--a) a) :3 4Ji 0 _-0.3 0.5 0.7 1.0 0.3 0.5 0.7 1.0
Roll -off factor : k Roll-off factor: k
(A) (B)
Fig. 3. (A) Full cosine roll-off. (B) Half cosine roll-off. Quadrupler outputcarrier and noise power; noise power is evaluated in thevicinity of thecenterfrequencywithfrequencyslot of 1/256THz,whereT=
symbol length.
I. INTRODUCTION
[1] Gardner,F.M.(1980) Self-noise insynchronizers.
IEEE TransactionsonCommunications, 1980, COM-28,
1159-1163.
[21 INTELSAT Document(1980) TDMA/DSIsystemspecification. BG-42-65E, June 1980.
[31 Davenport, W.B., and Root, W.L.(1958) RandomSignalsand Noise.
New York:McGraw-Hill, 1958. [41 Bennett,W.R., andDavey, J.R.(1965)
Data Transmission.
New York:McGraw-Hill, 1965.
CTD MTI RADAR FILTERWITH
CHARGE-TRANSFER INEFFICIENCY
COMPENSATION
The design and implementation ofa second-order nonrecursive
moving target indication (MTI) radar filter using commercially availablecharge-transfer devicesasdelay linesaredescribed. Asimple
techniqueis included tocompensateforthedevicecharge-transfer
in-efficiency and its sensitivityisanalyzed.
Experimental laboratory tests and results in an operating radar system are reported showing the good performance of the realized
MTI radar filter.
Manuscriptreceived May21, 1981 revised December 1, 1981.
0018-9251/82/0900-0704$00.75 1982IEEE
The operation of present-day tactical and
surveillance radars in heavy clutter environments
re-quires efficient means to extract moving targets from
the clutter background. The criterion of discrimination
between thetwo typesofradar returnsignalsuses their
associate Doppler frequency shifts. Amoving target
in-dication (MTI) filter is employed to suppress the low
frequencies of the clutter echoes. The MTI filter
operates on a pulse train modulated by the received
Dopplerfrequencies obtained after demodulation ofthe
echo signals to the video range and is able to discern
moving targets even in the presence of clutter echo
signals many orders of magnitudegreater.
Inparticular, digital recursive andnonrecursiveMTI
filters have been proposed for clutter suppression in
operatingradar systems [1-3] for theirinherent stability
and flexibility. The performance of digital MTI filters
can be very satisfactory if digital words of sufficient
length are employed, especially when a large dynamic
range is required. However, the presence of
analog-to-digitalconversion andthenecessarydigital circuitry give
riseto a solution forthe MTI problem that insome
ap-plicationsmayhaveanexcessivecost orwhose technical
realizationmaybeprecludedbecause ofphysical
dimen-sions, weight, power consumption, or pulse-repetition
frequency (PRF)value.
Morerecentlythe advances incharge-transfer device
(CTD) technology have shown the technical feasibility
IEEE TRANSACTIONSON AEROSPACEAND ELECTRONIC SYSTEMS VOL. AES-18, NO. 5 SEPTEMBER 1982
REFERENCES
ofan analogdiscrete-timesolution to the MTI problem
[4,
51,
which may provide satisfactory performancewhile maintaining the characteristics of flexibility and
good performance at a low cost. CTDs can be used as
the delay lines required in first- and higher order MTI filters. However, they are not ideal delay lines and one
of the major contributions totheir nonideal behavior comes
from the charge-transfer inefficiency (CTI), i.e., from
the fraction e of the stored signal sample that remains
behind at any transfer step [6].
The design and realization of a second-order MTI
filter implemented by commercially available CTD
delay lines is described here. In Section 1I the problem
of theCTIisconsidered inparticular.Asimple and
effi-cient technique for compensating the CTI effects is il-lustrated and its sensitivity is analyzed. Section III
shows the performance of the CTI-compensated
second-orderMTI filter obtained in laboratory tests and
in connection with an operating radar system.
