Skip Navigation

Openreach appears to be taking lower noise margin to trail out across nation
Friday 17 March 2017 14:41:04 by Andrew Ferguson

A trial has been running looking into whether Openreach can squeeze more speed out of VDSL2 for some lines with no hardware changes in the cabinet or by the customers VDSL2 modem and it seems on Monday 20th March, this limited area trial is set to go national.

The trial involved the Openreach DLM system identifying lines that were stable at a 6dB target noise margin, and then lowering the target noise margin in 1dB steps over an extended period and monitoring the error rates and connection speeds. The monitoring is not new, the Openreach DLM (Dynamic Line Management) is constantly running, the change is that while 6dB is currently the lowest target noise margin modems are told to negotiate the connection at and now for lines that look like they can cope with lower margins without undue numbers of errors lower margins will be allowed. The bonus is thus more download speed without any cost.

It is thought the trial involved some 40,000 lines and we presume Openreach has seen positive results from the trial, we are monitoring what we believe is the trail area but to date have not seen an observable difference, but this may be that the 40,000 lines is swamped by those not in the trial but still in the area we've been monitoring. It may be once we run the Q1/2017 analysis in a couple of weeks we can see some indication of a difference, or it may take many more months once things go national.

The way the target noise margin's work mean that if your line is stable, then it does not matter if you connect at 21,232 Kbps or 44,536 Kbps its all about if the DLM thinks the line is stable and suitable for trying a lower margin, what will differ is the amount of benefit different people see and many people may see no change at all. Its important to emphasis that the drop is in 1dB steps down to a minimum of 3dB, and the margin may return back to the 6dB level if things do not look stable.

Sweating the assets to squeeze what might be just be a 2 Mbps average increase seems like a lot of work for little extra speed, but for individuals it might be worth much more speed, and given that across the UK if most users getting 24 Mbps were to see speeds jump to 30 Mbps this could mean that models on performance versus distance would have to change and a 6 Mbps jump for lines in the 1km long region would be worth 0.6 percentage points to the UK superfast total. CAUTION Don't take the 0.6 change as been what we expect from the roll-out, it is just an illustrative figure and we will only change our model if we see a significant change, in the past monitoring experience against our model was more difficult but since Autumn 2016 we have automated this so we can keep a better eye on how all the various changes are affecting the model we use.

Update 7pm While the trial is moving national, it is not happening overnight, but will be done in phases. The phasing is to allow Openreach and the broadband providers to assess how things are progressing. The phased roll-out should complete by September.


Posted by gt94sss2 about 1 month ago
Where do you get the stat that this "might be just be a 2 Mbps average increase"? - I would expect the attainable rate to increase by a lot more on lines which will benefit
Posted by TheEulerID about 1 month ago
It would be interesting to know if this would make a noticeable differences to the coverage at 24mbps & 30mbps. It might be quite modest as marginal lines will gain the least, but I suppose a gain is a gain (and, presumably it would also impact on the proposed 10mbps USO).

Alternatively, it might just be that this recovers some of the loss of increased x-talk as take-up increases. More damage limitation than gain.
Posted by bpullen about 1 month ago
On shorter lines, I've seen as much as an ~8mbps+ increase in sync speed as a result of these trials.
Posted by mr_mf about 1 month ago
I am not in a trial area and my line was reduced to 4db margin on Monday so it is already live.

According to Kitz.

~ For existing Sync speeds of around 20Mbps - each 3dB is worth 3Mbps
~ For existing Sync speeds of around 40Mbps - each 3dB is worth 6Mbps
~ For existing Sync speeds of 60Mbps or more - each 3dB is worth 11Mbps.

I've gone from 64MBps (6db) to 75MBps (4db) so suggests that is about right.
Posted by mr_mf about 1 month ago
Sorry should have said line length is 300 meters.
Posted by andrew (Favicon staff member) about 1 month ago
Average speed increase, indicating that it was more than one line, notice the bit later where we say for an individual it may be a lot higher.

Wanted to avoid saying it will be worth precise figures in item, since its going to vary per user and its not going to be an across the board drop, i.e. some will see no change.
Posted by R0NSKI about 1 month ago
I can tell you one set of users that this will make absolutely no difference to, and that is anyone on the wretched ECI cabinets such as me.
Posted by SlimJ about 1 month ago
Mine dropped to around 4.5db when I checked last night, syncing at 22Mbps. Was over 6db syncing at just under 20Mbps. I'm around 1.25 - 1.5km from the cabinet.
Posted by CarlThomas about 1 month ago
Had a resync at 00:24 today. Have an extra couple of megabits.

