Why is OTA DTV audio easily broken with weak signal?

[chengbin]

I use OTA HD, and sometimes I watch a weak channel. When the picture is even slightly garbled, the audio is intermittent. Why?

Audio is 448Kbps Video is 19200Kbps. It is 2.3% of the total amount of data for HD. So why is even a slight garble in video results in bad audio? I understand if bad reception causes me to lose say 10-20% of data, but what's the likelyhood that ALL of 2.3% of audio information is lost in there ALL THE TIME?

For analog OTA, the picture can be ridiculously fuzzy, and the audio would remain crystal clear.

I'm using a LG 50PK550, with the antenna plugged directly into the TV.

[Jake]

The audio stream and video stream are combined (muxed) at the source and broadcast. The TV tuner splits them apart (demuxed). When a data packet is received but corrupted both the audio and video are effected. As you know with digital data you can easily corrupt the data by changing a few bits. So you lose not only the few bits but the entire package of information. Plus with MPEG video it requires a few packets to build an entire image. With analog there are no packets. Analog signals are a different beast. Basically your TV can lock onto the analog audio signal easier than the analog video signal because there is less information encoded in the signal to extract.

[stampeder]

Yep, ATSC uses a single data stream across the alloted bandwidth. It is an all-or-nothing system, but it uses Forward Error Correction so if you are getting dropouts you need to look at better reception gear and/or aiming if possible.

The old analogue NTSC system uses the alloted 6MHz of channel bandwidth by chopping it up into specific parts for video and audio, and not only that but a station's own video and audio signals would interfere with each other if they used the same modulation so NTSC was deliberately specified to use AM (Amplitude Modulation) for the video while the audio portion uses FM (Frequency Modulation).

That's why it is possible to receive NTSC analogue audio and/or video separately.

[ScaryBob]

FM signals are more robust than AM, which explains why NTSC sound is sometimes present when the picture is unwatchable. ATSC sound cuts out when the picture is slightly garbled because there is a loss of information that cannot be corrected and decoded. Dropouts in the audio is more annoying than a slight loss of picture information. In the early days of ATSC (and other forms of digital HD), the picture would go black when there was a video dropout. That was very annoying. Today's video decoders buffer the last frame so that video dropouts are less noticeable. Unfortunately, there is no counterpart for buffering audio so audio dropouts are as annoying as ever. There may be other reasons why audio dropouts are more noticeable, such as encoding/decoding methods, but that's my take on it.

[timlocke]

My question then: As ATSC is a fairly modern invention, that is invented AFTER the availability of cheap fast digital processing, why did the implementors add some seriously robust error correction to the bit stream?

[JamesK]

so NTSC was deliberately specified to use AM (Amplitude Modulation) for the video while the audio portion uses FM (Frequency Modulation).

Originally, prior to WW2, AM was also used for audio. It was later switched to FM to improve sound quality. This is the first time I've heard any mention of interference between audio & video being the reason for FM audio. There were interference issues that dictated the frequencies used for colour which also required a very slight change in sync frequencies. Also, with analog TV, there were often separate transmitters for audio & video, instead of combining both into one transmitter.

[Jake]

seriously robust error correction

That would require a return signal. Much the same way cable systems use the return data channel.

[majortom]

atsc standards are located here for download.
http://www.atsc.org/cms/standards/a53/a_53-2010-r0.zip

containing:

A/53, Part 1:2009, “Digital Television System”
A/53, Part 2:2007, “RF/Transmission System Characteristics”
A/53, Part 3:2009, “Service Multiplex and Transport Subsystem Characteristics”
A/53, Part 4:2009, “MPEG-2 Video System Characteristics”
A/53, Part 5:2010, “AC-3 Audio System Characteristics”
A/53, Part 6:2010, “Enhanced AC-3 Audio System Characteristics”

[JamesK]

That would require a return signal.

Actually, no. There is also "forward error correction" where you provide sufficient redundant data so that the original can be recovered despite errors. That's what's used on CDs, DVDs etc. It's also used on space probes, where the long return trip precludes retransmission requests. This contrasts with error detection and retransmission, as used on computer networks. Also, I think you'll find the cable data return channel is for actual user data, not error correction.

[Jake]

I think you'll find the cable data return channel is for actual user data, not error correction.

Besides the FEC data, the RDC diagnostic page on my cable STB lists the numbers of retransmissions. Just saying.

[ScaryBob]

seriously robust error correction

Return signal or not, high levels of error correction require high amounts of redundancy, lots of overhead and time. For example, deep space probes send pictures hundreds of millions of miles but require days to send one picture frame. That's not a good option for ATSC TV. Error correction is always a trade off and will always fail under conditions it was not designed to handle. I'm sure the ATSC standards committee took all factors into consideration and came up with an acceptable trade off for what is required with terrestrial TV broadcasts.

