Ericsson SPR1100, Handbook, Page: 94 / 256

Forward Error Correction or FEC, enables packets lost during transmission over IP networks
to be recovered by adding extra information to the transmitted data. The particular type of
FEC used on the unit is that specified by SMPTE 2022 1/2.
The key features of ProMPEG FEC are:
• Open standard FEC scheme.
• Increased robustness of transmission.
• Increases network reach through FEC on high loss links.
• Highly configurable depending on bandwidth vs. robustness.
• FEC transmitted separately to media stream.
• Independent of video compression standard (MPEG-2, MPEG-4).
ProMPEG FEC can help to solve the problems caused by missing packets. It is an open
standard for protection of contribution broadcast real-time transmissions over IP networks
by facilitating real-time lost packet recovery at the receive units. It permits flexible
configurations for optimisation requirements. The scheme uses an RTP layer which adds
timing information for sequence correction. FEC packets are transmitted in separate IP
packets. It uses a matrix of media packets to calculate the FEC packets. The matrix size
defined by columns (L) and rows (D). FEC packets are calculated along columns and rows
using the XOR function. FEC can be 1D (Column Only) or 2D (Column and Row). Media and
FEC packets are transmitted on separate IP streams with the Column FEC stream offset
from media stream and has a UDP port number which is the media port 2. The
Row FEC stream is offset from Media stream and has UDP port number which is the media
port 4. This arrangement means that non-enabled FEC receivers can simply ignore
FEC streams and decode the media packets. The FEC data stream is off-set from the media
stream to protect against burst error loss and jitter. At the receiver, lost packets recovered
using the FEC data packets. The Column FEC protects against burst errors and the Row FEC
protects against random errors. ProMPEG FEC recovers lost packets using column and
(optionally) row FEC packets using the XOR function on the remaining packets. Depending
on the distribution and severity of the pack loss not all errors are recoverable.
The overhead which results from ProMPEG FEC transmitting extra packets depends on
whether column or column and row FEC is selected and how many columns and rows there
are. (Note that L = number of columns, D = number of rows.)
Column FEC: 1D FEC Overhead = (L+(D*L))/(D*L) = 1/D + 1
Worst case, 4 rows = 25%
Best case, 20 rows = 5%
Column and Row FEC: 2D FEC Overhead = (D+L+(D*L))/(D*L)
Worst case, 4 x 4 = 50%
Best case, 10 x 10 = 20%
FEC offers two methods of block alignment (also referred to as FEC linearisation) for use
when generating FEC packets: Non Block Aligned and Block Aligned. Both are guaranteed of
being able to correct L errors, sometimes more. The Block Aligned method can however
correct 2L+2 errors; this never happens with Non-Block Aligned.
Non-Block Aligned can in theory have a lower latency at the decoder if it can be guaranteed
that the mode of operation will never change.