ADSL2+ slight disappointment

Switched from ADSLMax to ADSL2+ today on TalkTalk LLU via XILO.

On ADSLMax i had an 8128 sync @ 6dB noise margin with a downstream line attenuation of 33dB (i am 2.4km from the exchange).
Prior to moving over i was hoping for 12-13mbps sync on ADSL2+ looking at the graphs and inputting my ADSLMax line details on a couple of websites.

Coming home tonight my sync was barely over 8128 @ 12dB noise margin (XILO default) but since that over 2 hours my noise margin had dropped ~3dB (monitored with Routerstats). Is that normal? On max it varied by 1dB maximum over a 24 hour period.

I've rebooted my router (Billion 7800N) and used its onboard SNR tweaking to reduce the target noise margin to 6dB. My new sync speed is 9770kbps, about 2000-3000kbps short of what i was reasonably expecting (websites and graph said expect anything up to 14000kbps).

Current stats:-
DMT Status No Defect
Operational Mode ADSL2+
Upstream 945
Downstream 9770
SNR Margin(Upstream) 12.4
SNR Margin(Downstream) 6.3
Line Attenuation(Upstream) 20.2
Line Attenuation(Downstream) 37.0

I'm using the latest 7800N firmware (1.06e) and have ADSL2+ and Annex L set only (did have G.DMT etc also selected but unticking them has made no difference).

I dont see how i can get anymore......am i either unlucky or are the ADSL graphs pretty inaccurate?
Or will a resync in the daytime give me another 2mbps...wishful thinking no doubt!!

(note:- I'm using a BT ADSL faceplate filter and ADSLNation Pro+ cable to the router, a few phone extensions are run from the front of the faceplate).

PS
Just went to post this and it dropped sync for a few minutes.....maybe i should of stuck with Max!!

Noise margin variation can be higher as you are pushing the line closer to its limits, and thus more variation is possible.

Your speeds are in the zone for ADSL2+ and the attenuation, 9400Kbps to 12800Kbps from www.farina1.com/adsl

Daytime resyncs may well give you more. I assume you have double checked the stats at the test socket, since with ADSL the ring wire was less of an issue, and with ADSL2+ it is more sensitive to wiring issues. VDSL2 is even more sensitive, hence the engineer installed faceplates on that product.

And just found out my landline and its not receiving incoming calls......(line is being taken over by TalkTalk/Xilo).

Edited by deleted (Tue 17-Jan-12 18:52:13)

Inputting my ADSLMax details ie 8128 sync, 6dB, and 33dB line attenuation i get from that link:-
Your ADSL2+ speeds are likely to be in the range 11500Kbps to 15000Kbps

I've tried my Netgear DG834GT in the filtered faceplate with the normal phone lines detached and its the same. I dont have any ringwire connected.

Tweaking the modem might confuse things.
What is the target margin if the modem has no tweaks applied? I'd usually prefer to let the ISP tweak things down.

So yes would expect better, but it is not at disaster levels. It is worth unplugging the filtered faceplate and checking at the test socket, just in case the extra connectors are making a difference, long shot though.

In reply to a post by InQ:

websites and graph said expect anything up to 14000kbps

Yes, for 33 dB attn., but attn. increases by about 3 with ADSL2+. So, I'd expect your line to hold about 11 Meg.

But the graphs and checkers say you cant check it when you actually have ADSL2+ as they take your ADSL line stats and tell you what to expect on ADSL2+.

Will be 12db with interleaving on a TTB service.

Matt

In reply to a post by uno:

Will be 12db with interleaving on a TTB service.

Matt

Yep Matt....i am impatient and running the SNR Tweak on 7800N.

Router Telnet stats:-

> adsl info --stats
adsl: ADSL driver and PHY status
Status: Showtime
Retrain Reason: 8000
Max: Upstream rate = 944 Kbps, Downstream rate = 11656 Kbps
Path: 0, Upstream rate = 944 Kbps, Downstream rate = 9789 Kbps

Link Power State: L0
Mode: ADSL2+
TPS-TC: ATM Mode
Trellis: U:ON /D:ON
Line Status: No Defect
Training Status: Showtime
Down Up
SNR (dB): 6.5 12.3
Attn(dB): 37.0 20.2
Pwr(dBm): 0.0 13.0


Is the Max speed quoted at the top an achievable target? Whats this based on?

In reply to a post by XRaySpeX:

In reply to a post by InQ:

websites and graph said expect anything up to 14000kbps

Yes, for 33 dB attn., but attn. increases by about 3 with ADSL2+. So, I'd expect your line to hold about 11 Meg.

