NTIS Ri

BY PIRM

,.,iCTION
;0I -N OF

INFORMAIIN CANAOA

Communications Research oo Centre

ITERATIVE JECHNIQUE FOR < ANJ /DESIGNING _INITE IMPULSE RESPONSE CHEBYSHEV W JILTERS WITH NONLINEAR DELAY,/
by
Go =.W.HerringZ

This document has bwon a.rov

for public relccyv.' c-. d salI; iLa

distribution is u dinted.

R"

"

v /./-

DEPARTMENT OF COMMUNICATIONS MINISTERE DES COMMUNICATIONS

TEHNICAL NOTE,/NO. 691

This work was sponsored by the Department of hational Defence, Research and Development Branch under Project No. 33C69.

CAN[~

OTTAWAMAR78

(
COMMUNICATIONS RESEARCH CENTRE

//

DEPARTMENT OF COMMUNICATIONS CANADA

AN ITERATIVE TECHNIQUE FOR DESIGNING FINITE IMPULSE RESPONSE CHEBYSHEV MTI FILTERS WITH NONLINEAR PHASE DELAY by R.W. Herring

(Radio and Radar Research Branch)

'o'

CRC TECHNICAL NOTE NO.691

March 1978

OTTAWA
This wcrk wss sponsored by the Department of National Defence, Research and Development Branch undsr Project No. 33C69. CAUTION

The use of this Information is permitted subject to recognition of proprietaly and patent rights.

AA,

.........

TABLE OF CONTENTS ABSTRACT. .. 1. ..... ...... ...... ...... ...... ....... ....... ...... ....... ........ ....... ...... ...... ....... ....... ...... ...... ...... ...... ...... ....... ....... .. .. 1 1 2

INTRODUCTION .. ...

2. BENEFITS OF NONLINEAR PHASE .. ...... 3. OUTLINE OF DESIGN PROCEDURE .. .....

.. 9

4. DESIGNING THE PROTOTYPE FILTER .. ... ...... 5. GAIN NORMALIZATION .. ... ...... 6. COMPUTER PROGRAMS. ... ....... ...... ... ..... ....... ...... ...... ...... ....... ...... ...... ......

... 10 ... 13 .... 13 14 14 14

7. SUMMARY .. ...... 8. ACKNOWLEDGEMENT. 9. REFERENCES. ..

....... ..... ....

APPENDIX A - PROGRAM MPMTIPSF. .... APPENDIX B

-

....... 16 ..... ..... 25 34

PROGRAM MPMTIPSFO .. ... ...... ......

APPENDIX C -SUBROUTINES. .. .....

Se-lon

AN ITERATIVE TECHNIQUE FOR DESIGNING FINITE IMPULSE RESPONSE CHEBYSHEV MTI FILTERS WITH NONLINEAR PHASE DELAY
by R.W. Herring

ABSTRACT A technique for designing finite-impulse-response (FIR) equiripple

high-pass digital filters with nonlinear phase, suitable for use in radar
moving-target-indication (MTI) systems, is described. The optimization of such filters in both the time and frequency domains is discussed, and comparisons are made between the spectral perf mnance vf nonlinear-phase filters and that of linear-phase filters designed to meet the same stopband bandwidth and attenuation specifications. In 92 out of the 94 cases exam ned, the transition bandwidth between the stopband is narrower fo. .'e nonlinear-phase filter. In the MTI application, this characteristic makes it possible to detect targets of low.r radial velocity. Listings of Fortran programs for designing these nolinear-phase filters are included.

1. INTRODUCTION The object of this note is to describe a technique for designing finite-impulse-response (FIR) digital filters with nonlinear phase-delays for use in radar moving-target-indication (MTI) systems. The optimization of such filters in both the time and frequency domains is discussed, and comparisons are made between the spectral performance of nonlinear-phase filters and that of linear-phase filters designed to meet the same stopband bandwidth and attenuation specifications. In 92 out of 94 cases examined, the transition bandwidth between the filter stopband anu passband is narrower for the nonlinear-phase filter. Ir the MTI application, this characteristic makes it possible to detect targets of lower radial velocity. It is well known from digital signal theory (e.g., (1]) that delaying a sampled data sequence by a fixed amount imposLs a shift or delay in the

2 relative phases of its spectral components which is linear with frequency. One of the great advantages of FIR filters is that they can be designed to have such a linear phase-delay, so that their only effects are to modify the magnitudes of the spectral components and to delay the sequence by a fixed amount. In contrast, infinite-impulse-response (IIR) filters and nonlinear-phase FIR filters have nonlinear phase versus frequency characteristics. The use of nonlinear-phase filters may increase the complexity of any coherent post MTI signal processing, but since the phase characteristics of these filters are deterministic, any undesirable effects due to the phase nonlinearities can be compensated. Houts and Burlage [2] have described the advantages to be derived by using Chebyshev eqiiiripple FIR filters in MTI systems, and they have published computer programs [3] for the design and evaluation of Chebyshev filters having linear phase-delay. Their design procedure is based on a computer program for designing equiripple FIR linear phase digital filters [4] using the Remez exchange algorithm.

2. BENEFITS OF NONLINEAR PHASE

The removal of the linear-phase constraint can result in improved MTI performance in both the spectral and the time domains. The parameters of interest in MTI filter design and defined in Table 1 and depicted in Figure 1. The standard definitions of equiripple FIR filter parameters (e.g., (5]) are defined in Table 2 and depicted in Figure 2. Note that the standard definitions refer to a filter designed to have unity gain in the passband, whereas for MTI applications, it may be desirable to have non-unity passband gain in order to have 0 dB white noise power gain and/or 0 d3 minimum passband gain. The improved spectral performance obtainable from nonlinear-phase equiripple FIR filters relative to linear-phase equiripple FIR filters can be realized in three different ways. First, smaller ripples in either the stopband or passband ripples (or both) can be achieved for given values of fSTOP, and N. Smaller passband ripples mean increased ASB and thus greater fPASS clutter suppression. Decreased RPB means that a higher detection threshold can be used without a loss of target visibility due to signal attenuation in the filter passband. a Second, the transition band, or the band of frequencies between fSTOP and fPASS, can be made narrower. Such a narrowing of the transition band is useful when enhanced low-velocity target visibility is desired in the presence of stationary clutter of finite bandwidth. Third, the performance of a given linear-phase filter can be approximated, except for phase, by a nonlinear-phase filter with smaller N. Thus ~enhanced incoherent integration gain can be achieved from a given fixed number of radar pulses, since a greater number of independent filtered output pulses are then available for integration [3].

TABLE 1 Definitions of Equiripple MTI Filter Parameters N h(n) f IH(f)I H(f) fSTOP fPASS fPR RpB ASB number of weights in filter impulse response (0 < n < N-i) filter impulse response Doppler frequency (Hz) absolute magnitude of the MTI filter response complex magnitude of the MTI filter response frequency of upper edge of filter stopband (Hz) frequency of lower edge of filter passband (Hz) radar pulse-repetition frequency (Hz) peak-to-peak gain ripple in filter passband (dB) minimum filter attenuation - measured from bottoms of passband ripples to peaks of stopband ripples (dB) white-noise-power gain (dB) minimum gain in filter passband (dB)

GWNP GpBM

IH(f) (d)

0 GPBM_

RPB

AS8

FILTER GAIN IN d8 VM~US

DOPPLER

FREQUENCY IN Hz

I 0 fSTOP fPASS 05 fPR

(Hz)

Figure 1. Definition of Equiripple MTI Filter Parameters

(I

4
TABLE 2

61 amplitude of passband ripple (linear) c2 amplitude of stopband ripple (linear) (other definitions as in Table 1)

it

FILTER GAIN VERSUS DOPPLER

-" , FREQUENCY IN Hz

02OII

"

(Hz)

fSTOP

fPASS

05 fPR

Figure 2. Standard Definition of Equiripple High-Pass Filter Parameters In Tables 3 and 4 there can be found examples showing each of the three forms of benefit described above. The filters summarized in Table 3 were 50 dB and values of fST0P ranging from 50 Hz to 400 Hz. designed to have ASB 4 were designed to have ASB = 30 dB and values of Table in Those summarized 2500 Hz. For fSTOP ranging from 100 Hz to 400 Hz. In both tables fPR = filters such since shown, are results no parameters design of some sets could not be designed subject to the constraints GWNP = 0 dB and GpB M = 0 dB (vide Tables 3a and 3e). The relaxation of the linear-phase constraint also allows some latitude in the choice of the filter impulse-response function (IRF), because there are several possible IRFs corresponding to a particular spectral amplitude function. Each of these IRFs corresponds to a different phase-delay characteristic, but since phase delay is of no concern, it now becomes possible to select a particular IRF on the basis of its time domain characteristics. Three possible criteria for this selection are: (I) a mini-max criterion, which attempts to equalize the magnitudes of the IRF weights by selecting that IRF having the smallest value of the ratio of the largest to smallest weights; (Ii) a criterion which minimizes susceptibility to numerical overflow by selecting that IRF having the minimum absolute sum of its weights; or (III) another criterion which attempts to equalize the magnitude of the IRF weight by selecting that IRF having the minimum variance of the magnitudes of its weights. Criteria I and III should therefore select IRFs which

