Kitz Forum
Broadband Related => ADSL Issues => Topic started by: jon_ on March 14, 2017, 05:30:53 PM
-
Hi
Bit of background - I live in the sticks, approx 5km from the exchange. I have 2 lines from BT, both with ADSL enabled. Both come in on the same piece of wire on different pairs from the junction point which is at the bottom of my drive.
Up until fairly recently, both were rock solid on a 6dB noise margin (set exchange side). Line one would sync at about 3megs. Line 2 just under 4M. Connection times could be weeks at a time between disconnects. They'd been like this for 3 or 4 months.
Then something changed. I get massive amounts of HEC errors on both lines. The ES count goes up fairly quickly due to the HEC errors. DLM kicks, and both lines get throttled back. Line 1 is currently 2.5M on a 9dB target, Line 2 2.5M on a 15dB target.
Attenuation is around the same on both lines as it's always been. No noise is evident on a quiet line test. Both modems are plugged into the master sockets.
Any suggestions on how to approach this - Uno (my ISP) can request a DLM reset, which results in stable connections that are much quicker, until DLM throttles me back down again. As there's no audible noise present I can't log a voice fault, and obviously don't want to pay for an OR engineer visit!! Unfortunately I can't swap to modems that support a target SNR override as they are Sharedband enabled (bonded DSL connections) :(
Thanks!
-
Hi jon_
That's fascinating. Background: I have a similar setup, much longer lines 7.3 km times three which are IP-bonded in both directions by my ISP and router working together. A single TCP download goes at triple speed. A single upload is less successful, at something like 70% of triple speed. I get 2600-2800 kbps d/s sync with a d/s attn of ~66dB and a supposed downstream target SNRM of 3dB (actual is 0.6-18dB), upstream sync varies between lines - 400k to 550k at an actual SNRM of 6dB.
Having bored you with that. I now admit that I can't offer too much insight. Questions: are the errors sporadic? - what's time distribution of them? How much can you gather in the way of stats?
I'm wondering about crosstalk - is it possible that you have a new disturber somewhere near you? A neighbour with a new service or one who has changed their kit. Or else a source of non-DSL noise has started up? BT can do an RF noise identification service if they feel like it, but they can't fix it for you as they have no powers to police the noise environment.
Desperation option: If you get sufficiently fed up, you could always talk to Andrews & Arnold and see if you can take advantage of their fix-your-line-or-your-money-back scheme. But realistically if this is an RF noise or crosstalk problem then it may not be a fix-line issue.
I can use any modems I like. I'm ignorant - didn't know anything about Sharedband, to my shame - and didn't understand your point about restricted modem choice. My router knows nothing about modem models, they're just some black boxes that speak PPPoEoE.
-
Thanks Weaver. I remember your setup from some other threads.
The Sharedband setup works in a similar way - my two lines have their own connections, the modems are linked and present a virtual IP to my router behind. I have it set up as a routed connection, you can also present as a NAT address and have the modems act as routers if wanted. The additional firmware on the modems takes care of the aggregation of the links, failover etc, and a server at the ISP takes care of the other end. Single thread downloads go at approx the sum of the two lines, minus some overhead as do uploads. I could add more lines up to a max of approx 20Meg total, but that's dependant on the hardware (TP Link TD-W8968) and it's processing capability, heavier duty routers can support bonding FTTC and faster connections. I don't think Uno support linked PPP sessions, they offer sharedband as their solution for aggregating links.
In terms of stats - the modems are locked down to stop 'fiddling' - I've got a ticket open to gain telnet (& SNMP) access, but it doesn't seem to be working as expected. At the moment all I can get is the stats from the admin page, which don't give a breakdown over time. From my checking, it doesn't seem to be time of day dependant, it's just a percentage of headers have errors, which increments the ES count and annoys DLM.
As far as I am aware nothing round here has changed - but the cable run is long so it could be something further away. Everyone else I know locally seems to also suffer from rubbish connections (if anything worse than mine) and aren't technical so can't offer any insights!
