top - download
⟦b0078148a⟧ Wang Wps File
Length: 52231 (0xcc07)
Types: Wang Wps File
Notes: SDPS,Seismic Data
Names: »2135A «
Derivation
└─⟦2c4664e78⟧ Bits:30006226 8" Wang WCS floppy, CR 0173A
└─ ⟦this⟧ »2135A «
WangText
SEISMIC DATA PREPROCESSING
SYSTEM
1982-05-13
SECTION I
Page #
TECHNICAL PROPOSAL
T̲A̲B̲L̲E̲ ̲O̲F̲ ̲C̲O̲N̲T̲E̲N̲T̲S̲
1. TECHNICAL SUMMARY ..........................
2 R̲EQUIREMENTS ANALYSIS .........................
2.1 System Overview ...........................
2.2 Interfaces ................................
2.2.1 Input .................................
2.2.2 Output ................................
2.3 Operation .................................
2.4 Environment ...............................
2.5 Functions ................................
2.5.1 System Management .....................
2.5.2 Input and conditioning of data ........
2.5.3 Preprocessing ........................
2.5.3.1 Estimation of Missing Values ......
2.5.3.2 Beam stearing (option) ...........
2.5.3.3 Resampling in the scan direction ..
2.5.3.4 Resampling in the trace direction
2.5.4 Demultiplexing and output ...........
2.6 Performance .............................
2.6.1 Data Rates ..........................
2.6.2 Processing Rate .....................
2.7 Support .................................
3. PROPOSED TECHNICAL SOLUTION ...............
3.1 Baseline System .........................
3.1.1 System Operation ....................
3.1.2 Data flow ...........................
3.1.3 Options and Expandability ...........
3.1.3.1 CRT Input and HDDT Output .......
3.1.3.2 Beam Stearing ...................
3.1.3.3 Increased Processing Speed .....
3.1.4 Performance .........................
3.2 Hardware specification ..................
3.2.1 Processing Unit .....................
3.2.3 Data Sheets .........................
3.3 Software Specification ..................
4. TRAINING ..................................
1. T̲E̲C̲H̲N̲I̲C̲A̲L̲ ̲S̲U̲M̲M̲A̲R̲Y̲
Christian Rovsing A/S is pleased to submit this proposal
for design and implementation of a Seismic Data Preprocessing
System.
The proposal is in compliance with your specification
of 82-03-23:
Preprocessing specifications for Seismic Data from
a 1000 Channel System.
The Seismic Data Preprocessing System proposed here
meets all the requirements of above specifications.
The very modular architecture, as well in hardware
as in software, provides extensive expansion capabilities,
as well for increasing the throughput as for adding
optional functions.
The proposed Seismic Data Processing System is a baseline
system comprising the following elements:
a) High Density tape recorders
Honeywell Model HD96 recorder has been assumed
for the purpose of this offer.
We assume, for the purpose of this proposal that
the customer procures these recorders.
b) Processor Unit
with operator's terminal and printer
c) Storage and I/O Unit
with 10 Mbyte Random access memory and 40 MB disk
storage
d) 6250 bpi CCT recorders
The baseline system is capable of performing full preprocessing
at real-time speed.
The preprocessing is automated as far as possible to
achieve both a minimum of operator assistance and a
maximum utilization of the system.
The modular architecture of the CR80M system provides
extensive growth capabilities, both for additional
functions as for higher throughput.
Processing speed is achieved by adding additional hardware
for parallel processing.
Beam steering is performed without lowering the processing
speed by adding an additional Floating Point Processor
(FPP) in the pipeline.
The optional computer compatible tape input and high
density tape output will be implemented in the baseline
system hardware such that the implementation of the
functions will require only additional software packages.…86…1
…02… …02… …02… …02…
figure 2 SDPS W. CCT input
figure 3 SDPS W. CCT input + HDDT out
figure 4 parallel processing
figure 5 full-blown
The production of preprocessed CCT's involves the tasks
outlined below:
The operator is "readying" the next High Density Digital
Tape (HDDT) to be preprocessed on the free High Density
Digital Recorder (HDDR) while the previous is being
processed.
This involves the following operator tasks:
o load the HDDT
o specify HDDT and file identification (terminal
input)
o specify the preprocessing to be done, normally
by selection from pre-programmed menu'es
The system automatically checks for consistency between
operator specified identification and header. The
tape is also positioned at the requested file while
the processing of the previous HDDT is proceding
The preprocessing starts with set-up of the specified
process, e.g. loading the HD Tape Controller (HDTC)
and the Floating Point Processor (FPP) modules with
new parameters. Upon this, the HDDR is started automatically
and continues until the requested file(s) have been
preprocessed.
The HDTC separates the headers from the Seismic and
auxiliary data. The headers are transfered to the
Processor Unit (PU) main store while the Seismic and
auxiliary data are routed to the FPP module via a local
bus. This module performs the resampling in the scan-direction
and conveys data to the next FPP module for resampling
in the trace-direction.
A memory of 128 K bytes internal to the FPP module
provides the necessary buffer capacity for the "corner
turning" operation on segments of a reasonable size.
Subsequently, the data are transferred to the storage
and I/O Unit and intermediately stored to achieve the
demultiplexed format.
The updated headers are added by CPU controlled transfers
to form the SEG-D format, ready for output to the CCT.
A dual-buffer concept is used to achieve a continuous
operation when processing multiple contiguous files.
As for the HDDR's, the free CCT will be "readied" while
the other is running, such that the switching can be
performed virtually instantaneously.
The proposed preprocessing system has an exceptionally
high availability, achieved both by a high reliability
and very low mean time to repair.
The high reliability is achieved by careful component
selection and screening, followed up by burn-in of
each module.
In the event of failure, maintenance philosophy is
based on module replacement. The modules are accessible
from the front or rear and can be replaced by non-skilled
personnel without the need for special tools.
Built-in self-checking and diagnostic aids provides
for the location of faults.
The Seismic Data Processing System will be fully supported
by Christian Rovsing A/S. The following items can
be offered.
o Transportation and installation
o Full documentation
o Training of maintenance personnel
o On-call engineering assistance
o Product improvement and upgrading according to
future needs, including use on-board ships.
