Skip to main content

Full text of "DTIC ADA086665: Optical Interarray Processing."

See other formats


AO-AO06 665 HONEYWELL CORPORATE MATERIAL SCIENCES CENTER BLOOMIN—ETC F/« *0/6 
OPTICAL INTERARRAY PROCESSING*(U) 

MAR SO 


UNCLASSIFIED 


N000U-B0-C-0216 



























































Office of Naval Research 


Contract No./NjDM 14-8p-C-^)216 J 

_ i O— ' ..... 

e/pTICAL INTERARRAY PROCESSING 


(H 4 

i V 


Mar4HMR80 / 





Honeywell %?. / 

Corporate Material Sciences Center. 

10701 Lyndaie Avenue South ' 
Bloomington, Ml m sesa ta 55420 






TABLE OF CONTENTS 


Section p age 

1 INTRODUCTION 1 

2 OPTICAL DESIGN 2 

2.1 Passive Optical Tau-Shift (POTS) Approach 3 

2.2 Linear Phase Shifter (LIPS) Approach 8 

3 NON-REAL TIME EXPERIMENTS 16 

3.1 Experimental Results 19 

3.2 Resolution and Accuracy 22 

4 REAL TIME SYSTEM DESIGN 24 

4.1 System Overview 24 

4.2 Front End Electronics 26 

4.2.1 TBC Buffer 28 

4.2.2 RF Driver/Modulator 32 

4.2.3 Data Preparation 33 

4.3 Rear End Electronics 33 

4.3.1 Two Dimensional Detector Array 33 

4.3.2 Video Memory 34 

5 SUMMARY 35 

REFERENCES 37 


iii 








LIST OF ILLUSTRATIONS 


Figure Page 

1 Geometrical relation of the two functions 

to be multiplied 4 

2 Ambiguity function generation from one 

dimensional input transducers 5 

3 POTS optical layout for ambiguity function 

generation using Bragg cells 7 

4 LIPS optical layout for ambiguity function 

generation using Bragg cells 10 

5 Construction of space variant linear phase 

shifter from conventional optics 15 

6 POTS experimental layout 17 

7 LIPS experimental layout 18 

8 Ambiguity function of NOSC data produced 

by POTS 20 

9 Scanning traces along the range axis (a) and 

the frequency axis (b) 20 

10 Ambiguity function of V-FM signal produced 

by LIPS 21 

11 Scanning trace along one ridge of the V-FII 

ambiguity function 21 

12 Real time system overview 25 

13 TBC buffer block diagram 29 


LIST OF TABLES 


Table 


1 System Design Specifications 


27 


iv 






SECTION 1 
INTRODUCTION 


This final report covers work performed during the period from 
October 1, 1979 to March 31, 1980 under ONR Contract No. N00014- 
80-CPD216 oh the subject "Optical Interarray Processing". 

The objectives of this contract were to design and demonstrate an 
optical architecture capable of non-real time generation of ambi¬ 
guity functions from one dimensional input transducers, and design 
the front end and rear end electronics necessary to support real 
time operation of such a processor. In the following sections we 
will describe the work we have performed to achieve these objec¬ 
tives. The actual construction of the front end and rear end 
electronics is not part of this contract. However, it will be part 
of the newly awarded ONR Contract No. N00014-80-C-0429. 


1 


SECTION 2 
OPTICAL DESIGN 


The ambiguity function x(v.'r) for two given signals f 2 (t) and 
f 2 (t) is defined by 

t* 00 

x(v,t) - J f x (t) f^(t-T) e -j2irvt dt. (1) 

In the past (1-5), the spatial integration methods used for opti¬ 
cally computing the ambiguity function have all required two 
dimensional input transducers. This requirement effectively placed 
a ceiling on the throughput of the optical processor in that it 
could not exceed the frame rate of the two dimensional input trans¬ 
ducer that was chosen. Currently available two dimensional input 
transducers have maximum frame rates of around 30 frames/sec. 

One dimensional input transducers such as acousto-optic Bragg 
cells have much higher frame rates. For example, a TeOg Bragg 
cell 30 mm in length has a frame time of only 48 nsec. Bragg 
cells also offer the advantages of electronic data composing and 
high diffraction efficiency. 

