UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

A multi-level digital correlation spectrometer Whyte, Don Andrew 1972

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
831-UBC_1972_A7 W39.pdf [ 8.4MB ]
Metadata
JSON: 831-1.0101497.json
JSON-LD: 831-1.0101497-ld.json
RDF/XML (Pretty): 831-1.0101497-rdf.xml
RDF/JSON: 831-1.0101497-rdf.json
Turtle: 831-1.0101497-turtle.txt
N-Triples: 831-1.0101497-rdf-ntriples.txt
Original Record: 831-1.0101497-source.json
Full Text
831-1.0101497-fulltext.txt
Citation
831-1.0101497.ris

Full Text

A MULTI -LEVEL DIGITAL CORRELATION SPECTROMETER by DON ANDREW WHYTE S c . , Queen ' s U n i v e r s i t y a t K i n g s t o n , O n t a r i o , 1969 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF APPLIED SCIENCE We a c c e p t t h i s t h e s i s as c o n f o r m i n g to the r e q u i r e d s t a n d a r d R e s e a r c h S u p e r v i s o r Members o f Committee Head o f Department Members o f the Depar tment o f E l e c t r i c a l E n g i n e e r i n g THE UNIVERSITY OF BRITISH COLUMBIA J a n u a r y , 1972 I n p r e s e n t i n g t h i s t h e s i s i n p a r t i a l f u l f i l m e n t o f the r e q u i r e m e n t s f o r an advanced degree a t the U n i v e r s i t y o f B r i t i s h C o l u m b i a , I agree t h a t the L i b r a r y s h a l l make i t f r e e l y a v a i l a b l e f o r r e f e r e n c e and s t u d y . I f u r t h e r agree t h a t p e r m i s s i o n f o r e x t e n s i v e c o p y i n g o f t h i s t h e s i s f o r s c h o l a r l y p urposes may be g r a n t e d by the Head o f my Department o r by h i s r e p r e s e n t a t i v e s . I t i s u n d e r s t o o d t h a t c o p y i n g o r p u b l i c a t i o n o f t h i s t h e s i s f o r f i n a n c i a l g a i n s h a l l n o t be a l l o w e d w i t h o u t my w r i t t e n p e r m i s s i o n . Department o f ^ 1 i^c T72-(C/=)L- AS<£T AJE: Z F / e V ^ g The U n i v e r s i t y o f B r i t i s h Columbia Vancouver 8 , Canada ABSTRACT The design of a 256-channel cross-correlation spectrometer i s described, which i s to be used i n the study of line-spectra i n radio astronomy at the Dominion Radio Astrophysical Observatory, near Penticton, B r i t i s h Columbia. The correlator i s a dual instrument, providing 128 channels for the determination of each of the Co- and Quadrature-spectra of the signals from . two paraboloidal antennas, over a bandwidth ranging from 1/4 MHz to 8 MHz. D i g i t a l c i r c u i t r y i s used, for the long-range s t a b i l i t y and free-dom from d r i f t s that i t provides. The instrument employs a unique design of a simple d i g i t a l m u l t i p l i c a t i o n and accumulation c i r c u i t , which minimizes the cost of implementing m u l t i - l e v e l d i g i t i z i n g . One signal i s quantized i n three-levels, the other i n f i v e l e v e l s . This produces a degradation i n the signal-to-noise of the cross-correlation c o e f f i c i e n t s , over that ob-tained i n an analog correlator, by a factor of 1.16. Previous d i g i t a l correlators of wide bandwidth have employed one-bit quantization, with a degradation factor of 1.57. The cross-correlation products i n each channel are accumulated i n fourteen-stage counters. Extended accumulation i s provided by a 4,096-bit c i r c u l a t i n g glass memory. The coe f f i c i e n t s are p e r i o d i c a l l y transferred to a PDP-9 computer, where a fast Fourier transform i s performed to y i e l d the complex cross-spectrum. The design of the analog-to-digital, arithmetic, timing, pulse d i s t r i b u t i o n , and control c i r c u i t r y i s described i n th i s thesis. i i TABLE OF CONTENTS Page ABSTRACT i i TABLE OF CONTENTS . i i i LIST OF ILLUSTRATIONS v i LIST OF TABLES . v i i i ACKNOWLEDGEMENT i x I. SUPERSYNTHESIS SPECTROSCOPY 1 1.1 INTRODUCTION 1 1.2 ASTRONOMICAL BACKGROUND 2 1.3 APERTURE SYNTHESIS AND SUPERSYNTHESIS 3 1.4 SUPERSYNTHESIS RECEIVER 9 1.5 THE D.R.A.0. SUPERSYNTHESIS TELESCOPE 1 1 II . INSTRUMENTATION FOR SPECTRAL ANALYSIS.... 1 3 2.1 MULTI-FILTER AND CROSS-CORRELATION SPECTROMETRY 1 3 2.2 THEORY OF THE CROSS-CORRELATION SPECTROMETER 1 8 a. Fourier Relationship 18 b. The Sample Cross-correlation and Sample Cross-spectrum 21 c . Preservation of Phase 21 2.3 THE DIGITAL CROSS-CORRELATOR . 2 2 a. Introduction 2 2 b. Sampling 2 2 c . Number of Spectral Points Calculated 2^ d. Application to Present Instrument 2^ 2.4 MULTI-LEVEL QUANTIZATION.... 26 a. Quantization 2^ b. Signal-to-noise Degradation.. 2 9 c . Three-level by F i v e - l e v e l D i g i t i z a t i o n 31 III. THE CROSS-CORRELATION SPECTROMETER .: 35 3.1 INTRODUCTION TO THE DESIGN. 35 i i i Page 3.2 REQUIREMENTS...... 39 a. Modes of Operation 39 b. D.C. Correction 41 c. Normalization and Gain Control 42 d. Continuum Correlators ^4 e. Overflow S t a t i s t i c s ^4 f. Correction for Quantization 46 g. Encoding the Quantized Signals , ^7 h. Special Channels ^ 8 i . Glass Memory 48 IV. DESIGN OF THE CROSS-CORRELATOR... 5 1 4.1 ANALOG-TO-DIGITAL CONVERSION 5 1 a. Introduction 52 b. Sampling and Quantization.. ^6 c. Performance C r i t e r i a 61 d. Encoding the Quantizer Outputs 62 e. D i s t r i b u t i o n of Pulses 64 4.2 ARITHMETIC UNIT 64 a. Function of Unit 64 b. Selection of Scheme 64 c. Design of the Revision Arithmetic: General Features... 66 d. Details of the Design 7^ 4.3 TIMING-UNIT 7 5 a. Requirements 7^ b. 32 MHz Clock 7 9 c. Master Clock Card ...... 8 0 d. Bandwidth Selector 8 0 e. Synchronization Problem 83 f. Timing Pulse Generators 86 g. Temperature Compensation 8 9 h. Channel Address Code Generator 90 i . Computer Transfer Timer..... 91 i v Page 4.4 MULTIPLEXING AND SERVICING OF CHANNELS 92 a. Channel Address Code Generator 92 b. Channel-Readout Multiplexing 95 c. Prompt Pulse I n h i b i t i o n 98 d. Reset D i s t r i b u t i o n 102 e. Timing of the Three Channel-Services 104 4.5 COMPUTER TRANSFER CONTROL 105 a. Readout Control - Computer Transfer Routine 105 b. Design 1°5 4.6 BUFFER REGISTER - INTERFACING TO THE PDP-9 COMPUTER 109 4.7 OPERATION CONTROL: RESET CONTROL 112 a. Operation Modes 112 b. Reset Control . 1 1 4 c. General Reset H4 d. Automatic Reset 116 e. I n i t i a l Accumulation H 8 f. Memory Reset. H9 4.8 OPERATION CONTROL: PROMPT PULSE SUPPRESSION ... 120 a. PAUSE and FREE-RUN Modes !21 b. D.C. Subtraction Control 125 c. Prompt Pulse Suppression.... 125 V. CONCLUSION 130 5.1 TESTS ... 130 5.2 RECOMMENDATIONS ................ 131 a. The Three-level by F i v e - l e v e l Correlator 131 b. Improvement i n S/N. 132 5.3 CONCLUSION... .. 1 3 3 REFERENCES. 135 APPENDIX . ... 136 v LIST OF ILLUSTRATIONS Figure Page 1.1 The 'u-v Plane" 5 1.2 Supersynthesis Radiotelescope for H^ . Spectroscopy 8 1.3 Supersynthesis Receiver 8 2.1 M u l t i - f i l t e r Spectrometer 14 2.2 Cross-correlation Spectrometer 14 2.3 Effect of Sampling on the Spectral Estimate 17 2.4 Suitable Grading Function, and corresponding Smoothing -Function 17 2.5 Basic Quantization Scheme 26 2.6 Quantization Viewed As an Area Sampling of the Input Prob a b i l i t y Density D i s t r i b u t i o n 22 2.7 Bivariate Gaussian Pr o b a b i l i t y Density D i s t r i b u t i o n 28 2.8 Multiplier-Accumulator-Module 32 2.9 Five-valued and Three-valued Quantization 33 3.1 Sim p l i f i e d Block Diagram of the Cross-correlation Spectrometer 36 3.2 Block Diagram of the 256-Channel Dual Cross-correlator 37 3.3 Multiplier-Accumulator Module, Schematic Diagram 38 3.4 Cross-correlation Coefficient, P x t t ( f ) , of the Quan-ti z e d Signals vs. p (T) of the Unquantized Signals Schematic Diagram of the Glass Memory 47 3.5 49 4.1. 1 Schematic Diagram of the Prompt of Semiprompt Quan-t i z e r 53 4.1. 2 Schematic Diagram of the Delay Quantizer 54 4.1. 3 Response of the Comparators 55 4.1. 4 Encoding of S e r i a l Outputs of the Quantizers 55 4.1. 5 Prompt and Semiprompt Pulse Buffers i n CPU Rack 63 4.1. 6 Channel Rack Buffer cards 63 4.2. 1 Two Possible Schemes for the Revision of the Memory-words From Data i n the Channel-Accumulators 65 4.2. 2 The S e r i a l Revision Scheme Adopted for the Arithmetic Unit 73 4.2. 3 Si m p l i f i e d Schematic Diagram of the Arithmetic Unit 74 4.3. 1 Timing Requirements for the Correlator 76 Page 4.3.2 Sequence of Events i n the Revision-Cycle 77 4.3.3 O s c i l l a t o r C i r c u i t for 32 MHz Master Clock Signal 78 4.3.4a Schematic Diagram of the Master Clock Card 81 4.3.4b Schematic Diagram of the Bandwidth Selector 82 4.3.5 Outline of Units Whose Timing Must be Synchronized 84 4.3.6 Sychronized Timing and Requirements of the Master Clock 85 4.3.7a Timing Signals Generated i n the Timing Unit 87 4.3.7b Timing Signals Generated i n the Timing Unit (continued) 88 4.3.'8 Monostable C i r c u i t s for Generating Pulses of Variable Phase and Pulse Width 89 4.4.1 Arrangement of Channels i n Memory Locations 95 4.4.2 Schematic Diagram of the Channel Address Code Generator 96 4.4.3 Block Diagram of the Channel Read-Multiplexing Network 97 4.4.4 Block Diagram of the Prompt Pulse D i s t r i b u t i o n Network 99 4.4.5 Schematic Diagram of a Prompt Pulse Inhibitor Card 100 4.4.6 Block Diagram of the Reset Distr i b u t o r Network 103 4.5.1 Block Diagram of the Readout Control 107 4.6.1 Schematic Diagram of the Buffer Register 111 4.7.1 Block Diagram of the Reset Control 115 4.8.1 Block Diagram of the Prompt Pulse Suppression Control 123 B.2.1 Schematic Diagram of the Arithmetic Unit 138 B.2.2 Timing Relationships for the Arithmetic Unit 139 B.3.1 Operation of the Bandwidth Selector 140 B.3.2 Revision-Cycle Counter, Timing Unit Card 1 141 B.3.3 Reference Waveform 16 Mhz 141 B.3.4 Pulse Generators, Timing Unit Card 2 142 B.3.5 Pulse Generators, Timing Unit Card 3 143 B.3.6 Schenatic Diagram of the Computer Transfer Timer 144 B.3.7 Shi f t Pulse Buffers 145 B.3.8 Read-Multiplexer Strobe Pulse Buffers 145 B.4.1 256-State Pattern of the Eight Channel Address Code Digits 146 B.4.2 Channel Address Code Buffers 146 B.5.1 Schematic Diagram of the Readout Control 147 B.5.2 Schematic Diagram of the Divide-by-41 Counter .148 B.5.3a I n i t i a t i n g Timing of the Readout Control 149 v i i Page B.5.3b Terminating Timing of the Readout Control 150 B.6.1 Timing Relationships for Buffer Register 152 B.6.2 Block Diagram of the Interface Between the Correlator and the PDP-9 Computer 151 B.7.1 Schematic Diagram of the Reset Control 153 B.7.2 Timing Relationships for the Reset Contrl 152 B.8.1 Schematic Diagram of the Prompt Pulse Suppression Control 154 LIST OF TABLES Table Page 2.1 Single-Channel Degradation Factors 31 2.2 System Degradation Factors 31 2.3 M u l t i p l i c a t i o n Table for Three-level by Fi v e - l e v e l Quantization 33 3.1 Modes of Operation 40 3.2 Encoding of Quantizer Outputs 40 3.3 R.M.S. Values of Quantized Signals 40 3.4 Overflow S t a t i s t i c s 45 4.2.1 Use of Digits from the Channels i n the Revision Process 67 A.3.1 Assignment of Special Channels 137 v i i i ACKNOWLEDGEMENT This work was supported by the National Research Council of Canada (N.R.C). Funds for a Research Assistantship for the author during the year 1970-71, and for the purchase of some components, were made available through Research Grant 67-3295. Most of the components for the correlator were purchased through the Dominion Radio Astrophysical Observa-tory., operated by N.R.C. The author i s grateful also for an N.R.C. Postgraduate Bursary for the year 1969-70. The author i s indebted to his supervisor, Professor F. K. Bowers, f o r h i s very h e l p f u l suggestions and contribution of ideas. The i n i t i a l developmental work upon which this project i s based was done by him and by Mr. D. S. Sloan at the University of B r i t i s h Columbia. The support given to the author by the entire s t a f f at the D.R.A.O. i s greatly appreciated. In p a r t i c u l a r the author i s grateful to Dr. T. L. Landecker, technical supervisor for the supersynthesis receiver system, for his collaboration i n the design of the correlator; and to Mr. M. D. Robinson, who assisted i n the wiring and construction of the instrument. Thanks are due to Dr. J. S. MacDonald for reading the manuscript and to Miss Norma Duggan for typing the f i n a l d r aft. The author i s also sincerely grateful to his family and friends, f o r t h e i r moral support. i x \ 1 I. SUPERSYNTHESIS SPECTROSCOPY 1.1 • INTRODUCTION One of the most productive endeavours i n radio astronomical research i n recent years has been the study of the s p e c t r a l line-emission of atomic hyd^gen (H^) at the wave-length 21.1 cm (1.420 Ghz). Obser-vations i n our own galaxy, of 21 cm l i n e i n t e n s i t y , frequency s h i f t and broadening s have helped to reveal the s t a t i c and dynamic properties of i t s spiral-arm structure [17], [19]. External g a l a x t i c objects have also been observed, both i n the line-emission of atomic hydrogen and as features i n the absorption spectra of continuum radio sources. Spectroscopic mea-surement of the 21 cm l i n e i n the Andromeda Nebula, for example, have yie l d e d information about gross structure, hydrogen d i s t r i b u t i o n and t o t a l content, and large scale dynamics ( r o t a t i o n a l motion) [1]. In addition, .the discoveries of l i n e emissions from excited hydro-gen, and from OH, CO^, CH and other r a d i c a l s and molecules i n the galaxies, has spurred much i n t e r e s t i n the f i e l d of radio spectroscopy i n astronom-i c a l observations. At the Dominion Radio Astrophysical Observatory, at Penticton, B r i t i s h Columbia, an antenna system, employing the technique of 'Aperture Supersynthesis' has been constructed for the study of the 21 cm hydrogen l i n e . For the s p e c t r a l analysis of the signals received by the antenna system, a 256 - Channel D i g i t a l C r o s s - c o r r e l a t i o n Spectrometer has been designed and constructed. Employing a unique t h r e e - l e v e l by f i v e - l e v e l d i g i t a l multiplication-and-accumulation scheme, developed by D.S. Sloan and Prof. F.K. Bowers of the Univ e r s i t y of B r i t i s h Columbia [22], the instrument produces 256 c r o s s - c o r r e l a t i o n c o e f f i c i e n t s which are Fouri e r -Transformed, i n a PDP-9 computer^ to generate an^ estimate of the input power spectrum. 2 The author was r e s p o n s i b l e f o r the design of a n a l o g - t o - d i g i t a l , a r i t h m e t i c , c o n t r o l , timing and i n t e r f a c i n g c i r c u i t r y of the spectrometer and f o r the assembly of the whole of t h i s instrument. A d e t a i l e d account of the design i s given i n Chapters I I I and IV and i n the appendices. For the completeness of the d e s c r i p t i o n of the c r o s s - c o r r e l a t : 0 r . the d i s c u s s i o n s i n these chapters i n c l u d e the c o n t r i b u t i o n s of other persons to the o v e r a l l design and acknowledgements are made as ap p r o p r i a t e . The mathematical b a s i s f o r the design of the c r o s s - c o r r e l a t o r ^ * ' i s presented i n Chapter I I . In general, r i g o r o u s developments sup p o r t i n g s t a t e d r e s u l t s are not presented, however, references to t h e i r treatments i n the l i t e r a t u r e are given. In the present chapter, an i n t r o d u c t i o n w i l l be given to the 'Supersynthesis' technique, and i n p a r t i c u l a r a b r i e f d e s c r i p t i o n of the "Spectroscopic Super-synthesis Radio Telescope, at D.R.A.-O. w i l l be presented. 1 . 2 ASTRONOMICAL BACKGROUND For s p e c t r a l l i n e work, f i n e r e s o l u t i o n i s r e q u i r e d i n both of the s p e c t r a l (frequency) and s p a t i a l domains. Developments i n m u l t i c h a n n e l , f i l t e r and a u t o c o r r e l a t i o n spectrometry, capable of measuring 1 0 0 or more po i n t s on a spectrum have e a s i l y provided the r e q u i r e d improvements i n s p e c t r a l r e s o l u t i o n . However, instruments f o r ac h i e v i n g the r e q u i r e d an-gular r e s o l u t i o n , w h i l e s t i l l p r o v i d i n g adequate s e n s i t i v i t y have been developed only r e c e n t l y . I n c r e a s i n g the aperture of a conventional para-b a l i o d a l ' d i s h ' antenna meets a l i m i t a t i o n i n the design and cost of the correspondingly l a r g e c o l l e c t i n g s u r f a c e . The method of u n f i l l e d a p e r t u r e s , i n which a l a r g e array ( r e c t a n g u l a r , T-shaped or M i l l s - C r o s s ) of s m a l l c o l l e c t i n g elements, w i d e l y spaced and interconnected e l e c t r i c a l l y , has provided a means of c r e a t i n g almost u n l i m i t e d r e s o l v i n g power at moderate cost, but w i t h the s a c r i f i c e of c o l l e c t i n g area and hence s e n s i t i v i t y . 3 Any ma jo r i n c r e a s e i n t he depth o f r a d i o - a s t r o n o m i c a l s u r v e y s r e q u i r e s a c o m b i n a t i o n o f b o t h l a r g e c o l l e c t i n g a rea s and l a r g e a p e r t u r e s , e s p e c i a l l y i n t he ca ses o f s p e c t r a l - l i n e o b s e r v a t i o n s o f d i s t a n t g a l a x i e s , i n wh i ch the v e r y weak s p e c t r a l s i g n a l s a r e b u r i e d i n a l a r g e b a c k g r o u n d n o i s e . The t e c h n i q u e o f A p e r t u r e S y n t h e s i s i n t r o d u c e d a t the U n i v e r -s i t ^ o f Cambr idge , and d e s c r i b e d i n p r i n c i p l e i n 1960 [ 2 0 ] , p r o v i d e s an i m p o r t a n t r e v o l u t i o n i n the q u e s t f o r l a r g e , h i g h l y s e n s i t i v e an tenna s y s -tems. T h i s method i n v o l v e s the use o f two s m a l l p o r t a b l e a n t e n n a s , c o n n e c t e d as an i n t e r f e r o m e t e r , whose r e l a t i v e p o s i t i o n and o r i e n t a t i o n i s v a r i e d to c o v e r a l l p o s s i b l e s p a c i n g s w i t h i n an a r e a o f l a r g e a p e r t u r e , s i m u l a t i n g the p e r f o r m a n c e o f a s i n g l e a n t e n n a o f the same a p e r t u r e and l a r g e c o l l e c t i n g a r e a . When such an i n s t r u m e n t i s p r o v i d e d w i t h a s p e c t r o m e t e r r e c e i v e r , i t p r o v i d e s a p o w e r f u l , t o o l f o r i m p o r t a n t a s t r o n o m i c a l r e s e a r c h , w i t h h i g h s p a t i a l and s p e c t r a l r e s o l u t i o n and r e a s o n a b l e s e n s i t i v i t y . 1.3 APERTURE SYNTHESIS. AND SUPERSYNTHESIS C o n s i d e r an a r e a on the g round the s i z e o f the a p e r t u r e r e q u i r e d to p r o v i d e a d e s i r e d r e s o l u t i o n , and d i v i d e i t i n t o segments the s i z e o f a s m a l l movable a n t e n n a . I f we move a p a i r o f antennas c o n n e c t e d as a m u l -t i p l y i n g i n t e r f e r o m e t e r to a l l p o s s i b l e c o m b i n a t i o n s o f p o s i t i o n s w i t h i n the l a r g e a r e a , the p r o j e c t e d b a s e l i n e c o n n e c t i n g the an tennas w i l l change i n l e n g t h and o r i e n t a t i o n , ffift'm the p o i n t o f v iew o f a r e g i o n o f sky to w h i c h the antennas a r e d i r e c t e d . At each s p a c i n g the o u t p u t o f the p a i r o f antennas measures one s p a t i a l F o u r i e r - c o m p o n e n t o f the sk}7 b r i g h t n e s s d i s t r i b u t i o n , h a v i n g a s p a t i a l f r e q u e n c y e q u a l to ' the l e n g t h o f t h e p r o -j e c t e d b a s e l i n e ( i n w a v e l e n g t h s ) [ 2 0 ] , [ 2 1 ] . 4 The components from a l l such spacings comprise the s p a t i a l F o u r i e r transform of the sky b r i g h t n e s s d i s t r i b u t i o n , - an image of the l a t t e r may be obtained simply by a mathematical F o u r i e r i n v e r s i o n of the components. The highest s p a t i a l frequency obtained, and hence the s p a t i a l r e s o l u t i o n , i s equal to the longest b a s e l i n e p r o j e c t e d , or the o v e r a l l dimension of the l a r g e area. The region of sky which may be so mapped i s that which i s contained w i t h tie primary beam of the sm a l l antennas used. This technique 'synthesizes' the performance of an e q u i v a l e n t l a r g e antenna, w i t h diameter equal to the l a r g e s t b a s e l i n e , and of c o l l e c t i n g area equal to twice the geometric mean of the areas of the sm a l l antenna and the l a r g e aperture. In 1 s u p e r s y n t h e s i s ' , one telescope i s moved w i t h respect to the other to a l l p o s s i b l e separations i n one l i n e a r dimension, and the r o t a t i o n of the earth changes the angular o r i e n t a t i o n of the b a s e l i n e , as i t i s p r o j e c t e d towards the region of sky bein g observed. Figure 1.1 i l l u s t r a t e s t h i s technique f o r two antennas movable along an East-West l i n e . The p r o j e c t i o n of the b a s e l i n e , onto a plane p e r p e n d i c u l a r to the l i n e from the antennas to the source, i s represented by a v e c t o r whose coordinates (u,v) are the two-dimensional s p a t i a l f r e -quency measured by the antennas. This plane i s c a l l e d the 'u-v s p a t i a l frequency plane'. Thus, when the source i s at an hour angle H, a component has s p a t i a l frequency (u,v) given by u = 7> cos H (1.1) v = ' 7,';, sin6 s i n H where, s7 i s the s e p a r a t i o n of the antennas i n the East-West d i r e c t i o n IN ALONG LOCI OF VECTORS, PRODUCED BY 12 HOURS OF OBSERVATION AT EACH OF SPACINGS I Figure 1.1 The 'u - v Plane'. The projection of the vector l i n k i n g the two antennas changes as the Earth rotates through 12 hours. 6 (H = -6^), having maximum value L, 6" i s the d e c l i n a t i o n of the source or region of sky observed. The coordinates (u,v), though dimensionless q u a n t i t i e s , are usually termed cycles per radian to lend p h y s i c a l s i g n i f i c a n c e to the quantities, (that i s , u i s a measure of the number of cycles of a frequency component which are contained i n one radian of arc subtended by the source). The locus of the s p a t i a l frequencies sampled by the antenna-pair, at one spacing 1, i s an e l l i p s e of semi-major axis 1 and e c c e n t r i c i t y cos 6. The Fourier r e l a t i o n s h i p between the output of the interferometer, V(u,v), and the sky brightness d i s t r i b u t i o n , B(x,y), i s given as follows: x 12 y 12 VCu,v; =J J s B(x,y) exp [-j2ir (ux + vy) ] dxdy -x s/2 -y s/2 and (1.2) Um Vm B(x,y) - E E V(u,v) exp [+J2TT (UX + vy)] AuAv, -Um -Vm where (1.3) -2 -1 B(x,y) i s the sky brightness d i s t r i b u t i o n , i n Watts m Hz steradian \ over the region of sky being observed, x,y are rectangular coordinates, measured i n radians of arc sub-tended at the antennas by the source, V(u,v) i s the averaged output of the m u l t i p l y i n g interferometer, at the projected spacing u,v. I t i s c a l l e d the 'complex v i s i b i -- 2 -1 l i t y function' , Watts m Hz To map a source confined w i t h i n an area of sky of angular dim-ensions x and y (usually corresponding to the extent of the main lobe s s of the primary beam pattern of the small antennas), the s p a t i a l frequen-cies need be sampled only at i n t e r v a l s Au = 1/x , and Av = 1/y , i n s s 7 order to obtain a l l of the s p a t i a l frequency information up to a maximum defined by, u = L, v = L sin6 m m Au,Av are usually somewhat smaller than the diameter of each small antenna. In addition, the map obtained by Fourier i n v e r s i o n , needs to be derived only at points on a rectangular g r i d , of spacing Ax, Ay, given by -1 Ax = U , Ay = , m m i n order to provide the. r e s o l u t i o n a v a i l a b l e . These are roughly h a l f the main-lobe beamwidth of the synthesized secondary r a d i a t i o n pattern. A convenient s i m p l i f i c a t i o n i s the fact that samples are required only i n one h a l f of the u-v plane; since the brightness d i s t r i b u t i o n B('x,y) i s a r e a l function, the complex v i s i b i l i t y function must be Hermitian, that i s , V(-u, -v) = V* (u,v). Therefore observations at each antenna separation are required only f o r a twelve hour period. Although the mathematics of the Fourier Transform i s performed most e a s i l y i n rectangular coordinates, the observations are a c t u a l l y made i n polar coordinates, (s,<J>), given by T~2 2 s = u + v <j> = arctan (v/u), and then converted to the rectangular coordinates. As one antenna rotates (with the earth) about the other, t r a v e l l i n g a distance A(j>, the output of the mu l t i p l y i n g receiver i s integrated to comprise one d i s t i n c t component, The time required f o r the antenna to traverse t h i s distance i s greater for small separations than for large ones. Consequently the r.m.s. noise on the low s p a t i a l -frequency components (longer i n t e g r a t i o n time) w i l l be much less than that on the high s p a t i a l frequency components. 8 SUPERSYNTHESIS RADIO-TELESCOPE FOR H I SPECTROSCOPY Figure 1.2 Supersynthesis Radiotelescope for Spectroscopy, at the v Dominion Radio Astrophysical Observatory, Penticton B.C. I.F. SINE RECEIVER x(t) COSINE RECEIVER L.O. < xy . > SINE S ™ G PRODUCT y • (t) ' sine A/4 y(t) cosine R.F. PREAMPLIFIER I.F. DELAY INSERTION LsinB < xy . • > COSINE 'cosine P R 0DUCT Figure 1.3 Supersynthesis Receiver. 9 The brightness d i s t r i b u t i o n obtained from the Fourier i n v e r s i o n must a c t u a l l y be regarded as the true sky brightness d i s t r i b u t i o n smoothed by convolution with the reception pattern of the synthesized beam of the instrument. One advantage of the supersynthesis system i s that s u i t a b l e grading may be applied to the amplitudes of the various components sampled by the d i f f e r e n t antenna o r i e n t a t i o n s , i n order to minimize the sidelobe structure of, and reduce the main-lobe beamwidth of the synthesized antenna pattern. The observation time required to completely map a region of sky with a supersynthesis instrument i s of the same order as, but somewhat 1 longer than, the time required to scan the same region with an equivalent, large paraboliodal antenna of the same reso l v i n g power. However, the ad-: d i t i o n of a t h i r d antenna to the system, e f f e c t i v e l y creating two i n t e r -ferometers, reduces the time by h a l f . Most e x i s t i n g and proposed super-synthesis instruments involve three or more antennas (e.g. Green bank, Va. [10, 15]; Cambridge, England [21]; Parkes, A u s t r a l i a [2]). 