The decision to bid on the Seismic Data Processing
System represents a definite commitment on the part
of Christian Rovsing A/S to devote its resources and
technological talent to ensure the successful implementation
of an efficient and flexible tool for Seismic Data
Reduction
2 R̲E̲Q̲U̲I̲R̲E̲M̲E̲N̲T̲S̲ ̲A̲N̲A̲L̲Y̲S̲I̲S̲
The requirements to the Seismic Data Preprocessing
system (SDPS) are expressed in the "Preprocessing specifications
for Seismic Data from a 1000 channel system in terms
of
o Input format
o Functional requirements
o Turn-around time
o Priority on optional functions
This subsection represents a review of above requirements
in view of the system architecture outlined in the
previous subsection.
2.1 S̲y̲s̲t̲e̲m̲ ̲O̲v̲e̲r̲v̲i̲e̲w̲
The collection and preprocessing of Seismic Data involves
the following activities.
o Data acquisition and recording onto HDDT on-board
the research vessel
o Replay of the HDDT and simultaneous preprocessing
and output of demultiplexed data onto 6250 bpi
CCT.
This is initially to be performed on a ground based
processing facility, but with an on-board application
in view.
o Full processing using as input the preprocessed
data on CCT.
This proposal describes the system for generating preprocessed
data on CCT from raw data on HDDT.
A functional overview of the SDPS is shown overleaf.
The functions can be divided into the four principal
parts
o System management
o Input and conditioning of data
o Preprocessing
o Output in demultiplexed format
fct overview
2.2 I̲n̲t̲e̲r̲f̲a̲c̲e̲s̲
This section describes the SDPS data input and data
output
2.2.1 I̲n̲p̲u̲t̲
The input is High Density Tapes to be reproduced on
recorders which for the purpose of this proposal has
been assumed to be two Honeywell recorders model HD
96 with the following interface characteristics.:
o TTL compatible levels for both data and remote
control lines
o 24 bits Parallel data I/O
o Built-in deskew and error detection
o Remote control of tape speed (or channel rate)
in small increments (steps of a few percent of
loss)
The input data are files, one for each shot.
Each file has the logical sequence:
General header
scan type header
extended header
external header
followed by a number of scans, each comprising
Sync/time code
auxiliary data
seismic data
end of information
Zero-fill may have been introduced anywhere in the
data stream.
On the HDDR parallel data output, all data are presented
as 24 bit parallel words consisting of a 4 bit code
and a 16 or 20 bit data word. All header information
along with sync/time words are 16 bit words each containing
2 bytes (1 byte = 8 bits) as described in SEG-n. In
SEG-D, all information is described in pairs of bytes.
All 16 bit words on the HDDR output contain the first
byte of a pair in the 8 MSR of the 16 bit word, and
the second byte of a pair in the 8 LSR of the 16 bit
word. The 20 bit word format applies to sampled data
only and replaces the 2 1/2 byte binary exponent multiplexed
data format as described in SEG-D.
The following bit pattern applies for all data on the
HDDR output:
MSB
LSB
Start of information 1 1 1 0 x x x x x x x x x x x x x
x x x x x x x
Header (excluding
external header) 0 0 0 1 x x x x b b b b b b b b b
b b b b b b b
External header 0 0 1 0 x x x x b b b b b b b b b
b b b b b b b
Sync/time code 0 0 1 1 x x x x b b b b b b b b b
b b b b b b b
Aux. channels 0 1 0 0 c c c c s q q q q q q q q
q q q q q q 0
Seismic data 0 1 0 1 c c c s q q q q q q q q q
q q q q q q 0
Killed seismic trace 0 1 1 0 x x x x x x x x x x x x x
x x x x x x 0
Zero fill 1 0 0 0 x x x x x x x x x x x x x
x x x x x x x
End of information 1 1 1 1 x x x x x x x x x x x x x
x x x x x x x
x = don't care. May be forced to zero by recording
unit.
b = binary bit in 16 bit word according to SEG-D
as previously described
c = binary exponent. This is a 4 bit positive binary
exponent of 2, written as 2…0e…cccc…0f… where cccc can
assume values 0 to 15.
s = sign bit. One = negative number.
q = fraction. This is a 14 bit one's complement
binary fraction.
The radix point is to the left of the most significant
bit with the MSB being defined as 2…0e…-1…0f…. The sign and
fraction can assume values from 1 - 2…0e…-14…0f… to -1 + 2…0e…-14…0f….
Note the 0 in LSB of aux. and seismic data. This
is to ensure unique sync/time code in SEG-D.
Optionally, input from 9 track 6250 bpi CCT could be
provided from one or alternatively two drives in ping/pong
mode. The driver should be capable of running with
a speed of 200 ips in reproduce mode.
2.2.2 O̲u̲t̲p̲u̲t̲
The preprocessed data are recorded onto 6250 bpi 9-track
CCT in demultiplexed (trace-oriented) SEG-D format.
The files will be written sequentially on the tape.
Tape change will occur only when the tape has been
filled to an extent such that room for the next file
cannot be guaranteed.
The CCT's shall be able to run at 120 inch per second
as a minimum.
High Density tape output to be produced from CCT could
be provided as an option. This (archiving) operation
will not include any processing
2.3 O̲p̲e̲r̲a̲t̲i̲o̲n̲
The preprocessing unit shall be operator interfaced
using a computer terminal. The operator communication
shall be in high level commands, using menues with
standard values.
The operator shall be able to edit, create and delete
menues.
A minimum of two CCT's for output must be provided
such that the one can be loaded with fresh tape while
the other is recording.
2.4 E̲n̲v̲i̲r̲o̲n̲m̲e̲n̲t̲
The environment for the first SDPS system will be normal
computer environment, 110 V, 220 V, 50 Hz or 60 Hz,
but the system should be amenable to a shipbased environment.
2.5 F̲u̲n̲c̲t̲i̲o̲n̲s̲ ̲
The functions of the SDPS can be divided into the four
groups:
o System management
o Input and conditioning of data
o Preprocessing
o Demultiplexing and output
2.5.1 S̲y̲s̲t̲e̲m̲ ̲M̲a̲n̲a̲g̲e̲m̲e̲n̲t̲
The system management includes Operator communication,
processing set-up and all system control and monitoring.
