JPRS ID: 10466 USSR REPORT CYBERNETICS, COMPUTERS AND AUTOMATION TECHNOLOGY

• Document Type: CREST
• Collection: JPRS
• Document Number (FOIA) /ESDN (CREST): CIA-RDP82-00850R000500050042-5
• Release Decision: RIF
• Original Classification: U
• Document Page Count: 144
• Document Creation Date: November 1, 2016
• Sequence Number: 42
• Content Type: REPORTS

APPROVED FOR RELEASE: 2007/02/49: CIA-RDP82-40850R040500054042-5
FOR OFFICIAL USE ONLY
JPRS L/ 10466
- 19 April 1982
USS~ R~ ort
p
CYBERNETICS, COMPUTERS AND
AUTOMATION TECHNOLOGY
CFOUO 5/82)
- FB~$ FOREIGN BROADCAST INFORMATION SERVICE
_ FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5
NOTE
JPRS publications contain information primarily from foreign
newspapers, periodicals and books, but also from news agency
transmissions and broadcasts. Materials from foreign-language
sources are translated; those from English-language sources
are transcribed or reprinted, with the original phrasing and
other charact~ristics retained.
Headlines, editorial reports, and material enclosed in brackets
- are supplied by JPRS. Processing indicators such as [Text]
or [Excerpt] in the first line of each item, or following the
last line of a brief, indicate how the original information was
processed. Where no processing indicator is given, the infor-
mation was summarized or extracted.
Unfamiliar names rendered phonetically or transliterated are
enclosed in parenthesPS. Words or names preceded by a ques-
tion mark and enclosed in parentheses were not clear in the
original but have been supplied as appropriate in context. ~
Other unattributed parFnthetical notes within the body of an
item originate with the source. Times within items are as
given by source.
The contents of this publication in no way represent the poli-
cies, views or at.titudes of the U.S. Government.
,
COYYRIGHT LAWS ANP REGLTLA,TIONS GO'JERNING OWNERSHIP OF
MATERLALS REPRODUCED HEREI:~ REQUIRE THAT DISSEMINATION
OF THIS PUBLICATION BE RESTRICTED FOR OFFICIAL USE OiVLY.
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5
FOR OFFICIAL USE ONLY
3'PRS Z/10466
19 April 1982
USSR REPORT
CYBERNETICS, COMPUTERS AND AJTOMATION TECHNOLOGY
C~oUO ~/s2~
CON7ENTS
HARDwARE
; Improved YeS-1055M Computer 1
PS-2000 Compu~er Complex ci
Questions of Simulating Large Boundary Problems by
Appro~dmate IJiakoptics Method 7
Associative Processor for Problems of Controlling Affiliated
Systems 11
Analog Memory Unit of Analog Computer 18
,
P'hysicotechnological Problems of Computer Engineeri.ng 20
Broad Application Memory Uni.t 26
SOFTWARE
Adapting Machine Graphics Program Packs to Graphic Information
Ilisplay Devi.ces 33
Centra'1-Level-Data Manipulation Language in Heterogeneous
Data Base Integration Syste~ns 38
Collective-Use Computer Center~ Software, Use of Natural
i,a.nguages !~2
Analysis of Space of Rank Estimatesand S~mthesis o~ Structural
. Analytical Models for Automating Expert Decision MalQ.ng 43
Creation, Application of Automated Ilialogue Training Systems.. !a.l4
= Multicxiterion Optimization Problems 1~5
- a- [III - US~K - 2].C S&T FOUO]
rnn ~+r~rr~~ . T r tn~ na~r v
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/49: CIA-RDP82-40850R040500054042-5
FOR OFFICIAL USE ONLY
Development and Introduction of Packages of Applied Programs.... 1~6
Selected Items From Journa7. tAI~GORITFII~IS AND PROGRAMS~,
September 1981: El~brus !~7
More Selected Items From Journal ~AIGOKITHMS AND PROGRAMS',
September 1981 49
Selected Items From Journal ~AI,CzORITHMS AND PROGRAMS~,
October 1981 53
APPLICATIONS
Automated Sup~rvisory Control System in llth Five-Year Plan 59
~ Survey of Work on Syntactical Component of Machine Translation
Systems 64
PARALI.Ei, . COMPUTING
Parallel Computer and Prograrrun.ing Systems fi8
Realization of Operating Sysf,em for Multiprocessor Complex
of Superprocessors of E1'brus-1 Multiprocessor Computer
Complex 73
Parallel Processin~ in Estonia .......................A.......... 81
OPTICAL CHARAC'PER READERS
- General Review of Optical Character Readers 86
Main Charactzristics of Soviet Character Readers 90
Design, Operation of Ruta-701, Sever-3 Character Readers 95
Design, (lperatioi, of Chars Character Reader 113 '
Equipment Costs, Wages for Character Readers 118
CON~RENCES
International 'I7iagnostic F~cilities for I}igital S~stems'
Conference 119
Optimization of Control Systems ?22
- - b -
FOR OFFICIAL USE ONL~'
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/49: CIA-RDP82-00850R040500050042-5
PUBI,ICATIONS
Methods of Numerical Simulation of Atmospheric Processes 124
Selected Items F`rom Journal ~INSTRUMENTS, MEANS OF AUTOMATION
AND CONTROL SYSTEM5', February, 1981 131
Selected Items F`rom Journal ~INSTRUMENTS, MEANS OF AUTOMATION
AND CONTROL SYSTENIS~, April 1981 133
Selected Items From Journal ~INSTRUMENTS, MEANS OF AUTOMATION
AND CONTROL SYSTEMS~, May 1981 134
Selected Items From Jourr,al 'INSTRUMENTS, MEA13S OF A?JTOMAT?~N
~ ANP CONTROL SY~TEMS~, November 1981 136
Selected 2tems F`rom Jo-arnal 'INSTRUMENTS, N1~ANS OF AUTOMATION
AND CONTROL SYSTEMS~, December 19$1 138
~
a
- c -
FOR OFFICIAL USE ON~.Y
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5
FOR OFFICIAL USE ONLY
HARI7WARE -
UDC 681.3:65.011.8
IMPROUED YeS-1055M COMPUTER.
Moscow PRIBORY I SISTEMY UPRAVLENIYA in Russian No 11, Nov 81
PP 7-9 .
[Article by engineer-economist A. Gruenewald, Germar. Democratic Republic]
[Text] As every year, the VEB Kombinat Robotror. presented its latest artieles
in the field of information processing and office equipment at the Leipzig
Spring Fair in March 1981.
The main exhibition among them was the improved YeS-1055M computer of the
Unified Computer System (Ryad-2j. The YeS-1055M computpr (see figure) was
shown at the fair in the followizg configurations
the YeS-2655 central processor (German Democratic Republic);
a~specialized processor--the MAMG matrix m~duTe (German Democratic
Republic);
~ operator console with YeS-7069 display (German Democratic Republic);
YeS-5567/5067 plug-in disk store with capaci~y af 100 Nmytes (Peoples
Republic of Bulgaria);
YeS-5517/5017.02 magnetic tape store (German Democratic Republic);
~
YeS-7033M parallel printer (Polish Peaples Republic);
YeS-6016 punch card input device (CSSR);
YeS-8371.01 remote data processor (Polish Peoples Republic) conr_ected
through modems to the YeS-8577 programmable multif:znctional user statinn
(German Democratic Republic);
YeS-7902M equipment complex (German Democratic Republic); '
YeS-7920M display system (German Democratic Republic);
YeS-5075 floppy disk store (CSSR).
- 1
- FOR OFF[CIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/42/09: CIA-RDP82-00850R000500054442-5
FOR OFFICIAL USE ISNLY ~
The new SVr~/YeS operating system and the improved OC/YeS system are used in
the YeS-1O55M computer.
The innovations related to modernization of the YeS-1055 are described in this
article.
The YeS-2655M Central Processor
The YeS-2655M central processor differs from the YeS-2655 processor in the
~ following improvements.
1. Internal storage (OZU) is realized on the basis of highly integrated
microcixcuits, its capacity was increased to 4 Nmytes and all 4 I~ytes are
based on two panels and are located in the processor rack. .
- The doubled capacity of the internal storage (4 compared to 2 I~ytes) facil-
itates organization of the user operating mode, contributes to an increase of
task processing productivity and thus ensures an increase of computer operat-
ing efficiency. When functioning in the virtixal memory or working file re-
- write mode in the internal storage, the external stores, for example, plug-
in disk stores, are freed.
Due to the absence of a special cabinet for the internal store, the tot~l
v~olume of devices and the area occupied by the central processor are reduced
to 1.74 m2 (compared to 3.7 m2). The power supply unit is designed on the
~ransfonnerless principle and due to using new integrated circuits for the
internal storage, t:~e consumed energy was reduced to 4kW (compared to 7.4 kW).
The required output of air ~onditioners was reduced in this case.
The microprogram store based on PPZU [semipermanent memory] (memory for stor-
age of constants) was replaced in the YeS-2655M processor by a reprogrammable
microprogram storage device based on statistical components of the PZU Lread-
o*~ly memory]. It consists of a main rigidly programmed part and a freely
loaded auxiliary part wi~h capacity of 1K microinstructions (unlike the 32
microinstructions of the YeS-2655). Loading speed is considerably increased
- due to the presence of a loadable microprogram storage device.
The Matrix Module (MAMO)
The matrix module is a specialized processor for expanding the capabilities
- of the YeS-2655M pmcessor when performing certain calculat;ing ~perations
Cfor example, matrix and vector calculations, correlation calculus, rapid
_ Fourier transforms and so on). The MAMO module is not an independent device
but operates under the control of the YeS-2655M processor. To do this, the
corresponding control programs are provided in the SVS (system with vertic~l
memory) and OS/YeS operating systems.
The MfiMO module controls the instructions in machir~e code. To do this, the
set of instructions of the YeS-2655/2655M central proceasor is supplemented
with nine control. i.nstructions (for debugging and accounting) and 28 instruc-
tions for prep~ration of fields and lines.
2
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/49: CIA-RDP82-00850R040500050042-5
'rOR OFFICIAL USE ONLY
Debugging micraprograms, direct error-detec+~ion programs, error analysis sys-
tem, debugging programs and special components of the operator's console have
been developed for convenience in operating the matrix module. A diagnostic
' bus~ new in its solution, provides access to all flip-flops of the matrix
module and to all cells of the read-only inemory of the matrix module. The
presence of a special diagnostic device i.n the matrix module creates good con-
ditions for servicing and diagnosis of the matrix madule.
The structure of the matrix module permits parallel and independent perfar-
mance of the following functions: microprogram control, data storage in a

~ buffer, opera~ions on numbers with floating point and operations nn exponents.
An increase in the productivity af the matrix module is achieved by realizing
parallel performance of operations and the assembly-line method of processing
= instructions, storage of data in a buffer, induced processing (which provides
a saving of inemory and machine time), selection of versions without increas-
ing machine time and programming simplicity.
In this case productivity is increased compared to usual fulfillment of pro-
grams approximately 10-50-fold and depends on the file.size, loading density
and comanunication algorithm.
New YeS-7902M and 1'eS-5075 Peripheral Devices
The new YeS-7902M complex is an input-output device with standard interface
of the Unified Computer System using papertape and cassette magnetic tape as
data carriers. The YeS-7902 papertape complex used until now has been re-
placed by this device. The device is produced in ~wo versions: M1 as paper-
~ tape devices and cassette magnetic tape store and M2 as only papertape devices.
- The main advantages of the YeS-7902M device compared to the YeS-7902 papertape
complex. are the fUllowing: higher reliability due to use of modern component
base, less occupied ar~a (0.6 instead of 1.5 m2), lower mass (260 inste~d of
' 325 kg) and lower v~olume (0.8 instead of 1.45 m3).
The YeS-5075 device--a floppy disk input-output device (CSSR)--was first of-
fered in the closed opera~ing mode and permits direct reading fram floppy
disk carriers.. A device of this type is used widely in the decentralized
computer equipment devices of the VEB ICc~mbinat Robotron.
Improved OS/YeS Operating System
The expanded veision 6.1 of the OS/YeS can operate on the YeS-1040, YeS-1055
and YeS-1O55M modems. Part af the number of the expanded functions is real-
ized in configuration of SVS control programs and can therefare be used in
the YeS-1055 and YeS-1055M. The functional expansions are compatible in a
"bottom to top" operating system or are altornatives in use.
The follawing modifications have been made.
1. Functional task servicing expansion. The "Spool" system, which offers a
number of advantages to users, has been introduced: maximum operating speed
3
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02109: CIA-RDP82-00850R400500050042-5
FOR OFFICIAL LiSE ONLY
of devices with systems input-output for punch card input or punch car~ output
devices and printers, increased capacity for tasks, better utilization of disk
sections of direct access devices to memory, new capabilities of control of
YeS-5075 floppy disk :store and new functions and improved reliability with
systems output to pun~�h card output printer.
2. The capabilities in sorting and fusion of data have been expanded. The
sorting process is accelerated due to new methods. Howe~rer, t'Zese methods
can be used only in new models of the Unified Computer System since they are
implemented only with expanded instruction reserve.
3. Joint use of OS/YeS and SVM/YeS systems. The function of combining sys-
tems in expanded version 6.1 of the OS/Ye5 was developed to incz�ease the oper-
ating efficiency of the SVS under the control o~ the SVM/YeS. This function
is realized in several components of the SVS control program and ensures ef-
ficiency of the SVS both in real and virtual machines.
4. Software of user s~ations of the Unified Computer System. The VEB K,ombinat
Robotron has developed software for IIniiied Computer System user stations us-
ing the di:ect access method during remote data processing (VTAM and TSAM)
within the framework of the production program "Decentralized computer
equipment."
The use of new devices in the system permits a considerable expansion of user
stations (terminals) and thus permits on~ to take into account the most diverse ~
user requirements.
These devices include the YeS-8577 programmable multifunctional user station,
YeS-7925M and YeS-8556 programmable display user stations and YeS-8551 univer-
sal programmable user station designed to solve scientific, engineering and
economic problems.
Realization of the Concept of Virtual Machines Using the New SVM/Ye5 Operating
System
The general operating principle in virtual computers, being the basis of the
S~M/YeS system, consists in the fact that parallel operation of several users
on the same computer is also carried out in addition to multiprogram process-
ing. In this case a virtual machine is offered to each user. The internal
storage is dete~nined so that each user has his own. The central processor
is offered to each user in the tima-sharing mode for equal intervals. Input-
output devices are shared or designated so that each user has at his disposal
a specific number of input-output devices.
The SVM/YeS operating sy~tem adjusts virtual machines to the real configura-
tion of com~uters and provides servicing and control of parallel operation of
virtual machines. Every virtual computer operates for the user as an inde-
pendent computer with its own resources and under the control of its own op-
- erati.ng system selected with regard to user requirements.
4
FOR OFFICIAt USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R004500054042-5
FOR OFFICdAL USE ONLY
Because of parallel operation of several virtual machines, the user can con-
trol the selection of operating modes--batch processin3, dialogue, debugging
and system servicing.
The SV/Ye5 operating system is distinguished by its broad capabilities:
it permits a) parallel operation in dialogue and batch processing modea,
_ b) performance of systems programming ind technical servicing tasks parallel
with the production process and c) parallel processing of the most diverse
problems on the same computer. Its advanta~es are the absence of idle times
related to replacement of inoperating systEm, high operating reliability and
reliability of data due to independence of virtual machines and simplicity
for operators to master.
The following can be used as operating systems operatinq under the control of
the SVM/YeS system for ~irtual machines:
1. The OS/YeS operating system~ which can be used not anly for batch process-
ing but for dialogue problems as well. It can function in parallel with all
configurations of control programs (MFT, MVT and SVS). Adjustment to operate
under the control of the SVS/YeS system has been realized for the corifigura-
tion of SVS control programs to impro~ve the efficiency of the system as a
_ whole.
- 2. The DOS/YeS (Disk operating system of t3nified Computer System] oparating
' system, which can be used under the control ~f the SVM/YeS as a batch process-
ing system. Tn this case the users having program packs for the DOS/YeS do
not have to rework them for new versions of the operating systems. .
3. ~'he PTS programming and debugging system related to the SVM/YeS file. The
PTS system is designed to d~velop, debug and fulfill programs in the dialogue
operating mode, for which the PTS system has been provided with its own in-
structi~n language and control files. The PTS system is distinguished by
high flexibility in use and simplicity to master.
4. The FRTS remote file transmission system, being a component of the SVM/YeS
and controlling file transmission between distant input-output devices and
virtual machines and between RFTS dat a stations. The main designation of the
RFTS consists in transmission of control data, programs and results of debug-
ging when working out proc~rams under the control of the PTS system at remote
displays.
5. The PDAS error analysis system--a component of the SVM/YeS system, having
special modules and procedures for error analysis and usPd to supplemant the
PTS system. The PDAS system is used as a valuable auxiliary device for the
syeteme programmer for gnthering, eervioing and analysis of errore of the
SVM/YeS opera~ing systems and devices.
The YeS-1055M computer displayed at the Leipzig Fair corresponds to modern
international trends in development of computer equipment~with extensive use
- of microelectronics and the related increase of productivity with a simultan-
eous reduction of consumed energy, occupied area and volume and also its own
modern software and common characteristics of the system.
COPYRIGKT: Izdatel'stvo "Mashinostroyeniye". "Priboxy i sistemy upravleniya",
1981
6521 5
CSO: 1863/69
FOR OFFICIAL USE ONLY .
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/49: CIA-RDP82-00850R040500050042-5
FOR OFFiCiAL USE ONLY
_ PS-2000 COMPUTER COMPI,E~C
Kiev ELEKTRONNpYE MOD~LIROVANIYE in Russian Vo~. 4,No 1, Jan-Feb 82 inside front cover
: .
;r y ~ , ;
tr ~ , ~
. r: ; -
, ~
. �
E..,
,~'Y ~ a?F~�.
~ ` ~ '~:1..
.
fi,.7"r.4 . ~
5~i~',~~.~ :
i
!
f,~'
~
~
~
[Text] FS-2000 computer compJ.ex developed by the Scientific Research Institute of
Control Co~4puters a~d fihe S~vexadone~sk~~~.'atpul,~s [Pu~.s.e] Scie~tific ~xQduc~ion
Association together with th~ T~oscow Tnsr~tute o~ Control Qz'oble~s~. Designed
for high-efficiency~processing o~ data obtained as the result o~ geophysical and
oceanographic research, te~ts of various complicat~d equipment and the 1ike.
CQPYRIGt~T: Izdatel'stv~ "Naukova dumka", "Elektronnoye modelirovaniye", 1982
8831
_ CSO: 1863/95 6
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5
FOR OFFICIAL U~E ~NLY
UDC 537.212.001.24
QUESTIONS 0~' SIMULATING LARGE BOUNAARY PROBLEMS BY APPROXIMATE DIAKOPTICS
METHOP
Kiev ELEKTRONNOY'E MODEZIROVANTYE in Russian Vol 4, No 1, Jan-Feb 82 (manuscript
received 13 Jun 80) pp 40-45
r:=h
[Article by Yakov I1'ich Tete1'bau ,~unior scientific associate, All Union
Petroleum and Gas Scienti~ic Resea Tnsti~ute, 1Koscow] ~
[Excerpr.s] In [1] a description is given of the method of approximate diakoptics
for a machine solution with limited equipment of large two-dimensional boundary
problems represented by means of electrical nets with a very great number of
nodes. This method is applicable for "net - digital computer" ("Saturn-1") hybrid
computer complexes or for digital computers with a worki.ng storage capacity in-
sufficient for accommodating the entire field. Direct methods of calculating
complex electrical and electronic circuits have been developed on the basis of
diakoptics [2-6]. By the meth.od of diakoptics a large system (electrical circuit)
is broken down into several sma11 subsystems each of which is calculated separately
~nd then the particular solutions are united by means of a"skeleton intersection
circuit" (skeleton diagram). This is a general ma.jor system described only by means
of variables included in the conditions for joining subcircuits, i.e., variables ~
(e.g., voltages) at boundaries at points of sections. The skeleton diagram of the
general circuit is made up of the skeleton diagrams of subcircuits, whose parameters
are determined from experiments in short-circuiting and breaking the circuit (no-
load) for the boundary points of the joining of subcircuits.. These experiments are
performed on a computer by means of digital or analog simulation.
The boundary problem for a system with distributed parameters is xeplaced with the
accuracy of a finite difference approximation by a system of linear algebraic equa-
tions simulated by means of a resistive electrical net representing an electrical
circuit for which it is possible to obtain a precise solution by the method of dia-
koptics. As compared with a camplete system of finite difference equations for the
entire region repxe~ented by the electrical net, the equations o~ the skeleton
intersection diagram contain considerably fewer variables (~ust for points of the
joining oP subregions), which s~~mplifies the solution.
Methods o~ diakop~ics are ef~ective when the circuit i,s di,vided i,ntv paxts at a
limited number o~ points. Tn the case of the fin~.te difference xe~resentation of
_ boundary problems subregions are divided along boundaries regularl}* containing a
great number of nodal points. The advantage of ~he direct method of diakoptics
7
FOR OFFICIAi. USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5
FOR OFFiC(AL USE ONL'Y
here i~, practica7.7~r xeduce,d ~a nougt~~ b~caus~e, a~ tt~e ~.ahAx ~,nt~ns~.v~nes~s. o~ a
precise ,3ezermi.nation a~ the p~xame~~,rs o~ ~t1e ske,~,eton d~,agxama .of ~.nd~,v~idua7.
su'~regions. ~ons~quently, a n~w method became necessatxq~ Cca:~l~ed the method of
approximate d~,akopt~,cs) wh3;ch 3,s bas~d vn s3.mp7.i~icatio~ o~ the ske7.etan i:ntex~
section circuit i'or de~exmining v'oltages at tlie boundztxfies o~' ~oined subregions [1] .
