New fibre connection: Do you agree about these stats?

At the end of last week I changed my ISP, also taking the opportunity to swop from ADSL to VDSL (FTTC). I've also updated my router-modem firmware (Billion 8800NL).

The changeover went reasonably smoothly and I'm now running on a much faster DSL connection than before (previously 4M bps downstream). With the new ISP, I chose to have a 40M/02M bps package.

I've now lifted a few of the figures from the 8800's xDSL Stats page:

*******Down**Up

SNR (dB) 9.1 18.9
Attainable Rate (Kbps) 46841 8655
Rate (Kbps) 39999 1999
Interleaver depth 16 1
RS correctable errors 59795 0
RS uncorrectable errors 0 0
HEC errors 0 0
OCD errors 0 0
LCD errors 0 0

Strangely, though, the other day I downloaded a 9.9MB file and the download itself reported that that took 3 secs. If my arithmetic's right, that works out to a data download rate of just over 26M bps.

So, where's all my speed gone? 26M bps seems slow, for the actual data. Agreed?

For, say, a nominal 40M bps connection, what's the data profile? I'd have thought the data rate would have been around 37M bps, not 26.

The copper leg is approx 700 metres long. That's the walked route (underground ducts), so initially I was expecting the line to sync at 30 - 35M bps, with a commensurate data rate somewhat below that. The Openreach speedtester had predicted 37 - 40M bps.

I'd been aware that Openreach normally sets the target SNR on downstream to 6dB and, sure enough, that was the SNR that the line was working at just after I first got it all working. But a couple of days later I had to restart the 8800 a number of times in the course of doing its firmware update, and therefore I'm not now surprised to find that the DLM has bumped up the SNR to 9dB. Could that at least be a partial explanation for the 26 figure? I assume that, in time, the 9dB will drop back down toward 6dB again.

Frankly, I can't believe the 8800's figures for the Rate. Both the Down and Up Rate figures are recorded as all but maximum! And notice that the attenuation figure for Up is zero, and clearly that's impossible. Interleave depth was originally 1 and 1, but is now 16 and 1.

Do you agree that some of these 8800 stats are decidely questionable? Should I just leave the 8800 to get on with its background co-operation with the DLM and hope that everything just settles down later on?

Incidentally, within the 8800's user settings there's something called 'Fullcone NAT'. Anyone know anything about that? Given that my 8800 connection is straighforward PPPoE (no bridging or additional firewalls), do I need to select and use Fullcone NAT?

I think a 3 second download is way too short to draw any conclusions, so I suggest you try a much larger filesize say 500MB- 1GB.

You could always try the speedtester on this site and also the BT speedtest should give you your IP profile if you proceed to 'further diagnostics'

Oh, and the stats look fine

Edited by deleted (Fri 10-Feb-17 13:04:03)

There are SO many variables that play a part in a download.

A 3 second test doesn't really give much chance to let the download ramp up.

Try downloading a file that's several Gig and see how long it takes.

Make sure you're downloading from a site that can provide the data at that rate. Use a wired connection to the router and not on an old machine that might struggle with that data rate.


Have you ran any speedtests using a wired connection?

You're connecting at the max rates for the service you're on. Not sure why you're questioning if the router is lying about the connection rates.

The reason the SNR is 9 is because you're service is limited to 40meg, but your line seems capable of around 46. So there's "room" for a higher margin, which in your case, is a good thing, as it gives more protection to the connection.

Edited by deleted (Fri 10-Feb-17 13:06:04)

try the 1GB test file download here and then let us know what speed you got.

In reply to a post by meditator:

Do you agree that some of these 8800 stats are decidely questionable?

Should I just leave the 8800 to get on with its background co-operation with the DLM and hope that everything just settles down later on?

No.
Yes.

In reply to a post by meditator:

The copper leg is approx 700 metres long. That's the walked route (underground ducts), so initially I was expecting the line to sync at 30 - 35M bps, with a commensurate data rate somewhat below that. The Openreach speedtester had predicted 37 - 40M bps.

My first FTTC line was 650m of 0.5mm copper.

On the original 8MHz profile used by BT, it couldn't quite manage 40/10 speeds (the top packages at the time), and tended to sync at around 38/10.

When BT swapped to 17MHz profiles in 2011 (but before they started selling the 80Mbps packages), it managed 40/10 easily, and reported an attainable of around 60Mbps.

Increased takeup would have almost certainly caused crosstalk, and reduced those numbers, but I moved out soon after.

In reply to a post by meditator:

I'd been aware that Openreach normally sets the target SNR on downstream to 6dB and, sure enough, that was the SNR that the line was working at just after I first got it all working. But a couple of days later I had to restart the 8800 a number of times in the course of doing its firmware update, and therefore I'm not now surprised to find that the DLM has bumped up the SNR to 9dB. Could that at least be a partial explanation for the 26 figure? I assume that, in time, the 9dB will drop back down toward 6dB again.

You should only expect the modem to be running with an "actual SNRM" value that matches the "target SNRM" value of 6dB when it is synchronised at something below the package maximum.

You are sync'ed at the maximum, on a line capable of a higher speed. The modem is thus under-utilising the line, leaving it with spare SNRM capacity over and above the target: 3.1dB spare at the moment.