We have invented a spatial integration approach called Passive 
Optical Tau-Shift (POTS) that uses one dimensional input trans¬ 
ducers. We have also invented another spatial integration approach 
which uses a space variant Linear Phase Shifter (LIPS) and also 
accepts one dimensional input transducers. Both approaches will 
be described in the following paragraphs. 


2 


t 


2.1 PASSIVE OPTICAL TAU-SHIFT (POTS) APPROACH 


If we look at equation (1), it is clear that the product of the 
signals f^(t) and f|(t-x) will reside in a two dimensional plane 
(t,x). However, each signal itself is a function of t only and 
hence one dimensional in nature. The representation of f-(t) in 
a (t, t) plane (Figure 1) can be thought of as the one dimensional 
signal f-(t) stretched vertically along the x axis. Similarily, 
the representation of f 2 (t-x) in a (t,x) plane can be thought of 
as the one dimensional signal f 2 (t) stretched along the line t-x, 
making some angle 6 with the x axis. 

Stretching one dimensional signals can be accomplished easily by 
cylindrical lenses. An implementation of this idea for producing 
the ambiguity function is shown in Figure 2, where Bragg cells are 
assumed as input transducers. 

The one dimensional signal fg (x') in Bragg cell I is diverged 
along the y' axis by a cylindrical lens placed along the x' axis, 
where the x'-y' coordinate system is tilted with respect to the 
x-y coordinate system by some angle 0. The stretched version of 
f 2 (x') is then demagnified to preserve the proper scale with 
respect to the x-y coordinate system, and converged along the 
y axis to form a line on Bragg cell II in a geometrical sense. 

The field immediately after Bragg cell II will be the product 
function f 1 (x)fj (x-t). By Fourier transforming this field with 
respect to x and imaging the intermediate plane in y, the desired 
ambiguity function of equation (1) will be produced. 


The focal lengths of the cylindrical lenses Lg and L 4 are chosen 
to be identical, while the focal lengths f 2 and fg of the spherical 


t 










lenses L 2 and L g are chosen such that their ratio is the demagnifi¬ 
cation factor,cos ( 0 ), or 


7 ^ = cos ( 0 ). 
r 2 


( 2 ) 


A Bragg cell has a finite window height in the vertical (or trans¬ 
verse) dimension, and this imposes a limitation on the processor 
bandwidth and x-shift. Bragg cell I receives a converging wave- 
front from cylindrical lens whose line width is negligible. 

The wavefront converging on Bragg cell II will have a spread in 
y due to the tilt of Bragg cell I with respect to the x-y coor¬ 
dinate system. The signal at plane A in Figure 3 will have a band¬ 
width BW a along the x' axis, and the spread Ay at Bragg cell II 
is due to the y component of BW^, or 

Ay = Xf 5 BW a sin (0) (3) 

where A is the wavelength being used. Also, BW a is related to the 
bandwidth of the signal in the Bragg cell (BW) by 

(4) 

and (4), we obtain 

sin (28). (5) 


BW. = t 2 - BW. 
A f 3 


Combining (2), (3) 


Ay 


,2 

* ia. 

2 f o 


BW 


6 



Figure 3. POTS optical layout for ambiguity function 
generation using Bragg cells. 


7 












It is easily seen from Figure 2 that the maximum r-shift (t „ ) 

max 

is proportional to the time window, T, of the signal and the 
tangent of the tilting angle 0. 


T max 


= T tan 


( 6 ). 


Since tan 0 s 1/2 sin (20) for small 6, we get 


Ay 2 A 



T max 
T * 


( 6 ) 


(7) 


The spread in y at Bragg cell II is, therefore, proportional to 
the maximum r-shift of the processor and the signal bandwidth in 
Bragg cell I, and a tradeoff will exist between the two quantities 
whenever Ay is larger than the effective window height of the one 
dimensional input transducer chosen. 

We will use an acousto-optic modulator as a shutter to define the 
exposure time, which must be shorter than the time window of the 
Bragg cell. The acoustic signals in the two Bragg cells are imaged 
upon each other and travel with the same acoustic velocity so there 
is no misregistration during the exposure time. 