II. CTDMTI FILTER
in
_|~~CTD
DELAYLINE-T
EAYLNout
Fig. 1.Structureof second-ordernonrecursiveMTIfilterusingCTDs.
which corresponds to an actual impulse response
d'(t)
=(1
- NE) (t -NT)
+ N£ d(t - NT-
7)
A. CTI Compensation Technique
Inaradarsystem atthe inputof thedigitalordiscrete
MTI filter the whole-range echo signal is
sam-pled at an appropriate rate IIT, in order to prevent
un-acceptable loss indetectability [1]. Hencethe MTI filter
must delay the echo signal samples by multiples ofNT,
where N = TIT isthe total number ofecho signal
sam-ples contained inoneradarinterpulse period
T,
= /PRF.Inparticular it is known [I] thatthe operation ofa
non-recursive second-order (three-pulse) MTI filter is
char-acterized by a z-transfer function
G(z) = 1 - 2z-N + z-2
instead of the desired one
d(t) = d(t -
NT)
(5)where d(t) is the Dirac delta function.
Asa consequence of the attenuation 1 - NE of the
firstpulseandthepresenceof thesecondpulsein the
ac-tual response(4), clutter residueswillremainattheMTI
filter output of Fig. 1.
The CTI effects on the CTD delay line can be
eliminatedby cascadingasuitableequalizerfilterE(z)to
the transfer function D'(z). The equalizer E(z) has to
satisfy the relation (1)
where z1 is the delay operator associated with the
sampling frequency 1/T.
The filter (1) can be implemented by using two
N-stage CTD delay lines as shown in Fig. 1. However
theirCTIEcausesthe actualtransfer functiontodeviate
from itsideal valueZ-N.Assumingtheproduct NE< I (a
condition almost always satisfied in practical
applica-tions), it was shown in [7] that an N-length transversal
filter, designed to havea desired transfer function H(z),
when realized using CTDdeviceshaving aCTI value £,
givesrise to an actual transfer function H'(z) expressed
by
H'(z) = H(z) + E(z -
1)[dH(z)/dz].
(2)Hence for anideal delaylineD(z) = zN, the actual
transfer function D'(z) becomes
D'(z) = (1 - NE) Z-N + NFZ-(N+1) (3)
D'(z) E(z) = zN.
From(3) it follows for E(z)
E(z) = 1/[(1 - NE) + NE
z1
Z(6)
(7)
which can be easily realized as a first-order recursive
discrete filter. The resulting structure of the equalized
CTDdelay line is shown inFig. 2. The inputandoutput
waveforms ofaCDTdelayline with andwithout the CTI
compensation accomplished bytheequalizer (7)areshown
in Fig. 3. In Fig. 3(A) the output signal distortion due
to the device CTI is very clear, while in Fig. 3(B) the
outputsignal appearstohave been wellequalized. In this
test T was equal to 2 ,s and the measured product NE
was 1/12, with Nequal to 256.
0018-9251/82/0900-0705 $00.75 1982 IEEE
CORRESPONDENCE
(4)
I N STAG
- l 1 L NP
Di
Let US SLppose tnat in the CTD delay line of Fi! 2.
etqalized for a spectfic CTI value t, hechargec
Ltransf-se
loss changes to a nce x,aleI ¢ i- An. The overall
transfer function A4(z; of tthe systenm oft Fig. 2 is the
cascade
of (3),exalatetaad
for the newvalue}
, and of i4s 4E%I 11(z) (RI 1VE ) Nit z
- I1
whose inveise tranristorTn is [8)
n(4
t) [(I )/ )jd6(1
NT)C
'\,lr) "
>2 \c'l/)--eI 6K(; NI '-k
Thie
assumption Vt « i. and of course also NE7tm1pisee that ()id\ thic t'Irs' aXv U rn. ire tt practcal a
tereNt Hence )9c n-)
approx
mated
ats'n(t) oJ(1 \5 A)-: a4t;s -NT
NA A (t! ---- NJi -- II
B. Sensitivit, Anaki.Y'SI
tie
cqujatw
. -. irtnersate, otaf.,xed va'tuet-4 lt"ii w-oduct \Nir HoVQ""lC..chanPges 'in te
g&x1Ing
totemper.-atu.reo .t, A ariatiocrs, -art p
pr~pra~el a(":~'-lf. ,rCroua,ec by mewasuring at n1ierkalsthe
product
\`E anddhangintg-
the feedback oc-op weight and,ruea'nliticr ata. _a e-}rcbngix
u2A
alt akitomnalIL"'exter-fiall> ,xtnim.otieu
vIn tihe
fo~-llowling.
