Print

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

[WWWombat]

I'm using an 8800NL (v1) which works as well as the HG612. I think the v2 works better, and I wouldn't expect the 8900AX to be any worse.

It is hard to say how DLM would have reacted, as this is all new behaviour. The old DLM would tend to only react to the error rates, rather than SNRM values. The high FEC rate wouldn't have affected the old DLM decisions, but it could play a part in the 5-4-3dB decision-making.

I see your line looks to have jumped to a new setting today, at 3.5dB. Sync is 76Mbps, a gain of 10Mbps overall.

The progression (ignoring the blip) has been 66, 69, 72, 76Mbps with SNRMs of 6.4, 5.4, 4.4, 3.5dB.

Those are in line with what I'd expect.

[displaced]

Sadly, no SNRM drop for me, yet.  Perhaps it'll just take a while longer to kick in, or maybe my line's performance doesn't match some criteria I'm not aware of.

Oh well!

[BigJ]

True, we are in new territory now. Hopefully my line will stay nice and stable at the new 3.5 SNRM though I'm not going to complain if it needs to go up a little

[gt94sss2]

As the 3dB is now going national I have two questions:

1. Are lines that are at the maximum sync for their service ub scope? For instance, I am on a 55/10 service but my attainable is 60/20, will DLM lower my target SNRM (assuming the line is stable) to produce a higher attainable figure?

2. Obviously, the big beneficiaries  are those closer to the cabinet, but do those further away benefit "by default" if those closer are now running with a lower SNRM target (does it make their signal strength 'stronger' even if their own SNRM target doesn't change) ?

[ejs]

In theory, I suppose it would be the opposite. Running at a lower target SNRM could allow the same line rate at a lower power output, and so with lower power output, there would be less crosstalk. Relatively speaking, the crosstalk reduction would benefit shorter lines more than it benefits longer lines.

Of course, the answers to your questions are: I don't actually know, you'll have to wait and see.

[DMZ]

This probably has been answered already;
Will this rollout be for ECI too? I'm unfortunately on an ECI cab. 

[Chrysalis]

probably not as I expect g.inp is considered a requirement.

[DMZ]

Oh dear, thanks for answering. 

[skyeci]

This probably has been answered already;
Will this rollout be for ECI too? I'm unfortunately on an ECI cab. 

Openreach has confirmed to ISPreview.co.uk that the roll-out is still due to commence this month as planned (phased roll-out) and some updated performance data from their extended trial is due to be shared with ISPs over the next couple of weeks.

We also asked whether ECI based VDSL2 DSLAMS (Street Cabinets) would continue to be excluded from the roll-out (as first reported here) and they confirmed that the enhancement will only be applied to Huawei based DSLAMS. ECI kit has become notorious for causing problems with various VDSL2 enhancements and changes.

Remember, not everybody will benefit from the new SNR profiles and Openreach’s DLM technology won’t apply the lower margin(s) unless your line is deemed likely to benefit.

[Moderator edited to add an attribution to the quotation.]

[WWWombat]

1. Are lines that are at the maximum sync for their service ub scope? For instance, I am on a 55/10 service but my attainable is 60/20, will DLM lower my target SNRM (assuming the line is stable) to produce a higher attainable figure?

2. Obviously, the big beneficiaries  are those closer to the cabinet, but do those further away benefit "by default" if those closer are now running with a lower SNRM target (does it make their signal strength 'stronger' even if their own SNRM target doesn't change) ?

1. We don't know. I haven't seen one happen, nor seen discussion about it anywhere.

2. It seems unlikely. SNRM values, by themselves, don't alter the signal strength nor the amount of crosstalk inflicted on other lines. The only difference i can think of is when some extra tones get added into the medley because they are now viable. But that would require the modem to shift some bits off a high-bit-count tone onto a low bit-count one, and I don't know that it would happen. If it does, then the impact will be additional crosstalk on the new tones.

What @ejs says ...

In theory, I suppose it would be the opposite. Running at a lower target SNRM could allow the same line rate at a lower power output, and so with lower power output, there would be less crosstalk. Relatively speaking, the crosstalk reduction would benefit shorter lines more than it benefits longer lines.

... is a possibility, but I thought the aim at the DSLAM was to transmit every tone at the same power on every line, to spread crosstalk evenly (different tones can be different power, but still the same power on every line).