I
K

are less susceptible to disruption by impulsive noise, whereas Criterion II is based on minimizing susceptibility to numerical overflow. The evaluation of the suitability of these criteria or the proposal of others remains as a topic requiring further investigation.
TABLE 3a Comparative Examples of Equiripple Finite-Impulse-Response MTI Filter Characteristics LINEAR PHASE Number of Weights N Passband Low.. Edge (Hz) fPASS Passband Ripple (dB) RPB NONLINEAR PHASE Passband Lower Edgef(Hz) PASS Passband Ripple (d8) RPB Reduction of Passband Lower Edge (Hz) Reduction in Passband Ripple (dB)

5 6 7 8 9 10 11 12 13 14 15 16

477 537 392 421 386 350 352 301 317 265 286 237

2 71 3.30 2.48 2.52 2.16 2.05 2.00 1.73 1.82 1.51 1.65 1.34

390 376 347 319 292 270 251 234 219 206 195

2.12 2.07 1.93 1.79 1.64 1.51 1.40 1.30 1.22 1.14 1.08

147 16 74 67 58 82 50 83 46 80 42

1.18 041 0.59 0.37 0.41 0.49 0.33 0.52 0.29 0.51 0.26

ISTOP

25Hz

=2500 Hz

AS8

50 dB

GWNP OdB

GPBM

OdB

TABLE 3b Comparative Examples of Equiripple Finite-Impulse-Response MTI Filter Characteristics LINEAR PHASE Number of Weights N 5 6 7 8 9 10 11 12 13 14 i 16 Passband Lower Edge (Hz) fPASS Passband Ripple (dB) RPB 4.30 3.33 3.44 2.54 2.80 2.07 2.35 1.76 2.03 1.66 1.79 1.65 NONLINEAR PHASE Passband Lower Edge (Hz) IPASS Passband Ripple (dB) RpB 3.48 2.87 2.45 2.14 1.90 1.72 1.60 1.60 1.57 1.52 1.46 1.39 Reduction of Passband Lower Edge (Hz) Reduction in Passband Ripple (dB), 0.82 0.46 0.99 0.40 0.90 0.35 0.75 0.16 0.46 0.14 0.33 0.26

718 539 556 423 455 352 386 303 337 288 300 284

576 482 416 367 332 300 282 280 274 263 252 240

142 57 140 56 123 52 104 23 63 25 48 44

STO P

50 H Z

f R PH

2 5 0 0 H z.

. .... A' S8=5d

WPN

dBGPBM

0d

TABLE 3c Comparative~ Examples of Equiripple Finite-Impulse-Response M TI Filter Characteristics LINEAR PHASE Number of N 5 6 7 8 9 10 11 12 13 14 15 16 Passband Lower Edge ( Hz) f PASS 793 542 583 537 469 477 397 419 346 372 340 335 Passband Ripple (dB) RPB 5.38 3.37 3.78 3.33 2.96 3.01 2.47 2.64 2.13 2.33 2.09 2.09 NONLINEAR PHASE Passband Lower Edge (Hz) fPASS 591 532 511 462 417 380 349 332 330 321 308 293 Passband Ripple (dB) R P8 3.67 3.37 3.13 2.84 2.56 2.31 2.11 2.01 2.00 1.95 1.88 1.79 Reduction of Passband Lower Edge (Hz) 202 10 72 75 62 9707 48 87 16 51 32 42 036 0.63 0.13 0.28 0.21 0.30
Reduction

VWeights

in Passband Ripple (dB) 1.71 0.00 0.65 0.49 0.40 1

1ST

100OHz

fP

=2500 Hz

AS13=50dB

GWNOdB

GPB

OdB

TABLE 3d Comparative Examples of Equiripple Finite-Impulse-Response MTI Filter Characteristics

Number of

Passband Lower
58

Passband Ripple (dB) R

8

Passband Lower Edge (Hzi IPA 807 648 634 580 522 479 479 458 431 407 405 397 ASB SdB

Passband Ripple (dB) RFBEdge 5.61 4.28 4.16 3.81 3.40 3.30 3.10 2.97 2.80 2.64 2.62 2.58

Passband Lower (Hz) 3 170 20 47 95 40 57 35 40 65 18 39 GPB

in Passband Ripple (dB) 0.04 1.40 -0.61 0.45 G.i4 0.15 0.50 0.28 0.34 050 0.18 0.32 OdBZ 0
'

N (z)~P Weiht Ede 5 6 7 V8 9 10 11 12 13 14 15 16 810 818 654 627 617 519 536 493 471 472 423 436

5.65 5.76 3.55 4.26 4.14 3.45 3.60 3.25 3.14 3.14 2.80 2.90 fP 2500 Hz

~SO 200 Hi
fSTOPPR

GwNPOdB

S:,

WP

PB

TABLE 3e Comparative Examples of Equiripple Finite-impulse-Resoonse MTI Filter Characteristics LINEAR PHASE Number of Weights N Passband Lower Edge (Hz) f PASS Passband Ripple (dB3) R PB NONLINEAR PHASE Passband Lower Edge (Hz) f PASS Passband Ripple R PB (dB3) Reduction of Passband Lower Edge (Hz) Reduction in Passband Ripple (dB)

I--871 8 10 11 12 13 14 15640 16 607 873 815 716 723 703 636 647 6.65 5.58 5.28 5.35 5.16 5.39 4.74 4.68 3.97 789 754 684 686 646 636 618 596 593

6.66
4.89 5.50 5.58 4.92 4.61 4.52 4.39 4.07 4.20

84 61 32 37 57 0 29 44 14

1.76 0.08 -0.30 0.43 0.55 0.87 0.35 0.61 -0.23

=TP400 Hz

f P=2500 Hz

A S8

50 dB

G WNP OdB

GP8M

OdB

TABLE 4a Comparative Examples of Equiripple, Finite-Impulse-Response MTI Filter Characteristics

4..

Number of Nmeof

LINEAR PHASE Passband Lower PA8 Passb3nd Ripple dB) RPB 2.65 3.27 485 422 428 353 374 306 332 278 298 275 100 Hz f 2.79 2.53 2.53 2.09 2.23 1.80 1.98 1.49 1.78 1.61 =2500Hlz A8

NONLINEAR PHASE Passband Lower Edge (Hz) fPS 489 455 407 365 330 302 279 270 268 262 253 243 30d(3 Passband Ripple (dB3) RP 2.78 2.61 2.35 2.12 1.91 1.75 1.62 1.56 1.55 1.53 1.48 1.43 GWNP0 d8

Rdcino Passband Lower Edge (Hz) 3 79 78 57 98 51 95 36 64 16 45 32 G B

euto in Passband Ripple (dI3) -0.13 0.66 0.44 041 0.62 0.34 0.61 0.24 0.43 -0.04 0.30 0.18 OdB

Weiht Ndg (*) 5492 6534 7 8 9 10 11 12 13 14 15 16

rr

fo,

TABLE 4b Comparative Examples of Equiripple Finite-Impulse-Response MTI Filter Characteristics LINEAR PHASE Number of Weights N 5 6 7 8 9 10 11 12 13 14 y 15 16 1 Passband Lower Edge {H) I
1

NONLINEAR PHASE Pasband Lower Edge 1Hz) fPAS Pasband Ripple (dB) R Reduction of Passband Lower Edge (Hz) 172 46 74 53 46 85 22 48 29 37 so 22 Reduction in Passband Ripple (dB) 1.26 -0.18 0.61 0.37 0.36 0.58 0.19 0.35 0.21 0.27 0.34 0.17

Passband Ripple (dB) R

PASS

RB
4.97 3.46 3.73 3.28 3.03 3.05 2.61 2.74 2.52 2.48 2.46 2.28 594 503 505 469 428 394 387 380 366 349 334 331

AS

PB
3.71 3.64 3.12 2.91 2.67 2.47 2.42 2.39 2.31 2.21 2.12 2.11

766 549 479 521 474 479 409 428 395 386 384 353

STPPG8WPpBM fSTOP 200 Hz

fPR

500 Hz

ASB

30 d8

GWNP

0dB

Od8

Comparative Examples of Equiripple Finite-Impulse-Response MTI Filter Characteristics LINEAR PHASE Number of Weights N 5 6 7 8 9 10 11 12 13 14 15 16 Passband Lower Edge IHz) fPASS 822 829 773 600 669 645 688 597 584 548 555 548 Passband Ripple (d8)RPe 5.85 5.95 5.06 4.66 4.75 4.55 4.17 4.24 4.15 3.91 3.96 3.90 NONLINEAR PHASE Passband Lower Edge (Hzj fPASS 822 759 697 630 630 593 579 567 543 542 530 518 Passbaisd Ripple (dB) RpB 5.84 4.66 4.85 4.97 4.35 4.10 4.00 3.92 3.77 3.75 1.68 3.55 Reduction of Passband Lower Edge (Hz) 0 70 76 -30 39 52 9 30 41 6 25 30 Reduction in Passband Ripple (dB) 0.01 1.29 0.21 -0.31 0.40 0.45 0.17 0.32 0.38 0.16 0.28 0.35