2.2 LINEAR PHASE SHIFTER (LIPS) APPROACH 

In the previous paragraph we showed how the POTS approach can 
create the two dimensional field fg(t-r) from one dimensional 
input transducers. This is done essentially by coordinate rota¬ 
tion followed by demagnification. In the LIPS approach we gener 
ate the r-shift by shear instead of rotation. 


8 




This shear concept is depicted in Figure 4. The telecentric 
spherical lens pair Sg And forms the image of Bragg cell I 
and f 2 (t) onto Bragg cell II and f^(t) through a linear phase 
shifter in the Fourier plane. The presence of the linear phase 
shifter causes a position shift of the image, and this misregistra¬ 
tion accomplishes the x-shift. By spatially varying the slope of 
the linear phase shifter along the vertical direction, the system 
spatially scans continuously in the x axis. Lens performs a 
spatial integration to yield the desired ambiguity function. We 
will now analyze this system in formal Fourier mathematics to 
prove that this concept is sound. 


The Bragg cells accept temporal signals f 2 (t) and aDd convert 

them into a running transmissivity function f 2 (t - ~)• At an 
instant in time we can consider them as the spatial transmissivity 
function B1 and B2 with an appropriate scaling factor. 


B1 = f 2 (x) (8) 

B2 = f 1 < x) (9) 


A linear phase shifter is placed in the Fourier plane to shift 
the phase according to = 2 tt£ti (see Figure 4) where £ and n 
are the coordinates in the Fourier plane. Thus the transmissivity 
function of this linear phase shifter (LPS) is 

LPS = e J2ir?n . (10) 


9 













The process of this cascade optical system can be explained 
effectively using mathematical manipulations to show how this 
system generates the ambiguity function in the final plane. The 
optical fields are notated by U Q , U^, ... corresponding to plane 
0, plane 1, .... The superscript - and + indicate the field im¬ 


mediately before and after the device. 

First, U" can be approximated by a horizontal line, 

IT = S(y). (11) 

The first Bragg cell modulates this line into 

U+ = f 2 (x) U“ = f 2 (x) 6 <y). (12) 

Lens Sg takes the Fourier transform of this field to give 

U~ = ff U* e" j27r(5x+ny) dx dy = F 2 (£). (13) 

This goes through the linear phase shifter to become 

U+ = e J27r5n V~ « F 2 U) e j2lT£n (14) 


11 




Lens takes the Fourier transform to give 
d; - It H+ e _ J2iT(5x + ny) dE dI) 

. I f 2 (C)[/ e J2 ’ ,5n e- 32 ’"' 1 ' dn) e- J2,Jx d£ 

- I F 2 (£) »(E-y> e' J2 " 5x d£. 

Uj - r 2 (y) e" J2 " xy . (15) 

Equation (15) indicates that the height of the pattern is the band¬ 
width of the signal f^x). If the height of the Bragg cell's ef¬ 
fective window is larger than the bandwidth, there is no loss of 
information due to the narrowness of the Bragg cell window. 

U+ = f 1 <x) U~ - f^x) F 2 (y) e~ J2irxy . (16) 

Lens S e takes the Fourier transform of this field and displays 
it in the plane 3 

0 3 - It u 2 e -J 2x <5* - ny) dx dy . 

- / (!<*)[/ F 2 (y) e-J 2 ’ xy e 32 ”ny dy] e - 32x « x dx. 

«3 - / lj(x) t 2 (x -n) e~ i2vix dx. (17) 

Equation (17) clearly shows that the ambiguity function defined 
by equation ( 1 ) is achieved in the spatial frequency space (£,n). 
The conjugation of signal f g (x) can be obtained by putting the 
signal on a carrier and evaluating the first diffraction order 


12 







with the aid of a vertical slit in plane 1. The mathematics 
manipulated in equation ( 8 ) through equation (17) are essentially 
the same to achieve 

U 3 = / f 1 (x) (x -r,) e _J2lTCx dx. (18) 

It is clear that equation (18) is a spatial representation of the 
desired ambiguity function, and we can obtain equation ( 1 ) by con¬ 
verting the spatial variables into the temporal variables with the 
appropriate conversion factors. 

