The interleaving process in LRPT is a technique to help recover from burst errors by spreading the interference-
striken bits out over a large range of bits from before and after the burst and hoping that the viterbi decoder can
correct them.
The Metop LRPT specification describes the interleaving process onboard the spacecraft as:-

What does this mean for us on the ground? We have to build the reverse of this structure. The 36-way switches
are still there and they run in the same direction as shown but the shift register delays are arranged in reverse
order. At the top (position 1) is the 35x2048 bit long register all the way down to the 2048 bit long register at
position 35, with position 36 being the straight through connection.

In my software I literally have 35 buffers, from 2048 bytes to 35x2048=71680 bytes, to store the symbols in.
This amounts to nearly 1.3Mbytes! (1+2+3+....+34+35=630, 630x2048=1,290,240bytes). Remember though,
each byte holds one 3-bit soft decision sample at this stage (not actual data bytes). The longest time delay
through this structure is through the 35 M buffer which itself is only fed every 36th symbol, so:-
-- buffer length is 35 x 2048 = 71680 bits
-- bit rate before viterbi decoder is 2 x 72000 = 144000 bits/second
-- buffer feed rate = 144000 / 36 = 4000 bits/second
-- therefore the duration is 71680 / 4000 = 17.92 seconds
This means that a 1 second burst of errors will be spread across 18 seconds worth of data. The bad news is
that if a long burst of interference floods the de-interleaver memory it will take 18 seconds to clear it out...

BACK