1.4 SUPERSYNTHESIS RECEIVER A t y p i c a l receiver system for a supersynthesis instrument, i s shown i n Figure 1.3. Two features are to be noted, (i) Delay Insertion Signals c o l l e c t e d by the two antennas are c a r r i e d by trans-mission l i n e s to a;central point and m u l t i p l i e d . The d i f f e r e n c e between the paths t r a v e l l e d by the two s i g n a l s from the sources to the antennas must be compensated i n two ways. F i r s t , i n order to render the i n t e r f e r e n c e pattern of the interferometer stationary with respect to the region of sky observed, so that the phase information of the sky brightness d i s -10 t r i b u t i o n i s preserved, the phases o f the two signals must be continuously equalized, within an i n t e g r a l number of wavelengths. This i s provided by a continuous phase r o t a t i o n of the l o c a l o s c i l l a t o r signals to the r . f . mixers. Secondly, whenever the path difference becomes greater than a c e r t a i n f r a c t i o n of a wavelength, (depending on the bandwidth of the signals) a reduction i n c o r r e l a t i o n between the signals occurs. The path diffe r e n c e continuously varies as the source i s followed through a range of hour angles, and i s equalized by the i n s e r t i o n of delay cable i n one path. The i n s e r t i o n i s done i n small steps equal to the t o l e r a b l e path di f f e r e n c e , and i s c o n t r o l l e d along with the phase r o t a t i o n of the l o c a l o s c i l l a t o r s i g n a l s , by the computer, ( i i ) Sine and Cosine Receivers The v i s i b i l i t y funciton, V(u,v) i s a complex function, having both amplitude and phase which must be preserved i n order that the Fourier inversion may reproduce an image of the sky brightness d i s t r i b u t i o n . This requires that one s i g n a l be divided into two separate s i g n a l s , one of which i s phase s h i f t e d by 90°, and that each s i g n a l be c o r r e l a t e d with the t h i r d s i g n a l . Consider an example of a s i n g l e monochromatic s i g n a l a r r i v i n g at ;the two antennas with phase di f f e r e n c e <j). The two s i g n a l s to be cor-r e l a t e d may be written as cosine waves, v x = A 1 Cos (2-rrft) v 2 = A 2 Cos (2irft + <j0 (1.4) The product of these signals may be expanded, to f i n d v^v = A A 2 Cos iJ) [Cos 2t\£t Cos 2uft] -A1 A 2 Sin<{> [Cos _2uft Sin 2irft]. (1.5) The product consists of two parts. The f i r s t part, the 'Cosine' product, 11 i s produced by d i r e c t m u l t i p l i c a t i o n of the signals and contains i n -formation about the cosine of the phase. The second part, the 'Sine' product, i s y i e l d e d by m u l t i p l i c a t i o n a f t e r one s i g n a l has been s h i f t e d i n phase by 90°, and contains information about the sine of the phase. In general the phase information i n the Sine and Cosine products for polychromatic s i g n a l s i s recovered i n the process of Fourier i n v e r s i o n . A good review of the techniques of aperture synthesis was given by Swenson [24] i n 1969; the same author p a r t i c i p a t e d i n a presentation of the gen-e r a l theory of radio-interferometry [23], which deals with the d e t a i l e d mathematical aspects of the antenna and receiver systems. 1.5 THE D.R.A.O. SUPERSYNTHESIS TELESCOPE -' The instrument being constructed at the Dominion Radio Astro-p h y s i c a l Observatory, Penticton, B r i t i s h Columbia, i s depicted i n Figure 1.2. The i n i t i a l operation w i l l involve the two movable antennas (8.6 m diameter), already constructed, and plans c a l l f o r the addition of the other two stationary antennas (9.6m diameter) i n the near future. The basic 300 m instrument w i l l produce a r e s o l u t i o n of 2 arc minutes at a wavelength of 21 cm while the proposed 800 m instrument w i l l improve the r e s o l u t i o n to 45 arc seconds. The extent of the primary beam of the antennas, and hence the area of sky which can be mapped i n an ex-periment i s about 2 degrees. Because the u-v plane degenerates to a narrow e l l i p s e f o r sources at low de c l i n a t i o n s , plans also c a l l f o r the addition of a North-South arm extending from the mid-point of the East-West l i n e , i n order to broaden the u-v plane f o r such sources. As was mentioned e a r l i e r , the D.R.A.O. supersynthesis instrument i s to be applied l a r g e l y to the study of the 21 cm emission l i n e of atomic hydrogen, H . That i s , a s p e c t r a l p r o f i l e i n a range of frequencies near 12 1420 Mhz i s to be produced at each of the map points resolved by the i n -strument. For this purpose, each of the two c o r r e l a t i o n receivers gen-erating the Cosine and Sine products must be replaced by instrumentation for performing a s p e c t r a l analysis on each.spatial component sampled. The t o t a l system noise temperature i s expected to be approxi-mately 100°K, and the s t a t i s t i c a l f l u c t u a t i o n s on the map points produced at each frequency are expected to have an r.m.s. value of 1.6°K. 13 I I . INSTURMENTATION FOR SPECTRAL ANALYSIS 2.1 MULTI-FILTER AND CROSS-CORRELATION SPECTROMETRY The measurement of the cross-power spectrum P (f) of two s i g -xy nals x(t) and y ( t ) , bandlimited'to B Hertz, may be performed i n e i t h e r of • two ways as follows. ( i ) M u l t i - f i l t e r Spectrometer Each of two signals are applied to a bank of n contiguous narrow-band f i l t e r s , as i n Figure 2.1, whose centre frequences are spaced by Sf Hertz over an o v e r a l l bandwidth B equal to n6f. The outputs of the corresponding f i l t e r s i n each bank are m u l t i p l i e d and averaged, to y i e l d an estimate of P (f) at frequencies f r t + i 6 f , i = 0, n - 1, where f_ xy 0 0 i s the lower l i m i t of the band B. Each, f i l t e r has a half-power bandwidth Af, approximately equal to f, so that each point i n •the -estimate i s 'smeared' over a region of frequencies equal to Af. ( i i ) C r o s s - c o r r e l a t i o n Spectrometer The cross-power spectrum of the signals x(t) and.y(t) may be ca l c u l a t e d as the Fourier transform of the time c r o s s - c o r r e l a t i o n function-defined by, +T lim 1 f r x y ( x ) = 2T x ( t ) y ( t + T ) d t ' ( 2 > 1 ) where x represents the l a g of x(t) with respect to y ( t ) . The Fourier r e l a t i o n s h i p i s a form of the Weiner-Khinchin theorem; thus, P x y ( f ) = ^ ( T ) e " j 2 l T f T d T (2.2) x( 2n PHASE-MATCHED FILTERS BANDWIDTH Af = 6f CENTRE FREQUENCIES f0 f Q + 6 f f„+2<5f f Q + ( n - l ) 6 f Figure 2.1 M u l t i - f i l t e r Spectrometer x(t) TAPPED DELAY-LINE OR SHIFT-REGISTER LAG STEPS A T V y(t) PHASE-MATCHED FILTERS, BANDWIDTH B = n6f P (6f) xy P (26f) xy P (36f) xy r ( ( I I - I ) A T ) xv P (n6f) xy FOURIER TRANSFORM Figure 2.2 Cros s - c o r r e l a t i o n Spectrometer 15 In p r a c t i c e , the c r o s s - c o r r e l a t i o n function i s calculated over a truncated range of d i s c r e t e values of the l a g f > defined by, T = KAx , K = 0, n - 1, as i n Figure 2.2. An estimate P' (f) of the cross-spectrum i s then c a l -xy culated as a modified Fourier transform, P' (iSf) = A T E r' (iAx) e" j 2"' i < 5 f ( K A t ) (2.3) X y K=-n X y where r' (iAx) i s the estimate of the c r o s s - c o r r e l a t i o n function r (x). xy xy ' The s i z e of the delay unit Ax i s r e l a t e d to the o v e r a l l bandwidth, through the a p p l i c a t i o n of the Nyquist Sampling Theorem, by - • i The spacing 6f between estimated points on the cross-spectrum w i l l be found l a t e r to be, 2(n-l)Ax ' where n i s the number of c r o s s - c o r r e l a t i o n c o e f f i c i e n t s measured. A comparison of these two methods follows, ( i ) In observations with the supersynthesis telescope, data samples ( s p a t i a l Fourier components) c o l l e c t e d several months apart are combined to derive the f i n a l map of the sky brightness d i s t r i b u t i o n . The s t a t i s t i c a l f l u c t u a t i o n s i n the c r o s s - c o r r e l a t i o n c o e f f i c i e n t s for one sample are of order .01 percent, and the s p e c t r a l analyzer must be s u f f i c i e n t l y stable and d r i f t - f r e e to measure co r r e l a t i o n s down to t h i s value over an e n t i r e observation period. Both the f i l t e r and the c r o s s - c o r r e l a t i o n methods may be r e a l i z e d i n analog or d i g i t a l form. However, the f i l t e r technique i s s u i t e d mostly to analog c i r c u i t r y , i n which s t a b i l i t y and d r i f t - f r e e operation are d i f -16 f i c u l t to achieve over long periods of time. On the other hand, d i g i t a l instrumentation, for which s t a b i l i t y and d r i f t c r i t e r i a are v i r t u a l l y e l -inimated, i s su i t e d more to the c r o s s - c o r r e l a t i o n method. ( i i ) The n, phase-matched f i l t e r s required i n the m u l t i - f i l t e r method are much more costly than the n-section d e l a y - l i n e or s h i f t - r e g i s t e r of the c r o s s - c o r r e l a t i o n apparatus. In the supersynthesis instrument, moreover, a dual cross-spectrum i s to be measured, between x(t) and y (t) and between x(t) and y ( t ) , as i s explained i n s e c t i o n 2.2. The m u l t i -s p f i l t e r system would then require a t h i r d , complete set of n f i l t e r s and a second set of m u l t i p l i e r s and accumulators, whereas the c r o s s - c o r r e l a t i o n instrument can share i t s d e l a y - l i n e between the two sets of m u l t i p l i e r s and accumulators. ( i i i ) With d i g i t a l implementation the o v e r a l l bandwidth of the cross-co-rrelation instrument may be changed merely be a l t e r i n g the data sampling -rate f = 1/A T and replacing a few band-pass f i l t e r s . A change i n band-width i n the m u l t i - f i l t e r system requires complete replacement of a l l of the narrow-band f i l t e r s . (iv) The bandwidth of d i g i t a l equipment i s l i m i t e d by the speed of oper-ation of the l o g i c c i r c u i t r y . The emergence, i n recent years, of very high speed integrated l o g i c c i r c u i t s has made possible bandwidths of 32 MHz, or greater. (v) D i g i t i z a t i o n of the data causes a degradation i n the s i g n a l - t o - n o i s e (S/N) performance of a c o r r e l a t i o n instrument over that obtained for analog c i r c u i t r y (disregarding i n s t a b i l i t y and d r i f t ) . Increasing the f i n e -ness of the d i g i t i z a t i o n decreases the S/N degradation, but increases the cost and complexity of the instrument. The importance of the s t a b i l i t y and cost f a c t o r s , ( i ) and ( i i ) , 17 favours the use of a d i g i t a l c r o s s - c o r r e l a t i o n spectrometer. Past work has sought to improve the S/N performance of c o r r e l a t i o n instruments while main-t a i n i n g " the economy of simple design. The f i r s t d i g i t a l a u t o correlation instruments, [25], [8], employed one-bit quantization of the input s i g n a l . The s i m p l i c i t y of the design i s countered by a degradation of the S/N performance, over that of an analog instrument, by a factor of 1.57. Two-bit c o r r e l a t o r s have also been b u i l t [7], [22], with considerable improvement i n the S/N. An i n v e s t i g a t i o n of the commercial market reveals that the maximum bandwidth c a p a b i l i t y a v a i l a b l e , i n d i g i t a l instrumentation, i s of the order of. 250 KHz, whereas bandwidths up to 8 MHz are needed for the present ap-plication.' The unique design of a simple, t h r e e - l e v e l by f i v e - l e v e l d i g i t a l m u l t i p l i c a t i o n and accumulation scheme by F.K. Bowers and D.S. Sloan at the University of B r i t i s h Columbia, has made possible a s u b s t a n t i a l reduction i n the signal-to-noise degradation, with a small increase i n complexity, and operation at a bandwidth of 4 MHz. This m u l t i - l e v e l scheme w i l l be described i n s e c t i o n 2.4 along with the e f f e c t s of quantization on the S/N of the c r o s s - c o r r e l a t i o n mea-surement. Section 2.3 deals with the general features of a d i g i t a l cross-c o r r e l a t o r . F i r s t , however, a b r i e f summary of the theory of the Cross-c o r r e l a t i o n Spectrometer w i l l be given. 2.2 THEORY OF THE CROSS CORRELATION SPECTROMETER a. Fourier Relationship Most radio signals of i n t e r e s t i n astronomy o r i g i n a t e i n a s t o c h a s t i c p h y s i c a l process which i s stationary ( i t s s t a t i s t i c s are time-18. invariant) and ergodic (time averages taken on one sample are equivalent to averages taken at one instant over the ensemble of samples). In a d d i t i o n , the random process may be assumed to have Gaussian s t a t i s t i c s , since i t i s formed as the sum of a large number of random processes of a r b i t r a r y p r o b a b i l i t y density d i s t r i b u t i o n s . Thus, two signals x(t) and y(t) received by the antennas must be regarded as samples or r e a l i z a t i o n s of the r e a l , b i v a r i a t e , Gaussian random process ( X ( t ) , Y ( t ) } , which has the j o i n t p r o b a b i l i t y density d i s t r i b u t i o n , 1 P X Y ( x t ' y t + x ) . n 2 , ' J./2 ( 2TT a x a y [ l - p X Y ( x ) J -I exp - 1 2 ( 1 " P XY ( t ) } X 2P x Y ( T ) x t y t + x y t + T 'X CTX °Y "Y The f i r s t moments of the process are the expected values of X(t) E [X(t)] = y v (2.4) and Y ( t ) , and E [Y(t)] = p y , which are zero (in 2.4) for radio astronomical s i g n a l s . The second moments are. the auto-covariance functions, r x x ( x ) = E [(X(t) - ^ x ) ( X ( t + T ) - u x ) ] , r Y y ( T ) = E [(Y(t) - M Y ) ( Y ( t + r ) - M y ) ] , and the cross-covariance functions, r ^ d ) = E [(X(t) - u x ) ( Y ( t + x ) - y y ) ] , r y x ( x ) = E [(Y(t) - p Y ) ( X ( t + T ) - U x ) ] . (2.5) When the means are zero, these become the auto- and c r o s s - c o r r e l a t i o n functions; these terms w i l l be used henceforth. The following properties of these functions are e a s i l y proved. (1) r ^ C O ) - ax 2 = Variance of X ( t ) 2 ( i i ) r y Y ( 0 ) = a y ( i i i ) r x x ( - x ) = r x x ' ( T ) (even function), s i m i l a r l y f o r I C ^ ( T ) J (iv) f o r r e a l processes X and Y , R X Y ( " T ) = R Y X ( T )  R Y X ( _ T ) = R X Y ( T ) so that the covariance between the two processes may be completely des-cribed by the s i n g l e c r o s s - c o r r e l a t i o n function r (x) , XY where -°° £ x £ ro , which i s not ne c e s s a r i l y an even function. The term p (f) i n equation (2.4) i s the normalized cross-corre-Xx l a t i o n function, R X Y ( T ) / > X Y ( T ) = ' _ 1 * P X Y * X ' Now, the function which i s of i n t e r e s t i n the s p e c t r a l analysis i s the cross-power spectrum, P (f) which shows how the c r o s s - c o r r e l a t i o n Xi o between the two processes i s d i s t r i b u t e d with frequency. I t i s commonly shown i n text books on communication theory [e.g.-.11], that P (f) i s XY rel a t e d to r (x) by the Fourier transform X i P X Y ( F ) ^ ( x ) e j 2 ^ f dx , (2.6) where r (x), i n this case, i s the c r o s s - c o r r e l a t i o n function, XX Lim . r (x) = — X Y W • 2T 20 X(t) Y(t+x) dt. (2.7) Equation (2.6) i s usually used to define the cross-spectrum of a s t o c h a s t i c process. The p h y s i c a l s i g n i f i c a n c e of P (f) w i l l be seen s h o r t l y . - ' -b. The Sample Cr o s s - c o r r e l a t i o n and Sample Cross-spectrum Since only samples x(t) and y ( t ) , of f i n i t e duration, of the processes X(t) and Y(t) are a v a i l a b l e for the averaging process of equation (2.7), only estimates of the t h e o r e t i c a l C r o s s - c o r r e l a t i o n and Cross-spec-trum may be found. These estimates are c a l l e d the sample C r o s s - c o r r e l a t i o n r' (T) and the sample Cross-spectrum, P' ( f ) . Successive 'estimates' are xy r xy to be considered as r e a l i z a t i o n s of the two 'estimators' r' (T) and P' (f) xy xy which are themselves random v a r i a b l e s . The estimators are d i s t r i b u t e d over ranges of values whose s t a t i s t i c a l means are r e l a t e d to the. t h e o r e t i c a l values. A 'consistent' estimator i s one whose mean approaches the t h e o r e t i c a l value, and whose variance approaches zero, as the length of the record, T, approaches i n f i n i t y . I t may be shown [3], [11] that i f the range of values of x over which the c o r r e l a t i o n i s determined, i s l i m i t e d to + x , and — m T >> x , the s p e c t r a l estimator i s consistent. This condition i s met i n m a p r a c t i c a l c o r r e l a t o r . c. Preservation of Phase The cross spectrum P (f) i s i n general a complex quantity, and can be expressed i n terms of r e a l and imaginary parts as P X Y ( f ) = L X Y ( f ) " j Q X Y ( f ) > ( 2 < 8 ) I t may e a s i l y be shown that these parts are r e l a t e d to the c r o s s - c o r r e l a t i o n function, r (x) by the Fourier Transforms L X Y ( f ) = 1 X Y ( T ) C O S ( 2 l T f T ) d T ' ( 2 - 9 ) 2I> and Q x y ( f ) = J IxyC-O S l n (2lT f T) d t , (2.10) where 1 (x) and q (x) are the even and odd parts or r (x), as A X A l A l . ^ ( x ) = ^ ( x ) + q x Y ( x ) . (2.11) The quantities 1 (x) and q (x) are the cosine and sine pro-XY X i ducts of section 1.4 r e s p e c t i v e l y , calculated at lag x. Since they are known to be even and odd, they need only be measured for p o s i t i v e values of x. 2.3 THE DIGITAL CROSS-CORRELATOR a. Introduction The r e s u l t of the previous section i s that an n-point estimate of the complex cross-power spectrum of the input signals X(t) and Y(t) from the antennas may be formed by r e p l a c i n g the sine and cosine receivers of Figure 1.3 by a dual n-channel c r o s s - c o r r e l a t o r , whose outputs, 1' (x) Xx , (estimate of 1 (x)) and q' (x) are to be Fourier transformed, to y i e l d X i X I the r e a l and imaginary s p e c t r a l estimates L ' w ( f ) and Q' ( f ) . A l X l The c r o s s - c o r r e l a t o r i s to employ d i g i t a l c i r c u i t r y , and there-fore the input s i g n a l s are to be sampled i n ti:me, and d i g i t i z e d . The remainder of this chapter i s to deal with various aspects of these two processes. A s i m p l i f i e d configuration of the d i g i t a l c r o s s - c o r r e l a t o r i s shown i n Figure 3.1. b. Sampling The considerations to follow deal i n terms of the quantity. r' (x); however, i t i s to be assumed that the measured qu a n t i t i e s X i 1' (x) and q 1 (x) are to be treated separately i n i d e n t i c a l manners. X Y XY 2'Z The d e f i n i n g equations of the c r o s s - s p e c t r a l estimate, as i n ( 2 . 1 ) and ( 2 . 2 ) , must be modified f o r the p r a c t i c a l case of the d i g i t a l c r o s s - c o r r e l a t o r : P* (i6f) = Ax. E r * (kA T)d(kA T) e~ j 2 l T i 6 f ( K A X ) ( 2 . 1 2 ) k=-(n-l) ^ M r* ( k A T ) = ^ E x ( m A T ) y ( m A T ' + | k | A T ) . ( 2 . 1 3 ) X y m=l Three l i m i t a t i o n s are introduced: (i ) The l i m i t s on the times T and T (record length and time delay) are f i n i t e . ( i i ) The function r ' ( T ) i s determined at a f i n i t e number of d i s c r e t e xy values o f t , rather than over a continuous range. T = IC A T , - ( n - 1 ) s'k £ n-1 , ( 2 . 1 4 ) where n i s the number of c r o s s - c o r r e l a t i o n channels. Only c o e f f i c i e n t s with p o s i t i v e values -of kA'r need to be measured. The maximum delay'is T = (n-1) A T ( 2 . 1 5 ) m ( i i i ) The data signals are sampled i n time, at i n t e r v a l s At, so that samples; x(mAt) and y(mAt) are cro s s - c o r r e l a t e d , where m = 1, M and M >> n. I t i s usually convenient to make At = A T . The function d ( k A T ) i s introduced i n order to express the trun-cation of the c r o s s - c o r r e l a t i o n function at the maximum delay T . I t m has been c a l l e d a 'data-window' by Blackman and Tukey [ 3 ] and i s defined as d ( T ) = W ( T ) - T £ T $. T m m = 0 ITI $ T ( 2 . 1 6 ) m In the case i n which W ( T ) i s uniformly;' equals to unity, the data window s t r i c t l y defines the truncation of the c r o s s - c o r r e l a t i o n function. Otherwise, i t also prescribes a grading of the c r o s s - c o r r e l a t i o n c o e f f i c -i e n t s . The width and shape of the grading function determines the f r e -quency r e s o l u t i o n of the s p e c t r a l estimate, as s h a l l now be shown. Following procedures commonly ou t l i n e d i n text-books on sampled-data systems [e.g. 3], performing a Fourier transformation of the r i g h t -hand side of (2.15) y i e l d s P* (f) = Z xy J = - C C P' (f) D ( f — \ i f s ) d« (2.17) where, D(f) i s the Fourier transform of d(x) and i s c a l l e d the 'spectral window'. Examples of the functions P' ( f ) , D ( f ) , and P* (f) are shown xy xy i n Figure 2.3. From this diagram i t i s noted that: (i ) The s p e c t r a l estimate P' (f) i s smoothed by convolution with a narrow function D(f) whose half-power width i s i f 1 l / i , so that s p e c t r a l features m narrower than Af are not di s t i n g u i s h a b l e . The smoothed spectrum i s repeated p e r i o d i c a l l y along the f r e -quency axis, about i n t e g r a l values of f , the sampling frequency (f = 1 / A T ) , s s This e f f e c t i s due e n t i r e l y to the sampling of the data and of the cross-c o r r e l a t i o n function. Sampling, i n e f f e c t , i s a mixing operation, so that energy entering the c o r r e l a t o r w i t h i n the repeated bands or ' a l i a s e s ' along the frequency axis, are mixed down to the baseband of the instrument, represented by the ' p r i n c i p a l ' a l i a s . In order to reproduce the smoothed spectrum, i t i s required to re j e c t a l l but the energy i n one a l i a s by p r e - f i l t e r i n g of the s i g n a l s . In p r a c t i c e , to minimize d i s t o r t i o n a r i s i n g from harmonics of strong l i n e 24 a) Cross-power Spectrum of Input Signals D(f) —jr V- HAL F-POWER WIDTH, Af b) Spectral Window, or Smoothing Function, Af =l/x m r * - » f ALIAS USED IN PRESENT INSTRUMENT c) P ^ ( f ) , equation 2.17 Figure 2.3 E f f e c t of Sampling on the Spectral Estimate d ( T ) a) .! b) \ B ( f ) V- Af of order 1/T \ m Figure 2.4 Suitable Grading Function,a), a Truncated Gaussian Curve; and Corresponding Smoothing. Function, b ) . 25 spectra, i t i s preferable to use a f i l t e r with upper and lower cut-o f f frequencies equal to f 12 and f . s s Another problem evident i n Figure 2.3, c a l l e d ' a l i a s i n g ' , occurs when the convolution of P ' (f) and D(f) i s non-zero for | f | > f /2. In t h i s case adjacent a l i a s e s overlap and errors are caused i n the s p e c t r a l estimate at the edge of the band. This e r r o r may be reduced i n two ways. (i) The o v e r a l l bandwidth B of the signals passed by the p r e - f i l t e r s must be r e s t r i c t e d by B £ f g / 2 . (2.18) This i s the f a m i l i a r Nyquist Sampling Theorem. The passband of each p r e - f i l t e r i s to be f l a t , f a l l i n g o f f very sharply at the band-edges. ( i i ) The shape of the window function D(f) may be t a i l o r e d i n order to reduce the l e v e l of i t s sidelobes by adjusting the shape of d(x). Because d(x) must be zero outside the range |x| <^ m> i t i ' s n o t possible to produce a D(f) of minimal side-lobe structure, whose main-lobe width Af, (and hence the r e s o l u t i o n ) , i s much less than 1/x . A grading function which m provides a s u i t a b l e compromise i s shown i n Figure 2.4a. I t i s a truncated Gaussian curve, which y i e l d s the smoothing function of Figure 2.4b^. c. Number of Spectral Points Calculated Since the r e s o l u t i o n of the s p e c t r a l estimate i s equal to Af, i t i s s u f f i c i e n t to produce estimates of P * (f) only at d i s c r e t e values of f, spaced at i n t e r v a l s 6f. In order to recover s p e c t r a l d e t a i l to a r e s o l u t i o n Af, the value of the spacing i s 6f = 1/2 Af » 1_ ~ 2 x m ~ 1 ~ 2(n-l)Ax t S e l e c t i o n of these functions i s due to C.H. Costain, D.R.A.O. 26 ~ n-1 Therefore the number of points c a l c u l a t e d , over the bandwidth B, i s n. d. A p p l i c a t i o n to Present Instrument The r a t i o of B/ f chosen appropriately f o r observations of external galaxies with the Supersynthesis instrument i s 64. Therefore the number of channels required i n each side of the dual c r o s s - c o r r e l a t o r i s 128. Several of the channels are to be used f or s p e c i a l accumulations f o r normalization and D.C. corrections, so that only 123 channels are to be involved i n the c r o s s - c o r r e l a t i o n . 2.4 MULTI-LEVEL DIGITIZATION a. Quantization Quantization occurs whenever a p h y s i c a l quantity i s expressed num-e r i c a l l y ( d i g i t i z e d ) . Figure 2.5 i l l u s t r a t e s A OUTPUT 2d --d --5d/2 -3d/2 -d/2 d/2 3d/2 5d/2 INPUT - d -2d. QUANTIZATION INTERVAL Figure 2.5 Basic Quantization S cheme 27 the simplest of quantization processes, i n which the value of an input variable x occurring somewhere within a quantization i n t e r v a l of width d i s rounded-off to the value at the centre of the i n t e r v a l . The round-off error, c a l l e d 'quantization noise,' i s distributed over a range of values, + d/2. I t s presence degrades the l e v e l of r.m.s. fluctuations on the results of any manipulation of the quantized data. The f i n e r one makes the quan-t i z a t i o n i n t e r v a l s , the smaller i s the noise degradation. However, the corresponding increase i n the s i z e of the numbers required to describe the d i g i t a l c i r c u i t r y . A compromise between economy and accuracy must be made. Quantization i s a sampling process, s i m i l a r to that of amplitude sampling (time sampling). However, the l a t t e r i s an operation on a function i t s e l f , while the former may be viewed as an operation on the probability density d i s t r i b u t i o n of the function. As i l l u s t r a t e d i n Figure 2 . 6 PROBABILITY, P (x) Figure 2.6 Quantization Viewed as an Area-sampling of the Input Pr o b a b i l i t y Density Distribution 'the round-off process i t s e l f i s non-linear, however the probability density d i s t r i b u t i o n of the quantized signal i s seen to be a l i n e a r area-sampling 28 on that of the unquantized si g n a l . In the present application, two random variables are quantized, so that the sampling i s on the bivariate probability density d i s t r i b u t i o n , equation (2.4) as depicted i n Figure 2.7. The p r o b a b i l i t i e s of j o i n t quantized outputs are calculated by means of d i f f i c u l t integrations over the quantization squares. Figure 2.7 i s drawn for low values of the correlation c o e f f i c i e n t p O T. for Low Correlation of Input Signals The analysis of quantization i s described i n terms of the Stat-i s t i c a l Theory of Quantization [5], [26]. The "Quantization Theorem' states that i f the fineness of the quantization process, 2 /d, i s greater than twice the highest frequency component i n the shape of the input pr o b a b i l i t y density d i s t r i b u t i o n , then the d i s t r i b u t i o n may be succes-s f u l l y recovered from that of the quantized output. Moreover, i f the fineness i s equal only to the highest frequency component, moments such as the cross-correlation may be recovered and the quantization noise i s uni-29 formly d i s t r i b u t e d between + d/2'. Gaussian signals quantized with an i n t e r v a l s i z e of as much as three standard deviations s a t i s f y the l a t t e r condition to a good approximation. I m p l i c i t i n the 1 successful'recovery i s the fact that successive estimates w i l l f l u c t u a t e about a mean value which approaches the theoret-i c a l value as the length of the record increases. Moreover, the added s t a t i s t i c a l f l u c t u a t i o n s which are due s o l e l y to the quantization are causally r e l a t e d to the fl u c t u a t i o n s i n the input s i g n a l s . Therefore large c o r r e l a t i o n s i n the input process induce corresponding c o r r e l a t i o n s i n the quantization noise, which add a bias e r r o r to the c r o s s - c o r r e l a t i o n c o e f f i c i e n t s . This e f f e c t i s small for input c o r r e l a t i o n s below 20%. b. Signal-to-noise Degradation The c a l c u l a t i o n of the sig n a l - t o - n o i s e r a t i o for the cross-c o r r e l a t i o n c o e f f i c i e n t s , defined as, S/N <x'y'> c < ( x v ) 2 > 1 / 2 ' requires the c a l c u l a t i o n of j o i n t p r o b a b i l i t i e s , as i n Figure 2.7. S/N^ depends on the signal-to-noise r a t i o of the input s i g n a l s , S / N 0 = p x y ( x ) o s a + a S n a s o n where, s + n1> s.