Print

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

[Weaver]

I set about comparing three different flavours of Netgear DG834 router - v1, v3, v4 - on the same extremely long line. As has been remarked earlier the v1, v3 are TI AR7-based, the new v4 has a Broadcom chipset. The v1 was tested for a week. The v3 is a long-term favourite.

In an earlier post I wrote that three different flavours of Netgear router from the same family reported different line attenuation figures on the same line:
    DG834v1 = 60dB
    DG834v3 = 61dB
    DG834v4 = 63.5dB

v1 vs v3: The v1 performed badly in comparison to the v3 despite having a slightly better looking reported attenuation figure. Restested over a week the v1 synced at a slightly lower speed than the v3 yet was a poor performer. The v1 recorded a very high number of uncorrected errors even during the first hour of uptime, no resyncs were evident but the high error rate was thought to be so high as to actually lower throughput a bit. The v3 in comparison was running with a very low uncorrected error rate and at a sync rate well below its potential (due to DLM paranoia on all three tests) with the v3 showing an SNR of 10.5-12dB on an initial SNR margin of 12dB. And this was true despite the fact that the slightly lower sync rate should have made the error rate even lower than the v3. The v1 reported an SNR of ~0.5-1.0dB worse than the v3 long term despite starting off also with a margin of 12dB. Looking at the bits-per-bin allocations of the v1 vs v3, the v1 seemed to be frequency averse and stuck to a narrower range of DS bins with the high end marginal ones being absent and one or two lower DS bins being less utilised. I could speculate that the v1 has a less successful analog front end design?

v4 vs v3: Crickey! The v4 was a disaster compared to the v3. Initially syncing at a rate 500k below that of the v3 (at v4=1440k, v3=1920k both  with SNR margin=12dB), even when the v4 was tweaked using the adslctl command and pushed mercilessly hard to force it down to a SNR margin of <9dB (showing SNR of ~8.5-9.0dB) it to near the sync rate that the v3 held naturally, and later attempts to push it even further down from SNR margin=9dB down towards 6dB simply failed completely. It was so bad that further testing was deemed pointless. The v4 seems to be a catastrophe on ultra-long lines, yet I know from personal experience that the v4 can be an excellent performer (in ADSL terms) with short lines and strong signal.

Netgear seem to have made a horrendously bad decision here with the v4, a device that has attracted criticism in other areas too concerning software reliability in higher layers.

It needs to be pointed out that the v4 could be at a disadvantage because the DSLAM used was TI-based giving a possible synergy with the v3, but that isn't all there is to it as the v1 did not do as well as the v3.

The DG834v3 emerges bathed in glory as extremely fast and reliable in ADSL terms on an ultra-long line. Aside from the ADSL modem, it still has a few design weaknesses in terms of the GUI design, especially being not as kind to non-NAT users as it should be, and has one known bad firewall state mistake if you are unlucky with the UI.

The firmware versions were the latest available as of the end of Dec 2008.
DG834v1 Firmware Version 3.01.38  [ - ?  latest as of Dec 2008]
DG834v3 Firmware Version 4.01.40  [- rel Oct 2008 ]
DG834v4 Firmware Version 5.01.09  [- rel May 2008 ]

[roseway]

That's very interesting (thanks), and supports our long-held belief that the DG834v3 is an excellent choice on very long lines. Tests which I ran on it a little while back on my medium length line (~41 dB) also gave good results in terms of connection speed, but the Broadcom-based routers are more stable on my line.

[Weaver]

It is quite frightening

(i) how useless the v4 is on extremely long lines, even those that are exceptionally clean, and

(ii) that Netgear as a company now don't have a solution at all any more.

I'll need to find a volunteer in the 60db+ club who has access to other mfrs (non-TI) DSLAM exchange to see if the v4 is useless everywhere.

I'm also going to give a few other routers a try, starting with the SpeedTouch, in the "extreme length" challenge.

I'd be interested in any other tips for other worthy contenders. (Which should poss go in a separate thread.)