Not sure what to do with it all. My life is now complete and I can die happy. I'm nearly up to 2/3rds the download speed I had 6 years ago.
Posted by adslmax about 1 month ago
It won't make any different in lower SNR if the line is full sync at 39999/1999 or 9999 and 79999/19999
Posted by zyborg47 about 1 month ago
Is it worth the risk of instability for a little bit extra speed? BT DLM is not the best to be honest and I remember the problems on ADSL when they mucked around with it, which is why I went onto ADSL24 as they used Cable and wireless network, far better than BT, but with Fibre, most of use have no choice but to use BT system.
Posted by philipd about 1 month ago
Papering over the cracks of cross-talk as vectoring is never going to happen for the masses on VDSL. If Long Range G.FAST goes live using the VDSL frequencies, that pulsing time division multiplexing of G.FAST is likely to knock more off our speeds and 3db margins will vanish up and down the country.
Posted by underzone about 1 month ago
Schoolboy spelling mistakes throughout.
e.g. noise margin "trail" national
Posted by Aaron_01 about 1 month ago
I sync at 71233 so looking forward to seeing it max out to 79999 if mr_mf's post holds true! I reckon it will hold stable, had no noticable errors/line drops in the past year since moving into new property.
Posted by steve14 about 1 month ago
I'm more interested in a stable connection and low ping times because I'm a gamer. Will having a lower noise level cause packet loss or make any difference to ping times?
Posted by mr_mf about 1 month ago
You will only get packet loss and higher pings if G.INP fails to keep up with the increased noise on the line. With 4db margin my FEC errors have increased considerably but most importantly I don't get any Error Seconds on the downstream.

Previous 1 day time = 24 hours 0 sec
FEC: 127159 712
CRC: 0 156
ES: 0 141
SES: 0 0
UAS: 0 0
LOS: 0 0
LOF: 0 0
LOM: 0 0
Posted by Terranova about 1 month ago
The article should be edited to state that those on a ECI cab wont be seeing any of this It requires G.INP and ECI doesn't have that and with no real indication as to when or if it will be re-rolled any time soon.
Posted by kijoma about 1 month ago
glad i use a broadband system with typically 30db of margin. a SNR of under 6dB is asking for problems on any data comms system.

TANSTAAFL rule applies always. There ain't no such thing as a free lunch.
Posted by CarlThomas about 1 month ago
30dB of margin? That is an awful lot of wasted bandwidth.

30dB SNR and 30dB margin are pretty different things. Sure you don't mean 30dB SNR, good for 64QAM or 128QAM comfortably, maybe 256QAM safely with good error correction?
Posted by JDPower about 1 month ago
Kijoma, beware the double negative ;)
Posted by davidcrocker about 1 month ago
Perhaps this explains why I was getting a lot of disconnect/reconnect events about 2-3 weeks ago. I've just done a new speed test and it's giving me 25.9Mbits/sec. Previously the most I ever had was 24Mbits/sec. Noise margin is showing as 10.6db up. 4.0db down. Attenuation is 38.1 up, 25.1 down.
Posted by chrysalis about 1 month ago
I am on my own exclusive ECI 3db trial, after a power cut I found myself synced at 2,8db SNRM due to connecting before other crosstalkers. :D Went from 71.8mbit (which was at 4.8db SNRM due to crosstalk) to 79987, I dont know what my max attainable was at time of sync tho.
Posted by andrew (Favicon staff member) about 1 month ago
You'll have synced to the usual 6dB margin, but as one of the early ones reconnecting, as others reconnected your noise margin reduced as their signals started crosstalking to yours
Posted by chrysalis about 1 month ago
yes thats whats happened, it just so happens the snrm after everyone else resynced has put me just under 3db, hence me calling it my own exclusive 3db trial. :)

I got 499ES with 1 day and 1 hour uptime, well within DLM limits, previously with a 4.9db SNRM I was averaging about 220ES per day. This is on fast path.
Posted by WWWombat about 1 month ago
So broadly 8Mbps gained from that 2dB drop. Not bad...

I so hope ECI have sorted out their issues...
Posted by Blackmamba about 1 month ago
Hi Wombat
I have notice that a customer on a post code in Surrey has moved from 23 meg to 26 meg at a range of 1.1 miles this is very close to Elstead it is starting to look like the Dorking figure 99.7% at 15 meg is going to be achieved. Could it be that the Mamba Project is gain momentum time will tell.
Posted by _Mike_B_ about 1 month ago
@Terranova Nothing has been said that makes a distinction between ECI and Huawei cabinets. The briefing to engineers just says it's being rolled out so to be aware of it.
You must be logged in to post comments. Click here to login.