Routers and Broadband Speeds

[The Flying Pram]

First, a personal disclosure: I worked as a lineman for the GPO/PO Tel & BT between 1973 - 1996, so wasn't involved with broadband. However 1+1 carrier systems and ABC alarms used the extra bandwidth for similar purposes, so I'm familiar with how line faults & noise can affect performance.

Now to my plea for help: a friend is having trouble with poor download speeds - the same thing occurred a year ago, but that was a line insulation fault. This time round the line tests O.K. and there is no noise audible. He normally gets ~3.5meg download speed, and ~3-400k upload, but for the last week the D/L speed has varied from barely 100k up to 750k although the upload is seemingly unaffected. I spent an hour carrying out comparisons between his ASUS WL600G and a cheap and cheerful Belkin I keep for just such occasions. Both report almost identical downstream and upstream values on their respective status pages - 8meg / 450k. Yet the best actual throughput I got using the same server on Speedtest.net was 750k / 350k.

I repeated the comparison with my laptop rather than his desktop (both using a wired connection), and got exactly the same. I had already tried re-booting the ASUS, and disconnecting the extension wiring, with no change. The only thing I haven't tried is a separate microfilter - this is built into the front plate on the latest style linebox, and is (as far as I'm concerned) BT / Openreach's responsibility. However, as both routers appear to be communicating with the exchange at more than adequate speeds, I fail to see what difference this would make.

Am I right in thinking that those reported up/down speeds are based on the initial "handshake" carried out when booting up? If so, wouldn't this point the finger at a network fault - i.e. the data packets are being throttled beyond the exchange line-card? As the upload speed seems to be O.K. this furthers my suspicion. It looks as if I will have to do the waiting for Openreach tomorrow, as my friend is unavailable, so I want to be absolutely sure of my ground before any haggling begins...

TFP

[mixture]

Am I right in thinking that those reported up/down speeds are based on the initial "handshake" carried out when booting up? If so, wouldn't this point the finger at a network fault - i.e. the data packets are being throttled beyond the exchange line-card?

Assuming his distance from the exchange is reasonable (broadband speed decreases exponentially with distance from exchange), then you've either got a case of :

- Rate adaptive DSL not working properly
- Some sort of ISP tricks (contention, packet shaping, rate limiting etc. etc. etc).

Either way,you've tested with an alternative router, so if it remains the same after a new filter, and its not a distance problem, then its likely to be a network issue.

Tip from the wise: Ring up your ISP and make sure they ensure the imbeciles at Openreach send out a broadband engineer, not a line engineer ! Otherwise you'll end up with a worthless visit from someone who's only interested in a crackle-free dial-tone ! I've seen it happen more times than I care to repeat !

[The Flying Pram]

I had a look at my own router status page, and it shows ~18meg D/L and I normally get about 12-13meg on speedtest.net. Hence my concern at the huge discrepancy my friend is getting. He's about a mile and a half from the exchange, and the usual ~3.5meg speed is fine for his purposes. Yes, I forgot to try a different filter, but figured the reported up/down speeds would indicate this wasn't an issue. I'll take a known good one with me and do a quick swap first.

IIRC, on the previous occasion the Openreach "line engineer" didn't seem able to hear the quite obvious "hum" that signalled to me an earth fault, and I had to be quite insistent that he looked further. The first span of dropwire was going through a tree FFS - the very first place I would have looked...

[mixture]

The Flying Pram,

Just a thought, if your router has a stats page is there anything interesting happening on there ? (counters incrementing that look like they ought not to be).

[The Flying Pram]

@mixture - do you mean my own router, or my friends, and what, specifically do I look for? I don't remember exactly how much information the ASUS shows, but I'm going over to wait in for the engineer this afternoon, so I will have time to study it more closely. My friend has been asked by BT Wholesale to carry out speed tests at different times, and what he told me just now is very illuminating: apparently at 5am he was getting 6meg. This is within the upper range suggested for his line length (2km as the cables run) and on the basis that the exchange supports ADSL & ADSL Max (but not ADSL2+). Yet at 7am it shot right down, and during the daytime it can be as little as 100k. If that doesn't confirm your "Rate adaptive DSL not working properly" I don't know what does!

I'm taking another microfilter and the spare router and laptop, but I don't think they will make any difference. Whether I'll be able to post on PPRuNe from his house remains to be seen...

[mixture]

@pram

Doesn't really matter which router as long as its plugged into the affected line.

Basically there may (or may not) be a line stats page on the router, and you'll generally be most interested in any of the counters with "error" in their name, things like :

BER Errors (Bit Error Rate)
CRC Errors (Cyclic Redundancy Check)
FEC Errors (Forward Error Correction )
HEC Errors (Header Error Check)

What you want to do is make a (rough ballpark) note of the errors at T1 ... then go away for a reasonable Tx (maybe to make some tea :E) ... come back and T2 and see which (if any) have incremented and the extent of the increments.