[HPsauce]

I'll need to find a volunteer in the 60db+ club who has access to other mfrs (non-TI) DSLAM exchange to see if the v4 is useless everywhere.

It's interesting, though not relevant to the specific long-line problem, that the v4 (and GT) are very well-regarded for use on Be's ADSL2+ service. Two points are notable though:
1. Be won't accept very long connections anyway, and ADSL2+ isn't good at extreme distances.
2. Be's DSLAM's are Broadcom-based and work very well with the Broadcom-based v4/GT (but NOT TI-based routers)

This does make me wonder how much of the v4 issues you've found are down to chipset incompatibilities. 

[Weaver]

Indeed, I thought about the issue of whether later protocols pushed the designers' priorities in the direction of accomodating higher speed services at the expense of lower bit-rate use-cases.

Speculating wildly, it seems to me that there might be a number of factors in play in a hypothetical design. A design that knows it will never have to cope with throughput of more than say 2Mbps, or 8Mbps or whatever, but whatever 24 or even 50Mbps is never going to happen can use a lower-speed processor. This could mean a lower speed processor with an easier job achieving an extremely quiet local self-noise environment. An analog front end that is relatively bandwidth limited is one that lets less noise in, and so a design with a variable analog window might be a step towards the state of the art.

However, the v1 possibly has a slower CPU, possibly not, but it certainly dates from a period before the high speed ADSL2+ was on the spec sheet, and so the v1 might be considered a candidate for "excellence by not catering for the irrelevant". But the v1 didn't do too well on test, so so much for that theory.

It could be simply that the v4 is an example of skimping/costcutting (aka "production engineering") where good design was sacrificed on the altar of cost, or it could be a simple bug. And indeed there is still also the modem-vs-DSLAM synergy issue.

Worst fear is that this might reflect a future attitude that "all the important people have low attenuation" !

What we need is a system of Michelin stars for ADSL modems that is widely promoted, and in this a specific category for achievement on high attenuation lines, and/or valiant performance against horrid noise and awkward time-varying characteristics.

Politicians have rightly pushed BT to do the right thing in extreme rural areas, but seem to have neglected the fact that hardware and software play an absolutely critical role, and that designers need to be pushed to offer premium electronics for demanding customers.

[dieselglider]

I was* on a 51dB line, with a Texas Inst. DSLAM, a 9dB Target noise margin since May '08, and now have a small collection of routers  gathered in my quest for line stability.

My mainstay router has been the 585v6 (7.4.3.2 f/w), which generally gives a decent sync and is pretty stable.

The Netgear DG834gv2 sync'd a fair bit quicker, but the Routerstats SNR trace was a bit wild, especially at dark, and it dis-connected quite regular.  The DG834gv3 (4.01.37 f/w) was a lot more stable compared to the v2, with sync figures and stability similar to the Speedtouch, although I seem to remember the 585v6 had slightly better stabilty, maybe.  The DG834gv4 was a disappointment with the supplied V5.01.01 f/w, syncing at only 3,300kbps during daylight (the 585v6 was syncing at 4,500kbps immediately after), but upgrading to V5.01.09 brought it up to about 4,300kbps sync rate, and stability seems ok.  The main reason for buying the v4 was to play with DMT  , which set at 50%, gave about 5000+kbps during the day, although I sync'd during darkness for more stability, and trying to judge a re-sync to hold an IP profile of 3,500.

I also had a brief fling with a 2Wire 2700HGV, this sync'd roughly a couple of hundred kb faster than the Speedtouch, but with less stability, ie, more disconnects.

I've recently downloaded the latest DGteam firmwares for the Netgears, and was particularly thinking about trying it on the v3 to see how that goes.  Has anyone tried that yet?

One thing I've noticed, is since replacing my dect phone with a new one, the FEC's seem to have dropped, but still with bursts now and again.  Last night I noticed they were racking up again and switched off the dect at the plug resulting in the FEC count hardly changing.  I switched the dects back on, and it's been ok so far, with only about 10,000 in the last 10 hrs uptime.  I rebooted the Speedtouch last night after 25 days sync, to see if the Target had dropped, but alas, no.  Maybe the 360 million FEC's had something to do with that, although some think that these don't count as errors...