This method either d~,rectly or w~th re~inem~nt th~ough i.teration tnakes possible a
substantial reduction 3:n machine ti~tne, and fihe dfirect method of approximate dia-
koptics is not in opposition to iteration methods, but organ~.cally supplements
them when the initia7, approximation is unknown. The method of approxtmate dia-
koptics in terr.~s o~ the number oi exchanges urlth tfie external storage is consistent
~ with just a single block iteration, since experiment~ in breaking the circui~
(no-load) and ~hort--circuiting require onlp a one-tim~ ca11 of a subregion to the
para11e1 processor (net) or to the working etorage.
Experience has demonstrated the h3gh accuracy of approximate diakoptics solutions,
which results in a considerable saving of machine ttme. This is obvious from the
graph shown in fig 1.
a QQ
, o,~ ~
a~
0~0! Nu 2 ~
OAOJ /yd 1 ~
? 4 B O f0 12 A~
Figure l. Curve Comparing Convergence of Solutions by the Method of B1ock
Iteration and the Method of Approximate Diakoptics with Refinement
Key: ,
1. Nd [diakoptics] 2' Ni [iterationJ
The rate of convergence of iterations [i] has the form
r=-N lnss"',
k
~1~
where 8 and b are measuxes of the exror vector (k ~ N) and of k itera-
tions (mean values)k
Designafiing the initial. error o~ the iteration proce~s as $ , the initial error
after diakoptic approximation a~ b , the max~mum permissi~~e error of the
analog para11e1 processor or o~ dig~~a1 re~inement as a, the number of iterations
wittiout diakoptics as Ni , and the number o~ re~ining i~erations w3,th the existence
of a diakoptic so~ution as Nd ,~rom (~1) we get
8
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/49: CIA-RDP82-40850R040500054042-5
FOR UFFiC1AL USE ONLY
N�--~1nSH'. NA~~r~ao
0 A
Let flor the boundaxy~ node~ o~ ~he jo:~ning o~ subreg~ons (~rom the experience of
- solving prob7.eiqs~ x A 0.45 , dd ~ 0,003 , a A 0.001 and 0.5 ; then,
taking into account the ~act tY?at the ti~ne fgx a d~akoptic so~ution corresponds
to a single iteration, we have the ~ollowing sav~ing o~ time for a hybrid computing
- complex of the net - digital computer type: N/(1 ^t~ N)= 4 times and with r=
~ = 0.45 , dP = dod = 0.003 , we have Nd = 0 and Ni a 10 times.
Relatively high accuxacy is obta~.ned ~or approxfimate dfiakoptics in spite o~ the
_ rough approxiatati.on of the skel~ton diagram~ of subregions a~ compared with the
precise skeleton diagram~ determined by~ the method of diakoptics for electrical
nets. According to the procedure described in [1] the connection of subregions is
performed not at all boundary nodes, but with a certain spacing of the "lattice
of net electrodes," whereby in determining the parameters of the skeleton diagrams
of subregions only the ma~or connections between electrodes are taken into account.
The rough approxitaation of skeleton diagrams substantially simplifies machine cal-
culation of the field; however, the simplifications cannot be universal since from
this it would follow that the compl2x problem would be equivalently reduced to a
sirsple one. ThereYOre, the use of skeleton diagrams of a specif3c type is justi-
fied only with a corresponding restriction of the class of problems which can be
solved: The broader the class, the more important the organization of iteration
control of the accuracy of the method and, when necessary, the performance of
_ additional refinement of the results. With this the use of t'he approximate dia-
koptics method always gives a good initial approximation.
Al1 practical calculations and the introduction of the method have been performed
_ for multipolar fields (containing a great number of sources of different signs
corresponding, for example, to in~ection wells and operating holes of oil fields).
In this case the problem is tormulated as a problem with sources of the first and
third kind distributed by region and exerting a specific shielding effect on the
boundary potentials of joined subregions and the less the exchange of current
(nodal transfers) be:tween subregions and the more identical the subregions ~oined,
the higher the accuracy.
Below a theoretical verification is presented of the metho~ of determining voltages
at the boundaries of ~oined subregions, a procedure is presented for solving a
boundary problem in a large region part by part and questions relating to the inter-
polation and iteration xefinement of the boundary voltages of ~oined subregions
are discussed.
The method descxibed can be extended easi7,y ~o sevexa~ Subxegiorts, ~s qui.te simp7.e
for purposes o~ a7,goxithmization and possesses good con~vexgence. ~xactical.l}*
speaking, probl~ms in which a reg3;on was dfivided into 2 to 15 $ubxegions have been
solved [1]. The progxam which has been urritten for the machine solution o~ ~arge
problems wfith the '~ner digital comput~r" ("Saturn-1") hybrid computer complex
9
FOR OFFICIAL U5E ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5
FOR OFFICIAL USE ONLY
with a paxa~,7,~~, ~xoc~~Kd7C ~nx 7.424 nad~s~ ~s daaign~d ~ox ~oux subx.~$ions, i.e.,
~or 4096 nodes o~ the ne~.. fin~oxma,~ion on some pxob~~s wh~,ch have b~en so~.ved
is presented in ~he tab.l,e.
Tab1e 1.
Number of prob~eru 7. 2 3 4 5
Number of net nodes 730 ~,500 1500 1300 4096
Number of subregtons ` 4 2 2 15 4
Number of sources 60 19 19 36 172
Number of net electrodes 25 8 4 38 33
Spacin~ of net e7.ectrodes 2 4 8 4 4
Mean arithmetic error o~
determination o~ boundary
voltages, perc~nt 0.3 0.93 0.3 1.96 0.85
Bibliography
1. Tetel'baum, Ya.I. "Solution of Difference Equations for Boundary Problems
= of the Elliptical Type by Circui~ Theory Methods," ELEKTRTCHESTVO, No 5, 1978,
pp 57-60.
2. Kron, G. "Issledovaniye slozhnykh sistem po chastyam--diakoptika" [Study of
Complex Systems Part by Part--Diakoptics], Moscow, Nauka, 1972, 544 pages.
3. Khepp, Kh. "Diakoptika i elektricheskiye tsepi" [Diakoptics and Electrical
Circuits], Moscow, Mir, 1974, 344 pages. ~
4. Pukhov, G.Ye. "Direct Methods of Calculating Electrical Circuits," ELEKTRI-
CHESTVO, No 10, 1963, pp 23-28.
5. Nagornyy, L.Ya. and Zhukov, I.A. "On the Question of Simulating,Low-Density
Systems of Nonlinear Differential Equations of Large Dimensionality on a
Digital Computer and Hybrid Computer Complex," ELEKTRONIKA T MODELIROVANIYE,
No 15, 1977, pp 9-11.
6. Baranov, A.V. "Elements of the Theory of Coupled Systems," Moscow, VESTNIK
MGU, SER. 3, No 4, 1965, pp 36-40.
7. Vazov, V. and Forsayt, Dzh. "Difference Methods of Solving Partial Differ-
ential Equations," Moscow, Izdatel'stvo tnoatrannoy Iiteratury, 1963, 487 pages.
S. Danilina, N.Z., Aubrovskaya, N.~., Kvasha, O.P. et a1. "Chislennyye metody"
[Numerical Methods], Moscow, Vysshaya shkola, 1976, 368 pages.
9. Tete1'baum, I;.'M. a,nd ~@te1.''baum, Xa.T. F"Modeli pryamo}* analogi~," [Models of
Dizect Anal,ogy~J, ~ioscow, Nauka, 1.979, 384 pages.
COPYRIGHT: Tzdatel'stvo "Nautcov~a dumka'~, "E~~ktxonnoy~e modelirovaniqe", 1982
- 8831 .
CSO: 1863/95
10
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2407/02/09: CIA-RDP82-00850R000500450042-5
FOR OFFICIAL USE ONLY
UDC 681.32
ASSOCIATIVE PROCESSOR FOR PROBLEMS Og CONTROLLIN~ AFFTLTATED SYSTEMS
Kiev ELEKTRONNOYE MODELIROVANTYE in Russian Dol 4, No l, Jan-Feb 82 (manuscript
received 8 Dec 80) pp 55-58, 96
[Article by Gennadiy Aleksandrovich Nikitin, candidate of technical sciences and
assistant professor, Boris Vasi1'yevich Vinnikov, ~unior scientific associate,
Yevgeniy Vasil'yevich Kapranov, ~unior scientific asaoc3ate, and Tgort Leopol'do-
vich Kaftannikov, instructor, Chelyabinsk Polytechnical Institute]
[Text] Among problems of controlling large numbers of systems in real time can be
distinguished those which are characteristic of systems possessing the following
features:
All controlled systems are homogeneous (of the same type) and form a single hier-
archical level. ~
Local control of an individual system is independent and is carried out according
to one and the same algorithm.
The data files describing each controlled system have identical structures and
dimensionality and form the system's data base.
Global control requires taking into account the interrelationships of systems and
forming from their data bases a unified file of data to be processed.
We will call such controlled systems affiliated, e.g., machine tools with program
control when processing identical parts (performing one and the same operation).
Problems of the local control of affiliated systems (checking parameters, detect-
ing impermissible deviat~c~~, etc.) are reduced to processing various data bases
according to the same algorithm and create the prerequisites for system multipro-
cessing [1], which is most effectively impl.emented with a parallel computing
structure. Global control (ordering the operation of systems and displaying the
progress of processes) is based on the sorting of files formed frnm data bases,
which requires the use of associative processing methods [2].
, Associative Processor with the Classical Structure
- The classical organization of a parallel associative processor (PAP) assumes
structuring its memory on the basis of associative memory elements (fig 1).
11
FOR OFFiCIAL USE 01+1LY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02109: CIA-RDP82-00850R400500050042-5
- FOR OFFI('IAL USF; nN1.Y
By means of an input unit containing data and mask registers access is made pos-
sible to one or more bit slices of inemory words. Any operation (arithmetic,
logical, search) on multibit operands is performed by microprogram "simultaneously"
on all words of the memory by means of elementary polling and write micro-opera-
tions. Control signals are formed by a central control unit (TsW) and a loc2ll
control unit (MUU) makes possible the modification of microinstructions to be
executed by taking into account the information to be processed in each word of
the associative memory, whfich fn this case tt is appropriate to ca11 the associa-
tive decision field (ARP).
~
1)
Bd,a~
oe yrmpaitrnl~ ~
_ 2
~
- i _ 1 i ~ 5
~
~ ' ~ i a �
I - - ~
j
I - ~
i ~ II ~ i
' -
~ i K .
L_
~ Brodwp yanpairm6o 0 ~
Figure 1. Classical Structure of Associative Processor witch, Modular
Organization
Key: ~
1. Output unit 4. Local control unit (MUU)
2. Module 5. Gentral control unit (TsW)
3. Input un~t
For the purpose of organizing parallel processing, the data base for each system
must be accommodated in a single w~rd of the associative memory. Obviously, this
kind of organization for a PAP is not distinguished by structural flexibility,
since changing the number of controlled systems or the size and structure of their
data bases results in the necessity of changing the dimensions of the associative
d.ecision field. A modular structure with a layer-by-layer arrangement of data
bases is preferable (cf. fig 1). A layer i~ formed by memory words of the same
type of all modules, which makes it possible relatively easily to increase the
capacity of the PAP when changing the dimenaionality and nature of pxoblems
solved. A chann~l for regular cou~nunicatton between modules of the ARP in solving
problems of a given class can be executed most simply in th~ form of lines for
the parallel transfer of the results of associative processing successively from
one module to another (in parallel for all words).
Generally, control problems include not only logical and search operations, but
also a~ithmetic processing. Since any processing in an ARP is performed in -
12
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/49: CIA-RDP82-40850R040500054042-5
FnR OFF(CIAL IISE ONI.Y
synchronism, then for the purpose o~ characterizing the speed o~ any operation it
is possible to use the number of elementary polling and write microinstructions
required to execute it.
Investigations have demonstrated that the algorithms for arit'~me*_~c processing in
an ARP developed in [2] are insufficiently fast. Shortening the length of micro-
programs as the result of combining polling and write micro-operations or adding
a one-bit adder to each string of the associative memory {3] somewhat improves the
algorithmic speed of the ARP. However, formation of the mass result of an arith-
metic operation is performed serially bit by bit, and in parallel (simultaneously)
only for subsets of words activated by a given microinstruction. This results in
the fact that multiprocessing at the microinstruction level proves to be consider-
abl~ inferior to that potentially possible.
Of interest is the development of algorithms which make possible the fruitfulness
of each polling and write micro-operation simultaneously for the entire set of
operands to be processed. One of the simplest is the inverse write-in method [4],
based on the possibility of writing into a location of the associative memory the
direct or inverse write-in code depending on the results of associative polling.
This method makes possible the arithmetic processing of a single operand bit
after a single microinstruction, but the operation is performed serially bit by
bit. The synchronous implementation of the principles of async~?ronous summing [5)
makes it possibZe to form with a single microinstruction no less than one bit of
the mass result, but requires additional hardware costs and hampers the use of a
standard associative memory as the basis for constructing a decision field.
Thus, with considerable amounts of arithmetic processing in problems of controlling
affiliated systems the use of the classical structure for an associative processor
proves to be inadequately effective.
Processor with Parallel-Associative Processing
Development of the arithmetic capabilities of an associative decision field has
resulted in the structure of a processor with parallel-associative data processing
[6]. This processor is in the form of a set of independent processor elements
(PE's) operating under the control of a central control unit (fig 2) which forms
control signals for the next microinstruction and feeds them simultaneously to all
PE's. A local control unit for each PE, taking into account the data to be pro-
cessed, modifies the microinstruction to be executed and forms for the central
control unit signals informing of the result obtained.
The data to be processed from controlled systems is entered via an interface by
means of an input unit into the processor's memory, formed by the memory units of
individual processor elements. Data bases are arranged layer by layer. The mem-
ory addressing mechanism implements both addreas and associative methods of ac-
cessing. Tn address accessing the principle of separation into layers with sec-
tionalization of the memory ia employed, which makes it possible to process in
parallel data bases placed in the memory i�nits of various processor elements with
an identical relative address. Associative addressing makes it possible to acti-
vate for subsequent processing data bases located in any laj~ers of the processor's
memory. When performing associative and nonassociative operations with an
operand identical for a11 words to be proces~ed, the processor is placed in a
13
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/49: CIA-RDP82-40850R040500054042-5
FOR OFFICIAL USE ONLY
general register. The input unit contains r~gisters used not only for inputing
information but also ~or organizing associative processing. Tnformation is output
through a readout buffer. Each processor element is assigned its own readout
register for the pur~~ose of increasing speed. Both address and associative access
to data to be read out are possible; furthermore, any word is identified by ite
address.
The basic functional unit is the processor element (fig 3), containing a section-
alized memory and independent logic for performing arithmetic and associative
processing. The entire memory field of a processor element can be broken down
into the folloc~ing areas in terms of type of processing: arithmetic processing
(RAM memory area); associative-arithmetic processing (associative memory areas
connected to an adder); associative processing (associative ?aemory areas not
having access to an adder); and the working area (associative memory area used for
the local control unit).
n3a 4~ - --o I
i 2~ ;
~n mprramu ~
I
n3,
n3~ uyv
8.~oduqp Ob4ue 6t~~ep
ycrnpaiCmdo 7 tucm/aa $~.vmdAou~
yCm/puC~nJ~U ~
C~ ~(ro+puue C A7rFrnro~u
Figure 2. Structure o~ Processor with Parallel-Associative Data Processing
Key:
1. Arithmetic-logic unit (BAL) 6. Input unit
2. Memory unit 7. General registers
3. Local control unit 8. Readout buffer
4. Processor element 9. System interface
5. Central control unit
Arithmetic-logic processing is performed in the arithmetic field formed by the
arithmetic-logic processing units (BA,L's) of a11 processor elements. In this case
processor elements operate in para11e1 (according to a single algorithm), making
possible system multiprocessing of the problem to be solved for data bases prelim-
inarily activated' by the associative method or those located in a single layer of
the memory. The arithmetic-logic superstructure equipping the processor elements
contains a 2-byte combination adder (SM) with an output register (RSM). Operands
for summing can be supplied ~rom various areas o~ the memory unit. ~or the
associative processing of tiles (sorting, ordering, search and the like) the
14
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-04850R000500050042-5
FOR OFFICIAL USE ONLY
- associative memories oP a11 groceasor elements are united into an assoc.iative
deciaion field making pos~ib~e access to any bit alice of operanda. ,
~M O
3 ['/y Derutmy
NU/rk~QOvuv
Q~uuu ~
~j ~-----$>Myy_i -
~rar noeronw i ~
OJ A7y i
6) 7) ~ 1
~ , ~
~ '
_ _ ~ ~
a~~:~~
~ OmPerurmpo0 13~ ~c~�auomuEknu ad/nbanru
,wiwcu
Figure 3. Structure of Processor Elements
Key: ~
1. Co~on operand 8. Local control unit
2. Output register (RSM) 9. Coincidence indicators
3. Combination adder (SM) 10. Priority chain
4. Readout register 11. Address coder
5. Memory unit 12. From write-in registers
6, g~ 13. From associative processing ~
7. Associative memory registers
This organization of the processor, as thei:result of some restructuring of the
- decision field, ma.kes it possible to create the computing structure best suited to
specific algorithms to be run. An increase in the dimensionality of problems to
- be solved and in the number of controlled systiems, without altering the basic struc-
ture of the processor, is compensated either by a proportional increase in the
- capacity of each memory unit or by the addition of additior~al processor elements.
Instruction Set Features
The familiar pr~.nciples of informativeness and logical complexity form the basis
for developing the inetruction set [7], which has made possible the high effective-
ness of programs and has znade it possible to take into account the features of
parallel, group and associative processing. The instruct~on set 3ncludes "rapid"
instructions of the "register-register" type and instructfion~ for a complicated
search in a f~.1e.
By taking into account the architectural features of a processor with parallel-
associative processing it is possfiblc to create economical formats for instructions
for processing in the dectsion field which in terms of their content are closer to
15 ~
FOR OFFICiAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/49: CIA-RDP82-00850R040500050042-5
FOR OFF[CIAL USE ONLY
- the statements of a high-level pxogranaaing language than to tzadit~onal computer
instructions. The tnstruction set has been structured as an open one with great
capabilities for expansion as the result of the addition of new instructions and
modifications of those alreadq implemented. The use in the processor of an automa-
tic check of the 1eve1 of the location of data bases (of the size of the area
occupied) frees the programmer fram the need to take into account the size of
files to be processed.
The data addressing unit is an area of the processor's memory (cf. fig 3), and a
single address field of an instruction can address several areas of the memory
simultaneously. By changing the number of address fields instructions oi the same
type of various lengths a~e obtatned (the most used instructions have the shortest
format). ~
According to the kiiid of operations performed, all instructions are divided into
arithmefiic, associative-arithmetic and associative processing instructions and
auxiliary instructions. Operands can be files or subfiles (groups) and individual
words which are placed either in the memory or in the general register or directly
in an instruction. For the purpose of programming kinds of data processing differ-
ing from "statement" instructions, in the instruction set associative polling and
write instructions are provided for any area of the associative memory through a
random control character with masking, and these instructions have the longest
format .
A distinctive feature of control instructions involves the fact that in parallel
processing oP a file in the processor a multiple transfer condition is formed
(e.g., the pre~ence of ~ust a single word in the memory with the required control
character). Tn this case it is a good idea to separate instructions for forming
conditions and for organizing direct transfer.
A program is formed as a file which can be placed in a program memory with page
organization. Here a transfer instruction makes possible the transfer of control
within the limits of just the current page, and for the purpose of transfer between
pages an instruction for setting the page register is used.
Thus, problems of controlling affiliated systems make possible a high degree of
multiprocessing. The information bases of a group of systems are proces~ed most
e�fectively by means of a para11e1 processor and it is appropriate to use associa-
tive processors for the mass comparison and sorting of files.
A parallel-associative processor with a modular structure of the associative de-
cision field, which makes it possible to increase its capacity with a change in
the dimensionality and nature of problems solved, has the greate~c structural
flexibility. The disadvantages of an associative decision fieJ.d are associated
with the relatively low degree of muitiprocessing at the micr~-operation level in
arithmetic processing. Employment of the inverse write-in m~ethod or o~ the syn-
chronous implementation of the principle of asynchronous au~raning make it possible
to increase algvrithmic speed 3n forming the mass result of arithmetic operations.
A computing structure with parallel-associative data processing unites the ad-
vantages of para11e1 arithmet3.c and associative decis3.on fields. Reordering of
the structure of the decision field as a function of the processing algorithm is
16
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/49: CIA-RDP82-00850R040500050042-5
FOR OFFICIAL USE ONLY
~
= made possible. Employment of ti~e principles of direct and associative addressing
of data make it possible in addition to the usual vector processing to impleme:it
group processfing according to control characters. The modular structure of the
d~cision field makes it possible to increase its computing capacity by adding
additional processor elements.