Note that DLM does not bump up your "actual SNRM"; changes in the "actual" reflect the combination of "spare SNRM" from the time of the sync and subsequent changes in noise (which can be more than 3dB). DLM doesn't even bump up the "target SNRM" in FTTC at all. Trials are starting to try out reductions to 5, 4 or 3dB though.

What DLM would do instead, if it saw the need, is to "band" you: setting an artificial maximum sync speed. When the modem syncs, you'd see it reach a few killobits below a round figure (eg 36997 kbps), and a higher SNRM.

Having spare SNRM capacity does not cause a slowdown of throughput.

The throughput figure will still be based on the sync speed. You have a 40Mbps sync, so will have an "IP profile" of around 37.7Mbps. That's the maximum throughput of IP-level packet data (including IP headers); a download of a large batch will result in a user throughput in the region of 36-37Mbps ... provided the server can keep up.

In reply to a post by meditator:

Interleave depth was originally 1 and 1, but is now 16 and 1.

With an initial interleaving depth of 1, you were on fastpath with no error correction. With a change to 16, DLM has turned on error protection in the form of G.INP retransmission.

This change can improve speeds for a line, because there is a perceived coding gain from the protection. That is probably why your actual SNRM has increased from 6dB to 9dB, and your attainable speed has increased.

Note: DLM actually sets the "INP" and "delay" parameters, and the modem/DSLAM choose an appropriate depth during synchronisation. It can be better to monitor INP and "delay", rather than depth.

In reply to a post by meditator:

Frankly, I can't believe the 8800's figures for the Rate. Both the Down and Up Rate figures are recorded as all but maximum! And notice that the attenuation figure for Up is zero, and clearly that's impossible.

It certainly looks like the sync figures are right. For more details, you probably want to connect via telnet, and get figures out from the command line.

If you want to monitor 24x7, there are monitoring programs available here:
http://forum.kitz.co.uk/index.php/board,46.0.html
Both "DSLstats" and "HG612 modem stats" are capable of monitoring an 8800NL. In fact, the 8800NL behaves very similarly to the HG612.

The attenuation figure is always shown as zero, upstream.

In reply to a post by meditator:

Incidentally, within the 8800's user settings there's something called 'Fullcone NAT'. Anyone know anything about that? Given that my 8800 connection is straighforward PPPoE (no bridging or additional firewalls), do I need to select and use Fullcone NAT?

I just have "NAT" ticked, with "fullcone NAT" left off, as the default setting. I've needed nothing else.

Whether that is what you want depends on your own uses. Take a look here:
https://www.think-like-a-computer.com/2011/09/16/typ...

mediator

Your 9.9Mb file taking 3 sec is about right for that size.

The download could have been 2.1sec ( rounded up!). The download takes a few ms to start and the distant end speed also comes into account. Your PC also has some lag at the start and end of file.

If the file was much bigger these factors all tend to be the same size so disappear ibto the actual download time. Leaving just the distant end speed ( can be shared across many downloaders!) and the core network Contention to slow your download.

A 500Mb file would give a better estimation or a speedtest from here which is dedicated at both ends to you just leaving the core network to slow things down, the x6 test usually removes this leaving with an accurate test of your connections speed.

See a speedtest graph from the site for the effects at the start. it takes a little while to ramp up.

Thanks, everyone, for enlightening me over these stats.

Yes, I knew that my new FTTC account was specifically capped at 40M bps max but I truly didn't anticipate my line being capable of supporting anything much beyond about 35M bps, because that's what the Openreach speedtester and various published graphs predicted, before going FTTC, for my 700m Cu leg.

Yeh, by "bumped up", the word I was probably looking for when I wrote was "banded".

I run my setup specifically on wired Ethernet, BTW. I might run a proper speedtest to get a better confirmation of the present running rate.

I missed quoting the attenuation figures in my list. The downstream attenuation is 22.6dB. Glad to find out from you that the upstream is always given as zero.

I'll leave Fullcone NAT unchecked.

Batboy, KitKat et al,

Have just used the link you gave to the TBB speedtester, the one where you just insert your postcode. I was still logged in to the forum at the time but apparently I should have also logged in to the site, for the results to have been more widely recordable as a stat.

Anyway, it was all over very quickly and the results were:

Downstream 37.3M bps
Upstream 1.7M bps

so I'm well pleased.

The test didn't state the size of the file. Anyone know it? It appeared to run two tests per direction, presumably with different filesizes.

mediator

The two tests should have been labelled on the graph as 1x and 6x.

The difference between them is usually due to contention in the ISPs network. So the x6 test will show closest to what your connection can give and the x1 test will show what you will normally get on a single download. ( Steam etc will use many downloads until it saturates your connection.).

The file size will depend on the speed of ypur connection, faster the bigger , to overcome the issues that a small file over a fast connection would give or a big file over a slow connection would take too long.

You should be able to post a link to the results so that others can see the graph.

I'm now thinking of upgrading later this month to 10M bps upstream nominal. It'll cost me comparatively little extra. I'll then re-run the TBB speedtest and publish the results here.