Error Correction in ATSC Mobile DTV
I've covered the basics of ATSC A/153 Mobile DTV in previous columns, (see "Mobile DTV 101
," June 11, 2008; and "What to Look for in A153 Mobile DTV Tables
," April 3, 2010).
This month I'll provide a simplified overview of the error correction used in A/153 and offer some practical tips on rolling out Mobile DTV based on what I've learned helping stations with their buildouts. I'll also reference a Rohde & Schwarz white paper on A/153 that describes the nuts and bolts of Mobile DTV.
TURNING DATA IN A ROBUST RF SIGNAL
Conventional A/53 ATSC includes error correction—both Reed-Solomon coding of the stream and trellis coding with an interleaving—to help it recover from burst-errors like those caused by impulse noise. Considering the A/53 terrestrial DTV standard was drafted 15 years ago, it does a pretty good job for fixed reception. The error correction allows ATSC to work with signal-to-noise ratios as low as 15 dB. Since A/53 was adopted, there have been major advances in coding that allow significant improvement in both signal-to-noise ratio and the ability to recover from burst-errors and fading.
In an A/53 transmission system, data from the multiplexer containing video, audio and data (PSIP) is randomized and fed to a Reed-Solomon (RS) encoder that adds bits to allow forward error correction. The output of the RS encoder is passed through an interleaver that spreads the data over time to make it more resistant to burst-errors.
Finally, the data is fed to a trellis encoder that creates 3 bits for every 2-bit input and assigns them to the eight specific amplitude levels in the 8-VSB system. On the receive side, the decoder takes a sequence of the received symbols; analyzes all possible paths through the trellis; rates the paths for "likelihood;" and then keeps the path with the maximum likelihood.
Because the decisions in this system are based on the entire received sequence the "maximum likelihood" coding doesn't work well for Mobile DTV because the fading due to mobile multipath creates errors that the decoder, even with the A/53 interleaver, can't handle.
For Mobile DTV, it isn't possible to replace the RS encoder, interleaver and trellis coder with one more suited for mobile service due to the need to maintain compatibility with existing ATSC fixed receivers. Instead, the Mobile DTV packets, identified by their ATSC-reserved PIDs, are processed by two convolutional encoders separated by a long interleaver using turbo codes. Error correction is added in the A/153 mobile multiplexer at the studio along with the known training data and the transmission parameter channel (TPC) and fast information channel (FIC) data, (see "Monitoring Signals with ATSC USB Tuners" in the Oct. 6, 2010 issue for the tables for information on TPC and FIC).
The ATSC Mobile DTV receiver, unlike the A/53 receiver, uses inner and outer decoders to use and generate "soft" information. The "soft" information is a measure of the probability that the received bit is a "0" compared to the probability that the bit was a "1". The output of the inner decoder is fed through a de-interleaver to the outer decoder, which is fed to an interleaver that sends the results back to the inner decoder. The data goes back and forth through the coders, the de-interleaver and the interleaver until the soft information indicates a high probability that the decoders are correct.
Until I read the Rohde & Schwarz white paper "Understanding ATSC Mobile DTV Physical Layer," written by Mike Simon, I didn't realize that turbo codes got their name after the original inventors likened the process to a turbo-charged engine, where, as Mike explains, part of the power at the output is fed back to the input to boost the performance.
The interleaver spreads data across the MH frame. The frame is almost a second long (about 0.968 seconds), allowing it to handle the many deep fades encountered with Mobile DTV.
An ATSC A/53 signal has a known training (PN511) signal inserted every 313 data segments (about 24.2 milliseconds). This is an eternity for a receiver in a moving vehicle attempting to deal with rapidly changing multipath as it moves past buildings. The A/153 mobile DTV signal has known training sequences inserted more frequently in the A/153 data stream at known locations.
Realize how difficult this is—because the A/153 data stream has to go through the trellis coding and interleaver in the DTV exciter (see Fig. 1), the multiplexer at the studio has to know exactly how the exciter will map the training signals in the actual RF signal created.
