ATSC (US Digital (HD)TV standard) Basics

[Doom9]

ATSC Basics

The ATSC DTV (Digital TV) standard uses a 6 MHz channel to transmit a 19Mbit/s stream via terrestrial broadcast, and up to 38 MBit/s using the same 6 MHz channel but using RF cable as support. The modulation scheme used is 8 VSB (Vestigial Sideband Modulation), and 16 VSB respectively (the latter is used for the higher bitrate mode). There is also a DTV extension for satellite transmission (A/80: http://www.atsc.org/standards/a_80.pdf). The standard modulation scheme for ATSC over satellite is QPSK. This means, a single transmitted symbol contains two bits, and every symbol has the same amplitued and is 90° apart from the next. Alternatively, 8 PSK or 16 QAM can be used. The latter modulation schemes allow more data per transmitted symbols, but are also more error prone and optional in the standard. The modulation scheme used, along with the type of error coding decide how many channels can be transmitted on a satellite transponder (circuits that receive the signal from the earth, modulate, amplify and re-transmit it back to the earth. There are about 20 - 30 transponders on a single satellite, and each has a bandwidth between 27 and 72 MHz (most transponders use 36 MHz, which offers about 38 Mbit/s).

A DTV transmission system concists of 4 parts: The video subsystem, where video data is encoded (using standard MPEG-2), the audio subsystem where audio is encoded (AC3 is used), the Service Multiplex and Transport part, where the output of the video and audio subsystem for the different TV chains is coupled together along with ancillary data and control data to create a transport stream. The transport stream is then being coded for error resilience and finally modulated for the appropriate support channel.

Based on elementary streams that fullfill the bitrate/resolution and framerate requirements, the service and transport part takes a number PES (Packetized Elementary Stream) streams that are first being multiplexed together, then combined to a single stream - a so-called transport stream. While creating the transport stream, PSI (Program Specific Information) data tables are added, allowing a receiver to associate audio and video streams belonging to the same TV channel. The transport stream itself is divided into 188 byte long PES packets, and each packet has a 4 byte header. The header contains a 13 byte PID (packet identifier), that uniquely identifies a unique (audio or video) stream within the transport stream. The header also has a scrambling control field, telling the receiver if the packet payload has been scrambled, and a continuity field (a 4 bit rolling counter) associated to TS packets carrying the same PID. This counter allows a receiver to determine if packets of a certain stream have been lost. The transport error indicator field is not commonly used, but could be used to indicate that the packet contains at least one unrecoverable bit error.

ATSC uses a common reference clock, the System Time Clock (STC) for all encoding and decoding processes. The encoder samples the STC to create two timestamps: the PTS (Presentation timestamp) and DTS (decoding timestamp), that tells a decoder when to output / decode a packet. Why do we have two timestamps? Because MPEG-2 allows for bidirectional encoding (b-frames), which requires certain frames to be decoded out of order (as an example b frames reference previous and future frames, and in order to decode, both referenced frames have to be available. So if frame N references frames N-1 and N+1, and N is a b-frame, the decoder has to decode the frames in the following order: N-1, N+1, N, and sends them to the output in the following order: N-1, N, N+1). The STC is represented using a 42 bit counter, and each bit stands for a 37 ns period - so the clock frequency is 27 MHz. For a decoder to reconstruct this clock, the adaptation field in a TS packet header is used.

Transport stream packets are scrambled, then FEC (forward error coding) is applied for error resilience (20 parity bytes are added to each packet). Finally, the packets are interleaved, then trellis coded (using a 2/3 rate trellis coder), before being modulated for transmission.

The ATSC standard supports the following input resolutions and framerates:
1920x1080 at 23.976/24/29.97/30 fps progressive, and 29.97/30 fps interlaced. Allowable display aspect ratios (DAR) are 1:1 and 16:9
1280x720 at 23.976/24/29.97/30/59.94/60 fps progressive with an 1:1 or 16:9 input DAR
704x480 at the same framerates as 1280x720, but in addition to that, 29.97 and 30 fps are also supported in interlaced mode. Both 4:3 and 16:9 DAR is supported
640x480 supporting all previously mentioned framerates in progressive mode, and 59.94/60 fps in interlaced mode, at both 1:1 and 4:3 DAR