sensitivity analvs;ls ot htheloin-I~~~~
nIEl1
'iue: "lIMlspensatro-
eco O tito\4h ro n'imormitria (a'to ia Value is gi-V-'
j -4 _0 W*th i,u' H- '19 IFsIUt
i-(It))0)
v%hich is ot the samie ''r o
iwounequalized
d C P'delay line imptlse 1esp.-nse (4), ;xrt rcplacede by A
The restult showin in (10) max be exploited o find a
bounid
for the desiationAr0
oftthe
device ('Ti fromv.s
nomrinal salie 1:used ir the
eqiualizer,
according
to theacceptable levtl of'clutter residuets at the TI titter oun,"
put
HI1. IMPLEMENTATION AND PERFORMANCE
OF THE CTD MTII[EUER
'The complete
snie,C_ture
o-f' e-ie second-orde-rClTL)
MWrl filter including
tilc,
CIIc-ompensation
is illustratedin Fig 4 and its
perfa-rnmance w.was
verified by laboratorScts)ts arnd
in an operatingradar environmenr.
It can beobserved that
th-ie
final adder of the filterofFig
4worksor sampled
signals,
differenttly
fi-om other reported approaches [4, 51 that add signals after their reconstrutIc
I-oni)
continruouis
fo-Drm-r The
solutior of' Fig, 4 waspreferred because it leads to an accurate time quantiza
tion and
alignmnenit
of the sigrals at thetihree
adderif-puts without
requitniag
thle
presence of equaltransmrsi,_
sion networks tor the reconstruction ofthe continuous
signals along
thie
three ditferent pathsThe actual
implemenltation
of the equalizer E(z) asthe
first-order
recuLiSiVe filte-r of Fig, 2 i'sstraightfo-ward. The
operationl c`--t
does not require anyadditional
delay element like a CTD cell but is
appropriately
realized by means of a sample and-hold circuit at the
differential
atmplifier
outputOne also observes that ;n response to asynchronous
interference
pulses
theonlveffectol+f
the recursivestruc-Si 'iMRtN 9v f( I I\TE) -+ ;NT.- II 1 I -,. llcla, 1..,.-, 12 t .-,.I t. Al C. .-' c.ir,rcnsatscl-..-C' II1-1 7, I U " 1'I" 1, \- !-.) -11 -. A I I" % ., "N " I ,. . . ,. ., !; "). --.1,_". F-w() N,I '- ..-,, -,,"i N'!`..1
Input
.5
out
Fig.4. StructureofCTI-compensatedsecond-orcier nonrecursive MTI filter.
() 7.8125
Fig. 5. Frequencs response ofMTI filter with I/PRF
- l_ | lOutput
, | | _
i2 is~
Fig. 6. MTI filter iniput and output signals, simulating simultaneous presence oftarget and clutter echoes.
Kb' 0.512 ms.
ture of the equalizer is to possibly introduce unwanted
signals into distance-contiguous range bins, in case of
nonperfect delay lineequalization, leaving unaltered the
transversal MTI filter performancewith respect toradar
echoes from the same range cell at subsequent sweeps.
However, according to the results of the compensation
technique sensitivity analysis, these unwanted effects
can be made negligible.
A Reticon SAD 1024 device was used for the two
CTD delay lines of the filter of
Fig.
4, for which N256. This CTD delay line is not the best choice for radar
applicationsdue to its inherent low sampling frequency.
However, its relatively high transfer inefficiency and its
immediate availability allowed us to test the proposed
method of CTI compensation. Fig. 5 shows the
measured frequency response of the realized MTI filter
with a sampling
period
T = 2 ps, and Fig. 6 shows anexample of thefilter input and output signals. The input
waveform simulates a moving target echo with a
Dop-pler frequency equal to half the PRF embedded in a
clutter return signal extended to several range bins. The
output waveform clearly retains only the moving target
signal component. The achieved MTI filter cancellation
ratio was about 56 dB.
Tests in an operating environment were carried out
including this filter in a radar system (SMA APS-705)of
the noncoherent type. Figs. 7 and 8 show the
plan-position indicator (PPI) displays without (Figs. 7(A)
(A)
(B)
Fig. 7. PPIdisplay ofarea around Florencewithmaximum rangeof 20nm. (A) without MTI filter. (B) with MTI filter.
and 8(A)) and with (Figs. 7(B) and 8(B)) the
second-order CTD MTI filterfortwodifferent maximum ranges.
The straight lines in Fig. 7(B) are derived from echoes
coming from
adjacent
radars that wereoperatingon slightlydifferent carrier
frequencies
at the moment the test wasperformed. The straight lines are not present in Fig.