The instruction set is oriented primarily toward the vector and group processing
of files and has great capabilities for expans.ion. A~sociative polling and
_ write instructions for any control character are provided for the purpose of
progra~ing any~ kind of pTOCessing.
Bibliography
r 1. Vilenkin, S.Ya., Intenberg, I.T. and Nabatov, A.S. "Multiprocessing in
- Processors with a Single Stream of Instructions and ~ Multiple Stream of
Data," UPRAVZYAYUSHCHIYE SISTEMY I MASHINY, No 3, 1980, pp 61-65.
2. Prangishvili, I.V., editor. "Odnorodnyye mikroelektronnyye assotsiativnyye
protsessory" [Homogeneous Microelectronic Associative Processors], Moscow,
Sovetskove radio, 1973, 280 pages.
3. Medvedev, I.L., Prangishvili, I.V. and Chudin, A.A. "Mnogoprotsessornyye
vychislitel'nyye sistemy s perestraivayemoy strukturay" [Multiprocessor
Computing Systems with a Rearrangeable Structure], Moscow, 1975, 74 pages
(Institute of Control Problems Preprint No 85).
4. Vinnikov, B.V., Kaftannikov, I.L. and Nikitin, G.A. USSR Patent No 700863,
"Processor Module," published in BYULLETEN' IZOBRETENIY, No 44, 1979.
5. Kaftannikov, I.L. and Nikitin, G.A. "Asynchronous Method of Performing a
Group Addition Operation" in "Tnformatsionnyye i upravlyayuslichiye elementy
i sistemy" [Data Processing and Control Elements and Systems], Chelyabinsk,
1979, pp 149-151.
6. Nikitin, G.A., Vinnikov, B.V., Kapranov, Ye.V. and Kaftannikov, I.L. "Pro-
cessor with Para11e1-Associative Data Processing" in "Parallel'nyye mashiny"
[Parallel Machines], Kiev, 1980, pp 32-34. (Preprint No 223, Ukrainian SSR
Academy of Sciences Institute of Electrodynamics).
7. Arlazarov, V.L., Dinits, Ye.A. and Yemel'yanov, N.Ye. "One Approach to Con-
structing a High-Lev~l Instruction Set" in "Mnogoprotsessornyye vychislitel'-
nyye sistemy" [Multiprocessor Computing Sy~tems], Moscow, 1975, pp 45-53.
COPYRIGHT: Izdatel'stvo "Naukova dumka", "Elektronnoye modelirovani,ye", 1982
8831
CSO: 1863/95
17
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2047/02/09: CIA-RDP82-00850R400504050042-5
FOR OFFICIAL USE ONLY
ANALOG MEMORY UNTT OF ANALOG COMPUTER ~
Kiev ELEKTRONNOYE MODELIROVANIYE in Russian Vol 4, No 1, Jan-Feb 82 inside back
cover
[Text] The analog memory unit develo~ed at Kazakh State University is a unified
component of special-purpose analog computers in which resistive networks are
used. The unit's circuit contains access and storage circuits connected in
parallel and operating alternately over time to store and compile analog informa.-
tion obtained in~the resistive network. Information is stored in the form of
voltage of a specific magnitude in the plates of the memory's capacitor. The
analog memory unit is designed for the cyclic precis3on transfer of the analog
voltage value from one point in the resistive network to another.
~ i~e avt~r >,y,ak ,~q'r,~+t
c - '3'F~','~'~.F ~4
ry . ~
,;r ,vF: ~ ' .
k
_ ~'_tin
i.',
a
' i:�'
:f'~+
i
~ u;, .
~
� ~yh~
`~Y::~~~:
1~`y ~
I +
.'~f~'. ~r' .i
t~'- t' . vC~ 7n~Vi S:^ t~
? ! %'C) C ) ~
6
Figure 1. The ROS-A (a) and ROS-B (b) Stylized Type Faces.
2. For pages of text;
~ 3. for a continuous. roll of paper ~i.th line printing;
4. general-purpose macfiines that work artth various
types of documents.
91

~ FOR OF~'ICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5
FOR OFFICIAL USE ON~.Y
Tatile 1.
_ Descri~tion o� Tniti~al
Document Sp~ed
- ~'umber and
- riodel Kind of Characters Documents
(Year of Type Faces per per ,
Production) Class* Stze, mm (Alphabetl** Second Minute
~ VIiJITI-2 (1971) 2 200 x 297 1NSh "Optima" . Up to 20
(TsBS)
~ ChARS (1972) 2 210 x 297 1NSh "Optima" 200 Up to 60
- (TsSS)
= Sever-3 (1969) 2 150 x 180- Several printing 35-100 Up to 60
280 X 355 and typecariter
nonstylized type
' facea (TsBhS)
YeS (1977) 2 210 x 297 1SSh ROS-B ('rsBbS)Up to 500 Up to 50
~ B.tank-2 (1969) 1 105 X 105- 1SSh ROS-A (TsS) N/A To 400 ~
210 x 297 '
Blank-3 YeS6031 1 105 x 105- 1SSh ROS-A (TsS) N/A Up to 400 ~
(1976) 210 x 297
B,lank-4 (1977) 1 110 x 75 1NSh Printing T-8 Up to 60 Up to 400
(TsS)
Ruta-701 (1968) 2 210 x 150- 1NSh "Optima" Up to 180 Up to 20
210 x 297 (TsS), RZ (TsS)
R711 1 65 x 105- 2SSh ROS-A, Up to 1000 Up to 580 ~
105 x 210 ktOS-B, (TsBS),
RZ (TsS)
* According to the classification in subchapter 1.4.
Numeral number of type face; SSh - stylized type face; NSh - nonstylized
type face; RZ - manuscript characters. In the alphabet of character
~ readers for these types of type faces: Ts - numerals; B- capital letters;
b- small letters,� S-- other symt5ols (punctuati.on marks and the like).
[Russ~an text obscured, probaUly "50"]
[Table continued, next page]
92
FOR OFF[CIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPR~VED F~R RELEASE: 2007/02/09: CIA-RDP82-04850R000500050042-5
FOR OFFICIAL USE ONLY
~Table continued]
ReliaF~ilitp of Approxiviate
Description of Primary Block.~ Recognition Cost~ rubles
Block for Frequency
Feedi:ng and Frequency of Failures
Moving Recogn~.tion of Air. To Recog-
Documents Read Block Block % nize %
- VINITI-2 rod with "running correlatton 0.02-0.4 0.1-0.3 N/A
(1971) vacuum suc- bea~" system
tion devices, with CRT
roller con-

~ veyor
ChARS rod with photodiode correlation 0.01 0.1 140
(1972) vacuum suc- line
tion devices,
vacuum drum
Sever-3 manual feed, "running correlation 0.01--1 0.1-1 300
(1969) roller con- beam" system and analysis
veyor with CRT of struc- ~
tural fea-
tures of
characters
YeS vacuum feed, "running correlation 0.001- 0.005- 150-200
(1977) belt con- beam" system 0.01 0.05
- veyor with CRT
Blank-2 vacuum feed, photodiode analys.is of O.QO1 0.01 30
(1969) roller con- line structural
veyor features of
_ characters
Blank-3 vacuum feed, photod3ode analysis of 0.001 0.01 30-40
YeS6Q31 roller con- line structural
(1976) veyor features of
characters
Blank-4 vacuum feed, LI-601 analysis of 0.~1 0.1 N/A
(1977) belt con- dissector structural
veyor features of
characters
Ruta- rod with "runn~ng correlation 0.0~1, 0.02 65-90
701 vacuum suc- beam" syatem lanalysis or 0.14
(1968) tion de- with CRT of features manu-
vtces, for manu- script
roller con- script charac-
-veyor characters) ters
R711 vacuum drum photodiode analysis of 0.001- 0.01- N%A
line structural 0.01 0.1
features of
characters [Table continued, next pageJ
93
FOR OFFICIAL USE O1VLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/49: CIA-RDP82-00850R040500050042-5
FOR OFFICIAL USE ONLY �
[Table continued]
* ~ccording to the classii`i.cation in suhchapter 1.1.
COPYRIGHT: Izdatel~s,tvo "Tekhnika", 198Q
11,176
CSO: 1863/47
~
94
' , ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2407/02/09: CIA-RDP82-00850R000500450042-5
FOR OFFICIAL USE ONLY
DESIGN, OPERATION OF RUTA-701, SEVER-3 CHARACTER READERS
Kiev OPTICHESKIYE CHITAYUSHCHIYE AVTOMATY in Russian 1980 (signed to press
22 Nov 79) pp 95-146
[Chapter 4 of book "Optical Character Readers", by Doctor of Technical Sciences
Vladimir Antonovich Kovalevskiy, Candidate of Technical Sci~nces Georgiy
L'vovich Gimel'farb and Candidate of Technical Sciences Anatoliy Fedorovich
Voziyanov, Izdatel~stvo "Tekhnika", 4,000 copieg, 208 pagesl
[Excerpts] Chapter 4. Designs of Contemporary Character Readers
1. The Blank-2 Character Reader
The domestically produced Blank-2 character reader for reading forms with a few
lines was developed in 1968-1969 as a further development of the Blank-P label
reader. In addition to labels the Blank~2 reads a line of numbers and service
characters in the stylized type face 1tOS-A on blank forms in four formats:
210 x 297, 105 x 297, 148 x 210, and 105 x 148 millimeters. The characrer reader
is set manually for the format of the blanks and position of the line of text be-
fore reading begins.
The charact2r reader has a speed of 350-400 documents per minute for contin-
- ous reading of a package and lOq-125 documer.ts per minute if eacfi document is
fed following a request by an external computer to whtch the reader is con-
nected.
The block for feeding and moving documents has an ii~put pocket witfi a capacity
of 1,000 forms and three output pockets that held 750 forms apiece (one for
documents with unrecognized characters and two fur sorting documents according
to signals from the external computer). The input pocket rises as the package
is read so that the top sheet is presented to the feed assembly. A special
electromechani.cal tracking system with optical electric sensors which monitor
the position of the top of tfie package and correct feeding of the top sheet
(to see that it does not become crumpled and that no more than one sheet at a
time is taken) controls this risi:~g motion. The top sheet is separated from the
package liy a revo]ving disk arm. Then a small vacuum roller transfers the docu-
ment to a roller conveyor. T'he conveyor which moves the document is divided into
two parts between which there is a vacuum drum. T?:e drum holds the document while
it is read, maktng it possible to repeat the reading of a line of text if
95
' FOR OFFICIAG USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007142/09: CIA-RDP82-40854R040500050042-5
FOR OFFICIAL USE ONLY
necessary. In addition to the vacuum suction tfie shset is field in tfie drum bx
spring-activated rollers. When tfie sheet is being moved on tfie convepor it is
straiglitened out liy its leading edge: tlie res3dual misalignment of tfis sheet
in the drum does not exceed 0.3 millimeters for a lengtfi.o~ 297 millimeters.
The read bloclc of tfi e character reader Tias two separate heads: for reading
labels and for reading a line of text. Tfiey are located along the generating
line of the drimm; the :ext reading iiead can lie moved manually to adjust it ac-
cording to the position of the line on tfie sheet. This fiead contains an il-
luminator consisting of tu~o mirror ligfits, a lens, and a line of :i2 photo-
' diodes witfi pfiotocurrent amplifiers. The line of phDtodiodes is arranged
across the line being read and is approximately twice the height of a character
on the line: the image of a character projected on the line covers a height of
13-16 photodiodes. This makes it posaible to perform the reading operation
despite possible misalignments of the line being read with respect to the direc-
tion of movement of the document and vertical displacement of characters on the
line. After amplification the signals of the photodiodes are quantized on two
levels (black-white) and interpreted as binary variables (1-0 respectively).
The recognition block employs the structural logical method of recognition ac-
cording to which there is a given boolean decision function. from the binary
structural features of the character befng recognized for each character of
the alphabet. The presence o~ a certain vertical or horizontal lines (hachures)
is considered to tie a feature. A total of 10 vertical lines and three hori-
zontal lines located in five columns and three lines of the field of vision are
analyzed (see Figure 13). The choice of these features is dictated by the
characteristics of the design of the numbers and special characters of the ROS-A
type face. In the figure the upper vertical lines are designated BBl-BB5,
while the lower vertical lines are BH1-BH5 and the horizontal lines are ('1- ~ 3.
r~
~ m ~ m n
r2
.
~ ~ ~ ~ ~
ra
- Figure 13. Arrangement of the Main Lines (Hachures) for
the Numbers and Special Characters of tlie ROS-A Type Face
Recogntzed by tfie Blank~2 Character Reader.
96 .
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2407/02109: CIA-RDP82-00854R000500050042-5
- FOR OFFICIAL USE ONLY
To compensate for possible incomplete pr~nting of, tfia lines, the read zone
is divided on the vertical into 16 discrettzatton elements (cells). Corre--
sponding to each element are two adjacent pFiotodi.odes of tfie read line wh~s~e.
signals are logically summed, tfiat is, tfie element fs read as Filack if at
least one of tfie pfiotodiodes fias at i.ts output the correspondtng signal. Tfie
shortest vertical line s-hould consist of at 1Fast tfiree Iilack discretization
elements, while tfie widest liorizontal line s~liould not be ~ider than two ele~
men t s.
- The read head scans characters from right to lef t. The process of ineasuring
the features of tfie character begins wfien the first verttcal line of tfie char-
acter is detected and continues for five scanning cycles (eacfi cycle corre-
sponds to scanning one o� the columns shown in Figure 13 above). Eacfi cycle
tests whetiier tfiere is a vertical line and how it is positioned relative to the
first line found: higher, lower, or on the same level. Horizontal lines are
detected when the length of the "black'~ signal for the corresponding discretization
elements is greater than for the widest vertical ltr.e.
The boolean decision functionsfrom the features discove~~u take into account
several of the possible distorti.ons of characters during printing and reading.
_ Specifically, if we put a line over an expression to signify t~ie feature of
absence of the line and use a dot to indi.cate logical multiplication, the de--
cision functions for the numbers 1 and 2 used in the Slank-2 character reader
are written in the following form:
_ _ _ _
1-> (885 ~ BH5 � BB3 � BH3 � BB! � BH,~)
(BB4 � BB3 � BBI � BHI);
2 (BB5 � BH5 � BB4 � BB3 � BH3 � BHI � I'1 � 1'2 � 1'3)
V(BB5 � BH5 � BB4 � BB3 � BH3 � B81 � BH! � fl � l2)~/
(BB5 � BH5 � BB4 � B83 � BN3 � BB! � BHI � 1'3).
Several additional features are used to distingui.sh..certain similar characters.
In terms oF hardware the circuits to identify vertical features are constructed
on three-input logical AND circuits, each of whicli tests the existence of a
vertical line of minimum length in one of the three possible ad~acent discreti-
~ zati_on elements. This circuit includes two shift registers which make it
possible to evaluate the position of the vertical line fo~ind in ~~~h scanning
cycle relattve to the first line detected for the character. The output signals
of the AND circuits are accumulated in the fixst register durtng tfie cycle, after
which they are transmitted with a sfiift to ttie second register. This sliif t in
the first cycle is done so that the first (ini.tial) trigger of tlie second
register is set in state "1." The numher of sLeps of this sh~:ft is stored, and
in later cycles tfie shift is done on tHe same principle (Tiy the beginning of the
register}, but not for :1ore tfian tfie stored number of steps. The number of sF?ift
steps o~itained in eacfi cycle and the state of the first register trigger ("0" or
~ "1") indicates the position of tfie vertical line relati:ve to the first line de-
tected in tfie cfiaracter.
, 97
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2047/02/09: CIA-RDP82-00850R000540050042-5
FOR OFFICIAL USE ONLY
Controlled delay cixcuits ~rhi.ch ~oxm a unitaxy~ output s~gnal only whsn a
"black" signal of a definite length_i~ fed to thsir input are used to ~ind
_ the horizontal lines. Tfie output s~gnais oP tfie delay circu~.ts are copied into
the shift register and in the las~t (~iftfi~ scanning cycle are moved to tfie be--
ginning of the regtster for the.same numbex of ste~s- as signals for tfie vertical
features in tfiis cycle. The liori.zontal features are tlien identified using
three dou5le-input OR circuits wfiich are connected to pairs oF ad~acent trtg~
gers in the firs~ six liit positions ~f the regi~ter. Tlie 10 vertical and
three fiorizontal features obtained are transferred to the trigger register of
features, to wfiich a decoder ts connected. Tfiis decoder carri~s out the se-
lected decision functions.
W[lere there is an unamhig+sous answer at tfie output of the decoder the recog-
nition block forms the code of the cfiaracter recognized. If the answer is nAt
unambiguous (signals are received from two or more outputs of the decoder), it
_ issues a signal of failure to recognize.
The control lilo~k of the Blank-2 charscter :-ead~r establishes the read regime
- based on commands from an external Minsk-32 or Minsk-23 computer to whicTi the
~ reader is connected. Specifically, the computer may issue co~nands to begin
- ~r stop input, for continuous input of an entire package of forms, for indi-
- vidual input of the next document upon request, for document sorting, and so
on. The control block of the character reader independently monitors emergency
situations and stops the feeding process when two or more sheets are taken from
the packa,~e simultaneously, ~hen {.t is not possiBle to grip the next sheet,
~ when the receiving pockets are overflowing, when there are no more sfieets in
r_he feed pocket, and so on. When a signal of failure to recognize is given the
document ia thrown into a separate receiving pncket. A document also goes into
this pocket if a ma?_function signal is recetved from the external computer.
The control block is also capable of sorting documents into two pockets (by con-
, tent) in an autonomous wark regime.
Reliability of recognition. The factory specif i.~:ation for the error rate of
the Blank-2 cliaracter reader is 0.001 percent, with a rate of 0.01 percent for
_ failure to recognize. Experiments have shown that these factory ratings are
attained only when the quality of printing meets the ISO standard (see ChaptPr 1).
This quality can be acfiievec' when the text is printed using
electric typewriters that have inked rib~ions that are used just one time. As ex
periments have shown, the only domes~ti.cally produced inked riiiFion tfiat is su~t-
able for this purpose is the MKL--1 riBfion on an acetoliutyrate base. This ritabon
was originally developed to print documents for msanetic character readers.
With a carbon inked rilibon for one-~time use the frequency of e::�rors and unrecog-
nized characcers increases to 0.6--0.7 percent. A reusable silk ribbon raises
the frequency of errors and failures to recognize to 3-4 percent in tfie initi.l
stage of wear (first eight times tfirougfi tlie typewriter). As the ribbon becomes
heavily worn (,45-55 times) this ftgure reaches 40 percent.
Documents to tie fed to the Blank-:' character reader are prtnted on special paper
which was described in Chapter 1.
= 98
, ~
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2407/02/09: CIA-RDP82-00850R000500450042-5
FOR OFFICIAL USE ONLY
Comparative analysis of the ~rorking characterirtica of the B.tank--2 ciiaracter
reader shows tfiat in terms of speed and cost tt competes success�ully with
some foreign-made character readers~ of tii~ same class sucli as tfie Data

~ Recognition 700 (United States) and the ICL 82Q1/1 and TCL 8401/1 (England),
but is inferior to sucfi models as tfi~ Un~vac 2703 and Honeywell 243 (United
States). All these cnaracter readers were produced at approximately tTie same
time (1969-1970). The Univac 2703, unlike the otfier models mentioned, reads
numbers and special cfiaracters in three stylized type faces OCR--A, OCR-B, and
Univac H-14.
;
If we consider current trenda in automatic reading of documents witti few lines,
the Blank-2 has a number of shortcomings: obsolete basic elements; small
alphabet of cfiaracters (only numbers and four special cfiaracters of the only type
face ROS-A), the impossibil~.ty of reading manuscript cfiaracters, small capa-
bility for documer.t sorting, and so on.
The e::perimental model of tfie Blank-2 character reader was accepted by the inter-
departmental commission. At the present ttme., a new cliaracter reader for small
forms, the Blank-3, has been built for YeS computers (the YeS-6031) on the basis
of this development. The Blank--3 differs from the Blank--2 only in design
(new basic elements of electronic circuits, modernized mechanisms for feeding
and moving documents).
4. The Ruta-701 Character Reader
The domestically produced Ruta-701 character reader, which reads pages of text
with dimensions from 148 X 210 to 297 x 210 milltmeters was developed at the
Special Design Bureau of Computar Technology i~ti Vilnius in 1967-1968, and pro-
duced in a series of 20 untts. The Ruta-701 was successfully used for a
number of years in several organizations, among them tTie Scientific Research
Institute of the USSR Central Statistical Administration, the Computing Center
of Estonian Radio, and the All-Union Geological Institute. The character
reader makes it possible to read numbers, two letters (E, X), and the symbol
in the typewritten nonstylized typeface "Optima" or the same alphabet of
manuscript characters. Guide rectangles with two dots inside are printed typo-
gra}~hically on the page for manuscript characters. The factory rated speed of
the character readers is up to 200 cFiaracters a second or tetween 12 documents
p~r minute (for a page length of 297 millimeters) and 20 documents per minute
(for a length of 148 millimeters). The maximum leng~h of a readable line on a
document is 140 millimetera with a page wfdth of 210 mtllimeters.