The system management shall support
o Updating of header
o logging, labelling
o Software update and maintenance.
2.5.2 I̲n̲p̲u̲t̲ ̲a̲n̲d̲ ̲c̲o̲n̲d̲i̲t̲i̲o̲n̲i̲n̲g̲ ̲o̲f̲ ̲d̲a̲t̲a̲
The following functions are required for the input:
o Remote control of the HDDR
o Separation of Ancillary data (headers),
auxiliary data and seismic data
o Removal of filler data
o Data integrity check in terms of format check and
HDDR status monitoring.
o Truncation (windowing) of scans.
2.5.3 P̲r̲e̲p̲r̲o̲c̲e̲s̲s̲i̲n̲g̲ ̲
The required preprocessing functions are:
o Estimation of missing values
o Beam steering (option)
o Resampling in the scan-direction
o Resampling in the trace-direction
2.5.3.1 E̲s̲t̲i̲m̲a̲t̲i̲o̲n̲ ̲o̲f̲ ̲M̲i̲s̲s̲i̲n̲g̲ ̲V̲a̲l̲u̲e̲s̲
Operator-specified missing traces are estimated according
to the operator specified function, selected among
o 4 point reconstruction with sin x/x approximation
o Two-point interpolation
o Nearest neighbour
o zero
Estimation of missing values is applicable to seismic
data only.
2.5.3.2 B̲e̲a̲m̲ ̲s̲t̲e̲a̲r̲i̲n̲g̲ ̲(̲o̲p̲t̲i̲o̲n̲)̲ ̲
The beam steering function involves estimation of sample
values according to a time shift function by applying
a 4 point sinx/x reconstruction function.
The desirable time shift funcions in decreasing order
of priority are
o Constant in time and space
o Constant in time, trace-specific
o Trace-and scan-specific
2.5.3.3 R̲e̲s̲a̲m̲p̲l̲i̲n̲g̲ ̲i̲n̲ ̲t̲h̲e̲ ̲s̲c̲a̲n̲ ̲d̲i̲r̲e̲c̲t̲i̲o̲n̲
The resampling in the scan direction (spatial resampling)
shall be performed on the conditioned seismic data.
The following applies
n…0f…o…0e… = int ((n…0f…i…0e…-w)/s) + 1
for n…0f…o…0e…: Number of traces after resampling
n…0f…i…0e…: Number of traces before resampling
w: Number of weights
s: Step factor
The following is required
n…0f…i…0e…: up to 1008
w: up to 50
s: 2,3,4,5,6,7 or 8.
The resampling function itself is
p-j
y(n,m) = x (n, s x m + j) x k (s, m, j)
j = 0
for y(n,m): resampled value
sample no. n, resampled trace no. m
w: number of weights, w = 50
X(n, s x m + j) : Original sample value,
Sample no. n, trace no. (s m x
j)
K(s, m, j) : Weight factor no. j, a function of the step
factor s and (optionally) of output trace
no. m
2.5.3.4 R̲e̲s̲a̲m̲p̲l̲i̲n̲g̲ ̲i̲n̲ ̲t̲h̲e̲ ̲t̲r̲a̲c̲e̲ ̲d̲i̲r̲e̲c̲t̲i̲o̲n̲
Resampling in the trace direction shall be performed
on the output from the spatial resampling together
with the auxiliary data:
The following is required
Step factor : 1,2,3, or 4
number of weight factors: up to 65
The resampling function itself is analogous to the
spatial resampling function.
The system shall provide a standard set of weight functions
for each step values on request, but the operator may
also specify the values.
2.5.4 D̲e̲m̲u̲l̲t̲i̲p̲l̲e̲x̲i̲n̲g̲ ̲a̲n̲d̲ ̲o̲u̲t̲p̲u̲t̲
The preprocessed data shall be demultiplexed into trace
oriented format. Headers shall be updated according
to the preprocessing performed and added to the data
according to the SEG-D format specification.
External headers shall not be changed, but shall be
passed directly to the output tape, stripped off or
written onto a separate CCT as specified by the operator.
2.6 P̲e̲r̲f̲o̲r̲m̲a̲n̲c̲e̲
The performance requirements are in terms of turn-around
time.
Turn-around time is to be specified for two examples
of preprocessing
E̲x̲a̲m̲p̲l̲e̲ ̲A̲:̲
Spatial resampling from approx. 1000 channels to 250
(step = 4) with max. 20 weights, combined with a resampling
in time with 65 weights, step = 2.
E̲x̲a̲m̲p̲l̲e̲ ̲B̲:̲
Spatial resampling from approx. 1000 channels to 125
(step = 8) with max. 50 weights, combined with a resampling
in time with 65 weights, step = 2.
Auxiliary data are stripped off in both cases.
The requirement for examples A and B is a turn-around
time of minimum real-time with two times real time
speed highly desireable.
2.6.1 D̲a̲t̲a̲ ̲R̲a̲t̲e̲s̲
The requirements means that the mean-data rate from
the HDDR shall be a minimum of 1̲ ̲M̲ ̲s̲a̲m̲p̲l̲e̲s̲/̲s̲e̲c̲.̲
All these samples will have to be resampled in the
scan-direction. The output from this processing is
in
example A: 250 ksamples/sec. and
examble B: 125 ksamples/sec.
The subsequent resampling in the trace direction provides
the output rate in
example A: 125 ksamples/sec.
example B: 62.5 K samples/sec.
The mean output recording rate is slightly higher
due to the ancillary information:
Example A: approx. 400 kbytes/sec.
Example B: approx. 200 Kbytes/sec.
2.6.2 P̲r̲o̲c̲e̲s̲s̲i̲n̲g̲ ̲R̲a̲t̲e̲
The gross processing rate is achieved by multiplying
the number of output samples from each resampling process
with the number of elementary operations i.e. floating
point multiply (FMUL) and floating point add (FADD)
to be executed for each output sample.
With reference to subsection 2.5.3.3 and 2.5.3.4, each
output sample requires as many FMUL and FADD as the
number of weight factors, and this is valid for resampling
in both directions.
Hence, for example A:
20 weights x 250 ksamples/sec. gives 5 M(FMUL+FADD)/sec.
