Digital TV HD System Ilustrated - pág.21

odd number.  It knows from the top total that there is an error in either the
first or second bit, and from the second total that there is an error in either
the second or third bit.  It can therefore determine that the second bit is
wrong, and put it right.
Parity bits can eliminate single bit errors caused by
noise.  The more parity bits that are transmitted, the more chance the receiver
has of correcting any errors that occur in the transmission chain.
Impulsive interference can cause longer bursts of errors. 
To minimise this, bytes of data are resequenced.  The datastream is split into
12 different paths, each with a different delay, and then remultiplexed
together.  Each path takes a multiple of 17 bytes as shown.  The sync byte has
no delay and is used to synchronise the reverse effect at the receiver end of
the chain to put the bytes back into their correct order.
 

 
The bit stream is then further scrambled or «randomised» in
a defined way with every eighth sync byte being inverted to provide a reference
point.
These error correction techniques are applied in three
stages:
1. Outer Coding to minimise burst errors.   Each
MPEG2 bitstream packet of 188 bytes  (consisting of 187 bytes of data followed
by 1 sync. byte) is followed by 16 bytes of error correction data, thus
increasing the packet size to 204 bytes.  This is known as Reed Solomon (RS)
204:188 coding.
2. Interleaving and Scrambling. The bytes of data are
then resequenced to minimise the effects of longer bursts of errors.
3. Inner Coding. Further Forward Error Correction
(FEC) is applied to the input data, the exact amount being under the