The VEN-3000 FEC
The VEN-3000 is equipped with FEC which complies with the SMPTE 2022 and the Pro-MPEG Forum standards to ensure co-operation with third party encoders and decoders over private networks.
The VEN-3000 FEC can be configured for much larger matrix sizes and custom matrix shapes enabling the user to transmit video over the internet and other networks which typically exhibit high error rates and long burst errors.
The VEN-3000 FEC matrix can be configured up to 20 times the size (2000 IP packets) of SMPTE 2022 matrix enabling the user to configure error burst correction up to 250 IP packets with overhead varying between 12.5% and 50%. The user may also configure the overhead adapting the FEC matrix correct for most network errors. The user can configure any matrix shape by configuring number of columns and rows in the matrix as long as the Column x Row number does not exceed 2000.
The VEN-3000 FEC summary specifications:
- Max matrix size: 2000
- Max burst error correction: 250 IP packets
- Column range: 2 to 250
- Row range: 2 to 20
- Column*Row <= 2000
An overview of the characteristics of different FEC Matrix sizes and shapes can be found in Figure 1.
- The number of Matrix columns is equal to the error burst length the FEC can correct.
- A larger “square” Matrix has lower overhead but also has lower error correcting capability.
- Matrixes with “few rows” have large overhead and error correction capability.
- “Small” Matrix sizes have low latency, high overhead, no burst error correction capability, but can correct a high rate of single packet errors.
- “Large” Matrix sizes are required to correct burst errors.
- A “large square” Matrix can have low overhead, low single packet error correction capability for random packet loss, with the cost of longer latency. More and more IP networks today fit this profile enabling the user to trade overhead (save bandwidth) with increased delay.
Example: A powerful FEC comparative to the FEC + ARQ Hybrid discussed earlier, using 20% overhead which is capable of correcting a packet error rate up to 20% and correct burst errors up to 50 IP packets can be achieved with a 50 Column by 5 Row Matrix. The added FEC latency is 0.7 seconds at 10Mb/s data rate.
Guidance for shaping the FEC Matrix
- For networks with high number of “non-existent or very short bursts errors”: Use more Rows than Columns.
- For networks exhibiting infrequent but “Long burst errors”: Use more Columns than Rows
- For networks generating “Non-consistent” IP packet drops: Use a Square Matrix (equal Columns and Rows)
Note: Since the FEC processes the packets in rows, options 2 & 3 provide the same protection as option 1 provides.
- Set the number of column for the anticipated max burst length.
- Set the number of Rows to balance overhead and the need to recover random packet losses.
Start out with a 20 column by 5 row matrix and use the VEN-3000 statistics page to adjust the column and row number. If the statistics page shows Dropped Column packets larger than zero, increase the number of FEC columns on the MP configuration page. If the statistics page shows Dropped Row numbers different than zero, decrease the row number on the MP configuration page.
Comparison between the FEC + ARQ Hybrid and the Enhanced VEN-3000 FEC
The FEC + ARQ Hybrid technology double the burst error correcting capability of the Pro-MEPG Cop3 FEC without increasing the bandwidth consumed on the TCP connection. The figure below (Figure 2) shows how larger FEC matrixes can be used enabling the same or better burst error performance compared to the FEC + ARQ Hybrid.
Figure 2: Comparison between the FEC + ARQ Hybrid and the Enhanced VEN-3000 FEC
Figure 2 shows that Enhanced FEC Matrix shapes with 10% overhead can correct burst IP packet errors up to 100 lost IP Packets. A 40 x 10 Matrix can correct up to 40 consecutive lost IP Packets similar to the 20% (20 x 5) Pro-MPEG Cop3 FEC and ARQ Hybrid.
The Enhanced 100 x 5 FEC matrix shape can correct loss of up to 100 consecutive IP packets with a 20% overhead. Read more about the VEN-3000 here.