FEC me!

Had my line activated (80/20 package) 2 weeks ago now and have been getting a huge number of FECs. At first the line synced at 45mb and 6db SNRm, wasn't getting many errors at all. DLM gradually upped the speed to 62mb and lowered the SNRm gradually to now 3.4db and I have a huge number of FEC errors, around 20k per minute on the downstream, but very few on upstream.

Stats for the last 4 days

Error seconds is reasonably low, and I have not had any disconnections.

However, I believe I am only 350-400m from the cabinet, the attenuation at 19db would suggest I am more like 1,.4km from the cabinet (which I cannot be, I am only 2km from the exchange itself and there are multiple cabinets much closer than 1.4km), and in every speed test I have done since it first synced, I have got a real speed of 9mb less than it has synced at (consistently 9mb less). When it first synced at 45mb, I got speed tests around 36. Now it has worked its way to 62mb and my speed tests are around 53!

I have tested with a cat5 as well as 5ghz wifi and both are identical.

I have tried a Zyxel 1312-B10a, as well as a HG612, using an AC86U as the router (and have just tried the zyxel as modem and router). Both have all the same issues except the zyxel syncs about 3mb higher.

The master socket is an NTE5 with mk3 faceplate that I installed myself. The old master socket was not even an nte5, still had the bell wire attached, and had 4 extension sockets wired (in sequence, not all from the master) which I have disconnected so now I have the NTE5, no extensions and no bellwire, just A and B.

I have been on to IDNET support and they have got hold of the logs from their end and say they are not seeing this huge number of FEC errors on their side! This seems very strange (along with all the other strange things here)!!

When I do a quiet line test using a wired (not DECT) phone to the test socket, I can hear a feint hissing sound constantly, but IDNET have tested the line and it passed the test. When I was on the phone to IDNET support I could actually hear a much louder buzzing type sound on the phone which he could not hear from the other end.

We are at a bit of a loss as to what the hell is going on. The attenuation doesn't seem to make sense, my sync seems low (at 350-400m I should be pushing the full 80mb), the FEC errors are huge, only affect the downstream and do not seem to be visible on IDNETs logs, CRC's and error seconds seem fine, I get consistently 9mb less than whatever I am synced at, and we have been unable to figure out what is going on with it!

So experts here have any ideas? IDNETs tests all seem to find no real issues so they have nothing to get BT to come out without risking them sending an engineer, testing it for 2 seconds, saying its find and charging £140 for it. So unless I can figure out what is going on, I am either stuck with it, or have to play the BT engineer lottery.

1.4km on VDSL2 does not equate to a 19dB attenuation, it may have on ADSL2+ but much higher frequency so more attenuation per metre of wire, so your line is shorter than you think

FEC are corrected errors i.e. interleaving is doing its job

63 Mbps sync is around a best case download of 59 Mbps

What do the speed tests actually look like, if some spiking the value reported may not be as high as you are spiking too.

Ok the attenuation may not be relevant then. I think the line is 350-400m because I searched for the cabinet location via kitz and telecom-tariffs.co.uk and drew the route to my location from it on google maps (via roads not in a straight line) and google indicates 330m. Obviously there may be additional, from ground level up to my 1st floor flat, but add another 70m and its still only 400 which I believe is still in potential 80mb range.

I know FECs are corrected errors but that's one huge number of errors to begin with. I assume its not normal (particularly on a relatively short line) getting a million FEC corrections an hour? Even the IDNET guy said that was a lot (although they can't see anything like that many on the logs on their end which seems really strange!).

Speed tests indicate a very smooth 53-54mb from my 62mb sync, no spiking (except the upstream which yoyo's around between about 12 and 16 but I guess thats normal on fttc). I thought I might lose 1-2mb maybe but it seems its normal to lose more than that, friend on full 80mb only manages 74-75! If this is typical then it stinks of mis-selling to me if the speed you get is at best a full 5+mb below whatever you sync at! They should not be selling it as an "80mb" product if it cannot actually do 80mb in real world use.

I wonder if its worth whipping the wires out of the master and re-connecting them. Could the high FEC's be related to that or would it be a waste of time trying?

Edited by Ewok (Thu 03-May-18 22:38:52)

19dB attenuation suggests around 550 to 600m of cable length. It could be less if smaller gauge was used - certainly not 1.4km

Edited by MHC (Thu 03-May-18 23:00:00)

Yeah I am guessing its something around 400m including twists and turns but I am not sure what speed that would typically give. Just googling it seems to indicate my speed might be right in line with what a 350-450m line would give based on the couple of graphs I found but could go higher based on another couple of graphs.

Perhaps what I am getting, including the millions of FEC's and 7-9mb reduction in actual speed vs sync is all perfectly normal?

current 3.5hour uptime is showing 3.2million FECs on the downstream, but zero error seconds, zero CRC's, zero HECs. And on the upstream (which has no interleaving) 95 FECs, 10 CRC's, and 9 error seconds. With syncs of 62mb and 17mb and SRNm of 3.4 and 6.1.

Obviously the interleaving on the downstream is doing its job and correcting all them FECs, but it feels like a huge number of constant errors, but I have no basis for comparison. I just need to be satisfied I have no major issues and am getting the best I can expect to get given whatever my line length is.

I am also just wondering if the process of FEC corrections results in retransmission of is this something done instantly on the fly with little to no latency or bandwidth overhead?

Edited by Ewok (Thu 03-May-18 22:57:42)

https://www.thinkbroadband.com/guides/fibre-fttc-ftt...

No way is it that short - just tweaked my post to correct the numbers (finger trouble in the dark). My line is 440m - and I know te route almost exactly and attenuation is 16dB. Yours will be up at 550 to 600 m

You will be on the retransmission high DLM profile, as most on a 3dB SNRM target are on retransmission high.
This gives an IP profile around 91-92% of the sync speed.
So speed tests should be around 54Mb maximum.

I regularly get 10k -20k FEC per min. It's perfectly normal.
Powerline adaptors can increase FEC's dramatically, so if using them that might be the reason.

Ok thanks, perhaps everything is normal then. Just seemed odd that a combination of factors all seemed to point to strange things going on.

I don't have powerline adaptors, have avoided them as I know they can cause issues

What I am thinking is that if I lower the package to the 55mb one (since I am paying for 80 and effectively getting less than 55 anyway), and it then syncs at 55mb instead of 62.....I am probably going to get less than 50 in actual speed. So I would be paying for 55mb, on a line that can handle 55mb (real use) and only getting probably 48-49 at best?

And if going to that detail it may be with the sync capped that there are less errors to correct and DLM may behave differently so you might end up closer to the usual 96% rule

No one can predict which.