+ n 2, s i s the correlated signal- common•to the input s i g n a l s , (mean value zero), 30 n l ' n 2 a r £ "^ a r^ e' u n c o r r e 1 - a t e d noise s i g n a l s present i n x and y (mean values z e r o ) . a = a = 0 , n l n 2 and a >> a . f o r r a d i o - astronomical s i g n a l s , n s ° S/N^ a l s o depends on the d u r a t i o n of the averaging process. C a l c u l a t i o n s of S/N^ have been done f o r o n e - b i t , [25], t w o - b i t , [7], and h i g h e r - l e v e l [6] q u a n t i z a t i o n schemes. Bowers [4] has s t u d i e d the S/N behaviour of m u l t i - l e v e l schemes f o r a p p l i c a t i o n to the present instrument. The c a l c u l a t i o n s were g r e a t l y s i m p l i f i e d by the assumption of small c o r r e l a t i o n s of the input s i g n a l s ; i n t h i s case the b i v a r i a t e d i s -t r i b u t i o n i s separable i n t o u n i v a r i a t e d i s t r i b u t i o n " ' f o r x and y. He defines a degradation f a c t o r D, as, • S/Nn (analog system) S/N (quantized system) c For s m a l l c o r r e l a t i o n and Gaussian n o i s e , D i s separable i n t o D = D *D x y where D^ and D^ are s i n g l e channel degradation f a c t o r s depending only on the separate treatments of x ( t ) and y ( t ) i n the A/D converters. For a given choice of quantized values f o r l a b e l l i n g the output of an A/D converter, there i s an optimum choice of the q u a n t i z a t i o n , or d e c i s i o n , l e v e l s . The optimum values and t h e i r a s s o c i a t e d d e c i s i o n l e v e l s are u s u a l l y not spaced uniformly. However, m u l t i p l i c a t i o n i s much e a s i e r w i t h equi-spaced, i n t e g r a l v a l u e s , f o r which the optimum d e c i s i o n l e v e l s are also equi-spaced, and the penalty i n the S/N degradation i s s m a l l . Table 2.1 shows some s i n g l e - c h a n n e l degradation f a c t o r s f o r various q u a n t i z a t i o n schemes. Table 2.2 combines these f a c t o r s f o r two qua n t i z e r s . A current study by K l i n g l e r [12] i n v o l v e s complete i n v e s t i g a t i o n 31 of S/N dependence on the degree of quantization, and, i n add i t i o n , on the sampling rates. I t has been found that sampling rates greater than 2B r e s u l t i n a marked reduction i n the S/N degradation. A s i g n i f i c a n t im-provement occurs f o r f = 4B, but further improvement f or f > 4B i s very s s much l e s s . Sampling rates greater than f = 2B require the use of extra intermediate stages i n the s h i f t - r e g i s t e r , and f a s t e r logic,, so that a cost fa c t o r i s involved. The present instrument was designed to operate at f = 2B. s No. of Choice of Level ^ 2 values integral, values spacing " "x 2 -1,+1 - . 1.253 1.571 3 -1,0,+1 1.224xa 1.112 1.236 4 -3,-l,+l,+3 0 . 9 9 5 x 0 - 1.065 1.135 5 -2,-l,0,+l,+2 0 . 8 4 4 x 0 - 1.043 1.087 8 -7,-5, +5,,+7 0.585xa 1.019 1.039 Table 2.1 Single-channel Degradation Factors LOGIC D=DX'D 2 x 2 values 1.571 (Weinreb, n -. , , Davies) [25], [ 8 ] 2-value x analog 1.253 3 x 2 values 1.393 3 x 3 values 1.236 3 x 5 values 1.159*-Present Instrument 4 x 4 values 1.135 Tab le 2.2 System Degradation Faatov. (Cooper, Sloan) [7],[22] c. Three-level by F i v e - l e v e l D i g i t i z a t i o n The Multiplier-Accumulator l o g i c shown i n Figure 2.8, was 32 d e s i g n e d by F .K. Bowers and D. S l o a n a t the U n i v e r s i t y o f B r i t i s h C o l u m b i a , and u s e d s u c c e s s f u l l y i n an a u t o - c o r r e l a t i o n s p e c t r o m e t e r [22] u s i n g two-b i t q u a n t i z a t i o n and a s a m p l i n g r a t e o f 4 MHz. I t s d e s i g n may be seen v to be v e r y l i t t l e more c o m p l i c a t e d than t h a t r e q u i r e d f o r o n e - b i t q u a n -t i z a t i o n , y e t an improvement i s o b t a i n e d i n the S/N r a t i o , o v e r t h a t o f the o n e - b i t s y s t e m by a f a c t o r o f 1.36. x " , DELAY SIGNAL, 2-BIT BINARY CODE SHIFT PULSES P-SHIFT-REGISTER y SINGLE STAGE OF M.S.B, V V TO NEXT STAGE L. S . B . 1...11..1.11J1J1 . . y " ' P R 0 M P T P U L S E S -4 PROMPT LINE 'AND' FF 1 DIFFERENTIATE FF 2 FF 3 - 14 'AND' 'OR-' F i g u r e 2.8 M u l t i p l i e r - A c c u m u l a t o r - M o d u l e READOUT In the p r e s e n t i n s t r u m e n t , the M u l t i p l i e r - A c c u m u l a t o r l o g i c o f F i g u r e 2.8 i s to be u sed i n no rma l o p e r a t i o n a t s a m p l i n g r a t e s up to 8 MHz, u s i n g t h r e e - l e v e l by f i v e - l e v e l q u a n t i z a t i o n , as shown i n F i g u r e 2.9. Both q u a n t i z e d o u t p u t s a re encoded i n t o an a u g m e n t e d - l e v e l scheme, so t h a t o n l y p o s i t i v e p r o d u c t s w i l l o c c u r . T h i s e n a b l e s the use o f i n e x p e n s i v e U P - c o u n t e r s i n the A c c u m u l a t o r c i r c u i t , i n s t e a d o f t he more e x p e n s i v e and c o m p l i c a t e d UP-DOWN c o u n t e r s . 34 Correlator S h i f t - r e g i s t e r . Depending on t h i s code, the prompt pulses are gated into e i t h e r the f i r s t or the second stage of the Accumulator. At the sampling rate of 8 MHz, the maximum toggle-rate of both stages i s 32 MHz. The s e r i a l pulses delivered to the 'cosine' side of the MAM channel cards are c a l l e d the PROMPT PULSES, while those used i n the 'sine' side are c a l l e d the SEMIPROMPT PULSES. S i m i l a r l y the analog si g n a l s from which these are derived are c a l l e d the Prompt and Semiprompt s i g n a l s . 35 I I I . THE CROSS-CORRELATION SPECTROMETER 3.1 INTRODUCTION TO THE DESIGN A s i m p l i f i e d block-diagram of the dual Cr o s s - c o r r e l a t i o n Spec-trometer which has been implemented at the Dominion Radio Astrophysical Observatory, i s shown i n Figure 3.1. The design of the 256-channel Cross-c o r r e l a t o r , i s out l i n e d i n the o v e r a l l block-diagram of Figure 3.2. The basis of the design i s the use of the Multiplier-Accumulator Modules introduced in.Section 2.4c, and shown i n d e t a i l i n Figure 3.3. Fourteen stages of accumulation are included i n the MAM boards f o r each channel. However, the product rate at the rapid sampling frequency i s such that an extended accumulation f a c i l i t y i s required i n order to slow the data rate down to that which i s manageable by the PDP-9 computer. For this purpose, a Corning E l e c t r o n i c s D i g i t a l Glass Delay-line Memory was procured, which o f f e r s 4CJ96 b i t s of s e r i a l storage at a b i t -rate of 16 MHz, and with a t o t a l time-delay of 256 microseconds. This space i s conveniently segmented into 256, s i x t e e n - b i t Memory-words, and the outputs of the MAM channels are multiplexed successively at a 1 microsecond rate into the assigned storage s l o t s . An Arithmetic Unit i s required to s e r i a l l y update each s i x -teen-digit Memory-word every d e l a y - i n t e r v a l , on the basis of the mul-t i p l e x e d information from the channels. When the contents of the Glass Memory approach i t s capacity, the Memory-words are to be tra n s f e r r e d , v i a a Buffer Register and a Computer Interface, to a PDP-9 computer. Each Memory-word i s a c r o s s - c o r r e l a t i o n c o e f f i c i e n t . A f t e r each com-plete tra n s f e r of the c o e f f i c i e n t s to the Computer, they are a f a s t -Fourier transformed to y i e l d the Co- and Quadrature-spectral c o e f f i c i e n t s . The t o t a l accumulation i n t e r v a l between transfers i s to be approximately y ( t ) 30 MHz I.F. L . O . CONTINUUM CORRELATOR MIX PHASE-MATCHED BANDPASS FILTERS, BANDWIDTH B, CENTRE 33/2 SAMPLING RATE f « 2B y p ( t ) A/D M U L T A C C MIX x ( t ) A/D MIX F > A/D x ( t ) 123 SINE AND COSINE COEFFICIENTS AND 8 SPECIAL ACCUMULATIONS CONTINUUM PRODUCTS .CONTROL OF TELESCOPE AND FRONT END M E H M 0 R Y 16 LINES . GRADING AND FAST-ROUR1ER TRANSFORM MAGNETIC TAPE Figure 3.1 Simplified Block Diagram of the Cross-correlation Spectrometer (7> MULTIPLIER-ACCUMULATOR-MODULES ( MAM ) PROMPT CHANNELS PROMPT SIGNAL y >— P R O M P T O U A ' . T I Z E R I T V TT T. I FIVE-LEVEL SERIAL * X TT 123 MAM C A R D S I N C H A I N PROMPT PULSE DISTRIBUTION I 12 MHz S U P P R E S S " T H R E E - L E V E L D E L A Y _ P A R A L L E L S I G N A L D E L A Y Q U A N T I Z E R S E M I P R O M P T F I V E - L E V E L S E R I A L S I G N A L | ^ S E M I P R O M P T Q U A N T I Z E R T W O - B I T S H I F T -R E G I S T E R S E M I P R O M P T P U L S E D I S T R I B U T I O N 32 MHz S E M I P R O M P T C H A N N E L S m RESET DISTRIBUTOR I T D I G I T 13 (AQ) REATV-' M U L T I P L E X E R S -DIGIT 14 (Aj) 8 S P E C I A L C H A N N E L S F O R N O R M A L I Z A T I O N A N D C O R R E C T I O N S TT P R O M P T P U L S E =1 I N H I B I T O R w M A S T E R R E A D -M U L T I P L E X E R i i 2 CONTINUUM CHANNELS DIGIT 14 (A x) READ-MULTIPLEXERS DIGIT 13 (Ag) J1L SAMPLING PULSES . SHIFT PULSES. fs 32 MHz to Quantizers ' R E S E T D I S T R I B U T O R MASTER CLOCK CLOCK ENABLE 16 M H Z 8 - B I T < C O D E C H A N S E L A D D R E S S C O D E G E N E R A T O R 1/256 ' C O M T U T E R T R A N S F E R T I M E R T I M I N G U N I T U6 . R E S E T C O N T R O L T1M1NG SIGNALS TO ALL UNITS 16 s INTERVAL, NOMINAL ~1 AUTO STOP/START G E N E R A L R E S E T T O A L L U N I T S A R I T H M E T I C U N I T 16 MHz R E A D O U T C O N T R O L S U P P R E S S < -D . C . S U B T R A C T P R O M P T P U L S E S U P P R E S S I O N D . C . S U B T R A C T B U F F E R R E G I S T E R C I R C U L A T I N G G L A S S M E M O R Y , 4,096 B I T S : 256 S I X T E E N - B I T W O R D S SIXTEEN DATA LINES AND FLAG COMPUTER INTERFACE T O P D P - 9 - C O M P U T E R OJ Figure 3.2 Block Diagram of the 256-channel-Dual Cross-correlator. 38 SHIFT REGISTER SHIFT PULSES DELAY DIGIT 1, M.S.B. ** SHIFT PULSES RESET FROM RESET DISTRIBUTOR TO SECOND t* MULTIPLIER-ACCUMULATOR ON SAME CARD PROMPT PULSES IN ACCUMULATOR STAGE 1 9001 DIFFERENTIATOR ACCUMULATOR STAGE 2 9001 ACCUMULATOR STAGES 3 TO 14 7473 -3 UNITS OF FOUR T FLIP-FLOPS EACH DIGIT 12 & A DIGIT 14 DIGIT 13 • » DIGIT 11 Figure 3.3 Multiplier-Accumulator Module, Schematic Diagram. 39 sixteen seconds. Several such accumulations w i l l be combined, a f t e r the temporal frequency analysis has been performed, to constitute one S p a t i a l Fourier component of the sky brightness d i s t r i b u t i o n . The author was responsible for the design of a l l arithmetic, multiplexing, timing and control c i r c u i t r y i n the system. In addition he collaborated with T.L. Landecker, at D.R.A.O. i n the design of the A/D converters. The design of the Computer Interface was done by J.H. Dawson, at D.R.A.O. The 128 MAM channels are located i n eight c i r c u i t board enclosures, c a l l e d Channel Racks, along with those parts of the multiplexing and pulse d i s t r i b u t i o n networks associated with the s e r v i c i n g of the channels. A l l other c i r c u i t r y , i n c l u d i n g c o n t r o l , timing and A/D converters are located i n two enclosures which constitute the Central Processing Unit. The p h y s i c a l arrangement i s shown i n Plate 2 i n Chapter V. The d e t a i l e d design of each part of the system i s discussed i n Chapter IV. Several requirements on which the design i s based w i l l be presented i n the remainder of the present chapter. 3.2. REQUIREMENTS a. Modes of Operation The Cross-correlator i s to be operated at several d i f f e r e n t bandwidths as o u t l i n e d i n Table 3.1. Normal operation w i l l involve three-l e v e l by f i v e - l e v e l d i g i t i z a t i o n , and the maximum pulse rate into the MAM cards i s 32 MHz (4 pulses per sample). The nominal sampling rate i s 8 MHz, each sampling i n t e r v a l being of 125 r>s duration. For bandwidths less than 4 MHz, the only quantities to be changed are: the rate at which new samples are produced i n the quantizers; and the rate at which the delay d i g i t s are s h i f t e d along the. Correlator S h i f t - r e g i s t e r . For these slower sampling rates, r e p e t i t i v e samples are produced by the quan-t i z e r s i n each of the 125 ns sampling i n t e r v a l s which occur between 40 Table 3,1 Modes of Operation Bandwidth Delay and Delay D i g i t Maximum Prompt D i g i t Levels Prompt Rate Rate Delay x Sampling Prompt Rate, f s MHz MHz MHz MHz 0.25 0.5 0.5 32 3 x 5 0.5 1 2 1 2 4 1 2 4 32 22 *"Four pulses 22 per sample 3 x 5 3 x 5 3 x 5 4 8 8 32 3 x 5 8 16 16 32 -«-Two pulses per sample 3 x 3 (16) (32) (32) (32)-K)ne pulse per sample (3 x 2) Table 3.2 Encoding of Quantizer Outputs Delay Prompt (Semiprompt) Quantized Values x' -1, 0, 1 y' -2,-1, 0, 1, 2 Augmented Values x" 0, 1, 2 y" 0, 1, 2, 3, 4 P a r a l l e l Output (Code) x" 00,01,10 y" 00,00,01,10,10 S e r i a l Output (Pulses) x" 1, 2, 3 Y" 0, 1, 2, 3, 4 Table 3.3 R.M.S. Values of Quantized Signals. (for a ,a equal to 1 V o l t , (input s i g n a l s ) , x y and for optimum level-spacings) 2 * <(x') > = <(x' )(x' )> = 0.735 , measured i n s p e c i a l . ser. par. , ., • channel, ^^^24' <(y') 2> = 1.136 , desired quantity. <(y' )(y' )> '= 0.938, measured i n s p e c i a l channel w ser. J par. ' ^ ^126' ° r b F127' = 0.827 <(y') 2> 2 2 <(x') > and <(y') > are required for normalization. 41 sample pulses. Therefore the pulse rate i n t o the MAM c i r c u i t s i s main-tained so that the average product rate i s unchanged. The outputs of Accumulators can be sampled at a constant rate of f^ = 1/256 MHz. For bandwidths greater than 4 MHz, the prompt quantizers them-selves must be changed; the number of l e v e l s of quantization and hence the maximum number of pulses produced per sample must be reduced, i n order to maintain the maximum toggle rate, of 32 MHz i n the MAM c i r c u i t s . Quantizers were b u i l t only f o r the normal t h r e e - l e v e l by f i v e -l e v e l operation modes, however i n s t r u c t i o n s f o r the implementation of the faste r bandwidths are given i n Appendix A.2. b. D.C. Correction The accumulation i n a c o r r e l a t o r channel consists of products of x" and y", which are the augmented outputs of the.delay and prompt quantizers, according to the scheme introduced i n Section 2.4c. These quantities have D.C. o f f s e t s which accumulate as a f a l s e D.C. component i n the products, as follows: N* C " = E (x' + 1 ) (y 1 + 2) N* N* N* = E x'y' + 2 E x* + E y* + 2N* (3.1) The desired, unbiased accumulation i s then, N* C' = E x'y' N* N* = C.1 1 - 2 E x' - E y ' - 2N* (3.2) The unbiased quantity C1 may be obtained from the measured value C'' i f the bias terms are monitored. The second and t h i r d terms, the average values of the quantized s i g n a l s , should equal zero, but they w i l l have s t a t i s t i c a l f l u c t u a t i o n s . The fourth term i s an exact constant 42 value. Special channels are to be set aside to accumulate x', y' (one for each of prompt and semi-prompt quantizers), and N A (see Section 3.1b). For reasons to be explained i n se c t i o n 4.2, the constant term of the D.C. bias i s to be subtracted from the Memory-words as they are revised i n the Arithmetic Unit. The measured terms of the b i a s , w i l l be subtracted from each set of c o e f f i c i e n t s , a f t e r they are tra n s f e r r e d to the computer. Now, a s l i g h t e r r o r occurs i n the subtraction of the known part of the D.C. bia s . The number of samples, N*, which are a c t u a l l y m u l t i p l i e d i n the channels, exclude those samples which are suppressed under the f o l -lowing conditions, (i) while the delay cables are being switched (Section 4.8), and ( i i ) while the output of each i n d i v i d u a l Channel Accumulator i s being read by the Multiplexer (Section 4.4). However, D.C. subtraction i s only suppressed during the f i r s t circumstance. I t was found not f e a s i b l e to do so i n the second case, so that a c t u a l l y the number which i s subtracted from the Memory-words i s 2N. N i s greater than N* by an amount equal to the number of times the samples are suppressed i n ease ( i i ) , for each channel during one accumulation i n t e r v a l . This amounts to 125 nanoseconds i n every 256 microseconds, or about 0.05 per-cent. I t should be noted that the s p e c i a l channel set aside f o r the accumulation of 2N* also has the D.C. subtraction of 2N performed on i t , so that the corresponding Memory-word i s a measure of the d i f f e r e n c e AN = 2(N A -N). The accumulation of this net negative D.C. bias occurs i n a l l Memory-words; therefore each Memory-word should be corrected i n the computer, by the subtraction of AN. c. Normalization and Gain Control The cro s s - c o r r e l a t i o n , c o e f f i c i e n t s produced by the c o r r e l a t o r 43 must be normalized to a constant scale during an e n t i r e supersynthesis observation program. This may be accomplished by d i v i s i o n of the coef-f i c i e n t s by the standard deviations of the quantized s i g n a l s . That.is C' /N* p(kAx) = ^ 1c (3.3) N* N* Now, the quantization process i n the A/D converters requires that the standard deviation of the input signals be set at a nominal value of one v o l t . That i s , a gain control (automatic or open-loop) must be implemented to maintain the l e v e l of the input s i g n a l s . This regulation ensures that the mean square outputs of the quantizers, (re l a t e d to the input mean square l e v e l s by the quantization degradation f a c t o r , D) i n the denominator of equation (3.3) are constant, except f o r fl u c t u a t i o n s , and also that the c o e f f i c i e n t s are themselves always measured on the same scale. Therefore the d i v i s i o n of the c o e f f i c i e n t s i n (3.3) i s unnecessary and i n fact would only add to the noise on the c o e f f i c i e n t s . Now, the measurement of the mean square values of the quantizer outputs provides a convenient way of monitoring the gain of the r e c e i v e r s , to ensure thata = a = 1 v o l t . In addi t i o n these quantities provide a x y check on the operation of the quantizers, p a r t i c u l a r l y with respect to the s e t t i n g of the decision, l e v e l s . Therefore s p e c i a l channels are reserved for the measurement of these quantities f o r each quantizer. Because the receiver w i l l i t s e l f have s t a b i l i z e d l o c a l o s c i l l a t o r l e v e l s and an AGC'in the I.F. section, i t i s expected that a f a s t e r reacting monitor of the operation of the quantizers w i l l not be needed. I f the need does a r i s e , i t i s possible.to devise a simple averaging c i r c u i t to monitor the percentage of time i n which the decision l e v e l s are exceeded by the input 44 s i g n a l . d. Continuum Correlators In order to d i s t i n g u i s h the 21 cm s p e c t r a l - l i n e emission from a radio source, i t i s necessary to synthesize a map f o r the integrated continuum r a d i a t i o n emitted by the same source. Two continuum c o r r e l a t o r s are being designed by T.L. Landecker (D.R.A.O.), with p r e - f i l t e r s f o r r e j e c t i n g the s p e c t r a l information. The two continuum s p a t i a l c o e f f i c i e n t s are to be processed by the Glass Memory and Arithmetic Unit. (Through channels P-j.23 a n c* S P123' w*i:i-c*1 t l i e M A M w i l l be i n a c t i v e ) , and trans-ferred to the computer at the same regular i n t e r v a l as are the cross-c o r r e l a t i o n c o e f f i c i e n t s . The separate continuum map drawn from these c o e f f i c i e n t s w i l l provide a two-dimensional base, or reference plane, for the s p e c t r a l maps. e. Overflow S t a t i s t i c s In Section 4.2 i t w i l l be shown that the r e v i s i o n process i n the Arithmetic Unit r e q u i r e s that the l a s t two stages, 13 and 14, i n the Channel Accumulators must overlap with the f i r s t two l e a s t s i g n i f i c a n t d i g i t s i n the corresponding Memory-words. Therefore the t o t a l number of stages of accumulation i n the Channels and Memory i s 28. For the purpose of designing the Arithmetic Unit and the Read-out Control, i t i s useful to determine the time required f o r the capacity of the Memory to overflow, 'for the maximum product rate p o s s i b l e i n the m u l t i p l i e r s . Table 3.4 shows the r e s u l t s of t h i s determination for the case i n which the f a l s e D.C. component 2N*, of equation (3.2)-, i s allowed to accumulate i n the Memory, and for. the case i n which i t i s subtracted from the numbers i n the Arithmetic Unit. 45 Table 3-4 Overflow S t a t i s t i c s No D.C. Correction D.C. Correction Input 0.0 Correlat 1.0 ion -1.0 ' Input 0.0 Correlat P CO xy 1.0 , ion -1.0 Mean Product <x"y"> 2.0 2.744 1.256 0.0 0.744 -0.744 r.m.s. Deviation a ,, „ x y 0.836 1.076 1.076 0.836 1.076 1.076 Maximum Accumulation i n a Channel During One Multiplexing-cycle (256 us) L>(1.372 ± .048) x 2 1 2 (Before t D.C. Mean ± 4o ,, „ , Subtraction) x y • Capacity of Memory plus Channel Accumulator 2 2 8 - l ± 2 2 7 - l One stage for s i g n -b i t Number of Samples to Overflow, N (rounded o f f to next lower i n t e g r a l power of 2) 226 2 2 7 Overflow I n t e r v a l , N/f 8.4 seconds 16.8 seconds Range of S i g n a l - l e v e l s i n Memory-words (decimal equivalent of the sixteen d i g i t binary number) 0 to + 44,379 1 0 to ±24,379 1 Noise-level i n Maximum Accumulation, f o r 100% Corr e l a t i o n ~8 1 -12 t Note: f g = 8 MHz; however the table i s v a l i d for slower sampling rat e s , since for these, samples are repeated rather than-eliminated, and' the product rates are maintained i n the m u l t i p l i e r s , The accumulation of noise w i l l be increased (by a maximum fac t o r of f o u r ) , but w i l l be n e g l i g i b l e compared to the s i g n a l l e v e l . 4 6 The overflow c a l c u l a t i o n i s done f o r f u l l y c orrelated quantizer outputs x" and y", which produce the maximum product rate. The s t a t i s t i c s for f u l l y a n t i - c o r r e l a t e d s i g n a l s , as w e l l as uncorrelated s i g n a l s , are shown as an i n d i c a t i o n • o f the range of numbers which may accumulate i n the Memory. Although correlations of no greater than 15 percent are ex-pected i n radio-astronomical s i g n a l s , i t i s required to design for 100 percent c o r r e l a t i o n s , p a r t i c u l a r l y for t e s t purposes. The c a l c u l a t i o n of the number of samples required to overflow the capacity i s done according to the equation, <x"y"> N + 4o. „ „ YW $ CAPACITY. x y The s o l u t i o n for N i s rounded o f f to the next lower i n t e g r a l power of 2, i n order to f a c i l i t a t e the timing of the i n t e r v a l between tran s f e r s to the computer. f. Correction for Quantization The quantization of the data signals x(t) and y(t) induces a bias error.on the c o e f f i c i e n t s of the c r o s s - c o r r e l a t i o n estimate, which depends on the degree of c o r r e l a t i o n i n the input s i g n a l s , and which i n general may be eliminated by the a p p l i c a t i o n of a knox^n c o r r e c t i o n on the c o e f f i c i e n t s . For one-bit quantization, a r e l a t i o n e x i s t s , c a l l e d the Van Vleck c o r r e c t i o n [ c i t e d i n 26], between the quantized and unquantized normalized c o r r e l a t i o n c o e f f i c i e n t s : TT p (x) = s i n p , , (x) ] Kxy 2 Kx'y J For h i g h e r - l e v e l quantization, no simple a n a l y t i c r e l a t i o n may be found because c a l c u l a t i o n of p , . involves i n t e g r a t i o n s over f i n i t e x y ° square regions on the b i v a r i a t e d i s t r i b u t i o n s of equation (2.4). A num-e r i c a l i n t e g r a t i o n method was used however, to c a l c u l a t e P x t I ( T ) . versus p (x) for the t h r e e - l e v e l by f i v e - l e v e l quantization scheme, and i s 47 shown i n Figure 3 . 4 . P Figure 3 . 4 C r o s s - c o r r e l a t i o n C o e f f i c i e n t , p r, , ( T ) , of the x y Quantizied Signals vs. P (T) of the Unquantized Signals. For input c o r r e l a t i o n s l e s s than 0.7, the curve i s s u f f i c i e n t l y l i n e a r so that the correction applied to a l l c o e f f i c i e n t s would be a constant m u l t i p l i c a t i v e factor which would change only the normalization scale of the measurement. Since correlations greater than 0.2 w i l l be rare f o r radio a s t r o -nomical s i g n a l s , i t appears then that no correction process i s required, g. Encoding the Quantized Signals In order to accumulate the mean square output of each quantizer, using some of the MAM channels, the quantized outputs must be made a v a i l -able i n both s e r i a l (pulses) and p a r a l l e l (binary code) form. Since D.C. subtraction (2N) i s to be performed on the outputs of these s p e c i a l channels, the format of the input signals must be such that the mean product, per sample, i s 2, and must be compatible with the th r e e - l e v e l by f i v e -48 l e v e l design of the MAM c i r c u i t s . Table 3.2 outlines the formats adopted. The f i v e - l e v e l quantizer must produce a t h r e e - l e v e l p a r a l l e l output. Consequently the mean square product accumulated w i l l be d i f f e r e n t from that appropriate to the use of a l l f i v e l e v e l s . Hoxvever f o r an r.m.s. value^ of the input s i g n a l ^ of 1 v o l t , the two mean square products w i l l be r e l a t e d by a known factor, (Table 3.3). h. Special Channels A t o t a l of nine channels must be reserved for s p e c i a l accumula-tions of the following q u a n t i t i e s : mean and mean-square values of the outputs of the three quantizers (6 channels); unsuppressed sample pulses (] Two continuum c o r r e l a t o r channels (2). In addition, one channel i s l e f t blank. Therefore 123 c r o s s - c o r r e l a t i o n channels remain for each of the sine and cosine sides of the c o r r e l a t o r . The c o r r e l a t o r s h i f t - r e g i s t e r terminates at channels ^-^22 and SP^22' ^lae modifications of the MAM boards for the s p e c i a l channels are very s l i g h t . Table A.3.1 i n Appendix A.3 demonstrates the use of the s p e c i a l channels. i . Glass Memory The c i r c u i t of the Glass Memory i s shown i n Figure 3.5, modified s l i g h t l y with respect to the commercial design, for t h i s a p p l i c a t i o n . The device accepts d i g i t s at a 16 MHz rate, i n the non-return-to-zero format, and retimes them i n the input f l i p - f l o p . A p i e z o e l e c t r i c transducer i n j e c t s them into a quartz glass delay l i n e . The output of the delay l i n e ( v i a another transducer) i s amplified i n two stages, the second of which reshapes the s i g n a l with a Schmidt-trigger ac t i o n . The output of the delay l i n e i s subject to the interference of i n t e r n a l r e f l e c -t i o n s , however the response of the f i r s t a m p l i f i e r i s tuned to minimize Figure 3 .5 Schematic Diagram of the Glass Memory, Mod i f i e d from the O r i g i n a l C i r c u i t of Corning E l e c t r o n i c s f o r t h i s A p p l i c a t i o n . 