High FEC count masking underlying noise?

Hello all,

Just been taking a peek at my stats and seen that I'm having a high number of FEC constantly (around 1000 FEC/min) spiking at various intervals to ~5000/min. Is this any cause for concern?

Another thing I have observed is a high retransmission count: rtx_tx/rtx_c. Granted there are no uncorrected packets, however there seem to be a very high number of retransmitted packets which leads me to believe there are various noise ingresses somewhere along the line?

A relevant piece of information would be that my master socket was recently relocated by BTOR on my behalf and it now runs past power sockets and a light switch (the knob/dimmer variety). Don't want to seem like I am splitting hairs but would appreciate if people could shed some light on whether I have a serious noise issue being masked by G.INP or not.

I managed to get hold of an old Raspberry Pi so my all stats are available now on MWDS under 'fishpan.'

Reason why I ask all this is because I have done a bit of reading about the new SNR trials and wish to find out any possible reasons why the 3db target may not be applied to my line. (I want to reach that magical 78-80mbps zone)

Thanks!

No cause for concern in my opinion. When interleaving+FEC is enabled on my line and my main crosstalker is offline I get under 1000 FEC per min. When the crosstalker is online I get around 50000 FEC per min. Many lines see similar numbers and is perfectly normal.

We must remember that FEC's are errors that never happened. Even a small amount of noise on the line can increase FEC numbers dramatically. You should only be worried if there is an increase in CRC's or ES. It is indeed an indication there is additional noise on the line. The relocated master socket could indeed be the cause but so could 1001 other sources of interference. Did it start right after the master socket was moved?

With regards to the lower target SNRM, this is now being rolled out nationally. However it's being done in phases and may take some time to reach all lines. If the retransmission rollout is to be used as an indicator then OpenReach were able to do circa 45k lines per day last time I saw a figure placed on it. Perhaps someone with figures on the rough number of lines connected to Huawei cabinets could estimate how long it would take to cover all lines at a rate of 45k per day.

Thanks for your very informative and succinct reply. You made a great point about knowing the FEC stats before the master socket being moved. From telnet data I had before this was carried out - the line was experiencing an average of around 500 FEC/min but now this has gone to an average of 2500 FEC after moving the master socket.

I guess all I can do at this stage is to try and limit this increase in noise by eliminating potential noise ingresses - will start by replacing the dimmer switch with normal on/off and monitor if it makes a difference over time. I wish we could know more about whether DLM will ignore FEC errors when determining whether a line is stable enough to have its SNR dropped to 3db in the upcoming weeks/months.

In reply to a post by fishpan:

I wish we could know more about whether DLM will ignore FEC errors when determining whether a line is stable enough to have its SNR dropped to 3db in the upcoming weeks/months.

The parameters you should focus upon are the ES and SES. They are the significant counters, DLM-wise.

Yep, just wondering if ES and SES are well within parameter limits, then if DLM looks to a secondary set of parameters that could indicate instability such as high FEC? Not sure if this has been ruled out yet.

I will refer you back to what j0hn83 wrote, above --

We must remember that FEC's are errors that never happened.

The recorded FECs are just an indicator that the error mitigation code (error detection and correction) is operating correctly.

The purpose of the DLM process is to keep the circuit stable and operational. Clearly, if the FEC count is significant to the DLM process then we would not be seeing circuits still operational with "sky-high" FEC counts.

What sort of unproven and theoretical underlying noise is giving you cause for concern? I you are able to provide some hints then it may be possible to point you to such parameters in the circuit's operating statistics which may either prove or disprove your concern.