The feasibility of implementing the LIPS approach depends heavily 
on the manufacturability of the linear phase shifter element. It 
is essentially an optical wedge whose wedge angle linearly changes 
with height. The complex transmissivity function of this components 
in rectangular coordinates is given by 

g(x,y) = e JoXy (19) 

where a is a constant. 


Conventional manufacturing processes such as grinding and polish¬ 
ing a glass piece would be difficult if not impossible to apply 
to the fabrication of such an element. We have invented a method 
to fabricate this component out of conventional optics, hence 
high accuracy of the transmitted wavefront is possible. 


By modifying equation (19), we have 


(x+y) 2 

g(x,y) = e z 



, 2 , 2 , 
(x +y ) 


( 20 ) 


13 





2 2 

Define r ■ x + y and introduce a coordinate system (x',y') that 
is rotated from (x,y) by 45° (Figure 5). Then equation (20) can 
be rewritten as 


2 — i— r> 2 

g(x,y) = e^ ax e 2 


( 21 ) 


The first exponent in equation (21) is the complex transmissivity 
function of a cylindrical lens oriented parallel to the x' axis. 

The second exponent is a spherical lens. The cylindrical lens is 
twice as powerful as the spherical lens, and the sign is opposite. 

Therefore, the space variant linear phase shifter can be accurately 
fabricated by cementing a cylindrical lens and a spherical lens of 
opposite power together, and orienting them at 45°. The focal 
length of the cylindrical lens should be half that of the spherical 
lens. 


The LIPS approach has the advantages of fewer cylindrical lenses, 
fewer optical components, and much shorter optical path length. 
Experimental evaluations of both systems will be reported in the 
next section. 










SECTION 3 

NON-REAL TIME EXPERIMENTS 


To test the validity of the POTS and LIPS architectures, both 
systems were constructed, and experiments carried out in a non- 
real time mode using photographic transparencies as input trans¬ 
ducers. The focal lengths of the lenses for both test systems 
were deliberately chosen to be very long so that the necessary 
optical components could be quickly and easily obtained and aber¬ 
rations in the systems could be kept to a minimum. 

For example, the cylindrical lenses in the POTS archiecture had 
800 mm focal lengths, giving the system an f number of f/30, and 
a total system length of 12 meters. The cylindrical lens Cj in 
the LIPS architecture was also 800 mm in focal length, and the 
telecentric imaging lenses were 762 mm in focal length. The linear 
phase shifter element was constructed from a 250 mm single element 
cylindrical lens, and a 505 mm single element negative spherical 
lens. Both lenses were off-the-shelf components. Even with such 
long focal lengths, the total system length for LIPS was 5.5 
meters, considerably shorter than the 12 meters for POTS, although 
the two systems had nearly the same f number. The focal lengths 
of all the lenses for both systems are shown in Figures 6 and 7. 

The photographic transparencies used in evaluating the two sys¬ 
tems were created using a computer-driven precision CRT plotter 
and high resolution microfilm. Since the synthetic data on the 
NOSC tape is complex, it must be placed on a carrier frequency to 
preserve the phase information of the original signal. The NOSC 
data is, therefore, plotted according to the formula 


16 




















(22) 


R(t)cos(2Trv t) - I(t)sin(2irv t) + B. 
c c 

where R(t) and I(t) are the real and imaginary values of the 
signal at time t, v is the chosen carrier frequency, and B is 
a bias value chosen such that only positive values will be gen¬ 
erated for plotting. 

The physical size of a plotted transparency was chosen to simulate 
a Bragg cell, so it measures 24 mm long by 3 mm high. When used 
in the experimental optical processors the transparencies were 
placed in liquid gates filled with index-matching fluid. 


3.1 EXPERIMENTAL RESULTS 

An example of an ambiguity function generated by the POTS archi¬ 
tecture is shown in Figure 8. In this case, the input signal-to- 
noise ratio (SNR) was 0 dB. Scanning traces along the range and 
doppler axes are shown in Figure 9 and they show a good output 
SNR with the peak clearly standing out from the surrounding noise. 