Fig. 1: Simplified A153 Physical Layer Blocks. Courtesy of Rohde & Schwarz
This can only be done if the exciter coding is in a known state when the A/153 training signals arrive. A/153 includes a method for initializing the exciter trellis state to a known condition. If you look at the "ATSC Mobile DTV Standard Part 2 – RF/Transmission System Characteristics" (A/153 Part 2:2009), you will find a graphical representation of how the training signals at the input are positioned so they fall in a regular saw-tooth-shaped sequence at the output of the interleaver in the exciter.
While testing one of the early Mobile DTV proposed standards we noticed that some receivers couldn't handle the error generated when the trellis coder was reset. The final A/153 standard does not generate these errors and so far I've not heard of any legacy ATSC receivers having problems with it.
I've tried to summarize the 29 pages of detailed technical information in the Rohde & Schwarz white paper. I've left a lot out, but the details are in A/153 and, in perhaps a bit easier to understand format, the Rohde & Schwarz white paper. Even if you don't understand every step, having an idea of what's going on can help when trying to troubleshoot problems with Mobile DTV installations.
ROLLING OUT MOBILE DTV
While helping nine stations build out Mobile DTV I've found a few problems. I won't discuss problems encountered upgrading the MPEG-2 encoding gear to provide the bandwidth for a Mobile DTV signal without impacting the quality of the primary HD and multicast channels—that would require its own column—but instead focus on some of the issues directly related to Mobile DTV.
Many of the problems I've encountered adding Mobile DTV are intermittent, making them difficult to track down. Most problems can be traced to two things—insufficient null packets available in which to insert the Mobile DTV data and corrupted data on the path between the studio and the transmitter.
While it is easy to use a program like TSReader or the built-in monitoring on Rohde & Schwarz's AEM-100 mobile multiplexer to see how much bandwidth is available on the conventional ATSC DTV stream for Mobile DTV and the amount of bandwidth/null packets remaining after the mobile stream is added; it is also easy to ignore that instantaneous bandwidth needs (perhaps due to infrequent transmission of a EIT/ETT PSIP data for programs occurring three days from now) can result in not enough null packets for Mobile DTV.
Depending on how the mobile multiplexer handles the data shortage, the results will not be pretty for the mobile channel, the main channel or both.
The other problem is much more difficult to track down. In the early days of DTV, stations had to use converters to translate ASI streams into SMPTE-310 streams and back. Some microwave modems do their own conversions. Seamless ASI switches buffer packets and sometimes drop null packets. In short, the signal arriving at the transmitter exciter doesn't match the signal leaving the mobile multiplexer at the studio.
This isn't an issue for the A53 signals, which don't use the null packets, but it causes major problems for A/153 Mobile DTV exciters, not only for the mobile signal, but for the main MPEG-2 programs. As described earlier, the A/153 data is encoded for a known exciter coder state when it arrives. If a packet, even a null packet, has been dropped or inserted, the trellis coder in the exciter may not be initialized at the proper time, generating errors on the output stream. Depending on the standard ATSC receiver, the results could be loss of audio, video breakup, or complete loss of signal. Mobile DTV signals, of course, will take hits.
The Rohde & Schwarz white paper states the studio encoders/emission multiplexer must be locked to the same external reference as the exciter. This has not been done in most of the installations I've assisted, including those where the vendor aided in the installation, but I wonder if that could be responsible for some of the problems that show up a week or two after the A/153 mobile signal is put on the air. While there are exceptions where the microwave path has been very reliable, in most cases I've seen far fewer problems with ASI fiber links than with older digital microwave systems.
There's still a lot to learn about how A/153 Mobile DTV works with existing STL infrastructure and signal processing. If your station has run into problems implementing Mobile DTV, let me know, along with the solution you found (if any). It could help other stations.
The Rohde & Schwarz white paper is at www2.rohde-schwarz.com/file/mh1_wp-v1f.pdf and ATSC Standard A/153 Part 2 is available at www.atsc.org/cms/standards/a153/a_153-Part-2-2009.pdf.
Comments? Questions? E-mail me at firstname.lastname@example.org