The faster the numbers whizz up, the more likely a line fault .... if they are only incrementing modestly then it may just be nothing to worry about.

The counters won't tell you much about non-physical related issues, such as things happening the ISP network once it hits the exchange.

[The Flying Pram]

Here's the status page (can't easily get it formatted properly, or post as a screencap):

Downstream Upstream SNR Margin (dB): 7.7 22.0 Attenuation (dB): 30.0 14.5 Output Power (dBm): 19.8 12.1 Attainable Rate (Kbps): 9152 1064 Rate (Kbps): 8096 448 K (number of bytes in DMT frame): 254 15 R (number of check bytes in RS code word): 12 4 S (RS code word size in DMT frame): 1 4 D (interleaver depth): 64 8 Delay (msec): 16 8 Super Frames: 1543098 1543096 Super Frame Errors: 8491 6 RS Words: 209861346 26232632 RS Correctable Errors: 705373 8 RS Uncorrectable Errors: 211920 N/A HEC Errors: 6842 0 OCD Errors: 185 2 LCD Errors: 0 0 Total Cells: 500712569 0 Data Cells: 3791416 0 Bit Errors: 0 0 Total ES: 2136 0 Total SES: 181 0 Total UAS: 11 0

Having refreshed the page the HEC errors have changed slightly - 6842/0 but otherwise nothing dramatic.

Current D/L speed is ~0.6-0.8meg - U/L is pretty much unchanged at 0.35meg

[Keef]

I'll defer to mixture on the details, but
RS Uncorrectable Errors: 211920
would make me wonder.

I updated the firmware in my modem yesterday, and it's still showing zero errors, so I have no comparator.

[mixture]

I updated the firmware in my modem yesterday, and it's still showing zero errors, so I have no comparator.

Well, the usual thing you would do in diagnostics would be to zero the counters, leave the box alone, come back and see what the counters have done. So your firmware zeroing may be no bad thing for you in the long term !

So I guess the question in Pram's case, if you were looking at the numbers on their own, could be "how long has that router been turned on and active ?" (since I assume it will clear counters after a power cycle).

Otherwise looking at things over time is another way to do it, and Pram did say the numbers didn't appear to be incrementing rapidly after a period of time, and the line is stable other than being slow.

The SNR and Attenuation is quite good, so there's probably nothing substantial in the physical fault department.

It could be something as simple as a faulty line-card or port at the exchange, that a quick cable swap to a different card/port would sort out. Or it could be something at software level that's either got confused, or been configured that way, either accidentally or deliberately.

In terms of HEC errors:
I think they relate to uncorrected and discarded headers, so I guess if they are incrementing at a reasonable pace, this could possibly be correlated to slow line speeds and/or connection instability.

Are the OCD errors incrementing up alongside the HEC errors ? OCD is an incrementing counter that gets incremented when 7 consecutive HEC errors occur.

I think at the moment, going on the limited information we've got so far, my money's still on this being an ISP side issue. Certainly seems Pram has gone though most of the motions to cover the most common home-side problems different router/filter etc.

[Keef]

Yes, the update zeroed the counters. But then, everything's working well at the moment so I have no cause for concern. They're still at zero.

Sounds as if there's something "upstream" from TFP's friend's master socket that will need Openreach, ISP, or other techie input.

I have a spare decent quality modem/router (my old Draytek) that I take to friends' houses to check their systems. Sometimes it does a lot better than their ISP supplied job: a new one from the ISP fixes it most times. I doubt that would help in this case.

[Ancient Observer]

Pram,
At our local green box, and at our Exchange, GPO/Openreach folk have for ever been fiddling.
Before I moved to 2+ and then ftc, I had a similar sounding problem as your friend.
GPO, or whatever they were called then, tried lots of things over a period of time.
In the end, cable-swapping at both the green box, and (they told me..) at the exchange fixed it, as Mixture suggests. Only after stroppy letter to Chairman and CEO, when their "office" took ownership of the issue.

[mixture]

Only after stroppy letter to Chairman and CEO, when their "office" took ownership of the issue.

Indeed. Another one of my favorites.

Lets hope Openreach have sent Pram a mildly competent engineer who is actually trained to troubleshoot broadband and has brought the right diagnostic tools with him. Fingers, knees and toes firmly crossed !

Otherwise, Pram should throw his proverbial toys out. Getting the HLE team (a.k.a "chairman's office", but its not really ) on the case will get you results much quicker, or indeed sometimes give you results you would have never got, no matter how long you patiently waited !