An example of an ambiguity function generated by the LIPS archi¬ 
tecture is shown in Figure 10. In this case, the input signal was 

a V-FM signal composed of two symmetric wings of linear chirp of 

• , o 

the form e la joined at the middle. This is a very useful test 

6 7 

signal since its ambiguity function is well-known. ’ The ratio 
of the central peak height to the shoulder height immediately 
adjacent to the peak should be 4 to 1, and this is confirmed by 
the scanning trace in Figure 11. 


19 



Figure 8. Ambiguity function of NOSC 
data produced by POTS. 



(b) 


Figure 9. Scanning traces along the range axis (a) and 
the frequency axis (b). 


20 








Figure 10. Ambiguity function of V-FM 
signal produced by LIPS. 




Figure 11. Scanning trace along one ridge 
of the V-FM ambiguity function. 


21 


i ' 

r 






3.2 RESOLUTION AND ACCURACY 


The resolution of an ambiguity function is related to the signal 
bandwidth (BW) and the length of the time window (T) by the well- 
known relations 



The scanning traces in Figure 9 were taken from an ambiguity 
function whose input signals had signal bandwidths of 0.1 Hz and 
time windows of 600 seconds. Both traces were calibrated for scale, 
and their resolutions at the half power points of the central peaks 
were determined. For the frequency axis, the resolution was found 
to be 1.59 mHz, compared to 1.68 mHz predicted by theory. For the 
range axis, the resolution was found to be 9.7 seconds, compared 
to a theoretical prediction of 10 seconds. Both resolutions are 
therefore in excellent agreement with the relations in equation 
(23), a result which was typical of all the optically produced 
ambiguity functions. 

To check the accuracy of an optically produced ambiguity function, 
either the ambiguity integral for two given signals must be cal¬ 
culated digitally, or a signal whose ambiguity function structure 
is well-known can be chosen and processed optically, and the re¬ 
sults compared. The V-FM signal falls into the latter category. 

The ambiguity function of the V-FM signal produced by the LIPS 
architecture was previously noted to be in good agreement with 
the calculated ambiguity function in its peak height to shoulder 
height ratio. 


22 



Also present in the optically processed ambiguity function is 
excellent symmetry about the range and doppler axes, and a fine 
structure of ripples running between the diagonal arms of the 
ambiguity function, as clearly present in both the photograph 
and the scanning trace. The symmetrical presence of this struc¬ 
tural detail, which is predicted by calculation, indicates that 
a very accurate ambiguity function has been produced. V-FM ambi¬ 
guity functions having such excellent detail were produced by 
both the POTS and LIPS architectures. 

The two optical processing architectures investigated have the 
common advantage of using one dimensional Bragg cells as input 
transducers, and both can produce accurate ambiguity functions. 
The LIPS architecture also has the additional advantages of being 
inherently much shorter in optical path length, and it requires 
fewer optical components. Finally, the LIPS architecture requires 
only one cylindrical lens of quality, while the POTS architecture 
requires at least three. It is clear that the LIPS architecture 
has more potential to become a relatively small, compact, optical 
processor. 


23 








SECTION 4 

REAL TIME SYSTEM DESIGN 


In the previous section we showed that our optical system can 
generate the two dimensional ambiguity function using one dimen¬ 
sional nonreal time input transducers. This section will discuss 
the system we have designed to realize real time operation of the 
optical processor. The one dimensional input transducers that 
were simulated by transparencies will be replaced by acousto¬ 
optic Bragg cells which can accept real time electronic signals 
and convert them from the time domain to the space domain. The 
output of the optical processor will be detected by a two dimen¬ 
sional photodiode array and converted into an electrical signal 
for output from the system. 

4.1 SYSTEM OVERVIEW 

The real time system is designed around the optical ambiguity 
function generator utilizing two Bragg cells (Figure 12). The 
front end electronics, consisting of a time base compression 
(TBC) buffer, two digital-to-analog converters (DACs), and two 
RF driver/modulators, receives the input signals from the HP 
9825A calculator in digital form and converts them to analog 
signals which cause index of refraction modulations in the Bragg 
cells which are proportional to the input signals. The rear end 
electronics, consisting of a two dimensional detector array 
camera and a video memory, captures the ambiguity function, dis¬ 
plays it on a CRT, and digitizes and stores it for examination. 




2-D DETECTOR 



Figure 12. Real time system overview. 



