TABLE 4c

I STOP

00 Hz

fPR

500 Hz

A S=30 dB

GWN, =dB

G PM

0d8

9 3. OUTLINE OF THE DESIGN PROCEDURE

The technique for designing nonlinear-phase FIR filters is based on a procedure first suggested in [6]. This procedure involves the design of a prototype linear-phase FIR filter having an IRF h (n) of length (2N-1) and an amplitude response in the frequency domain Hp(f) equal to the square of the magnitude of the desired frequency response. It is shown below that this prototype filter cannot be designed directly from the desired parameters of the nonlinear-phase filter, but that an iterative technique must be used. A property of linear-phase FIR filters with real-valued IRFs is that if zo is a zero of the IRF, then so are zo-l, z* and (zol)* where * denotes complex conjugate ([11; page 159). Note that if zo is real, then zo = zo and also, if Izol = 1, then zo-1 = z*, so that roots can be real and single, or occur in conjugate pairs (if IZol = 1) or in reciprocal pairs (if zo is real), or in conjugate-reciprocal quartets (if Izol 0 1 and zo complex). In general, of the 2(N-1) zeroes of the (2N-l)-length prototype filter, there are k pairs of reciprocal real zeroes, 9 quartets of conjugate reciprocal zeroes and = (N-k-29.-l) pairs of double zeroes on the unit circle. To extract an iRF h(n) of length A, it is necessary to discard one zero of each of the m pairs of double zeroes, one zero of each of the k pairs of reciprocal zeroes, and one pair of conjugate 7eroes from each of the Z quartets of conjugatereciprocal zeroes. This leaves a set of (N-i) zeroes which is expanded to produce a real-valued IRF of length N. A set of M IRFs of length N can thus be derived from the (2N-l)-length prototype filter, where
M=

Lm

Note that one zero or pair of conjugate zeroes can be arbitrarily chosen to lie inside or outside the unit circle, since the only effect of this choice is to reverse the IRFs in the time domain; i.e., h(n) is replaced by h(N-l-n) for 0 < n < N-I. An upper bound on M as a function of N is given by
NB=2{ [(N-2)/2-1}

where (x] denotes the largest integer less than x, so it can be seen that the set of IRFs to be examined can contain of order 210 for N=24. Hence the optimization techniques described above can become quite time-consuming for filters of such length. This limitation, however, should not preclude the use of these procedures for designing filters of the lengths usually considered for MTI applications (e.g., N < 20; so that M < 256).

4. DESIGNING THE PROTOTYPE FILTER

In order to make use of the standard filter design algorithms, it is first necessary to define the filter ripple parameters (RPB,ASB) of interest to radar KTI designers in terms of the standard linear ripple parameters

Ita

1.0

(EI,"2)

It can easily be shown from Figures 1 and 2 that

10
_1 10 and

(RpB/20)
(R /20) +1

. I.

C ~ sB/20) 2 :10(A

(2)

N,

In the standard design procedure for linear-phase FIR filters [3], STOP' fPASS and fPR are specified, then only the ratio
W

if

ei/C 2 C

(3)

can be specified as a free parameter. This restriction can be circumvented by allowing fPASS to be varied in an iterative manner until that value for
fPASS is found which gives the desired values for C2 and hence el [3).4

A similar iterative procedure is necessary in the design of nonlinearphase FIR filters. For the design of the linear-phase prototype filter Ho(f) (see Figures 3(a)-3(c)), it is necessary to specify 2N-1, fSTOP' fPR, 61 and 62, where 61 and 62 have to be specified in terms of el and C2 . The outline of the scheme for relating the V's and c's is pictured in Figures 3(a)-3(c) and is similar to that of [6] except for one detail pointed out below. For clarity the scheme is described progressing from the original prototype filter Ho(f) (Figure 3(a)) to the intermediat.e filter HT(f) (Figure 3(b)) to the final prototype filter H (f) = IH(f)Il (Figure 3(c)). In practice, the actual. progression is from specifying H (f) to specifying Ho(f) in terms of H (f), since 1o(f) is the filter which Ts actually designed using the algorithm of (4]. H 1 (f) is related to H.(f) by the transfoimation H (f)
=

H (f) + 62"

(4)

In the time domain this is equivalent to

h h( (n)

=ho(n),

1 < Inl n = 0.

<

N-I

(5)

1ho(n) + 62,

H (f) is in turn related to H (f) by the transformation p H (f) pI which becomes in the time domain
= K H

(f)

(6)

11

H 0 (f)

"!

<

0 -- T2

f(HZ)
05 fPR

~PASS

Figure 3(a). Original Linear-Phase Prototype Filter Frequency Response

H1 lf)

HIM

00 fSTOP fPASS
0

f(Hz) .5 f PR

Figure 3(b)

Intermediate Linear-Phase Prototype Filter Frequency Response

--

------

~-

12

!Hp

fM - !H (f )I

A 2_

)y.

~~C

1-

---

?ii
. j

STOP

fPASS

0.5 f PR

Figure 3(b). Linear-Phase Prototype Filter Frequency Response

h (n p

Kh1 (n),

-N+1 <n <N-1

(7)

where K is a scaling factor. It is in the determination of the factor K that this development differs from that of [6], for it can be noted that a scaling by the factor (1+62)-l does not produce a filter IH(f)I with equal (linear) magnitude ripple excursions above and below unity gain. Since it is the par ameters RPB and ASB and thus el and E2 which are specified by the filter designer, it is necessary to invert the above functional relationship to derive expressions for 61 and 62 in terms of cl and E2. This being done, it is a straightforward matter to use the linearphase FIR filter algorithm to determine the (2N-1)-length IRF ho(n). It can be seen from eqn. (6) and Figures 3(b) and 3(c) that (1+ (1c and
C 2

el2
2

K(l + 61 + 6,,) =K(l -6 11 +6)

(8) (9)

2I

=2K 6 2

(10)

Some algebraic manipulation shows that

tv

1<13
61

2 +

2E 2

-E 2

(11)

1 2 2 and 2 + 2e2

E2

(12)

22

It is now a simple matter to derive hp(n) from ho(n) in terms of these factors as outlined above, and then to derive a nonlinear-phase N-length IRF h(n).

5. GAIN NORMALIZATION
It is often useful to design MTI filters which have 0 dB white-noise gain. This means that the white-noise power in the filtered signal is the same as that in the unfiltered input signal, so that in the absence of clutter, the false-alarm probability PFA is not altered. Such normalization is easily accomplished (3] by invoking Parseval's theorem

!
g2 2 =

n-i

0f.5f

Ih n)H j
n=O o

f2' PR

IHf f

df

(14)

to compute the white noise power gain g2 by means of a simple summation. Scaling h(n) by g-i produces a filter with unity white noise power gain (GwNp = 0 dB) and leaves ASB and RPB unaltered. Note that this procedure is valid for any FIR filter. It can also be useful to design a filter which has both GWNP = 0 dB and a minimum gain of unity in its passband (GPBM = 0 dB). This means that the probability of detection P in the passband will not be reduced by filter attenuation in these passband ripples. This goal can be achieved by an iterative procedure. In this case it is the passband ripple that is altered until the desired characteristics are obtained. The procedure is to specify N, ASB and fSB, select an arbitrary reasonable value for RPB and design a filter using the iterative procedure described above. Based on whether this filter has GPBM greater or less than unity gain, the value for RPB is increased or decreased respectively and another iteration is carried out. This procedure is repeated until convergence to GPBM = 0 dB is achieved to within an acceptable tolerance.

6. COMPUTER PROGRAMS
The program MI,,1TlPSF generates nonlinear-phase FIR MTI filters for specified values of N, fSTOP' fPR ASB and RpB. Parameters to specify the

14

grid density for the design of the prototype filter [4) and to select the impulse-response design criterion must also be provided. This program is listed in Appendix A, and the detailed operating instructions are given there. Typical times required to design a filter using a Xerox Sigma 9 computer range from 2.2 sec for N=5 to 28.4 sec for N-16. The program MPMTIPSFO generates nonlinear-phase FIR 14TI filters with GpB M = 0 dB. The same parameters as for the program MPMTIPSF must be specified by the user, but here the parameter RPB is used only as an initial value which is then modified by the program to converge to GPBM = 0 dB. Hence the choice of a starting value for RpB near its final value can significantly reduce the time required to design a filter. This program is listed in Appendix B. Typical running times on the Sigma 9 can be 1 to 10 or more times as long as for the program MPMTIPSF, depending on how good an estimate for the initial value of R B is used. Appendix C contains the listings of the subroutines required by the programs MPMTIPSF and MPMTIPSFO. The subroutine REMEZ and its ancillary subroutines have noz been included, since they are readily accessible elsewh re [4], but note the modifications required in the dimensions of the COMMON block variables.

7. SUMMARY
It has been shown in Tables 3 and 4 that, for the typically narrow stopbands used in radar MTI filters, nonlinear-phase filters can be designed which offer superior visibility for low-velocity targets, relative to that offered by linear-phase filters designed to the same stopband specifications. It has aiso been pointed out that the time-domain response of such nonlinear-phase NTI filters can be optimize~d without altering the filter power response in the frequency domain. The details of the design procedure have been described, and listings of Fortran programs for implementing the procedure have been provided.

8. ACKNOWLEDGEMENT
This work is supported by the Departnent of National Defence under Research and Development Branch Project No 33C69.

9. REFERENCES
1. Oppenheim, A.V. and R.W. Schafer, Digital Signal Processing. Hall, Inc. Englewood Cliffs, New Jersey, 1975. Prentice-

2.

Houts, R.C. and R.W. Burlage, Mfaximizing the Usable Bandwidth of MTT SignaZ Processors, IEEE Trans. Aerospace Electronic Syst-ms AES-13, No. 1, January 1977, 48-55.

15 3.
IMoving

Burlage, D.W. and R.C. Houts, Design Technique for Improved Bandwidth Target Indica*or Processors in Surface Radars. Report No. RE-TR-75-35, US Army 4issile Command, Redstone Arsenal, Alabama, June 1975. McClellan, J.H., T.W. Parks and L.R. Rabiner, A Computer Program for Designing 9ptimum FIR Linear Phase Digital Filters. IEEE Trans. Audio Electroacoust. AU-21, No. 6, December 1973, 506-526. Herrmann, 0., Design of Nonrecursive Digital Filters with Linear Phase. Electronics Letters 6, No. 11, 28 May 1970, 328-329. Herrmann, 0. and W. Schuessler, Design of Nonrecursive Digital Filters with Minim= m Phase. Electronics Letters 6, No. 11, 28 May 1970.