50 the i n t e r a c t i o n between adjacent d i g i t s . The output of the amplifiers i s passed through a tapped delay l i n e with a range of about 55 ns i n ten steps. The d i g i t s are re-clocked i n the output f l i p - f l o p s and r e c i r c u l a t e d through the Arithmetic Unit to a r r i v e , revised, at the input. The timing of the clocks and the adjustment of the tapped delay l i n e anable the t o t a l delay around the loop to be trimmed to 256 microseconds. 51 IV. DESIGN OF THE CROSS-CORRELATOR 4.1 ANALOG-TO-DIGITAL CONVERSION a. Introduction In the introductory sections on the design of the c o r r e l a t o r , the scheme of a n a l o g - t o - d i g i t a l conversion of the input s i g n a l s was pre-sented. We noted that the design of the m u l t i p l i e r s and of the accumulation channels was s i m p l i f i e d by the use of an "augmented l e v e l " system of encoding the d i g i t a l information, i n which only p o s i t i v e numbers (and products) are represented. The two p a r a l l e l 128-channel c r o s s - c o r r e l a t i o n systems, for the processing of the 'sine' and 'cosine' s p a t i a l Fourier components, require three 'quantization' units, as shown i n the block diagram, Figure 3 .2, along with appropriate systems for d i s t r i b u t i n g the outputs of those units. The f i r s t of these, the Delay Quantizer, i s shared by both c r o s s - c o r r e l a t i o n systems, as i t produces the t h r e e - l e v e l , two-bit binary code which i s en-tered into the Correlator S h i f t - r e g i s t e r . In addition, t h i s quantization i s to y i e l d a t h r e e - l e v e l s e r i a l output c o n s i s t i n g of one, two or three pulses per sample period, corresponding to the three quantization l e v e l s . This second output i s to be used i n the s p e c i a l channels required to ac-cumulate the mean and mean square output of the quantizer. The second and t h i r d u n i t s , the Prompt and Semiprompt Quantizers, -a-are to produce the non-delayed smaples for the two c r o s s - c o r r e l a t i o n sys-tem. The s e r i a l output of each of these i s to consist of zero, one, two, three, or four pulses per sample, corresponding to the f i v e l e v e l s of quantization. In addition, for use i n the s p e c i a l channels, a t h r e e - l e v e l , p a r a l l e l , • two-bit binary output i s required, as i n Table 3 . 3 . The d e s c r i p t i o n of the design of these quantization systems involves three separate problems: the sampling of the analog data and i t s 52 q u a n t i z a t i o n i n t o d i s c r e t e , d i g i t a l l e v e l s ; the encoding of the d i g i t a l i n f o r m a t i o n i n t o the p a r a l l e l and s e r i a l outputs as described above; the d i s t r i b u t i o n of these outputs to the various l o c a t i o n s i n the c o r r e l a t o r system. The r e s p o n s i b i l i t y f o r the f i r s t two of these concerns was i n i t i a l l y undertaken by T.L. Landecker at the Dominion Radio A s t r o p h y s i c a l Observatory. The author, i n developing the pro c e s s i n g i n t e r f a c e between the MAM channels and the computer, and i n assembling the c o r r e l a t o r system., became i n v o l v e d mostly w i t h the t h i r d problem. However, i n order to f a c i l i * -t a t e the d i s t r i b u t i o n of the q u a n t i z e r outputs, the author made some mod-i f i c a t i o n s of the encoding system, and consulted w i t h T.L. Landecker over p o s s i b l e improvements i n the design of the sampling and q u a n t i z i n g c i r c u i t r y . For these reasons, and f o r completeness, a d i s c u s s i o n i s given i n t h i s s e c t i o n of a l l three of the design problems, b. Sampling and Qu a n t i z a t i o n At i n t e r v a l s equal to T g = 1/2B where B i s the o v e r a l l bandwidth of the c o r r e l a t o r , each of the three input analog s i g n a l s ( b a n d l i m i t e d to B H ) i s to be sampled, and the v o l t a g e , so obtained, rounded o f f (or quantized) according to e i t h e r a t h r e e - l e v e l or f i v e - l e v e l scheme. These two operations are mathematically commutable. The a v a i l a b i l i t y of f a s t comparators i n i n t e g r a t e d c i r c u i t form i n d i c a t e d a preference f o r q u a n t i -z i n g f i r s t and then sampling the s i g n a l s . Except f o r the number of comparators r e q u i r e d , the b a s i c con-f i g u r a t i o n f o r the t h r e e - l e v e l and f i v e - l e v e l q u antizers are i d e n t i c a l . In Figure 4.1.1, the schematic diagram f o r the Prompt (or Semiprompt) Quantizer, the q u a n t i z a t i o n s e c t i o n c o n s i s t s of four 710 integrated, c i r -c u i t comparators. One input of each comparator i s set at a reference voltage as shown; as the other inputs f o l l o w the inp u t analog s i g n a l s , the PROMPT OR SEMIPROMPT SIGNAL I N , 1 VOLT R . M . S . 4 SAMPLING R E G . - 6 V PULSES Figure 4.1.1 Schematic Diagram of the Prompt or Semiprompt Quantizer. DELAY SIGNAL I N , 1 VOLT R . M . S . t> p s o u 1 R E G . - + 12 V , ~ 2 . 7 K.Q £ 15 YSI | 5 ™ 330 H P < 2 Kfi 710 330 ^ 1 - 3 : < " 5 ,< 2 K s COMPARATORS REG. - 6 V SAMPLING PULSES 6b — [ ) ^ y — » H E > — * — E E D — ^ , „ _ DELAY PULSES M. t>. B . 3 - L E V E L P A R A L L E L OUTPUT L . S . B . 9c M l O a P j 6d P i l O b P i l l c H l l d ) - i ) 5 b 5 - L E V E L S E R I A L OUTPUT A A A B GATING SIGNALS 32 mz FROM MASTER CLOCK Figure 4.1.2 Schematic Diagram of the Delay Quantizer. 55 OUTPUT COl&ARATOR B FLIP -FLOP REQUIRES STABLE INPUT DURING THIS INTERVAL INPUT SIGNAL TO QUANTIZER OUTPUT COMPARATOR A SAMPLING PULSE TIME Figure 4.1.3 Response of The Comparators. If t u and t s overlap, an error may occur i n the response of one sampling f l i p - f l o p . If t u + t s> t r an error may occur i n two f l i p - f l o p s simultaneously. 32 MHz SIGNAL JUTJlJirirL PULSES PER SAMPLE PERIOD '0,2,.: 4 1 i A J * 1 _ t 2. + 60 n " n . r rui J U GATING SIGNALS OUTPUT PULSES FOR ZERO P U L S E S , NONE OF GATING SIGNALS IS ENABLED Figure 4.1.4 Encoding of S e r i a l Outputs of the Quanti zers, 56 comparators turn on or o f f depending on whether the analog s i g n a l voltage i s higher or lower than the reference voltages. The outputs of the com-parators (TTL l o g i c l e v e l s 0 and 4 v o l t s ) are sampled by the Type-D f l i p -flops i n response to the r i s i n g edge (sampling instant) of the sample pulses at the clock terminals of the f l i p - f l o p s . . The values recorded are held at the outputs of the f l i p - f l o p s between sampling i n s t a n t s . Figure 4.1.2 of the Delay Quantizer, shows a s i m i l a r arrangement, c. Performance C r i t e r i a In an i d e a l quantizing and sampling configuration, the trans-i t i o n s of the outputs of the comparators would be instantaneous, as the input voltage t r a n s i t s the reference l e v e l s , and the time required for the copying of the comparator outputs by the f l i p - f l o p s would be i n f i n i t e s -imal. The non-ideal behaviour of the r e a l comparators and f l i p - f l o p s i s characterized as follows: 1. The response of the comparator to the input s i g n a l display a 'hysteresis' e f f e c t . That i s , the device turns on when ' and o f f again when y ( t ) > v r e f < + e + y(t) < V , - e J r e f . where e + and e_ are small o f f s e t voltages. The response when y(t) l i e s i n the range V f ± e + i s ambiguous. The values of e+ and e_ may also depend on the past h i s t o r y of y ( t ) , on power f l u c t u a t i o n s , or on tempera-ture changes. The d i s t o r t i o n of the 'true' quantized s i g n a l that t h i s re-presents and also the error caused i n the s p e c t r a l measurement, are d i f -f i c u l t to analyze or c a l c u l a t e . The e f f e c t of the hysteresis may be minimized by incr e a s i n g the mean-square l e v e l of the input s i g n a l , as well as the reference voltages, so that the o f f s e t s e + and z_ are n e g l i g i b l e i n comparison with the s i z e 57 of the quantization i n t e r v a l s . The 710 comparators have t y p i c a l values of 2 mV for these o f f s e t voltages, compared to the quantization i n t e r v a l of .844 v o l t s f or the Prompt Quantizer when 9" = 1 v o l t . It appears that t h i s e f f e c t then should not be. troublesome. A measurement of the e f f e c t might involve the observation of changes i n the spectrum obtained as the mean-square l e v e l of the input s i g n a l , from the same source, i s a l t e r e d (along with the reference voltages). 2. There has been observed a d i f f e r e n t turn-on and turn-off time measured between the input and output of the 710 comparator, an e f f e c t which may or may not be s o l e l y a manifestation of the h y s t e r e s i s mentioned i n 1. Such a dif f e r e n c e of about 4 ns has been observed i n simple tests with large input pulses. This amounts to about 10° of phase uncertainty i f applied to the maximum frequency component, 8 Mhz, contained i n the input s i g n a l y ( t ) . This phase s h i f t may produce a f a l s e c o r r e l a t i o n . The phase ambiguity should be reduced to a value l e s s than 5°, the tolerance placed on the p h a s e - s t a b i l i t y of the octave f i l t e r s at the input to the c o r r e l a t o r . I t i s not known to what extent the delay d i f f e r e n c e varies with the r i s e time of the input s i g n a l i n t r a n s i t through the reference voltages. Careful measurements are required to determine the seriousness of this e f f e c t . S p e c i f i c a t i o n s for the 710 comparator show that the pro-pagation delays and r i s e times of the output varies considerably with the extent of input overdrive (the amount by which the input voltage exceeds the reference voltage). For a 2 mV overdrive the propagation delay (measured at the TTL threshhold l e v e l of 1.5 v o l t s on the output waveform) i s about 60 nanoseconds, while the r i s e time (of the output waveform) i s about 50 nanoseconds; a 20 mV overdrive y i e l d s a propagation delay of 20 ns and a r i s e time of 15 ns. I t appears then that unless the 58 input s i g n a l c l e a r l y t r a n s i t s the reference voltage the response of the comparator may be slow, i n terms of a 125 ns sampling- i n t e r v a l (8 Mhz). Again, however, 20 mV i s small compared to the .884 v o l t quantization i n t e r -v a l . I t appears that any one comparator w i l l be i n serious uncertainty, as to whether i t i s on or o f f , a small percentage of the time. In the following paragraphs we s h a l l study the p o s s i b i l i t y of two comparators being "uncertain" simultaneously, a circumstance which could cause a large error. 3. When an unquantized s i g n a l i s sampled by a pulse of shape 1 p(t) , ( i . e . the s i g n a l i s averaged for the duration of the pulse), the spectrum of the s i g n a l i s m u l t i p l i e d by the square of the magnitude of the Fourier Transorm of p ( t ) . If the duration of the pulse i s very much smaller than the sampling period, this e f f e c t i s n e g l i g i b l e . A r e l a t e d problem i n the case -of the quantizers i n the present c o r r e l a t o r , i s the time required for the sample-and-hold f l i p - f l o p s to recognize the states at their, inputs. This property i s characterized by a 'set-up-time' and a 'hold-time', the i n t e r v a l s before and a f t e r the t r i g g e r i n g edge of the clock pulse (sampling pulse) during which the l o g i c a l value at the input must be stable. I f a . t r a n s i t i o n i n the output of a comparator occurs during this 'stable' i n t e r v a l , which may be c a l l e d t , s the response of the f l i p - f l o p w i l l be uncertain. I t i s expected that average value of the -comparator output during the time t , p o s s i b l y weigh-ted according to the previous state of the f l i p - f l o p , w i l l determine the response of the f l i p - f l o p . This e f f e c t i s s i m i l a r to those of 1. and 2. above, i n that i t contributes to a smearing of the 'decision l e v e l s ' (threshold voltages) of the comparitors. I f , i n f a c t , the response of the f l i p - f l o p s i n such a circumstance depends on i t s previous s t a t e , the error produced w i l l correspond to a f a l s e c o r r e l a t i o n between successive 59 samples. As suggested by S. Weinreb [26] i n h i s d e s c r i p t i o n of a one-bit c o r r e l a t o r , the error may be randomized by the a p p l i c a t i o n of a pulse simul-taneously to the SET and RESET terminals of the f l i p - f l o p a f t e r each sample. The f l i p - f l o p randomly, resets or sets. This precaution was not taken i n the quantizers i n the present c o r r e l a t o r ; however i f i t i s found necessary, i t may e a s i l y be implemented, provided that the pulse i s applied a few nanoseconds before the end of each sample i n t e r v a l ( i . e . before the suc-ceeding sampling edge occurs). This ensures that the encoding process, to be described below, i s not disrupted. An-advanced version of the sampling pulse might be used for t h i s purpose, with the appropriate edge d i f f e r e n -t i a t e d to provide a narrow pulse. There i s some concern that the sample-and-hold f l i p - f l o p s a v a i l a b l e i n the ordinary TTL families of integrated c i r c u i t s , cannot meet the need for a small stable period t . The type-D f l i p - f l o p s (MC 3060) used i n the present c o r r e l a t o r have a t of about 15 ns, which i s only about 1/8 of the shortest sampling period,(125 ns). Type J-K f l i p - f l o p s have a t of about 8 ns, and may be used i n conjunction with the complemen-tary outputs of the comparators i n a sample and hold c o n f i g u r a t i o n . Howr ever, with the recent advent of the f a s t e r , Schottky-clamped TTL c i r c u i t s , the time t may be reduced to about 3 ns. (The comparators themselves a c t u a l l y have a strobe f a c i l i t y , but i t does not o f f e r a f a s t e r reaction time than those mentioned above). The i n t e r p r e t a t i o n of the term 'stable' may also be questioned. The 710 comparators have r i s e times at t h e i r outputs of as much as 60 ns, but more t y p i c a l l y of about 15 ns. However there may be a region about the TTL l o g i c threshold l e v e l of 1.5 v o l t s , i n which the output of the com-parator may be int e r p r e t e d neither .as a stable '0' or a s t a b l e 60 We may define the time-interval during which the output of the comparator, when i t changes state, t r a n s i t s t h i s voltage region as t , which must be added to the time t , as an i n t e r v a l of uncertainty. That i s , i f the s t r i g g e r i n g edge of the sampling pulse occurs at such a time that the i n t e r -v a l t g associated with i t overlaps with the region of uncertainty, t , on the r i s i n g (or f a l l i n g ) slope of a comparator output, the outcome of the sampling w i l l be subject to doubt. This s i t u a t i o n i s i l l u s t r a t e d i n Figure 4.1.3. '.The l a r g e r this i n t e r v a l of uncertainty' i s , the more frequent w i l l be the occurrence of erroneous samples. I t i s evident that the i n t e r v a l t i s reduced by a f a s t e r r i s e time on the output of the comparator. I t i s estimated that for the 20 ns rise-time for the 710 comparator, t i s about 5 ns i n duration. The p r o b a b i l i t y of the occurrence of erroneous samples i s d i f f i -c u l t to c a l c u l a t e . However, a more c r i t i c a l requirement i s that (when t h e i r inputs are sampled) no two f l i p - f l o p s (In a quantizer) should exper-ience the uncertainty, described above, simultaneously. The errors produced i n such a circumstance would not only be frequent, but large i n magnitude (possibly ranging over three quantization l e v e l s ) . This s i t u a t i o n could a r i s e i f the t o t a l ' i n t e r v a l of doubt' (t + t ) for the comparators s u and f l i p - f l o p s i s as large as the smallest time (t ) required for the en-velope of the analog input s i g n a l to traverse the voltage range between the reference voltages of two comparators ( i . e . one quantization i n t e r v a l ) . The f a s t e s t time for such a t r a n s i t i o n would be produced by an 8 MHz (highest frequency) component of large amplitude i n the input s i g n a l . For example, such a component of amplitude 3 v o l t s (3o*) would traverse the smallest quantization i n t e r v a l of .844 v o l t s ( i n the Prompt Quantizer) i n a minimum of about 6 ns, (cf. t = 5 ns), while for a:x amplitude of 1.0 v o l t the time i s 17.6 ns. In the present comparators and f l i p - f l o p s , t + t ~ 20 ns. 61 I t appears, then, that frequent large e r r o r s , of the kind under consid-e r a t i o n , may i n fact occur. There are a number of devices a v a i l a b l e which may be used as comparators, with f a s t e r rise-times than that of the 710, (for example, theu A703 l i m i t i n g - a m p l i f i e r , rated for frequencies over 100 Mhz). Such devices should be considered for improvements i n the performance of the quantizers. It should be r e a d i l y possible to reduce t + t to 6 ns. At s u the same time the d i f f e r e n t i a l turn-on and turn-off times of the comparator should be reduced to a t o l e r a b l e 2 ns. Presumably the h y s t e r e s i s e f f e c t s may also be minimized. d. Encoding the Quantizer Outputs The d i g i t a l outputs of the sample-and-hold f l i p - f l o p s of the quantizers are to be translated into, the a p p r o p r i a t e : ' p a r a l l e l and • s e r i a l codes, as o u t l i n e d i n Table 3.3. The pulses of the s e r i a l outputs must be uniform i n shape and spacing i n order that t h e i r q u a l i t y may be preserved i n the trans-mission from the Central Processing Unit to the i n d i v i d u a l MAM's. There-fore, i n the quantizers, they are produced by the s e l e c t i v e gating of a common 32 KHz pulse s i g n a l by means of s p e c i a l gating waveforms generated i n the Master Clock. The arrangement i s shown i n Figure 4.i.4. One of the gating waveforms i s selected f o r each sample accord-ing to the outputs of the f l i p - f l o p s i n the quantizer. For example, i n Figure -4.1.1, the gating signals are ORed through gates 8a to 8d and 13a. However only one (or more) of the enabling gates 6a, to 6d i s on during each sample period. Thus the appropriate number of pulses of the 32 VJRZ s i g n a l , i n one sample period, are passed through gate 13b. A s i m i l a r scheme i s used for the Delay Quantizer. 62 The two d i g i t s of the binary code for the p a r a l l e l output of each quantizer, are taken from the appropriate outputs of the f l i p - f l o p s , or l o g i c , to s a t i s f y the requirements of Table 3.3. e. D i s t r i b u t i o n of Pulses The p a r a l l e l outputs of each quantizer are d e l i v e r e d d i r e c t l y to the f i r s t MAM s h i f t r e g i s t e r stage, i n the case of the Delay Quantizer, or to the appropriate s p e c i a l channel boards (Table A.3.1). The s e r i a l out-puts of the Prompt and Semiprompt Quantizers are delivered along 100 ohm coaxial cables to each of the eight Channel Racks, where they are d i s t r i -buted through the Prompt Pulse I n h i b i t o r Network (Section 4.4) to the channels. The pulses are buffered at both the sending and r e c e i v i n g ends of the transmission cables, Figures 4.1.5 and 4.1.6. The Prompt Pulse Buffer c i r c u i t of Figure 4.1.5 provides also f or the b u f f e r i n g of the s e r i a l output .of .the Delay Quantizer, which i s delivered to the s p e c i a l channels through the gates i n the eighth Channel-Rack Buffer card. 63 SUPPRESS 12b 4a 100 U PAUSE *\—\ . 1 A E3—dD~^  PROi-tPT SIGNAL FROM QUANTIZER 12" SEMIPROMPT SIGNAL FROM QUANTIZER 4b OUTPUTS TO EIGHT CHANNEL RACKS packages 4 ,3 ,7 ,8 3b 2a 2b OUTPUTS TO EIGHT IT)-—t^ H)~IJ~^ ~ ^ > CHANNEL RACKS ! packages 2,1,5,6 6b JD—^ 9b » j l Q ^ 1 1Q<?) [) ^ — !> TO EIGHTH CHANNEL RACK SERIAL DELAY SIGNAL FROM QUANTIZER Figure 4.1.5 Prompt and Semiprompt Pulse Buffers i n CPU Rack. + 5 V | 1 KO 100 a COAXIAL CABLE PROMPT •»—c> i r r J F > IN 4- .2 u  <T 1 ' 2" SEMIPROMPT It) SIFT SHIFT PULSES CP-RMP 250 n <1 4-J -'-a -£> PROMPT PULSES OUT SEMIPROMPT PULSES OUT -£ SHIFT PULSES OUT CP-RMP SERIAL DELAY PULSES P124 1/6 THIS SECTION. IS CONTAINED IN CARD FOR EIGHTH CHANNEL RACK ONLY + 5 V ' a 10 K f J 5 :!HHJE)HIL){E)^ 5 6 P F 4 1 Kfl £> to P 124 SERIAL DELAY T O S P m ° U T 5d 5a H 5 ! > SAMPLE COUNT INHIBIT PULSES FROM PROMPT PULSE INHIBITOR ' TO N* COUNTING CHANNEL S P 126 Figure 4.1.6 Channel Rack Buffer Cards. 64 4.2 ARITHMETIC UNIT a. Function of the Unit In the 256-Channel Cross-correlator, two modes of accumulation are to be used: the fourteen-stage counter of each Multiplier-Accumulator Module, and extended accumulation f o r each channel i n the C i r c u l a t i n g Memory. The memory has a capacity of 4,096 binary d i g i t s which i s divided into 256 segments or si x t e e n - b i t words. The arithmetic unit i s the core of the processing i n t e r f a c e between these two accumulation f a c i l i t i e s . Its function i s to p e r i o d i c a l l y update the contents of the Glass Memory on the basis of information received from each of the Accumulator channels, sampled i n turn. . In th i s section, the design of th i s arithmetic operation i s to be discussed. Two general schemes are f e a s i b l e : "(1) The Glass Memory i s capable of accepting or presenting, at i t s input or output, binary d i g i t s i n s e r i a l , non-return-to-zero (NRZ) format. In t h i s f i r s t scheme, i l l u s t r a t e d i n Figure 4.2.1a the Arithmetic Unit operates on the d i g i t s as i t receives them, updating them s e r i a l l y , and del i v e r a i n g them to the input of the memory. (2) A l t e r n a t i v e l y , as i n Figure 4.2.1b, an e n t i r e s i x t e e n - b i t xrord may be loaded into a temporary arithmetic r e g i s t e r , updated combinatorially, and then s h i f t e d out to the input of the memory as a new word i s s h i f t e d into the r e g i s t e r . b. Selection of Scheme The time required for the r e v i s i o n of each binary d i g i t , i n scheme (1) i s included i n the t o t a l c y c l i n g time of 256 microseconds, consisting of propagation time i n both the Glass Memory and the d i g i t a l e l e c t r o n i c s . Synchronism between the operation of the memory and that of the channels i s e a s i l y maintained i n the 256 microsecond c y c l e . In scheme 65 OUTPUT SIXTEEN-BIT MEMORY-WORDS GLASS MEMORY 255 15/16 u s -INPUT DELAY • 255.92 us REVISION ELECTRONICS DIGITS FROM MULTIPLEXED CHANNEL-ACCUMULATORS REVISED MEMORY-WORD TOTAL CYCLE TIME, 256 us SERIAL ADDITION: SINGLE DIGITS PASS THROUGH REVISION LOGIC a) S e r i a l Revision OUTPUT SIXTEEN-BIT MEMORY-WORDS GLASS MEMORY 256 us-— SHIFT-REGISTER AND REVISION ELECTRONICS .1 V S 2 [Ja: 312 1 2 DIGITS FROM MULTIPLEXED CHANNEL-ACCUMULATORS -I INPUT REVISED MEMORY WORD TOTAL CYCLE TIME 257 us PARALLEL ADDITION: ENTIRE MEMORY-WORD IS LOADED INTO THE REGISTER BEFORE REVISION IS DONE b) P a r a l l e l Revision Figure 4.2.1 Two Possible Schemes for the Revision of the Memory-words from Data i n the Channel-Accumulators . 66 (2) the loading of the r e g i s t e r requires that each, d i g i t of a s i x t e e n - b i t word i s available outside the quartz-glass d e l a y - l i n e for at l e a s t 1.micro-second per t o t a l cycle. Since the propagation time i n the Glass Memory i t s e l f i s 255.92 microseconds, the t o t a l cycle time must expand to at l e a s t 257 microseconds. Synchronism between the• operations of the Glass Memory and the multiplexing of the channel outputs would be maintained be reducing the master clock-rate by 1/256. Scheme (1) was chosen for i t s s i m p l i c i t y over scheme (2), and a design was produced f o r an arithmetic r e v i s i o n c i r c u i t which could operate on the d i g i t s at the 16 MHz d i g i t rate. c. Design of the Revision Arithmetic: General Features Before discussing the s p e c i f i c design of the Arithmetic Unit based on scheme (1),above, we w i l l discuss some general features o f the. r e v i s i o n routine. 1. Revision Algorithm In Table 3.4 we noted that i n the 256 microsecond i n t e r v a l be-12 tween readings of a channel, an average number, (1.372 + .048) x 2 w i l l 13 accumulate. The 14th stage of the Accumulator (weight 2 ) w i l l therefore change state at most once during that time. In successive i n t e r v a l s the 14 channel w i l l o c c a s i o n a l l y accumulate to i t s capacity of 2 . In extending the accumulation to the higher order d i g i t s contained i n the Memory, i t i s ^necessary to determine, at each reading, whether capacity has been reached and surpassed, and having detected that event to generate a Revision-14 Carry b i t into the operation on the d i g i t of weight 2 , contained i n the Memory-word. Table 4.2.1, shows that the overflow event can be unambiguously detected by the comparison of 'old' ( l a s t reading) and"'new' (current reading) values of the states of the l a s t two stages of the Channel-67 Table 4.2.1 Use of D i g i t s from the.Channels i n the Revision Process Previous Current Number of Revision-Carry Readin g Reading Changes i n * D i g i t , C D, into ** D i g i t 2 1 4 'Old' Values 'New' Values D i g i t 2 213 2 1 2 213 212 0^) <v ( A ^ (A Q) 0 0 0 0 . 0 0 .0 1 1 0 1 0 2 0 1 1 3 0 0 1 0. 0 3 1 0 1 0 0 1 0 1 0 1 1 2 0 1 0 0 0 2 1 0 1 3 1 1 0 0 0 1 1 1 0 1 1 0 0 1. 