Print

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

[burakkucat]

I have recently come across a posting by WWWombat in one of the ThinkBroadband fora.

It reads --

What it really means is that DLM is a bit of a catch-all name.

There are *many* different schemes that can be used under a broad title of "DLM" whose job is to react to differing line conditions, and attempt to bring stability to the digital comms going on over the line.

Some schemes are more dynamic than other, and can react faster to quick changes to conditions. Some are slower, or even static. Some schemes work well with other lines, while some schemes have a detrimental effect on other lines (I've seen one described as "zap your neighbour first"). Others have a green agenda, and attempt to reduce power usage.

Our understanding of various DLM schemes is advancing all the time, and comparisons made by the likes of Ofcom. An Ofcom subgroup (NICC) has a report on DLM techniques in the UK, but it only reports on a subset of the possible schemes. The Broadband forum has a newer report on DSL Quality Management techniques that gives a broader overview.

The major scheme used by BT Wholesale for ADSL Max and ADSL2+ is one that adjusts the target noise margin after errors (from the 6dB norm to 9, 12 or 15 dB). This scheme (which is the one mentioned by RobertoS) is known as AMA, or Automatic Margin Adjustment.

NICC prefers the newer scheme, known as TRA or Tiered Rate Adjustment, which aims for stability by reducing the sync speed targets into tiers. This is the DLM scheme used by BT Openreach in FTTC.

While BT Wholesale's ADSL Max products only use AMA, their ADSL2+ products start by using AMA but for intractable problem lines, they also have a form of TRA - but this is very much used as the last resort. Openreach FTTC *only* employs TRA.

Both TRA and AMA schemes utilise forward error correction as part of their strategies. The FEC and interleaving changes bring more stability, but the overheads reduce speed and add latency.

For FTTC, however, the specification requirements for the VDSL2 modems bring an alternative mechanism into play - the use of "PHYR", or retransmission at the physical layer. This will do away with the "static" overheads of FEC and interleaving, and replace them with a retransmission overhead that only happens when errors actually occur.

It seems likely (to me) that PHYR will be introduced either alongside vectoring, or afterwards.

Alike? Not even close.

There are a number of explicit statements made but without any references. Perhaps those statements will provide suitable starting points for further searching by one of our 'ferrets'? 

[Chrysalis]

I dont know if he has inside knowledge (so as such he cant give references) or its just down to observation of how lines are behaving.

ADSL DLM controls lines via both adjusting snrm and interleaving, it seems to have a preference to snrm.
VDSL DLM controls lines via banding and interleaving, it seems to have a preference to using interleaving.

Banding and snrm adjustments aim to do the same thing, cap the line speed, but the latter can be overriden by the user's modem, the latter also can be ineffective at capping speed even without end user's playing with margins.

Personally if my line has to be managed to give stability I would almost certianly prefer a higher snr margin to give that stability (banding) over interleaving as the affect of higher latency on performance is very apparent on most activities.  eg. a 10mbit line with 10ms latency to an endpoint should feel faster than a 20mbit line with 50ms latency to the same endpoint. 

This is why I have already decided if I ever get stuck on interleaving due to line instability I will downgrade to 40/10 to get enough snrm buffer to help DLM revert to fast path.

[ColinS]

Thank you for that B*Kat.  I have downloaded it for further study.

Personally, I can't disagree in principle with anything that the respected Wombat has said, albeit that in some cases it's a matter of semantics,
e.g. he says

... TRA or Tiered Rate Adjustment ... is the DLM scheme used by BT Openreach in FTTC.

whilst noting that

Both TRA and AMA schemes utilise forward error correction as part of their strategies.

Our esteemed colleague Asbokid (please come back, all is forgiven!) said (talking about Huawei FTTC DLSAMs) back in May that

It's understood there are nearly 200 different channel profiles. (192, I think he has subsequently sent me)

Many combinations of downstream and upstream transmission rate bandings don't exist in the wild.   

E.g. there is no channel-profile with [ 0.128 - 2 Mbps Downstream and 10 - 20 Mbps Upstream ]

In fact, as I recall, it was your good self that prompted that by republishing here a set published elsewhere by Plusnet. 

So, the Huawei channel profiles combine rate banding with INP and Interleaving (amongst other parameters), so that e.g. there are n 80/20 profiles with values of INP=0,3, etc, and Interleaving=1, x and so on, and my experience suggests it selects amongst them depending upon the extent to which it sees the line as suffering from instability (too many resyncs), or performance issues (too high an error rate).  Just my observations atm. 

[Chrysalis]

has asbokid left?

[kitz]

Thank you for the link...   I havent looked any further or done any research, but I would tend to agree that interleaving tends to be dealt with differently with FTTC than on adsl2+.

With having an intermittent fault, I appear to be seeing signs of what I understand PhyR to be all about.  In other words error correction without interleaving and only at times when the noise is bad.   

[kitz]

has asbokid left?

Not that Im aware of, I just assumed that perhaps he was away.

[Chrysalis]

Thank you for the link...   I havent looked any further or done any research, but I would tend to agree that interleaving tends to be dealt with differently with FTTC than on adsl2+.

With having an intermittent fault, I appear to be seeing signs of what I understand PhyR to be all about.  In other words error correction without interleaving and only at times when the noise is bad.   

FEC can be done without interleaving on adsl also, just I have never seen an isp deploy it that way.  For openreach to be considering it now, I assume its enhanced to the point to make it viable to use as when I tried it on my old ukonline adsl connection it barely made an impact on crc errors.  If this new algorithm has no latency penalty alongside dynamic speed penalty then it is indeed interesting.

[ColinS]

For anyone who can work out the bitmap, and wants to try it to see if the DSLAM will swallow it (This is where we need Asbokid):

xdslcmd configure [--phyReXmt <0xBitMap-UsDs>]

[Edit] it may be as simple as 0x01, see here:
# xdslcmd profile --show

Modulations:
        G.Dmt   Enabled
        G.lite  Enabled
        T1.413  Enabled
        ADSL2   Enabled
        AnnexL  Enabled
        ADSL2+  Enabled
        AnnexM  Disabled
        VDSL2   Enabled
VDSL2 profiles:
        8a      Enabled
        8b      Enabled
        8c      Enabled
        8d      Enabled
        12a     Enabled
        12b     Enabled
        17a     Enabled
        30a     Enabled
        US0     Enabled
Phone line pair:
        Inner pair
Capability:
        bitswap         On
        sra             Off
        trellis         On
        sesdrop         Off
        CoMinMgn        Off
        24k             On
        phyReXmt(Us/Ds) Off/On
        TpsTc           AvPvAa
        monitorTone:    On
        dynamicD:       On
        dynamicF:       On
        SOS:            On
        Training Margin(Q4 in dB):      -1(DEFAULT)

Which suggests, as far as the modem is concerned, it has the capability switched on in the DS already
Broadcom PhyR