*Att'n was in region of 49-52dB (reported 5.1km length) depending on router, but BT did some general work across the road recently, with a result of 44dB, and increased sync rate.  I'm currently syncing at night, presently at 4,224kbps with the 585v6, in order to get a decent uptime with relativly low-ish errors, to try and get back to a 6dB Target margin.

[Weaver]

I'm very grateful to dieselglider for her/his comments [that's the problem with these usernames and English pronouns' requirement for gender marking], very interesting.

> The DG834gv3 (4.01.37 f/w) was a lot more stable compared to the v2

I didn't look at the v2 devices. I have a few, and was just lazy, I'll admit. I was assuming they would not perform worse or better than the v3. Which v2 firmware did you use? (Essential to update firmware.)

For the DG834v3 You used the 4.01.37 firmware and 4.01.37.40 is now publicly available. The .40 version IMHO may bring important reliability improvements but I don't believe that it delivers any changes related to ADSL, it's more to do with TCP/IP and firewall bugs. (Vista users should make sure to upgrade.) The .40 release certainly passes the Microsoft Internet Connectivity Evaluation web-based router test suite, which may be a first, but I can't remember if the .37 or .40 was the first to get a full score.


I'm very interested in what you say about the SpeedTouch and the 2Wire. Might be worth changing microfilter type in my experience, as getting more speed but without stability in my experience is sometimes helped by filter choices.

Perhaps you might wish to experiment and try double-microfiltering your DECT phone. Also try the shielded twisted pair dial-up modem type posh cable (BT-to-RJ11) from Belkin to go to your DECT phone. This stopped my DECT phone from creating a moment of havoc with a big temporary drop in SNR when starting a call. From the sound of it though your errors are not following that pattern.

As the lucky owner of a mere 44dB line attenuation [!] you live in a different world to me and so some of my comments won't be appropriate. After all, you get 7 times more volts than I do (all things being equal, which they probably are not).

[dieselglider]

dieselglider = he 

Hi, looking back at my notes, ahem, I believe the v2 firmware was V4.01.20, that was back in Oct '07 and probably the last time it was used.  I can't even find it at the moment.

Last night my FEC errors only racked up a count of about 32k over a 24hr period from reboot, on the 585v6, but suddenly this morning they started shooting up again.  I happened to notice this while the missus was having a natter on the phone.  However, after she had finished nattering, I disconnected the phones and the FEC's still racked up, hitting well over a million last time I checked, so that sort of ruled out the dect as being an issue in this instance.  Btw, I had double filtered the phones last night as an experiment.

So I brought the DG834gv3 out of storage, installed the DGteam '.40' firmware, and connected at 5344kbps for about a half-hour.  However, the SNR was having constant dips from 9dB down to 6dB, which didn't look great to me regarding stability.  So I re-installed the Netgear '.37' f/w, and tried again, with the same result (and nearly the same sync rate), so back on now with the 585v6 at 5,376kbps.  It's been on for about an hour showing only the usual half-dB deflection from 9dB, and 1700 FEC's.  Seeing the noise margin trace for the 585v6 reminds me that the DG'v4 has a similar trace but with smaller movements, since it can log down to 0.1dB, whereas the 585v6 logs down only to 0.5dB difference.

Funny you should mention that cable, I forgot I had bought something similar for the router, and I had tried it, but was experimenting at that time with some clamp type ferrites I had bought at Maplin.  Unfortunately, things proceeded to go a bit hetic with the SNR trace doing some crazy jumps up and down, so I decided it wasn't such a good idea to put a ferrite on a modem cable.. 

Talking about routers, I've just recieved a new 585v7 from Plusnet this morning, as a result of having signed on for another 12 months.  I've no immediate plans to push it into service as any reports I've heard haven't been too upbeat about its long line performance...unless somebody knows different.  The main issue with swapping routers over at the moment though, is that the wireless printer tries to ignore me, it doesn't like change.

Also, I have various microfilters about the place, but tend to favour using the ADSLNation ones I paid good money for.