1 0 1 2 1 1 0 3 1 1 1 0 0 *Note: A count of 4 or greater (2 ) during one i n t e r v a l between readings i s s t a t i s t i c a l l y very rare and may be considered impossible. The maximum count for 100% c o r r e l a t i o n , adding four standard deviations to the mean accumulation, i s (1.372 + .048) x 2 1 2 =• .36 x 2 1 4 (see Table 3.4) The Boolean expression f o r C D, from t h i s table, i s C = M A + M A A + M M A R 1 1 0 0 1 0 1 0 = M 1A 1 • M QA 0 • ( M ^ ) Accumulator, i f one assumes that d i g i t 2 changes state at most three times during the i n t e r v a l (four changes corresponds to a s t a t i s t i c a l l y very rare count or 1 x i . 68 This comparison requires that the Memory-word must overlap the Accumulator stages, i n the d i g i t s 13 and 14, so that the 'old* values of these d i g i t s are retained i n the memory. At the time of a reading, the old and new values are to be compared, a Revision-Carry d i g i t generated, and the new values stored i n place of the o l d . To conform with the fa c t that these d i g i t s w i l l occupy the two l e a s t s i g n i f i c a n t places i n the memory word, and that they derive from -the channel Accumulator, we w i l l l a b e l them AQ and A^ (Accumulator d i g i t s 13 and 14 respectively) as they enter the Arithmetic Unit. We have seen that the unambiguous r e v i s i o n of the Memory-words requires that only stages 13 and 14 of the Accumulators need be read. Now, the neglect of stages of l e s s e r s i g n i f i c a n c e introduces s t o c h a s t i c round-o f f e r r o r s ; however, i t w i l l be shown, as follows, that these errors are small r e l a t i v e to the t o t a l r.m.s. noise i n the accumulations and that t h i s neglect i s j u s t i f i e d . I f stages 13 and 14 of the Accumulators are read, and also the coarse D.C. subtraction i s performed i n the Arithmetic Unit, the t o t a l cap-27 a c i t y ox the Accumulators plus the Memory i s (2 - 1) per channel. The maximum p r a c t i c a l accumulation i n t e r v a l before t h i s capacity i s overflowed 27 i s 2 sample periods, or 16.8 seconds1. To constitute the minimum indepen-dent accumulation i n t e r v a l of 96 seconds for one s p a t i a l Fourier component i n the s k y - d i s t r i b u t i o n , at l e a s t f i v e such 16.8 second accumulations w i l l be. combined, by addition. A c t u a l l y a nominal number of 8 such accumulations per independent set, or s p a t i a l sample, has been adopted. In t h i s t o t a l time, the accumulated r.m.s. noise l e v e l , f o r 100% correlated input si g n a l s to a channel, w i l l be approximately. r.m.s. noise = ^a^rn nr,<y\ x vxTumber of samples 69 = 4 x 1.076 x 8 x 2 12 ' = 36.8 x 2 . I f the l e a s t s i g n i f i c a n t d i g i t recorded i n the channel accumulation 12 i s to be 2 , the round-off error due to neglect of l e s s e r d i g i t s i s 12 11 evenly, d i s t r i b u t e d between 0 and 2 , with an r.m.s. value of (1//3) x 2 = 12 .29 x 2 . This er r o r i s induced at the beginning and/or at the end of a set of accumulations which are grouped together and treated as one independent sample. At the moment at which one accumulation i s separated from a subse-quent one, an uncertainty i n the values of e i t h e r arises due to the unknown contents of those stages of the channel Accumulators which are not read. In the normal operation of the c o r r e l a t o r , the Accumulator stages of each.channel are to be reset at the end of each separate set of combined accumulations. The e f f e c t i s to induce the round-off e r r o r , above, at the end of each set. I f , as we w i l l decide l a t e r , the subtraction of the f a l s e D.C. bias i s to be done to the same accuracy, the t o t a l round-off e r r o r w i l l 3 2 be il l a r g e r , equal to .41 x 2 ' . This amounts, though, to only about 1.1% of the t o t a l r.m.s. noise, above, accumulated i n the same i n t e r v a l . I f r e s e t t i n g i s not performed, the separation of sets of accum-ulations w i l l induce the .round-off error at both the beginning and the end of each set, so that the t o t a l round-off error i s increased by a f a c t o r of v7!, to .58 x 2 . This option w i l l be desired, for example, when more than one set of accumulations are themselves to be combined as one indepen-dent sample, as i n the case of small antenna spacings i n the supersynthesis instrument. In t h i s case, however, the t o t a l r.m.s. noise i n the accum-u l a t i o n w i l l also be increased, so that the round-off e r r o r remains small i n comparison. It appears, then, that the neglect of the d i g i t s of lower order 70 than 2 i s j u s t i f i e d on these grounds. 2. Memory Overflow and D.C. Subtraction The choice of whether or not to subtract the f a l s e D.C. component i n the accumulated products, i n the Arithmetic Unit, influences the over-flow time of the capacity of the Glass Memory. Because of the overlap of the Channel-Accumulators with the Memory-words, required-by the algorithm just described, the t o t a l capacity of the combination of these i s 2 8 b i t s per channel. I t i s shown i n Table 3.4 that at the rate of accumulation of products for 100% c o r r e l a t e d input s i g n a l s , the time required f o r this capacity to be overflowed i s 16.8 or 8.4 seconds, depending on whether or not the D.C. component i s subtracted. Complex c i r c u i t r y i s required to. compute the exact value of t h i s D.C. component and to perform the.subtraction. However, i t w i l l be r e c a l l e d , from Equation 3.1, that the D.C. component consists of a large known p o r t i o n , equal to twice the number of samples i n the accumulation, as well as a small unknown portion whose s t a t i s t i c a l mean i s zero. E f f i c i e n t use of the memory capacity may be achieved by the subtraction of the known portion i n the Arithmetic Unit, while complex c i r c u i t r y can be avoided by the separate accumulation of the unknown portion, to be subtracted l a t e r i n the computer. Three s p e c i a l channels are reserved f o r the purpose. 12 -The known portion i s equal to 1 x 2 counts per M u l t i p l e x i n g -cycle, and therefore on each passage of a Memory-word through the A r i t h -metic Unit, t h i s q u a n t i t i y may be subtracted from the f i r s t d i g i t , of 12 weight 2 . This i s inconvenient, since the desired optional use of the D.C. subtraction feature would complicate the r e v i s i o n algorithm of Table 14 4.2.1. A l t e r n a t i v e l y , the quantity 1 x 2 may be subtracted on every fourth 71 passage of a Memory-word, operating only on the t h i r d d i g i t , of weight 14 2 . This i s e a s i l y implemented with the use of the 2 s complement format adopted for the arithmetic. 3. Format of Storage When the coarse D.C. subtraction i s performed as described above, the numbers i n the Memory can cover a range of p o s i t i v e and negative values about zero. Therefore one of the storage d i g i t s must be used to designate the sign of the number: l e t us assign t h i s r o l e to the most s i g n i f i c a n t d i g i t . We already know that the two l e a s t s i g n i f i c a n t d i g i t s are stored values of the sampled stages of the Channel Accumulators, and that these two d i g i t s alone comprise a p o s i t i v e number. The stored Memory-word can thus be viewed as a combination of two parts as follows: i ) M M M M xj n 1 5 u2 1 0 Z X X X X — X X O.'O . „ 7 i / • i o + 2 2 7 2 1 4 2 1 2 i i ) 0 0 0 0 0 — O O Y Y Z X X X X X X Y Y , inhere, Z i s the s i g n - d i g i t , Y i s one of the stored values derived from the two output stages of the channel Accumulator, X designates the other d i g i t s of the Memory-wordm determined by. the continual updating of the number of the basis of changing values of the Y d i g i t s , number i ) i s eit h e r a p o s i t i v e or negative number, and number i i ) i s always a p o s i t i v e number. There are three formats which suggest themselves f o r the repre-sentation of the f i r s t part of th i s number: ' si.gn-plus-number 1 , 2's com-plement, and l ' s complement. 72 A comparison of the c i r c u i t r y required i n each case decidedly favours the use of the 2's complement format, which i s also, compatible with the requirements of the PDP-9 computer, d. Det a i l s of the Design The use of 2's complement arithmetic enables the complete re-v i s i o n process to be implemented by a binary f u l l - a d d e r , as i n Figure 4.2.2. The subtraction of the binary number 0000 0100, where the s i n g l e d i g i t 12 1 i s of weight 2' , i s performed by the addition of i t s 2's complement 1111 1100. This number i s s e r i a l l y added to the d i g i t s of the Memory-word on every fourth passage of the l a t t e r through the adder, under the control of the v a r i a b l e D ( f o r D.C. subtrac t i o n ) . The Revision-Carry d i g i t Cn, generated by a comparison of the R f i r s t two d i g i t s of the Memory-word x^ith new d i g i t s from the Channel Ac-cumulator, i s selected by the switch to act -as the CARRY-IN to the operation 14 on the t h i r d d i g i t (2 ) of the Memory-word. The CARRY-IN to the operations on a l l other d i g i t s i s the CARRY-OUT from the preceding d i g i t . When both D and C have the value 1, the net change of the Memory-word i s n i l . U is. A high-speed binary f u l l - a d d e r , F a i r c h i l d Semiconductor Incor-porated' s MSI 9304 was used, whose delay-time from the input to the sum output i s les s than 30 ns. A d e t a i l e d d e s c r i p t i o n of the c i r c u i t of the Arithmetic Unit i s given i n Appendix B.2, however the present discussion may be completed by the study of the s i m p l i f i e d arrangement i n Figure 4.2.3. The f u l l - a d d e r receives the s e r i a l , d i g i t s of the Memory-words from the clocked output of the Glass Memory. The f i r s t two d i g i t s , M^ and M^, are sampled and held i n f l i p - f l o p s C and D. These are the old values of d i g i t s A and A from the previous sampling of the p a r t i c u l a r channel. MEMORY-WORD,^ ^-L.S.B., 2 SERIAL INPUT -ZXXXX XXYY 12 SIGN DIGIT M 15 DIGIT M. D =0 00000—000 ; =1 11111 110 r— -a* CARRY-IN C =0 KKKKK K00O =1 KKKKK K100 ' SWITCH TO C„ FOR DIGIT M„ A SUM FULL B ADDER C I N COUT -> REVISED MEMORY-WORD CARRY-OUT DELAY,ONE BIT TIME CR DC OPERATION a) 0 0 NO CHANGES TO DIGITS IN ADDER b) I 0 ADDITION OF 0000 00100 VIA CARRY-IN TERMINAL c) 0 1 ADDITION OF 1111 11100 VIA TERMINAL B d) ' 1 1 b) and c) COMBINED: NET EFFECT - NO CHANGE Figure 4.2.2 The S e r i a l Revision Scheme Adopted for the Arithmetic Unit. DIGIT A Q GATED, BECOMES NEW M Q v DIGIT Aj GATED. BECOMES NEW Mj V DIGITS S„ to S,c GATED NEW MEMORY WORD, DIGITS M Q TO M 1 5 TO MEMORY TO BUFFER REGISTER FOR CARRY INTO DIGIT M„ D.C. SUBTRACT 1MHz CLOCK Figure 4.2.3 Simplified Schematic Diagram of the Arithmetic Unit. 75 The nev; values of these d i g i t s are clocked i n t o f l i p - f l o p s A and B and compared with M and M to produce the Revision-Carry d i g i t C„. • U 1 R In i t s normal condition, switch G allows the carry-out from the addition at one bit-time to be ca r r i e d into the addition at the succeeding b i t - t i m e . However at bit-time t^, the switch i s pulsed to sample the value of C^, providing i t as the c a r r y - i n f o r d i g i t M^. When the D.C. component i s to be subtracted, at every fourth r e v i s i o n of the Memory-words, D = 1, and terminal B of the adder i s set at the l o g i c a l '1' value. This d i f f e r s from the scheme prescribed i n Figure 4.2.2; however the only e f f e c t i s that the d i g i t s MQ and M^ w i l l be a l t e r e d i n passing through the adder.. These are rejected i n gate k, however, and the remaining d i g i t s are i s o l a t e d from t h e i r e f f e c t by the forced i n j e c t i o n of the Revision-Carry d i g i t i n t o the operation on d i g i t M^ -The ..gates H,J.,.K., -and .L., combine the current samples of AQ and A^ with the sum-digits S ^ , to form the new, revised Memory-word. This output i s deli v e r e d e i t h e r through gate M to the input of the Memory, or through gate N to the buff e r r e g i s t e r . The l a t t e r occurs at the time of transfer of a Memory-word to the computer, under the control of f l i p -f l o p F which detects the Transfer Command. F l i p - f l o p E i s required to re-time the carry d i g i t s fed back to the carry input. 4.3 TIMING UNIT a. Requirements The requirements f o r clock signals and timing pulses f o r the cor-r e l a t o r are presented i n Figure 4.3.1. Three basic timing cycles are operative: (i) Revision-cycle The Arithmetic Unit services each s e r i a l s i x t e e n - b i t Memory-word i n a fun c t i o n a l routine l a s t i n g one microsecond. Controlled by a MASTER CLOCK CARD r~ EXTERNAL 32 MHz 32 MHz 8 MHz SYNCH. PULSES STOP/START ENABLE HINT 2il GATING 32 MHz TO SIGNALS TO QUANTIZERS QUANTIZERS 16 MHz 1 1 MHz PULSES TO I ARITHMETIC UNIT AND ALL ' CONTROL UNITS „_„ A-CODtvj / INES 5 / CHANNEL ADDRESS CODE GENERATOR 16 MHz TO 1 G L A S S MEMORY j 8 MHz SHIFT P U L S E s ' AND SAMPLING PULSES REVISION CYCLE COUNTER REFERENCE 16 MHz BANDWIDTH SELECTOR PULSE GEN., *16 256 COMPUTER I TRANSFER TIMER V CONTROL SIGNALS TO READOUT CONTROL TO READOUT CONTROL SHIFT PULSES TO SHIFT REGISTER AND SAMPLING PULSES TO QUANTIZERS ;ure 4.3.1 Timing Requirements for the Correlator. 77 16 MHz c l o c k , t h i s R e v i s i o n - c y c l e may be conveniently d i v i d e d i n t o 62.5 ns time s l o t s , defined by the s t a t e s of a s c a l e - o f - s i x t e e n counter.. This . counter may be used to generate a s e r i e s of 1 MHz pulses r e q u i r e d to c o n t r o l the a c t i v i t i e s w i t h i n each c y c l e . •t- t t t t t t t t f t t t 8 10 12 14 0 2 4 6 8 10 12 14 0 t9 t l l t13 t15 C l C3 t5 t7 t9 t l l t!3 fc15 1. 2. 3..5.4. 6. 7. 1. - SELECTION OF CHANNEL 2. PROMPT PULSES TO SELECTED CHANNEL ARE INHIBITED 3. CHANNEL ACCUMULATOR, STAGES 13 and 14, ARE READ 4. CHANNEL IS RESET 5v DIGITS FROM CHANNEL ARE AVAILABLE AT INPUT OF ARITHMETIC UNIT 6. REVISION CARRY C R IS GENERATED 7. NEXT -CHANNEL IS SELECTED. Figure 4.3.2 Sequence of Events i n the R e v i s i o n - C y c l e . Figure 4.3.2 shows the events a s s o c i a t e d w i t h the r e v i s i o n of a Memory-word, i n approximate r e l a t i o n to a Reference-cycle, whose epochs t g , t ^ - — t are the negative t r a n s i t i o n s of the s i g n a l which cloc k s the counter. The phases of a l l pulses and events are des c r i b e d i n terms of these epochs. For example the phase of an event o c c u r r i n g f i f t e e n nanoseconds a f t e r epoch t ^ i s l a b e l l e d t ^ + In a d d i t i o n , a l l pulses and s i g n a l s are given mrvemonic l a b e l s , p r e f i x e d w i t h the l e t t e r s CP (e.g. CP-MEM 1 i s a clock pulse which i s used i n the Glass Memory). 78 .75 yH .01 pF _ L J l .100 K« - 33 pF 47 Kfi Q, .001 pF Ih i 2 Kfi 32 KHz MASTER CLOCK SIGNAL 32.000 MHz Q 1 = MPF 107 Q 2 = MPS 834 Figure 4.3.3 O s c i l l a t o r C i r c u i t for 32 MHz Master Clock Signal. ( i i ) M ultiplexing-cycle The r e v i s i o n process of Figure 4.3.2 i s to be repeated within a 256-state Multiplexing-cycle of one microsecond steps, i n which the channels are addressed and serviced successively, under the control of the Channel Address Code Counter. ( i i i ) Computer Transfer-cycle At i n t e r v a l s determined by the capacity of the Memory, (e.g. 16.8 seconds) the Memory-words are to be transferred to the PDP-9 computer, i n synchronism with the f i r s t two cycles. A long counting chain i s to time the i n t e r v a l , at the end of which the Readout Control w i l l execute a Computer Transfer Routine, d i r e c t i n g the Memory-words from the Arithmetic Unit into the Buffer Register. In addition to the timing f a c i l i t i e s f o r the above functions, s e r v e r a l 32, 16 and 8 MHz signals are required, for the A/D conversion and clocking of the d i g i t a l data. Therefore the Master Clock i s to operate 79 at a 32 MHz rate, and the three cycles above are to be derived from count-downs from t h i s rate. These clocks and counters w i l l be described i n the following d i s -cussions, along with two p a r t i c u l a r problems: synchronization and temper-ature s t a b i l i t y . b. 32 MHz Clock The operation -of sampling a waveform by means of an impulse t r a i n at the Nyquist r e p e t i t i o n rate, i s equivalent to mixing i t s frequency components i n t o a m u l t i p l i c i t y of adjacent bandpasses, spaced i n units of the bandwidth along the r e a l frequency axis. The subsequent f i l t e r i n g operation passes only one bandpass segment. - In the reproduction of the o r i g i n a l spectrum, a high degree of s t a b i l i t y i n the mixing operation i s required, i n order that the frequency structure be accurately defined. To obtain a frequency s t a b i l i t y of 0.5% of the narrowest bandwidth of the spectrometer the sampling frequency must be stable to 0.002%. This also i s the s t a b i l i t y required of the 32 MHz Master Clock t The c i r c u i t designed f o r use as the i n t e r n a l master clock f o r the c o r r e l a t o r i s shown i n Figure 4.3.3. The o s c i l l a t o r consists of a j u n c t i o n f i e l d - e f f e c t t r a n s i s t o r i n a Pierce-type c i r c u i t u t i l i z i n g the gate-source and gate-drain capacitances f o r i n t e r n a l feedback. A source follower provides d r i v i n g c a p a b i l i t y into TTL l o g i c c i r c u i t s . The 32.000 MHz c r y s t a l has a thermal c o e f f i c i e n t of -26 Hz per C°, i n the range 20°C to 40°C, y i e l d i n g a t o t a l d r i f t of +520 Hz (+ .0017%). For t e s t i n g and i n i t i a l operational purposes the c r y s t a l was enclosed with the o s c i l l a t o r i n a metal box, i n which the expected temperature v a r i a t i o n t Due to D.S. Sloan, University of B r i t i s h Columbia 80 of l e s s than 10°C w i l l l i m i t the frequency d r i f t . However, the required s t a b i l i t y of +325 Hz w i l l be r e a d i l y achieved when the c r y s t a l i s replaced by a high-temperature c r y s t a l mounted i n a c o n t r o l l e d 75°C oven, as i s planned. c. Master Clock Card In the section 4.1 the requirement f o r a 32 MHz s i g n a l , along with three s p e c i a l gating s i g n a l s , for the quantizers was established. The Master Clock c i r c u i t was designed to produce these gating s i g n a l s by counting the 32 MHz d r i v i n g s i g n a l down i n a divide-by-four counter and combining the r e s u l t i n g 16 MHz and 8 MHz s i g n a l s . In Figure 4.3.4athe l o g i c c i r c u i t r y of the Master Clock card i s shown. In addition to the 16. MHz clock pulses f o r the Revision-cycle Counter, t h i s card provides sampling and s h i f t pulses f o r the quantizers and C o r r e l a t o r v S h i f t - r e g i s t e r . The frequency of the l a t t e r signals are to be a l t e r e d f o r d i f f e r e n t bandwidth modes by means of an extension to the Master Clock Card, c a l l e d the Band-width Selector. d. Bandwidth Selector The maximum sampling and s h i f t rate required f o r the normal op-era t i o n of the c o r r e l a t o r i s 8 MHz, a v a i l a b l e from f l i p - f l o p 7b i n the Master Clock c i r c u i t . By counting this rate down, the bandwidth i s a l t e r e d by factors of two. The counter in! the Master Clock i s extended i n the Band-width Selector Figure 4.3.4b to include four a d d i t i o n a l stages, operating synchronously to preserve the synchronized timing r e l a t i o n s h i p s described i n the next subsection. The sampling waveform i s selected from the outputs of the stages of the counter, and used also to generate a s h i f t pulse enabling s i g n a l . The 125 ns pulses of the l a t t e r enable the passage of every nth pulse of the 8 MHz s h i f t s i g n a l generated i n the Master Clock c i r c u i t , when the 3b EXTERNAL CLOCK 32 MHz SIGNALS 16 MHz CLOCK ENABLE P-SHIFT PULSE GATING SIGNAL, FROM BANDWIDTH SELECTOR rS>-» FLIP-FLOPS - NEGATIVE EDGE TRIGGER 14a T 0 Q U A N T I Z E R S -QJ-f "j 3 ^ 32 MHz TO BANDWIDTH SEL. 0-FFMC7b E>~rE) - ^ )—&—i^>-&—ED - * 13a -p. q-FFMC7a 'A B GATING SIGNALS TO QUANTIZERS 50S2 , 4 ns delay AUXILIARY 16 MHz 16 MHz TO REVISION-CYCLE COUNTER, TIMING UNIT CARD 1 SYNCHRONIZING PULSES, 8MHz SHIFT PULSES Figure 4.3.4a Schematic Diagram of the Master Clock Card. co Figure 4.3.4b Schematic Diagram of the Bandwidth Selector. CO 83 selected bandwidth i s 4/n MHz. This i n d i r e c t generation of the s h i f t pulses i s necessary for the establishment of the desired phase relationships between these and other timing signals, e. Synchronization Problem A s i m p l i f i e d diagram of those portions of the correlator system whose timing must be synchronized, i s presented i n Figure 4.3.5. With res-pect to timing control, the system may be divided into three major sections: 1. The Multiplier-Accumulator channels and the Correlator S h i f t -r e g i s t e r . 2_. The Prompt and Delay Quantizers, and the Master Clock c i r c u i t which controls the sampling of data and the delivery of samples to the m u l t i p l i e r s . 3^. The Timing Pulse Generator; the various multiplexing and d i s t r i b u t i o n networks which read, reset, and i n h i b i t the inputs of, the m u l t i -p l i e r channels and the Arithmetic Unit which processes the i n f o r -mation from the channels. The Multiplier-Accumulator Modules receive information under the control of the c i r c u i t r y i n section 2, and deliver i t to the Arithmetic Unit under the control of the Timing Pulse Generator i n section 3. Exact synchronism between the three sections must be ensured. There i s an ambiguity of 180° in the phase of the 16 and 8 MHz clock signals produced by the master clock, so that the transitions of the states i n the Revision-cycle counter can be aligned with-either the leading or the t r a i l i n g transitions of the 8 MHz sampling signals. Thus the i n h i b i t pulse (CP-PD1), which i s produced by the counter to i n h i b i t entire samples at the inputs of each channel (see section 4.4), may or may not be aligned with the pulses of a complete sample. SYNCHRONIZATION PULSE GEN. S>i 32 MHz GATE I 'START' SIGNAL 16 MHz •f a "[GATING SIGNAL 16 MHz GATE MASTER CLOCK COUNTER ANALOG 'PROMPT' SIGNAL ( OR SEMIPROMPT ) SECTION 2 8 MHz SHIFT PULSES ANALOG DELAY SIGNAL 5-LEVEL PULSE CODE A/D PROMPT QUANTIZER 8 MHz SAMPLING PULSES 3-LEVEL BINARY CODE A/D DELAY QUANTIZER *16 COUNTER T " REVISION CYCLE CONTROL PULSES RESET PROMPT PULSE INHIBITOR I I -READ MULTIPLEXER r-SECTION 3 RESET | DISTRIBUTOR T 1 SECTION I MAM CIRCUITS CORRELATOR SHIFT-REGISTER Figure 4.3.5 Outline of Units Whose Timing Must be Synchronized. co A) REFERENCE PHASE B) OUTPUTS OF. MASTER CLOCK cy COUNTER D) MASTER CLOCK SIGNAL E) SYNCHRONIZING PULSES F) FFRES 12a ENABLED G) CLOCK SIG. TO REVISION CYCLE COUNTER H) REFERENCE STATES OF REVISION CYCLE COUNTER J) 8 MHz SAMPLING PULSES K) L) TYPICAL PROMPT SAMPLES i W 8 MHz Q - F F M C 2 J T J T J T J T J T J T J T J T J T J 1 _ 16 MHZ Q-FFMCI Tru in jwuwinnru i^ 3 2 MHZ GATE MCSO 1 i n__n 1 8 n s n n r G A T E Mcisb 1 \ - 4 k — 38 ns | Q - F F R E S 1 2 a , PASSES SYNCHRONIZATION PULSES n.j^^Lnr^l'LiiJijmjL . ie MHZ Q-FFRES2af T > 1 1 r T — 1 — r T — • — 1 — 1 — r TO FFRES2a T STANDBY QUANTIZER OUTPUT " »l PHASE t , - 2 , etc. AT INHIBITOR GATES M) INHIBIT PULSE CP-PD1 N) PROMPT SAMPLES INTO MULTIPLIERS P) SAMPLE INTERVALS INTO MULTIPLIERS I I Q) TYPICAL DELAY SAMPLES ^ R) SHIFT PULSES, AT MAM CARDS S) DELAY SAMPLES INTO MULTIPLIERS t g + 36, etc. FALLING EDGE t 1 Q + 28 T) CP-RMP, READ PULSE INTO FIRST STAGE OF SHIFT REGISTER t + 24, etc. RIPPLE DELAY THROUGH ACCUMULATOR 278.5 ns «t P 70 ns U) CP-RD2, RESET PULSE t Q + 20 20 ns Figure 4.3.6 Synchronized Timing of Quantizers, Pulse D i s t r i b u t i o n , Multiplier-Accumulators, and Arithmetic Functions; and Requirements of the Master Clock. 