Kitz Forum

Broadband Related => Broadband Technology => Topic started by: waltergmw on November 06, 2011, 10:12:35 AM

Title: ADSL2 on long lines
Post by: waltergmw on November 06, 2011, 10:12:35 AM

I had always thought that it was not recommended to use ADSL2 on long and bad quality lines.

However I have now observed, via a highly respected and efficient ISP, that it is (sometimes ?) possible to achieve a better result if a BT Wholesale line with a 2700HGV modem is set to:-

WBC 160K - 24M No delay (INP 1) 6dB Downstream, UC No delay (INP 1) 6dB Upstream (ADSL2+)

Whether or not you can convince all other providers to organise such a change is another matter entirely !

The picture attached shows very good stability since the configuration change was completed even though we had torrential rain overnight last Thursday.

EDIT Here's the details as reported by the ISP:-

The new xdsl details as below;

xDSL Status Check Result

Sync Speed: 1902 Kbps
SNR Margin: 6.6 dB
Loop Loss: 70 dB <==
Errored Seconds: 77
HEC Errors:
Cell Count: 25281

Sync Speed: 554 Kbps
SNR Margin: 6.5 dB
Loop Loss: 39.3 dB <==
Errored Seconds: 0
HEC Errors: 0
Cell Count: 26659

BIP Interface Data
FTR: 1152
MSR: 1440
BRAS Profile: n/a

I noticed the loop loss jump up slightly higher than before but this can be attributed to the regrade as MSAN is trying to push more of the signal through.

DLM has automatically added INP, so thats some time saved.

IP Profile Result

Product Type: 21CN - ADSL2+

Downstream Line Characteristic
Profile: 160K - 24M
INP: 1
Interleaving: Medium
Target SNR: 6 dB

Upstream Line Characteristic
Profile: UC
INP: 2
Interleaving: Medium
Target SNR: 6 dB

Kind regards,
Title: Re: ADSL2 on long lines
Post by: burakkucat on November 06, 2011, 06:45:17 PM
The interesting thing to note is that the 2700HGV first tries at ADSL2+ mode but quickly settles on ADSL2, due to an inconsistent signal with the former.

I see exactly the same with my own "not so long" line. (See Screenshot-1 & -2, attached below.)

DSL                                           Down       Up

Current Rate:                        3873 kbs       872 kbs
Max Rate:                             3873 kbs       872 kbs
Current Connection:
   Current Noise Margin:         10.0 dB       12.1 dB
   Current Attenuation:           46.4 dB       27.8 dB
   Current Output Power:        18.9 dB       12.4 dB

Looking at my SNRM graph, I see I have had an unusually "noisy" past 24 hours.  :( 

I wonder what external event has caused that . . . surely not fireworks! :-\

This is an appropriate moment to acknowledge the help I have received from Bald_Eagle1 with guidance in the use of gnuplot to create the graphs.  :)
Title: Re: ADSL2 on long lines
Post by: kitz on November 08, 2011, 11:02:00 PM
>> first tries at ADSL2+ mode but quickly settles on ADSL2,

There could be a valid reason for that.
Adsl2 uses the same frequency ranges as adsl 1..  but adsl2 (and adsl2+) utilises something called S=1/2 mode.
The effect of this is less overheads due to more efficient RS code words.
s=1/2 mode is what enabled some people (as long as they were on an exchange with MSANs that supported it) to be able to sync at 8128 when supposedly on adsl1 even if they were interleaved.

adsl1 supports up to 8Mb and although adsl2 uses the exact same frequencies by using s=1/2 mode increases the available max speed.

adsl2+ uses the extended frequencies and s=1/2 mode which is why the max speeds are up to double that of adsl2.

I should imagine it also entirely depends on the line and which particular frequencies cause the most noise..  but its entirely possible that by using adsl2 you are getting the best out of both technologies without opening the line up to the extended frequencies (and higher attenuation?) available in adsl2+ 

Edited to add. 

adsl2 and adsl2+ also has the added benefit of being able to support a minimum of one bit per tone during the initial bit loading process, so a channel will still work down at 3dB SNRM as opposed to 2 bits in adsl1 (6dB).. so if the line isnt doing loads of bit swapping all over the place it may well hold sync for longer.

Edited to add more.

I've just recalled something in the distant past (about 8-10 months ago?) when Im pretty sure me and leo had a convo on the phone..  and I spouted at him the exact same theory why adsl2 may work better on some lines.
Title: Re: ADSL2 on long lines
Post by: kitz on January 25, 2012, 01:12:43 PM
btw...  whilst doing something else I stumbled across a highly technical ITU document and amongst a pile of other stuff it actually listed the enhancements of ADSL2 over ADSL1.

Basically the gist of it is that the modulation method used by ADSL2 is more configurable and more efficient at handling overheads.. specifically optimised use of the RS coding gain and frame structure.

This actually confirms something Ive speculated on now for some time and thats in relation to something referred to as S=1/2 mode.

Despite adsl1 and adsl2 having the same frequencies available to them,something Ive always said is the constraint of adsl1 down to 8Mb is because of the frame size needed for RS encoding.  s=1/2 mode makes this more efficient and rather than the maximum frame size being constricted down to the RS frame size, its back into the ATM realm... more efficient use data of the bins and therefore a possible slight increase in speed.

ADSL2 is giving you the best of both worlds, the better modulation method but without opening up all the additional frequencies which may attenuate the line that comes with adsl2+.  It may not work in all cases as each line is different.. 

....but it was nice to see something in black and white about this..  and I'd say it also confirms my theory as to why  some adsl1 interleaved lines are able to sync at 8128.. and thats if you have a router that supports s=1/2 mode (efficient RS modulation for ADSL2/2+) and the dslam does.. or far more likely...  the BTw MSANs..  which of course will because they are designed for adsl2+.