4.

5. 6.

k
,'

16

APPENDIX A

PROGRAM MPMTIPSF

iI

THIS PAGE IS BEST QUALITY FBCTICABLE D~~~n= okIy BUBtSHSD TO DQQ G

17

*z

U-Jrw CM3 LUZ u 04 US-

:> WE r Li u-

EL

j 0)

cc

LLm

Q:
C CC cr-

*..Dw * 0 o --

j
LUc. NJC * 1 *2: w-

a a

<r

nw Ic

z
CH40 < C C

ii ..i~ CI
-1
-

*)*

-jm E
*C*N LU ; r 54* j M La = D* i:,

zU t
(. Li. z
*7

u CC

* C

C Wd

Z CC-4

cN0 ,

-.

N~ xW -4 Z U-t CD M CN.-

i,-

c
ii-'

rnr
t-'_ WD
I- i

* cccc *LUWLUD =!

-IH
-C

in.a...UcnZ

*m W:>

*S mf iZiWC * owo -wc *L *XZwLC

rSO

owM J

UWW

*ST M Z

*i*ED F-i -CWLTU...J.3N >--iC*

-itr

i-i

r-

ccW*

CC a C*

F -

CCWL*~ F- U*cz*

*w

C CL 0 C -i*'i

zc~ CD0)

U
w.

-Sf

f*1-o

0 CO .1-WIL 0

zi iXU-ScolWC w~i.1 wzz WQ z 0-CS u-LZ~' C; "-w~rc -Z6 *C X-i -4r4C 1 W04 WZ-C 0 CC
004

W )

c3c:c 0'Z

ccc
xC

w
*

CC

CSESm*-l

0C

D
U

--C Sr m

~*ZC ZC
*W
*tl0

Z C

W3
1-

Li.

IXJCCCO I-'-*-i-i-*C0C 7 Ul no fC~ -i--zC< ccLw zzz* wcc i aa. CW C-L C0)flL C i-l"---J*X1ii4i -

(n -A1a.

1-4

LC iZW 4T C -~ e LIW. ~ C ZU)~ C u )W WilCi> J0)-cJ *a ca H 0)o WW il 0. IWL)C C 0-WaZ - Or)n _ co-i "cc r~a co ccui iLi z _C 0)0)i !C. CULL C 0~ < -L0) 1 *n . Z>-ICC-ccfc *C D O Li. WI-:,l0Z 74WJ '-4 I- =WCZ WWW-f Z -C CC: CCC-1 * 14 Y aCD ZDrLU C1 CC)CC " .rw CC CCI--0)siw jC*L3ILU1 *C4 (0D _j -iC0 LW CD(DSCCWreWm0)0OJOO*LL-J LCc 5L 50) *C)C[ z S C -- 1-10J -j 0 acf! l-fS * -i -iw CCF-a.DF-CWSLL1..L*a.Zai DCC *0 LUc-40)ZUC _ * :3 3 *I-qi--i iWw w-LULU WZ D?-a C-L <.Ca Ma0-4-ii t z 5. * ;--a-C ITC *N." ii.-4 WZ *L * ~ * C: *C z~~~~ *iS S

a.

cc
L S 1-iW 14
WI Z

*C *C

i <-COw -i-1cc Z m-C-4mi> <l- iC

!X

Li. Ci I-- a. C C Cm cc cr

*L. *o C * *z t *0 *w *C*

zC

- I C C W DL C * C C WECC U *aLC----S <C MLU a r.. * a WI D cf LiU a*0 u D

liCL0~i'.I) uj SWO

-o 05w
I

F -4 zuL 5 z-

C, CC a.

)-

cc :3

CCC 4

CCCLU 01Ci-N' m: 0I-aii c-i 0 *

- -

* *, O' *i *** ** oNmo * * T0 aNm0 c. j ,m ; q

*uuu.Uuuuuuuuuuuuuu *TW oNm0 ,m e mv

c**** .**********jo

18 ThIS pAGE IS BEST QUAIiITy pRACTICABLB &-ODCOPFY FUIJlSHMD TO DDQ

F-

3z

<1

LI-I

0'~~C
-0

LJ C

L
C

C Lu im

:> (n

CL IN,. Z~ -

X-L

~L
**
-

0'0

0-

WC~4~

4'n 0

NN-CZZL <

W~'r U,
C4

1-N
L-

'r

~C~
.4 L*M-

WE

U 1 z

co<

X4

0 0

-iF-

Z J

U- 1.0 0C0C i..L.L. wW La)-

~Dfl-E~
*EO- I iM-41..U i- -i.L. 4w liz C C-4ll L
L
L

C 0
;c
-~-4 F - X

N L

NC74 UW U Z wC xilcc--fl~ J-. z CCC z ZZZ

t~~~E-

W-N J

*cI,-jt-, cWiw F .0- , * -X-CwNw 7 "I -C, *c-.-;-iq '1- * r -W -4I- LL *

0 *C -

CNEj

IX

I-F

lZ
J C4.

CD

C0 LZ

CZ a
U

Di.

-Z

>I4Ji

ww

w
=~
1"

~
F4

~
X -

1111 1 11 11 W3
C
: iN I-U

_jZjc
L 0

C).
fI

Mf

C41

?' f)'

N M 0' 0

-1

n'

in 0 N

0'

0as

.I'

0f

NI

n M 4) rr I'0

N 0M 0. 0

IT4 V4'It4

TV

0000

o,.-C

0G

-_

aNNNNNNNNNNM

~
C

*.f3.

*3~o-

BEST QUALITY

17

z*

*7

17-

-a

*n

I-

Q coiif 93 0zl'-0

U0

C.;

XTl

0 CC~ Z~~* 0nC

M LLJO-

L.!6.-

*- LL WLU -- F- nZ 0
L)

I0

z~
6
M 0 W'.i

0
N- N
C.C-MCi14'-4

--a~

FLi. <r N --

0 ILL

0
W
-

iC. D 0~c

Lii En EDC

0:: ED

I-HO C; < -i

J&C NdUlZ +

CL

~i~l~...:>*
<:u(. L

+ 0L M i2 :0

1i CD

Q0[I I'- b)

0 z
I-

'+
1f0~i.

-C

--

If 11C CD wu)
LLI F-

FLi) l11-00-4' C LiW 0 I UWr H

-'--)

H-.1
e

O*

-0 *L

i
Il

**4

LL 0. IIIW-

*II'-t C;_j LU-4 0

z
*r

F-L. wjW LLLi

HLj..DLC

Li) Liq 0

Li.

H S ,-

J 0 0

UH.J.Jiii. UiJ

000-

O L-0--3 I-4

--

-4

% -q 1?

LL "~ -4LC.)~C H -

,i N

V-10 11) q-

20

THIS PAGE IS BPST QUALiITY PEACTICABLE YPROM COPY FLW'JISHa&TO DPC

'00

ui

zI

+
w M%

'

D: m)

U. C
a--

0
N, w.

N i C

C CL 0
cc +i

*C

C-

CC

N)

C.
-

0Y) z

w WD
X m

cCL C

Z 2

Vi

0
-

w
00

0C '- JHnH riC < -. a:

F- Zi C

xi

IX

-; CL !X' sO (r C MCCIM 0. Cl 0 -wo wi -_1 N ix=mF -wL CC -t W jL CC H CL UawI

iI

<

jL

C:C'
.

q4
a . M;. *

r
IM- C4

r Si' uo
r4

i j

H 2
1 N-

o
.
L

x
0
c

3u
-

CC4 a
m

1 Hi

Ce CL4
C

0T
-_ y

r
;-1

34 xC *0
-

11
na

()

C Lu
LL

F-

N
)f
-'--C--

+ -- L)
U E

CC
-

C4~

w . CLi
ZC

1-qCri.20 ix0C<

w C

w l aL ci. UC -I1iL 1)

Lu <E-

CC

i-C

-i C -.

4x -

F+

)
-44

CL

Lu

.0

ui H

Ui H

1 U

I U

C .

ui u

-2

(0C -4

'-CC v- C'U 1- -'--

C1 --

O'I

LD

!C~I

Irr-4CC

UI

U)U02

TjjjSPAGE21
r4W OY Ua$HWT

X L

t
if4 CON

zI'
Ic

I-i C'

ix ULL
ca1 w

C!
L.J

U-<-

rr

x
-

.4
:-,L.

Hi'
r,
44C

C Cd
i; Z
Z

ZCC H

N,--CC

L-C

cLt

>-

-J

m--

N-- i.

Z
LL Lt

zU

f-r 0 -. 1

__z

-.
-SE

LCE:.ZC HH~ ~~~

cc
f:C L,

H~
LL; -CA CWs

cc
-k-

--

'

z>CZ
Ef CC CC

C-Z Z

'

z xoa
'--i7*

~
..

Z
' L

uC a

-, Z F-

z--i

ZJs-0X

0H4'7

-H'-S

(C

U-If:1'--

s.o C45 enx CD 17 O* WI

LL LL H-,Z

11
*!.

LOF
'DC

6"Z M

a.