Mine line is 34db and during the summer it can hold 15Mb/s fairly reliably. Over the winter I've had to raise the target margin and it's been holding 12.8Mb/s for nearly two months now.

But yeah it depends on the particular line being considered.

So you add 3 to the attn. before you give it to the checker, knowing full well the usual effect of ADSL2+.

OK, well it wasnt my understanding of how to gauge the ADSL2+ speed. I thought you took your ADSLmax line attenuation, pump it in and it tells you what to expect from ADSL2+.

Still not to bad at the moment:- http://www.speedtest.net/result/1713459300.png

Looks like Interleaving is on, have to check what the telnet command is to find that out o nthe 7800N.

Edited by deleted (Tue 17-Jan-12 20:30:55)

Max is what the modem itself thinks is possible and something like DSL or target noise margins, interleaving are holding you back

The figures are so fuzzy, that either approach puts you in the ball park. Which you are

Yes, some of those checker sites imply that attenuation is invariant under ADSL Mode.

I prefer: http://www.kitz.co.uk/adsl/max_speed_calc.php

Thanks Andrew.....forgot about Interleaving and i think it is on currently but not sure how to check it other than telnet into my router.
Any ideas if the Interleaving depth is the "D" row i have highlighted below?
(note i've resynced after *ahem* replacing the cable between the RF3 filter and master socket as per my post the other week, existing one had split open in a couple of places after years of being stood on under the carpet).

> adsl info --stats
adsl: ADSL driver and PHY status
Status: Showtime
Retrain Reason: 8000
Max: Upstream rate = 948 Kbps, Downstream rate = 11408 Kbps
Path: 0, Upstream rate = 945 Kbps, Downstream rate = 10000 Kbps

Link Power State: L0
Mode: ADSL2+
TPS-TC: ATM Mode
Trellis: U:ON /D:ON
Line Status: No Defect
Training Status: Showtime
Down Up
SNR (dB): 5.5 12.4
Attn(dB): 37.0 20.2
Pwr(dBm): 0.0 13.0
ADSL2 framing
MSGc: 59 10
B: 61 21
M: 4 8
T: 5 6
R: 6 8
S: 0.7909 5.9116
L: 2569 249
D: 64 4
Counters
SF: 193525 175325
SFErr: 21 0
RS: 15723934 2101873
RSCorr: 2371 0
RSUnCorr: 431 0

HEC: 21 0
OCD: 0 0
LCD: 0 0
Total Cells: 73318727 6900793
Data Cells: 2008232 426540
Drop Cells: 0
Bit Errors: 0 0

ES: 125 7
SES: 69 0
UAS: 1421 1421
AS: 3108

INP: 0.59 0.51
PER: 16.06 17.73
delay: 12.65 5.91
OR: 32.36 7.21

Bitswap: 508 0

Deleted the rest but there were "FEC" counts so think this is "Forward Error Correction" and hence Interleaving is on?

With what NMs? All my quotes assume 6 dB in absence of any further info.

RF3 filter which is reducing noise may also have a small affect on the ADSL2+, or just indicate your line is never going to excel speed wise

Interlease depth of 64 makes sense

In reply to a post by MrSaffron:

RF3 filter which is reducing noise may also have a small affect on the ADSL2+, or just indicate your line is never going to excel speed wise

Interlease depth of 64 makes sense

OK well at some point i want to get BT to move the master socket from the middle of the hall and put it on the external wall which the drop wire comes to so might not bother with an RF3 and take it direct into the master socket.

Will get the wife to resync the router tomorrow during the day.

Thanks all.

In reply to a post by XRaySpeX:

With what NMs? All my quotes assume 6 dB in absence of any further info.

15Mb/s on 3db, 12.8 on 6db. The margin reported on my router is currently varying from nearly 9db mid morning to just over 6db in the evenings. I suppose that fits in with the fact that at the moment with a target of 3db it'll drop two or three times a day.

In reply to a post by InQ:

Thanks Andrew.....forgot about Interleaving and i think it is on currently but not sure how to check it other than telnet into my router.
Any ideas if the Interleaving depth is the "D" row i have highlighted below?
...
Deleted the rest but there were "FEC" counts so think this is "Forward Error Correction" and hence Interleaving is on?

Yes, the D row is the interleaving depth. It is the number of packets over which real data packet is spread.

1 = Fast Path, as that means the data is all in one packet.

FEC's are another way of detecting interleaving, yes. Though some routers, IIRC the BT Home Hub being one, can mis-report the upstream, making it always equal to the downstream.

Thanks Bob, helpful as always!

Is putting it on Fastpath from Interleaving likely to give me much sync wise?

Edited by deleted (Tue 17-Jan-12 21:34:53)

In reply to a post by InQ:

Any ideas if the Interleaving depth is the "D" row i have highlighted below?

Yes, they are the Interleaving depth.

there were "FEC" counts so think this is "Forward Error Correction" and hence Interleaving is on?

EDIT: Yes, also shown as "RSUnCorr" (I think, but you can confirm with the deleted FEC figure). {I always confuse FEC with HEC. Thanks, Bob}

Your ES (Error Secs) are a rather high for Interleaving. Nothing to worry about unless you notice an adverse effect. I'd expect 1 ES in every 5 mins or longer (AS/ES).

Edited by XRaySpeX (Tue 17-Jan-12 21:51:47)

In reply to a post by XRaySpeX:

It's the RS errors that indicate Interleaving.

LOL! I didn't notice those. Yes, I agree, but would also expect RS corrected and uncorrected to equal the FEC figures. As I understand it, Reed-Solomon is just a particular algorithm for implementing FEC. (Probably the predominant one these days).

In reply to a post by XRaySpeX:

Your ES (Error Secs) are a rather high for Interleaving. Nothing to worry about unless you notice an adverse effect. I'd expect 1 ES in every 5 mins or longer (AS/ES).

Think they are high cus i've been messing about, this is the stats since the last sync;

Since Link time = 1 hours 34 min 25 sec
FEC: 5097 72
CRC: 49 13
ES: 32 6
SES: 0 0
UAS: 0 0
LOS: 0 0
LOF: 0 0


So 1 ES every 3 minutes on average but we are over the noisy period of the day, over a 24 hour average it should be alot less.

Edited by deleted (Tue 17-Jan-12 22:00:46)

Deleted the rest but there were "FEC" counts so think this is "Forward Error Correction" and hence Interleaving is on?

Not sure anybody picked this particular point up.

Yes FEC = Forward Error Correction.

It's "forward" in the sense that a error correction block is calculated and sent along with the data so that the other end can correct transmission errors - up to a point - without having to "NAK" the data and ask for a retransmit. It's always on (AFAIK) regardless of interleave setting.

What interleave does is increase the usefulness of FEC in the face of noise spikes. E.g say the FEC is set up to correct a one bit error per byte and a noise spike covers a whole byte - FEC won't correct that block of 8 single bit errors. Now suppose that you send a block of 8 bytes. First byte contains all the bit 0's, second all the bit 1's etc. Now a noise spike covering one byte causes 8x 1 bit errors and FEC can correct it.

The downside is that you have to accept all 8 bytes before you transmit any of them which increases latency.

The real block lengths and error correction capability are, of course, larger but I hope that this illustrates how it works.

Thanks.
I'm pinging bbc.co.uk at 33ms so not bad considering interleaving is on.

Sorry, I can't accept that.

You will never find FEC counts non-zero on Fast Path, ie no interleaving. There is however also Impulse Noise Protection which I know little about, which is generally used in conjunction with FEC, and does not require interleaving to be on. I don't recall any counters for errors corrected by that.

Interleave depth 1 definitely means the packet is transmitted integrally, ie Fast Path, thus avoiding the extra latency of FEC. What happens regarding error correction within the many bytes contained within a packet is another matter, but is not FEC.

For instance Be users have the ability in their control panel to turn interleaving on and off on their line, but only support can turn INP off. I did skim through a writeup about INP once but remember little.

INP settings are one of the reason Broadcom routers and DSLAM line cards get on well together on Be, as there is an extension to the INP specification that Broadcom have implemented and Be use. I don't know of any non-Broadcom routers that have this.

Just got the wife to reboot the router as its daytime and sync has gone up to 10735kbps at noise margin of ~6dB. Will see how that goes then might get interleaving switched off.

Can anyone harbour a guess on how much sync wise interleaving would be holding me back?

Edited by deleted (Wed 18-Jan-12 11:29:28)

In reply to a post by InQ:

Can anyone harbour a guess on how much sync wise interleaving would be holding me back?

Not significant in my opinion. Geek stuff.

There are two effects. One, a possible small change in connection speed, and the other a potential change in throughput at any given sync speed.

The first is the one I would mainly apply "insignificant" to. The second you can think about yourself, by watching how your FEC/RS Corrected ramp up, and whether increases occur fairly evenly or in definite bursts. By removing interleaving every one would require a request for and retransmission of the corrupt package, which again could be subject to corruption ....