Video bitrate can be up to 19.4 MBit/s in standard mode, and 38.8 MBit/s in high data rate mode. Audio bitrate has to be <= 128 kbit/s for a mono audio track, <= 192 kbit/s for stereo audio, and the bitrate of any audio to be decoded simultaneously shall not exceed 576kbit/s (the effective limit is 448kbit/s for a main audio service type)

As you can see, these framerates are only suited for NTSC. However, the ATSC has since created an addendum to the standard (A/63: http://www.atsc.org/standards/a_63.pdf), that adds PAL compatible framerates:

1920x1080 at 25 fps both progressive and interlaced, 1:1 or 16:9
1280x720 at 25 or 50 fps progressive, 1:1 and 16:9
720x576 at 25 or 50 fps progressive or 25 fps interlaced, 4:3 or 16:9
544/480/352x576 at 25 fps progressive or interlaced, 4:3 or 16:9
352x288 at 25 / 50 fps progressive, 4:3 or 16:9

Video and audio constraints are not affected by these changes.

When it comes to satellite transmission, the A/81 standard for Direct-To-Home Satellite Broadcast defines a series of additional resolutions, like 1440x1080 (progressive and interlaced, and supporting all beforementioned NTSC compatible framerates, at 16:9 DAR), and resolutions smaller than D1 (720x480) as 704/640/544/528/480/352x480. Except for 704/640x480, all formats only support 23.976p and 29.97i framerates and 4:3 DAR (the exceptions support all NTSC framerates where applicable: 29.97/30 fps is for interlaced content only, progressive content can have all the beforementioned framerates).

Audio streams contain certain additional information, the service type, which is stored in the audio descriptor of the PSI. There are 2 main and 6 associated service types: complete main (CM), music and effects main (ME), visually impaired (VI), hearing impaired (HI), dialogue (D), commentary (C), emergency (E) and voice-over (VO). If supported by a TV station, the audio service types offer some interesting possibilities to the watcher. An audio program may consits of a combination of a main and an associated service, or a main type alone. Decoders are not required to support simultaneous decoding of a main and associated service type, but knowing that both can be done, it explains the rather vague definition of a maximum bitrate (now we can say that the bitrate of the main service and associated service may not be higher than 576 kbit/s, but since an audio program can consist of only a main type, the maximum AC3 bitrate is effectively 576 kbit/s) In addition to all this, an elementary AC3 stream can have a language code (so multi-language broadcasts are possible). The service types also allow for an interesting way to transmit multiple languages: Rather than using a main service type for a full blown 5.1 audio, the voice track could be removed, and provided using a low bitrate associated service type track. This would allow to broadcast a number of such mono tracks.
The ATSC standard also allows for data transmission (A/90: http://www.atsc.org/standards/a_90-with-att.pdf) and CA (Conditional Access) modules for Pay-TV (A/70: http://www.atsc.org/standards/a_70_with_amendment.pdf). Both extensions are based on the Program and System Information Protocol (PSIP, A/65b: http://www.atsc.org/standards/a_65b.pdf). PSIP is "the glue that holds DTV together". In less fancy terms, it is a collection of tables carried in a TS stream for terrestrial broadcast and contains two main cateogories of information: system information and program data. The former allows navigation and channel access within a TS, and the latter provides information for channel browsing and even selection.

As previously mentioned, PSIP is a collection of tables: There are 4 base tables: System Time Table (STT), Rating Region Table (RTT), Master Guide Table (MGT) and the Virtual Channel Table (VCT). All those tables are labelled with a PID known as base packet identifier. The MGT links to two sets of secondary tables the EIT (Event Information table), and the Extendet Text Tables (ETT).

The STT serves as a time reference for time-of day functions. The RRT can be used for a rating system (thus if properly programmed, your kids can no longer watch violent shows, or whatnot). The MGT is an indexing table for all the other tables. the VCT contains a list of all channels that are currently being broadcast, and their attributes (like channel number and channel name). The EIT tables (referenced by the MGT, and elements within an EIT table are referenced by the VCT) describe events (TV programs) for 3 hours. There can be 128 such tables, allowing for a TV program of up to 16 days (so the TV program for the next 16 days can be carried if all EITs are used). At least 4 EIT tables (4 x 3h = 12h) must be present.

These tables are commonly used to support an Electronic Program Guide (EPG, or also known as Interactive Program Guide - IPG).

Sources: http://www.atsc.org/standards.html