IF L

oT 0L

*Jr
W

C
ZCl L-C
Cc

-a

wr-'

C'

3!3S3wCw
74
'-

i-

C x

-4

r
*'-

C w'-

" G

' H'-

Ci

Sx a

x
?

"c"Mr
LC

N
E5 Lo u.

CL
L

oI

ON

HL

HO~11 ir.

Z-Hn

-M.

z U-

'

100110--0O,

N CLOCC 000C
-

-iU NNu

''1 4'1
N

U-N N~

N N NL m L

CCCMCTZ-3C
) U

T 0
xm0

MC'

HmwI 'C O_ 0-c-

)C

N MO&

-4

22

THIS~ PAGE IS DtST QUALITY PRACTICABLE Q momQ Gary IU.cclziw TO P)9

xA
L.7

4-

4'C

,.

LL

ED

--

c-; m

N-C

-l

CdN
LL 4C En

le

L-L

-L

_j'
w7 N

nc
3.

Ci *; W
C
W LL.-A<

'-CC

c. c.
N

Z~

L t
U) ;--r,

-cir)
.

3
o.W

I-~~:
I-

C.'- usiUr '014C

z--

CC LL'. 0NL :3-

)L
C

17 1- 1
cr

cC-1
Li

0-

WL
CL-ti

LLJO

UOz...

-*Z)
C

= ZZ C s U
N+.OZ

Z-17 :40
-. I

(2' c-

WE
LLL -C 0U0h '-a I
LLc c

CO

L 0,'4 X CO C..
*-r-

0
<

Z-C
--

LL 7 '. L-1LI '

0.4 A-q

-C

C(2j

dM Z0Ci X

W~-

WI

'-

L I0
z

-i
i'--

0* 3- ceC O'Lt H
-C

W N.
r--

-Z

MZ 11

L.

+
'--

N!jflZZc 7S'-i z C*--CifN cZ. J-- Oc :3

X

N CcZZ S z
-

WCH'-

Cc

CS u+1

W4 *L INW

-c-xC'a

>r

ill: U)ICOF- z-- CIN

Z Z~-~'

oi-x* LL.

_r .' a CU C4- C-a.--x'c

N 0
-3

ono-' UL3zn

00oZDC =

Z z z
N

0000J'-*N 04C)-W ,C

'0 ;71NC

-~ -DI . '4 N

--

'

--

C-.

CCNN **

-i'*
N7;

C<'4
Cz

XX.V!f'
'-N)If N

Totsl PAGE IS BEST Q LIT ?ATIAL CYBUISa DoDO ...-

23

aa

-0

LL4

Nz

z4

c nUct

CL

.z

co

C 2.2
z

NZ
C ' :

2. --

., I-- s-.-

ILL aC
-4

-nz c

'

c.
N

--

'S ~Z,.szzo

NI-NNK c

c1

4-.~
_ _ 'm -4

4.7-Z

r-

N.

oxr--

4L-

LE:-

Ca

a,

:%

NN

IIV

24

THIS PAGE IS BtST QUALITY MR&QXCABLE

RhM COrY RalbH.Md TD DDC
.

u-

F-ii

[M
W
_jL

W XZ-C
(' ' ' z '1(4C

:1
z3
z 11 z k-i I

1'
t
A

25

54

4. N

I Il

.5'

(.1

A'

APPENDIX B
I ~

PROGRAM MPMTIPSFO

L
Is

I.

K K

~5

(5

~....~

,~w

-~

26

TUIS PAGE ijo DtTQALT

i~CIAB4

*z

<E U WI *L E *-* L * 1.*W Z

M-

~
Z. *
"-4

174.

C LO r i. :> 0Z
ce~* '-l LLI-

i (le
LL

* *~
z

zC
Co ix'CL)

t~.
mee

~~

W3
zC

>w
..0
nU z

-i_j -a D. Z

lwce
CLJcflZ= 1:41mu.
i-i

* e C -*w.a

jl:

0.W co

CW*-4-CC

~
IUN

Li

iL
ze

W Z
0h x

Co Leo
WW:

*C L

'Wcc cDU)in Q *U).d

C3 W f-izW En Z U C J

74i e-Z
'xw

CC
a~-C~ C
aMZE I

(n'i*
r-

wUC~UX

W-

U *W

IW-Li

SM U
i-

W M 0 U

*WC-0 UCZ=w 0 W*17 X0 Ix -I " H e zZ * i= Z LL. r. CO*<tC * PCL w *Z1rzi w *'zU <Z CJC -< f-

U X U CF* LW U ~Z Z - W ZIiC M U)-J M ~ e~ W0 -141- x14* ~ ~~~C M* LL Z C4I Z. '-'U M iiC 0 za X e 0- a4 r-- 00~ I-ILOCD 3 z
4.J

X*
0-4I-i =* *

w CeC-m CD <0

Z
'i

P L W0L

WZ

M<l
C

o pUt'~ WO, *W *:41=1x -4

ZU)m

I-i - -- 4z W LL Z

F- I

OWC

le-

Lz.iW CL
< -WWC~

'-

Z-.LLZXXX*I--~

017
Z i*Z-i

W XQN
CAIi4

-4

~Z LLU
Q: U- n
0

11>- 10-1 "-.i

WW3 C

U 0 0 -I

F- F- H-

CLWL =

C i")

'N a
*

.Cc:

mD W

*0cI4 < * Le. LT4

U)U

-1

11 p 7-

CD (D

0 z CLO
Ce

j ..

L.. W CIL

-.

0C

W"a', mw

I-- mw

0CL

pZZ7

iX i

WW

1! i 11

aJW*

LL L

c LL CL 3

C Z m X LLLLo 0i~ mCL e <L-~ -LED eW C *0 mD C LW - oM U Z -*-O14j 4Z CO Z U) M U M~~ ~ ~ ~ 1:Zq I... C WZ xCL~fcI--4.ZE W x

cc: urj 40 La

0CrWCL x 3mu w ~w *O U 10MC .wfI

! r Liq1-

c*0
z

0-L *U U

*iLL

L-4

'-ow

i-M--a lr(rL.0 L> IL>

U U Ur-UU U- X z 1 1 CC

U U

"W<'

U U U U U U) U4C U U4 U U-U U UL 0 W0 Jt NW0 x w *u -1 = f--

UUUU NI)

')

*0 W

TIS PAGE IS BBST QUA~iITY MnOiicIAB27

4C
i I

-j

1-

z
0

:c

Cz

X I-':3 0D 0 - L0 40H=

1:)

LL C U) z zzz c uj4 c .4 o

<W 30 l'J -0W 0 Lai W * 0 0f L

05-Cm zC4 N-~-C z r0

CL I U,

i->HE4)UUU)) C
C WWW LoEn!M-1

i- W _Q-j-~. 0Z C
*

CC
Nw Ze :

Z Z

~
** Z

I-

. LL L V) 1L In LLL t V) LLL -JW

a.CK-<* I L OW I -) C +

WN f4UUW U Zl-O w C N L J U. 0 cl.J Z Z~ Z ixC~~ clL w OH NIHZ 0 0 0 0 C L CL CL M ~-f-4 x CLI 110 :3ji j0 C-1 11 1- L11 0 Ix43 CL.

~WN,
q

I- -4

L )J -41HLL-i X

L (-4Lr

L. cc: f4 l*

rk--

28

THIS PAGE IS BtST QiJALITY PRACTICABLE FBA~ QOPY FUB=ISHED TO DDCQ

4-

CL

CJ
X4 0
(L InI

-I

to.

--

0 C(n
CL 1

*
-

* -I

1'-'.
0 !:u L IP

um
103

0-Z0 -'
**0

Li Z IL

N 1: + ' ><0 IL 0-'OI

0.

11.
174 JtZ aC

-++rF
01

IN

II

;-L

X
J M0 M0

CC. LL

14

+F- 0 cIIx -)<).Z

co+ i

C*-Y_

X. U)

7*1*16

z U IL
"I

>( C z
z

0 IL
W
1:4

N~ JCO LL U3 -4 W*
CLL

F--JZO

0 !7 I-z
;T

-cr 0
0

Uc5)'CO

+ +ImI- F-C !X LAJ14 i i~'-U 0

1 1z

o~+* 0 0...j
74L

4 I LOL

**Z-C3.O

0
-4

CD
4-q

It) L 111 ,:4

-. 1" W -- i.4 "t _jP" C. J < NzIx c 0 0 -CD f.C 0 _ 1= _j 1-._CA*.1-~~ C IL. Er
1-'-

iMW 0

Ct) L < CL U1-

3
>

-.4-.1

'0 0 X
.1 -1 4

(D zI

CO

Li I.D (.

j4

C47 CC3 co) + I If BIIZIX

0

i-i ICt W, z
<1

O0 -04 jW - W 4 -(. -4 -w CDC,-OCJL+C*. Too 000'-ozi -411011

- -0*

in

--

-j

-.1

11
Z-

U CL C Mi I ix

(.t)
M >-J.

C 0X 0
CL

LOW

-I

(N~ Ci

rtC). 1i LL
IL-

w U) I-ILDIxLCDI
14 -4 IIIW

IU
U X

M--O~J co
L) U 1:'

.4

M-

(fU IIIII 1 Z1 .!JC1L !I--'1-jaJ-..JIx .1.1 *J.1 3.) I C4-(.iC L iL a IL4LiWILL

N- M" -t nf

"0 N XO0 C., .w m0m mw

N\ M~ '~T U) -0 N or, 0'- 0

Of0% 0 -

q- N~ M~ T V) 'a N CO0, 0
'sos

N~ M'* rl-

If 0t

N MO 0' 0

TISZ PAU IS BEST QUABiITY MRCTkqA FMG~ OPY FUMIS1DTO DDQ

29 29

LL

CL 3: Cr
U. r:

z Ld CL

zZ

c -4

L' 1-4Cf

fL

II L

CL

Zcr =
0
L

LCi *c
CL

W,

>(q
-

o",LZ
:3

U).
.

e
U

Z''
*a

Cr '-x-z-m
WF

I-i-<

Li
F4

Lj

IL

lz

-%-

+-

CIC

LL.

