pergamon Press, 1953), p. 27.