65 weights x 125 ksamples/sec. gives 8̲ ̲M̲(̲F̲M̲U̲L̲+̲F̲A̲D̲D̲)̲/̲s̲e̲c̲.̲
Total, example A approx. 1̲3̲ ̲M̲(̲F̲M̲U̲L̲+̲F̲A̲D̲D̲)̲(̲s̲e̲c̲.̲
̲
Similarly for example B:
50 weights x 125 ksamples/sec. gives 6M(FMUL+FADD)/sec.
65 weights x 62.5 ksamples/sec. gives ̲4̲M̲(̲F̲M̲U̲L̲+̲F̲A̲D̲D̲)̲/̲s̲e̲c̲.̲
total, example B approx. 1̲0̲M̲(̲F̲M̲U̲L̲+̲F̲A̲D̲D̲)̲/̲s̲e̲c̲.̲
2.7 S̲u̲p̲p̲o̲r̲t̲
In addition to delivery of the system the following
support functions are to be considered:
o Training of maintenance personnal
o Installation
o On-call engineering support
3. P̲R̲O̲P̲O̲S̲E̲D̲ ̲T̲E̲C̲H̲N̲I̲C̲A̲L̲ ̲S̲O̲L̲U̲T̲I̲O̲N̲
The proposed technical solution is presented as a baseline
system which fulfills at least the basic requirements
to turn-around time, functions etc.
It is also shown how the options and expansions are
easily added on to the basic system.
Finally is described the overall performance.
3.1 B̲a̲s̲e̲l̲i̲n̲e̲ ̲S̲y̲s̲t̲e̲m̲
The baseline SDPS is a CR 80 M computer with dedicated
modules for the high speed data processing.
The system is divided into two units.
The Processor Unit (PU) performs
o System control and monitoring
o Retrieval of raw data from HDDT
o Preprocessing of raw data
o Ancillary data update
The storage and I/O Unit performs, under supervision
of the PU:
o Intermediate buffering and demultiplexing of data
to output format
o Recording onto 6250 bpi CRT
o Background storage on disk of program, system log,
standard parameters etc.
3.1.1 S̲y̲s̲t̲e̲m̲ ̲O̲p̲e̲r̲a̲t̲i̲o̲n̲
The operator controls all system function from a VDU
terminal.
The only manual interventions in normal operation is
mounting and demounting of tapes.
Selection of HDDR, start/stop and search for specific
files is performed under software control. The operator
is alerted by a message on the terminal and an audible
start when a tape has to be changed, operation completed
etc. The same applies to the 6250 bpi CCT's. The
program, running under DAMOS, the CR80 mapped operating
system, is stored on the disk together
with standard parameters, menues and system log. The
system log on disk reflects all commands and system
responses given, all tagged with real-time and date.
The log can be recalled for display on the VDU and/or
printed on paper.
Paper labels for the CCT-reels are printed for each
recorded reel with e.g. tape identifier, time and date
of production, preprocessing code etc. Optionally,
the labels could be printed on a separate printer with
adhesive labels.
3.1.2 D̲a̲t̲a̲ ̲f̲l̲o̲w̲
The normal operation comprises retrieval of raw, multiplexed
data from HDDT, preprocessing and recording of preprocessed
data in demultiplexed format onto 6250 bpi CCT.
See figure overleaf.
Data are retrieved from the HDDT at a constant rate
through the HDDR Controller Adapter HDTA.
The HDTA provides for the adaptation of the signals
and the switching between the HDDR's.
The HTA connects directly to the HD Tape Controller
(HDTC), which performs the separation into Ancillary
data (headers) auxiliary data and seismic data. Filler
data are removed and truncation is performed, if applicable.
The ancillary data are transferred to the PU main store
by the CPU while seismic and ancillary data, with the
latter re-aranged to form the trailer of the scan,
is transferred to the Floating Point Processor (FPP)
module via a dedicated, high-speed data bus.
The FPP module performs the conditioning of data (estimation
of missing traces) and the resampling in the scan-direction.
The resampling in the trace-direction is performed
in the next (hardware -wise identical) module.
Prior to this, data have been buffered in the built-in
RAM, capable of storing the necessary number of scans
which must be available for the processing.
The processing parameters, weight factors etc. have
been loaded into the FPP modules prior to the start
of processing, fully software-controlled, while the
high-speed processing of the FPP modules is fully controlled
by the built-in processor, with the only CPU-load being
a regular status monitoring.
The fully preprocessed data are transferred via a high-speed
DMA channel to the storage and I/O Unit.
The transfer is executed by hardware in the DACA and
DCI once a data buffer has been released by the FPP.
The updated headers are transferred to the memory by
CPU-initiated DMA transfers.
The Mag Tape Controller performs the demultiplexing
to trace-oriented format by reading data in the proper
sequence.
The memory size of 5M 16 bit words allows double-buffer
operation for continuous operation of files of size
according to approximately 10 seconds of data with
a 1 - to - 8 reduction in the preprocessing.
3.1.3 O̲p̲t̲i̲o̲n̲s̲ ̲a̲n̲d̲ ̲E̲x̲p̲a̲n̲d̲a̲b̲i̲l̲i̲t̲y̲
Options and expansions are easily accomodated due to
the very modular concept of the CR80M architecture.
3.1.3.1 C̲C̲T̲ ̲I̲n̲p̲u̲t̲ ̲a̲n̲d̲ ̲H̲D̲D̲T̲ ̲O̲u̲t̲p̲u̲t̲
The necessary hardware will be available in the baseline
system. Implementation of the functions requires only
the corresponding software modules.
3.1.3.2 B̲e̲a̲m̲ ̲S̲t̲e̲e̲r̲i̲n̲g̲
The process of beam steering can be performed as a
separate process to be made before the resampling in
the scan direction. The option can be implemented
by adding one more FPP in the chain.
In addition, software will be needed for calculation
of the weight functions. Beam steering with constant
delay and beam steering with different time shift for
every spatial resampled trace, but constant in time
are both possible
3.1.3.3 I̲n̲c̲r̲e̲a̲s̲e̲d̲ ̲P̲r̲o̲c̲e̲s̲s̲i̲n̲g̲ ̲S̲p̲e̲e̲d̲ ̲
The processing speed can be increased by adding more
FPP's in parallel. Up to four FAP's in parallel is
possible.