The block for feeding and movi:ng documents is designed in the form of a roller
conveyor that works in a start-stop or continuous regi~ne. The capacity of tlie
block's input pocket is 1,000 sheets. The next sfieet is fed to tlie conveyor
t~y a moving arm with vacuu:n suction that operates- or, the principle of tfie auto-
~ matic feed in printing machines. A special photoel~ctric sensor that measures
the transparency of the sheet being fed monitors whetlier the suctions have mis-
' takenly taken two or more sheets at once. When the ne.xt sheet enters the field
of vision of the read block, the document feed block switches to a start-stop
regime. The sheet does not move wtiile the line is read. Then it is moved until
the next line appears in tfie field of vision. An electromagnetic coupling
99
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5
FOR OFFICIAL USE ONLY
included 6etween the drive and the roller of the conveyor is used for start-
~ stop work.
The read block tn the Ruta--701 cfiaracter reader is a'~running fieamK system on a
protection CRT wiiicli_ can work in two modes: search. and line s~canning. To s~m'
plify and speed up tiie search. for tfie. l~ne to fie read a tag is used, a black
rectangZe printed in tfie margin of the page oppos3te tfie start of eacfi line.
The maximum lengt~i. of a readatile ltne ~s 140 m~.l.limeters, wfiicfi is determined
by tfie capab~lities of tfiQ cfiaracter reader's read block.
During scanning the rectangular area of the field of vision corresponding to one
character is expanded into 24 x 20 discrettzation elements. Tfie read signals,
" which are recetved in the line search regime, are used to control the block for
feeding and moving documents. A compensatton FEU [amplifier for photoelectric
cell] directed at the screen of tfie tube is used to reduce the unevenness of
the brightness of the scanning spot on the CRT screen. The feedback signal
from this FEU is fed to the modulator of the CRT. Read.signals from tfie working
FEU directed to the surface of the sheet being scanned are quantized on two
levels ("black"-"white") and transmitted to the recognition block. Specia~
feedback to stabilize the level of "white'~ in the videosignal and adjust the
threshold of quanttzation are introduced to reduce tfie effect of fluctuation in
the reflective capabtlity of the paper and the contrast of printed characters.
The recognition block of the Ruta-701 consists of tmo assemblies. The first
realizes an elementary variation of the correlatian method of recognizing type- .
wri~ten characters. Tlie second assembly performs structural logical recogni-
tion of typewritten and manuscript cfiaracters. The two assemblies can work to-
gether or isolated from one another. The correlation recognition assembly
completes the resemblance between the character being identtfied and standards
- for each character of the alphabet. The criterion of similarity used in t~e
~ character reader is an approximation to the simple iinear criterion
24 20
g(k) ~i v~~ 106 P~vy = 1 I k) -f- (1- vc~) l~~ P(vri ~ 0 ~ k),
which is easy to derive, assumin~ that the components_oi isetheg'lilack~' signala
tistically independent, quantized on two levels (vij
vi~ = 0 is the "white" signal), and appear for character k of the alphabet
with probabilities of p(vi~ =.l~k) a~d p(vi3 = O~~k) = 1-p(vij = l~k). The com-
' ponents of the signal are measured for elements (i, of tY?e 24 x 20 discreti-
- zation network into wh3:ch tfie field of vtsi.on is hroken wfi.en scanning a char-
acter.
The approximation used in the Ruta-701 character reader is based on the fact
that the pYObab.ilities p(vi~ = l~fk) and p(vi � ~Ik) quantize on three levels
which correspond to the interval [0; 0.25], '~0.25; 0.75], [0.75; 1] of proba-
- hility values and set each element (i, in correspondence to a component of
the three-fold standard of tF~s cfiaracter:
1'FOY p\=1l~ � 1 ~ k~ E~~~ ~r25~ ~
e~j~ - 2�For~ p(vr~ = t ~ k) ~ f0,25; 0,75~;
3 for p(vr~ = 1 ~ k) E[~~75; 1).
100
- FO~ OFFICIAL USE OMLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500054442-5
FOR OFF[C[AL USE ONLY
When proba~i.iltties are quantized in thia inanner th~ stmilarity Fietween the
character being identified and tfie standaxd a~ tlie characters is evaluated Fiy
the quantity _
s+ so
_ g (k) _ ~ ~ 4 ~Ur~~ e~~~),
(a~ la~
where the elementary similarity q(y~ , e~?]~ be.tween signal vij and the c~ompon-
ent of the standard witF~ the coded value ei`K ) is given in Table 7 below.
~
Tafile 7.
e~k)
~ (v~~, e~~~ ~
~ I 2 I 3
0 I I I 0,845 0
i
vi;
- 1 I 0 I 0,845 I 1
The character being recognized is assigned to class k� if for this cliaracter ttie
similarity is maximal g(k�) = mak g(k), and two conditi.ons are met:
- 4 ~k�) > Hi~ -
g - ~k g (k) > 0,~
where the set thresholds O1 and 02 can be ad~uated during the work of the char-
acter reader.
If tl~ie condi.tions are not met, a failure to recognize signal is formed and the
unrecognized character is scanned again several times. If after tfiese repeti-
t.ions the character remains unrecognized, line reading is stopped and an image
of the character is pro~ected ~n tfie screen of the CRT monitor on the console
of the control b'ock of the character reader so tfiat the correct answer can be
fed from the console key~oard manually. The method of centering the character
~eing identified, that is, the mutual arrangement of the corresponding components
of the signal and the standard, is not presented tn detail in descriptions of the
Ruta-701 character reader. To ~udge by tfie putiliafied structural diagram, auto-
matic centering ts done before scanntng signals that compute similarity with
standards are transmitted. The criterion for this kind of centering is fiow the
black points of tfie character are arranged relative to the boundaries of tfie
raster. If the centering conditions are not met, the character is scanned again
(the number of possilile repetitions is limited).
101
- FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2407/02/09: CIA-RDP82-00850R000500450042-5
FOR OFFICIAL USE ONLY
' In hardware terms, the correlation recogni.tiQn assemfily is constructed of
several litnary counters. The sums of elementary~similari.ties for points ~n
the middle interval of probab.iliti.es;j0.25; o~75], tfiat is, for points wfi~re
e~~~ = 2, do not depend on tfie signals v~~ and there~ore are written into
their counters tn advance f or eacfi.standard. Tfius, computation of tfie simi-
1~j.ty g(k) amounts to cok~ entional '~1'~ addition in tfie situations v~j = 0 and
e~j 1 or vi~ = 1 and e~~ = 3.
The standards of tfie characters of tfie alphaFiet are shaped on the computer
which has a Ruta-701 read 61ock. The discretized images of the cliaracters of
one class are automatically centered witfiin ttie boundaries of tfie 24 X 20 raster_,
after whicfi the frequencies of "1" for each cell of the raster are determined.
Thsse frequencies serve as estimakes of probab�lities p(vi~ = l~k), and then the
the components of tfiE standard e~~ ) are computed.
- To switch to reading 14 characters of a new type face it is necessary to re-
place the standards of the cfiaracters in the memory of the recognition as-
semUly. The design of the assembly makes it possible to copy the standards
written on punched cards in a few hours. Judging by publications, the problem
of optimizing the selection of atandards for characters has not lieen solved.
In the second recognition assembly of the Rui:a--701, the structural logical as-
sembly, the features which indicate the continuity or discontinuity of the
lines at certain points of the character lieing recognized are determined.
This recognition technique has been patented ~iy character reader developers in
England, the Untted States, France, and elsewhere.
The raster is divided into vertical zones for identification of features. The
boundaries pf the zones coincide with the columns in which the number of inter-
sections of the vertical column of the raster witfi the lines of the sign
- changes. An intersection with a line is noted when the number of sequen-
tially arranged adjacent black points in a column exceeds a set threshold, and
two adjacent lines (that is, ad~acent groups of such black points) are separ-
ated from one another by at least a given number of white points. To avoid
random distortions of the signal, the places of transition from one number of
intersections to the next are recorded only when the new number of inter--
sectj.ons is consistently re~eated a given num~er of times in adjacent columns
of the raster.
The columns of the raster included in tlie identif ied zone are ~oined into one
_ column by logical summtng of the scanning signals (vi = 0 or 1) located in
the same posittons of thes.e columna on the vertical. ~It is considered that
a line in this zone is continuous if in the combtned column between tfie indi-
cated points of interaection there ts no sequence of zero (white~ positions
longer than tlie given tlires.hold. Otherwise the ltne is considered discon-
tinuous.
Two types of features are introduced: for zones. in which tfie number of inter-
sections is two, and for zones in wiiich the number of intersections is tfiree.
For features of the first type continui.ty or discontinuity of the line is
checked between tfie potnts of intersection. For fEatures of tTie second type
102
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5
FOR OFFICIAL USE ONLY
a separate check is made of the nature of tfie line T~etween t~ie ~irst and
second intersection starting from the top do~ in tfi~ raster col~~mn, and
between the second and third.
Features of tFie first type are conventionally- des~gnated iTA.n, 11Bn, Ti , nD
~n n
(n = 1, 2, 3} and can Iiave tfiree values: cont~nuity ~m = 1), discontinuity
(n = 2), and absence o� a zone of tfie featuxe in the cfiaracter (n = 3}.
The zone that joins the raster columns for tfie feature TTAn extends from the
initial column of the raster to the first column with two intersections (in-
cl~sively). Tfie feature TiBn is formed in two cases:
1. if after the first group of columns detected with two
intersections, no more such columns are found until tfie
end af the raster;
2. if after the first group of columns with two intersections
there are columns. with one intersection, and then again a
group witfi two intersections (th.at is, wt.th a sequence of
transfers 2-1-2).
In the first case the zone for feature TTBn is formed by the columns begir.ning
with the last one that has two intersections and going to tFie end of the ras-
ter. In tfie second case the zone includes all the columns with one inter-
section and both enclosing columns witTi.two intersections.
The feature 1TCn is formed if the second case for nBn repeats again, that is,
for all repeated transfers..., 2-1-2,
Finally, the feature TTDn is formed in the last repetition of this situation;
in tfi is case the zone of the feature, as in the first case forTfBn, begins
from the last column with tvo intersections and extends to the end of the
raster.
Features of the second type nEn,1TFn, TTGn, 1TH n(n = 1, 2) assume only two values:
continuity (n = 1) and discontinuity (n = 2). The zone for features ~En~ ~n
begins from the first column of the raster and ends at the first column with
three intersections. The zone for features TTGn, TTHn extends from the last
column with three intersections to the end o~ the raster. The features T1En and
7fGn are characterized by a line located between tli~ first and second inter-
sections in the column from the top; the features ?fF~ andTfHn are identified
by a line between the second and third intersections.
Recognition by the resulting features tak,es place in twro stages. In the first
stage tfie cfiaracters are divi.ded into two groups by a maximum number of i;~ter-
sections of the raster columns with lines of the characters - two or three.
In the second stage the measured features of the character are substi~uted into
logical decision functtons that describe the characters of the alphabet of the
particular group. The numeral 0 is an exaatple of a character of tfie first
group
~na~ � nai � nc3 � nos~ v(nAt � ne2 . ~cs � nDS~
v ~~Ai � nea � nc~ . noi>;
103
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5
FOR OFFICIAL USE ONLY
the numeral 2 is an example of tF~ cha~racter of tfis second group
(lIE2 � nFl � IIG1 � TIH2) (IIE2 � /1F1 � I7G2 � IIH2)
(IIA2 � ITE2 � I1F1 � IIGI � lIH2) V(f1A2 � RE2 x
x 11F1 � IIG2 � IIH2)
Figure 17 ~aelow sh~o~s examples of tFie characters Q and 2 obtained from
scanning real printe.d cfiaracters. For . tfi.e numeral 0 tfie Taoundaries of the
zones are tfie columns 7 and 16: in zone 1-7 feature TTA1 forms, while in zone
16-20 feature nBl forws. For th~ numeral 2 the Fioundaries of tfie zones are
the columns 3, 7, and 15: feature TfA2 forms in tfie zone 1-3,features?fE2 and
QF1 form in the zone 1-7, and features nGl and TTH2 form in the zone 15-20. In
these examples tfie neW number of intersections oras recorded if it repeated at
least two time~, and the feature of disccntinuity was fozmed wfien more than two
zeroes were repeated in order in thE. combined coluum in the sequence of ones be-
tween the intersection points.
The control block of the Ruta-701 controls the interaction of all the other ~
blockc of the character reader. Information that has been read is either
transmitted directly to the Ruta-110 computer to which the reader is connected,
or to punched cards or punched tape (in an autonomous work regime).
Reliability of recognttion. The factory rated frequency of error for manuscript
_ and typewritten characters is 0.01 percent, while the frequency of failure to
recogntze is 0.1 percent. Interdepartmental testing ~f the character reader in-
volved multiple readings of 350 typer~ritten pages done on 20 standard Optima
typewriters witfi aimple cotton inked ribbons (a total of 202,420 characters) and
300 manuscript pages filled in by 14 trained persons at the ~cientific Research
Institutes of the USSR Central Statiatical Administration (a total of 100,800
characters) and produced the following results:
Frequency of Errors:
For Typewritten Characters . . . . . . . . . . . . . . . 0.003%
For Manuscript Charactera . . . . . . . . . . . . . . 0.001%
.
~ Frequency of Failure To Recognize:
For Typewritten Characters . . . . . . . . . . . . . . . 0.015%
For Manuscript Characters . . . . . . . . . . . . . . . 0.004%
Frequency of Skips (During a Search in a Document):
Character . . . . . . . . . . . . . . . . . . . . . . . . 0. 0002y
Lines . . . . . . . . . . . . . . . . . . . . . . . . . 0.0018%
It should fie note.d tfiat careful preparation of personnel and training them to
- fill out manus~cript documents~ correctly played a decis~.ve role here. Wfien manu-
script documents filled out by 13 persons at the All Union Geological Insti-
tute who did not fiave adequate preliminary training were read the sum frequency
of error and nnrecogni.zed cfiaracters rose immediatelp to 2-3 percent.
Comparative analysis of tfiE working characteristtcs of the Ruta-701 character.
reader shows tfiat in 1968, the first year of its production, it was able to
io4
~
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2407/02/09: CIA-RDP82-00850R000500450042-5
FOR OFFIC[AL USE ONLi~
read manuscript characters from pages of text much more cfieaply tfian tfiQn--existing
comparable foreign-made character readers sucfi. as- ti~ Tffi`'I 1288, But for otfier
characteristics suchasalpfiaUet of characters, number o~ type spaces~ and speed tlie
Ruta-701 is inferior to all otfier well-~Cno~m models. Specificallp, tFie laser
OCR-1 character reader built Fiy tfie American Optical Business Macfiines C~~mpany,
wfiich came out in 1971, for roughly tfie. same cost makes it possible to read the
complete alphaliet of typefaces for OCR-~ or OCR-H (numerals, capital letters, and
special symbols), tlie numerals of the typefaces Selfcheck 7B, IB':~i 1428, or
NCR NOF, and manuscript numerals at a ~peed of up to 40Q cfiaracters per second.
The recognition technique used in tfie correlation recognition block of tlie Ruta-
701 does not permit a significant broadening of tTie alphabet of characters. This
is because, first of all, this method fails in principle to consider the inter-
relationships that exist l~etween signals in ad3acent elements of the cfiaracter.
ThiG leads to a low reliaLility of recognition of characters that fiave similar
ourl.ines (for example the Russian "sh" and "shcfi"). The second drawback of this
recognition technique is that the centering of the cfiararters is almost unre-
lated to subsequent recognition. This imposes too-rigid requirements on the
qua~ity of character printing and ultimately also reduces the reliability of
recognition (the more efficient way used in many other correlation character
readers is to center the characters by checking the possitile positions of the
character relative to the standard and identifying the similarity that is maxi-
mum for all of these positions).
The Ruta-701 remains in the history of domestic computer technology as the first
series-produced automatic machine for reading pages of typewritten and manu-
script character ter.t.
S. The Sever-3 Character Reader
The domestically produced Sever-3 c.haracter reader parmits reading printed and
typewritten pages written in various Russian or Latin typefaces. Tfie proto-
type of this character reader was produced in 1971.
The set of recognizable type fac.s is determined by the customer. Each type face
can have up to 150 characters in :Cts alphabet: printed and cursive small and
capital letters, Arabic and Raman numerals, marks of punctuation, and special
_ symbols. The vertical dimensions of the printed characters can vary from 1.35
to 2.5 millimeters (type sizes 8, 9, and 10), while typewritten characters range
from 2.2 to 3.8 millimeters.
The text being read may be on one or b.oth sides of a page and may consist of one,
two, or three columns. with headings, figuras, and formulas.
Befor.e reading a page of text or series of simi~ar pages, instructions on tfie
reading procedure are fed to the control hlock of tfis character reader by means
of the console keytZOard or punched tape. The instructtons indicate the number
of columns in tlie text, the number of pages, the kind of typeface, and the width
of the columns and contain descriptions that make it possible to identify places
not to be read (figures, formulas., and the like) and places for reading to stop
automatically so that essential additional information can lie fed from the
io5
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007142/09: CIA-RDP82-40854R040500050042-5
FOR OFFICIAL USE ONLY
console of the character reader or ~rom puncfied tape. Journals and t"i.ooks are.
torn apart into separate pages b.efore automati~c reading. One column on a slieet
being read sliould not be.wider than 180 mtllimetexs. During readins,~ th~ lines
may be misaligned relative to tfie le~t edge of tlie page, Fiut no more than tlie
height of a small letter for a line length of 180 millimeters. It is also per-
. missitile for cfiaracters to "drop" out of tha line, ~ut not more tfian one-third
of the height of a ~mall letter. .
Each svccessive page to be read is fed to the cfiaracter reader manually. Pages
with text on tioth sl:des are also turned manually. A page that has been fed
goes to a r.oller conveyor with. clamp and feed rollers and a guide rule. TTie
rollers are set at a slight angle to the direction of movement of the sheet so
that the slieet is pressed against the rule. During tlie reading process the
~ rollers are rai:sed by means of an electromagnet. TTiis prevents the next page
from being fed until tfiis~page has been read.
For line reading the sheet is moved in the field of vision of the read block in
steps of 5.1 millimeters. The page automatically returns to the initial posi-
tion for reading a new column or turning the page. When being returned and
brought up to the read zone the sheet is. moved at a speed of 0.3 meters per
second. In the field of vision of the read block the sTieet is held tiy means of
a vacuum tray with two pairs of clamp rollers at tlie beginning,and end of the ~
tray. The lower rollers are the guide rollers and are driven by a step motor.
The passing or the beginning and end of the sha.et in tfie fie.ld of vision of the
read block is monitored by photoelectrtc cells azhicli.consist of an illuminator
and a~hotodiodF. The dimensions of the field of vision of the read block are
50 millimeters in length and 160 millimeters in wtdth.
After the sheet has been read it is e~ected into the receiving pocket by two
rollers ~oined liy an electromagnet. The work of alI the mechanisms and as-
semblies of the block for feeding and moving documents in the Sever-3 character
reader is coordinated in time with the length of reading pages.
Tha read block of the character reader is a T'running heam" system on a CRT with
h~gh resolution. A rectangular point raster with dimensions of 28 X 90 milli-
meters (up to 800 x 2,600 raster points) is formed on the screen of the tulie.
The diameter of the scanning spot on the acreen i.s 70-100 microns. The step of
the raster. may range from 35 to 270 microns. The raster is projected on the
surface of the sheet being read with a 1.8 times magntfication. To reduce the
unevenness of illumination of tfie s.creen of the tube,part of the scanning beam
of lignt is diverted F~y a semi.transluscent mtrror to a compensation FEU wFiose
signal goes to the modulator of tlie tuhe, TTie liglit reflected from tlie surface
of the sheet is recei.ved by aix worktng FEU~s connected tn parallel to the total
- load and tnstalled so that they reduce the unevenness of th~ output scanning sig-
- nal caused by the position of the acanning spot on the sheet. Dynami.c focusi.ng
of the CRT beam on the field of the raster ts u$ed to compensate for changes in
the dimensions~of the scanning spot. Geometrical distortions of the raster are
_ reduced to one percent liy means of special correcting circuits whos�e signals are
_ "mixed'~ into tfie signals of the generators of the CRT scan.
106
FOR OFFICI~L USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5
FOR OFFICIAL USE ONLY
The video signals received ~rom the azorking FEU~s, are quantized on two levels
(black--whtte) . If tfie value of the video signal does not exceed 0.15 of tfie
standard maximum level "lilack~~ tfie signal ~or tfie raste.r point ~s considered
to tie wfiite. ~f the value of the video signal ~s greater tfian 0.70 of tTie
same standard level the signal is considered bl~ck~ For inte.rmediate values of
_ the video signal (lietween 0.15 and Q.7C1 of tIie standard level~ an additional
circular sc~n is engaged with the center at tfie raster point being quantized.
The video s~ignals for the resulting ci.rcular enviromnent are averaged and tfie
average signal is c~~nsidered the tfire~hol.d of quantizatton for this point. The
lens of the read block. can li.e. automattcally moved across the page and occupy
one of three pos~~ib:ie posttions (according to the number of columns in the text)
for scanning tiie entire area of the page.
The recognition block of the Sever-3 character reader uses a two-stage recogni-
tion algorithm. In the first stage tTie structural features are identifted and
measurements are made according to which.the cfiaracter is classified in one
- of several groups of characters of the alphabet (up to a total of 15 such groups)..
_ In the second stage tfie character is finally recognized by computing the minimuli?