0018 9251 82090t)0707 $00.75 (eC 1982 IEEE Co)RRESP()ND NCN
7i-i
plex zeros can he synthlesized irstead of two real coinci
denit onies at z-z Therefore a wider clUtter rejectiona
region can be obtained by positioning Womplex zeros
around the point Z I
Experimental tests carried out in both a
laboiator-and ani operating raciar evironment hlave shown the
geod perfor-ma'nIce achiteved bK thec (TI-compensated
CTD second-order MT1 filter. F;'rther specific features
of this filter ire a low power consurnptioni, simall
[)h]N sicail
diilerit-t
ionsiwctuhc.h
- md(ht
ctsu c hane
Ing} the
sampimpng
rate. hmt,amh io the filtet to he usediM3 radars einphI{n tiocerid PpRho
-KNOWIlEDGMIE:NI
The authors t-hank
Florence, for his helpful
f'B1
1 g 8 PPI displayofarea paodf to Coce vit rxironm .i's i '
$(B) because the test wNas performed at a differer' tine,
In these tests the
samphlng
intersal was f I 2 us. Atthis ciock rat-
rhxaiue
the ofthe productiN
was L C EvenLhough this value is c'rnparable with mily the caipen sated MTul filter attained a good cluutter rejectwnA.m as
demonstrated by
FigsA
and 8,'where
the retaiiedechoesc-orrespoad to
b-;De
mnoving
clotids of a windsvdayxIV.- CONCLUSIONS
A second-order transversal
MITI
filter-l1
baseda,
tXeuseof
comnmrercially
asvalable CT[) delaylinses
has been-idescribed, The filter includes
s,pecfic
compensationnet-w 'rks for elirntingn3 X,he cit A c.1devilc 1
which
oftherwvise
souhdgreat]'
educe;he
Nil] filter-performnance. it$
>'mcmiIa Fn.rt'peii.aF1on
(allowsKsLCcorid orde M1T
c{-anticeler
to berel
ahzedK byithcecompaCt suruCtu t.Fi1g-2tnsteadC. twe.a-ascaded
wiltht TI UtifA'%J'i3bit() be unfiatsi5l hei.a<iuse
of, its creatermn liTD CT m
An attractixeortN wsfor str utre Ci I-ic ;5
tohat, by chan^1rigine e weghtring coeffIcients in (1 om0 M
Dr. MN. Bernabo of
SMA,
cooperatiorn.
STIEFANO BOZZI
MBER1T() (.A XilATT11 \ 110 ('.APPELI,IN1 ENRI(O() DEL RE
IARR(X() NXIARGHFRI
:titO!ti di I}LtttrO]iiv1
vtiz\/ ItIi'J ii'i-!Iic
a.[ S. NMarita
IX AN SAiJLIN i55t.1 tr,i,rj'
REFEREN(IS
Linder, RA and kiaii. (1967)
Dlgitra' ot!iVg N tiidLa
Supplement to-'E JianScwron on Aerospace and lee ironic Sr.tisers, :Ne it I96 if S-3
E1] Roecke. A,V (I9271
The- applrtn- ofregital filters thr mnovingtargetl indica tion
IEEI-c'EJascr' AIdic and Electroacusts, Ma. I 92 s--1I9
h1] MarL W alidm Wood, 1A- (19'2-t
A recursive dlgla± M1 I radar fitrer. Pro(ceeding.s-< thetIEEEL,7 Junic 972, 60. [41 Lotbenstvin, 13. and 1Ludinrgton, 1) 11975)
\ enarge .rafe UiSnpiemC11ATI) rneniation).
irn r3roceeditngs oK:) the lIE i'nternataonai' Radar Con-ferer.e VsWashingtroir, V i. >91975
i51 Btieler J eLigght e'-I odherg H.S_ Pckeuttc, CIN
and tobenstelin H_ (19 6)
Chargeransfe t 410 ieniorts for radar and EGNI
svste'tA
IEEE Journal fN Sitl-State " rcunds, laeb, 1976, SC
'-?6] Gersho, A ¼979)
Clbarge-iarin,ser tltlerngn.
Proceeding.s tI-tie EEE Feb, 1979, 67.
lt)i Re, F (1980)
Anatais a pe-sdai m on . CITi ffects an C(&C trarnisversali f [tier aepotme
[EEI' Tra'nor n)cousac, Speeckh and .SIgnal Pro cessing, Dec. 198(1'1,ASSP28.
(8h jury, .L (1964)
The'r. an AIpplication * --Trant/erra Vfefirt,ho
Nesk Yrmok Wilel's 964
'.'f t ' I ) t) 1 Fy"Ii fli R t)'