I 0W O

Ce +CJ En
+

L~

LL ZLL

ZZ

Li

F-0

IC

L: CLC I 0 C-

0,. 0' T

*0
crL)

1- L CL~a - C
Xf

<E 0

I
'a

0 0 0 1111111~~~

0 0 1

0 I

0 0 1 LL1

-T,. I

0 o11I

COF

MO= Cr

a-

U LLX

U,

30

THIS PAGE IS M~T QUALITY PAQTXA=B

cr;

c OW
1-4I4

2 cf

CL

-'Wc'

_
LL

z
N~~~x Lu W Ii

*t

i--

CL.
C-

0
WL.

I (.

WL rCr +i .. W=4

W (I C4 C
C *4

_e
cn
U-

2
CL

0J

w
n.
EC~

to
mt~-

i74.CZ 02 ziE C30 0t. -

.- L.
=

tr
2
CL-C

0~
'--

0Z. EW.J W.) w-l 0

0 cz
u. Li..

QW'41

CL

Cflrr WW +! iij

ClW
+ W

mlf -0 0 ~ 0 Lo CoE
EEW * LLL 0-. cr.u

W.J

>-W C.. M-' i.N

-)

WO WI i ce 1W 1-0 WW D r.. L *Lu Ii)O

1

1 0' 0 Li. 0~ U-f CL CICCJ 0 0 Z:

.-

.. 0

01 0 WI-CExC

w 0 0

CO

... i3-.*CJ WC 0 L.C0 CUHW 0IiJCLC (CXLW U*XJD ml If

ctJ

Z
OCZ z<C M CD r w
_j3 -jcr-

ix0Q b-J 0''Z

C -+*l <) 174 111 Wu

i < C

-C. I

Li.. W3

* 000z

' ccCICJ-0I
-I 0 11 Q 0Jw :0 -,'c -iL

CL

LL(\ N. -j i NI10()1I' .JL) 01014(n1 U

z ED D

0 C .

.- -E r

-xL CL)

L L
Fi

-4 cr 0 MCL OZO 3 LL * C .Jii-i m -

I.-

L.

U..
CL.
wI

OL) Z

c:2 M. u1L-C uwcz: mCL

HL.

0(1 En

E
ix

NW

Q r*u W CL.I0

NC JCL 00

0 EnEn

ez

=0

--

'4--p

L~

EFOM OPY FUlRNISIE

THIS PAGE IS BEST QUALITY PRACTICABLE TD DDQ

.~

31

S.

<

W, w -cr
z~'

Nr
w w. w' ;=

N Z-C14c;L
0
'L *

ciw

I--

ED

to ww z2
Z

CC f-, z LiL 0f
V

NW, Z C~ Li ixii
.
N

CL

Cf- -'. CLrN I. CL WW r- ui

X..

C-4i4 In
-

ii

CL U.

-I
-iC

z<~. '
L L
i.

z r-i.L H.W
-EEr

. z

-W= X DW)'
Il N,-.Cl* Ze

-W, 0

-ZLA..-3 cp X4C~~u N

0r 14 WCL
W

)X

oG I I

2.j 0
-

LC Z C

-1 000

-CL3uj-i

z
C

CL COU

N
0~ Cc

o1 U'-1 En'-1W O
-SL

ml >(JZ2_. L H
En

xW

0.2 000 i xWC C--K-ti>Ujf -.Z-xWLC 0

I.2'

LL LL~ U CL *COC 0.3 M Z4 Cp

ia,

a- CD

a 41 < Z Z-'C)- m D
",L X

Cc

N W1 COC M F- a, 00

Z-

'J

0-

:- 'CO 4:

WC

- C * v)4, ce W:;? N0001 m 0' X' '-0 -'N C;C Z~CO 0 L N U,.< * (0(3 3 0) 11 if =3< Howa H
0) CD

ZH 1- 00 m
om'

C .

(asOI-ZZCDC ON~ ~~2 -. 0' N LrC4 0'Cm-.to4T EnZq 0sdIM-xx

-0 *1

-- C0)X -~ u.0I 1--

N NOM '0 -. CiC'ON' <W w ,, w '.., <. i ", , -i . N a. aN) **in-iW0) --0 N) 1- 0F CI0N3O IO- I- ", 0 N. c w c a

~
IN N

LL to

= (Dc~lL CL 0 Ix0E
NWCWFf!~aLU3j-

Cc0 0

cl-0
3 LL

LL 0LL 0L
LLHu

.0 i50
L3L

!X00
:3 u

C0'o
3 LL :3 L

0
z

4
CL

co0 O

If0 n~ Lr). 10le m V.)

N0

IIM

In

U-lN)~I) '40N 00 0' 0T -

Cl~ ~ ~ l(~

'~ JC ~

C 10 N 00 0' 0 ~lN) M V~ M 00 m) n' f N- mlN 0% 0' (NC lC ~~~~~~' lC iC - l( N N(N N cj N- N. N- N- (N (N N- Nl N N (N N N CN (N (N

32 32&

PGHIS ST QjALITY FRACTICA=' ... CQ Uj4-ASaa TO DD,

LL

i*
LL

LL.

I-U

CL

".ij..

o~I-- F

--

M .

CF-

N~~ o
r

'~UC4 (n0 Z

i:4I~

T zU.. MU<L... CO U

No

I0m

m 1-i w w 0r 143Cr a: F-i:,: L CL W <E C t Ld< ---WU)tf (C~-r -I< X <> CL C~~LO C"'-Li. m-.'mN lC I-.L F-L 0 zz z
1. D

C r

z
-i

Gi

Z co ci ~

)'Eu
C-s -= CC

Z~ I z N'-'N U") N I UN M
- M

CD)

*C

Z~.'-ZN--.

c-S

M W N-LiUi~

)~.JLuX L: C ri M-J CL
E

'-CM)Z>C .'

XC!r Zl N x' N~

)I0i< 0Z'' L NI M CL

LL U4T " N

CI

-0 II V) W4~ 0CL in N
CO 14 a 14

.'.i M NC N ZM i <C ~ LL W ru0 00.

-C

X i4 N

0IT q
0 '
< T

C w
1 4-' 14 c aV

CL

IfT 0,. N 00 0' 0 C-ij 4-) Ct' 0 7-!C.' M1 T L! -0 N. 0) 0 0 ,0 N .10 4:) 0t 'I) V) '0N 0) 0s 0 N-. M~ AT) '0 '0 O 0'0 10 ' 0 N r N N N N N N N N MO 'T 4T)I 4T) IT) IT) 0f 0f n0 If) n',0 iT If) '0 10-0 - E NCC l. lN C. 04 C(I-. (N N '- N ' N l N Nl N. N- N (N (N (4 (N (N N C-i C. fC*jNC- N N NC-IC (4 'ICNN ' N N

-:I

THfIS PAGE I-S BEST QUALITY PRACTICABLE =9O OPY FU2ISHE~D TO DQ

33

-IA

LLu

LL Li -L

f-f
Cci

''rj

(, L i-

V-I Ni-.

Lid
HZ -I Li-11
Z.

LZ 3 LL O

0, 3

4q~'-o

Lrm CoCim 2'4 0

mJ D14

34

i
APPENDIX C

SUBROUTINES

THIS PAGE IS BEST QUALITY PRACTICABLE fGm COPY ]VRNISHM~ 1rQ D-Q

35

z w
w
LL.
..

2:z

v.

ZW LL

-L

w
W x->
1-D

04

c1-4

CJL LLI0)q
1%,

PIi 3

CC

0 *3
w

:r-

01
0*0-*)

Uo
ce

CC

1=
1:4
,*

D -e

w.J

U,.* CD WIx

Qw 00T

C) v, .' <E
(n.C
LLo

r0 QoLi. b4(J4WiW

.- LL0TCDC

CD

.zJ
+

0b1.4 01-r,-0

0~w

CL~ a-1 0
L n Z4L ww LZ
CL Mn

(n

*r

In -

W4

-.

Zt

"1

w 11 1 Ix 11

I: 11 1I7

I I 11 e

II

0(3

L 1

CL-1V - 0"4x h

36

MRIS PACE IS UEST QUALITY ?R&OXCABLE

TO DDOC~ WOM Z OFplsk1Jw

LdI

1-4z

Lr

00--4 0
Cr

0 cC4o

xZ0W -xrlwwr

0 Z

r
= -

-0

M
*

-W
fMQI
3 H- 1:

l t W fl'T
M o I-j

0 u
if-z4 -1
1.,ux -LC. f1

zz

1 '-Un- r-4Z ai rC 111:3"z >-Z -

00

CD,2.-L

CCIWC:DL''

1".4.4 '-4

pI :3 CD.4 P

LI p :3.
1,

1-4

IU-'

Q~ I 1-4 LI w W CL:Li

U-.
z

.L
Zb)

1cLL1-4 N 0 0

NII

ru

37 TIS PAGE IS BEST QUALITY PRACTICA~tL9 M9M COPY FUM3ISHED TO DDQ

zz z L
CC
J.