Hemming distance between the quantized signals for tfie character and the binary
standards for characters of the alphdhet from the particular group. The algo-
rithm that is used takes account of the variable widtfi of printed cfiaracters
across the line and possible contact between characters in the lin~ because of
printing or reading error. For each auccessive character in the line the sig-
nals are read in a segment whose beginning is determined by tfie result of recog-
nition of the praceding character (that is, by the widtli of the preczding ~
;:ecognized character) or by a clear space appearing in the field of vision after-
recognition of the preceding character. The horizontal dimension of the segment
read (width of a recognized cliar~cter) is selected to correspond to the greatest
width of a standard character or is limite~.i to tlie following clear space if the
distance to it from the beginning of the ~cegment does not exceed the width of
the widest standard.
In the first stage of recognition the standards are centered relative to the
- character being identified. ~ao types of structural features are used for this:
~ primary and auxiliary. The primary feature describes what is called the basic
line of the character. The basic line is tfie first (reading from left to right)
vertical rectangular segment or convex (to the left) arc, with the length of tile
line equal to tfie height of the character. The auxiliary feature characterizes
_ the presence of an outside element, tfiat is, the first line from the left that
is below the line level (as in the letter "p") or above i~ (as in tne letter "d").
The primary feature Tfl assumes three values;
0 if the prtmary line is absent;
r1" 1 if a rectangular segment is found;
2 if a convex (to the left) arc i.s found.
107
FOR OFF[CIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5
FOR OFFICIAL USE ONL~1
. The auxiliary feature re~resents the set o~ two measurements ~2 and ~3, eacfi.
of wfiich assumes tfiree values:
0 if there. is no outside element;
~Z = 1 if an a~iov~tfie-line element ~s fdund;
2 if a beloor~-t~ia-line element is found;
0 if tfiere is no outside element;
~3 = 1 if tfie outside element is located to tfis right;
2 if the outside element is located to the left.
At the same time as the features for the recognized character are being formu-
ldted measurements are made of its width L(tlie wtdth of the segment read),
the horizontal coordinate xl of the mt.ddle of the primary line relati~re to the
beginning of the character~ and tfie horizontal coordinate except two of tfie ~
beginning of the outstde element relattve to that same beginning of the charac-
ter (the widtfi of the segment and coordinates are measured by number of columns
of the scanning raster).
Eacli of the binary atandarc?s for characters of ti.e alphabet e~k~ is character-
ized by the same features and measurements: f/~.k, ;;~,R,./;3,k, x~,k, x2,k, Lk�
The first stage of recognition begina by checking the coincidence of primary fea-
tures and the correspondence between dimensions of the cfiaracter and the standard:
~1= ~~.k;
L ~ Lk�
If this condition is not met for t~e given group of standards, the next group is
selected. If the condition is met, the character is centered relative to each
of the standards in the group. The method of centering depends on tfie values of
the features found. ~
If there is no prfmary line (Tfl 1 k= 0), in the second stage of recogni--
tion the beginning. of the atandard e~~) is cami~ined arith. the bQginning of tfiE
character, after whtch the standard i.e "stretcfi~d out~~ over tfie segment of the
line being identifi.ed. Wt.th "stretcfiing" ths given numher of relative poaitions
of the character and standard on the vertical and horizontal are cfiecked and for
each such position the Hemming dis.tance between the cfiaracter and the standard is
computed. Centering is done to minimize this distance.
If a primary line is found (.T~1 1, k~ 0~, for centering the mi.dpoints of the
primary lines~ the character and standard are matcfi.ed (th~ known fiorizon*_al
coordinates x1 and xl~ k are used for this purpo~e). With this kind of centering
the standard e~k) ~overs a segment Lk wide of the cFiaracter. This se~ment is
"aligned" on tlie horizontal to the primary line of the character ahich hae been
found (the begtnning of thia segment is shifted on tfie horizontal by dk = xi -
xl k relative to the beginning of the character). For the isolated segment of
_ length Lk its own values are determined for the auxiliary featureslT2, Tfg. They
~ 108
FOR OFF[CIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007142/09: CIA-RDP82-40854R040500050042-5
FOR OFFICIAL USE ONLY
~ depend oh ~arZier measurements 1T2~ ~g for tfie outside elzment of the cFiaracter
being recognized:
- a. Tf'2 = 1T3 = 0 if Tj2 = 0 or ~2 ~ 0, but the existing
_ outside element of tfi.e cfiaracter does not fall witfi--
in the isolated segment
xz~-Sk-I-!_,,;
b. rf 2=~Tf 2 and 1Tg = 1 if ~2 # 0 and the. existing out-
stde element of the character falls in tfi,e right fialf
of the isolated segment
~5,; ~ .r, . _ i5~, /2;
c. ~i~2 = iT2 and Rg = 2 if n2 ~ 0 and tfie outside element
falls in the left half of the segment
- xz < Sk -i- L~/2. _
The values of ttie auxiliary features ~2, jj3 gotten for the isolated segment of
ttie character are compared witTi the corresponding values Tf2~ k,~3, k for the
standard e~k~, and if they coincide the recognition block mo~~es to the second
stage of recognition while the beginning of the standard is combined with the
beginning of the isolated segment of the character. To lessen the influence
of random distortions, the comparison in the second stage is done with both
standards in which the primary line is straight and standards in which this
line is curving. In hardware terma, the Hemming distance between the char-
acter and the standard is computed sequentially as the standard is reac: from
internal memory of the recognition block, which includes a general-purpose
_ single-address computer that performs 29 computing and control commands at a
speed of 430,000 ops per second. The internal memory on magnetic cores con-
tains 4,096 45-bit words. In addition, there is a small high-speed memory
consisting of 24 trigger registers (10 and 13 bit positions) and a specialized
assembly for fast computation of the Hemming distance between the character
and the standard. This assembly consists of .four 45-bit trigger registers and
an adder, connected by special logical control circuits, and makes it possible
to compute the Hemming distance immediately for 12 relative positions of the
standard and the character. The character standards for different type faces
_ are stored on magnetic tape and when necassary are copied into internal memory
either automatically or on instructions from the operator.
The Sever-3 character reader has the capability of being instructed for a new
type face, that is, the possibili.ty of conatructing standards and their struc-
tural features and storing this information on magneti.c tape during the process
of using the cfiaracter reader.
Another recognition technique in which the characters are recognized tay parts
may be used to reduce computations and increase the speed of the character
reader and recognition block. The part of the character is composed of several
adjacent scanning columns. The full standards of the characters are constructed
from parts ordered in the form of a tree, as shown in abstract form in Figure 18
below. For each unit of the tree the internal memory of tfie recognition block
l09
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/49: CIA-RDP82-40850R040500054042-5
FOR OFFICIAL USE ONLY
H - - -
_ ~y ~ ~ ~ I . 1
,
' E ~ V V I- r F L~~
6BEr I1VI~KK~-MP U b~
bH ~
W ~
6 B E' /1 M q R/O H,N P 4~l lU b
- Figure 18. Tree to Shape ths Full Standards of Characters of
the Alphabet by Theix Parts in the Sever-3.CTiaracter Reader.
stores a standard for the corresponding part of the character of tfie alphaliet,
the characteristics of this standard (structural features and width), and re-
lations to the preceding and following units of the tree. The standards for
parts of characters are divided into initial, intermediate, and final standards.
The tnitial standards recognize the left part of the character, and the final
standards complete recognition of the character. Tfie code of the corresponding
character of the alphabet is included in the characteristics of tlie final
standards.
The procedure for rec~gnition by parts largely resemblesthe procedure described
earlier for two-stage recognition of characters as a whole, with the exception
of the fact that in each step of recognition we are dealing not with the en-
tire character but with a particular part of it.
Tc find the structural features of the next part of a cfiaracter a definite number
of columns of the scanning raster are analyzed. This numlier is selected based on
the results of the preceding step of recognition or by a clear space in the field
of vision.
After the left edge of the next character in tfie. line is found, tlie group of
initial standarda u~hose features coinct.de with t:~e features of the cfiaracter is
. selected. A mi*~imum of 24 Hemming distances obtatned wTien this standard is
"stretched" over the character are computed for each standard. Tlie decision is
made according to tfi e minimum resulting distance. Then the next group of stand-
ards necessary to recognize the next intermediate part of the cfiaracter is se-
lected on the tree and the process continues until an answer is obtained for the
final standard. At this point recognition is complete.
110
FOR OFFIC[AL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/49: CIA-RDP82-40850R040500054042-5
FOR OFFICIAL USE ONLY
If the difference between tfie. minimum Hemming distance and tfie one ef ths 24
computed distances that is closest to it is less ttian an assigned thresfiold,
both answers are considered permi.ssifile and botfi.groups-of standards indicated
by transfers on ttie tree are sele~cted for recogni.tion of the next part of tfi~
charact~r. If ttie least distance e~xceeds the given threshold or the structural
features of the nearest standard and t~ie recognized parts of tfie character do
not coincide, the answer obtained is rejected and the standard is removed from
- consideration. If the answers for all standards in tfie group selected are re-
- jected, the recognition tilock returns to the preceding step of recognition on
the tree. Thi~ step is repeated for all standards of tTiE~. preceding group with
the exception of ttiose for wfiich. rec.ognition answers were received tfie first
time.
If a double answer is received for the final standard or if when returning on
the tree tfie process again ar~ives at the stage of preliminary recognition but
all permissitile variations of features of the. init~al part of the cfiaracters
have been exhausted, the line segment being recognized is scanned again.
After a given numtaer of repeated scans a signal of fa~lure to recognize is
- formed, the unrecognized character is pro~ected on the CRT screen located on
the control console of the character reader, and the character can be put in
manually using the keyboard of the console.
Control of tfie Sever-3 character reader is decentralized. Autonomous control
units are instal].ed:
l. in the read block - to scan any area of the field of
vision with horizontal or vertical point rasters;
2, in the recognition block - to organize the computation
of the Hemming distance between standards and tf~e char-
acter being recogni.zed;
3. in data input/outF~ut units when working with magnetic
tape or punched tape, and so on.
General control of the character reader which coordinates the work of all its
blocks and assemblies is accomplished by the program control unit of the com-
puter included in the composition of the character reader. The combination of
hardware and software control ma.kes it poasible to realize complex algorithms
for recognition and editing o� information read and makes it easier to adapt
the character reader quicli.ly to solving new prolalems.
The information that has heen read is transmitted to the external computer to
which the cfiaracter reader is conne.cted or outputted to punched tape.
Reliability of recognition. The ~actory rattng for frequency of errors for the
Sever-3 character reader is one percent for an alphanumeric text and 0.01 per-
cent for numeric typewritten text. These indicators should be achieved with
fluctuations in the thickness of character lines from 0.1 to 0.6 millimeters
in printed text and from 0.2 to 0.56 millimeters in typewritten text with a line
contrast of at least 30 percent.
111
FUR OFFICIAG USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2407102/09: CIA-RDP82-00850R000500450042-5
FOR OFFICIAL USE ONLY
' An experimental model of the cfiaracter r.eader arae tested on .typewritten pages
and printed sheets tn dif~erent foxmats a~i:th di~~erent numbers .f text columns~
(1-3) written on one and botfi ~tdes of the paper,.
The average reading speed (:converted per ~fiset~ aras 64 cfiaracters per s~econd for
printed text, 38 ch~.racters per second for alpTianumeric typewritten text, and
56 characters per second for numeri.c typewritten text. Tfie average ttme re-
quired to change tfie set oi standards wfien swi:tching to reading a new type face
did not average more thar. 20 seconds. During tTie reading of 40 printed book
and journal pages (106,000 cfiaracters~ the frequency of error was 0.6-0.7 per-
cent witfi a frequency of failure to recogntze of 0.1--0.2 percent. For alpha-~
numeric typewritten text (13 pages, 21,000 cfiaracters) tfie frequency of error
was 0.9 percent (witfi a frequency of fai~ure to recognize of 0.04 gercent). For
typewritten numbers (64 pages, 119,000 characters~, the frequencq of error was
= 0.02. percent, and the frequency of fatlure to recognize was 0.002 percent. ,
Comparative analysis of the working characteristics of the Sever-3,character
reader is made dtfficult by the fact that tlie well-known contemporary models are
rarely designed for reading full aets of printing type faces.
Nonetheless, it should be noted that the speed of the Sever-3 machine is about
an order lower than tfie speed of foreign made cfiaracter readers for reading
pages of text and general-purpose character readers of equivalent cost. Tfie
prototype of the Sever-3 character reader has been used successfully for severa].
years to printed texts to an information service system.
COPYRIGHT: Izdatel'stvo "Tekhnika", 1980
11,176
CSO: 1863/47
112
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5
FOR OFFICIAL USE ONLY

~
DESIGN, OPERATION OF CHARS CHARACTER READER
Kiev OPTICHESKIYE CHITAYUSHCHIYE AVTOMATY in Rus~sian 1980 (signed to press
22 Nov 79) pp 146-190 .
[Chapter 5 of liook "Optical Character Readers", by Doctor of Technical Sciences
~ Vladimir Antonov3:ch Kovalevskiy, ~and3date of Techni~al Sctences Georgip LFvovich
Gimel'farb and Candidate of Tecfintcal Sciences Anatolty Fedorovich Voziyanov,
Izdatel~stvo "Tek.hnika", 4,000 copies, 2C8 pages]
[Excerpt] Chapter 5. The Design of the ChARS Character Reader
- 1. Features of the ChARS character reader
The Institute of Cybernetics of the Ukrainian SSR Academy of Sciences fias de-
veloped several modif ications of character readers carrying tfie name CfiARS
(character reader with shift register) and designed to read multiline typewritten
documents. In 1965 a model of a character reader for 10 numerals using the corre-
lation method of recognition j16] witfi.grid dimensions of 8 x 11 was built. The
model worked at a low speed of 25 characters per second. Tfie reading unit used
a"running ~eam" type of scan on a cathode ray tube. The tirightness of the ele-
ments of the image was quantized on four levels. The probability of error was
not more than 0.01 percent. Based on this model a prototype character reader for
the full alphabet of the Optima typewriter was built in 1969.[1, 2]. The grid in
this reader was slightly enlarged (10 x 18). But unlike the earlier modification,
a column of photodiodes was used to acan the image. The masks for this machine
were constructed with due regard for the spatial discretization of the image
[18]. Tfie machine has a read mechanism that allows automatic scanning of an en-
tire package of documents.
Tests of the prototype showed that witiz a 10 x 18 grid a probability of error
lower than 0.1 percent could not Fie achieved~ even reducing tfie alphafiet to 48
characters instead of 53 (the Russian letters B, V, I, M, and ~T were eliminated).
The average speed of the reader was 24Q characters per second with a maximum
speed of 500 characters per second.
Because the results obtained with this prototype were unsatisfactory~, the 1969
ChARS was modernized. The new 1972 modification appeared~ and is the one
described in detatl lielow.
113
FOR OFFIC[AL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/49: CIA-RDP82-40850R040500054042-5
FOR OFF[CiAL USE ONLY
The most important cliaracteristic of a character reader is the requirem~nt that
it imposes on tfie document heing read. Ttie 1972 ChARS was desi.gned to read
typewritten pages printed on white.wri.ting paper in a format of 210 x 297 milli--
meters by an Optima typewriter. A filack cotton inked rihfion of ~medium thickness
- for typewriters and calculators sfiould ~e used for printing.
The aipfiabet of the machtne contain~ 53 cfiaracters, includfing 10 numerals (0-9),
the 29 capital letters of tfie Russian alphaFiet, ~tve signs of aritfimetic opera-
tions (�--x/=), tfiree marks of punctuation five service cfiaracters
and a "space" character.
Up to 30 lines can be printed on a page, ~uith up to 62 characters on eacfi line,
in otfier words a total of up to 1,86Q characters per document. Tfie lines are
printed across the short side of .the slieet. At tlie beginning of each line sub-
ject to reading a special character is put at a certain point in the ~argin of
the page.. This is a marker in tlie form of a fiorizontal line formed by four~
consecutively written signs The marker should begin not more than 13
millimeters from the left edge of the. slieet and end at a distance of 20--25 milli-
meters from it. The machine does not read lines that are not marked by tlie
marker. This makes it possilile to put additional information on the document
sucli as fieadings, notes, explanations, and tlie litce, whicTi can be printed ar~
bitrarily or entered by hand. Tre minimum permissible distance between lines
depends on the permissible misalignment of lines relative to the edge of tfie
sheet and should tie at least 8.5 millimeters (two intervals) if a line mis-
alignment of up to � milltmeters is permitted on the length of a line. Spaces
between words (letters) in the ltne with a maximum misalignment do not exceed 50
millimeters). The distance from the top edge of the sfieet to the top marker
should be at least 25 millimeters, and from the bottom edge of the sheet to the
lowest marker at least 18 mtllimeters. The thickness of tfie lines of the charac-
te~s is in the range 0.2-0.5 millimeters. The coefficient of reflection of a
character line should correspond to the condition [(P~ - Pn)/px] ~ 0.5
where P~ is tlie reflection factor of the liackgrcund paper.
There should not be tears, contrasting fitaers and impregnations, or very dirty
places on the sheets of paper.
At the start of developmen~ the goal was set of achieving the following indi-
cators of reliabiltty of recognition given compliance witli the above requirements
for printing quali.ty:
Type of Array Frequency, %
Fatlure To
Errors Recognize
- Numerical . . . . . . . . . . . 0.005 0.01
Alphabetic . . . . . . . . . . O.QS 0.1
Mixed (all 53 characters) 0.04 0.07
za.4
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/49: CIA-RDP82-40850R040500054042-5
FOR OFFICIAL USE ONLY
- The average reading speed counting time.losses searcfiing for lines and replacing
documents is 170 cliaracters per s~econd. The CfiARS as well as tfie. stand equip--
ment and measuring instruments are arranged ~n an area of 25 square-meter~~
The ChARS receivES power from a tfiree~fiase 224/380.volt current at 50 fiertz.
Its power usage is not more than 4.5 kilowatt-amperes.
The character reader set includes the following uni.ts:
a. document reader, wfiicli.consiats of a filock to feed and move
documents witfi.a control assemfily and read fiead;
b. recognicion unit, wfiich.includes the recognition block
with its control assemTily;
- c. .buffer memory unit, which coordinates tFie work of the
cfiaracter reader wi.tfi the computer and puncTi and insures
manual encoding of unrecognized cTiaracters.
As can be seen from these technical specifications~ the ChAR3.differs from
other domestically produced character readers sucli as tfie:Ruta-701 and Blank-2
considered in Chapter 4 tn the followtng ways. .
- Ordinary writing paper and ordinary multiple-use inked ribbon are used to prepare ~
documents. The documents are typed on typewriters with standard nonstylized type-
face and a fairly large alphabet (53 characters). No special requirements are
made for printing quality: all that is necessary is to avoid the use of an
inked ribbon that is too thick or too worn out.
Requirements for arrangement of the information on the document come down to
maintaining a minimum distance between lines and the need to identify lines with
markers.
In the first modifications of the ChARS a correlation recognition algorithm was
used which had been derived from the method of permissible conversion (see sub-
chapter 2.3) on the basis of a mathematical model of a set of images taking intu
account opttcal conversions, image carryover in tFie field of vision, and noise.
The 1965 model of the ChARS was buflt following this method and demonstrated high
recognition reliatiility for an alpha~et consisting only of numl~ers. Clearly
printed images of the numbers were chosen as initial star_dards for this model.
The same method was the tiasis for tbe next, 1969 modi.ficati.on of the ChARS. The
dimensions of the grid of this modi~ ication of the machine (10 X 18~ were chosen
on the basis of positive results obtained from tTi~ earlier model.
Because the image in this machine is: xepresented on a comparatively crude grid,
a methodology was developed to construct masks and took account of distortions
occurring during discretization. Tfiis methDdology ts described in j8], wliich
demonstrates that tfie set of images occurring during discretization of a certain
fixed image s~hifted relative to tfie raster witf~in tfie limits of a step of
discretization is a convex polyliedron whose apexes are the so-called nodal
images. For example, four nodal tmages distingutshsd bp a shift of exactly one
u5
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/49: CIA-RDP82-40850R040500054042-5
- FOR OFFICIAL USE ONLY
element of the grid are sufficient to descrifle. zacfi of tfiQ classe~ su~fi as tfie
Russian letters N, P, and Ts, wfiereas tTia descr~tption of ths set of images fox
letters~ s-scfi as~ tfie Russian letters H~ V, and tFiE: snft sfign is a polpfiiedron ori.tfi.
nine apexes~. Optical convers3ons o~ere also taken into account in constructing
tfiese sets.
The metfiodology involved constructing the difference masks necessary to dts--
tinguish similar classes, and isolated masks to cfiaracterize a class taken
separately. To construct a difference mask the problem of dividing two convex
polyhedrons tiy a fiyperplane whi.ch maximizes tfie dtstance ta tlie polpfiedrons ~aas
solved. TTie isolated masic was selected so that the leas~t correlation witfi the
images belonging to the particular convex polyhedxon was maximized. It sfiould
be noted that in order to simplifp the technical~ realizatian of the algorithm,
convex polyhedrons cliaracterizing tfie set of images were constructed only for
one fixed thickness of the lines of the character. This d~cision was made on
the assumption that the convex sets constructed would also include images with
line thicknesses differing from that for wfiich tfiey were constructed. It would
evidently have been better to construct masks for different thicknesses and
during the recognition process search for the maximum similarity not only Y,y
classes and different positions of the image, but also by different tliicknesses.
