Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Weaver]

This is a continuation of the thread
    http://forum.kitz.co.uk/index.php/topic,16501.msg305240.html#msg305240
and see also
    http://forum.kitz.co.uk/index.php/topic,15958.msg296825.html#msg296825

In the first of these two threads, I was wondering about the possibilities for designing an ultra-high quality DSL modem (a 'hi-fi modem'). I remembered reading about a Zyxel (VDSL2 ?) modem that had an analog input filter at the front end of it. The poster remarked how this gave Zyxel some advantage and was a component lacking in other designs.

This set me wondering. Revisiting the old thread, I wonder if it would be possible to design something rather less ambitious than a whole modem - a separate active analog input filter, tuneable to suit very long ADSL lines which have different degrees of attenuation of the higher frequencies. This could utterly block out the crap in a highly selective way. As to what order of roll-off, I'm not sure. And I'm taking it as given that it needs to be active so as to have unity gain. (Or optionally higher?)

Any thoughts?

[guest]

I'm having a strong feeling of deja vu here

I definitely remember discussing filters on here - butterworth/chebychev/etc and notch filters using op-amps. Probably a few years back as the resident nutter was still around. 2010-2011 maybe?

You don't need an active filter (ie an op-amp circuit) unless you need a notch-filter to filter a very narrowband noise source.

If you are really meaning "I want something to boost/pre-emphasise the higher frequencies dynamically to compensate for higher attenuation/crosstalk/noise on long lines" then forget it. You're at the wrong end of the line and even if you weren't the ANFP prohibits it. Assuming this is what you wanted you might like to know its called "vectoring"

Anyway have a look through my old posts - it might have had something to do with SDR, or that might just have been the resident nutter.

[Weaver]

I should have explained properly that the filter is to go in between the phone line and the modem. At the CPE end.

[Weaver]

Rizla wrote:
> I'm having a strong feeling of deja vu here

I had it first, as I'm repeating myself.

> Probably a few years back as the resident nutter was still around. 2010-2011 maybe?

Am I not the resident nutter?

> Anyway have a look through my old posts - it might have had something to do with SDR, or that

Will do.

[les-70]

  From the experience of messing about with many things I doubt any filter will make a difference unless the amplitude of some noise is so great as cause non linearity the intended adsl signal.  A lot depends on what sort of noise a line sees.  Do you get only a general tick over of errors or also odd large error spikes?  If it is odd large error spikes a common mode filter is worth a try.   With your low sync rates an BT RF3 should have no negative impact.  Depending on the error source running the modem via  a BT Mains Conditioner can also help.  Quite often nothing helps -- this is usually because almost all modems actually do a very good job.  Are your line stats on MDWS? if there were both you and others might be better able to spot issues.

[Weaver]

Quite often nothing helps -- this is usually because almost all modems actually do a very good job.

Good point. This is the kind of thing I need to know. One the longest usable lines, say 0.5 Mbps ADSL / G.Dmt what's the highest tone in use?

[ejs]

I think any noise on frequencies not in use, the frequencies of tones with zero bits allocated to them, isn't really going to concern the modem. If the modem is ignoring those frequencies anyway, I don't think filtering them off before the modem would affect anything.

[Weaver]

@ejs perhaps I've misunderstood what the Zyxel mentioned earlier is doing? What do you think?

[les-70]

  I believe the Zyxel has a common mode filter.   That filter is set so as to pass vdsl differential signal frequencies and is not optimal for adsl.  The RF3 is set to pass only adsl  differential signal frequencies and would be more appropriate for your line.  If you have a common mode noise issue it will help and if not it will make no difference.   

[guest]

The front-end of ALL xDSL devices is a differential amplifier so unless the line pair is very unbalanced (or the modem is complete crap) I see little point in a differential filter. The noise will still be on the line unless its generated locally.

[les-70]

so unless the line pair is very unbalanced (or the modem is complete crap) I see little point in a differential filter.