The parallel operation is easily accomplished by letting
each branch in the parallel path operate on a half,
third or a quarter of the number of traces.
3.1.4 P̲e̲r̲f̲o̲r̲m̲a̲n̲c̲e̲
The turn-around time is calculated for the exaples
A and B of subsection 2.6.
The turn-around time is determined by:
t…0f…s: time for set-up of process description etc.
t…0f…i…0e…: Initialization time
t…0f…sc…0e…: Processing time for each scan.
Change of Mag. Tape will influence only negligibly
since two drives in pin-pong operation are assumed.
Change of HDDT is not considered either, since this
will in any case be handled as a new job.
Depending on the streamer configuration, tape density
etc. a standard 9000 feet reel will last for about
half an hour real-time, so the contribution from HDDT
change on turn-around time will be negligible.
We then have:
Turn-around time:
T…0f…ta…0e… = t…0f…s…0e… + t…0f…i…0e… + m x n x t…0f…sc…0e…
for a number of m files, each with n scans.
This set-up time, t…0f…s…0e…, includes selection of menu'es,
input of tape identifiers, label text etc. This time
will be highly variable depending on how well the standard
menu'es are applicable.
It is assumed that a few minutes will be sufficient
as an average figure. It should be noted that the
set-up can (and, presumeably, normally will) be made
as a background task while the previous job is in progress.
The initialization time, t…0f…i…0e… covers the time needed
for setting up new parameters in the processing modules
and start up of HDDR. This time will be determined
by the acceleration time of the HDDR, again a function
of the speed.
Assuming a tape speed of 60 inch per second, t…0f…i…0e… will
have a value of 5 to 8 seconds.
The processing time, t…0f…p…0e… for each file is limited either
by the processor speed of 8 Mega multiply in parallel
with 8 mega additions, or by the channel speed of 3
mega 16 bit words per second.
For example A and B of subsection 3.6, the limit will
be set by the processor to a maximum rate of 1 Msmples/sec.
In the general processor-limited case, the processing
time t…0f…sec…0e… will be determined by the number of multiply/add
operations to be performed, and as such be a function
of the number of channels, weight factors, missing
lines etc.
Specifically, in example A we have:
for the resampling in the scan-direction, approx. number
of multiplications per scan
n…0f…m…0e… = 1/4 x 1000 x 20 = 5,000
Number of additions is the same figure
n…0f…a…0e… = 5,000
For the resampling in the trace-direction we have:
n…0f…m…0e… = 1/2 x 250 x 65 = 8,125
and n…0f…a…0e… = n…0f…m…0e…
Considering the architecture with the two resampling
processes pipelined and the performance of the Floating
Point Processor Module of 8 Mega multiply/add per second,
we have the processing time for each scan:
8,125
t…0f…sc…0e… = ̲ ̲ ̲ ̲ ̲ ̲ or approx. 1 milisecond.
8M/sec
Or in other words, the preprocessing can be performed
at r̲e̲a̲l̲-̲t̲i̲m̲e̲ ̲r̲a̲t̲e̲
For example B, the limiting process will be the resampling
in the scan direction.
The number of multiply per scan is
n…0f…m…0e… = 1/8 x 1000 x 50 = 6,2 50
This process will therefore be possible to perform
at a rate s̲l̲i̲g̲h̲t̲l̲y̲ ̲h̲i̲g̲h̲e̲r̲ ̲t̲h̲a̲n̲ ̲r̲e̲a̲l̲-̲t̲i̲m̲e̲ ̲r̲a̲t̲e̲.
Optionally, the processing rate can be increased by
parallel operation of the Floating Point Processors.
The limitation is ultimately set by the CCT recording
speed of maximum 781 Kbytes per second at 120 inch
per second.
3.2 H̲a̲r̲d̲w̲a̲r̲e̲ ̲s̲p̲e̲c̲i̲f̲i̲c̲a̲t̲i̲o̲n̲
The hardware covered by this proposal covers three
types of equipment
o In-house standard items
o Specially developed modules
o Bought-out equipment
The standard items are described by the corresponding
data sheet, formed in the appendix A of this section.
The modules to be developed is described by their
main functions and block diagram, formed in subsection
3.2.3.
3.2.1 P̲r̲o̲c̲e̲s̲s̲i̲n̲g̲ ̲U̲n̲i̲t̲
The constituents of the PU is a basic crate configuration
with the following modules:
1 CR 8050 M Power Supply
1 CR 8002 M CPU
1 CR 80 16 M RAM, 128 K
1 CR 8020 M MAP
1 CR 8071 MSA
1 CR 80xx MD Tape Controller
2 CR 80xx Floating Point Processor
1 CR 80xx HD Tape Adapter
1 CR 80xx Data Channel Adapter
The modules designated CR 80xx are modules to be developed
for this project.
The peripherals are
2 High Density Digital Recorders. The assumption
is that these recorders are processed by the customer
1 CR 8350 VDU terminal
1 CR 8390 Matrix printer
3.2.2 S̲t̲o̲r̲a̲g̲e̲ ̲a̲n̲d̲ ̲I̲/̲O̲ ̲U̲n̲i̲t̲
The storage and I/O unit consists of a basic crate
configuration with the following modules:
1 Power supply
5 1 MW RAM modules
1 Mag. Tape controller, G 250 bpi
1 Disk controller
1 6250 Tape Adapter
1 Disk controller adapter
The peripherals are
2 6250 bpi tape drives
1 Disk, 40 mbyte.
3.2.3 D̲a̲t̲a̲ ̲S̲h̲e̲e̲t̲s̲
On the following pages if found a brief description
of the modules/equipment to be developed.
3.2.3.1 H̲D̲D̲R̲ ̲T̲a̲p̲e̲ ̲A̲d̲a̲p̲t̲e̲r̲ ̲H̲D̲T̲A̲
The HDTA module is located in the rear magazin and
constitutes the interface between the HDDR tape stations
and the Tape Controller module. The HDTA module contain
a data interface circuit for data to be read/written
to the tape and the interface circuit required for
control of the tape station.
Monitoring and control of the adapter and thereby the
tape station is performed from the controller Module.
Characteristics:
o Dimensions comply to CR80 adapter module standard
o Electrical and physical interface to HDDR Tape
Station
o Transfer rate higher than 2.5 mega 24 bits words
per second.