[The Flying Pram]

Sorry for the delay in replying, but my friends Jack Russell didn't want to go back in the house when I was leaving and the little b******d bit my hand making typing somewhat difficult yesterday...

Lets hope Openreach have sent Pram a mildly competent engineer who is actually trained to troubleshoot broadband and has brought the right diagnostic tools with himFortunately it turned out to be a former colleague and "team member". He joined a few years before I took the money and ran, but at least did a reasonable amount of training, unlike the herbets they take on now. We had a good catch-up conversation, and I was pleased to hear he's following in my footsteps - as far as questioning every stupid idea put forward by the current breed of "managers"! Indeed, he had a blazing row with one, leading to him being banished to the sticks (which suits him just fine).

I digress.... I explained the testing I had already done, and he agreed that it wasn't likely to be a local problem. Nevertheless he carried out a full line test, which verified no insulation, excessive resistance or imbalance faults. Then he plugged in an all singing/dancing tester which incorporates a basic router - this happily synced and reported exactly the same downstream/upstream speeds as both the units I had tried. He replaced the Linebox, because (apparently) the previous numpty had not connected the incoming pair to the right terminals. I didn't think it was right at the time, but the latest version is not something I had ever worked with. This wasn't a fault as such (and made no difference to the speeds), but at least we now have the basics right. He then said he would raise a "Lift and Shift" job - which is BT slang for moving the line to a different card slot (or possibly a different card altogether).

Obviously I will have to wait for my friend to report back - I can see two possible outcomes: 1) it solves the problem, or 2) nothing changes, in which case it would seem like an ISP issue. He's using one of the smaller companies and paying quite a bit over the odds, but whether it is a supposedly non traffic shaped line I don't know. We are talking about a small rural exchange without ADSL 2+ or LLU equipment, but a (reliable) D/L speed of about 5-6 meg would be perfectly adequate for his purposes. So far I've never seen more than about 3.5, so clearly something is going on...

One interesting snippet I gleaned from our chat is a relatively recent problem of "crosstalk". In my day this was invariably due to "split" pairs in exchange to customer cabling, but now it seems that the near universal take-up of broadband is having a similar result. Apparently, any particular cable is O.K. until about 80-90% of the pairs have broadband running over them, at which point troubles start occurring. Of course the solution would be investment in extra cable capacity, but that costs money...

I'll post back with any news.

TFP

[mixture]

Hello TFP,


Thanks for the update and sorry to hear about the canine injury.

Sounds like your ex-colleague is one of those rare breed of engineers that BT could do with (a LOT) more of ! Sounds like he's firmly on the ball.

[The Flying Pram]

Contention....

Well a week's gone by, and you might think that my friends broadband speeds are now what they should be..... and you'd be wrong! Unbelievably, he has been run around still more, and another Openreach Broadband engineer came out today. This one connected a BT Hub and (surprise, surprise) got more or less the same results as last time... He blames contention as the issue - no s**t, I had sussed that out ages ago. Apparently, my friends ISP Unicom use TalkTalk to provide the actual service, and they are too damn tight to pay for sufficient bandwidth! This guy said he regularly has to fend off angry customers because of the same thing. Once more the issue has been "passed back" and daytime speeds continue to vary from ~600k to 1.2meg. Friends neighbour has just changed to a FTC service and gets about 23meg, but having read many posts about BT's implementation of this service it's only a matter of time before contention starts showing up there...

This weeks announcement by Ofcom about improved customer rights, as regards to poor speeds and difficulty getting out of contracts, is perhaps timely. The aforementioned ISP has a particularly bad name as far as the latter goes, so it will be interesting to see what transpires...

TFP

[BEagle]

The Flying Pram wrote: Apparently, my friends ISP Unicom use TalkTalk to provide the actual service...

Ah, TalkTalk....:mad:

When Virgin meeja flogged off its ADSL service to TalkTalk, I wanted nothing more to do with my Virgin 'National' service as I had absolutely no faith in the ability of TalkTalk to maintain the service.

With my 'up to 8 Mbps' service, my last test showed that I was getting 4.12 Mbps down / 1.28 Mbps up on the old system. But I was then able to change to Virgin's Superfast broadband as the road on which I live has legacy ComTel fibre optic cabling.

Apart from the abysmal customer service and administrative incompetence which I suffered from Virgin, the Internet service has been very good. I signed up for their 'up to 50 Mbps' service and was regularly getting 60 Mbps down / 2.9 Mbps up. But since early March I'm now seeing 109 Mbps down / 6.0 Mbps up, which is excellent.

My firm guess is that TFP's friend's problem (apart from that JR) is with TalkTalk's ISP policy.....