The optical source is a one watt Argon laser, and an acousto¬ 
optic modulator acts as a shutter. Shutter operation, camera 
activation, and data transfer into the video memory are all con¬ 
trolled by timing control signals from the TBC buffer. 

The system design specifications are listed in Table 1. The design 
processing speed of the real time system is 500 frames/sec and 
the system time-bandwidth product is 240. This real time system 
is interfaced to an HP 9825A calculator for system performance 
evaluation purposes. Details of the subsystems are discussed in 
the following paragraphs. 

4.2 FRONT END ELECTRONICS 

The most important design criterion in the front end electronics 
is the system time-bandwidth product (TBW). An acoustic shear wave 
in a Te0 2 Bragg cell will propagate with a speed of 617 meters/ 
sec, while the practical physical window size is 30 mm. Therefore, 
it takes only 48 ysec for the acoustic wave to propagate through 
the window. A sufficient system TBW therefore requires a fast data 
rate and high speed DACs. The fastest 8 bit DAC commercially avail¬ 
able operates at a 50 MHz data rate, and a digital TBC buffer of 
comparable speed must be constructed to operate with it. The fol¬ 
lowing paragraphs describe the TBC buffer design based on a concept 
of multiplexing emitter-coupled-logic (ECL) memories. Also des¬ 
cribed will be the RF driver-modulators and data preparation. 


26 


Table 1. System Design Specifications 


Bragg Cells 

Physical Length 
Time Window 

Time Base Compression Buffer 
Size 

Data Rate 

Maximum Signal Time Window 

Data Preprocessing 

Bandwidth 
Subcarrier 
Sampling Rate 

RF Driver 

Driver No. 1 
Driver No. 2 

System Processing Rate 

System Time-Bandwidth Product 


30 mm 
48 ysec 

1920 bytes x 2 channels 
50 Mbytes/sec 
38.4 ysec 

0.25 Hz 
0.4 Hz 

5 points/carrier cycle 

Center frequency 45 11Hz 
Center frequency 75 MHz 

500 Frames/sec 

240 


27 




4.2.1 TBC Buffer 


This buffer is an electronic system designed to provide the re¬ 
quired time base compression between very slow data storage devices 
such as magnetic tape, disk, or digital memory, and very fast 
optical devices such as Bragg cells. The component design used for 
this task is ECL circuitry which can provide the speed capability 
needed in this buffer. The TBC buffer is designed to accept a 16 
bit parallel direct-memory-access (DMA) word with handshake and 
control lines from a HP 9825A calculator through a HP 98032A inter¬ 
face connection. 

The TBC buffer has the following nine basic circuit blocks (see 
Figure 13): 

1. I/O Translators 

2. Clock 

3. Controller 

4. Address Generator 

5. Memory 

6. Parity and Display 

7. Demultiplex 

8. Line Driver 

9. Digital-to-Analog Converter (DAC). 

Circuit blocks 1 through 8 are contained on one Augat 3 layer, 

5 level, wire-wrap circuit board which provides low noise, high 
frequency, high density, and accurate system performance. 

1. I/O Translators — Translator circuits are used at all I/O 
ports to convert from TTL logic to ECL logic, and vice versa. 

These ports include command byte, data word, peripheral signals, 
handshake, and control line activity. 


28 




HP98032A 



I INPUT TRANSLATORS 
0 OUTPUT TRANSLATORS 
CB CONTROL BYTE 
MAB MEMORY A BYTE 
MBB MEMORY B BYTE 
OR OR'S A1 AND A2 
L LINE DRIVERS 
T TIMING 


Figure 13. TBC buffer block diagram. 


19 























2. Clock — A 25 MHz frequency generator is used to clock and 
synchronize the TBC buffer. This generator is an RC-coupled 
voltage controlled multivibrator. The clock signal is divided by 
two to provide the two phases required for multiplexing. 

