Broadband DLM system in operation

News and views, and should we control them?

Moderator: embleton

Broadband DLM system in operation

Postby embleton » Wed Oct 31, 2018 3:15 am

There are five main figures in operation with DLM. These are the target synch speed (reported on router stats), the target noise margin, desired and actual noise margin (reported in the router stats), and the day period. The exchange (ADSL2+) or cabinet (VDSL2) equipment will initially use a figure known as the target noise margin and the day period.

The desired noise margin in a router will adjust the target noise margin by negotiating with DLM and the actual synch connection speed will change to the target. The actual target synch speed on training (learns the correct stable speed) is the ultimate and is stored in the exchange or cabinet equipment, not the target noise margin after a stable target synch speed is established in a period. The actual target synch speed is the governing factor in the short period and even a longer period if the actual threshold errors and line drops step outside a range DLM action is taken in the longer period on speed because of fault or raised actual noise margin on a good line (yes, that is the right way).

On desynchronisation the synch will not be changed immediately, it will be delayed until drops or errors exceed a threshold and it is important to know the noise margin will drop (yes, that is right before desynch). Or the line conditions improve with an increase in the actual noise margin. The actual noise margin can go up or down because of the fixed target synch speed and especially line condition but will return to a target noise margin which is not necessary 6dB. With most good DLM systems, the desired noise margin is adjusted in a wide sweeping scale up and down depending on good line conditions or bad, on the router itself by reporting back and DLM adjusting the target speed. Then the actual noise margin will report accordingly around target noise margins or below and synch speed is fixed and is in-band speeds in DLM

The actual noise margin is raised (because synch speed is fixed) because of good conditions or bad and decides which way the synch speed goes, either up or down when DLM takes action. But DLM takes action over a period and is in day periods on VDSL2. DLM will monitor the line continually on VDSL2. It is important to know there is no fixed 6dB target noise margin once a line is stable, it's the synch speed that is fixed in the short period and over a longer day period target noise margin *may* either go up or down. This is proved by the fact that the fixed synch is only changed at the period and you see a raised actual noise margin on a fault that because it needs more range or lower target on good lines.

On ADSL2+ the only way is down in speed until rebooting the router then the speed is renegotiated during training to a target noise margin that is not 6dB but in a scale stored in dynamic memory from target noise margin bands & synch speed, and that is clear from the fact it shows actual noise margins ranging from 1dB to 18dB in a sweep to synch speed fixed target.

When talking about target noise margins they are between a range. When talking about that range its actual 6dB or lower reported on the router actual noise margin when good line conditions exist not the target noise margin which doesn't really exist its synch speed. If you think about it what happens when synch is lost and you look at the Support recent activity log in the Sky router, it will not drop the synch speed until those thresholds exceed set ranges and it always bands at lower and higher speed figure.

Target noise margin: a threshold usually 6dB or lower.
Noise margin: actual noise margin on the line reported on the modem router.
Target synch speed: usually fixed unless errors or drops occur then it drops. And up in speed when it's a stable line automatically on VDSL2 and manually on ADSL2+ on reboot.
Desired noise margin: Reported back to the DLM system in negotiation when training with a modem.
Day period: When DLM takes action on target synch speed usually a day or longer.

Just think about it if target noise margins and target synch speed are not in-bands then it would train forever! And if the target noise margin was used the target synch speed would change absolutely everytime a reboot occurred forever call racing synch speed, and that is not what the router logs show and prove.

Open for discussion.

1. If it is a sticky actual noise margin at 6dB then diagnostics most be used you can call a fault on target synch speed when outside range.
2. If you have a raised noise margin that is because of the line condition whether good or bad.
3. Call a fault if the synch speed is below par but before doing so reboot the router to see if synch goes up or down.
4. If actual noise margin goes down and the synch speed goes up then don't call a fault.
5. If the synch speed remains the same don't call a fault to wait the period day or days is then necessary to see noise margin and whether it's 6dB or lower; 6dB or lower isn't a fault.
6. A reduced synch speed immediately then call a fault on target synch speed after those diagnostics tests on target synch speed or reported in the support logs.
7. If the line races and continually drops and ups synch target speed (seen in the Support recent activity log) call a fault if it occurs more than 3 times in the day and 7 resynchs at night by DLM. But the maximum before Sky considers a fault is 10 drops or x number of errors in a period in milliseconds on the line without resynch.
8. And call a fault if the attenuation changes by a small 0.1dB on any of the 6 attenuation figures on VDSL2.