BT/OR are *supposed* to be rolling out FTTH in the next 3 months, so I don't really want to be changing too much as I don't want to get tied into contracts! Worst case I can request to be switched to TT LLU with DLM disabled, which may not fix the HEC errors but will let me stay with a quicker connection, or if it drops much more I'll ditch the aggregated connections and go back to my old Billion router and force it to connect at 6dB!
-
TT LLU might indeed be a better option with their less obnoxious (or is it non-existent) DLM.
-
Talktalk have a new DLM, which seems to have done more bad than good. I've no personal experience using it, but I believe it can't be "turned off" like it could before. Early impressions from users on kitz are not good with complaints of previously stable lines having interleaving applied and sync reduced.
-
Burakkucat will be the man to ask. Wonder if this is the same for TT Retail, TT Business and TT’s Wholesale arrangements with ISPs? Three cases possible to consider. AA user TT Wholesale a lot so I wonder if this affects them.
-
Slight development... I've managed to get telnet access, and the modems support SNR tweaks..
Once DLM has put you on a 15dB SNR target with interleaving, is there much else it can do? If I force it to connect at a lower SNR is it likely to start doing really nasty stuff like rate limiting the connection?
-
Is it 20CN ADSL1 or 21CN ADSL2+?
-
21CN ADSL 2+...
-
Burakkucat will be the man to ask.
Hum? I haven't seen any sign of a new DLM process. :-\
Still bumbling on, much as always, in The Cattery --
$ xdslctl info --show
xdslctl: ADSL driver and PHY status
Status: Showtime
Last Retrain Reason: 8000
Last initialization procedure status: 0
Max: Upstream rate = 960 Kbps, Downstream rate = 5548 Kbps
Bearer: 0, Upstream rate = 960 Kbps, Downstream rate = 5120 Kbps
Link Power State: L0
Mode: ADSL2 Annex A
TPS-TC: ATM Mode(0x0)
Trellis: U:ON /D:ON
Line Status: No Defect
Training Status: Showtime
Down Up
SNR (dB): 5.6 6.7
Attn(dB): 45.0 28.4
Pwr(dBm): 0.0 12.7
ADSL2 framing
Bearer 0
MSGc: 59 12
B: 160 120
M: 1 1
T: 1 1
R: 16 10
S: 1.0000 4.0000
L: 1416 262
D: 32 8
Counters
Bearer 0
SF: 1296987 238951
SFErr: 129 0
RS: 84304060 3898477
RSCorr: 56793 82
RSUnCorr: 2334 0
Bearer 0
HEC: 2192 0
OCD: 2 0
LCD: 2 0
Total Cells: 254503015 47629536
Data Cells: 7402239 553626
Drop Cells: 0
Bit Errors: 138447 0
ES: 43 0
SES: 2 0
UAS: 46 46
AS: 21076
Bearer 0
INP: 1.00 1.00
INPRein: 0.00 0.00
delay: 8 8
PER: 16.25 18.00
OR: 32.00 8.00
AgR: 5131.95 964.23
Bitswap: 8100/8101 0/0
$
-
Early impressions from users on kitz are not good with complaints of previously stable lines having interleaving applied and sync reduced.
j0hn -- you might like to edit your post, above. I've highlighted in red the word in question . . . ;)
-
What modems are they ?.
Can you not get the stats via a web browser page ?. http://www.kitz.co.uk/routers/index.htm
I am doing some tests with my modem locked to ADSL instead on what it should be for my long line of 4.3kl of ADSL2*, I'm finding that it's making the line more stable with less errors HEC=4.86 /Hr, if you look at the Upstream on the graph it is like I have drawn it on with a ruler.
*= the 21CN exchange settings are still set to ADSL2+ for my line and have not been changed by my ISP.
-
What modems are they ?.
TP-Link W8968 Mk3's. I can get web based stats, but that just gives a snapshot, there's no easy way of seeing how the values change over time. I *may* be able to get CLI based stats now I have telnet access, but I need to play around with my setup as it won't connect from my internal network, I have to patch through to the dirty side and set a static IP..
Currently Line #1 has been up for 2 1/2 hours since I reset the stats, 4704 HEC errors, 1761 ES.. So that's an HEC error approx every 2s, 1761 ES out of 9000 seconds passed...