3. Controller — The internal operation of the TBC buffer and the 
response of the peripherals connected to it are determined by the 
controller circuits which operate under program control and per¬ 
form the following tasks: 

a. Function — The command instruction from the HP calculator 
that selects the peripheral configuration, e.g., write memory, 
read memory, camera start, etc. 

b. Handshake — Each time the HP calculator communicates with 
the TBC buffer, a signal is sent and returned that the instruction 
has been received and completed. Four handshake conditions are 
used: 

• Command byte 

• Write with each data word 

• Read 

• Read and EOF (end of frame-camera). 

c. Timing — Accurate simultaneous and sequential signals are 
generated in this section for internal buffer usage and to properly 
time peripheral activity. These timing signals affect addressing, 
chip select, write enable, DAC strobing, the A/0 modulators, and 
the camera and video store. Reset restores all buffer registers 

to zero. 


30 





d. Multiplex — The access time of the memory chips is long 
enough to be a problem at 50 MHz. The memory is therefore multi¬ 
plexed, with this circuit controlling chip access to easily 
satisfy the speed requirement. 

e. Peripherals — These are the two A/O modulators, the 100 x 
100 array camera, and the video memory. These devices are initial¬ 
ized, started, stopped, and reset by program control. 

4. Address Generator — The address generator is a 12 bit syn¬ 
chronous binary counter of which 10 bits are used to provide 
addresses A0-A9. These address lines take two paths. The first 
path is through line buffers directly to memory A^ and memory 
The second path is to a 10 bit register which at a later time is 
accessed (during multiplexing), then through line buffers to 
memory A g and Bg. 

5. Memory — Memory A consists of sixteen 1024 word, 8 bit memory 
chips in two sections (A^ A 2 ) of 8 chips each, as does memory B. 
Memory A^ and B 1 are simultaneously addressed, chip selected, 
and write enabled; Memory A 2 and B 2 are simultaneously delay 
addressed, chip selected, and write enabled. The delay is the 
multiplex operation. Each section is alternately written (and 
alternately read) to the program-selected number of words. Writing 
the memories is done at HP 9825 DMA speed (200K words/sec). Read¬ 
ing has the same format as writing but the output data rate is 

50 mega-wcrd8/8ec. 

6. Parity and Display — As each input data word is processed, a 
parity bit is generated and placed in a small parity memory. When 
data is output, the parity checker compares the newly generated 


31 


I 



parity value with the value in parity memory. If these are 
different, an error bit is counted and translated to an LED 
display. 

7. Demultiplexer — Data from the output ports of memory A.^ and 
A 2 are logically OR’d to provide a combined output data rate of 
twice the 25 MHz clock frequency. The same thing is done for 
memory and Bg. The results are sent to data line drivers. 

8. Line Drivers — The line drivers provide true and complimentary 
outputs which can be transmitted on twisted pair leads. Twisted 
pair transmission lines of several inches length can be used to 
supply these fast signals to the CACs which are mounted adjacent 
to the TBC buffer board. 

9. DAC —The ECL digital-to-analog converters are Tektronics DAC 
850s. These are 8 bit converters capable of a 50 MHz update rate. 
The analog output is gain and offset adjustable. 


4.2.2 RF Driver/Modulator (RFDM) 

RFDM I drives Bragg cell I at a center frequency of 45 MHz. 

RFDM II drives Bragg cell II at a center frequency of 75 MHz. 
Both drivers were designed for suppressed carrier amplitude 
modulation. If we place a signal with a bandwidth of 0.25 Hz on 
a 0.4 Hz subcarrier, and choose a sampling rate of 5 poinds/ 
carrier cycle, the 50 MHz data rate from the TBC buffer will 
upshift the carrier frequency to 10 MHz and the bandwidth to 
6.25 MHz. 


32 






4.2.3 Data Preparation 


Initially, the NOSC synthetic data resided on a magnetic tape. 

This tape was read by a Honeywell Level6 minicomputer at CTC and 
its contents transferred to the Level6 disk storage, where it can 
easily be accessed for data preparation. A data preparation pro¬ 
gram has been written to place this data on a carrier frequency, 
add a bias to generate only positive 8 bit values from 0 to 255, 
combine an 8 bit data point from each signal into one 16 bit word, 
and store the prepared data in a disk file for transfer to the HP 
982SA calculator. 

The data transfer from the Levels to the HP 9825 will occur through 
a special HP interface by way of a 300 baud telephone line. A 
transfer program has been conceived allowing the HP 9825 to decode 
the data and store it on its floppy disk system. Once the data is 
on the floppy disk system, it will be ready for multiple write 
operations to the TBC buffer. 


