CDM-600 Satellite Modem Revision 7
Forward Error Correction Options MN/CDM600.IOM
7–3
lower data rates there are sufficient number of computational cycles per input symbol to
permit the decoding process to perform optimally. However, as the data rate increases, there
are fewer cycles available, leading to a reduction in coding gain. This is clearly illustrated in
the performance curves that follow. For data rates above ~1 Mbps, Viterbi should be
considered the better alternative. The practical upper limit at this time is 2.048 Mbps.
Table 7-2. Sequential Decoding Summary
FOR AGAINST
Higher coding gain (1-2 dB) at lower data
rates, compared to Viterbi.
Pronounced threshold effect - does not fail
gracefully in poor Eb/No conditions.
Higher processing delay than Viterbi
(~4 k bits) - not good for low-rate coded voice.
Upper data rate limit approximately 2Mbps
Coding gain varies with data rate - favors lower
data rates.
7.4 Reed-Solomon Outer Codec
IMPORTANT
It cannot be emphasized strongly enough that the purpose of the
concatenated Reed-Solomon is to dramatically improve the BER
performance of a link under given noise conditions. It should NOT be
considered as a method to reduce the link EIRP requirement to produce a
given BER. Factors such as rain-fade margin, particularly at Ku-band, are
extremely important, and reducing link EIRP can seriously degrade the
availability of such a link.
The concatenation of an outer Reed-Solomon (R-S) Codec with Viterbi decoder first became
popular when it was introduced by Intelsat in the early 1990's. It permits significant
improvements in error performance without significant bandwidth expansion. The coding
overhead added by the R-S outer Codec is typically around 10%, which translates to a 0.4 dB
power penalty for a given link. Reed-Solomon codes are block codes (as opposed to Viterbi
and Sequential, which are convolutional), and in order to be processed correctly the data must
be framed and de-framed. Additionally, R-S codes are limited in how well they can correct
errors that occur in bursts. This, unfortunately, is the nature of the uncorrected errors from
both Viterbi and Sequential decoders, which produce clusters of errors that are multiples of
half the constraint length. (This is particularly severe in the case of Sequential, where the
constraint lengths are considerably longer than Viterbi). For this reason, the data must be
interleaved following R-S encoding, and is then de-interleaved prior to decoding. This
ensures that a single burst of errors leaving the Viterbi or Sequential decoder is spread out
over a number of interleaving frames, so errors entering the R-S decoder do not exceed its
capacity to correct those errors. In the case of the CDM-600, different R-S code rates are
used, according to the mode of operation: