The continuing saga of my strange speed fluctuations.

Some of you may have read my continuing issue with speed drop since moving from a pretty stable Plusnet connection on 25th April to BT. Here and Here. From the second I migrated, I lost 10Mbps in speed, going from a sync speed of 68Mbps to 58Mbps. Next day, my sync dropped again and has slowly dropped every day since until the connection dropped dead on 7am this morning and I reconnected at 49Mbps with interleaving now turned off. Speedtests are abismal at around 40Mbps and the TAP3 test fails but BT are adamant there is nothing wrong.

Last week I purchased a HG612 to replace my Billion 8800nl (just to rule out modem issues) and my line was reporting an attainable of about 55Mbps even though the actual sync never got above 45Mbps and throughput was never above 40Mbps.

This morning's drop was the last straw so I plugged the 8800nl back in. Hey presto attainable no back up to 60218Kbps and sync'd at 60214Kbps with an snr of 6.2.

Not sure what is going on but think I might have bought a dodgy HG612. It definitely had the latest firmware.

I'm now hoping G.INP might kick back in now (got taken on on my migration) and I could regain the extra 8Mbps I had before I moved from Plusnet.

Latest stats incase anyone is interested in commenting or can see something strange:

adsl info --stats
adsl: ADSL driver and PHY status
Status: Showtime
Last Retrain Reason: 0
Last initialization procedure status: 0
Max: Upstream rate = 17076 Kbps, Downstream rate = 60222 Kbps
Bearer: 0, Upstream rate = 17080 Kbps, Downstream rate = 60214 Kbps

Link Power State: L0
Mode: VDSL2 Annex B
VDSL2 Profile: Profile 17a
TPS-TC: PTM Mode(0x0)
Trellis: U:ON /D:ON
Line Status: No Defect
Training Status: Showtime
Down Up
SNR (dB): 6.1 6.1
Attn(dB): 19.2 0.0
Pwr(dBm): 12.6 7.5

VDSL2 framing
Bearer 0
MSGc: 18 21
B: 239 237
M: 1 1
T: 64 64
R: 0 16
S: 0.1268 0.4434
L: 15136 4583
D: 1 1
I: 240 127
N: 240 254

Counters
Bearer 0
OHF: 840724 241393
OHFErr: 8408 1
RS: 0 2562523
RSCorr: 0 1
RSUnCorr: 0 0

Bearer 0
HEC: 440 0
OCD: 0 0
LCD: 0 0
Total Cells: 198358854 0
Data Cells: 7094544 0
Drop Cells: 0
Bit Errors: 0 0

ES: 1259 1
SES: 0 0
UAS: 30 30
AS: 1710

Bearer 0
INP: 0.00 0.00
INPRein: 0.00 0.00
delay: 0 0
PER: 2.03 7.12
OR: 94.23 30.32
AgR: 60308.42 17110.39

Bitswap: 25/25 4/4

Total time = 29 min 3 sec
FEC: 0 1
CRC: 8408 1
ES: 1259 1
SES: 0 0
UAS: 30 30
LOS: 0 0
LOF: 0 0
LOM: 0 0
Latest 15 minutes time = 14 min 3 sec
FEC: 0 1
CRC: 4181 1
ES: 600 1
SES: 0 0
UAS: 0 0
LOS: 0 0
LOF: 0 0
LOM: 0 0
Previous 15 minutes time = 15 min 0 sec
FEC: 0 0
CRC: 4227 0
ES: 659 0
SES: 0 0
UAS: 30 30
LOS: 0 0
LOF: 0 0
LOM: 0 0
Latest 1 day time = 29 min 3 sec
FEC: 0 1
CRC: 8408 1
ES: 1259 1
SES: 0 0
UAS: 30 30
LOS: 0 0
LOF: 0 0
LOM: 0 0
Previous 1 day time = 0 sec
FEC: 0 0
CRC: 0 0
ES: 0 0
SES: 0 0
UAS: 0 0
LOS: 0 0
LOF: 0 0
LOM: 0 0
Since Link time = 28 min 32 sec
FEC: 0 1
CRC: 8408 1
ES: 1259 1
SES: 0 0
UAS: 0 0
LOS: 0 0
LOF: 0 0
LOM: 0 0
>

Edited by WelshWArrior (Sun 15-May-16 11:32:13)

Post deleted by WelshWArrior

Looking on mdws after 7 am (DLM re-synch) your sync was 53,969 on the hg612, with no error seconds

sync on the billion 60,214 with LOADS of error seconds.

Try the hg612 again.

I'll give it a go once I get home.

I see you have interleaving (1121) after all the error seconds the billion caused yesterday.

You still had a reduction in error seconds at 7PM to 10 PM last night and I'm sure we will see a reduction in FEC's at the same time tonight.

Put the hg612 back on until you are back on fastpath (or interleaving = 1)

then resync at this "quiet time" (between 7 and 10 PM)

Ian

Hi Ian

Yes by the time I got back last night , all I wanted to do was go to sleep (kids had worn me out haha) so didn't get chance to swap them back over. I'll do it later.

I'm still puzzled about why G.INP hasn't kicked in? I've never had interleaving turned on until I migrated from Plusnet to BT - I always had G.INP and it not only increased my sync speed to 65Mbps, but seemd to keep all the errors at bay.

The errors have been happening since 20th November and die each night at exactly 19:00hrs until 22:30hrs when they appear again.

Some funny things going on, certainly.

I looked at MDWS for the last 10 days, where it is quite obvious the DLM had currently left FEC+Interleaving activated, with INP=3 and delay=8 quite consistently. It doesn't matter that the depth has altered - that results from a variety of calculations and compromises. The INP & delay values stayed consistent from the 8800 into the HG612.

The fact your speed dropped with the HG612 is strange ... as it ought to have resulted in a high SNR value, but it didn't. Your SNR stayed at 6dB. We don't know why this happened, but the result is an "artifically low" speed for your line, compared to previous expectations.

The other impact that an "artificially low" speed should have is a reduction in errors ... and you certainly seem to have seen that, all through the initial placement of the HG612. MDWS reports that, through the 12th, 13th and 14th, you suddenly lose almost all CRCs (there weren't many to start with) and the FECs reduce to 1% of what they were. A huge, huge reduction in FECs!

Things changed markedly as of 7am on the 15th (Sunday). Presumably as a consequence of this huge reduction in FECs, DLM has taken away the need for FEC+interleaving: INP gets set to 0, delay gets set to zero, and the line resyncs.

Up until 7am, this was certainly all good news - the HG612 was doing better, and (perhaps by that low speed) helped your line to reduce the error rate enough to get DLM to recover.

From 7am until 10:30am, the HG612 kept going at a new speed of 54Mbps, and with almost no errors again - obviously no FECs now, but still very few CRCs and therefore few ES's - only 30 accumulated in those 3+ hours. Things were looking promising, error-wise, even if this was not top speed ...

Unfortunately, as you say in you OP, "This morning's drop was the last straw so I plugged the 8800nl back in." Instead of seeing the HG612 behaviour as positive, you threw it out ... and quite possibly threw the baby out with the bathwater.

Once the 8800NL was plugged back in, it gave you that good speed - 60Mbps - but it also gave you a terrible error rate. Hundred of CRCs from 11AM to 7PM, and more from 22:30 onwards. You ended up accumulating over 22,000 ESs by midnight, and another 15,000 between midnight and 7AM. This is a long way over the 24hr threshold of (max) 2,880 that would apply to a Plusnet line.

The result isn't a surprise ... DLM kicked in again, immediately. INP=3, delay=8. FEC+interleaving turned on again, speed reduced again, and hundreds & thousands of FECs.

It looks to me like your line is inherently noisy, in a way that causes the 8800NL to see a lot of FECs, an awful lot of FECs, resulting in DLM giving you FEC+interleaving-style intervention. The HG612 seems to see the noise too, but treats it significantly differently - giving you a lower speed that can overcome the noise better, resulting in much lower error counts.

I think that plugging the 8800NL back in was a bad move.

Who knows ... if you had left the HG612 in place for longer (possibly much longer), DLM may have analysed your (free/open/non-interleaved) line differently, and chosen to activate G.INP and we'd see a completely different outcome.

I agree that the behaviour of your line is puzzling - and the seeming coincidence of migrating to Plusnet adds to the puzzle. But the behaviour I'm seeing has nothing to do with the ISP, or a switch of ISP.

As your line looks like it (now) picks up a lot of noise, you either
- Have to persuade BT of this fact, and get it fixed or changed
- Have to choose to use an 8800NL, which gives you higher speed with interleaving
- Have to choose an HG612 which gives lower speed, potentially without interleaving
- Try other modems.

Thanks for that WWWombat - must have taken a while to write Very interesting though. I did plug the HG612 back in during my lunch break today but DSLstats seems to have crashed and I can't remote to it from work. Not sure if you even can remotely manage the HG612 like you can the 8800nl??

I think I'll leave it plugged in for a few weeks and see what happens. I'll try ian007jen 's suggestion above initially - wait for interleaving to switch back off and then resync during the quiet time to see it that gives me some speed back.

I'm planning on doing some more scouting around to try to find the source of this noise - possibly with an old AM radio (I think??) around my property first in the faint hope that it's something that i can definitely do something about.

Not that it matters but I went from Plusnet to BT.

Edited by WelshWArrior (Mon 16-May-16 15:51:04)

Grab some "Bitloading Graphs" from DSL Stats ensuring that under <Config>, <Items to monitor> you have <SNR per tone> and <Include with Bitloading> ticked.- reasonably expanded so that you get all tones shown across 5 images and host those somewhere. We can have a look and see if there are any strange notches and how it compares to "normal" plots.

If there is a noise source you may well see a large notch as the approximate frequency with the effect spreading each side. If it is very wideband then everything will be suppressed.

Thanks MHC - I'll get these done in the next hour and post a link to them.