With interleaving on, a million or so a day is considered not excessive, but I don't know what effect using Fast Path would have at that volume. A consistent few thousand a second, which I have experienced in the past, is a line fault.

Edited by RobertoS (Wed 18-Jan-12 12:26:23)

OK well i'll see how the line goes stability wise.
It dropped from circa 6dB to just above 4dB noise margin overnight @ 10000kbps sync so a 10735kbps sync and it might see 3dB tonight.

As my post above i had ~5000 FEC's within 90 minutes during the evening so over 24 hours thats 90000 FEC's although i'd expect less due to the lower daytime noise.

Well i'm hovering at mid 3dB's down from 6.5dB at the resync at 11.15am this morning.

adsl: ADSL driver and PHY status
Status: Showtime
Retrain Reason: 8000
Max: Upstream rate = 960 Kbps, Downstream rate = 11312 Kbps
Path: 0, Upstream rate = 960 Kbps, Downstream rate = 10763 Kbps

Link Power State: L0
Mode: ADSL2+
TPS-TC: ATM Mode
Trellis: U:ON /D:ON
Line Status: No Defect
Training Status: Showtime
Down Up
SNR (dB): 3.6 12.3
Attn(dB): 37.5 20.2
Pwr(dBm): 0.0 13.1
ADSL2 framing
MSGc: 53 10
B: 122 44
M: 2 4
T: 3 3
R: 8 6
S: 0.7293 5.9520
L: 2786 250
D: 64 4
Counters
SF: 1896406 774206
SFErr: 191 181
RS: 167831936 3384029
RSCorr: 306567 673
RSUnCorr: 4578 0

HEC: 179 196
OCD: 0 0
LCD: 0 0
Total Cells: 776858336 69140802
Data Cells: 5211682 716249
Drop Cells: 0
Bit Errors: 0 25703

ES: 119 82
SES: 0 0
UAS: 30 30
AS: 30601

INP: 0.73 0.38
PER: 16.13 17.85
delay: 11.66 5.95
OR: 29.24 7.16

Bitswap: 3993 15

Since Link time = 8 hours 30 min 1 sec
FEC: 306567 673
CRC: 191 181
ES: 119 82
SES: 0 0
UAS: 0 0
LOS: 0 0
LOF: 0 0

OK the FEC's are up compared to what i thought i might see. ES averaging 1 every 4-5mins.

Sorry, I can't accept that.

Which bit?

I'm assuming that you're taking issue with the statement that FEC is present regardless of whether you're using fast or interleaved path.

I must admit that the standard is impenetrable as far as I'm concerned but look at this summary, especially figure 8 which shows the superframe structure together with FEC overhead in the fast path.

Also there's a better explanation of interleaving, with diagrams over on kitz

FEC itself would not introduce latency on the transmit side - you can do reed-solomon effectively on the fly accepting the first byte byte and starting transmission/computing the error correction block simultaneously, on the receive side it's no worse than adding a CRC interms of latency. Interleaving does, because you have to accept a bunch of data before transmitting it all.

As to practical implementations, I couldn't say, so how the BE setup works exactly I don't know - I do note that the kitz page says "interleaving and error correction are turned on together" so it's possible that I interpreted the FEC bit of the superframe as mandatory but, in fact, it's optional.

I probably should have mentioned that the interleaved path as more error correction, though (again as far as I understood).

Anyone know for sure?

Edited by mr_bean (Wed 18-Jan-12 20:48:04)

In reply to a post by mr_bean:

Also there's a better explanation of interleaving, with diagrams over on kitz

See the bottom of this page, which hasn't been altered since the site was created in 2009.

Well it did sit between 3dB and 4dB overnight (dipped to just under 3dB for a short period) but router held on so that was a 3.5dB drop....is this kinda normal on ADSL2+? My ADSL1 swing was 1dB.

One oddity is that the noise margin didnt rise to ~6dB once it was daytime which is what it was at the day before when it had been resynced, instead it kept around 5dB or a touch over.....odd?

Also couldnt resist so got XILO to reduce my target noise margin to 9dB and fastpath from their default of 12dB and interleaved (i had it tweaked down to 6dB previous using the routers SNR tweaking capability). Currently down to 6.5dB from 9dB......will see how it goes overnight!

Those changes are normal.

A possible reason it didn't rise as high next day is that some frequencies had been eliminated from the "usable" spectrum, so when the noise fell they didn't come back into use. Which is fine if the noise the next night affects the same frequencies, so the margin might fall to the previous low level but not much lower.

That won't apply to tonight of course as you've had it all reset.