As pointed out above, during testing of the 1969 CIiAP,S unsatisfactory values
were recei~ed for the parameters of recognition re;.iability. Tfie basic reasons
for this were the following.
The masks in the machtne were constructed for one fixed thickness of lines.
But with actual printing quality line thickness varies in a wide range.
The machine did not have adequate resolution capability. Even if the masks taking
into account discrettzation had been constructed for the full range of variation
of thicknesses, the raster adopted for the mach~ne would still have been a crude
one. The di:mensions of elementa of the grid (0.27 X 0.30 millimeters) exceeded
the minimum thickness of lines of a character by 50 percent. This led to such
serious distortions of images with thin lines that the images of similar classes
_ could hardly be distinguished from one another after discretization.
So-called crude di~placements of alien (competing~ mask:s were not taken into ac-
count in constructing tlie masks. As researcfi showed, because of the distor-
tions caused by discretization, the maximum stmilari.ty in the correct and an
alien mask may be attained for different crude dtsplacements, that is displace-
ments corresponding to carryover of tmages for an entire cell of tiie grid.
In this case it may prove that similarity with an alien mask in an adjacent dis-
placement is greater than ~imtlarity arith th.e correct mask. Because the answer
is taken for maximum stmilarity, in this ~ase an error will lie recorded. There-
fore, wfien constructing the masks it is necessary to consider tfie different
positions of tlie alien mask. The positions of masks wfitch differ by a displace-
ment of one cell are especially critt.cal.
The algorttfim used tn the machine to divide lines into characters had low noise
resistance because it was based on the asaumpti.on tTiat false maximums of simi-
larity at tfie intersectiona of two characters do not occur. Given real printing
quality using conventtonal type faces this assumption is not ~ustified.
116
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2407/02/09: CIA-RDP82-00850R000500450042-5
FOR OFFICIAL USE ONLY
The prototype was developed on the. bas.ia of testing a model uzhicFi fiad dif-
ferent parameters, without preltminary comput~r znodeling of tfie new modifica-~
tion. Before tliis model was built the correlation ~me.ths~d of recognit~on ~ras
modeled and s:tudtes were made of its ef~ectiveness for an alpiiabet of 10
numerals witti a grid of 8 X 16 cells and brigfi.tness quan~ization on 16 levels.
In addition, to improve the prectsion of centering fialf--step sTiifts~ of tfie
image were done during modeling, tliat is, eh~fts equal to Tialf a discretization
step. When tlie paramerers of the machine were ~expanded alpfiabet,
quantization on four Ievels, a6sence of half--step sliifts, and tfie like), it was
also necessary to mak2 a detailed study of tfie recognition algorithm.
The metfiod used to construct the mask does not permit an estimate of the proba--
bility of recognition error for any given masks. Althougli tfie masks constructed
on this methodology insure mtnimum protiabi.l.ity of error give~i the assumptions
made, tlie value of this minimum was unknown.
The sfiortcomings of earlier modifications were taken into account in develop-
ment of the 1972 ChAR~ and the following improvements were made. The grid was
doubled in size by reducing the discretization step on the horizontal (20 X 181
with a cell size of 0.135 x 0.31 millimeters. But then it was necessary, for
purposes of equipment economy, to reject the four levels of quantization and,
consequently, use of the correlation metfiod of recognition in the form con-
sidered in Cfiapter 2. To avoid a situation where the results of quantization
of the image on two levels would depend strongly on changes in the contrast
and darkening of the background, a quantization procedure witli a variafi le
threshold depending on tfie tirightness of adjacent cells was employed.
A new methodology of constructing masks was developed, based on representing
a set of images ditinguished by thickness of lines and position relative to
the raster in the form of multidimensional distril5utions. Tlie magnitude of
the independent variable (argument) of proTaability of error. was used as the
optimization criterion.
COPYRIGHT Izdatel'stvo "Tekhnika", 1980
11,176
CSO: 1863/47
- 117
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/49: CIA-RDP82-40850R040500054042-5
FOR OFF[CIAL USE ONLY
EQUIPMENT COSTS~, C,TAGES FOR CHARACTER READERS
Kiev OPTICHESKIYE CHITAYUSHCHIYE AVTOMATY in Russian 1980 (signed to press
22 Nov ?9) p 191-205
[Appendix from Hook "Optical Character Readers", by Doctor of Technical Sci-
ences Vladimir Antonovich Kovalevskiy, Candidate of Teefinical Sciences Georgiy
L'vovich Gimel~farb and ~andtdate of Teclinical Sciences Anatoliy Fedorovicfi
Voziyanov, Izdatel `'stvo "Tel~hntka", 4, OOQ copies, 208 pages]
Table P1 Nominal Speed
of Text Coding,
Characters
Unit Cost, rubles per Second _
YeS 9015 Punched Card Data Entry Unit 8,000 5
YeS 9018 Punched Card Verifier 7,000 5
YeS 9001 Magnetic Tape Data Preparation Unit 20,000 5
SPD~-9000~Magnetic Tape Data Preparation 120,000 Plus 5
System with Keyboard Consoles cost of All
Consoles
(5,000 rubles
apiece)
ChARS Character Reader 140,000 150
Table P2
Service Personnel Monthly/Annual Wages, rubles
Operator l0A/1,200
Senior Operator 120/1,440
DtspatcTier-Operator 150/1,800
Technician 90/1,080
Mechanic 140/1,680
Engineer 12Q/1,440
Group Leader 160/1,920
COPYRIGHT: Izdatel~stvo "Tekhnika'~, 1980
11,176 ~
CSO: 1863/47 ~
118
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5
FOR OFFICIAL USE ONLY
CONF~RENCES
INTERNATTONAL 'DTAGNOSTTC ~'ACTLTTTES FOR DTGTTAL SYSTEMS~ CONFERENCE
Kiev ELEKTRONNOYE MODELIROVANTYE in Russian Vol 4, 'Vo l, Jan-Feb 82 pp 105-107
[Article by V.A. Gulyayev]
[Text] The international "Diagnostic Facilities for Digital Systems" conference,
organized by the Czech committee of the Electrical Engi~learing Sc~entific and
Engineering Society, the Czechoslovak IMEKO [expansion unknown] national committee
= and the Diagnostics in Electronics Central Trade Group, was held on 28-30 September
1981 in Brno (CSSR).
Diagnostic facilities for digital systems represent part of the program for improv-
ing the quality and reliability of digital electronics products. Questions relat-
ing to the area of the testing and reliability of digital systems from the view-
point of their design, production and servicing were discussed in individual sec-
tions. Fifty-one reports from 14 countries, including the USA, France, 1~RG, Italy,
GDR, Polish People's Republic and Hung~rian People's Republic, were presented at
the conference.
The following sections were at work at the conference: 1) designing of systems
insensitive to failures; 2) simulation and evaluation of high-reliability systems;
3) software reliability; 4) test generation; 5) teat simulation; 6) designing for
easy testability; 7) self-testing systems; 8) system diagnosis; 9) self-verifying
check circuits; 10) diagnostic instruments and systems.
At the morning plenary session Professor (E. Dzh. Mak. Klaski) (USA) delivered the
paper "Survey of Methods of Designing for Testability," in which difficulties
originating in testing were pointed out--the inadequacy of models, the 1ow-per-
centage of the disclosure of malfunctions, great input of time, complexity and
the considerable amount of documentation. An alternative is guaranteeing test-
ability at the design atage. Three approaches to guaranteeing testability are
discussed in the paper. Among them is the technique of installing and observing
the state of circuits. A rather general approach is discussed, based on dividing
a circuit into a combination circuit and one with a memory. A great number of
specific examples are presented. A1so discussed in the report are question~ re-
lating to guaranteeing the testability of programmed logic arrays and questions
relating to the generation of j.nput sets for testing by means of shift registers
with feedback.
-J
1.19
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/Q2/09: CIA-RDP82-00850R000500050042-5
FOR OFFICIAL USE ONLY
In a paper by Belorussian SSR Acade~my o~ Sciences Corresponding ~Iember A.D.
Zakrevskiy (USSR), Diagnosis o~ Malfunct3ons o~ Regulax Technologica~. Structures,"
theoretical and practical questions were discussed x~elating to testing of a new
class of large-scale integrated circuits--programmed lvgic arrays (PLM's). A
distinctive feature of these structures is the presence of a special class of
malfunctions, viz., of modifying functtons performed by a combination circuit.
A tabular representat~,on of functfi.ons implemented with programmed logic arrays
is used for the purpose of analyzing c3rcuits and constructing tests.
At the evening session a paper by (I. Siray) (Hungarian People's Republic) was
heard, "Survey of Methods of Generating Fun..tional Level Tests," in which existing
approaches to constructing tests at the functional level were discussed, including
methods developed on the basis of the modular approach, Boolean differences, and
development of a certain d-algorithm. The speaker presented an algorithm developed
by him and based on the development of a d-algorithm and which takes into account
repeated malfunctions, and he also compared two methods of simulating units with
malfunctions--the parallel and concurrent.
In his paper "Evaluating the Reliability of Computing Systems" (Zh.-K. Lapri)
(Toulouse, France) presented information on the percentage of errors of various
types For individual design stages and on the life of elements of compsting sys-
tems.
A classification was given of models for describing the state of systems and their
behavior. The possibilities of using methods of Markov and semi-Ma.rkov processes
and of simulation for the purpose of studying the reliability of computing systems
were discussed. Various methods of simplifying analysis--consolidation of states
and the like--were discussed. Information on the use of these methods for estimat-
ing the reliability of severa.l classes of systems was presented.
In a paper titled "Methods of Reducing Diagnostic Data" V.P. Chipulis (USSR)
presented a method of compressing source data obtained by simulation an a digital
computer, by using so-called masks. Two variants of reducing the amount of source
data were suggested: minimizing the amount of information (defined in a 0, 1, x
alphabet) with the required search resolution, and reducing the amount of informa-
tion with an acceptable reduction in degree of resolution.
The following papers were presented in section 1: (M. Kateyama) (Japan), "High-
- Reliability Digital Signal Processor with Insensitivity to Failures;" (P. Cardini)
(Italy), "Mechanisms for Protection Against Failures in the (MyuTIM) Multipro-
- cessor" and ('F. Biardi) (Ttaly), "Protection and Limitation of Errors in an In-
formation Transfer Medium: the Nucleus of the (MyuTT'M) System;" and (L. Simoncini)
(Italy), "Integrated Procedure for Designing Systems Insensitive to ~ailures."
The following papers were discussed in section 2: (P. Belardi) (Italy), "Compari-
son of Four Modular Redundant Circuits in a Computing System;" (L. Saloriutta)
(Finland), "Hierarchical Mode1 of the Productivity of a Multimicroprocessor In-
sensitive to Failures;" (P. Kaspi) (~'rance), "Methods o~ Evaluating Systems Re-
sistant to Failures;" A. Babak (Polish People's Republic), "Functional Reliability
Mode1 of the Hard Copy Log of a Computer Network;" and (ri. Tutoveanu) (Rumania),
"Improvement o� the Readout of Tnformation from a Magnetic Disk Storage by Means
of Redundancy."
120
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2407/02/09: CIA-RDP82-00850R000500450042-5
FOR OFFICIAL USE ONLY
The following papexs wexe discussed ~,n aectian 3: (A. Lev~na'k~.) (Polish People's
Republic), "Method o� ~reating Arithmetic lriicroprograms;" (L. Saloriutta) (~inland),
"Evaluation b;r Means oP Petri Nets of a Sy~nnnetric Operating System Tnsezs3.tive t~
Failures;" and B. Kore1 (Polish People's Republic), "Stream of Data in Testing
of Programs."
The following reports were read in section 4: (Sh. Vaxsegi) (Hungarian People's
Republic), "Functional d-Algorith~cn;" (Y'a. Rada) (CSSR), "Adaptive Pseudorandom
Test Generator ~or Logic Circuits;" and V'.A. Gulyayeva (USSR), "Adaptive Method
of Simulating Digital Equipment with Malfunctions."
The following papers were presented in section 5: (P. Marchal) (France), "Simula-
tion of the Time for Detecting Ma.lfunctions of the Integrated Circuit of a Pro-
cessor;" (D. Raynerta) (GDR), "Two-Way Simulation--a New Approach to Test Genera-
tion;" (K. Valchak) (Polish People's Republic), "Deductive Simulation of Repeated
Malfunctions in Logic Circuits;" (Y'a. Piotrowski) (Polish People's Republic),
"Graphic Language for a Functional Description of Zogic Circuits." Tn section 6
a paper by (V. Koy) (FRG) was read, "Distribution of Test Points for Simple Genera-
tion of Tests."
Two papers were read in section 7: (G. Kravchik) (Polish People's Republic),
"Algorithm for Steady Self-Testing of a Multiprocessor System" and (E. Mikhta)
(Polish People's Republic), "Self-Testing Microprocessor Devices."
Section 8 was represented by the following papers: (Ya. Zeleny) (CSSR), "LIrIDS
[expansion unknown]--a Microdiagnostic System for Local3.zing Malftinctions;"
(A. Plugachek) (CSSR) ,"St.ructure and Properties of Tests of the 1'~T~IDS System;"
(P. Zhak) (CSSR), "Testing the Tesla TsNTs [expansion unknown] SysLem;" and
I. Smishek, "Use of Diagnostics in Graphic Systems."
The following papers were read in section 9: (V. Rabira) (CSSR), "Designing a
Self-Verifying Check Circuit for 1-of-n Code;" (M. Kotochova) (CSSR), "Designing
Self-Veri~ying Check Circuits for Some 1-of-n Codes;" and (P. Go1an)(CSSR), "New
Se.lf-Verifying Check Circuit for 1-of-3 Code."
The following papers were presented in section 10: (P. Shtolle) (CSSR), "The
ADT/ZKD [expansion unknownJ Automatic Testing System;" (V. Shkvor) (CSSR), "Local-
- ization of Malfunctions by Means of the ADT/ZKD System;" (Ya. Revitski) (Hungarian
_ People`s Republic), "Classification of Integrated Circuits in TEZ's [Thermocouple
Probes];" (B. Chaban) (Hungarian People's Republic), "Strategy for Testing TEZ's
and Facilities for Designing Technical Equipment;" (K. Shviyeru) (Polish People's
Republic), "The 'ANISS-20' Logical Analyzer;" and ~A. Khlavichka) (Polish People's
Republic), "Improving the R~liability of Error Detection by the Analysis of
Control Characters."
During the conference opinons were exchanged on various questions relating to
the design of diagnostic facilities. The organiz~ng committee has published a
collection of the conference's proceedings.
COPYRIGHT: Tzdatel'stvo "Naukova dumka", "Elektronnoye mod~l~.rovaniye", 1502
8831
CSO: 1863/95
121
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2407/02/09: CIA-RDP82-00850R000500450042-5
FOR OFFICIAL USE ONLY
OTIMIZATTON QF CONTRQL S`~'S~E1~S
Vilnius OPTI'MATTZA~SIYA STS~EM KONTROLX`A ir. Russiaa 1981 (afigne~ ro press 13 Jan 81)
pp 4, 240-241
[Foreword and table of contents ~rom collection "Optimization of Control Systems",
edited by Professor k1. Bikyalis (B~kelis), Lithuanian SSR Ministry of Higher and
Specialized Secondary Education, 500 copies, 247 pages]
[Text] Foreword
The "Applicati~n of Probability Theory and Ma.thematical Statistics" republic con-
~ ference dedicated to the 400th anniversary of Vilnius University was held at the
start of 1979 at the mathematics department of Vilnius State University. The con-
ference made it possible to make an evaluation of the state of applications of
mathematical methods in the republic's enterprises, to exchange know-how and to
hold a discussion on questions relating to problems, as well as to determine the
most promising trends in the development of inethods and applications of probability
theory and mathematical statistics, taking into account the opportunities for the ~
extensive employment of computers. The active participation of representatives of
enterprises and design bureaus, who delivered half of the papers, made it possible
to ascertain the requfrements of industry in cooperation with mathematicians and
to specify the framework of this cooperation. The ma~ority of the papers 'elivered
at the conference are printed in this collection.
CONTENTS Page
Foreword 4
Vaitkus, P. and Kruopis, J. "Mathematical Methods of Optimizing Quality
Control Systems" 5
Kruopis, J. "Minimization of Quality Control System Loss Functions" 12
Barcevicius, R. "Inte111gent Selection of Product Output Control" 28
Kepezinskas, A. "Structure of an Effective Plan for Selective Output
Quality Control of Consumer Electronics Products" 35
Kalnius, R. "Optimum Ranges of Acceptance of Products in a Multistage
Control System" 44
Adomenas, V. "~i~ding the Limits of Channel Selector ~ailure Rate ~unction
- Parameters" 55
Adomenas, V. "Methods o~ Choosing a P1an for Testing the Reliahi.lity o~
Channel Selector~" 61
122
FOR OFFICIAL US~ ONLY ~
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5
FOR OFFICIAL USE ONLY
Kruopis, J. and Ne~an~.s, P.-A. "Stat~;at~:ca~ Method o~ Contxol.l~n$ the Pxe~
cision o~ Technol.ogica7. N(easuxements" 76
Bagdonavicius, V. "~valuation of Zongevity ~n Fast ~ests~' 89
_ Vaitkus, P. "Opt~~um Aesign of Quality Control ~h$rts`~ 98
Purtulis, R. "Use af 'Methods of Mathematical Statistics ~or Controlli.ng
a Technolo~ical Pxocess" 107
Gaigalas, 3., Kaminstci~, V. and Kukuraitis, M. "Predict~on o~ the ~mis~ion
Current of B1ack-~and Whi:te Picture Tubes~' 110
Elkoninas, V.T. and Baxauskas, S. '~Study o~ Empixica7. Models o~ ~echnolo-
gical Systews ~'~ox the Purpose o~ Optiutizatton" 114
Mi"seikis, F. "'Mathematical Experimental Aesign Method~s " 123
Miseikis, F. "Sinrulation Methods fn Deaign Problems" 127
Sabaliauskas, R. "Models of Queueing in Production" 130
Baublys, A. "Stochastic Models in an Automated Control Spstesn for a
- Machinebuilding Enterprise" 134
Skakauskas, V., Maknys, M. and Jasinavicius, R. "Mathematical Analysis of
Narrowly Localized Fields of Magnetic Heads" 138
Kruopis, J. and Lekevicius, R. "Estimating Parameters of Ce11 Cycles in
Asynchronous Ce11 Populations" 142
Bagdonavi"cius, V., Pauliukonis, A. and Jusas, J. "Analysis o~ Di~fusion
Effects of an Tmmobilized Polyenzyme System" 156
Lukoseviciene, 0. "Mathematical Model of Dynamics of a Number of Highway
Accidents" 163
Gar;;va, J. "Problem of Forming a Production Program for the Republic
Ministry of Construction" 181
Gorbunov, N.I. "Bayes Statistical Tnference Procedures When It Is Possible
to Refuse to Make a Decision" 186
Bistrickas, V. "Method of Studying Discrete Control Processes" 201
- Skakauskas, V. "Motion of a Flexible Nonstretchable Thread Along a Plane
Curve" 205
Rupsys, P. "The Basis Maximum Principle" 216
Jusas, J., Bagdonavicius, V. and Dienys, G. "Method of Determining Reaction
Rates of Coupled Chemical Reactions" ~20
Bikelis, A. and Kiminius, J. "Probability Distributions Concentrated in
Several Lattices" 22~
Milasevicius, J. "Asymptotic Expansions for Probabilities of Great Devia-
tions in the Poisson Approximation Case" 240
_ COPYRIGHT: Ministerstvo vysshego i srednego spetsial'nogo obrazovaniya
Litovskoy SSR, 1981
8831
CSO: 1863~123
_ ~~3
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2407/02/09: CIA-RDP82-00850R000500450042-5
FOR OFFICIAL USE ONLY
PUBLICATIONS
UDC 551.509.313
METHODS OF rIUMERICAL SxMULATTON OF ATMOS~HER~^v PROC$SSErS
Leningrad METODY CHTSLENNOGO MODELTROVANTYA ATMOSFERNYI~i PROTSESSOy in Russian
1981 (signed to press 20 Aug 81) pp 2-10
[Annotation, foreword and introduction from book "Methods of Numerical Simulation
of Atmospheric Processes"y by Vladim3r Viktorovich Penenko, Gidrometeoizdat, 1350
copies, 352 pages] ~
[Text] This book contains a systematic presentation of inethods of numerical simu-
lation of the hydrothermodynamics of atmospheric processes. Ma~or attention is
devoted to the structural algorithmic aspects of this problem. Discrete models
are constructed on the bas3s of the variational principle and the splitting-up
method. Finitp-difference and spectral-difference models are diacussed, as well
as methods of studying the sensittvity of discrete m6dele to variations in input
data and of identifying the parameters of models, assimilating and adjusting mea- .
surement data, etc.
This book is intended for specialists in the fields of computing and applied mathe-
matics, meteorology and geophyaics, as well as for students at universities and
hydrometeorological WZ's specializing in methods ~f numerical simulation of the
atmosphere and ocean.
- Foreword
This book is devoted to methods of numerical simulation. It is intended as a
description of the structure of and of the principles of the structure of a ~
- complex of mathematical models and computing algorithms forming the basis of a
package of applied programs for solving problems in atmospheri~ physics and ~
physics of the ocean and environmental protection. Tta main goal is a systematic
discussion of the structural algorithmic aspects of this problem. Therefore
general questions relating to the theory and physical fundamentals of mathematical
simulation are touched upon only in part in tt. The central place in the book's
logical system is occupied by the fourth chapter, in which methoda of studying
the sensitivity of discrete models and their application~are described, and the
mathematics needed to implement these methods are discussed in the first three.
