The Real Time Wind (RTW) system employed an Apple II computer and Commodore PET
(the one with the tiny kbd, BASIC, and 7K of RAM). A non-coherent
reciever was switched to the
4 antennas in the "Y"
array in sequence at 7.5Hz. Incoming 8 bit data at 32
(3Km) height gates were converted to bit amps (0,1) in 30sec blocks relative
to the average signal in the block. As data were being read in and converted,
running auto and cross correlations of the bit amplitudes continued in the
background. At the same time, bit correlation data (that is, numbers of
"1-matches") from the previous record were being sent out one height at a time
to the PET for numerical full correlation analysis, and its analysis
for the previous height received. As these were received by the Apple
they were written to the incremental 9 track tape drive.
A 5 minute record was 592 bytes (32 D/T etc. header, 32 signals, 32 gain + 32 heights
x 16 bytes).
Bit correlation produces reduced correlation magnitudes, so these had
to be corrected before analysis [rho'= sin(pi/2 rho) ]. Another
problem was that the receiver main bandpass filter came after the
gain control, so when gain was changed (8dB steps) it would "ring".
To mitigate this gates were controlled in fixed groups of 4 height
gates, so only one of the four would be upset by a gain change.
The upset caused a significant loss of data at the affected height
gates - always the same gates- but what was left seemed normal
(and if a gain change was not required, then no problem.)
Because of the gain switching problem, calibrated signals
are not available.
The system was junked in 1999 due to fire damage. There was already
a 10 year overlap with (the much better) new coherent (CRTW) wind
The analysis process is illustrated here