In theory, though, it would be possible to allow an "over-provisioned" line to use lower power; just low enough (once extra crosstalk inwards is allowed for) to drop the attainable speed to the package speed. That lower power would, in turn, reduce crosstalk outward to other lines. Reducing power is mentioned in DSM guides as an option, but I don't think I've seen it mentioned for quite such a dynamic case. The case when a new disturber gets added would cause a headache on deciding whether to increase power.

Upstream power backoff does allow lines to use different power upstream ... so different target SNRMs could have an effect there. However, this 5-4-3dB change appears to be only downstream.

[kitz]

Modems are supposed to take the Target SNRm into account when they calculate the attainable rate.

[j0hn]

We had a big discussion about this on the thinkbroadband forum
I was arguing that the maxattainable uses the actual target SNRM. MrSaffron and Robertos seemed to think modems use a fixed 6dB in all cases. WWWombat seemed to clear it up linking to VDSL2 specs showing 3 possible ways a modem can work out maxattainable. All 3 ways use a variable target, backing up my argument.

As the trial has gone on that point is now beyond debate. The maxattainable clearly rises as the SNRM drops each time.

edit: tag fail

[underzone]

Mr Saffron seems to have zero technical knowledge, I think he is just an employed/contracted journalist on that site.

[WWWombat]

WWWombat seemed to clear it up linking to VDSL2 specs showing 3 possible ways a modem can work out maxattainable. All 3 ways use a variable target, backing up my argument.

I'll just add...

As a software engineer, whose job has been to interpret & implement specs like this, I will say that shortcuts can be taken during development, and things that ought to be variable can get hard-coded. With the intention to come back and fix it, of course

Trouble-shooting is the art of checking what the system is actually doing, rather than what you think it ought to have done. Spoken as the proud creator of hundreds of bugs

[kitz]

I was arguing that the maxattainable uses the actual target SNRM. MrSaffron and Robertos seemed to think modems use a fixed 6dB in all cases.

Never saw that - its been a while since read anything on TBB otherwise I could have said something earlier to point you in the direction of ITU standards to back up your claim.

I already knew that it is uses the target SNRm from when I attempted to do a bit of digging a few years ago to see if there was any clue why max attainable went so skewy when Interleave (or rather INP) was applied to the line. 

I'm still none the wiser why INP causes the discrepancy.  I assume that most of the modem stats we see on here use the basic method (in fact I can't even recall there being a choice of methods at the time I looked).  I'm also assuming that because most of the stats we do see tend to come from BCM chipsets then it could be down to the way that BCM interpreted the original recommendation.

I note the latest version of ITU G.993.2 says 

The conditions of the basic method in this version of the Recommendation are unchanged compared to the conditions for ATTNDR calculation in previous versions of this Recommendation. This set of conditions did not specify a number of conditions to calculate ATTNDR, which lead to vendor discretionary behaviour in the reported ATTNDR values.

Looking at an archived version, which was the pre-published version of these changes relating to the max attainable rate was dated 05/2015 (amendment 6).  Most of us here are used to seeing stats from the HG612/Zyxels/Billions all of which were around before that date.

I will say that shortcuts can be taken during development, and things that ought to be variable can get hard-coded.

Im talking about the BCM based modems that most of us use,  but with this stat being part of the standard BCM commands then I would think its unlikely as we see enough movement of SNRm and target SNRm for it to be fairly accurate on lines which dont have INP.   These modems are also used on adsl lines and with other SPs which have had variable target SNR (both more than and less than 6dB) long before Openreach decided to apply it.  It only appears to go pete tong when INP is applied.

I dont know here and perhaps ejs may have some thoughts, but the other important figure used is SNRGAP. ITU states
The SNRGAP value is defined for a 10−7 bit error ratio on 4-QAM (no coding gain, INP_min0 = 0).

Is it something to do with the use of INP=0 in the calculation when say INP 3 is applied to the line which is throwing the figures out. Whatever way the figures that Ive noticed always seem to over compensate by about double what they should be. 

The modem has to calculate a [true] SNRGAP to get the actual sync speed which will be calculated using any existing DLM params.  Therefore is it possible that BCM is using the actual SNRGAP figure rather than one with INP=0 in the ATTNDR calculation that is throwing the figures out for interleaved lines.