MPEG Section Filtering

In MPEG-2 digital TV transmission, the video and audio are transmitted in MPEG transport streams. In addition to video and audio, there should be a mechanism to describe the audio/video programs that are contained in the stream, so that a receiver can receive a program and decode it. This is achieved using MPEG sections.

MPEG sections, or ‘private sections’ are nothing but elementary streams that carry data, instead of audio/video. A audio/video elementary stream is packetized into a packetized elementary stream (PES) but a data elementary stream is packetized into ’sections’. A section has its own format with header and other fields, but for this discussion, it is sufficient to know that a section is nothing but a packetized data stream.

MPEG sections can contain any type of data that the broadcaster wishes to insert in the MPEG stream. A section filter is used to filter out the desired sections from the MPEG stream, by matching a set of parameters with the section’s header or data. Section filtering is normally a sub-block inside an MPEG-2 transport demultiplexer and typically implemented in hardware.

Section data is normally send as section tables and the MPEG-2 standard defines a number of them as mandadoty. The mandatory tables describe the programs in a transport stream so that a receiver can find out which PIDs contain a specific program. A program, in this context, would typically include a video PID, audio PID and a PCR PID.

Section Filter example:

A section filter parameter set typically consists of 3 sets of byte arrays; each set is n bytes long, where n is the number of bytes in the message to compare against. The parameter sets are:

1. Coefficients
2. Inclusion Masks
3. Exclusion Masks

Each bit in the coefficient/mask corresponds to the corresponding bit in PSI message section. For example, the first byte of the section is the table ID, so the first byte of the parameter array corresponds to the table ID. For MPEG sections, the third byte of the PSI section is the lower 8 bits of the section length. A typical section filter skips this byte so that the fourth byte of the parameter array corresponds to the fifth byte of the PSI section.

For both inclusion mask and exclusion mask bits, a ‘0′ enables the corresponding message bit to be compared against the coefficient. Disabled mask bits (=’1′) are ignored and not used for comparison.

For a particular section filter to accept a message, all of the message bits marked for inclusion (Inclusion Mask bit = ‘0′) should match the corresponding coefficients and at least one of the message bits marked for exclusion (Exclusion mask bit = ‘0′) should mismatch the corresponding coefficient bits. In other words, an exclusion mask discards a message only if the message bits marked for exclusion exactly matches all the corresponding coefficient bits.

In a typical hardware implementation of section filters (for e.g., in a Settop SOC), the number of section filters are limited – a typical number is 64. The correct use of inclusion masks and exclusion masks allow the user to achieve as much filtering as possible, with minimum number of filters.