Know-how in so7.ving applied problema gained at the USSR Academ}* of Sciences Siberian
Division Computing Cente~r was used in writing this book.
l24
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2407/02/09: CIA-RDP82-00850R000500450042-5
FOR OFF[CIAL USE ONLY
A considerabl.e paxt o~ the algoxithma pxe~ented hexe has heen innp].emen~.~d on
- computers by A.Xe. Aloyan, N.N. Ohraztaov and ~1.V. ~xotasov. ~'hey~ hav'e a~.so
completed a long cycl.e o~ methodological. s~udiss and numerica]. experiments on
solving speci~ic pxobletns o~ atmosphexic phpsics and ph}*sics o~ the ocean.
The author considers it his pl.easant duty to express his pro~ound gratitude to
his teacher, Academician G.T. Marchuk, who exerted a greaic influence on Pormation
of the aLthor's scienti~ic interests.
At individual stages in this work considerable as~istance in the ~orm o~ advice
and comments was rendered by N.N. Yanenko, N.T. Buleyev, L.T. Matveyev, Sh.A.
Musaelyan, M.I. Xudin, L.S. Gand~.n,'V.P. T1~in, A.N. Konovalov, V.P. Kochergin,
G.P. Kurbatkin, V.P. Dynmikov, Y'e.Ye. Kalenkovich, V.g. Kim, G.R. Kontarev, G.S.
Rivin and especially A.Y'e. Aloyan, N.N. Obraztsov, A.V. Protasov and Ye.A.
Tsvetova, for which the author expresaes his sincere gratitude to them all.
The author is deeply grateful to S.L. Belousov, S.M. Yermakov and L.V. Rukhovets,
who read the book in manuscript and made a number of important comments, as well
as to G.T. Ivanova, N.I. Kolker and Z.K. Urazalina for their great help in pre-
paring tlie manuscript.
Introduction
The problen~ of mathematical simulation of the dynamics of the atmosphere and ocean
has attracted the attention of scientists of various countries for a long time.
This is due to the fact that studies of processes taking place in the atmesphere
and ocean are closely allied to solving problems in the theory of climate and wea-
ther forecasting which are of dirPCt practical importance. In recent years interest
in this has increased more and more in connection with the problem of the inter-
action of man with his environment. Taking into account the complexity of arrang-
ing full-scale experiments under real conditions, the most natural approach to
studyir.~ a^d evaluatinQ rhP infl,~Ance of man's act~r~t~es on the atmospherA and
ocean is the creation of mathematical models and methods of mathematic~l simularion
making it possible by means of numerical experiments on high-efficiency computers
to evaluate perturbations o.f. key parameters and functionals characterizing the
state ~nd conditions of the atmosphere and ocean.
For solving this class of problems the following are first necessary:
Physically complete mathematical modeis of climate, general circulation o~ the
atmosphere and ocean and local atmospheric and oceanic processes.
A theory for the transfer of aerosols and other impurities in the atmosphere and
methods of evaluating the influence of transformation of these impurities on the
composition of the atmosphere and on changes in hydrometeorological conditions on
the local and global scales.
Methods of numerical simulation o~ meteorological and oceanic processes and prin-
ciples of their use in estimating the influence oP man~s activities on hydrometeoro-
logical condit~ons.
125
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5
FOR OFF'ICIAL USE ONLY
Perturbation theoxy s~ethod~ ~ox es.ti~~i~ng ~Paxiatians~ o~ tqa~qx ~unctiona,~a ~n
problems o� atn~ospheri~ and oceanic physics~, ~or study~ing th~ Sensiti~vity o~
models to variafiions in input da~a and ~o'r esti~qat3,ng c~,ima~ic anoma7,ies.
Models of followi,ng tY~e state o~ a climat~c sy~tem and of inethods o~ assimilatiing
data from measuring complexes.
This book is devoted to the development of the last three o~ the trends enumerated
above. Obviousl.y success in solving the problems posed depends to a considerable
- extent on the physical models used as a basis and on their mathematical �ormulation.
But since the main goal of this book is the development of the algorithmic aspects
of the problem of numerical simulation, the theoretical fundamentals and the
physical meaning of the ~ormulation of hydrothermodynamics problems are practically
nowhere discussed in it. These questions are repres~nted sufficiently completely
in s~udies by T.A. Kibel'; Ye.N. Lorents; S. Manabe and K. Brayen; G.I. Marchuk;
L.T. Matveyev; A. Arakava; A.S. Monin; M.T. Yudin; in a collective monograph; etc.
Ideawise the book is closest to the studies by G.I. Marchuk; therefore in the
discussion the author relied on the definitions and terminology used j.n his mono-
graphs.
The variational-difference method of constructing energy-balanced discrete models
of hydrothermodynamics o� the atm~~;~here and ocean is discussed in the first
~ chapter. Its main idea and applications to equations of mathem.3tical physics
are described very well in ch 6 of O.A. Ladyzhenskaya's book. Tn order to use this
method, the mathematical model must be written by means of an integral identity
corresponding to the initial system of equations and boundary ar.d initial condi-
tions. This double description of the model by means of systems of equation~ and
integral identities has a number of advantages for numerical simulation.
After derivation of the identity, further operations are accomplished formally
and in principle can be run on a computer. First a discrete, i.e., a summer,
analog of the identity is constructed and then the conditions are written for
stationarity of the summer functional with random and independent variations in
trial functions at the nodes of a net region. Systems of basic and con~ugate
equations in discrete form are thus obtained. It was found that by special con-
struction of approximations of the identity over time it is possible to obtain
automatically from it splitting diagrams, also.
The integral identity is so constructed to be able to introduce di~~erent net re-
gions, to combine regular and irregular nets, to approxiinate integrals and inte-
grands differently, to use diagram~ with an enhanced order of accuracy, etc., al-
ways keeping invariant for discrete analogs the energy balance property.
The atmosphere-ocean-continent system is unified in the energy sense. Therefore,
- if an energy functional is introduced for it and an integral identity is con-
structed on its basis, some conditions at interfaces are shown to be natural for
� a variational ~unctional. Because of this, in obtaining discrete approximations
a number of questions involving matching scalea and approximating boundary condi-
tions are done away with and the possibtlit}r appears of automatically monitoring
energy and mass exchange at interfaces. However, let us r.ote that the problem of
126
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2047102109: CIA-RDP82-00850R400504050042-5
FOR OFF[C[AL USE ONLY
parametrizing ~rp~es.~~s o~ intexa~~ion o~ the atIIlosphexe with the ocean and the
sur~ace o~ dry~ 1,at~d mus.~ sv~n w~th ~hi;s appxvact~ b~ con~~dsxed ~3e.paxate,".y.
The second chapter contains a desc~iption of inethods of constxucting bases ~or
the representation o~ multidi~ens~onal ~te1d~ and spectral-d~~~erence models of
atmospheric hydrothermodynamtca.
The problem of constructing inforatatfi~e haaes ~s o~ independent 6ign'~ficauce in
hydrometeorology. A numbe~ o~ approaches to sol~ririg it are d~scribe.d in the
, literature. On1y some o# them are discussed here. The mosfi promiaing for app13,-
cations is the method based on camb~:ning an approximate solution to the particular
spectral problem for linearized hydrothermodynamics operators with the isolation o~
the natural orthogonal camponenta of the f3elds of ineteorological elements ("pri-
mary components" in term~ of factor analysis and mathemafiical statistics). Opera-
tors are linearized in the neighborhood of a certain assigned primary state, which
is introduced as one of the input parameters.
With the existence of bases discrete modela are derived from the integral identity
or its summEr. analogs by an approach etandard for variational difference methods.
This results in closed energy-balanced systems of basic and conjugate equa.tions in
spectral-difference form. If the basis is sufficiently informative, i.e., the
fields of unknown functiona can be represented approximately by a truncated Fourier
series with a small number of terms, then a model of this sort is customarily
called few-component or few-parameter. Tn particular, the basis can be made up of
a group of functions with finite carriers.
In view of the nonlinearity of the initial model the equations for Fourier coeffi-
cients are also nonlinear. And since in constructing the basis an approximate
solution to the spectral problem is used for the operator obtained from the same
initial system, then there is present in them also a linear part with a diagonal
matrix made up of approximate eigenvaluea of this operator, which has been shown to
be convenient to take into account analytically when integrating in terms of time.
This approach makes possible the integration of all equations simultaneously with
the same time interval in spite of the difference in eigenvalues. Finally, for
reconstructing unknown fields from Fourier coefficients A.N. Tikhonov's method ~or
the generalized summation of series, based on th~ regularization idea, is~used.
- It must be emphasized that in spectral models Fourier coefficients represent
generalized characteristics of the entity being simulated and with the successful
~ choice of basis functions, even with a small amount of them, it is possible to
carry a considerable shar~ of useful information. Eigenvalues of the problem's
linearized operator also come under the heading of generalized characteristics.
From eigenvalues it is possible to ~udge the time scales of processes being simu-
lated and from the sign of their real half the atability of the.se processes in the
neighborhood oP the primary state.
In the third chaptex the ~rinciples o~ the practical implementation o~ nwnerical
models based on the splitting-up method are discussed. The basic tenets of this
method are widely ~amiliar at the present ti~me because o~ atudies by the Soviet
mathematicians Ye.G. D'yakonov, G.T. Marchuk, A.A. Samarskiy, N.N. Yanenko, etc.
The splitting-up method is o~ fundamental importance in constructing mathematical
. 127
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5
FOR OFFICIAL USE ONLY
model5 of compl.icated phys.ical procesaes. ~h~,s; has b~~n cqnv~ncingl,y demonStxated
in G.I. Marchuk's books ~or problems o~ dynami,c~ o~ the atmoSphere and acean.
The development o~ nu~qexical, simulat~on methods is ~~.osely a113,ed w~th the tech-
nology of their practical. implementation, which is be3,ng steadil}r improved togetiher
with progress in the ~~e1d o~ cmnputer technology. Already at this time it has
become gener~lly accepted that it is ntost fittirig to emplo}r the modular principle
of prograu~ing and to create packages o~ applied programs. The modular structure
of algorithms and programs makes it possible to approach a~olution to entire
classes of problems and the existence o# facilities for adapting packages to spe-
ci.fic conditions makes it possible to make active use of accumulated experience.
The practical l.mplementation of numerical models is organized according to the
following system. First a"modular" analysis is made of the entire complex of
problems. Then a certain basis set of general-purpose algorithms is singled out,
which are implemented in the form of program modules in algorithmic language.
From these modules a special library is constructed, comprising the basis of the
package of applied programs. ~
- Of course, it is impossible within the scope of a single book to describe the
entire set of basis algorithms; therefore, it was decided to limit ourselves to
a presentation of only those of them which in our opinion are most interesting
for the class of problems discussed. These are direct algoritlims for solving
transport equafiions and the problem of the dynamic matching of the fields of
meteo.rological elements. Preference was not given to direct algorithms accidental-
ly. They are fairly economical in terms of the number of arithmetic operations
and as compared with iteration algorithms�they reduce the degree of uncertainty
in ~olving a problem. The advantages of direct algorithms are evidenced especially
clearly in constructing perturbatian theory methods, where it is a question of
- estimating slight variations in the solution depending on variations in the model's
parameters and where it is necessary to filter out fictitious perturbations and
noise caused by errors in the method of solving the problem.
The fourth chapter is devoted to methods of studying the sensit:tvity of d~iscrete
models of dynamics of the atmosphere and ocean. The idea basis of these methods
is the perturbation theory, the principles of whose structure for the class of
problems considered were developed by G.I. Marchuk. Functionally they are designed
for studying the behavior of models in a space of parameters and for identifying
mod?ls in terms of ineasurement data.
The concept of state functions and of parameters can be introduced for each mathe-
matical model. A model essentially specifies the transformation which places
a group of values of state functions in correspondence with a group of values of
parameters. The difference between parameters and state functions is in many in-
stances purely hypoChetical and depends on the specific formulation of the prob-
lem. An example o~ this can be the interrelationship between a~titude and pressure
in a Cartesian and isobax system of coordinates. For the purpose of describing
the behavior of the model, in addition to state functions it is possible to use
a certain set of functtonals determined in sets of values of state functions and
parameters. An analysis oP current approaches in numErical simulation has demon-
strated that models constructed on the basis of ~ust the basic system of equations
' 128
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/49: CIA-RDP82-00850R040500050042-5
FOR OFFICIAL USE ONLY
of atmospheric and oceanic hydrothermodynamics are pooxly~ aui.t~d to studying
their behavior in a space o~ parametera. Tn practice the numbex o~ para~neters
is shown to be too $xeat and the organizatton o~ num@xical. exper~~nents according
to the traditional system o.~ dixect simulat~on, wRich reduces to a u~ultiple so1u~
tion to the prob7,em with dif~erent input data, fis poorly~ ef~'ective even when using
very high-ouput computers~ The numer~cal model structux~ described in this mono-
graph proposes the introduction o~ at least tliree more elements---a system of con-
~ugate equations, a system o~ var~at~onal equations, and a block ~or the calcula-
tion of sensitivity ~unctions, ~ox e~t~mates of var~ations in ~unctionals and for
identifying the parametexs o.~ modele. A11 elements of the model are distinguished
in terms of their Punct~.ona1 purpose. For example, b}r solving the b~sic spstem o~
equations or variational problem, it is possible to predict how the influence of
perturbations of any ~actor wi11 be spread fin space and over tiTae. By solving the
conjugate problem and calculatfing sensitivity funct3,on~ with a apecfific choice of
Punctional it is possible to estimate whence perturbations w311 arrive in a spe-
cific region and to reveal region~ of heightened sensitivfity vf the model to vari-
ations in parameters. It is especially important to have an opportunity to obtain
such quantitative estimates in problems of controlling the quality of the environ-
ment.
The theory of sensitivity represents a new and actively developing scientific
trend in cybernetics. An analysis of its key ideas and state-of-the-art can be
found in R.M. Yusupov's review article. At the pre~ent time it is developing as
a set of inethods and approaches for studying th~ influence of the properties of
operators and their perturbationa on the quality of the functioning of various
automatic control systems; but the range of application of sensitivity :heory me-
thods is considerably broader--it includes also various trends in mathewatical
- simulation.
_ Three types of problems arise in studying the sensitivity of models. To the first
type belong direct problems, when variations in parameters are known and it is
necessary to estimate variations in state functions or functionals, and to the
second inverse problems, when from known variations in functionals or state func-
tions it is necessary to estimate variations in certain parametera. Problems of
the mixed type, including elements of direct and inverse problems, are also known.
In relation to the computing aspects of the sensitivity theory it should be men-
tioned that at the present time the solution of inverse problems is encountering
great, and in certain cases insurmountable, difficulties (cf., for example, R.P.
Fedorenko's monograph).
Sensitivity functions are introduced from physical considerations and implement
a relationship between variations in unknown and input quantities. Pormally they
can be defined as derivatives of some characteristic of the model tn terms of para-
meters of the same model and can be computed by solving basic and con~ugate prob-
lems and variational problems. Although in nonlinear models the use of perturba-
tion theory methods involves limitations on the magnitude of variations, neverthe-
less sensitivity �unctions play a ma~or role in the numerical simulation process.
By their means it is possible to estimate a trend in the influence of perturbations
of parameters, to plan an observation experiment for estimating functionals, and
in estimating variations o� ~unctionals to reduce the in~luence of ~ictitious noise
produced by a discrete mode7..
FOR OFFICIA~ USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2407/02/09: CIA-RDP82-00850R000500450042-5
FOR OFF(CIAL USE ONLY
Formalization o~ the individua~. stepS~ ~n const~ruct~;n$ d~:scxete appxox~tions and
computing algax~thms snakes poss:Cb7.e mutua7. compati;bi7.ity~ o~ ~a,x~.ous paxt~ o~ the
model and ~aci~~tat~s the simu7.at~on process, but ~n n4 case subs~itutes fox the
intuition and exper3;ence o~ invegtigzttors in creation o~ the mode7, itsel~. The �
mor~ apriori 3,n,Eoxmation and physical ~n~aning which are put into the model, the
greater the hope o~ a successful solution to the problem. ~or a long time now
the problem has been posed ot introducing a two~ray relationship between models and
measurements o~ tfie character~;stics o~ ent~ties to be simulated. How fs this to
be done? Especial,l}* it is taken into account that measuxements are made by .
many measuring sy~stems distributed unevenly in spaee and fiaving various precision
tolerances. The dtrect inclusivn o~ mea~urement data in the model results in the
so-called "shock" e~~ect, the rea~on for which is the mismatch between the scales
of perturbations contained in observation data and in simulated fields.
Here one way is suggested for implementing feedback through the minimization of
functionals determining the degree of devi2~ions between fields computed by means
of the model and measured under actual conct~tions, by using models as space-time
interpolators on the basis of the variational principle. Modern models are very
complex and fitting them to actual data generally can prove to be too expensive,
and the measurement data themselves are insufficient for identifying all the para-
meters of models. Aut if the cost of these studies is compared with iche cost of
full-scale, then it becomes c'lear that the development and use of inethods of solv-
ing inverse problems are justified, since even for the most.ideal models there will
always be the problem of estimating the degree of their correspondence to real
physical systems and of the intelligent use of ineasurement data. Combining c~mplete
models with few-component ones represents a reasonable compromise in solving this
problem. Few-parameter models are simpler to identify and they are economical to
implement. The spectral structure of these models makes it possible to separate
perturbations in terms of scale and to take into account primarily the part of them
most full of interest. At the same time complete models are more convenierit for
a local description and parametrization of various facto~s and it is advisable to
use them for refining solutions obtained by means of few-parameter models.
COPYRIGHT: Gidrometeoizdat, 1981
8831
- CSO: 1863/118
130
FOR OFFIC[bI, U`'~ ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2047/02/09: CIA-RDP82-00850R400504050042-5
FOR OFFICIAL USE ONLY
, SELECTED ITEMS FROM .TOURNAL FINSTRUMENTS, MEANS OF AUTOMATION AND CONTROL SYSTEMS',
FERR.UARY, 1981
Moscow PRIBORY, SRIDSTVA AVTOMATIZATSII I SISTEMY UPRAVLEI~TIYA, . TS-2: SRED"aTVA VYCHI-
SLITEL'NOY TII{HNIK7 I ORGTII{HNIKI (REFERATIVNYY SBORNII{) in Russian No 2, Feb 8]
pp 1-14~
[Excerpts]
15332~ "The complex 'Ele~on-D-2M' for measuring arid monitoring integrated digital
dynamic parameters of microcircuits. Bondarevskiy, A. S., Pankov, Ye. D., Ostapenko,
A. A., K ozlov, A. A., and Klimenko, L. V. ELF~CTRONNAYA TII{HNIKA .~SERIYA UPRAVNENIYE
KACHESTVOM~ METROLOGIYA, STANDARTIZATSIYA~ 1980, No 2-3 (80-81~, p 272 (Central Scien-
tific Reseaxch Institute "Elektronika"). .The urgency of development of the complex
"Elekon-D-2M" is substantiated. Its ma,in technica,l cha.racteristics are compared with
the Soviet and foreign analogues. The working principles, compositiori and metrologi-
cal specifica.tions of the complex are described. (Input by GPNTB SSSR, 103031, Mos-
cow, K-31~ ulitsa Kuznetskiy most, 12~.
5� 15338� "A~angement of components on computer structures by the method of branch-
es a.r~d bonds. " Lebedev, B. K., and Kalashnikov, V. A. EI,II{TRONNAYA TII{HNII{A . SERIYA
M~{ROELII{TRONNYYE USTROYSTVA, 1980, No 1(19), pp j8-67, Central Scientific Research
Institute "Elektronika". An algorithm is proposed for the arrangement of components
of computer structures, taking into consideration the real conditions of layout by
channel algorithms. In para11e1 with the arrangements, a forest of very short, ortho-
gonal, connecting trees is constructed. (Input by GPNTB SSSR~ 103031, Moscow, K-31,
tilitsa Kuznetskiy most, 12~ .
8. 15386. "Control and correction of a single error during cyclic shifts on the
basis of Lagrangian codes." Karpukhin, A. I., and Nugmanov, R. N. ELII{TRONNAYA T~{H-
NIKA. SERIYA MIKROELII{TRONNYYE USTROYSTVA, 1980, No 2(20)~ pp 13-19. (Central Sci-
entific Researah Tnetitute "Elektronika"), It is ehown that a single packet of errore
can be detected and corrected on the basis of Lagrangian codes during cyclic shifts.
Selection of the nodes of interpolation and the systema,tic cha,racter of th~ Lagrangian
codes permit leaving the code word coded for the indica.ted operations. In tha,t ca.se
the ne cessa.ry equipment for conversion of the certifying symbols is not complex ana
the time is approximately-the time of shift of the information part. A simple decoding
algorithm is presented for detecting and correcting a single errnr with consideration
of t~os operations. (Input by GPNTB SSSR, 103031, Moscow, K-31, ulitsa. Kuznetskiy
most _2~.