LI

a
L.

u-~ 0. CL

x
a m
.-

CL0x osL

Ce
LI 14-C 11---'
4 1L

LI
LIC z ZCL z'.J Z0n Z C
ix Ce.1

Z zI a-.Jc
Go ix VIP'

Ce
O

Ir

:3

0I

aCX fCZ'- MI--(J

CC CLL.. 1 CC 0C O74
--

~'i' ~eCC >-C0

w17

C.

L1

Ce ~~cl
C

C 0Z C

Er
C. cc 0

In

C0I'-. CfO-C~ * I-C I-- D ccef4T

... -j
-.-

+-

-' N N LL.

Cw * Z -0 Z'-LI

X' ixC
i. N*

ED-IN-

.!x XC-'' : C-

CC

LIC :_kCa

ixLa-a LL

LI C0-C1. cC

I
0

LL C.1--

Ce x
z; c C I-j M Z a-

cc
?i

I-

0 J M * cca-a e

Xi

L1Z N N Z Z CL 0 0 0 L- . Z4 aMI I-7 -iZZZ eMeM ZCL 0 CCC LC M a 1,' -4 -- - 174 U U -L zr Z.JJ. WL CZ o ix 15c: L

-C0

.J Oi. i I z Z il.. LL0~ C. Ce

Ci -

LI4

Ca-a0-0i-CC

a.

C-CC i Z -a' !L LLI C- * 0) 01-4--' loE i0.nJcCe~w >--C W -j MCL 1z, eC z It CCL U CC XC Z Er C~Cj CC -CfL--CWI! a-a lI W ti l t-Z C 1C4 C1Cl-CCL0I 'C'10- L! LI Of~ W 0 LI ix LOamI0 .I C Ce Cl

CN1
C Fele aC~'C

0l-lt

111 0Z

1!

-j

IllW

IM

I.-i I

C 1 -4 - =411

9 IC

12

e zwwM

x0

0 -j =0_ 174

38
F=O

THIS PAGE IS BEST QUALITY PACA= OWPY FURNISHM Tf, MDC'

