ntinuous 
mpera- 
the track 
d later 
sed for 
y other 
ter, etc. 
depth at 
re an- 
27, three 
und 
ause of 
'S at the 
measu- 
no other 
ter sam- 
re avail- 
mpera- 
ration 
nt are 
ents can 
vailable 
s diffi- 
sition 
| mea- 
itten 
On 
ver- 
s Divi- 
each 
  
¢. each set of four CCTs is merged on the IBM 370/165 computer of the 
JRC-Ispra Establishment into one tape. 
On the merged tape data the following operations have been performed: 
channels on different PCM tracks have been aligned, missing frames or 
lines have been substituted with adjacent frames or lines, 
synchronized, and the roll of the aircraft is corrected for. 
For each generation of CCTs visual quality control of the digital image data 
can be performed on a COMTAL 8300 image display unit which is inter - 
faced to a PDP 11/05 minicomputer. 
lines have been 
For the analysis of the OCS data two of the mentioned installations were 
at the disposal of the Applied Informatics Division. The IBM 370/165 
offers more than sufficient processing capabilities in terms of computa- 
tion speed, main w.emory and mass storage devices, but is accessible 
only through batch job entry or inte ractively through slow transmission 
lines and alphanumeric or graphics terminals, which are not well uted 
for display of imagery data. The Division“ own installation, the PDP 11/05, 
is slow and has a restricted storage capacity, but is better suited to in‘er- 
active processing and has a powerful image display unit autacisd to it. 
These advantages of the smaller installation justified its use in the ana- 
lysis of the OCS daia, even if special programming techniques had to be 
developed in order to arrive at reasonable execution times for the pro- 
cessing of images. 
The PDP 11/05 word length is 16 bits, and integer operations are based 
on this word length; floating point operations are mot implemented. A 
FORTRAN compiler is available which does suppcrt arithmetic operations 
on numbers of more than 16 significant bits (sign bit included) czly in 
floating point operations, realized through software routires. Since arith- 
metic operations on numbers with 10 significant bits require Landlinz o: 
intermediate results 20 bit wide, a technique for implemeniing efficicntly 
algorithms using extended pr^cision integer arithmetic, had to be deve- 
loped. Essentially a virtual machine is simulated through a set of subrou- 
tines which translate symbolic virtual machine instructions into PDP 11 
machine code, provide communication between calling programs and the 
generated machine code, and transfer control to the generated code. 
These subroutines can be used in any FORTRAN (or Assembler) program, 
and are partially written in FORTRAN themselves. The characteristics 
of the virtual machine are: 
- Last in - first out arithmetic stack of 32 bit integers. Arithmetic ope- 
rations between two numbers on the stack or between top of the stack 
and l6 bit numbers in memory. Arithmetic shift operations on the stack. 
- Indirect addressing of 16 bit or 32 bit items in memory through "address 
registers" (which are 16 bit words in main memory); instructions for