Relationship between frequency and bandwidth?

[Francis_Vaughan]

The reason ADSL uses many separate bands has nothing to do with some notional ratio of useful bandwidth as a ratio of the carrier/centre frequency. It would be quite reasonable to create a 20Mb/s encoding that worked perfectly inside a 2MHz bandwidth with no sub-band encoding.

The reason ADSL uses many sub-bands is so that it can dynamically choose which bands are good and which are not in the face of noise and signal losses across the spectrum. So it can drop a sub-band from use if it can’t get enough signal to noise on it. This enables it to get the absolute best performance out of a situation where there is high amplitude noise, or too great a loss, in various parts of the spectrum it uses. A simple wide spectrum encoding would fail, or be reduced to vastly poorer data rates than ADSL can manage in the same situation. This is why you see varying data rates on your connection, and why the data rate tends to drop with distance. The line sync works out which of the sub-bands it can use, and that determines your data rate. In general it will be dropping the higher frequency bands first, as they will exhibit the greatest loss with distance, and hence the worst signal to noise, but it can and will drop intermediate bands if there are issues with them, such as interference.

[JWT_Kottekoe]

@Francis_Vaughan

The reason ADSL uses many separate bands has nothing to do with some notional ratio of useful bandwidth as a ratio of the carrier/centre frequency. It would be quite reasonable to create a 20Mb/s encoding that worked perfectly inside a 2MHz bandwidth with no sub-band encoding.

The reason ADSL uses many sub-bands is so that it can dynamically choose which bands are good and which are not in the face of noise and signal losses across the spectrum. So it can drop a sub-band from use if it can’t get enough signal to noise on it. This enables it to get the absolute best performance out of a situation where there is high amplitude noise, or too great a loss, in various parts of the spectrum it uses. A simple wide spectrum encoding would fail, or be reduced to vastly poorer data rates than ADSL can manage in the same situation. This is why you see varying data rates on your connection, and why the data rate tends to drop with distance. The line sync works out which of the sub-bands it can use, and that determines your data rate. In general it will be dropping the higher frequency bands first, as they will exhibit the greatest loss with distance, and hence the worst signal to noise, but it can and will drop intermediate bands if there are issues with them, such as interference.

I agree, but there is another advantage that you did not mention, i.e. greater immunity to intersymbol interference. If you break a band into 1000 subcarriers, the data rate on each carrier (if spread equally) is 1000 times less, which means the symbols are 1000 times longer. If multipath or dispersion in the transmission medium smears out the symbols in time, they may overlap enough to destroy the transmission. The subcarriers improve the tolerance to this type of distortion by a similar factor of 1000.

[Francis_Vaughan]

@JWT_Kottekoe

I agree, but there is another advantage that you did not mention, i.e. greater immunity to intersymbol interference. If you break a band into 1000 subcarriers, the data rate on each carrier (if spread equally) is 1000 times less, which means the symbols are 1000 times longer. If multipath or dispersion in the transmission medium smears out the symbols in time, they may overlap enough to destroy the transmission. The subcarriers improve the tolerance to this type of distortion by a similar factor of 1000.

Yeah, multipath or dispersive effects is a very good point. That does lead to a general trend in radio comms to want to restrict individual channel bandwidth. ADSL mostly doesn’t have these issues, rather line echo, loss and noise, so it is different set of tradeoffs.

As usual, the devil lives in the details.