6̲2̲5̲0̲ ̲B̲p̲i̲ ̲T̲a̲p̲e̲ ̲A̲d̲a̲p̲t̲e̲r̲ ̲(̲6̲2̲5̲0̲ ̲T̲A̲)̲
The 6250 TA module is located in the rear magazine
and constitutes the interface between the 6250 BPI
tape station and the Tape Controller Module. The module
contain a data interface circuit for data to be read/written
to the tape and the circuit required for control of
the tape station. Monitoring and control of the adapter
and thereby the tape station is performed from the
Tape Controller Module.
Characteristics:
o Dimensions comply to CR80 adapter module standard
o Electrical and physical interface to 6250 BPI Tape
Station.
o Transfer rate higher than 1.25 mega bytes second.
3.2.3.2 T̲a̲p̲e̲ ̲C̲o̲n̲t̲r̲o̲l̲l̲e̲r̲ ̲M̲o̲d̲u̲l̲e̲ ̲(̲T̲A̲P̲E̲C̲T̲R̲L̲)̲ ̲
The TAPE CTRL is located in the front magazine and
constitutes the interface between the CR80 transfer
bus structure and the Tape Station via the Tape Adapter
module. The modules for HDDR and 6230 BPI are identical
in the hardware. The module is designed around a bit
slice controller (16 bit) Master Control which monitors
and controls the module operation as a function of
setup commands loaded into the communication RAM from
the CR80 CPU via the Channel Bus/Databus A. The Communication
RAM is also used for status messages to the CR80 CPU
and for temporary storage of data to be transferred
to/from RAM memory on the Channel Bus/Data Bus A.
Data transfer between memory and tape station is performed
by means of DMA controllers, one for F.P. Data Bus
I and another for Channel Bus/Data Bus A, controlled
on a monitoring from the Master CTRL.
To avoid synchronization problems between the CR80
buses and the Tape Station a four words first in first
out (FIFO is included.
Characteristics:
o Dimensions comply to CR80 standard front module
o Two different versions, one for HDDR and one for
6250 BPI Tape Station
o Dual CR80 Bus interface Channel Bus/Data Bus A
and FP Data Bus I.
o DMA Transfer rates
Channel Bus/Data Bus A: higher than 1 mega 16 bit
words
F.P. Data Bus: higher than 2 mega 32 bit words
per sec.
o Communication with CR80 CPU via RAM are DMA Buffer
descriptors
o Multiple commands from CPU by chaining.
Figur
Figur Block Diagram
3.2.3.3 D̲a̲t̲a̲ ̲C̲h̲a̲n̲n̲e̲l̲ ̲A̲d̲a̲p̲t̲e̲r̲ ̲(̲D̲A̲C̲A̲)̲
The DACA module is located in the rear magazine and
constitutes as the interface between the F.P. Data
Bus II (Floating Point Processor) and the CR80 Data
Channel. The transfer is bidirectional controlled
from the Floating Point Processor and is either by
single or block transfer. Data transferred to and
from is parity checked.
C̲h̲a̲r̲a̲c̲t̲e̲r̲i̲s̲t̲i̲c̲s̲
o Dimensions as CR80 standard adapter module. Supports
single word or block transfers.
o Transfer rates with data channel length 1 m:
o Single word: higher than 1 mega 16 bit words
Block transfers: higher than 3 mega 16 bit Words
during data transfer time on
data
channel. The throughput is deter-
mined by memory access time.
3.2.3.4 D̲a̲t̲a̲ ̲C̲h̲a̲n̲n̲e̲l̲ ̲I̲n̲t̲e̲r̲f̲a̲c̲e̲ ̲(̲D̲C̲I̲)̲
The DCI is located in the rear magazine and constitutes
the interface between the Data Channel and Data Bus
A and Data Bus B.
The module is controlled from the Data Channel and
can operate either in single word (16 bit) or block
transfer mode.
To accomodate for the access time variation when accessing
the Data Bus memory, a FIFO is included. The conversion
from the 8 bit Data Channel format to the 16 bit format
on the Data Bus is also via the FIFO. The synchronization
and address control is performed by the Memory Access
Control.
Characteristics:
Dimensions as standard CR80 channel interface module
supports both single word and block transfers.
Transfer rates:
Single word: higher than 1 mega 16 bits words
per second
Block transfers: higher than 3 mega words per second
on the Data Channel. Actual
throughput determined by memory
access time.
Addressing of up to 15 mega words on the Data Bus.
Fig. Block Diagram.
3.2.3.5 1̲ ̲M̲W̲ ̲R̲a̲m̲
The RAM module is located in the front magazine and
provides the bulk memory in the system.
The module is dual ported to allow for access from
both Data Bus A & Data Bus B.
Characteristics:
o Dimensions comply to CR80 standard front module
o Dual ported access
o Memory size 1 Mega (16 + 2) bit words
o Mean access time 500 MS
Fig. 1MW RAM Block Diagram
3.2.3.6 F̲l̲o̲a̲t̲i̲n̲g̲ ̲P̲o̲i̲n̲t̲ ̲P̲r̲o̲c̲e̲s̲s̲o̲r̲ ̲(̲F̲P̲P̲)̲
The FPP is located in the front magazine and performs
high speed floating point multiply/add. The module
interfaces to three CR80 buses, the Channel Bus for
loading of factors and module set-up commands from
the CR80 CPU, the F.P. Data Bus I for loading of input
data from the Tape Controller or the other FPP and
the F.P. Data Bus II for output of results to the bulk
memory or the other FPP.
The module contains two RAM areas for storage of sensor
data (64 K) and weight f̲a̲c̲t̲o̲r̲s̲ (32K). When data have
been loaded into these two RAMS the MULT/ADD SET UP
CTRL transfer the data to the two CACHE memories (double
buffered).
Execution of the arithmetic operations is controlled
from MULT/ADD CTRL, which generate the CACHE memory
address sequencies and the necessary signals for the
multiplier, adder and pipelining registers.
The results are transferred to F.P. Data Bus II controlled
from the DMA CTRL which by means of setups from the
MASTER CTRL can maintain more data buffers simultaneously.