4.3 REAR END ELECTRONICS 

4.3.1 Two Dimensional Detector Array Camera 

The detector array camera is a Reticon MC520 with a RS520 con¬ 
troller. The camera consists of 10,000 segmented photodiodes in 
a 100 x 100 matrix; it has a dynamic range of 200 to 1. The 
maximum pixel scan rate of 5 MHz translates into a frame rate 
of 500 frames/sec. The camera gives a positional signal for x and 
y and a video signal for z. It can, therefore, be used with a 
CRT with a z input terminal. The camera control signals are 
provided from the timing control circuit in the TBC buffer. 


33 





4.3.2 Video Memory 


A video memory made by Applied Micro Technology will be used. It 
is designed to interface exclusively with the Reticon detector 
array camera, and can capture a single frame when the frame rate 
is 500 frames/sec. The value of each of the 10,000 photodiodes 
in the detector array will be digitized to 8 bits accuracy and 
stored. The stored frame can be continuously displayed on a 
refresh CRT. The digitized data can also be sent to the HP 9825A 
calculator for further analysis. 


34 



SECTION 5 
SUMMARY 


This project had two major objectives. The first objective was to 
demonstrate nonreal time generation of ambiguity functions from 
one dimensional input transducers. The second objective was to 
design front end and rear end electronics capable of supporting 
real time operation of the optical processor. Both of these objec¬ 
tives were successfully achieved, and in addition, a new optical 
architecture was invented (LIPS), and its feasibility and superior¬ 
ity was experimentally demonstrate d.^ 

**The optical system demonstrating the proposed Passive Optical Tau- 
Shift concept (POTS) was designed with a large f number (>?0) for 
aberration-free performance, and the system length became 12 meters. 
One dimensional Bragg cell-sized photographic transparencies were 
used to generate ambiguity functions. The results of processing 
both analytic V-FM signals and synthetic data from NOSC showed 
excellent agreement with theoretical predictions in terms of the 
correlation peak height and resolution. 




Another optical architecture based on an entirely new concept of 
a space variant Linear Phase Shifter (LIPS) was independently 
developed. It generates a, x*-shift by a shear operation instead 
of the rotation used in pojrs. Nonreal time experiments demonstrated 
good agreement with theory/ i*M;erms of correlation peak height and 
resolution, and excellent/ disrbrtion-free output. The advantages 
of the LIPS approach are/compactness, fewer optical components, 

and the elimination of the rotated\coordinate systems which are 

/ 

in the POTS approach. / 


35 






The real time system design can achieve a 50 MHz digital data rate 
in the front end electronics and a 500 frames/sec frame rate in 
the rear end electronics. The processing system will have a system 
TBW of 240, and any of the frames being generated at a 500 frames/ 
sec rate can be captured, digitized, and stored for later evalua¬ 
tion either by a minicomputer or a large computer system. 


36 





REFERENCES 


1. L.J. Cutrona, E.N. Leith, C.J. Palermo, and L.J. Porcello, 
"Optical Data Processing and Filtering Systems", IRE Trans. 
IT-6, 386 (1960). 

2. K.A.K. Said and D.C. Cooper, "Crosspath Real Time Optical 
Correlator and Ambiguity Function Processor", Proc. IEE 120, 

423 (1973). 

3. R.J. Marks II, J.F. Walkup, and T.F. Krile, "Ambiguity Function 
Display: an Improved Coherent Processor", Appl. Opt. 16, 746 
(1977). 

4. D. Casasent and B.V.K. Vijaya Kumar, "Optical Image Plane 
Correlator for Ambiguity Surface Computation", Appl. Opt. 18, 
1673 (1979). 

5. P.N. Tamura, J.J. Rebholz, and T.C. Lee, "Ambiguity Map 
Generation with Coherent Optics", J. Opt. Soc. Am. 69, 1451 
(1979). 

6. A. Rihaczek, Principles of High Resolution Radar (McGraw-Hill, 
Hill, New York, 1969). 

7. C.E. Cook, M. Bernfeld, Radar Signals (Academic Press, New 
York, 1967). 


37 







END 


DATE 

FILMED 


DTIC 


tit