86 The desired alignment i s obtained by ensuring that the Revision-cycle counter starts i t s 16-state cycle at a w e l l defined instant i n a sampling period when i t i s f i r s t enabled. The f i r s t pulse of the 16 MHz signal toggles the Revision-cycle counter into i t s tg state. The propagation of Prompt and Delay pulses from the Quantizers to the MAM cards and the timing of the i n h i b i t , read, and reset pulses are also shown. The r e l a t i o n -ships shown i n Figure 4.3.6 are v a l i d also for bandwidths less than 4 MHz. f. Timing Pulse Generators Figures 4.3.7a and 4.3.7b shows the timing requirements for a l l 1 MHz signals which are generated i n the Timing Unit for the control of other units. The tolerance of + 5ns stated for the phase and pulse-widths of these signals i s p a r t i c u l a r l y important for those signals which control the Arithmetic Unit. Several 16 MHz signals used i n the Glass Memory and Buffer Register are also shown. Although i t i s possible to produce a l l of the timing pulses from one basic 1 MHz s i g n a l , many large delays would be needed. I t proved more economical to use the sixteen^-state counter to make available sixteen basic pulses at 62.5 ns i n t e r v a l s . The timing pulses are then derived from these by means of short adjustable delays for phasing, and by further adjustable delays for the control of the pulse-widths. Four types of delay devices were considered: 1) distributed-parameter delay l i n e s , 2) lumped-parameter delay l i n e s , 3) monostable multivibrators manufactured i n Integrated C i r c u i t form, 4) monostable c i r c u i t s using TTL l o g i c gates. The fourth type was found to be the most p r a c t i c a l and inexpensive. Figure 4.3.8 shows the arrangement, employing seven l o g i c gates. The REFERENCE CYCLE REFERENCE CLOCK 16 MHz CP-TUl AND CP-CA1 CP-TU2 CP-TU3 CP-RMP LOCATION OF TEST POINT t 5 1 0, V 2 . > . i ° , \ 2 1 V 1 5 6 1 I 8 1 \l\ f 1 2, t° -UK- 62.5 ns 'REFERENCE WAVEFORM' TEST POINT t , + 58 X T "LT t 1 5 + 4 -»1 [ —*{ |*— 100 ns •^u -5l |f-30 ns t 1 2 + 58->Lf TIMING UNIT CARD 1, PIN 3 " PIN 5 K- 30 ns L T t, . + 58 . _ 1 4 -* 1^30 ns " PIN 7 INPUT OF MULTIPLEXER CARDS CP-PD1 CP-RD2 CP-RD3 CP-FFAR3 CP-AR1 t . + 20 t 1 Q + 28 7a r INHIBIT GATES IN PROMPT PULSE 125 ns INHIBITOR ( PHASE OUT OF TIMING UNIT CARD 2 - t . , + 20 ) OUTPUTS OF RESET rH \<~ 30 ns 15 DISTRIBUTOR 'P + 1 6 -j>n -*'| 30 ns —5*1 l 1^ - 30 ns t n + 28 ARITHMETIC UNIT, PIN 23 PIN 6,7 PIN 14 —*j K— SLIGHTLY GREATER THAN 62 ns N.B. TOLERANCE ON PULSE PHASES AND WIDTHS - +5 ns Figure 4.3.7a Timing Signals Generated i n the Timing Unit. ho c i 2 c i 4 co H H £6 'a ho ha * u co L 0 C A T I 0 N 0 F TEST POINT I i i i i i t I i i i i r i i i i i i i i I i i CP_AR2 0 z£ 1 ARITHMETIC UNIT, P I N 17 ->j \*~ 60 ns t . + 30 , , CP-AR3 1 -»< I • " P I N 18 -*} | * — 6 2 ns ._ ; . " P I N 19 CP-AR4 t . + 20 i ' ' ' 0 ^-145^1 C P - R C 3 1 5 "-J-Fl ; I~J READOUT CONTROL -*i H~ 30 ns CP-RC4 a I n ' -'1 K— 30 ns CP-DC2 0 t„ + 30 PROMPT PULSE SUPPRESS 30 ns CONTROL BUFFER REGISTER ADVANCED 15 ns U.R.T. REFERENCE PHASE CP-MEM1 GLASS MEMORY ADVANCED 32 ns W.R.T. REFERENCE • CP-MEM2 GLASS MEMORY ADVANCED 10 ns W.R.T. REFERENCE N.B. TOLERANCE ON PULSE PHASES AND WIDTHS - ±5 ns Figure 4.3.7b Timing Signals Generated i n the Timing Unit (continued). oo 89 Pulse In i f Pulse Delay Pulse Width Figure 4.3.8 Monostable C i r c u i t s For Generating Pulses of Variable Phase and Pulse Width. f i r s t two NAND gates at the l e f t of the diagram form a monostable f l i p -flop whose time constant can be changed by adjusting i t s RC network. This determines the phasing of the output pulse. The width of the pulse i s s i m i l a r l y determined by adjusting the RC network of a second f l i p - f l o p formed by the l a s t two NAND gates. The middle gates are used to produce a short inverted pulsie which causes the second f l i p - f l o p to be triggered by the t r a i l i n g edge of the f i r s t one. Details of the Revision-cycle counter and of the pulse generators are given i n Figures B.3.2, B.3.4 and B.3.5 i n Appendix B.3. g. Temperature Compensation The correlator w i l l be operating within an environment maintained at room-temperature, and the excess heat w i l l be drawn away by fans. Never-theless, i t i s expected that the operating temperature of the correlator could vary by as much a + 5°C. Propagation delay variations i n the cor-r e l a t o r over this range may cause phase d r i f t s of some control signals i n excess of the tolerances placed on them, necessitating compensation for the temperature dependence. Although NPO capacitors were used i n i t i a l l y i n the RC networks i n the dual-monostable c i r c u i t s , d r i f t s i n propagation 90 delays do occur i n the d i g i t a l integrated l o g i c c i r c u i t s . For example, a measurement was made on the control signal CP-AR4, which services the adder output of the Arithmetic Unit. In the temperature range 15°C to 50°C the phase of the pulse varied with respect to the reference phase by +0.8 nanoseconds/C°, and the width varied by +0.7 ns/C°. Relative to the t o t a l delays produced by the RC networks i n the dual-monostable c i r c u i t which shapes- t h i s pulse, these d r i f t s represent delay variations of +0.4%/C° and +0.6%/C° respectively. The values of these d r i f t s are t y p i c a l of those occurring i n the other timing pulses, and i t was also found that the percentage changes for a l l pulse delays and widths could be approximately represented by the figures given above. I t was found convenient to provide compensation for these d r i f t s .by replacing the NPO capacitors by ones with large negative temperature c o e f f i c i e n t s . Thus the decrease i n the RC time constants with increasing temperature could be made to balance the increased propagation delays i n the integrated c i r c u i t s . Using combinations of N5600 capacitors (temperature c o e f f i c i e n t s of -0.56%/C°) and NPO capacitors, the required compensation d r i f t s of -0.4% and -0.7% per C° were approximately obtained. Table A.2.1 shows the values of capacitance used for each timing pulse. S u f f i c i e n t compensation was provided i n th i s manner so that the absolute d r i f t rates for the pulses were t y p i c a l l y reduced to values less than +0.2 ns/C°, and the phase d r i f t of the pulse CP-AR4 was reduced to +0.1 ns/C°. Therefore t o t a l d r i f t s i n the expected + 5 C° range have been l i m i t e d to values much less than the assigned phase tolerances, h. Channel Address Code Generator Included i n the timing chain of,Figure 4.3.1 i s a 256-state 91 counter which serves the dual purpose of generating an e i g h t - b i t binary code for the addressing of the channels , and of dividing the clock rate down to 1/256 MHz (see Section 4.4). i . Computer Transfer Timer The 1/256 MHz signal CP-TC1 i s further divided by a sixteen stage ripple-timing counter, to produce a s i g n a l , CP-TC2, of long period, for timing the i n t e r v a l between transfers of the contents of the Glass Memory to the computer. The maximum accumulation i n t e r v a l i s to be 16.8 seconds, the overflow time of the memory for f u l l y correlated signals. Several shorter intervals are desired: 8*4 seconds for shorter accumulations; 1 second for monitoring the action of the readout routine i n the memory and for other test purposes; and 16 milliseconds for monitoring the op-eration of the Readout Control and Buffer Register. Ripple timing i n the counter i s adequate for the slow toggle rates involved, and v a r i a b i l i t y i n the phase of the readout command-pulse (CP-TC2) i s accounted for i n the design of the Readout Control. The details of the Computer Transfer Timer are shown i n Figure B.3.6. 92 4.4 MULTIPLEXING AND SERVICING OF CHANNELS The interaction between the Channel Accumulators and the Glass Memory i s necessarily on a channel-by-channel basis: each channel must be addressed and serviced i n turn. The services required are described under three categories: (i ) Channel-Readout Multiplexing: The outputs of the l a s t two stages of each Channel Accumulator must be read, and the values delivered to the Arithmetic Unit, ( i i ) Prompt-Pulse I n h i b i t i o n: The input of each channel must be in h i b i t e d some time before the outputs are to be read, i n order that errors do not occur i n the reading, ( i i i ) Reset D i s t r i b u t i o n : The accumulator chain i n each channel i s to be (optionally) reset immediately after i t s outputs are read, at the time of reading coincident with the transfer of data to the computer. This action i s required to establ i s h the indepencence of separate integrations. The performance of these services requires that the 256 channels must be addressed i n sequence. The implementation of the three multiplexing and d i s t r i b u t i o n services permits the use of an ei g h t - b i t binary code for addressing the channels however, due to the nature of the transfer routine between the Glass Memory and the PDP-9 computer, the address se-quence must be of a non-natural order. The design of the Channel Address Code Generator i s discussed below, following which each of the service networks w i l l be described. a. Channel Address Code Generator At the time of transfer of the contents of the glass memory to the PDP-9 computer, i t i s not possible for the computer to receive all4 , 0 9 6 d i g i t s sequentially at a 16 MHz rate. I t i s required that 93 sixteen d i g i t s (one word) at a time be transferred i n p a r a l l e l , and that the computer be allowed s u f f i c i e n t time between such transfers i n order to perform some immediate manipulations, and to store the numbers. The transfer of a Memory-word v i a the Buffer Register requires one microsecond of time and an i n t e r v a l of forty-one microseconds i s desired between transfers. The readout routine then involves loading every f o r t y - f i r s t consecutive channel passing through the Arithmetic Unit from the c i r c u l a t i n g memory, into the Buffer Register, for transfer to the computer, s t a r t i n g from the f i r s t channel and proceeding through several Multiplexing-cycles u n t i l a l l 256 channel-words have been transferred. In order that i n this process the computer may actually receive the channel-words i n a natural order (e.g. Prompt channels 0, 1, 2, 3 127, followed by Semi-prompt 0, 1, 2, 3 127), the channels must be arranged i n the memory i n a non-natural order such that every f o r t y - f i r s t channel addressed i n the Multiplexing-cycle f a l l s into the natural sequence. This arrangement i s i l l u s t r a t e d i n Figure 4.4.1. I t i s possible to allocate a l l 256 channels i n this c y c l i c sequence i n the memory because the numbers 41 and 256 do not have a com-mon factor. Beginning with channel P Q and counting every f o r t y - f i r s t channel continuously around the cycle, one counts a l l 256 consecutive channels a r r i v i n g back at P Q after completing exactly forty-one cycles. The exact sequence of adjacent channels i n the memory i s given i n Table B.4.1 i n the Appendix B.4. The assignment of the 256-state binary code to the channels i s shown i n Table B.4.2. Consecutively num-bered channels along the correlator s h i f t r e g i s t e r are located adjacently i n the channel-racks. The multiplexers and d i s t r i b u t i n g decoders that 94 service those channels (with p a r a l l e l l i n e s going to a l l channels) require that the channels be assigned consecutive binary code numbers. This as-signment i s summarized as follows: ' Channels Codes - Assigned  Prompt 0 to 127 0 0 0 0 0 0 0 0 . t o 0 1 1 1 1 1 1 Semi-prompt 0 to 127 1 0 0 0 0 0 0 0 to 1 1 1 1 1 1 1 I I M.S.B. L.S.B. Thus the counter which generates these codes must count i n the non-natural sequence defined by the order of channel-word storage i n the memory, The design of this counter involved the use of a 256-state difference-equation truth table (defining the consecutive states of the eight f l i p - f l o p s i n the counter); the derivation of eight application equations for each f l i p - f l o p by means of eight-variable Karnaugh maps [16]; and the solving of these with the c h a r a c t e r i s t i c equations of the JK f l i p - f l o p s , to determine the input equations of each stage. The process was s i m p l i f i e d by the extensive presence of symmetries in the Karnaugh maps. The Q-outputs of the eight f l i p - f l o p s constitute the code l i n e s and they are l a b e l l e d as A 3 A 2 A 1 A 0 ' a n d B 3 B 2 B l V I I Most s i g n i f i c a n t - Least s i g n i f i c a n t stage stage The input equations for each of the f l i p - f l o p s are given i n Table B.4.3 in Appendix B.4. The resulting implementation i s shown i n the schematic diagram of Figure 4.4.2. The eight-stage counter i s synchronously clocked by the 1 MHz signal CP-CA1, and the Q outputs of the f l i p - f l o p are taken out as inverted code li n e s (they are re-inverted i n buffer gates and delivered 9 5 to a l l of the Multiplexers and Decoders). In addition, three special 1 / 2 5 6 MHz outputs are required by the Readout Control card, and these are produced by the states of the counter corresponding to the addressing of channels P Q , P^g and S P ^ g (see Section 4 . 5 ) . Figure 4.4.1 Arrangement of Channels i n Memory Locations, I l l u s t r a t i n g the Non-natural Sequence i n which the Channels are Addressed. b. Channel-Readout Multiplexing Two i d e n t i c a l multiplexing networks are needed for the reading of the two output d i g i t s of the Channel Accumulators. In each network i t i s necessary that 2 5 6 l i n e s from the channels converge through a MEMORY-WORD FOR CHANNEL P,, 96 97 ONE INPUT _ SELECTED * FROM SIXTEEN, CHANNELS IN RACK (0 - 1 5 ) B ^ CHANNEL ADDRESS CODE POWER GATES IN RACK BUFFER CARD S T _ C P - R M P LLll (16-31) (32-47) (48-63) (64-79) ( 8 0 - 9 5 ) ( 9 6 - 1 1 1 ) P. RACK .1. (0-15) CP-RMP . CHANNEL / 4^ 4-i RACKS ( *i fit FROM CP-RMP BUFFERS SN-74150 ' S I X T E E N - B I T S P \f~ M U L T I P L E X E R 1 , RACK 2 L U l p p-3 (16-31) JUUL (32-47) LLU, (48-63) p o — ILL (64-79) * J E L / 4, -CP-RMP BUFFERS T CP-RMP' CHANNEL ADDRESS CODE r < A i < A„ rem um p |o-7 (80-95) (112-127) i_LL (96-111) (112-127 DIGIT OUT TO ARITHMETIC UNIT 'SN-74150 MASTER READ-MULTIPLEXER ONE INPUT SELECTED READ-MULTIPLEXERS IN CHANNEL RACKS Figure 4.4.3 Block Diagram of the Channel Read-multiplexing Network. 98 selection 'tree' on one l i n e into the ARithmetic Unit. A two-rank multiplexing tree was conveniently implemented, using the Texas Instru-ments, MSI sixteen-bit multiplexer SN74150. The configuration of the network i s shown i n Figure 4.4.3. The f i r s t rank of the tree consists of sixteen multiplexers, located at the Channel Racks, each receiving l i n e s from sixteen channels. The second rank i s a master multiplexer which receives the outputs from the f i r s t sixteen multiplexers, and i s situated i n the Central Processing Rack. The four least s i g n i f i c a n t d i g i t s of the ei g h t - b i t channel ad-dress code ( B ^ , B^. B ^ , BQ) serve a l l of the f i r s t - r a n k multiplexers, so that one channel-line into each of these i s enabled. The four most s i g n i -f i c ant code-digits (k^, k^, A^, AQ) select one of the l i n e s from the rack-multiplexers into the master multiplexer, so that only one of the 256 input l i n e s has through-path to the Arithmetic Unit. When this path has been es-tablished by the code l i n e s , a read-strobe s i g n a l , CP-RMP, i s delivered to a l l the multiplexers, and the l o g i c a l value on the selected i n p u t - l i n e i s transmitted to the output l i n e , for the duration of the strobe pulse. The two p a r a l l e l networks for the d i g i t s A^ and A^ are to be strobed simultaneously, i n order to minimize the duration of the i n h i b i t i o n of the prompt-signal inputs to the Channel-Multipliers, c. Prompt Pulse I n h i b i t i o n While the output stages of a Channel Accumulator are being read (the read strobe i s 70ns i n duration), i t i s important that t h e i r states should not change. This requires that the a c t i v i t y of the mul t i -p l i e r be frozen a certain time before the reading, accounting for the r i p p l propagation-time of the accumulator chain. The easiest method i s to i n -h i b i t the prompt (or semi-prompt) signal at the input to the channel. 99 CHANNEL ADDRESS CODE W l B 0 F 9311 ONE-OF-SIXTEEN DECODER INHIBIT GATES TO CHANNELS CHANNEL RACK .1 CHANNEL ADDRESS CODE MASTER DISTRIBUTOR •^T,— - F93U .CP-.EDl PROMPT PULSE BUFFERS, IN CENTRAL PROCESSING UNIT TO RACK BUFFERS . SP » PROMPT PULSES . SEMIPROMPT PULSES FROM QUANTIZERS J (0-15) SEMIPROMPT L-LUC^ PULSE BUFFERS (IN RACK BUFFER CARD) s RACK 2 (16-31) (32-47) .5 -P 5 1 1 1 P A 1 1 1 P 7 7 ""' E H ,(64-79) 1 I I I SP 5 K 5 (48-63) (64-79) (80-95) i_l_L SP 6 (96-111) LLL GH T i i n . SP 7 (112-127) M I L SP 8 (80-95) (96-111) (112-127) f T f DISTRIBUTORS IN CHANNEL RACKS Figure 4.4.4 Block Diagram of the.Prompt (or Semipromt) Pulse D i s t r i b u t i o n Network Showing Use of the Inhibito r Cards. 100 INHIBIT GATES TO CHANNELS f3tT)-^WV-fr ( V CHANNEL ADDRESS CODE F9311 ONE-OF-SIXTEEN DECODER 65 p F - ^ PROMPT PULSES INHIBIT STROBEP FROM MASTER DISTRIBUTOR 11 * P or SP, JJTZ.- *  o-CHANNEL * FOR SPECIAL CHANNELS NOTED, INHIBIT GATE IS BYPASSED : INHIBIT LINES ARE TAKEN DIRECTLY TO RACK BUFFER CARDS, FOR USE ON SERIAL DELAY SIGNAL AND SAMPLING PULSES Figure 4.4.5 Schematic Diagram of a Prompt Pulse Inhibitor Card Located i n the Channel Racks. 101 Since this i n h i b i t i o n of prompt-pulses i s to occur channel-by-channel, i n the same order as the read multiplexing, separate prompt signal l i n e s are required to each of the 256 channels, each passing through an 'inhibit-gate'. An i n h i b i t - p u l s e must be directed v i a a decoding-tree to the one inhibit-gate corresponding to the selected channel. A two-rank decoding tree, using the F a i r c h i l d Semiconductor integrated c i r c u i t sixteen-line Decoder., was used for this purpose. The arrangement i s shown i n Figure 4.4.4. In the f i r s t rank the master i n h i b i t - p u l s e d i s t r i b u t o r , or decoder, has sixteen outputs going to prompt-signal inhibit-cards i n the channel racks. These constitute the second rank; each has sixteen prompt-signal output l i n e s to the channels. The four least s i g n i f i c a n t d i g i t s of the binary address code select one output l i n e i n each of the sixteen rack-decoders, while the four most s i g n i f i c a n t d i g i t s select one of these decoders to be served by the master decoder. The diverging decoding-tree thus opens one of 256 d i s t i n c t paths, along which the i n h i b i t pulse CP-PD1 i s delivered to the i n h i b i t gate for the selected channel. A detailed schematic diagram of a t y p i c a l Prompt-Pulse I n h i b i t o r card i s shown i n Figure 4.4.5, i n which the use of the inhibit-gates i s shown. Into each card i s delivered a prompt- or semi-prompt signal l i n e , which supplies the pulses, v i a buffer gates 2a and 2b, to a l l sixteen i n h i b i t gates. A l l gates, except that one selected by the four d i g i t s of the channel address code are enabled, allowing the pulses to pass to the channels. The prompt/semi-prompt signals originate i n the Central Pro-cessing Unit, i n the Quantizer c i r c u i t s , and are delivered to each of sixteen i n h i b i t o r cards. The task of maintaining or regenerating the 102 quality of these prompt signals (maximum pulse rate i s 32 MHz) proved to be a d i f f i c u l t problem i n the correlator design. I t s solution i s described i n section 4.1e. d. Reset D i s t r i b u t i o n In the operation of the supersynthesis telescope, each i n -dependent integration, corresponding to the sampling of one s p a t i a l Fourier Component of the brightness d i s t r i b u t i o n i n the f i e l d of the primary beam, w i l l consist of several sixteen-second accumulations i n the Glass Memory. In any one complete survey the number of accumulations used per point w i l l be constant, (for example, eight i s a t y p i c a l number). At the end of the integration period i t i s required to reset the channels to establish independence of adjacent samples. In previous discussions we noted that this resetting c o e f f i c i e n t s , but which are n e g l i g i b l e com-pared to the t o t a l r.m.s. noise. I t should be noted, however, that i f a constant number of ac-cumulations per point i s used, then at small antenna spacings, the s p a t i a l Fourier-plane (u-v plane) w i l l be considerably over-sampled, and adjacent integrations w i l l not be independent (regardless of resetting) but rather w i l l be somewhat correlated. Thus i n the Fourier inversion pro- ... cess some of these adjacent points w i l l be combined. I t appears then that i n some circumstances the resetting of the channels w i l l be harmful, whereas i n others(wide spacings) i t may indeed be desired. I t i s then required to provide i t as an option. When enabled, the reset routine w i l l operate on the channels during every 'nth' transfer to the computer, (where n i s t y p i c a l l y 8). The transfer routine occupies forty-one multiplexing cycles — every f o r t y - f i r s t channel i n the memory i s transferred, one at a time, and 103 F93U ONE-OF-SIXTEEN DECODER I B 3 > CHANNEL ADDRESS CODES A3 t — V A, MASTER DISTRIBUTOR IN *0 CPU — ' CHANNEL RESET 5 •V-I J TO CHANNELS IN RACK (STROBE)^ j 1 SP fc=?==5z*5 RESET ENABLE-PAUSE + MEMORY RESET-PAUSE T O LL (16-31 11 (32-47; T O 11 a (48-63) 1LU T O a (64-79) (80-95 iii. (96-111) (96-JJ. (112-127) | LOCATED IN RACK 1 I 1(0-15) 111 LLll a LL (16-31) (32-47) (48-63) (64-79) 111 SP N 6 (80-95) M l SP — s 7 (96-111) (112-127) DISTRIBUTORS IN CHANNEL RACKS Figure 4.4.6 Block Diagram of the Reset Di s t r i b u t o r Network. 104 the resetting of a channel i s to occur only at i t s time of transfer. The reset d i s t r i b u t i o n tree can be constructed i n a s i m i l a r fashion as the Prompt-Pulse Inhib i t o r Network, with some additional controls on the issuance of the reset pulse. The network i s shown i n Figure 4.4.6, and consists of two ranks of sixteen-line decoders operating on the two parts of the eig h t - b i t binary address-code. The strobe input to the master reset decoder consists-of two control l i n e s . The f i r s t one, l a b e l l e d CHANNEL-RESET i s a pulse issued from the Memory-Reset Control whenever a channel i s being transferred from the memory to the computer v i a the Buffer Register. I t i s controlled then by the Read-out Control, and by the Revision-cycle pulse generator. The second strobe-control l i n e i s a com-bination of control variables, created i n the Reset Control c i r c u i t . The variable RESET ENABLE takes the l o g i q a l '1' value on every nth occurrence of the transfer routine i f , and only i f , the RESET OPTION switch i s on. When the correlator i s i n the FREE-RUN or PAUSE mode, then, this second control l i n e enables the reset d i s t r i b u t o r on every nth readout to the computer. When the correlator i s i n the PAUSE mode, the Reset D i s t r i b u t o r i s normally disabled except when the Memory Reset button i s depressed. In thi s case both control l i n e s go to continuous low states, so that a l l channels are reset as they are addressed. A more thorough discussion of t h i s reset f a c i l i t i e s i s given i n the section dealing with the Reset Control. e. Timing of the Three Channel-Services Waveforms M,T, and U, i n Figure 4.3.6 i l l u s t r a t e the r e l a t i v e phases of the three events involved with the reading of a channel: prompt-pulse, i n h i b i t i o n , reading of output stages, and resetting of the channel. 105 The approximate propagation delay between the input of a pulse to the f i r s t stage of an MAM channel and the response of the l a s t stage, i s 256 ns. I f an entire sample (of possibly four pulses) i s i n h i b i t e d , the safe reading-interval during which the output f l i p - f l o p s may be read, without any danger of th e i r changing state, i s reduced from 125ns to 70ns. This i s due (in addition to v a r i a b i l i t y i n the response-times of the f l i p - f l o p s ) to the facts that the input pulses may enter either of the f i r s t two stages, and that two output stages are to be read simultaneously. 4.5 COMPUTER TRANSFER CONTROL , a. Readout Control — Computer Transfer Routine. In Section 4.4 a discussion was: presented of the need for a l l o c a t i n g channel-words i n the Glass Memory, so that every f o r t y - f i r s t channel, taken continuously around the cycle, f a l l s into the natural sequence [P^, P^ ... P^y' S p 0 S p 1 2 7 ] ' The Computer Transfer routine, i n i t i a t e d at the end of a desired accumulation i n t e r v a l , involves the loading of every f o r t y - f i r s t channel i n the memory, st a r t i n g with channel P Q , into the Buffer Register, from whence the d i g i t s may be delivered on p a r a l l e l l i n e s to the computer, v i a a suitable I/O interface (see Section 4.6). This present section deals with the device which was designed to regulate the transfer process, the Readout Control. b. Design The following requirements apply to the design of th i s u n i t : 1. The device must be capable of responding to a command-signal from the Computer Transfer Timer (CP-RC2) to begin the transfer-routine. 2. I t must generate a pulse of one microsecond i n duration, delivered to the Arithmetic Unit every forty-one microseconds. At the Arithmetic Unit 106 this s i g n a l , CP-R01, i s sampled by f l i p - f l o p 9b once every microsecond, and when i t s l o g i c a l value equals 1, the output gates of that unit d i r e c t the current Memory-word into the Buffer Register instead of back into the memory. 3. After 256 issuances of the pulse CP-RC1, the device must turn i t s e l f off and await the next command to re-start the routine. In respect of these requirements, then, i t appears that the device must consist of four d i s t i n c t sections: ( i ) A turn-on control, responsive to the command signal CP-TC2 from the Computer Transfer Timer. ( i i ) A turn-off control, responsive to an i n t e r n a l i n d i c a t i o n that a l l 256 channels have been transfered!. ( i i i ) A scale-of-41 c'ounter, operating at a clock rate of 1 MHz, controlled by sections ( i ) and ( i i ) , and issuing pulses CP-RC1/ (iv) A scale-of-256 counter driven by the output of the scale-of-41 counter. The output of the larger counter i s the indicator to the turn-off section ( i i ) , occurring 10,496 microseconds after the event of turning-on. However, i n place of the 256-counter proposed above, the e x i s t i n g scale-of-256 counter i n the Channel Address Code Generator, can be used i n a different manner. Given two counters, one a scale-of-256 and the.other a scale-of-41, operating on the same clock signal of frequency 1 MHz, any two states, one from each counter, w i l l coincide only once i n t h e i r lowest common period of 10,496 microseconds. This event then can be used to both i n i t i a t e and terminate the transfer-routine. The Readout Control c i r c u i t was designed on this basis, modified to account for problems associated with propagation delays. The three main sections i n the unit can be discerned: the scale-of-41 counter, the i n i t i a t i n g c i r c u i t and the terminating c i r c u i t , NORMALLY IN 'ZERO' STANDBY STATE CP-RC3 1MHz CLOCKING SIGNAL INITIATING SIGNAL CP-TC2 rn. P " J l 'READY' PULSE J INITIATING CIRCUIT PULSE DIFFERENTIATOR CP-RC5 SELECTION OF CHANNEL P„ 1 -ft CP-RC8 TO RESET CONTROL (ZERO . STATE) *41 COUNTER RESET CP-RCI TO ARITHMETIC UNIT AND BUFFER REGISTER (FIRST STATE) CP-RC2 (SECOND STATE) SET ' RESET TERMINATING CIRCUIT CONTROL FLIP-FLOP CP-RC4 1MHz STROBE PULSE CP-RC6 SELECTION OF CHANNEL P 24 Figure 4.5.1 Block Diagram of the Readout Control. 108 and these are depicted i n block form i n Figure 4.5.1. • The i n i t i a t i o n command s i g n a l , CP-TC2,- i s the output of the selected slow stage of the Computer Transfer Timer. The t r a i l i n g edge of this signal i s used to generate a short i n i t i a t i o n pulse. The phase of this pulse i s uncertain due to the ripple-timing of the Computer Transfer Timer, so that the pulse i s used merely as a 'ready' signal which alerts the i n i t i a t i n g c i r c u i t . The control f l i p - f l o p i s set by the strobe pulse CP-RC4 at the f i r s t subsequent occurrence of the one microsecond pulse CP-RC5 (selection of f i r s t channel, P Q ) . The output of the control f l i p - f l o p enables the gate to pass the 1 MHz clock-signal. CP-RC3 to the counter. The counter i s normally reset i n i t s standby or 'zero' state. When activated, i t generates the one micro second pulse CP-RC1, i n i t s f i r s t state (and a s i m i l a r pulse CP-RC2 i n i t s second s t a t e ) , at a re p e t i t i o n rate of one pulse per forty-one micro-seconds. CP-RC1 i s the transfer command signal used by the Arithmetic Unit to d i r e c t the currently revised Memory-word i n to the Buffer Register rathe than back into the memory. I t must be a f u l l microsecond i n duration and the operation of the counter must be synchronous with that of the A r i t h -metic Unit. I t i s the coincidence of the pulse CP-RC2 with the state of the channel-address counter corresponding to the selection of channel P^g (CP-RC6), which causes the termination of the readout routine, by allowing the strobe pulse, CP-RC4, to reset the control f l i p - f l o p , the counter, and the i n i t i a t i n g c i r c u i t . This coincidence occurs one microsecond after the transfer of the l a s t channel, SP-127, has taken place, during the l a s t of the forty-one multiplexing cyles required i n the complete transf e r a l of a l l of the channels. 