The arithmetic sequences/operations to be performed
as controlled by the Master CTRL as a function of set
up commands loaded into the communication RAM. This
RAM is also used for status to the CPU.
Characteristics:
o Dimensions comply to CR80 standard front modules
o Interface to Channel Bus and F.P. Data Bus I and
F.P. Data Bus II
o Arithmetic Section:
o M̲u̲l̲t̲i̲p̲l̲i̲e̲r̲ operands: 8 bit exponent, 24
bits mantissa, results:
8 bits exponent, 48 bit
mantissa.
o A̲d̲d̲e̲r̲.̲ Operands: 8 bits exponent,
48 bits mantissa, results:
8 bits exponent, 48 bits
mantissa.
o Basic speed for each section 125 ns giving 8 mega
floating point operations per second by utilizing
pipe lining.
o Data transfer with buffering between the two E.P.
Data Bus'es can be performed without involving
the arithmetic sections.
Fig. Block Diagram
3.3 S̲o̲f̲t̲w̲a̲r̲e̲ ̲S̲p̲e̲c̲i̲f̲i̲c̲a̲t̲i̲o̲n̲
This section describes the software which impelemnts
the required SDPS functions. The software is implemented
in four groups:
- Standard Software
- Basic Application Software
- Preprocessing software
- Optional Software
The overall structure and the software are shown on
the fighre overleaf.
Section 3.3.1 gives the details of the standard software
which forms the basis for the applications. The basic
application software is described in section 3.3.2
while the preprocessing software is described in section
3.3.3.
A modular structure has been adapted for the application
software in order to achieve a solution which allows
maximum reuse of software during later implementation
of any of the options as well as stepped `upgrading
of the baseline system to the maximum configuration.
3.3.1 S̲t̲a̲n̲d̲a̲r̲d̲ ̲S̲o̲f̲t̲w̲a̲r̲e̲
DAMOS, the CR80M Advanced Multiprocessor Operating
System, is a virtual memory operating system kernel
for the mapped CR80M series of computers. DAMOS fully
supports the CR80M architecture which facilitates fault
tolerant computing based on hardware redundancy. DAMOS
supports a wide range of machines from a single Processing
Unit (PU) having one CPU and 128K words of main memory,
up to a maximum configuration of sixteen PUs each PU
having five CPUs and 16.384K words of main memory plus
a virtually unlimited amount of peripheral equipment
including backing storage.
DAMOS is ideally suited for use in real time systems
but also supports other environments like software
development and batch. The main objectives fulfilled
in DAMOS are high efficiency, flexibility and secure
processing.
DAMOS consists of many layers of software where each
layer offers a service to the higher layers. The lowest
level is the DAMOS Kernel which implements fault tolerant
processes and interprocess communication.
The DAMOS Page Manager is responsible for memory allocation
in a local Processing Unit and for activating appropriate
disk processes to transfer data from disk pages to
memory pages and vice versa.
The next layer consists of DAMOS device handlers, which
are software processes handling physical devices like
communication lines, line printers, terminals, disks,
and magnetic tape drives.
The DAMOS file Management System offers logical structuring
of physical disks into files
The DAMOS I/O system provides normal application programs
with a standardized and device independent interface
to all peripheral devices including files on disk storage.
All devices are handled as block oriented devices.
A comprehensive description of DAMOS is available upon
request.
3.3.1.1 C̲R̲8̲0̲D̲ ̲S̲u̲p̲p̲o̲r̲t̲ ̲S̲o̲f̲t̲w̲a̲r̲e̲
The CR80M support Software consists of a variety of
program development and support tools. This section
describes those that have been found most important
for the performance of SDPS software. A more comprehensive
description of the CR80M support Software is available
upon request.
3.3.1.1.1 T̲e̲r̲m̲i̲n̲a̲l̲ ̲O̲p̲e̲r̲a̲t̲i̲n̲g̲ ̲S̲y̲s̲t̲e̲m̲
The Terminal Operating System (TOS) is a high level
operating System that supports multiple interactive
terminal users during program development and maintenance.
3.3.1.1.2 L̲a̲n̲g̲u̲a̲g̲e̲ ̲P̲r̲o̲c̲e̲s̲s̲o̲r̲s̲ ̲
The CR80M language processors include the following:
a. PASCAL is a high level block-oriented language
that offers structured and complex data and enforces
well structured programs. The CR80M implementation
is based on standard Pascal as defined by Kathleen
Jensen & Niklaus Wirth, with only minor deviations.
The CR80M implementation provides for bit mask
operations in addition to standard PASCAL data
structures. Furthermore, the CR80M implementation
provides the following powerful additions:
1. Compile time option enables mergin assembly
object directly into the Pascal module.
2. Overlay technique is supported.
3. Built-in Trace of program execution may optionally
be switched in/out for debugging purposes.
4. Sequential and random file access is available
from run time library.
b. The CR80M COBOL compiler is an efficient industry-compatible
two-pass compiler, fulfilling American National
Standard K3.23-1974 level 1 as well as most of
the level 2 features. COBOL is not included in
the baseline offer.
c. SWELL 80 is a Software Engineering Low level Language
for the CR80M minicomputer. SWELL offers most
of the data and program structures of PASCAL, and,
by enabling register control, is without the efficiency
penalties experienced in true high-level languages.
The main purpose of SWELL is to combine efficient
program execution with efficient program development
and maintenance.
d. The assembler is a machine-oriented language for
the CR80M. The language has a direct correspondence
between instructions read and code generated.
e. ADA compiler. A project has been launched for
implementation of the new DOD standard programming
language ADA on the CR80M machine. The project
is planned for completion in 1983 and includes
development of an ADA compiler hosted on and targeted
for the CR80M as well as of an ADA programming
support environdment. The programming support
environment is based on the Stoneman report.
3.3.1.1.3 S̲y̲s̲t̲e̲m̲ ̲G̲e̲n̲e̲r̲a̲t̲i̲n̲g̲ ̲S̲o̲f̲t̲w̲a̲r̲e̲
The utility SYSGEN-EDIT generates object files -- based
upon a set of directives, a system source, and command
files -- for subsequent compiling and linking. A BINDER
the binds the system object together with the application
object based upon a command file from SYSGEN-EDIT.