[GeorgeMx]

Originally Posted by Jake

Besides the FEC data, the RDC diagnostic page on my cable STB lists the numbers of retransmissions. Just saying.

The cable digital video system has no retransmission as it was designed to operate over one-way cable networks feeding tens of thousands of customers. Broadcast video is like UDP packets on the internet. The retransmission data you are seeing relates to the reverse channel which is used for reporting PPV purchases, requests for channels in switched digital video and any TV interactive services.

Cable, satellite and OTA digital transmission are based on technologies over 15 years old. In computer terms, processing power was back in the 386/486 era and memory was hugely expensive compared to current prices. The system designers had to balance cost against performance so the level of FEC and other error correction techniques was constrained by cost.

With respect to the original question about audio failure in OTA, the problem has attracted the interest of the ATSC. People are used to the audio signal surviving in NTSC even when the video is seriously impaired but the ATSC design means they both fail at the same time. I believe that some effort was put into increasing the error protection for the audio stream PIDs but I don't recall if it was ever turned into a standard and implemented.

The original design could have included additional error correction for the audio but users listening to the audio without seeing good video was not considered. An alternative design for digital TV could have provided different levels of picture quality with different amounts of error correction. A basic lower resolution picture would have the most error correction to provide a very wide coverage area. The next level of resolution would have less error correction and full high definition would have even less error correction. The quality of picture received would have depended on signal strength and quality. This complex solution would have cost more and reduced the data capacity for a full high definition image but it would have provided a more robust signal for fringe area viewers or people using indoor antennas.

[Rick Dude]

There is some discussion in earlier posts about analogue TV. I will explain some aspects of it. Originally from the UK, I was in a TV development lab at the time of the introduction of the PAL colour system.

TV sets in the UK in the mid 1960's were known as dual standard. They had to work on both 405 lines, broadcast on VHF in the UK, and 625 lines broadcast on UHF.

The 405 lines system used in the UK was known as System A. It had positive AM modulation for vision and AM sound. The French 819 lines System E (think of it as early HDTV) was also positive modulation for vision and AM sound and used a channel width of 14 MHz!! The UK 625 lines System I uses negative AM modulation for vision and FM sound.

Positive vision modulation has the sync pulse as minimum transmitter output (almost zero) and peak white at around 90-95% output. Negative vision modulation has the sync pulses at 100% transmitter output and peak white at around 10-20% output - dependent upon the TV system.

In 1936, the 405 lines BBC TV was the first in the world, but positive vision modulation is not the best. To receive the sound properly, the channel had to be fine tuned on the channel selector to put the sound at the correct frequency of the IF amplifier which was 38.15 MHz. Low VHF channels can be subject to interference from car ignition systems which were not as well suppressed as now. This would produce white spots on the video and interference on sound. A white clipper was used to reduce the annoyance of visual interference. On negative vision modulation, any vehicle ignition interference would appear as a black spot and is far less annoying.

The FM sound channel in the TV receiver is produced as a mixing product in the demodulator. This is known as inter-carrier sound. In the UK, FM sound IF is 6 MHz and in North America on the NTSC system, it is 4.5 MHz. Because the output of the vision transmitter never drops below 10%, there is what is known as the residual carrier. This means there is always vision carrier at the demodulator to produce the inter-carrier sound. The FM IF circuit has to have excellent AM rejection otherwise there is a buzz on the sound that varies with picture content, known as inter-carrier buzz. No fine tuning is needed to optimise the sound.

So, to sum up. FM sound can only be used with negative vision modulation as the residual carrier is needed to produce inter-carrier sound. Conversely AM sound can be used with both positive or negative vision modulation.

[stampeder]

Thanks for that, Rick_Dude, very informative! Your post also makes an interesting corollary to GeorgeMX's post #3 in this thread:

http://www.digitalhome.ca/forum/showthread.php?t=130384

[ericball]

Actually, almost 43% of the data received by the TV is FEC.

0.8% is sync data (which is used for ghost cancelling)
33.3% is trellis encoding FEC
9.6% is Reed-Solomon FEC

I suspect the reason OTA loses audio before video is precisely because it has less data. So one uncorrected bit error can have a much greater impact.

MPEG-2 video is made up of three types of frames - I frames which are self-contained and P & B frames which are based on past and future frames. However, even within the P & B frames there's a lot of "same as previous frame" (which is compressed down to nearly nothing). So if the decoder was able to construct most of the previous I frame, then it has data it can simply copy to the current frame.

AC-3 audio on the other hand is made up of overlapping blocks of data, so the loss of data in one block propagates into the next block. There's very little ability to try and mask (or simply ignore) errors. (Whereas MPEG-2 video can either copy from the previous frame or produce garbage data.) While a bad block in video will produce an ugly picture, bad audio could damage speakers.