226 
  
  
  
  
  
227 
3.2 Ground Preprocessing 
  
This is the area where most work was necessary and where most 
progresses have been accomplished. 
In the 1979-80 period, the first SAR processors to exist were 
extremely slow in producing a Seasat image (typically 20 hours 
computing time per scene). These experimental facilities were 
mainly devoted to develop and validate the various algorithms, 
using modest computer facilities. The use of array processors 
(mainly the AP 120 B by Floating Point Systems) allowed for 
achieving a reasonable production on an operational basis 
{e.g. at JPL and DFVLR). 
However, the achieved throughput (typically a few hours per 
scene) was still far from corresponding to the requirements of 
the next SAR missions (e.g. SIR-B, Japanese ERS-1 in L-band, 
European ERS-1. in C-bDand, etc...) 
It is not possible here to report all endeavours to develop 
efficient SAR ground processors. Diversity is the keyword: 
from mini-computer plus multi-array processors to data flow 
systems, from dedicated facilities to general purpose super 
computers, all solutions are under detailed assessment. 
This diversity corresponds to various trade-off on 
technological level (e.g. dedicated operational facility as 
opposed to experimental (i.e. flexible) facilities) and also 
reflects the speed of the evolution in the computer industry. 
Without being a driving market, SAR processing is a good theme 
for testing novel computer architectures and to follow all 
these developments, a survey of all existing and/or under 
development SAR processors is continuously maintained (thanks 
to Mr. Gredel) and an updated version is distributed at every 
working group meeting. 
From this survey, it seems that "current" processors under 
development will produce a SAR image in the 20-30 mn, range. 
There is still a significant gap to reach real time operations 
(i.e. 15 sec. per scene). 
-From the meetings held at Frascati and Ottawa, it was clear 
that Japan is particularly active in this field. A meeting in 
Tokyo (November 1983) was mainly devoted to reviewing the 
diversity of the Japanese developments: dedicated processes 
(Mitsubishi), data flow machine (NEC), super computer (Fujitsu) 
abe Jas 
It should be noted that the tendency towards using super 
computers (as mentioned in the minutes of the Tokyo meeting) 
was based on an incorrect appreciation and that more recent 
information confirms that the diversity of the approaches in to 
be maintained.