In reply to a post by asbokid:

Openreach has configured a vdsl line-profile in which the upstream and downstream transmission rates are adapted at start-up only. After line activation, the transmission rates cannot change. Any change to the transmission rates requires the retraining of the line.

Openreach disables the feature, but the Huawei MA56xx DSLAMs can also dynamically adapt transmission rates at runtime (i.e. during showtime). If enabled, this does not require the retraining of the line.

cheers, a

SRA?

I read it , but obviously openreach have their reasons for not utilising it. Most likely hard to get proper CPE compatability.

Actually another little think I am guessing the IP profile system is a reason why they turn it off, currently if the speed changes it needs a resync and mostly also a ppp recconect which establishes a new IP profile. Put SRA into the mix and IP profile would never update as both sync and ppp wouldnt drop.

Edited by Chrysalis (Sun 19-May-13 06:38:33)

BE trialled SRA (obviously not for VDSL) some time ago, and decided not to go ahead; I don't know the reason. It looks excellent on paper, but I assume there are enough technical issues that the actual gains didn't make up for the pains.

I can't recall where I saw this, but it was a document on a variety of DLM/DSM techniques...

The impression I got for SRA was that it worked well on an individual line (ie DSM level 1 activities), but wasn't very good in terms of spectrum efficiency to other users (ie DSM level 2). Something where it effectively caused everyone to run at highest power, and therefore increase crosstalk.

I'll try to find it.

Edited by deleted (Sun 19-May-13 12:29:17)

The issues on ukonline is it only worked with a very small amount of equipment, typically the older billion 7402s. But when it did work it was a complete transformation for poor lines. On good stable lines probably not so much a benefit as I dont really miss it on FTTC, but on my adsl service it was a complete game changer, I went from 3 to 100 or more drops a week to having sync uptimes around a year (on 6db margin, fast path).

In reply to a post by Chrysalis:

In reply to a post by asbokid:

Openreach disables the feature, but the Huawei MA56xx DSLAMs can also dynamically adapt transmission rates at runtime (i.e. during showtime). If enabled, this does not require the retraining of the line.

SRA?

Yes, Seamless Rate Adaptation.

Although what Openreach does implement for dynamic rate adaptation is a theoretically superior method, any way.

By opting for a bespoke centrally-controlled DSM Level 1 / DLM algorithm, the network operator isn't stuck with the vagaries of a vendor-specific DSLAM-based solution. The algorithm can be updated or even replaced altogether at any time. And, fundamentally, it can tweak a range of parameters in the channel-profile, not just the data rates.

The SRA is limited to tweaking the minimum and maximum data rates (banding), but the Openreach DLM algorithm can also modify the interleaving delays and impulse noise protection levels (and theoretically the TSNRM too). And by centrally logging that line data on an individual basis, long-term trends for a line can be discovered heuristically by the algorithm, too.

When a subscriber port is initially activated at the DSLAM, a line-template is bound to the port. That template comprises three components: a line-profile, a channel-profile (and an alarm-profile).

The line-profile is configured identically on all lines across Britain since much of it is stipulated in regulations. For example, the line-profile defines the VDSl2 profile (17a), the PSD class mask (B8-11), the DPBO and UPBO PSD masks, as well as the target SNRM (6dB) and the rate adaptation method (set at startup), etc.

However, there is just that one, solitary line-profile in use in the copper access network.

In addition to the line-profile, Openreach uses nearly 200 different definitions of channel-profile. These channel profiles specify the minimum upstream and downstream bit rates (that's the banding part of the profile), the minimum reserved bit rates, the interleaving depths, and the minimum INP levels, etc.

During its daily iterations, the centrally-controlled DLM algorithm processes all of the individual line data that has been retrieved (by ANCP/SNMP) from DSLAMs under its control. This data is examined to discover the number of coding violations and the number of (downward) re-trains on a given line. The algorithm then classifies the line under one of those 200 different channel profiles. This could see, for example, the interleaving and INP levels raised and/or the introduction of fixed maximum and minimum data rates to increase stability. Or, ideally, the channel-profile will be left completely unchanged.

cheers, a

edit: here are the Huawei docs for configuring a TR129 VDSL Channel Profile [1]. The docs illustrate the range of parameters that can potentially be tweaked by the DLM algorithm.

[1] http://pdfcast.org/pdf/tr129-vdsl-channel-profile-ad...

Edited by deleted (Sun 19-May-13 22:29:30)

well which is superior I guess depends on what way you score it.

I look at outages and affect on latency.

The DLM system requires an outage to make an adjustment, SRA does not. SRA doesnt touch latency, and I consider a loss of speed superior to a bump on base latency, especially as interleaving tends to reduce speed on top of the bump on latency. Plus I think that DLM simply can go too far, not go far enough, act too late etc.

eg. if we think back to my previous issues on my FTTC line I had a 2 hour occurance in the early hours of the morning of noise on my line of some sort. DLM did nothing whilst that was ongoing, but then some time later after it stopped it applies interleaving to my line affecting the line performance when there is no longer a problem. Whilst SRA would (when it works) react near enough instantly to any problems.

I think unless one has suffered from a very bad line they cannot appreciate how good SRA is. On my old ADSL line DLM was near enough useless. Isp tech's consistently berate DLM as well saying when they compare they find bad lines perform better without DLM.

I am less fussed now as I no longer am suffering like I did on my adsl line. But I still find DLM annoying, makes me scared to do anything with my modem incase I upset DLM.

Edited by Chrysalis (Sun 19-May-13 19:40:48)

makes me scared to do anything with my modem in case I upset DLM. All part of their devious plan.

I found my references.

The thing that seemed to work worst on external effects - labelled "zap your neighbour first" in one paper [1] - was called "Virtual Noise". It causes DSL to end up like everyone shouting at a party.

SRA wasn't so bad [2]. It seems to be aimed at SLOW changes to the noise picture, not fast ones. The negatives were that it still gave a problem to power control of the line - As the line adapts down, it used less power, but as it adapts up, the power goes up too.

I think that prospect is probably the killer when DLM propagates to level 2 - when the aim is to manage the power level of all the lines, so spectrum (power and frequency) gets shared more equitably. It doesn't work in a predictable way if the lines are then "randomly" altering their power settings.

In the meantime, we've definitely shifted from the AMA style of DLM to TRA [3].

[1] http://www.stanford.edu/~achowdhe/Papers/Inderscienc...
[2] http://www.broadband-forum.org/technical/download/TR...
[3] http://www.niccstandards.org.uk/files/current/NICC%2...

I think it was designed to deal with gradual changes like daytime to nightime, however on my adsl line it did also work pretty well in dealing with large sudden noise bursts,which without SRA would have either left the modem holding onto sync but with unusueable internet or a resync. It was able to make instant large % swings to sync speeds without downtime eg. from 6400kbit to 4500kbit in one split second when a noise burst started.

However with that said, the SRA utilised on ukonline wouldnt have helped my FTTC fault, because that SRA worked by making the snrm stick like glue to the target snrm. So if eg. the snrm fell to 5db the sync speed was reduced to peg the snrm back to 6db. On my FTTC fault, the snrm did not change, I just had a burst of errors but no change in noise margin. But there is a 2nd type of SRA which monitors error rate and adjusts sync speed independent of the current snr based on error rate, that 'may' have helped my FTTC fault.

Edited by Chrysalis (Tue 21-May-13 17:16:55)