131
FOR OFF[CIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2047/02/09: CIA-RDP82-00850R000540050042-5
FOR OFFICIAL USE ONLY
20. 15388. "Integrated logical microcircuits noise st~,bility criteria." Ura.l'
skiy, Yu. A. ELII{TRONNAYA TIICHNIKA. SERIYA NID{ROELII{TRONNYYE USTROYSTVA, 1980, No
1(19), pp 3-16 (Central Scientific Resea.rch Institute "Elektronika". Vaxious noise
stability criteria of logical integrated microcircuits are examined for static and
dyn~,mic working conditions. Shortcomings of the presently adopted noise-intensity'
criteria are shown. An estimate of noise stability according to instantaneous power
is proposed,. sca.ttered by the noise generator on resistances of a logica,l section
constructed in the given system of elements. Circuits axe presented. An analytical
expression is obtained for the amplitude-time cha,racteristic, which determines the
noise stanility under c~ynamic conditions. (input by GPNTB SSSR, 103031~ Moscow, K-31,
uli.tsa KuznetsT~iy:?i~ost, 12~.
21. 15389� '~lanning operating storages on the ba,sis of supplementary NIDP-transis-
tor large-sca.le integrated circuit storages." Zhemeytsev, A. G., Goslavskiy, A. V.,
Kab~nov~ I. N., and Petrova, F. Sh. ELII{TRONNAYA TII{I~IldIKA. SERIYA NBfCROELIICTRONNYYE
USTf:cOYSTVA~ 1980, No 1(1~), pp 17-22. Central Scientific Research Institute "Elek-
tronika". On the ba.sis of analyais and systema,tiza,tion of experience in the develop-
ment of semiconducting operating stor=,:oe with MPD integrated memory circuits with
ca.pacities of 156 and 1024 bits are formed the requirements for elements for the con-
trol circuits and organization of the operating storages file. A number of struc-
tural and ciruitry solutions of large-scale integrated circuit storages is examined,
the solution of which will permit improving the cha.racteristics and simplifying the
development of storage devices on the whole. (Input by GPNTB SSSR~ 003031, Moscow,
K-31, ulitsa Kuznetskiy most~ 12~.
23. 1~00. "Questions of modeling and analysis of tests of digital devices." Feti-
sov, N. S. Preprint No 1(1980)~ Moscow, 1980, 23 pa,ges. The author examines ques-
tions about increasing the precision of logical modeling and ca.lculation of the state~
of inemory elements of sequential circuits during test analysis. (Input by GPNTB SSSR~
103031, Moscow, K-31~ ulitsa Kuznetsl~iy ritost, 12~.
COPYRIGHPs Tsentral'nyy nauchno-issledovatel'skiy institut informatsii i tekhniko-
ekonomicheskikh issledovaniy priborostroyeniya, sredstv avtoma,tizatsii
i sistem upravleniya (TsN.LITEIpriborostroyeniya), 1981
2174~
cso: 1863/99
132
~ FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2047/02149: CIA-RDP82-00850R000500050042-5
~OR OFFICIAL US~ ONLY
SF.LECTF.b ITF.MS FROM JOURNAL 'INSTRITMENTS, MEANS OF AUTOMATION AND CONTROL SYSTEMS',
APRIL, 1981
Moscow PRIBORY, SRIDSTVA AVTOMATIZATSII I SISTEMY UPRAVLENIYA, TS-2; SRIDSTVA VYCHI-
5LITE'L'NOY TEKHNIKI I ORGTII{HNIKI (REFERATIVNYY SBORNIIC) in Russian No 4, ~pr 81
pp 1-11
[Excerpts~
16. 15805. "Development.of fifth-generation computers." BYULLETEN' INOSTRANNOY
NAUCHNO-TF3CHNICHESKOY INFORMATSII TASS, 1980, No 29, p 19. The Japanese Ministry
of Foreign Trade and Industry ha.s decided to develop a computer of the fifth genera-
tion~ which is regarded as the computer of the 1990s and whose f~ulctions more resem-
ble those of the human brain than those of contemporary computers. The development
of the new computer will be started in the 1981 fisca.l year and completed in the
1990 fiscal year. The new computer will be able to ma~ce associations and analogies
and reason. [Input_b,y GPNTB ~SSR (USSR Main ''ublic Scientific-Technical Library)o
103031, Moscow, K-31, ulitsa Kuznetskiy most, 12~.
17. 15806. "Reduction of.noises created by key pressure stabilizers." Skegirev,
Yu. N., and Tomigas, 0. A. OBMEN OPYTOM V RADIOPROMYSHLENN~STI, 1980, No 6, pp 4~3-4~:
Some methods are suggested for reducin noises created by key pressure stabilizers ir.
apparatus obtai ning power from them. ~Input by GPNTB SS5R (USSR Main Public Scien- _
tific=Technical Library), 103031, Moscow, K-31, ulitsa Kuznetskijr.:Most, 12].
29. 15844. "The MIKRAS-system of microprogramming microcomputers constructed on the
ba.sis of series K589 large-scale integrated circuits." Bekasov~ A. A. In book:
"ALGORITMY" (Algorithms)~ Tashkent, 1980~ pp 9-18. UzSSR Aca.demy of Sciences. Col-
lecti.on of Scientific Works. No 4~0. A system is described for microprogramming com-
puters construected on the ba.sis of series K 589 large-scale inte ated circuits. [In-
put by GPNTB SSSR (USSR Main Public 5cientific-Technical Library~, 103031~ Moscow~
K-31~ ulitsa. Kuznetskiy most~ 12~.
COPYRIGHT: Tsentral�nyy nauchno-issledovatel'skiy ins~itut informa.tsii i tekhniko-
ekonomicheskikh issledovaniy priborostroyeniya~ sredstv avtomatizatsii
i sistem upravleniya (TsTTIITEIpriborostroyeniya), 1981
2174
cso: 1863/99
133
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5
FOR OFFIClAL t!SF: ONLY
SRLECTED ITEMS FROM JOURNAL 'INSTRUMFNTS, MEANS OF AU'FON~ATION AND CONTROL SYSTEMS',
MAY, 1981
Moscow PRIBORY, SREDSTVA AVTOMATT.ZATSIII SISTEMY UP?tAVLENIYA~ TS-2: SRIDSTVA VYCHI-
SLITEL'NOY TII{HNIl{I I ORGTII{HNIKI (REFERATIVNYY SBORNIK) in Russian No 5, May 81
pp 1-11
- [Excerpts~
9~ 1589~� "~ganization of storage of data files on magnetic carriers." Gaganov,
P. G. OBMEN OPYTOM.V RADIOPROMYSHLENNOSTI~ 1980, N o 7~ pP 3-5~ The author analyzes
the reasons for distortion of data on magnetic carriers durin lengthy storage and
use. Methods of increasing data storage time axe examined. ~Input by NIIWM (Scien-
- tific Research Institute of Control Computers)~ 34-9940~ Voroshilovgradskaya Oblast~
Severodonetsk, pl Pobely~ 2~.
22. 16005. "Installation for monitoring commutation plates on the hasis of
micro-
- computers. Nikulin, G. V. ELII{TRONNAYA TII{HNIKA. SERIYA 9~ Economics and Control
Systems, 1980, No 2(35), p 114. The paper presents the results of the development
and introduction of an installation for monitoring thin-film commutation plates of
hybrid integrated circuits, created on the basis of the microcomputer "Elektronika-
SB-O1 and an APK-1 automat3.c program contacting machine. [Input by G~'NTB SSSR (USSR
Main Public Scientific-Technical Libraryj, 103031, Moscow, K-31, ulitsa Kuznetskiy
most, 12~
23. 16006. ''Video terminal connector to computer." Gromov, Yu. K., Kekrasov, L.
T., and Tselyapin~ A. N. ELII{TRONNAYA TIICHNIKA. SERIYA 9~ Economics and Control
Systems, 1980, N~~'' (35)~ P 128. The paper describes the USO-5~ a device intended
for the organization of connection of the "Elektronika 100-I" computer with YeS com-
puters and various video terminals, particularly w~th the "Elektronika T1000". Ex-
amples of the accomplishment of such a connection in automa.ted control systems of
_ various configurations are presented~ (Input by GPNTB SSSR (USSR 2~fa,in Public Scien-
tific-Technical Library), 103031, Moscow, K-31, ulitsa Kuznetskiy most, 12~.
24. 1.6009. "Structural organiza.tion of large-scale integrated-circuit multicha,nnel
commutators." Zolotarevskiy, V. I.~ and Nekra,sov, V. M. ELII{TRONNAYA TII{HNIKA~
SERIYA 9. Economics and Control System~, 1980, No z(35)+ P 133~ The paFer examines
134
FOR OFFIC[AL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2407/02/09: CIA-RDP82-00850R000500450042-5
FOR OFFICIAL USE ONLY
structural circuits of multichannel commutators constructed on the basis of MOS
structures and intended for use in measuremant systems and systems of communica.tion
with an object in an automa.ted system foz~ control of technological processes. (In-
put by GPNTB SSSR (USSR I~ain Public Scientific-Technical Library)~ zo3o31, Moscow~
K-31, uli~~a Kuznetskiy most, 12].
COPYRIGHT: Tsentral'nyy nauchno-issledovatel'skiy institut informatsii i tekhniko-
ekonomicheskikh issledovaniy priborostroyeniya, sredstv avtoma,tizatsii
i sistem upravleniya (TsNTITEIpriborostroyenipa), 1981
2174~
cso: 1863/99
135
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5
FOR OFF(CIAL US~ ONL1'
SEI,ECTED ITEMS FROM.JOURNAL 'INSTRUMENTS, MEANS OF AUTOMATIOTT AND CONTROL SYSTEMS',
NQVEMfiER, 1981
Moscow PRIBORY~ SRIDSTVA AVTOMATIZATSII I SISTEMY UPRAVLENIYA, TS-2: SRIDSTVA VYCHI-
y SLITEL'NOY TII{HNIKI I ORGTII{HNIKI (~EFERATIVNYY SBORNIK) in Russian No 11, Nov 81
pp 1-12
~Excerpts~
21. 28936. "Complex technologica,l line for manufacturing fragments of fields of
circular fe~^rite cores." Bolunov, V. V., Dmitriyev, A. F., and Samiulina, A. A.
In book: "Razvitiye teorii i tekhniki sre~stv khraneniya informatsii" (Development
of the Theory and Technology of Means of Data Storage); Tezisy dokladov Vsesoyuznoy
nauchno-tekhnicheskoy konferentsii Moskva (Abstrac~ts~of Reports of 5cientific-Techni-
cal Conference, Moscow), Riga, 1980, pp 68-70 (Scientific and Technica,l Society of
Radio Engineering, Electronics and Communica.tion imeni A. S. Popov). The introductinn
of a complete set of technologica,l. equipment and instruments into production of frag-
ments of fields of circular ferrite cores permits reducing by two-thirds the labor
intensity of manufacturing fragments~ improving their quality and reducing expendi-
tures of materials. [Input by GPNTB SSSR (USSR Ma,in Public Scientific-Technica,l Li-
brary), 103031, Moscow, K-31, ulitsa.Kuznetskiy most, 12].
UDC 681.327.2
32. 28703. "Static storage unit with series AM-9131 random access." In book:
"Kratkiye tekhnicheskiye kharakteristiki na pecha,tayushchiye ustroystva~ nakopiteli"
(Brief Technica.l Cha,racteristics of Printers and Files), Severod.onetsk,,1980~ p 27.
NIIUVM [not f~.irther identified]. A brief technical characteriza.tion is presented of
a static storage unit with series AM-9131 random access. [Input by NIIWM, 34-994~0~
Voroshilovgradskaya Oblast, Severovodetsk, pl, Pobeli~ 2].
Unc 681.327.2
33~ 28704~. "Static storage unit with series AM-914~2 random access~" In booki
"Kratkiye tekhnicheskiye kha,rakteristiki na pecha,tayusYichiye ustroystva, nakopiteli"
(Brief Technical Characteristics of Printers and Files). Severod.onetsk, 1980, p 32.
NIIUVM. A brief cha.racteristization is presented of a static storage unit with ser-
ies AM-914~2 random access. [Input by NIIWM~ 3~99~0, Vorosh3.7:ovgradskaya.Oblast,
Sever~vodsk, pl Pobeli, 21.
136
~
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2407/02/09: CIA-RDP82-00850R000500450042-5
FOR OFFICIAL USE ONLY
34~. 28705. "Static storage unit with series AM-9140 random access." In book:
"kratkiye texhnicheskiye kha,rakteristiki na pechatayushchiye ustroystva, nak opiteli"
(Brief Technica,l Cha.racteristics of Printers and Files). Severodonetsk, 1980, p 27.
NIIUVM. A brief cha.racteriza.tion is presented of a static storage unit with series
_ AM-9140 random access. [Input by NIIUVM, 349940, Voroshilovgradskaya Oblast~ Severo-
vodsk, pl Pobeli~ 2~.
35� 28706. "Static storage unit with series AM-91jU random access." In book:
"Kratkiye tekhnicheskiye kha,rakteristiki na pechatayushchiye ustroystva~ nakopitali"
(Brief Technica,l Characteristics of Prin~cers and Files). Severodonetsk, 1980~ p 27.
NIIUVM. A brief cha,racteriza.tion is presented o~ a static storage unit with series
AM-9130 random access. [Input by NIIUVM, 34~99~ro~ Voroshilovgradskaya Oblast, Severo-
vodsk, pl P obeli, 2~.
~3� 2892~. "Some results of the development of a ae~i.c~3uc~o"rc~a~s~@~:.permanent
holographic memory." Yelkhov~ V. A., Klimov, I. I.~ and Shidlovskiy, P. P.. In
book: "Razvitiye teorii i tekhniki sreclstv khraneniya informatsii" (Development of
the theory and technology of ineans of data storage); Tezisy dokladov Vsesoyuznoy
na,uchno-tekhnicheskoy konferentsii Moskva (Abstracts of Reports of Scientific-Techni-
cal Conference, Moscow), Riga, 1980~ pp 102-103 (Scientific-Technica.l Society of Radio
Engineering, Electronics and Communications imeni A. S. Popov). The paper describes
prototype realiza,tions of a module of a permanent hoiographic memory and clevice for
recording holograms with use of semiconducting lasers. [Irput by GPNTB SSSR (USSR
Main Public Scientific-Technica,l Library~~ 103031, Moscow, K-31~ ulitsa Kuznetskiy
_ most, 12].
~9~ z8935~ "Semiconductor large-ca.pacity memory unit ba.sed on K565RU1 dynamic inte-
grated circuit." Nozik, I. R., and Vasilishin, S. M. In books "Razvitiye teorii i
tekhniki sredstv khraneniya, informa,tsii"; Tezisy dokla,d.ov Vsesoyuznoy nauchno-tekhni-
cheskoy konferentsii Moskva, Riga, 1980~ pp 63-65 (Scientific-Technical Society of
Radio Engineering, Electronics and Communica.tions imeni A. S. Popov~. A semiconductor
main storage ba.sed on K 56jRUl integrated mic~ocircuits has been developed and intro-
duced into the composition of a coml~uter. ~Input by GPNTB SSSR].
52� 289~. "Improvement of tne elementary base of optical storage ba,sed on inte-
grated optical devices." Pokrovskiy, Yu. A., Makaretskiy, Ye. A.~ and Khurkhulu~
Yu. S. In book: "Razvitiye teorii i tekhniki sredstv khraneniya informatsii"; Tezisy
dokladov Vsesoyuznoy nauchno-tekhnicheskoy konferentsii Moskva~ Riga~ 1980~ PP 53-55
(Scientific-Technical Society of Radio Engineering, Eleetronics and Communications
imeni A. S. Popov). Some questions of improvement of the elementary base of optical
storage based on integrated optical devices are discussed. [Input by GPNTB SSSR].
53� z89~5~ "Unused possibilities of computer ferrite morires." Burkin, Yu. A. In
book: "Razvitiye teorii i tekhniki sredstv khraneniya, informa,tsii"; Tezisy dokladov
V:,esoyuznoy nauchno-tekhnicheskoy konferentsii Moskva~ Riga, 1980~ PP 52-53~ (Scien-
tific-Technical Society of Radio Engineering, Electronics and Communica.tions imeni A.
S. Popov). A new technology is described for the production of ferrite stora,ges, one
developed by the Siberian Department~ tJSSR Aca.demy of Sciences. [Input by GPNTB SSSR~.
COPYRIGHT: Tsentral'nyy nauchno-issledovatel'skiy institut informatsii i tekhniko-
ekonomicheskikh issledovaniy priborostroyeniya, sredstv avtoma,tiza,tsii
i sistem upravleniya (TsT1IITEIpriborostroyeniya)~ 1981
= 2~7,~
Cso: 1863/99
137
FOR OFFIC[AL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2407/02/09: CIA-RDP82-00850R000500450042-5
FOR OFFICIAL USE ONLY
SELECTED ITEMS FROM JOURNAL FINSTRUMENTS, MEANS OF AUTOMATION AND CONTROL SYSTEMS',
DECEMBF.R, L981
Moscow PRIBORY~ SREDSTVA AVTOMATIZATSII I SISTEMY U~RAVLETTIYA, TS-2: SRIDSTVA VYCI~-
5LITEL'NOY TIICHNIKI I ORGTIICHNIKI (REFERATIVNYY SBORNIfC) in Russian No 32, Dec 81
pp 1-12
[Excerpts~
uDC 681. . 325. 65. 06
33~ 30b4~. "Software for the ELII{ON-SF system." Boxisova~ Ya. I.~ Vasil'yev, Ye.
P.~ Lyshenko~ V. I., Orlov, Ye. V., Popel', L. M., and Traskova, S: M. ELII{TRONNAYA
PROMYSHLE'NNOST'~ 1981, No 2(98), pp 28-33. The paper presents software developed
for the series-produced system for monitoring large-scale integrated-circuit memory
units and microprocessors ELF.~{{ON-SP, including the high-level monitoring language
ELII{ON F, a complete set of programs for statistica.l processing of the results of
monitoring the laxge-sca.].e integrated circuit and programs for monitoring working
capacity, dia.gnosis of defects and metrological certification~of the system~ [Input
by GPNTB SSSR (USSR Main Public Scientific-Technica.l Library~~ 103031, Moscow, K-31,
ulitsa Kuznetskiy most, 12~.
uDC 621.3.049.77:681.2
3G,. 3o6~r5. "The K59~PA1 infra-analog converter." Abraytis, V. B., and Klima.shau-
skas. ELII{TRONNAYA PROMYSHLENNOST'~ 1981, No 2(98~r PP ~9-50. The paper presents
- the trlain cha.racteristics of an infra-analog crnverter, intended for conversion of a
12-bit code into analog current with an error of 0.012 percent. [Input by GPNTB SSSR].
unc 621.3.04~9.77:681.327.67
35~ 30~� "K170AP3 and K170AP4 integrated semiconducting microcir_cuits for metal-
lic-oxide semiconductor storage control." Samsonov, V. 5., and Khmel'nitskiy, S. L.
ELIICTRONNAYA PROMYSHLENNOST', 1981~ No 2(98)~ PP ~8-~9~ The paper presents the main
electronic chaxacteristics of K170AP3 and K170AP~V microcircuits intended for metallic-
oxide semiconductor ~torage~ ~Input by GPNTB SSSR]~
uDO 621.3.04~9.77i.i4
3064~2. "Instal.lation for flzr?ctional monitoring of a large-sca,le integrated cir-
cuit main storage." Danilin, N. N., and Popel'~ L. M. ELII{TRONNAYA PROMYSHLENNOST',
~
138
' DR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5

APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5
1y81, No 2(98~, PP 15-18. A brief technical characterization is given of an in-
- .,tallation for fur,ctional monitoring of the ELEKGN F-ZUM large-scale integrated-
circuit main storage. [Input by GP~B SSSR~.
- unc 621.38z.6i9~3
- 55~ 30643. "Statistica.l analysis of parametric reliability of a 256-bit large-scale
integrated-circuit main memory." G orovoy, V. V., and Savotin, Yu. I. ELIICTRONNAYA
PROMYSHLENNOST', 1981, No 2(98), pP ~3-~5; Bibliography: 4~ items. The paper presents
the results of investigation of the parametric reliability of a static ma.in memory
with a capacity of 256 bits on ESL-structures and determination of possible ways to
increase ma.in memory reliability. Used for analysis was the physical-statistica.l
approach, based on consideration of the interrelation of the statistical laws of
distribution of the parameters of an article with the character of their variation
in time. [Input by GPNTB SSSR~.
UDC 621.3.0~9.77:681.3
56, 3064~~. "Data file KR555Rye4~ with a 16K ca.pacity for a symbol generator."
Nazar'yan, A. R.~ Neklyudov, V. A., and Shchetinin, Yu. I. ELIICTRONNAYA PROMYSHLEN-
NOST', 1981, No 2(98)r PP ~6-~8. The paper examines the purpose~ compositions and
functions of a KR555RYe4 da.ta file for a symbol generator. The main electrical para-
meters of the file are presented. ~Input by GPNTB SSSR].
C~PYRTGHT: Tsentral'riyy nauchno-issledovatel'~kiy institut informa.tsii i tekhniko-
ekonomicheskikh issledovaniy priborostroyeniya, sredstv avtoma.tizatsii
i sistem upravleniya (TsNIITEIpriborostroyeniya~, 1981
2174
CSO: 1863/99 ~D
- 139
FOR OFFICIAL USE ONLY
APPROVED FOR RELEASE: 2007/02/09: CIA-RDP82-00850R000500050042-5