Print

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

[Weaver]

Do other contributors have experience of seeing differing line attenuation figures shown by different router models on the same line (even when they come from the same stable)

On my line
  Netgear DG834v1 = 59dB
  Netgear DG834v3 = 61dB
  Netgear DG834v4 = 63.5dB (bogus?)

I'll compare with a Speedtouch too, and report back.

I would assume that the figures might depend on _which_ frequencies are being looked at as I assume we're talking about real RMS voltage levels at certain frequencies, but the need to decide on some way of combining them them? Simple average? Or weighted? Or ignoring the unused bins?

If that's a true picture then it's not really reporting a property of the line that's divorced from the way you choose to look at it, but I suppose that's either a good thing or a bad thing depending on what you need.

Is it fair to say that a router that thinks that it's having to cope with 63.5dB as opposed to one that has an idea for a "59dB plan" is pretty useless? :-)

[roseway]

I'm not sure how routers calculate the attenuation, but it's certainly some sort of weighted average, and I don't believe there's any standardised algorithm for doing this. So it's quite probable that different combinations of firmware and chipset will result in different values being reported. In my case I have seen values between 40.0 and 43.5 dB.

I don't think that there's any reason to assume that the one reporting the highest value is inferior. The router will negotiate the best connection it can, then calculate the attenuation; it doesn't use the measured attenuation to control how it makes the connection.

[kitz]

Do other contributors have experience of seeing differing line attenuation figures shown by different router models on the same line (even when they come from the same stable)

Yes its something I first observed a while back (even before 2Mb adsl) and reported elsewhere several years ago.  Although the original page has now long gone - some of those original stats from those observations now remain on the bottom of this page.

As eric rightly says, its down to the different ways in which each router measures attenuation.  Mostly its to do with how DMT works in that adsl is split into many different sub-channels.  Attenuation is always going to be higher at the higher frequencies, and therefore its not unusual to see your attenuation increase if say you move to a product that provides higher speeds.


Some routers will report the attenuation taken at 'x'kHz - typically 300kHz or and 1Mhz
Some routers will report the average attenuation across all the frequencies in use
Some routers will even report the average attenuation across all the available frequencies even if the router has marked them too low to carry sufficient bits.

>> The router will negotiate the best connection it can, then calculate the attenuation; it doesn't use the measured attenuation to control how it makes the connection.

Correct   -  for more info see Initialisation and Synchronisation

[Weaver]

I'd be very interested to hear if there's a general correspondence between rubbish performers and outlandish reported attenuation anyway.

I read with interest the numbers on that page that kitz pointed out. Is it the case that the sets of data in the last section "pre max stats - 2mbps" are all off the same one line with routers swapped over?

[Weaver]

Roseway wrote:
> I don't think that there's any reason to assume that the one reporting the highest value is inferior. The router will negotiate the best connection it can, then calculate the attenuation; it doesn't use the measured attenuation to control how it makes the connection.

Of course what Roseway is saying makes sense, presumably the real s/n ratio in a particular bin ought to be the decider, as long as the signal isn't so low in absolute terms that it poses a problem for the input stage of the dsl receiver, and I suppose that is just another way of saying that if it were low enough on a perfect line it would just hit the noise floor of the router's own input in the end.

But if a router is somehow insane and is picking inferior bins which are best left alone, then presumably it might report a higher attenuation figure, not because there is more noise in those bins but simply because there is greater high frequency loss. I say "insane" assuming that there is a router out there that achieves a good performance from allocating more bits per bin, either because it's less buggy or simply has some other beneficial feature in its software or hardware design.

So while I agree with Roseway, I would have thought that other software choices might affect the set of contributors to an averaging process or affect a set of weights?

[kitz]

I read with interest the numbers on that page that kitz pointed out. Is it the case that the sets of data in the last section "pre max stats - 2mbps" are all off the same one line with routers swapped over?

Yes same line - those are the only stats I have left now - I just dumped them on the end of that page since over the years and changes of computers I no longer have some of my earlier stats.

But when I did the tests orginally back in about 2004... yes those stats at the bottom are all from the same line.

Also of note worth pointing out is the SAR110 stats which atten went from 21.5dB to 11.95dB after an upgrade in firmware. 

Another note - the Voyager 2100 reports my attenuation as 7dB.  Im now on adsl2+ so of course my attenuation is going to be higher, but when using a ST585v6 it shows my attenuation as 12dB.. but if I fire up DMT that tells me that my attenuation at 300 kHz is 7dB.

[Weaver]

Kitz's remarks are very much bringing it home to me (finally :-) ) that these figures are very much a modem's _view_ of the characteristics of the line, not so much "properties of the line+environment". DSL modems are clearly not pieces of test equipment of measuring devices. We would need something standardised and more well defined independently, possibly in terms of various appropriate weightings to be able to talk about local loops' physical properties in a way that's independant of individual pieces of kit whilst still being sufficiently relevant to some possible piece of reasonable kit's ability to actually get the job done.

[kitz]

yw -

There is also little bit more info here:

~ What is Attenuation?

[audiolab]

thats funny you should mention that because my old netgear dg834pn (with latest dgteam firmware) syncs at least 500k lower than my new bthomehub 2  (speedtouch?) and seems more stable.which judging bt's record on these things is quite amazing,have tried swapping between the 2 just to test it twice on each and still the bthub shines through,didnt want to try anymore tests as didnt want to lose my noise margin profile

[Weaver]

Postscript: In an earlier post I wrote that three different flavours of Netgear router from the same family reported different line attenuation figures.

I rechecked these numbers, the flavours of Netgear DG834 on the same line second time around give
    DG834v1 = 60dB
    DG834v3 = 61dB
    DG834v4 = 63.5dB

As was remarked earlier the v1, v3 are TI AR7-based, the new v4 has a Broadcom chipset.

Accuracy: The v1 could actually be 59.5-60.5-ish, as no decimal point is ever displayed by v1-v3, and the v1 earlier read 59dB consistently.

But concerning the question of whether higher-lower attenuation figures actually mean anything in terms of likely performance or reliability it seems that the answer is definitely a NO. Better numbers do not mean better performance.

The short answer, the v3 is a star, fast and reliable, the v1 is not good, being noticeably slower and having a high uncorrected error count. The v4 is a complete disaster.

I have posted a fuller write-up on their relative ADSL performance separately, in its own thread.

[Azzaka]

Version 1 was not an AR7 chipset.

The attenuation can make a difference and the reasons for the performance issues is down to the way the chip works.

The v3 range pushes the line to its limits where as the Broadcom is conservative and hence slower speeds. But in saying this the Broadcom can be more stable due to the conservative nature.

[orainsear]

Version 1 was not an AR7 chipset.

According to this it was

[mr_chris]

Azzaka, I had thought it was an AR7 in the v1 too. In fact, given that Netgear release the same firmware file for the v1 and v2 (see the bottom of this page ), I have to ask if you're absolutely sure about that?

[Weaver]

Telnet-ing into my v1, its cli commands show it certainly is an AR7, if I recall correctly.