109 A more detailed discussion and a complete schematic diagram for the readout control are given i n Appendix B.5 4.6 BUFFER REGISTER — INTERFACING TO THE PDP-9 COMPUTER The accumulation of the 256 cross-correlation c o e f f i c i e n t s i n the Glass-Memory, involving the continuous re-cycling and r e v i s i o n of the data through the Arithmetic Unit, has been the subject of previous d i s -cussion i n this chapter. The present section new attends to the commun-i c a t i o n of this data to the PDP-9 computer. At the end of an appropriate i n t e r v a l of accumulation ( l i m i t e d by the over-flow time of the Glass Memory), the contents of the memory, consisting of 256 sixteen-digit numbers, are to be transferred to the computer at 41 microsecond i n t e r v a l s . The communication of the data from the correlator to the computer, i s to be effected by a Correlator Interface, which i s connected to the Computer's Input/Output Bus, and i s serviced by the Automatic P r i o r i t y Interrupt f a c i l i t y . The sixteen d i g i t s of a single Memory-word, pr cross-correlation c o e f f i c i e n t , are delivered to the Interface on p a r a l l e l l i n e s . On receipt of a command from the correlator ('DONE FLAG') the Interface requests a h i g h - p r i o r i t y interruption of the main program being executed, for the transfer of the data. Upon the granting of the requested service, the computer transfers the data (sixteen-digits) v i a the I/O Bus l i n e s , into i t s accumulator register. Successive numbers are to be transferred i n a s i m i l a r manner. The f i r s t of the 256 command flags achieves the diversion of the computer's attention from the main program. However, the operation of the main program i s actually not restored u n t i l a l l 256 numbers have been transferred. The complete transfer requires a duration of 10.496 milliseconds, a small 110 portion of computer time compared to, the 16.8 seconds accumulation i n t e r v a l i n the correlator. The Correlator Interface, l i n k i n g the correlator to the computer, was designed by J.H. Dawson at the Dominion Radio Astrophysical Observatory, Penticton, B.C., using standard DEC Chip-Modules; The design i s i l l u s t r a t e d i n the schematic diagram of Figure B.6.2. In addition of receiving the data from the correlator, the interface also issues the control commands AUTO-STOP/START and AUTO-SUPPRESS. The f i r s t l i n k i n the i n t e r f a c i n g system, which provides the para-l l e l output of the sixteen-digit Memory-words, i s the Buffer Register. I t i s described i n the following discussion. I t has been explained i n the previous section (Readout Control) that at the time at which a Memory-word i s to be transferred to the computer a control signal CP-RC1 i s issued for the duration of the passage of the' word through the Arithmetic Unit. The control-signal directs the output d i g i t s , of the Arithmetic Unit, to the Buffer Register and feeds zeros into the memory i n t h e i r place. The Buffer Register must, at the same time, be activated,.in order to load the d i g i t s sequentially into the r e g i s t e r stages. At the completion of this operation, a DONE-FLAG i s to be issued to the Cor-r e l a t o r Interface. Figure 4.6.1, i s the schematic diagram of t h i s u n i t . The control signal CP-RC1 i s used to gate sixteen pulses of the clock-signal CP-BR1, properly phased to s h i f t the d i g i t s at the input from the Arithmetic Unit into the stages of the r e g i s t e r . The t r a i l i n g edge of CP-RC1 generates a narrow pulse, at the output of gate 9c, which i s then widened to become the DONE-FLAG. SERIAL DIGITS, SIXTEEN-BIT MEMORY WORD FROM ARITHMETIC UNIT 4 4> CP-BRl (16 MHz) 5 6 7 12 t 13 T 14 T 15 A I A I 4k I 4> 1 2 b f " \ l l a ^ " \ llbf^ 3 DIGITS TO PDP-9 COMPUTER 16b — 3b j ~ZJ 2,T 2b \ LZ 7b L-J 6a j I— 6b ' T V T TYPE-D FLIP-FLOPS, POSITIVE-EDGE TRIGGERED GENERAL RESET 5V 4 7 0 p F ^ I K f ? — ' 4 2 0 0 n ..^PRESENT WHEN TRANSFER ROUTINE IS ACTIVE CP-RC1 (41P9 REPETITION PERIOD lys DURATION) ; ; Figure 4.6.1 Schematic Diagram of the Buffer Register. 112 The f l a g , and the s i x t e e n o u t p u t s o f the s h i f t r e g i s t e r a r e b u f f e r e d and d r i v e n i n t o the c a b l e s g o i n g to the compute r . 4.7 OPERATION CONTROL: RESET CONTROL In t he p r e v i o u s s e c t i o n s o f t h i s c h a p t e r , t he d i s c u s s i o n has d e a l t w i t h t h o s e p a r t s o f t he c o r r e l a t o r w h i c h aire d i r e c t l y i n v o l v e d i n the h a n d l i n g o f the a c c u m u l a t i o n s i n the c h a n n e l s — f rom the s a m p l i n g o f the c h a n n e l o u t p u t s , th rough s t o r a g e i n t h e t r a n s f e r o f i n f o r m a t i o n to the computer . In t he o p e r a t i o n o f the c o r r e l a t o r s e v e r a l s u b s i d i a r y c o n t r o l f u n c t i o n s a r e n e c e s s a r y o r d e s i r e d . These a r e c a t e g o r i z e d as f o l l o w s : 1. O p e r a t i o n Modes: STANDBY/ACTIVE and PAUSE/FREE-RUN 2. R e s e t O p e r a t i o n s : dependent and i n d e p e n d e n t w i t h r e s p e c t to the h a n d l i n g o f the d a t a . 3. S u p p r e s s i o n o f prompt p u l s e s d u r i n g the s w i t c h i n g o f d e l a y -c a b l e s , o r w h i l e the PAUSE mode i s i n e f f e c t . The s e c o n d and t h i r d c a t e g o r i e s w i l l be d i s c u s s e d s e p a r a t e l y i n t he n e x t two s e c t i o n s . However, as an i n t r o d u c t i o n to t h e s e f u n c t i o n s , and as a f a m i l i a r i z a t i o n w i t h the p r a c t i c a l o p e r a t i o n o f t he c o r r e l a t o r , t he f i r s t c a t e g o r y w i l l be d i s c u s s e d i n t h i s s e c t i o n , a . O p e r a t i o n Modes W i t h i n the c o r r e l a t o r s y s t e m , t h e r e a re s e v e r a l o p e r a t i o n s whose t i m i n g must be unambiguous ly s y n c h r o n i z e d . T h i s n e c e s s i t y was d i s c u s s e d i n s e c t i o n 4.3e and t h e r e a r o s e a n e e d f o r a ' s t a n d b y ' s t a t e f rom w h i c h a l l o p e r a t i o n s c o u l d s t a r t i n p r o p e r o r d e r , on command. In t h i s mode, a l l u n i t s o f the c o r r e l a t o r , e x c e p t t he M a s t e r T i m i n g c a r d (which g e n e r a t e s t he t i m i n g p u l s e s f o r the q u a n t i z e r s , t he c o r r e l a t o r S h i f t R e g i s t e r , and the i m p o r t a n t s y n c h r o n i z i n g p u l s e s ) , a re i n o p e r a t i v e , and i n a ' r e s e t ' c o n -d i t i o n . 113 A control c i r c u i t i s required to maintain this condition, which can be switched between the states 'STANDBY' and 'ACTIVE', the l a t t e r of which releases a l l units from the reset condition and sta r t s the Revision-cycle Counter i n the proper phase indicated i n Figure 4 . 3 . 6 . I t i s desired that this switching be performed manually (for testing and manual operation of the correlator) as wel l as automatically (under the control of the com-puter-which w i l l also maintain the operation of the entire sypersynthesis apparatus). After the correlator has been started, and a l l clocks are functioning properly, two additional, and optional, modes are desired for the purposes of testing the correlator. These have been termed 'FREE-RUN' and 'PAUSE'. The FREE-RUN mode i s that i n which a l l correlator functions are operating normally, and w i l l be used exclusively during a l l supersynthesis observation programs. Alternative to this i s the 'PAUSE-MODE' , i n which the operation of a l l m u l t i p l i e r s and accumulators ceases, but the numbers exi s t i n g i n the memory and i n the Channel Accumulator chains are preserved. Thus a l l 'reset' operations are disabled, a l l three quantizers are i n h i b i t e d and D.C. subtraction i s suspended. This mode i s to be established on the f l i p of a switch, without destroying any ex i s t i n g information i n the channels or memory. After allowing the correlator to accumulate cross-products for a time, one can then switch to t his mode and readily examine the contents^ of the memory, thus t e s t i n g , the operation of the system as a correlator. In the discussion of the Reset Control to follow, the implementation of the STANDBY/ACTIVE option i s discussed, while the de t a i l s of the PAUSE/ FREE-RUN modes are dealt with i n the succeeding section on Prompt Pulse Suppression. Included i n the l a t t e r unit i s the D.C. Subtraction Control. 114 b. Reset Control Those control operations which have a bearing on the resetting of various parts of the system are discussed i n this section. Most of the c i r c u i t r y involved i s located on one card termed Reset Control, while the remainder, the Memory Reset Control, shares space on another card with the 'CP-RMP Buffers'. Three types of resetting functions are to be discussed: 1. General Reset: This i s the condition of the correlator when i t i s switched to STANDBY mode. 2. Automatic Reset: Channel Accumulators and Memory-words are to be optionally reset as they are transferred to the computer. 3. Memory Reset: Independent resetting of the Memory and Channel Accumu-l a t o r s , i n response , to a manual command, c. General Reset The requirements of the Reset Control are outlined i n block form i n Figure 4.7.1. Of f i r s t i nterest are the two f l i p - f l o p s A and B, which constitute a two-part control over the mode options ACTIVE and- STAND-BY. In the STANDBY mode, with f l i p - f l o p B turned o f f , the 16 MHz clock from the Master timing Unit to the Revision-cycle Counter i s disabled by a'low l o g i c l e v e l on the l i n e , CLOCK-ENABLE. In addition, a l l those control units, counters and data-handling c i r c u i t s to which the GENERAL RESET l i n e goes, are held i n a reset state by a low voltage. Since i t i s desirable that the Quantizers, D.C. Subtraction Control, and channel-resetting op-erations be disabled while the correlator i s i n the STANDBY condition, the PAUSE mode i s included i n the General Reset condition and provides these' disablings i n response to the AUTOMATIC PAUSE signal delivered to 115 SYNCHRONIZING PULSES MANUAL START ! ! L - ^ MANUAL STOP l=£Dn 1 FF A J FF B -£> GENERAL RESET -fc. 16 MHz CLOCK ENABLE ^ AUTO-PAUSE READY PULSE CP-RC7 +-L _ RESET s T 8 CP-1 COUNTER ON ' ^ s RESET ENABLE OPTION p A U S E ^ MEMORY RESET CONTROL TO RESET DISTRIBUTOR CP-RC8, -TRANSFER > CP-RD2 P-GENERAL RESET O AUXILIARY 8 MHz > CP-CA1 b-I CP-CA1 + MEMORY RESET • (8MHz) MEMORY RESET CONTROL ON SEPARATE CARD WITH CP-RMP BUFFERS Figure 4.7.1 Block Diagram of the Reset Control. 116 the Prompt Suppression Control. When f l i p - f l o p B i s turned on, the correlator i s released from the STANDBY condition, and normal operation i s restored. However, as mentioned i n Section 4.3e this ' s t a r t i n g ' event must establish an unambiguous relationship of phase between the Revision-cycle timing and that of the Quantizers; therefore, f l i p - f l o p B i s to be turned on by a 'Synchronizing Pulse'. F l i p - f l o p A i s the l i n k between an asynchronous 'turn-on' command and the synchronous operation of f l i p - f l o p B. I t can be turned on and o f f by the manual switches on i t s set and reset terminals, or automatically by successive pulses from the computer to i t s clock terminal. When this f l i p - f l o p turns on, i t s output enables gate C and the f i r s t synchronizing pulse, passed by the l a t t e r , sets f l i p - f l o p B, releasing the control to the Revision-cycle Counter, d. Automatic Reset During the normal operation of the correlator, the Channel Accumu-lat o r s and corresponding Memory-words are to be optionally reset at the time of transfer to the computer. The requirements of t h i s operation are des-cribed i n Section 4.4d dealing with the reset d i s t r i b u t i o n network, and are summarized as follows: 1. When the correlator i s operating i n the FREE-RUN mode, each channel i s to be reset on every nth transfer (where n i s expected to be 8 for most supersynthesis observation programs). Therefore a divide-by-eight counter, operating on the 'ready' pulses i n the Readout Control, i s required. 2. The transfer of Memory-words to the computers occurs at forty-one microsecond i n t e r v a l s and i s commanded by the pulses CP-RC1, generated by one state of the counter i n the Readout Control/Because the resetting of 117 a channel i s to occur a few nanoseconds before the leading t r a n s i t i o n of this pulse, another pulse, CP-RC8, generated by the preceding state of that counter, i s used to command the resetting action, i n conjunction with the counter i n 1. above. Although CP-RC8 i s operated by the zero or reset state of the readout control counter, i t s value i s held low when the transfer routine i s i n a c t i v e , (by the Q-output of FF-RC14b) so that false resetting of the channesl at times other than those of the transfers, does not occur. 3. The resetting of the channels i s to be provided as an option, selected by a switch on the control-panel. 4. When the correlator i s switched to the PAUSE mode (or to the STANDBY mode), the reset ting of the channels i s to be suspended, except when a manual command i s given to reset the memory and a l l of the channels as i n Subsection f below. In order to s a t i s f y these requirements, two signal l i n e s are provided by the reset control, generated by l o g i c c i r c u i t s D and E, which go to the 'enable' terminals of the Master Reset Distributor. The d i s t r i -butor i s active (transmits a low l o g i c l e v e l or inverted pulse to the channel being currently addressed) when both l i n e s are at a low voltage l e v e l . The output of D i s low under the following conditions: a) the correlator i s free-running (PAUSE = 1), the reset option switch i s on, and the divide-by-eight counter has counted the eighth readout i n t e r v a l ( i . e . 'RESET ENABLE' = 1); or b) the correlator i s i n either the PAUSE or STANDBY mode, and the Memory Reset button has been depressed (as described i n subsection e below). The output of E i s normally a t r a i n of inverted pulses - replicas o f those of the strobe-signal CP-RD2, which coincide with the transfer-command signal CP-RC8. However, when the Memory-Reset button i s depressed, the output of E becomes a continuously low value. 118 e. I n i t i a l Accumulation I t i s useful now to consider the nature of the i n i t i a l a c t i v i t y of the correlator, after i t has been switched to the ACTIVE mode. I t i s evident that a l l control functions and data-handling c i r c u i t s w i l l become operative. Yet, the numbers delivered at the time of the f i r s t transfer to the computer, for a l l channels, w i l l have s l i g h t errors, for the f o l -lowing reasons. 1. The channel outputs are multiplexed, one at a time, into the Glass Memory, so that the 'accumulation i n t e r v a l s ' for successive channels are displaced i n r e a l time, one with respect to another, i n units of one micro-second. Because a l l channels s t a r t accumulating at the same instant when the correlator i s switched into the ACTIVE MODE, the l a s t channel read w i l l have had a longer accumulation i n t e r v a l than the f i r s t , by 256 microseconds, for the f i r s t number transferred to the computer. Over a sixteen second period, however, this error becomes n e g l i g i b l e , and smaller than the round-off error. 2. I t w i l l be noted that while the correlator i s i n the STANDBY mode, the p a r a l l e l output of the Delay Quantizer and the s h i f t pulses, are s t i l l oper-ative. Therefore the S h i f t Register f i l l s with data continuously, and cor-r e l a t i o n a c t i v i t y may resume without delay when the ACTIVE mode i s estab-li s h e d . However, i t w i l l be seen i n the next section that although spurious pulses a r i s i n g from the switching of delay-cables w i l l normally be prevented from entering during the f i r s t few milliseconds of operation i n the ACTIVE mode. Therefore, i n a p r a c t i c a l observation program, the f i r s t accumulation should be discarded by the computer (or used merely as a check on the operation of the correlator) and subsequent accumulations may be treated as v a l i d contributions. 119 I f the reset option i s used, so that groups of eight (or n) successive accumulations are combined as a single integration, then i t i s desired that the reset counter could be arranged to cause the resetting of the channels after the f i r s t transfer has been made to the computer, and after every eighth transfer thereafter, f. Memory Reset At any time i n the operation of the correlator, p a r t i c u l a r l y i n test circumstances, i t may be desirable to reset the Memory, and the Channel Accumulators, independent of the synchronized routines i n progress i n the correlator. This may especially be desired when the correlator i s i n the STANDBY condition and one wishes to clear the contents of the channels and memory i n order to s t a r t a fresh accumulation under test conditions. However, the Channel Address Counter, the Timing Pulse Generator, and the Readout Control are inactive i n this mode, -so that the operation of auto-matic resetting as i n Subsection d. above cannot be used. The task of the Memory Reset function, then, should be to re-activate the Channel Address Counter; to cause the Reset D i s t r i b u t o r to issue pulses continually to a l l of the Channels addressed sequentially; and to hold the input of the Glass Memory at a l o g i c a l '0' value. The minimum duration of this a c t i v i t y should be one Multiplexing-cycle or approximately 1/4 millisecond. The provision o f this function i s i l l u s t r a t e d i n Figure 4.7.1. The 'Memory Reset' routine i s activated by the Memory Reset Switch. When the memory reset switch i s depressed, the following effects r e s u l t : 1. C i r c u i t G, which normally delivers the 1 MHz clock-signal CP-CA1, to the Channel Address Counter, instead passes the a u x i l i a r y 8 MHz signal which 120 i s g e n e r a t e d i n t he M a s t e r T i m i n g U n i t . ( A c t u a l l y , the 8 MHz s y n c h r o -n i z i n g - s i g n a l i s u s e d f o r t h i s p u r p o s e ) . T h u s , t h e c h a n n e l s a r e s e q u e n t i a l l y a d d r e s s e d at a r a p i d r a t e . 2. The o u t p u t s o f c i r c u i t s D and E a re f o r c e d to the low l e v e l as d e s c r i b e d i n S u b s e c t i o n d , and the Rese t D i s t r i b u t o r i s c o n t i n u o u s l y e n a b l e d . T h e r e -f o r e , as each c h a n n e l i s a d d r e s s e d , as i n 1., i t r e c e i v e s an e x t e n d e d r e s e t p u l s e . I t s h o u l d be n o t e d t h a t t h i s t akes p l a c e o n l y when the c o r r e l a t o r i s i n e i t h e r the STANDBY o r the PAUSE mode. However, b e c a u s e o f accumu-l a t i o n r a t e i n the c h a n n e l s i s so r a p i d , r e s e t t i n g them a s y n c h r o n o u s l y w h i l e t he c o r r e l a t o r i s i n the FREE-RUN mode wou ld s e r v e l i t t l e u s e f u l p u r p o s e . 3. The o u t p u t o f c i r c u i t F , w h i c h i s t h e r e s e t l i n e to the i n p u t f l i p - f l o p o f t he memory, goes l o w , and z e r o s a r e f e d i n t o t h e q u a r t z d e l a y - l i n e . The u s e f u l n e s s o f t h e Memory Rese t f u n c t i o n i s p r i m a r i l y i n t h e t e s t i n g o r t r o u b l e - s h o o t i n g o f t he c o r r e l a t o r . Under p r a c t i c a l c o n d i t i o n s , i n w h i c h the computer t u r n s the c o r r e l a t o r o n , b u t r e j e c t s t he f i r s t a c -c u m u l a t i o n i t i s n o t n e c e s s a r y to r e s e t the Memory o r Channe l A c c u m u l a t o r s w h i l e the c o r r e l a t o r i s i n STANDBY. However, i n t e s t c i r c u m s t a n c e s i n w h i c h c a b l e - s w i t c h i n g i s n o t i n v o l v e d , the ' f i r s t a c c u m u l a t i o n ' may be q u a n t i t a t i v e l y u s e f u l , so t h a t one w o u l d w i s h to m a n u a l l y r e s e t the Channe l s and Memory b e f o r e s t a r t i n g an a c c u m u l a t i o n . The e r r o r s i n t h i s a c -c u m u l a t i o n , due to n o n - u n i f o r m i n t e g r a t i o n - t i m e s i n t he c h a n n e l s , a r e n e g l i g i b l e . 4.8 OPERATION CONTROL: PROMPT PULSE SUPPRESSION The u n i t t o be d i s c u s s e d i n t h i s i n t h i s s e c t i o n f u n c t i o n s a l o n g w i t h the Re se t C o n t r o l i n p r o v i d i n g the v a r i o u s mode o p t i o n s d e s c r i b e d i n . t h e i n t r o d u c t i o n o f t he p r e v i o u s s e c t i o n . I t s p r i m a r y f u n c t i o n i s t h e s u p p r e s s i o n o f t h e a c t i v i t y o f t he Q u a n t i z e r s d u r i n g the i n t e r v a l s i n wh i ch the c a b l e s i n the d e l a y s y s t e m 121 are being switched. At each switching, spurious pulses and transient signals are l i k e l y to be generated, which should not be processed by the correlator. Over a period of time comparable to the switching-time of the reed-switches used i n the delay-system, the Quantizers must be i n h i b i t e d . I t i s also required that during these i n t e r v a l s , the D.C. subtraction operation should be suspended. Therefore, i t was convenient to include the D.C. Subtraction Control on the same c i r c u i t - c a r d with the Prompt Pulse Suppression Control. Since one of the functions of the PAUSE mode i s the suppression of the D.C. subtraction, the implementation of this mode i s also included i n the present unit. a. PAUSE and FREE-RUN Modes. In the introduction of the preceding section, the usefulness of the PAUSE mode for test purposes was suggested: when the correlator i s switched from the FREE-RUN mode (normal) to the PAUSE mode the e x i s t i n g information i n the memory i s to be preserved and the accumulation of products i n the channels i s to be suspended. To achieve t h i s , the following conditions are required: 1. Prompt and Delay Quantizer outputs ( i n the s e r i a l form), and sample pulses, must be i n h i b i t e d , so that a l l Channel-Multipliers (including the special channels which count the quantizer outputs and the sample pulses) produce constant zeros. One might suspect that the information i n the memory may be pre-served merely by ceasing the operation of reading the outputs of the Chan-nel-Accumulators ( i . e . to i n h i b i t the read-multiplexing c i r c u i t s ) . However this action would feed constant zeros to the d i g i t l i n e s A Q and A^ entering the Arithmetic Unit. A review of the revision algorithm employed i n that unit w i l l reveal that the Memory-words would i n i t i a l l y suffer a destructive 122 al t e r a t i o n . 2. The resetting of channels (during transfers of Memory-words to the PDP-9 computer) i s to be suspended (discussed i n previous section). Normally, when a Memory-word i s to be loaded into the Buffer Register, zeros are directed into the Memory i n i t s place. In the PAUSE mode, the Memory-word should be re-inserted into the Memory as w e l l as loaded into the Buffer Register. 3. The subtraction of the D.C. component from the Memory-words must cease. 4. I t i s desired that the correlator be switched between the FREE-RUN and PAUSE modes manually. However the actual switching must take place at the beginning of a: Revision-cycle, l e s t the updating of a Memory-word be disrupted while i n progress. Therefore the manual switching (asynchronous) must be treated as an 'enabling' action for a synchronous switching between the .modes ,by some I n t e r n a l l y .generated s i g n a l . 5. When the correlator i s switched to the STANDBY mode, i t i s desirable to have the quantizers also i n h i b i t e d so that the prompt-pulse d i s t r i b u t i o n system and MAM's are inactive. The PAUSE mode should be automatically included i n the 'General Reset', condition. Synchronism i n this case, of course, i s nbt a concern. The signal for i n h i b i t i n g the operations mentioned above, termed 'PAUSE', can be provided as shown i n the block diagram, Figure 4.8.1. F l i p - f l o p A, whose two states correspond to the modes PAUSE and FREE-RUN, i s turned on or off by strobe-pulses of the signal CP-DC2, which occur a few nanoseconds before the beginning of the Revision-cycle, tg. These pulses are directed to either the RESET or SET terminal of the f l i p -f l i p , depending on the status of the manual switch and the 'AUTO-PAUSE' variable (occurs with 'GENERAL RESET'). AUTO-PAUSE FREE-RUN PAUSE AUTO-SUPPRESS CP-RC5 y CP-DC2 &~ -C> PAUSE 12 STAGES ADDITIONAL STAGE 0 F * 4096 RESET SUPPRESS D f T T } . D.C. SUBTRACTION CONTROL H ) CP-DCl ' D.C. SUBTRACTION OPTION I - O F F A . PAUSE SUPPRESS Figure 4.8.1 Block Diagram of the Prompt Pulse Suppression Control, (which includes the D.C. Subtraction Control). ho CO 124 When the c o r r e l a t o r i s i n t he ACTIVE mode, and the PAUSE /FREE -RUN s w i t c h i s t h rown , the f i r s t s u b s e q u e n t s t r o b e - p u l s e changes the s t a t e o f the f l i p - f l o p . The i n p u t e q u a t i o n s o f the RESET and SET t e r m i n a l s o f f l i p - f l o p A , p r o v i d e d by c i r c u i t s B and C a r e as f o l l o w s : (B) RESET = AUTO-PAUSE + PAUSE-SWITCH + CP-DC2 (C) SET = (PAUSE-SWITCH * CP DC2) + AUTO-PAUSE When the f l i p - f l o p i s i n the no rma l FREE-RUN s t a t e , (PAUSE-SWITCH = 0 , AUTO-PAUSE = 0) the RESET t e r m i n a l r e c e i v e s p u l s e s o f CP-DC2. The SET t e r m i n a l i s h e l d a t a h i g h l e v e l under the same c o n d i t i o n , and the f l i p - f l o p i s i n the FREE-RUN s t a t e . As soon as 'PAUSE-SWITCH' changes to t h e v a l u e ' 1 ' , the .SET t e r m i n a l i s h e l d a t a h i g h l e v e l . The f l i p - f l o p then s w i t c h e s to the PAUSE s t a t e . I f , w h i l e the f l i p - f l o p i s o n , t he AUTO-PAUSE v a r i a b l e change to ' ! ' ( r e p r e s e n t i n g the immed ia te r e s p o n s e o f t h e f l i p - f l o p 2a i n the R e s e t C o n t r o l to the ' STOP ' command), t he RESET t e r m i n a l i s i m m e d i a t e l y h e l d a t a h i g h l e v e l , w h i l e s w i t c h e d to the PAUSE c o n d i t i o n . However, when the v a r i a b l e AUTO-PAUSE i s r e s t o r e d to ' 0 ' , f l i p - f l o p A i s t u r n e d on a g a i n by the s ynch ronous p u l s e o f CP-DC2. The c o n t r o l s i g n a l ' PAUSE ' g e n e r a t e d by the o u t p u t o f f l i p - f l o p A , i s u s e d as f o l l o w s : 1. I t i s d e l i v e r e d t o the P r o m p t - p u l s e - B u f f e r c a r d , where i t i n h i b i t s t h e s e r i a l o u t p u t s o f a l l t h r e e q u a n t i z e r s . I t i s a l s o u s e d to i n h i b i t t h e sample p u l s e s d e l i v e r e d t o the s p e c i a l c o u n t i n g c h a n n e l f o r r e c o r d i n g the number o f samples u s e d i n an a c c u m u l a t i o n . . T h i s number can be u s e d to t e s t the s i z e o f t h e numbers a c c u m u l a t e d i n the o t h e r c h a n n e l s . 2. In t he A r i t h m e t i c U n i t , i t p r e v e n t s t h e o u t p u t ga te s f rom f e e d i n g z e r o s i n t o the Memory i n p l a c e o f t h e d i g i t s coming ou t o f i t . 125 3. The D.C. Subtraction c i r c u i t , on the same c i r c u i t - c a r d , i s i n h i b i t e d by 'PAUSE'. 4. 