I'll talk about those 6 attenuation figures and what they mean later for diagnosing a fault on this thread concerning the line to the cabinet on VDSL2.
My email address is mark @ domain of the forum. terms of use of site in announcements.
User avatar
embleton
Site Admin
 
Posts: 361
Joined: Sat Aug 02, 2014 1:40 pm
Location: Plymouth

Re: DLM system in operation

Postby embleton » Wed Oct 31, 2018 2:13 pm

The 6 attenuations and noise margin information are useful for diagnosing faults on VDSL2.

1. The 6 attenuation figures represent attenuation at the DLM training start.
2. Lossing a downstream or upstream noise margin for target speed on that band *may* be a fault or normal for the distance to the cabinet.
3. The 3 attenuation downstream and 3 upstream figures represent and go up the spectrum in frequency in the bands.
3. When you look at these attenuation figures they *rise* along the bands from DS1-DS3 in that order, following the expected frequency response of the line to the cabinet.
4. And the upstream attenuation figures rise the same from US0-US2" in that order.
5. The attenuation for each band has a useful target synch speed for carrying capacity depending on attenuation and noise.
6. About 20dB or lower results in the maximum capacity on the band.
7. Above 20dB lowers the capacity on a band and adjusts useful target synch speed.
8. A fault can be called when a band that is lower than the top band attenuation is higher than it.
9. 8 above is not normal and means a wavelength fault where the signal is reduced in capacity depending on a distance to the fault.
10. A loss or 0.0 noise margin on a band means it's not being used to carry anything in capacity target speed.
My email address is mark @ domain of the forum. terms of use of site in announcements.
User avatar
embleton
Site Admin
 
Posts: 361
Joined: Sat Aug 02, 2014 1:40 pm
Location: Plymouth

Re: DLM system in operation

Postby embleton » Wed Oct 31, 2018 4:04 pm

A DLM reset would result in the maximum target synchronisation speed with G.INP contrary to belief. That is why people ask for it even though DLM is operating all the time on VDSL2 and BTO don't do it often. A G.INP drop for interleaving or both is a *possible* fault which can be tested by testing latency to the Huawei cabinet. At first, a DLM reset on VDSL2 will result in it quickly actually setting on the target synch speed in less than 2 hours or quicker and it will then do it with G.INP enabled on Huawei cabinets. If G.INP is not enabled on DLM reset and DLM resynch then maximum target synch speed would never be reached you cannot go back in time for setting G.INP it would be pointless you have to reach DLM, and why to do it twice in the wrong order!
My email address is mark @ domain of the forum. terms of use of site in announcements.
User avatar
embleton
Site Admin
 
Posts: 361
Joined: Sat Aug 02, 2014 1:40 pm
Location: Plymouth

Re: Broadband DLM system in operation

Postby embleton » Mon Nov 05, 2018 1:32 am

This is related to VDSL2.

Line management initialisation G.INP on retransmission modem support and can G.INP retransmission based on latency <= minimum delay latency and low mean time between errors (MTBE) two bearers, bearer 0 carry data & bearer 1 1Kbps sizes (10 DMT symbols maximum) and buffers well suited to low errors. It is possibly *interleaving*, turnaround delay must be low and runs a buffer with data bearer which offers retransmission.

And it is the lowest latency, and highest throughput achieved with the retransmission only occurring when data is corrupted at low levels of errors at the physical layer on bearer 0 with bearer 1 the G.INP bearer doing retransmission when *necessary* in 1Kbps sizes decided at the line initialisation phase. This is why you see 79999 and 19999 on downstream and upstream on VDSL2, it's the 1Kbps minimum for G.INP. There are no counters on G.INP for DLM reset to decide it to my knowledge, it's delay and MTBE for G.INP line initialisation at startup that decides G.INP on or off.

The main reason G.INP is disabled on VDSL2 because of a DLM reset is that the frequency spectrum shaping hasn't been set up by DLM accordingly because of the DLM reset (bad idea) on bearer 0, but if the line spectrum analysis is good and meets the first paragraph G.INP will be enabled immediately on initialisation of line management. Usually, spectrum analysis is poor on DLM reset, but this isn't the case all the time and G.INP will then be enabled immediately if 10 DMT symbols don't get corrupted during first-time spectrum analysis, latency delay and MTBE when startup training meets paragraph 1 G.INP enabled.
.
My email address is mark @ domain of the forum. terms of use of site in announcements.
User avatar
embleton
Site Admin
 
Posts: 361
Joined: Sat Aug 02, 2014 1:40 pm
Location: Plymouth


Return to The Internet, Providers and Social Networks

Who is online

Users browsing this forum: No registered users and 1 guest

cron