Quote:
Originally Posted by
bobbyc 
Is there a good website that explains the woes of fringe digital reception.... and why digital/8vsb doesn't work like the good ol analog days of snowy/ghosty picture yet very watchable? I tried to google it real quick but didn't come up with much.
I've had two people this week ask for my help; but I wanted them to know what they're up against before I provide any help. They're both older gents and probably used to the good ol analog days of receiving a bounced signal off the hill etc...
Thanks, Bob C
This is the best page I know of but it probably doesn't provide the information you'd like in a concise manner:
http://www.hdtvprimer.com/
I'll take a stab at answering your question but I'm not sure I'll be successful.
The analog world is pretty easy to understand. A TV picture ranged from perfect studio quality reception to so snowy you could barely make out the picture. Signal to noise ratio was represented in the picture by how much snow and/or ghosts you saw.
8VSB is an analog modulation scheme that has to deal with noise too. What you're really asking about is the ATSC encoding scheme. This consists of data compression, data randomization, and a couple levels of error correction. This data stream is used to modulate the 8VSB signal. In order for this to work and the TV tuner to be able decode a picture and sound, there can only be just so much noise. The maximum level of noise that the system was designed to handle occurs at 15.2 dB signal to noise. At lower SNRs the entire encoding scheme breaks down and the TV cannot produce a picture or sound. The system was designed to produce a perfect picture and sound or none at all.
You almost always hear that since it's digital TV, it either works or it doesn't work, like the digital concept of 1 or 0. That's not the case. It's the ATSC data encoding scheme that causes this "digital like" behavior.
To illustrate this a bit further, I used to receive NOAA low orbiting weather satellites that used BPSK modulation - another digital modulation scheme. This data was very simple, no data compression and no error correcting. It did contain formatting data. When the SNR was too low, some of the bits were received in error due to noise. The resulting pixels assembled from that data would be the wrong brightness values. It looked just like noise. A weak signal could produce a very noisy image, similar to analog TV. To obtain a noise-free image, the SNR needed to be high enough so that noise had no impact on the data decoding. A picture would be produced until the receiver lost carrier phase lock otherwise it just assumed the data was correct.
I'm going to try to explain how noise affects digital modulation, but specifically 8VSB. 8VSB is amplitude modulation and transmits 8 discrete amplitudes. In this system each amplitude represents 3 bits of digital data which are called Symbols. Symbols represent digital 000 - 111. There is never any actual digital 1's and 0's transmitted like you think of it in a computer. It's an analog representation of a digital number. If the SNR is very high, the received amplitude will be the same as the transmitted amplitude. As the SNR goes down, noise begins to randomly alter the received amplitude. There's a small range of amplitudes (think of it as a window) that will still decode the same digital number. Even high SNRs above 20 dB do not produce a perfect received amplitude. As the SNR continues to drop, some symbols will appear out of the correct amplitude range and be decoded as a symbol different from what was transmitted. This is a bit error although it's actually an error of 3 bits. If there are not too many errors, the ATSC system can use its built in error correction to fix these and still produce a perfect picture. At some point there are too many errors to be fixed and the system collapses. An SNR of 15.2 dB is generally the lowest SNR where all the errors can be fixed.
Because of the complexity of the ATSC encoding system, it needs virtually perfect data for it to work. This is the reason why it cannot just fade into the noise like NTSC could.
Multipath acts very much like lowering the SNR. Some of it can be corrected but much of it cannot.
I hope this was helpful, but I don't make any guarantees.

Chuck