1741 zI
0co C~~
Ci C4

LLL
0

~~~~
-

~
CL

..

~
U

co-ZI.
1 -i.

Q
CLl il4 ii cc

A:4 CrCL
CC20 0

0
0

'---s

Z~'

+-4

174I'-* 'I C L 0

-'-'

~
z 00

zC F 0j 0
e C 0

Ii4m e.
-

CC

0-0

4 0 Z 00 C-'-i0LLF-

c111-0 I

0 cr 1!i ~ e
'-e -

01 11OZ 0 n

+O WO +D 0F-0m

- 00 li

i! I
:Ml

CO 0 li

0 !n

I-,
Ce
0 0 C-'J C 0: nCF0

5:DJ*

~
n~ e 0 j

Z~ 1-) 0+ Co C4 ) C-4 co ,0t Ze~ )C E. 0 0 0

'-i n-.~oCC i e

0 Z(10
I-

00

--

CCl CL li

il

-:rI

~~ ~ ~~~ ~ ~ 0wo

11.-Zc:

38

THIS PAGE ISBEST QUALIITY PRACTICABLE FMM COPY FMPIS11ED TO DDC

00 0 : 0L

z0 00 1-zD

r49'

C.2

Z
r4 x -.DC 1 x
U) 0 L~-i

+
0 C
4 -0Ii Q

-- l 0C

z , 0

0 X n 0

w *=
aJ
*

C5Z
0

--

wi

00 0-1
**

1=CC
00
-

~
00
Ix+ W 0O

M 0ll I
1

H
4

0' -+
4 I

1
-i
1Z

cl. W H
c

z.i

-z 0
= C

Hl1-C C
n

e- cz
o *1

-+

1e=

1 0

0c 0

LJ

H
0

?--

.)
<-ic

x-

-. 0

l: '-e

0H C+E..J... X~ H
11
0

Zcl

0 00 ce1 M x C
C.

L x -Z CL).
WC,0

'I-4W

0 0 U H 0

Z -1 (Z

cr.
cI-4

i-i
0~

'. J

0 Z 0 0 L-LL II I Cu

3 CUH ~0
17

~0I 1101

0 0 lo- 0 C 0 174 !Ht XU U

w.-i-

1.
lm

. H:0U Vj 0110::: 1-4

00~

0 LI.T.rIT

'ONO W C' IIn 0 000

CulZ0LLH00

'0 N M0W

0 000

0 0,--.0

O'

N NN 0

0' NN

LNN

VITT

-~~

THIS PAGE IS BEST QUALiITY PRACTICABLE FROM Copy kURAISHM TO DDQ

39

I-~

~
i---

-4

..

-4

0 y-

X4

-- i.i0

~J
14 1C' 0 C UJ
Le-

I ( 0 cr
<1

0
IL

C4
=Z
-'0

~
IWI-U..
CLb") 0 .U.

'-LL. 00 W0 C. I Z +wod
4 i 4

z
.

-A
4C

+
eCL z-

+
iU. x~

0~ z
i-4

LL IN

0 wr0
0

Z1

Ud

rICr

U W~4u I-1LL M M-Z-* I .. 440 c0 (-'1l: i. F- ly i- C I 11-0--N'w.4 C) -i 0 0 CL. 0 0IIH'0 9-

+ 0 -'00 +-i-

u~ CIr 1-r 0 *
1i

0 1-Ic- 0 00 Crj ix
-ji'

cr0

Io p 0 C-

c~-ce
-'-

w
T

i-0i

0
-.

L)
.-

I_

uj
3 Li

~
0

'Z cr; .1 'I

~ ~
z--. E .J
W

~
0 0J CL- M

~
'--fl 0
LL u0L.

0- +++ Ii !l

'-

l)

Cc:

E-I-IF- z w00w00z F- 000 -00

17 XI MI

'1-0H- 11X-) 0 0 00c

Li-0 C LC-Z - L

C:LZ~'

L
-e

-- C M LL. F0 0 z0 - Z L c IZ40 0 coo 00 )Z rzZ m CL xu 0

IT:

)~To-C J IN CC 0 0

lo

'

ri 'r
,O r

1-Il4 0 N0 0- 0 -4 C4
N0 l O

IT

j -o 0 N M C- 0

N4Ci! M

V)

"C

c010'

40

Ix

Its,

C'

~THIS pAGEP IS BIDST QUAn T

yaoM Copy FUMISHE TO DDQ

BA

n''

41

*J

ITM

4j
1-

W
*.w

LU

-C(1C

'-

F-

zz

C CL
zU j

il
C.
.-

CZF-F a.
CCZI
..

HC' LiJC 0'i IT WCWO CL r;LU

ix ,-

C(.-I 4 Lu- F--r -4( W w

c-

a. i

CC

4-

A': J.

0LI

CO

Ca L CC.
WC ' -4 Z

Z I-'

-i

W cr

.r

CC

CI-x
CLHWr -- i .JW: F-r-i '-C CH

ZL
I

m L X1 1 I--JUi _. --

W "~ LUC CC

CC

2'

CC.O S

C~C

WH

CLU1--. mZL

Z.0

CL

w- f u

uWx C0

N) St

*CCL

mJ

0>C(A

0 C
CC X CCLU

1 .li
HA

-IH

'-,i

(C-F-Il-4

Ui Q

-)L

00 r)~1! N

N00

~N

-4O ) C--100'4( -1

'ON

00'

tq m0 N m 0'T0 e

42

THIS PAGE I-S BEST QUALITY PRCTcI~r

F*9W COY )'URUISHBD TO) DDO

F'-

Ull

Ld L(J

J W 74

F- I-- 1---W l i

'T M

I-W-U

W WW 14C ****

QM 44 M a-1 *

04MF-

rW WWW

q 7 W WU

0.-1, i 474 -J * FI. F

'

CC

4c u ww

i +c-* - 1 4-4C

C < C Xi-i 1 1
ix ILCL=-I

n-'WWO W'-I- W F-i

FI-I W W Wl*CX-iCT

+W -'

-- W I--IW*W +C~iY i74 *+ I *

I

10i

il NX

ol~~~ 0111110 H

-J i-4T I qH 1

Wj 4T -T F1 11,h

-M W nJ N-WWWW--WWW-F-1-=iWL '-=4--C F~ *-- NWi

11 11 1!1 H- 1E011

cW -i If X) qm 0 c:

Ix 11

1L1
...

- - cCL CLi jj OW>

O... WWWWWWWWi.WWWWWWWWWWWWUWWL-WWL.IlaJwo

F- 1-- 14N0

41z 2~

4N &'--

q-4 14

N -

-rj

1r-0

II I r41 C- m TL

I I

II

IqI

0NO

10-1cjm I

- o1

m0

iC4V

,0

.N0

,0

C MI

.,

IT4r4Tq I qn qrvI

00r

nU

0N - 4j10-

1rr"N

N N

NN0

VN

"'o

43 S STQ1 AI PA [. Tj1OM CIAGY

0~

44.

IT.

4C

014

00

LL, 1--

IP W--

cc- w w LIJ w- lb---

co 174 CC

+ IN "' *

-0
a*

+
Ccn0M 'TM
i -

wF- I-- + +
--

-ce 4C;
Z:w1= C.IQI-

T-qNNmi
wwww

:
mr4 H

* Ni C

C-1 LIW-XX < I WI CLC HI

w
ri

11
LirI
TIlc s1414(1 Il0 N

NW

< J..iI IT I-~ I! LIf N

LLr;

iz iC

NMM WWwIIIII

C4C4
mHHHH1 HH1 IIll

W I- W w w w w

w -0 w U-.L W w Ill x x-.j

~IM r

T Ix -- CL CL

0WJCLL

a-CL-CLa-CL 0-l'-

,40

44THIS

PAGE1 I BST QUITY ?RAOTICAL$ WAWPY W~tdSII TO DDO

..

ItN

CL

Crn-

1- -

1:1

011

zj

-qrL4
-4

lwc
CO
-1

T
O

U L

m0+

fC-z
1
U0 "WF

- x z 0 HI
--

U1

C.

-+ I-zwm IT

-CiC-i

- 1-4 CL
.

LC WW

q
XU

nil 0 0

o.c-1
rI x

1
CL CCCq 1-LH

ifc'Ti 1r11t~ 1a *~ ~c
I-

1*

- F

X-IC-1

1 W

C-4

I MI r4' M-'C CC Ml-LOC-i

N%

Cm
mmmMMCCWJ'-'rlr4nCnOnoOno

l-~N

i CC#-4'-IM

Nr4N) NiNCC'

45 n=oM py FUMNSHED TO DIDQ

THIS PAGE I'S BEST QuALITY PACTICABLE

.

CN

LL;--

-- i
+ -4

U-7

*-4

S,1

r wd -*'

m*

II

01

1<

F--F

in

464

DFDCCBI 1JHS OWY AG 9UNSHM 70I~T

Nr

O I-

-%

J-

wz

C- 0-O

c-

N X

U-

U)

>

- UCL')-Z1 N
ro LC CC a

zL
N W

*I-COI-N~~~~ ~CL)L 01i -~ N '-~'J -Z0

Nll-~ N 0

--..

XN

CZ

-X

o ZC.
E

~
Z -N X0% N 'CJ *-"OZNZ
N

C
Ui -jI co L.

to~.C~W Z C4 a

z:

L X
-L

IX

C
W'-Z

N 4C W (J1-4

- 0N

Z7z1HH-r i---

*z1c.4 W CL

0-~~~~~~I

L-L

i-

i-U)n >( Cl I Lii ) -4Z ncoc

L0

Ci

!X

Lii W

L1

Um

4L)

U NI I

LJuL) U.U vI n -0 N M0

LUU

Uf2U

ULuU p 0a

UUL) MT n 'a NI m

I so N I; x N1 M 4T 0I

~"~-"MT

47 THIS PAGE IS BEST QUALITY P MMo cOFY iiWWSHWi~ TO DDC TZdiled

IxI
-

z
EnCl
-

0 W C40

LL LL.-

CL w Ce-

N z N z
z Z

X~ .J
LL w

CL w N z
CL

1-4U. .

U a
3*1

Ud
.

Ce
ILu

Z -

:3

rz

4x

N IUd
Z z Z :e)

CL r.. E.
*-'

N
CL f-

N
F -4
-j

U imIN

LU Ce

~-

cr- z WO d N
i1-+

rZd N
*

r)
m
CtCL

Z
0 Ci -1

Ix

Zf*U -

z
4x

c C

Z
:3

'-

LL

CL

LL*LLU

OZL

a.J1

wIx

I
i-I Ii

IT

U)
W'

iI N

0I 0I
Im N,

I
10 o -0 -o

(n

x)

N' r-

D'

48

IS PAGS 'I'SBts QUALITY PBACnCABLE

y~WI O()flJIskia TO DIJO

;E

<C

zz
-IL

O o c.C

zCKi
C

-.
L
-ji-

'

Li
JCL

N
I--

'0i..* :0 CO
i4

<r ILL

I-Er;C N Ci Mi N
I-co

'-

co rc
-

:d zI 0

C) N~f _j r4I-

CLi

LL.! C-U
xi

to. H x
w0N

Z
-4 f'H W CL

0 0 LIU C.. LC TI NC

.4U'

N z

i iL

6r LC Li 0N
<r-TLi

rZ i

~
0. -

<[~

L IW rZw ZW Li

-W

.Ji.

W N

DCL

Cj.

N~ lzi

I&N

u Wi-i

Li

CiL i

I;

Hi CL a. W

LL
LLi Nc coN
.

DH, P w
NX CH
w4 z. co w

+4
Wc : W

K
~ij

r LL UU

N C :4

rN

U I-U eU U l ll l ! I

UU:jC

0U

;E

1 11

--

1-411

1-4 (n

l 2 IS PAM -

QUa Ty4

47

zC

Z5

ci
-z

La
XC

z
0 +r coS

--

rcc
-.

z 4. 5.
NI

c
lrc

L..
C.

L.UM
(I

<1 crJN M) I. E z'4

N
IoD

Z LU

-f,

i.
H

n!

0.e.

ZL i C
M 0: xn el

5-4

HL
C. n 0~

N O.. *mxo
Lj

c cE

En

xo N

s4Z *s0)

zC-0"

If

CJ 0-

C...-sz

j
E.WD.

z Li

ix-i-s N

0-

0-b)b

010

:
-cc -

Is~l

- ,CNNC4

o~a w
mm MS M 1-)
'

'N-

'r~ I-V'fr.e4, -T

TI

- T bj

r- v u- c-H.S r,cc Wr ~ I W-h

:l3,
r 1

s-io

a.

10

II-

La

(0

J~I

c~FM

TO DAQ.

CCOP

I
N

2:1z Ncc

4I

51

TIS pAGE TS BEST qUAITy PlCICBL

~~1lVTS14Uf

TO DD,C

jL

U-

a:

0o

z-

ci

ci

cCz a~

nZ
-

N[

?-

U)
E M Cj

ZC 4 "i
U,

CL1<--l

W
:

>-

Z C.3
Z 0 C
*

Iz 0

Oz
m-

Z LL
Im

0
CC >-

M -a

(1)M. ci U) C;
c
L.

1-W
CL

w -.

~
-a N

uW ". W 0. U.

CL~0

ZI Co.. ixD

N" a'eI -0 CL CC 0 x0 Cr
*

L)CU MM U)~~0 '0ii r Lu. zL -aWLi>c U. a.0xc cc 4 WU.1 CCU Lz LLCT acr

CiL

0
X z

-C
IN 01
-

C-41:
T)U C0LU.C z . U 0 ULd C !x

C
52

ZC~C
*.*

>CC

Z *C
-

Wd>ZC
X :3CC 4x
..
*.

0 CL. CCr
d 1-*..

" -CC
-

C
I i
-

ITC
LU

Z
)< ) C
Ill

x r- r.= - C'-o Z c C

Z 0

Ci..~~

z.j
F-- Z

LP~ IF-CLE z 0 C)<cc LL W :C F>Z>- z2 i.

4--~ "

ITU) U) Z -3
I.F-

z Z
ix

~
*

c CL: r- CL 0
a.- D C zrc Y- X

DC 0 a
Li

~
Iz
I-r=

criIx E ct: IzN

0 ":. Z zc Xr aix 0 :3

CF-)

3
:Q

w
E
0~

ti-I

ta.

-ai.~~ *4I If IW i-JM CLa. L i.-l''Z O 1-L 11 (r.w >- M U) 10 C. 0 : D iZ4 0 O'- H ;4IX I m Z -,.i
W = r 3L f

1 <f

~ ~

)C-

4MI i, M -0NM4 . c0 PI T nG NI s0 C. et M0

K 52

~THIS PAGI IS 3tST QUALZTY pRACTICBL

fli)OOOYFRM3RCABLE

~U..
xt r

-'

I-i

CCa

jA0'
rR,

I. i.

C -4 0 i
0
I

-S ~i <W
frHz 000
00

-J

w 0
0*

*0

N 117 I

THI PAGE 1T5 S.E

?A'E S BEST QUALITY FBlACTCBa THI

I
wd

U'

z x
>U4

>

MJac

-~zx

W "H3

:r U;'-X C. m *xwrowr -IiWCDIM -q CC C

CC.

CC x 0 CC LDX
CC riLW

Hcrz

+cxc > a_ . ,Ic"

r 0i
>-

LL

CO

Ifi

I& Wm~h z 000 Q:

Cz

cc:

D
i---

u r-WHHWL

-j 002

-H

O c, CC

cur -aL.

r4S Su
H

1 CC ,k C-L CS
-3C

00u CW

00 55z : (Sn z IS

0 ;-Z =d 0 0 ci N
i-i

U0

*u

ItWXt aH j z NLLU)XzLL~o S T a: cCOLOZLujM . Hw.Za.I.L CC LL WC CCCCC C LSI CO w L E. :. CC4 ZWW014 or L N

x
-d

xx

4
N

~~~
L.J

DL.0 U- DC -o L. W

CLL a.'~a CjX

*

.jCCCZCC

lr,

UX

C>a

aLcr IL 0J0
X-

o D

U C ,H

* d' 1- 0CX

a
C r-Z -ZZ-N

z 1-4a
U2I H) r sx
H

z
WO

Lii
Z Scc

WOO
C00 F-C0zCC*74

0- x M M oa 1-1Q

HL

S1

a40

0 j -4 IT U'

0.j4: a0 to~~j

a.trn

0e

0N 1j). O

'T

0 N M N0 o0' MIT

xv0

~MIS MAOE IS US QUALITY PEACTICABLE

W~~~~~~F-

~j

1Q
ITI

GUNO OY FukwISHM~ TO D~o

NC

V.4q

T4j

55

IT

tN

41.

CL

iY-

-1

LL!

--

ZI -I

-.-

i-

Z-