'PAUSE' i s used i n the reset control as part of the control function governing the action of the Reset Di s t r i b u t o r . The detailed implementation of these and other c i r c u i t s discussed below, i s shown i n Figure B.8.1. b. D.C. Subtraction Control The subtraction of the false D.C. component from the accumulated products i n the channels was discussed i n the section dealing with the Arithmetic Unit. I t was established'that a s i g n a l , CP-DC1, must be provided, which assumes the value '1' during every fourth channel-multiplexing cycle. In Figure 4.8.1, the c i r c u i t G i s shown to be counting pulses of CP-RC5, which correspond to the addressing of channel P Q . Actually CP-RC5 i s ANDed with the CP-DC2 strobe-signal, timed so that the t r a n s i t i o n of CP-DC1 occurs at the beginning of the Revision-cycle. Because CP-DC1 i s to be i n h i b i t e d while the correlator i s i n the PAUSE mode and while the delay cables are being switched (see the next subsection), the output of the counter i s gated with the control signals 'PAUSE' and 'SUPPRESS'. In addition, the D.C. subtraction i s to be optional, so that a t h i r d gating s i g n a l i s provided by the option switch for D.C. subtraction. c. Prompt-Pulse Suppression During the switching of the cable-delays, which occurs about-once every second, the channels of the correlator must be is o l a t e d from the analog input signals. This i s o l a t i o n may be accomplished either by the suppression of the outputs of the quantizers or by the i n h i b i t i o n of 126 the sample-pulses which clock the sample-and-hold c i r c u i t s of the quantizers The f i r s t of these i s actually necessary, for the quantizers produce strings of i d e n t i c a l samples (rather than zeros) when the sample-and-hold c i r c u i t s are not taking new samples. Also, however, since there i s one special channel which records the number of smaples used i n an accumulation, the sample-pulses going to the channel must be i n h i b i t e d while the quantizer . outputs (samples) are suppressed. On the other hand, i t w i l l be shown, at the end of this section, that i t i s necessary for the Correlator Shift:— r e g i s t e r to continue receiving new Delay-samples (binary-coded, p a r a l l e l outputs of the Delay Quantizer) during the 'suppression' i n t e r v a l . There-fore the sample pulses delivered to the Delay Quantizer (and hence also to the Prompt Quantizers, for convenience) should not be suppressed. During the normal operation of the correlator, with uncorrelated inputs, the average products produced i n the m u l t i p l i e r s are equal to the false D.C. bias which i s l a t e r subtracted i n the Arithmetic Unit. This biasing i s due to the scheme of po s i t i v e number l o g i c which s i m p l i f i e s the design of the Multiplier-Accumulator Modules. Now, during the suppres-sion-period of cable-switching, the accumulation channels are to be con-sidered as having completely uncorrelated samples at t h e i r inputs, and should therefore be only accumulating the D.C. bias component. However, because the quantizers are being i n h i b i t e d , the outputs of the m u l t i p l i e r s w i l l be zeros. That i s , the D.C. bias component i s not produced i n t h i s c i r cumstance, and therefore D.C. subtraction i n the Arithmetic Unit i s not required and should be suppressed. 127 Ari s i n g from these considerations are the following requirements: 1. The suppression of the Quantizers and D.C. subtraction i s to be commanded by an asynchronous pulse issued by the PDP-9 computer. However, to avoid disrupting the revision of the Memory-word currently being processed i n the Arithmetic Unit, the suppression must be delayed a few nanoseconds and i n i t i a t e d by a synchronous strobe-signals, at the beginning of the Revision-cycle. 2. The duration of the suppression must at least equal the time required to complete the cable-switching operation (2 to 3 milliseconds). However, a l l channels must be deprived of an equal number of samples.in t h e i r ac-cumulations, so that suppression must l a s t for an i n t e g r a l number of com-plete Multiplexing-cycles. Further, D.C. subtraction i s involved In every fourth Multiplexing-cycle, and i t s complete i n h i b i t i o n for a l l changes can be achieved only i f the suppression occurs for four complete cycles, or any multiple thereof. The suppression can begin with the selection of any one channel i n the Multiplexing-cycle, but must also terminate with the same selection. In Figure 4.8.1, those requirements are met by the action of f l i p - f l o p s E and D and the counter G. F l i p - f l o p E i s turned on by the 'AUTO-SUPPRESS' signal from the computer. Its output enables gate Dl and the f i r s t subsequent pulse of the strobe CP-DC2 passed to the SET terminal turns on f l i p - f l o p D, at the beginning of the revision of one Memory-word. The pulses of CP-DC2 passed by gate Dl are counted i n c i r c u i t G. On the count of the 4,096th pulse, the l a s t stage of the counter turns on and i t s output transmits a 'reset' signal to f l i p - f l o p E. The next subsequent pulse of CP-RD2 then turns fl o p - f l o p D o f f , v i a gate D2, again at the beginning of the revi s i o n of the same Memory-word at which i t turned on. The 128 SUPPRESS signal thus generated by f l i p - f l o p D, which provides the i n h i b i t i n g control to the Quantizers and D.C. Subtraction Control, i s on therefore for a t o t a l time of 4.096 milliseconds, or exactly sixteen multiplexing-cycles. Occurring once every second, this suppression i n t e r v a l amounts to a .4 percent decrease i n the integration time for each s p a t i a l Fourier component sampled by the Supersynthesis Telescope. I t i s to be noted that the delay between the issuing of the command s i g n a l , 'AUTO-SUPPRESS' by the computer, and the turning on of the suppression signal i s one microsecond. Since the reaction-time of the mechanical switches i s about one millisecond, the suppression of the Quan-t i z e r s begins w e l l i n advance of the generation of any spurious pulses. At the completion of the suppression i n t e r v a l , i t i s desirable that the cross-correlation a c t i v i t y resume immediately. Tf the Delay Quantizer were to be i n h i b i t e d during this i n t e r v a l , the Correlator S h i f t -r e g i s t e r would contain only zeros and a delay would be incurred while the register r e f i l l s with legitimate Delay samples. The Delay Quantizer must be re-enabled sometime i n advance of the completion of the suppression i n -t e r v a l (at least the r e f i l l - t i m e of the s h i f t - r e g i s t e r ) . No harm however i s caused i f the Delay Quantizer i s not:suppressed at a l l , since products w i l l be formed i n the channels only when t h e - a c t i v i t y of the Prompt Quan-t i z e r s resumes. In addition, the transients induced i n the input signal l i n e s by the cable-switching a c t i v i t y w i l l have subsided by about one m i l -lisecond before the end of the suppression i n t e r v a l (4.096 milliseconds), and therefore the Delay samples i n the l a s t portion of the i n t e r v a l w i l l constitute v a l i d data. The 'SUPPRESS' signal generated by t h i s control unit i s required, then, to i n h i b i t only the s e r i a l outputs of a l l three 129 quantizers, and the same pulses going to the special counting channel. A l l channels, including the special channels, w i l l be inactive during the suppression i n t e r v a l , while the normal a c t i v i t y of a l l other un i t s , including the Readout Control, w i l l continue. There i s one circumstance i n which the correlator may not be protected from the influence of cable-switching transients. When the correlator i s i n the STANDBY mode, the Timing Pulse Generator i s inactive so that the strobe signal CP-DC2 does not e x i s t . Therefore the Supression-control c i r c u i t cannot respond to a command from the computer, so that i f the computer were to command the correlator to s t a r t up within 4.096 m i l l i -seconds of i t s commanding a switching of cables, the transients generated w i l l enter the correlator unsuppressed. The f i r s t accumulation, w i l l be erroneous. Hence, i t i s arranged that the f i r s t accumulation, i n any p r a c t i c a l observation program, i s to be discarded by the computer. 130 V. CONCLUSION 5.1 TESTS The assembled correlator i s shown i n Plate 2. The Central Processing Unit, located i n the two racks at the midsection of the assembly, has been tested and found to operate successfully, according to the design. D r i f t s i n the phases of a l l timing signals have been minimized. The MAM cards and thei r associated multiplexing and pulse d i s t r i b u t i o n c i r c u i t s are mounted i n racks above and below the CPU, and have been made f u l l y operational. Cross-talk and noise pickup on signal and control leads have been eliminated by the extensive use of either multi-wire l i n e s with alternate l i n e s grounded, of twisted pair l i n e s . The glass memory i s shielded by enclosure i n a grounded metal box (right hand side of the cor r e l a t o r ) . I n i t i a l tests involving the actual processing of data signals have been made only with sixteen channels operating. The system successfully interacts with the PDP-9 computer, and a program for the Fast Fourier Transform has been employed. (This was written by Dr. CH. Costain of the D.R.A.O.). A f u l l test of the 256-channel correlator w i l l be performed shortly. Such a test w i l l involve the auto-correlation of a monochromatic s i g n a l , at a frequency within the passband of the correlator, i n the presence of a bandlimited white noise. Faulty operation of the A/D c i r c u i t s or of any MAM channels w i l l be apparent i n i r r e g u l a r i t i e s i n the cross-correlation c o e f f i c i e n t s . The test i s to be performed with the CW signal at various frequencies within the passband. 131 5.2 RECOMMENDATIONS a. The 3-level by 5-level correlator In section 4.1, suspicion was placed on the performance of the comparators i n the A/D c i r c u i t s at the highest signal frequencies. Recom-mendations were given for th e i r improvement, i n case they are found inad-equate. In the testing of the MAM c i r c u i t s , i t was found that the response of some of the TTL l o g i c gates and f l i p - f l o p s to the maximum pulse rate of 32 MHz was c r i t i c a l l y dependent on the shape of the signal pulses. The pulse d i s t r i b u t i o n system has been designed to deliver uniform and noise-free pulses to the MAM cards. I f i t i s found that f a i l u r e s are frequent, the gates involved may be replaced by faster l o g i c (e.g. Schottky-clamped TTL). Several changes i n .the design .of the correlator are possible with respect to the contruction of additional models, as more antennas are added to the supersynthesis telescope. These are considered below. i ) Much of the design of the arithmetic and data processing c i r c u i t r y i s based on the use of the c i r c u l a t i n g glass memory. However, current costs of TTL integrated c i r c u i t s permit the addition of sixteen stages ( f l i p - f l o p s or multi-stage counters) to each Channel Accumulator. The outputs of these stages may be delivered through gates (open collector) to sixteen bus-lines servicing a l l channels, thereby requiring only one Read-multiplexing network. (Note.that this change i s also possible for the present correlator). The Arithmetic Unit would be eliminated and the timing c i r c u i t r y would be s i m p l i f i e d . i i ) Another type of replacement for the Glass Memory i s i n the form of multi-stage s h i f t r egisters. Sixteen 256-bit s h i f t registers operating 132 i n p a r a l l e l at a 1 MHz s h i f t - r a t e would store the Memory-words, and the s e r i a l revision process would be replaced by combinatorial l o g i c i n the Arithmetic Unit. I t i s suspected that this change would y i e l d l i t t l e saving i n cost, but would improve the r e l i a b i l i t y of the central memory. i i i ) A large percentage, about 15%, of the cost of the present i n -strument i s due to the p r i n t e d - c i r c u i t card connectors used (approximately 230 p a i r s ) . I t should be possible to reduce the number required by moun-ting several MAM channels together on larger cards, provided this does not reduce the r e l i a b i l i t y of the connections or the ease of servicing the channels. b. Improvement i n S/N The.S/N degradation factor of a correlator employing the simple design of the Multiplier-Accumulator Modules may be improved i n a number of ways, provided that faster l o g i c c i r c u i t s can be used. 1. The prompt pulse rate could be doubled, changing the quantization scheme from Three-by-Five levels to Three-by-Nine l e v e l s . This would y i e l d an S/N improvement of 2.6%. 2. The delayed signal could be quantized i n four l e v e l s , permitting the prompt pulses to enter the f i r s t and l a t e r stages of the MAM chain simultaneously. (Even more levels are possible, but require an increase i n the capacity of the s h i f t r e g i s t e r ) . The improvement i n S/N would be 4.3%. 3. The signals could be sampled at twice the Nyquist rate (four times the bandwidth), though this requires extra s h i f t - r e g i s t e r stages. None of these promises s i g n i f i c a n t gains i n S/N to offset the complexity of the d i s t r i b u t i o n of the faster pulses. A l t e r n a t i v e l y , f i n e r quantization might be obtained with the 133 use of up-down counters or d i g i t a l m u l t i p l i e r s using large scale integration, (e.g. Advanced Micro Devices' AM2505, 2's Complement M u l t i p l i e r ) . 5.3 Conclusion The technology of the manufacture of l o g i c c i r c u i t r y , i n small, medium and large scale integration i s advancing rapidly, making available faster and more d i v e r s i f i e d modules. The most e f f i c i e n t and economical design of an instrument such as the one described depends on the state of development of the technology at the time of consideration, and i s made rapidly obsolete by further developments. The cross-correlation spectrometer was designed on the basis of the most economical use of the devices available i n the period 1969-70. The experience of constructing i t enables one to improve future designs by the use of l a t e r components. However, the present instrument i s useful, and i s expected to be incorporated as an essential part of the supersyn-thesis system at D.R.A.O. I t i s hoped that i t s design w i l l render i t r e l i a b l e . and easy to use and service. Plate 1 Top: Prompt Pulse I n h i b i t o r card, front and back view; Bottom: Multiplier-Accumulator Module front and back view. Plate 2 The 256-Channel Cros 135 REFERENCES 1. Argyle, P.E. "Spectrometer Survey of Atomic Hydrogen i n the Andromeda Nebula." Astrophysical Journal, vol.141 (1965), p. 750. 2. Batchelor,. R. A., Cooper, B. F. C., Cole, D. J . , and Shimmins,A. J. "The Parkes Interferometer." Proc. I.E.E.E., v o l . 30, No. 10 (October 1969), pp. 305-313. 3. Blackman, R. B., and Tukey, J. W. The Measurement of Power Spectra. New York, Dover Publications Inc., 1958. 4. Bowers, F. K. " M u l t i - l e v e l Correlation Spectrometer For Radio Astronomy." 1971 I.E.E.E. International Convention Digest, New York, 1971, I n s t i t u t e of E l e c t r i c a l and Electronic Engineers, Inc., Paper No. 3D.2, p. 156. 5. Chang, K. Y. "On the Error Analysis of Correlation Devices." M.A.Sc. Thesis, Dept. of E l e c t r i c a l Engineering, University of B r i t i s h Columbia, 1969. -6. Cole, T. " F i n i t e Sample Correlations of Quantized Gaussians." Australian Journal of Physics, v o l . 21 (1968) p.273. 7. Cooper, B. F. C. "Correlators with Two-Bit Quantization." Australian Journal of Physics, v o l . 23 (1970), pp. 521-7. 8. Davies, R. D., Ponsonbyj J. E. B., Pointon, A. G., de Jager, G. 11 The J o d r e l l Bank Radio Frequency D i g i t a l Autocorrelation Spectrometer." Nature, v o l . 222 (June 1969), p. 933. 9. Elsmore, B., Kenderdine, S., Ryle, M. " The Operation of the Cambridge One-Mile Diameter Radio Telescope." Monthly Notices of the Royal Astronomical Society, v o l . 134 (1969). 10. Hogg, D. E., Macdonald, G. H., Conway, R. G., and Wade, C M . " Synthesis of Brightness D i s t r i b u t i o n i n Radio Sources." Astronomical Journal, v o l . 74 (1969), p. 1206. 11. Jenkins, Gwilym M., and Watts, Donald G. Spectral Analysis and i t s Applications. San Francisco, Holden-Day Inc., 1968. 12. K l i n g l e r , R. ( i n preparation) M.A.Sc. Thesis, Dept. of E l e c t r i c a l Engineering, University of B r i t i s h Columbia. -13. Kraus, J. D. Radio Astronomy. New York, McGraw-Hill Book Company, 1966. 14. Millman, J . , and Taub, H. Pulse, D i g i t a l and Switching Waveforms. New York, McGraw-Hill Book Company, 1965 15. National Radio Astronomical Observatory. " The VIA Concept." The VIA Proposal, v o l . I . , S p r i n g f i e l d , Va., CFSTI, Stock PB 173912, January 1967. 16. Phister, M. Logical Design of D i g i t a l Computers. New York, John Wiley and Sons, Inc., 1968 (copyright 1958). 17. Roberts, M. S. " Integral Properties of Galaxies." Astronomical Journal, vol.74, (Sept. 1969). 136 18. Roberts, M. S. "Hydrogen i n 32 Galaxies." Astronomical Journal, v o l . 73, (Dec. 1968). 19. Robinson, B. J. and Van Damme, K. J. "21 cm Observations of NGC 55." Australian Journal of Physics, v o l . 19 (1966), p. 111. 20. Ryle, M., and Hewish, A. "The Synthesis of Large Radio Telescopes." Monthly Notices of the Royal Astronomical Society, v o l . 120, No.3, 1960. 21. Ryle, M. " The New Cambridge Radio Telescope." Nature, v o l . 194, May 1962. 22. Sloan, D. S. "A Search for Galactic H 2 ." M.A.Sc. Thesis, Dept. of Physics, University of B r i t i s h Columbia, A p r i l 1969. 23. Swenson, G. W., J r . , and Mathur, N. C. "The Interferometer i n Radio Astronomy." Proceedings of the I.E.E.E., v o l . 56, No. 12, Dec. 1968. 24. Swenson, G. W.4Jr. " Synthetic-Aperture Radio Telescopes." Annual Review of Astronomy and Astrophysics, ed. L. Goldberg, 1969, v o l . 7, p. 353. 25. Weinreb, S. " A D i g i t a l Spectral Analysis Technique and i t s Applications to Radio Astronomy," M.I.T. Technical Report # 412, Cambridge, Mass., August 1963. 26. Widrow, B. " S t a t i s t i c a l Analysis of Amplitude-Quantized Sampled-Data Systems." A.I.E.E. Transactions on Applications and Industry, v o l . 79, 1961. 137 APPENDIX The technical appendix to this thesis, contains detailed notes on the operation, layout, and wiring of the various units i n the correla-tor, and i s intended for the use of the technician who maintains the instrument. Therefore, i t has been produced under separate cover. However, diagrams and tables which serve to complete the discussion of the design described i n this thesis are included here. Channel Nunber Inputs and Pin Numbers Output , before D.C. Subtraction Output, after D.C. Subtraction, 2N Use,after Subtraction o f A N p 123 . ir123 Continuum Channels P124 •t x" ;21 Ber. N* £ (x*+2) N* E (x1) + oN - Cp(124) Subtract 2C_(124) from Cp and Cgp (0 to 122). SP124 x" »19,12 par.* ' x" ,14 ser. N* E (x'+2)(x'+l) N* , N* E (x') + 3E x* + oN - Csp(124) Normalize Delay Quantizer < ( . * ' / > ?P125 t. y'; ;2i •N* t (y^+2) N* E (yj) + AN - Cp(125) Subtract from Cp(0-122) i yV> SP125 t "SP N* N* E (yj.p) + oN - Csp(125) Subtract from Cgp (0-122) P1Z6 y^ ' ;2i Y l ? ( p . r . , ' » . N* E (yp+2) (y,p(Par.)+1) N* , * .827E (yp)2 + 2E y p ( p a r < ) N* + £ yp + AN - Cp(126) Normalize Prompt Quantizer SP126 BLANK P127 Sample \Z\ Pulses 2N* AN - Cp(127) Subtract from all other coefficients 3P127 . ySP(par.)!^  Similar to P^ Normalize Semi-Prompt Quantizer. T»t>le. A.3.1 Ktivjamant •* &p*ci«l Channels Note: The quantities accumulated i n these channels are averages taken over a period of time which i s displaced by as much as 10 ms with respect to the accumulated products i n the correlator channels, due to the sequential transfer of the accumulations to the computer. Therefore, I f the mean of the output of a quantizer deviates from zero during that 10 ms i n t e r v a l , i t s effect w i l l appear i n the products accumulated i n the current Computer Transfer Interval but not i n the accumulation i n the special channel for that same i n t e r v a l . An error i n the D.C. Correction thus occurs. The maximum possible error, however, calculated as the accumulation of products which are four standard deviations above the mean product for a duration of 10 ms, amounts to much less than the round-off error incurred i n the readout process. A CP-FFAR3 ' ' CP-AR1 MEMORY R E S E T - G E N E R A L R E S E T H-1 LO Figure B.2.1 Schematic Diagram of the Arithmetic Unit. 0 0 139 i'10, f12, V\;°, i 2 , Vt i 6 , Is, i w , V2, v«. v MEMORY DIGITS AVAILABLE AT FLIP-FLOPS 8a, 8b CLOCK TO 8a M i l l hhh\ l l l I l l M I i 1 I \ l 0 ' Ziflt— 3 0 ns | 15 CLOCK TO 8b "DIGITS Ag,Aj 'FROM C15 + 31>| \ READ-MULTIPLEXERS . i ^—• 70 ns ^ DIGITS AVAIT.ABLE IN THIS INTERVAL CP-AR1, AT 11a t1 + 42 tj + 46 CURRENT VALUE AVAILABLE 62 ns MEMORY DIGITS INTO ADDER . CARRY DIGITS INTO ADDER SUM DIGITS OUT OF ADDER CIN - COUT FR0; PREVIOUS DIGIT T T I I I I I I k K h l I I I I | I I I 1 I I * 0 + 2 6 i i width 145 ns DIGITS A .A INTO T Z ^ 0 1 GATES lid, lie C0 + 2 tW C U R R E N T V A L U E S CP-AR2 CP-AR3 tn + 30 J ~ l <—60 ns tj + 30 »|—•^~ 62 ns MEMORY DIGITS OUT OF GATE la OT lb C0 + *\ f|Ao jAl 15 r t f t t NEW MEMORY DIGITS MQ Mj M.j M4 C P - D C 1 J SIGNAL TURNS ON DURING INTERVAL t , AT BEGINNING OF EVERY FOURTH MULTIPLEXING CYCLE CP-RC1 ' t. DURATION 1 vs CP-RD3 ADDER CARRY-OUT tp + 16 ~-*f| c0 + 10. CP-FFAR9 I I I M I I I M I | I J I i i w i j m f k i j T j T ^ ^ TT t„ + 46 Figure B.2.2 Timing Relationships for the Arithmetic Unit. CZ C4 £ 6 C 8 C10 C12 C14 C 0 C2 I | l l , | i t i c 8 ho hz C l I I I • t 3 +8. 1 I 1 I 1 I 1 J 1 I 8 MHz 4 MHz 2 MHz jori n n n n n n S A M P L I N G P U L S E S A T I N P U T TO Q U A N T I Z E R S 1 MHz Qi Mhz NOT SHOWN) '70 ns t 0 + 25 J—L_ _T~L n r r r Figure B.3.1. Operation of the Bandwidth Selector. 8 MHz 4 MHz 2 MHz 1 MHz S H I F T P U L S E G A T I N G WAVEFORMS, A T O U T P U T O F G A T E 8d I N BANDWIDTH S E L E C T O R O 141 16 MHz CLOCK U ^ l K f l 4^7on TO CLOCK £ INDICATOR IN TIMING UNIT CARD 3 +5V ? _ , 25 pF > 1KQ - r ^ - ~ A •IEM1 Q- 470f2 -jT)—[Ij)>- PULSE DELAY 1 |7k)---p/Jl) [~4b)-> CP-MEM2 H> CP-BR1 -fc CP-CDU GENERAL RESET Figure B.3.2 Revision Cycle Counter, Timing Unit Card .1. w Figure B.3.3 Reference Waveform 16 MHz 142 Figure B.3.4. Pulse Generators, Timing Unit Card 2. 143 DRIVER ure B.3.5 Pulse Generators, Timing Unit Card. 3. CP-TC2 Figure B.3.6 Schematic Diagram of the Computer Transfer Timer. 145 SHIFT PULSES IN SIXTEEN POWER GATES I I I L 4 © — > SHIFT PULSES OUT TO CHANNEL RACKS (TOO PER RACK) Figure B.3.7 Shif t Pulse Buffers, on single card. Figure B.3.8 Read-multiplexer Strobe Pulse Buffers. 146 V A2 * NUMBERED BRACKETS AND MARKINGS ARE AN AID TO THE RECOGNITION OF THE PATTERNS n M M F n i ,n.FTrn i in r ^ T u i n ^ M . i . r ^ i - - - ^ o A "'I I, ,^,,PTTT1,,flpTm,,,,rTT-rn,, , , | . . i | l | , , --> o 4 t 5 5 t-5 «• — J I—*. CONTINUING PATTERN 4,5,5,4,5,5 3 A l A 0 J~T_LTO_uJn i J r 1aJ^^ u T i j m i \ r i j x r u r i j ^ ^ e e c  3 1- t • t * B, - 1111 r~^~*~i it 111 * * 111 r etc. i J j ~ n J j n j T i x r ~ i Bj U_l ' JJ_l I I l_U ' etc. B 0 iinjianjTjai^  Figure B.4.1 256-State Pattern of the Eight Channel Address Code Di g i t s . Four gates la,b?5a,b B3 P — L L / — ^ B3 2 ;6 — N j. To Multiplexers, etc. i n 2 P [) 2 Channel Racks B i >—^y-+\ - 4 ;8 B ° p — — > B ° Figure B.4.2 Channel Address Code Buffers, Four L.S.B.s, (two i d e n t i c a l cards). Four M.S.B.s are on card with Computer Transfer Timer. CP-RC3 PHASE - PRECEDES CP-RC5 BY 440 -825 ns ~ 41 SYNCHRONOUS COUNTER RESET TO j COUNTER CP-RC2 CP-RC8 I CP-RC7 $ (TO D.C. SUBTRACT COUNTER) INITIATING CIRCUIT ' J IE) CP-RC4 (1 MHz) CP-RC5 (CHANNEL ? Q SELECTED) CP-RC6 (CHANNEL P„.) Z4 CP-RC1 TO ARITYMETIC UNIT 1 4 ^ I TERMINATING J CIRCUIT GENERAL RESET Figure B.5.1 Schematic Diagram of the Readout Control. 4S --4 148 0 FFRCUb Figure B.5.2 Schematic Diagram of the Divide-by-41 Counter i n the Readout Control. : 'O C 2 C 4 C 6 t 8 C 1 0 C 1 2 C14 C 0 t 2 '4 £ 6 C 8 'lO" " t12 C14 " C 0 • I ' 1 , ' 1 1 I ! • I I I I I I I I READY PULSE CP-RC7 }"J< 4QQ ns ( K i n . ) . J  0-FFRC14b , CP-RC5 CP-RC4 t g + 28 ^ r-t cl? + 3 0 Q-FFRC14a TOGGLE PULSE TO 14a_ Q-FFRC14a ^ 1 5 + 3 9 CP-RC3 (PASSED BY GATE 15d) J l J l CP-RC1 STATES OF tCOUNTER STANDBY STATE, 0 STATE 1 Figure B.5.3a I n i t i a t i n g Timing of the Readout Control. 1—1 STATES OF COUNTER STANDBY f rf, COUNTING INITIATED ± J f T r ? " " ' o ll '2 I »0 h l2 1 < — 41 ys -4-l s t CYCLE ->| 1 ys CP-RC8 CP-RC2 CHANNEL SELECTION CODES CP-RC6 CP-RC4 2nd 3 r d . COUNTING TERMINATED, Y COUNTER RESET ' o i l * 2 I 0 256th CYCLE TL_ n L T O ! FIRST MULTIPLEXING CYCLE O C-4 p-> CM Cv, C 4 39 CYCLES I 39 PULSES OF CP-RC6 I IN THIS INTERVAL I I MULTIPLEXING CYCLE T T T - i - -I 256 ys — 9.984 ys •> 215 ns->J J l ^.RISING EDGE t J 2 + 20 ; t c . • e t c . RESET PULSE OUT OF GATE 12c t +44 « 0-FFRC14a Figure B.5.3b Terminating Timing of the Readout Control. READ PROG STATUS INT I / O S K I P A P I R E A D I/O P R I O R I T Y ( ) R E Q U E S T S Y N C , * R O G N OCH O V E R * F L O W R O G N E N I 0 A O D R L I N E S Bl D I R E C T I O N A L I 0 B U S 4 I O P 2 D E V I C E S E L E C T I O N L I N E S fTTT fill-HMO mi ,1, t>"T* U /tl' — IB \ 03101 <r D£\)IC£ Sit-Lc-T-A, JV>T a nil f t HI**"* 3 DO U 1 1 » tort* T o n e A U T O fne/inxier 3 o T H » A * T » I W P i t C t f toff-fi.lt? C o * " * (pevtttri.) S U B D E V I C E S E L E C T I O N 0 7 Figure B.6.2 Block Diagram of Interface Between the Correlator and the PDP-9 Computer, Designed by J.H. Dawson at D.R.A.O. h2 C 1 4 C 0 C 2 C 4 C 6 '8 C 1 0 h2 C14 e 0 C 2 £ 4 t i l i I i i • i > I l i y " i , , , , , , , , , , , ,M l 5,  DIGITS AT INPUT t^W"*"' T"! j I 1 I I I 1 I I I 1 I £-CLOCK ON POSITIVE EDGE cP-BRi, INPUT R J i n j i n j i J T J i j n j ^ ^ GATE 10a ' tr +48 j : [ CP-RC1, DELAYED TO - — J 1 GATE 10a t DONE-FLAG i LTJ L_ Figure B.6.1 Timing Relationships for Buffer Register. e 0 t2 tA t 6 t 8 C 1 0 C 1 2 fc14 !o l2 C 4 l6 '' tB C 1 0 C 1 2 ' K ' O I I I I i i I i i i i 1 i i i i I cp-RC7 n CP-RES1 I OUTPUT OF COUNTER ON 8 t h COUNT OF CP-RC7 PULSES SELECTION OF t q + 2 8 _ J ~ 1 CHANNEL P. — 2 1 ' t + 20 r R E S E T S T R 0 B E PUISE CP-RD2 — ; — ; 1 5 ~-»n TL CP-RC8 I FROM READOUT CONTROL CP-RC1 . ; ' I TRANSFER OF CHANNEL P Q OCCURS Figure B.7.2 Timing Relationships for the Reset Control. SYNCHRONIZING PULSES > MANUAL START MANUAL STOP CHANNEL RESET OFF OPTION MEMORY RESET ON MEMORY RESET CP-CA1 Figure B.7.1 Schematic Diagram of the Reset Control. Figure B.8.1 Schematic Diagram of the Prompt Pulse Suppression Control 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            data-media="{[{embed.selectedMedia}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0101497/manifest

Comment

Related Items