Hacker News Reader

Comment by moultano | Hacker News Reader

Anyone have a link that explains (ideally with some math) how this works?

How it adapts to poor line characteristics? Seems that ADSL uses multiple frequency bins, probing each for signal-to-noise ratio and choosing parameters accordingly, because different lines have different attenuation/noise on different frequency ranges.

This article might be helpful: http://www.kitz.co.uk/adsl/adsl_technology.htm

The pair of parallel conductors (wet string) forms a balanced transmission line[1] just like the copper pair in traditional phone line. The ADSL transceivers at each end automatically adapt to the available transmission line; you'll note that instrument reports that this 2 meter line looks like 4.5km of copper; it is a very poor conductor, so the ADSL transceivers fall back to lower frequency operation, which is apparently sufficient for low speed communication.

I do wonder about impedance matching here. Traditional phone lines apparently have a characteristic impedance of either 600 or 900 ohms. That pair of wet strings look to be separated by about the right distance to produce impedance in that neighborhood, or maybe the line length is tuned to eliminate reactance. It isn't mentioned but I suspect that RevK understood this.

[1] https://en.wikipedia.org/wiki/Transmission_line

Whilst the impedance of the line under yesterday's test conditions wasn't measured, it was replicated again this morning, and from memory it was around the 900 ohm mark. I'll be testing again tomorrow (trying to run bonded data across two of them) so I'll try and get a recorded measurement of the line's impedance at some point rather than just something from memory.

Great! That will be very interesting. It was just guesstimation on my part but the diameter and separation of the strings looked similar to high impedance lines used in SW amateur equipment typically in the 400-600 ohm range. A couple sites I've read state that ASDL actually experiences about 100 ohms impedance as opposed to voice (600 ohms) due to the large frequency difference. If all that is true then bringing the strings closer together will lower impedance which should improve the match.

String wet with salt water conducts electricity similarly to wire.

I was curious about this line in particular.

> One of the key aspects of the technology is its ability to adapt to the length and characteristics of the line on which it is deployed.

DSL transfers data over specific frequencies. Certain frequencies will resonate better than others over certain lines. Here is an image of a line test conducted for a VDSL2 customer: http://wut.nz/images/spm/1027296.jpg. The redder the graph, the faster (up to the maximum VDSL2 speed) the connection is.

The vertical line is when the connection spontaneously resynced, and a new speed was attained based on the new conditions of the line.

For this particular customer, you can see that tones 410-460 show no transfer at all. This causes a reduction in speed, since those frequencies couldn't be used.

https://en.wikipedia.org/wiki/Asymmetric_digital_subscriber_...

https://en.wikipedia.org/wiki/Reed%E2%80%93Solomon_error_cor...

As I understand it, the endpoints test a range of frequency blocks (or "bins") and only use the ones where there is sufficient signal-to-noise. This is dependent on line conditions.

Heh, "line". That has a very different connotation in this case.

Saltwater conducts electricity.

It seems to conduct dsl frequencies pretty poorly - the first attempt had 56db loop loss, equivalent to that of a 4.5 kilometre copper line for just two metres of wet salty twine. Detailed line stats from the testset here: https://twitter.com/0x47DF/status/940634277967990784

Sounds to me like they just need to tune their electrolyte balance!

It's only a ratio of 2250 estimated metres for each actual metre of wet string, perfect for low cost broadband needs! :p

I'm still amazed that it works at all, but I guess that does show how fault tolerant ADSL is...?

The last mile approximator.

Reply on news.ycombinator.com