-
TP-Link W8968 Mk3's.
Can you try installing DSLstats http://dslstats.me.uk/index.html
Eric (roseway) could tell you if it will work with your modem.
-
I'm afraid I don't know anything about the W8968 Mk3. Whether DSLstats works with it depends on the usual conditions as given on the index page in your link.
-
On BTWholesale ADSL2+, the ISP should be able to disable the DLM or set custom thresholds which can be used to disable the DLM.
-
Guys don't we want HEC_errors / total_cells, not per unit time?
Or does it not matter because cells are sent when the link is idle, is that correct? If that notion is true, are HEC errors even counted in idle cells?
-
I think it does not matter, idle cells are transmitted when there's nothing else to send. Idle ATM cells would be identified by their header, so if the cell header fails the check, you don't really know if it was an idle cell.
In general the DLM doesn't seem to care about the actual number of errors, only the number of errored seconds.
-
From the faster line, which is now SNR tweaked back to around 8dB...
Internet Up Time: 1Day(s) 20:25:32
Mode: ADSL_G.dmt.bis
Traffic Type: ATM
Status: Up
Link Power State: L0
Downstream Upstream
Line Coding(Trellis): On On
SNR Margin (0.1 dB): 84 63
Attenuation (0.1 dB): 545 332
Output Power (0.1 dBm): 0 126
Attainable Rate (Kbps): 4,324 928
Path 0 Path 1
Downstream Upstream Downstream Upstream
Rate (Kbps): 3,449 922 0 0
MSGc (# of bytes in overhead channel message): 59 11 0 0
B (# of bytes in Mux Data Frame): 107 122 0 0
M (# of Mux Data Frames in FEC Data Frame): 2 2 0 0
T (Mux Data Frames over sync bytes): 1 1 0 0
R (# of check bytes in FEC Data Frame): 16 8 0 0
S (ratio of FEC over PMD Data Frame length): 1.9850 8.4666 0.0 0.0
L (# of bits in PMD Data Frame): 935 240 0 0
D (interleaver depth): 16 2 0 0
Delay (msec): 8 4 0.0 0.0
INP (DMT symbol): 1.00 0.00 0.0 0.0
Super Frames: 9,878,160 8,766,343 0 0
Super Frame Errors: 49,664 527 0 0
RS Words: 321,040,221 2,261,535 0 0
RS Correctable Errors: 1,665,153 1,389 0 0
RS Uncorrectable Errors: 218,213 0 0 0
HEC Errors: 89,588 518 0 0
OCD Errors: 11 0 0 0
LCD Errors: 11 0 0 0
Total Cells: 1,296,183,418 346,044,796 0 0
Data Cells: 238,758,184 33,533,844 0 0
Bit Errors: 0 41,334 0 0
Total ES: 37,778 355
Total SES: 2 0
Total UAS: 0 0
On the other line, there are a lot less errors (although I've not reset them since the connection last reset so the stats are meaningless), however this jumps out:
D (interleaver depth): 96 2
Not seen interleaving get that high before, theoretical max speed is reporting as 3ish megs, connecting at 2ish, and it's moved itself back to a 6dB SNR margin from 9dB previously...
-
@ejs - so you and I seem to be in agreement, I did get it right about idle cells. I remember about ATM repeatedly looking at cell headers in order to discover frame boundaries, instead of having the usual start-of-frame + end-of-frame symbols. Clever, in a perverse way.
-
As a slight update to this, both of my connections are now connecting with very high interleaving - 128 for the faster connection, 96 for the slower one, and both reverted to 6dB SNR targets. They both also appear to have been put on banded profiles - 4M attainable rate / 3M actual for the faster line, 3M attainable / 2.2M actual for the slower. Both have INP = 4.
From reading around, does INP=4 mean that the banding won't remove itself if the connection remains stable, and will require a manual reset ISP side?
-
Mine is .
INP: 8.42 D 0.37 U it went up to 15 the other week.
It varies with interleaving as I think they are linked.
-
I'm trying to understand the non-integral interleave factors. Need to look this up again, 8.416667 = 101/12