All the external references of the object modules
are resolved in the Binder output, which is a load
module ready for execution. The BINDER produces a
listing giving memory layout, module size, etc.
3.3.1.1.4 D̲e̲b̲u̲g̲g̲i̲n̲g̲ ̲S̲o̲f̲t̲w̲a̲r̲e̲ ̲
The software debugging facilities include:
o Test Output Facility
o On-line interactive debugger
3.3.1.1.5 U̲t̲i̲l̲i̲t̲i̲e̲s̲
The CR80 utility softwar package will include:
o Editor
o File copy and compare
o File merge
o Interactive propoer patch facility
o File Maintenance program
o FORTRAN-to Pascal converter
The FORTRAN - to - Pascal Converter will provide a
automated one-to-one conversion between FORTRAN statements
and Pascal statements.
3.3.2 B̲a̲s̲i̲c̲ ̲A̲p̲p̲l̲i̲c̲a̲t̲i̲o̲n̲ ̲S̲o̲f̲t̲w̲a̲r̲e̲
The basic application softwar consists of
o Man-machine interface
o System management
o HDDR control
o HDDR Data handler
o Quality monitoring, logging
o SEG-D output formatting
o Cataloguing and labelling
3.3.2.1 M̲a̲n̲-̲M̲a̲c̲h̲i̲n̲e̲ ̲I̲n̲t̲e̲r̲f̲a̲c̲e̲
The MMI provides the operator with a simple man-machine
interface which only requires the operator to select
one of the defined sets of operating modes followed
by the propoer requested parameters. It directs the
operator if and when he has tomount/dismount HDDT's
and CCT's or `during other manual operations which
the system does not control. It provides status information
and error messages on the operator's CRT. In all,
it implements a uniform approach to operating the system
and facilitates implementation of a minimum labour
intensive interface resulting from the reduced complexity
of operations required to control the system. Furthermore
the hierarchical structure of menues facilitates menus
to be added.
The menu is the central part of the MMI. Each menu
consists of a set of functionally connected `system
actions. Each action may identify a function or capability
of the system and is presented by one or more CRT lines.
This enables a characteristic identification to be
employed for each action. Alternatively each action
may represent a parameter associated with a preceeding
selection, e.g. speed selection on a CCT recording.
Default parameter values are employed to facilitate
the operator's data entry; only parametes different
from the default need to be keyed in.
The menu is displayed in the middle of the operator's
CRT-screen. The upper line is reserved for system
messages while the last two lines are reserved for
error messages and replication of the operator's input.
The operator keys in his selection among the `displayed
actions. The selection is checked for format and content
before a new (resulting) menu is displayed. An error
message, if necessary, is displayed on the last line
and an audible alarm is produced. The error message
provides a short description of the cause.
Data from the Man-Machine Interface is passed to the
System Manager module `when all selections and parameters
required to execute a production or alternatively a
background job have been entered.
A production menu provides the basis for schedule planning.
this menu porovides the operator with the estimated
time of start and completion for each production order.
Any production order may be interrupted or cancelled
even if it is ongoing. A special password has to be
used in order to cancel a production order.
3.2.5.4 S̲y̲s̲t̲e̲m̲ ̲M̲a̲n̲a̲g̲e̲m̲e̲n̲t̲
The System Manager implements central control of the
SDPS system by having overall responsibility of the
entire configuration, production scheduling and processing.
Production orders are contained in the order stak maintained
by the System Manager module. Up to twelve production
orders may be stacked at any time. The first of these,
the primary task, is the one currently in execution.
KThe System Manager module controls and monitors the
execution of a production order with the system mostly
operating in automatic mode. It provides the operator
with proper status information (e.g. which production
is running) as well as requests mount/dismount of HDDT's
and CCT's.
Once ae production order is completed, the System Manager
signals the products and order catalogues. The next
production order, if any, on the stack is automatically
initiated, dependent on its current state.
The System Manager allows background jobs, e.g. system
management, to be submitted concurrently with the production
jobs. However, the later of these are priority jobs
which are allocated resources whenever a conflict between
the two types exists
Other modules provide the System Manager with status
reports at regular intervals on their fitness. Lack
of this report or indication of an un-recoverable error
results in a error message.
3.3.3 S̲e̲i̲s̲m̲i̲c̲ ̲D̲a̲t̲a̲ ̲P̲r̲e̲p̲r̲o̲c̲e̲s̲s̲i̲n̲g̲ ̲S̲o̲f̲t̲w̲a̲r̲e̲
This software is comprising
o Ancillary data updating, i.e. updating of headers
to reflect the performed processing
o Resampling in space covers the handler for the
Floating Point Processor used for this resampling.
o Resampling in time covers the handler for the Floating
Point Processor used for this resampling
o Processing Parameter calculation is the SW package
for provision of the various constants for the
processing, either by table look-up or by algorithmic
calculation.
4. T̲R̲A̲I̲N̲I̲N̲G̲
The training section, which is part of the Integrated
Logistic Support Department, is responsible for the
development and conduct of lcustomer training.
The following `course description of the Operation
and Maintenance Course is a combination of standard
CR80 system training and specific customer training,
specially assigned personnel involved with the operation
and maintenance of the system.
4.1 C̲R̲8̲0̲ ̲O̲p̲e̲r̲a̲t̲i̲o̲n̲ ̲a̲n̲d̲ ̲M̲a̲i̲n̲t̲e̲n̲a̲n̲c̲e̲ ̲C̲o̲u̲r̲s̲e̲
4.1.1 S̲c̲o̲p̲e̲
After course the students are able to
o Operate the system
o Run applicable Maintenance and Diagnostic Software
o Repair the hardware to module (card) level
o Load and execute applications software
o Patch the sytem and application software
4.1.2 D̲e̲s̲c̲r̲i̲p̲t̲i̲o̲n̲ ̲o̲f̲ ̲t̲h̲e̲ ̲C̲o̲u̲r̲s̲e̲
System description
Operation of the system
- System Initialization
- Job Execution
- Controls and Indicators
System Software
Software Utilities
Hardware Module Description
- CR80
- Peripherals
System Troubleshooting
The number of participants is max. 10 persons.
The previous knowledge for the course shall be minimum
of 3 years experience as a computer technician and
good ability to communicate in English.
The course duration is two weeks.