   I think that is correct but on my line, for example, such filters seem to moderate occasional large noise spikes. I have assumed these spikes are part or all common mode and may be big enough to cause non-linearity in the modem input stages. Such non-linearity damages the input stage common mode rejection.  Separate from that I would say such often filters just have mild negative attenuation impacts.   The RF3 is at least cheap and easy to try, but as you say it will most likely make no difference. 

It is much the same story with a BT mains conditioner.  They give full mains isolation via a transfomer and are not like most mains conditioners.  One does  help my line but may make no difference to most lines. 

[guest]

If the noise spikes were of sufficient amplitude and wideband then it'd have an adverse effect on the diff amp stages as you say. Perhaps the Zyxel has a shunt for that situation built in.

However such spikes are more than likely local to the modem (eg conducted & radiated mains noise) and best dealt with at source if possible.

I wouldn't imagine Weaver has too much of that from neighbours so most of the noise on his line is likely to be locally generated - the remotely induced noise will suffer the same attenuation as his DSL signal anyway so not a lot can be done there in terms of improving things.

My usual starting point for this sort of stuff is to use a cheap UPS to run the modem (Sky router here has some ancient Belkin cheapo UPS feeding it) and a laptop then flick the mains switch at your fusebox.

Run your stats programs and look for a difference on the graphs. If there is then go play detective & find what's generating the noise - here it was some second-hand PS/3 with a very noisy power supply which I filtered with one of these : http://www.maplin.co.uk/p/tacima-2m-six-socket-mains-conditioner-plus-rfi-filter-for-home-cinema-and-hifi-l07bk (didn't pay that price though, they're on Amazon). Someone else on here used the same filter to sort out a noisy treadmill. Its got a very decent filter but is stupidly expensive for what it is.

In the case of locally generated mains noise its better to filter at source - ie don't connect the mains filter to the modem/router, connect it to the offending mains-powered items.

Not trying to teach anyone how to suck eggs but eliminating conducted/radiated interference isn't always as obvious as it seems, so its best to start from the point of "is there a problem in the first place?".

I'm also assuming that Weaver has a UPS - damn sure I would if I was living up there in Winter

[Weaver]

I have several UPSs, with mains surge protection before them, and double mains filtering on the output of the one serving the modems in case the UPS' synthesised mains is staircase-like (as I have seen described in articles on UPS quality).

I'm not thinking about designing a solution for my particular situation, rather musing if any beneficial general add-on component is doable for the case of the high attenuation, restricted band of bins in use.

[guest]

Not really. As said before you're at the wrong end of the line to do much about downstream attenuation.

Vectoring is unfortunately the answer as essentially its just a dynamic pre/de-emphasis system.

In physical design terms its very complex but at the analogue line-level all it basically does is dynamically increase/decrease power levels at various frequencies to adapt for noise/crosstalk/attenuation in cable bundles. Not likely to happen on ADSL in the UK although its technically trivial in terms of circuit density/power consumption when compared to VDSL2/g.fast.

In terms of your own situation - have you tried killing the power like I suggested? Noise on mains cabling goes from conducted to radiated (and hence induced into remote wiring of all sorts elsewhere) quite rapidly as the solid core cabling is an excellent antenna for both transmission and reception.

[Weaver]

> In terms of your own situation - have you tried killing the power like I suggested? Noise on mains cabling goes from conducted to radiated (and hence induced into remote wiring of all sorts elsewhere) quite rapidly as the solid core cabling is an excellent antenna for both transmission and reception.

No I haven't. It would be an interesting experiment. I'm not hunting any kind of noise problem, but if I had appropriate measuring tools and were well enough it would be very interesting.

I have filtered the mains coming into all UPSs, and double filtered the mains coming into the modems, as I mentioned before. I've also paid careful attention to keep mains cables physically slightly separate from modem DC cables and mains is not run along in parallel touching anything potentially interference-critical. The whole office is on its own mains ring; so no heating, motors relays or any such is on the same ring circuit.

Because of the extreme isolation in terms of lack of local disturbers along the latter three miles of metallic path, then I have always assumed that the house could be a prime interference contributor because of the amount of activity and the attenuated state of the signal. Nevertheless, I don't have any reason to assume that I have any kind of noise problem, in fact I rather assume that noise levels are reasonably low.