Kitz Forum
Broadband Related => Broadband Technology => Topic started by: Weaver on December 25, 2015, 02:25:03 PM
-
[Edit:: - note: valid criticisms mean the rest of the thread needs to be read concerning the unproven assumptions expressed in the following post]
I have three ADSL2+ long lines, all with a downstream attn of 67 - 68 dB. I made a comparison between modems configured to be locked in ADSL2 mode vs a config where they should default to choose ADSL2+ mode.
The sum of the downstream sync rates on the three lines was as follows:
ADSL2+ : 8107 k (note: assumed : mode='auto' )
ADSL2 : 8341 k (note: verified : mode='adsl2' )
As expected, ADSL2 performs better on a long line when ADSL2+ can not successfully exploit its additional octave of very high tones as they don't propagate. The reason, it has been suggested to me, that ADSL2 performs better is that the input bandwidth in ADSL2 mode being half the width means that less garbage is allowed in through the narrower input window.
-
G.Dmt uses 256 tones.
ADSL2 uses 256 tones. (But more efficiently than G.Dmt. Something to do with the bit loading -- Kitz has some details on a page at the main site.)
ADSL2+ uses 512 tones.
It would be interesting to compare the synchronisation speed for one of your lines with the modem configured to use each of those three modulation modes in turn.
I am uncertain if you would see --
ADSL2 > ADSL2+ > G.Dmt
-- or --
ADSL2 > G.Dmt > ADSL2+
-
Weaver, did you check they were actually operating in ADSL2+ mode? If they were, they would usually report the attenuation about 2-3 dB greater. I'm not sure what you mean by "should default to choose ADSL2+ mode", the default Auto mode often ends up operating as ADSL2 when no bits are allocated to any of the higher ADSL2+ tones.
I thought that for all the tones that get zero bits allocated to them when the modem connects, which would very likely be all the higher ADSL2+ tones, there would then be very little signal transmitted on these tones, and I wouldn't have thought the modem would be paying much attention to them either.
My own line, connected to TSTC 21CN equipment, does not automatically select ADSL2. Here are some historical stats and bitloading graphs:
ADSL2+:
US Connection Rate: 752
US Line Attenuation: 36.9
US Margin: 4.4
DS Connection Rate: 3347
DS Line Attenuation: 60.5
DS Margin: 6.0
ADSL2:
US Connection Rate: 882
US Line Attenuation: 36.9
US Margin: 5.3
DS Connection Rate: 3678
DS Line Attenuation: 58.6
DS Margin: 5.8
I'm not exactly sure why, but for my line, the difference between ADSL2+ and ADSL2 affects the upstream and the lowest downstream frequencies, but the rest of the ADSL2 tones look much the same.
I wouldn't be surprised if G.DMT mode gets about the same speed as ADSL2, or even slightly higher. My own line gained about 500k of DS bandwidth when switched from ALCB exchange equipment to the TSTC equipment it is on now (which happened in 2010), both using ADSL1 at the time. So some of Weaver's extra speed could just be due to being on a more modern DSLAM, and not just the change in mode from G.DMT to ADSL2.
-
The figure marked adsl2+ is from three modems using a default config mode of auto, presumably, I haven't checked this. The speed went up dramatically using these modems when the lines converted to adsl2+.
The figure marked adsl2 is from three new modems configured (and checked) to be in adsl2-only mode.
I finally learnt the true ds attn, by the way, as A & A's servers can report the higher numbers niw rather than just sticking on 63.5 dB.
-
> did you check they were actually operating in ADSL2+ mode
Indeed not, mea culpa. This is just an assumption. It may be that the new modems are just better than the old ones (I have seven units in stock, four existing ones plus three new units that turned up this week with the special ADSL2-only mode set up for me by A & A - I checked them)
So the adsl2+ line could simply be labelled "Old, Unknown" I'm afraid.
-
> So some of Weaver's extra speed could just be due to being on a more modern DSLAM
This would fit the bill very well. My (measured) speed total went up by a Meg from 5.5-6.2 as reported about a month ago in the "breaking the 6.0 thread". Then on the 23rd and 24th I could watch each of the lines gain yet more speed, and lots of it, live, with figures coming from A & A modem status reports. I got a second Meg combined over the period 23rd-24th.
So one Meg (divided by three) from a better DSLAM, while still on ADSL1, attn unknown because it locked out at 63.5. Then a second Meg combined after the ADSL2x switchover. All in all about a 40% improvement between the beginning of November and xmas day.
Note: I don't have a convenient way of interrogating the modems directly, in that I don't have two Ethernet connections into the back of one via a switch. I'm getting info via A & A ( via BT seemingly) interrogating them for me on request and adding the reports into the A & A clueless.aa.net.uk per-line logs.
The comparison shown in the original post could then be due to better modems, or even DLM.
I'll need to qualify the original post, indeed. No prizes for experimental work, (excuse is that my degree did say literally "theoretical" on it) :-[
-
> So some of Weaver's extra speed could just be due to being on a more modern DSLAM, and not just the change in mode from G.DMT to ADSL2.
It has to be so, and I'd say “about 50%”.
-
The new vs old modem results of today are consistent across the three modems. New one slightly faster than the old one in terms of ds sync. The upstream doesn't show this improvement.
-
Raw data: line 1
Old modem ::
BT Test xDSL Status Check:Pass Standalone sub test passed successfully.Pass OK. Circuit In Sync
BRAS=2161Kb/s FTR=2272Kb/s MSR=2272Kb/s ServOpt=1 I/L=I
Up Sync=637Kb/s LoopLoss=41.7dB SNR=6.1dB ErrSec=0 HECErr=0 Cells=0
Down Sync=2731Kb/s LoopLoss=67.4dB SNR=6dB ErrSec=9 HECErr=N/A Cells=0
New modem ::
BT Test xDSL Status Check:Pass Standalone sub test passed successfully.Pass OK. Circuit In Sync
BRAS=2161Kb/s FTR=2272Kb/s MSR=2272Kb/s ServOpt=1 I/L=I
Up Sync=637Kb/s LoopLoss=41.6dB SNR=6dB ErrSec=0 HECErr=0 Cells=0
Down Sync=2798Kb/s LoopLoss=64.7dB SNR=6dB ErrSec=0 HECErr=N/A Cells=0
And I definitely should have followed the anti-DLM procedure KFC (suggested in a long-winded post some months ago), to check for DLM effects.
-
I've just noticed the change in "loop loss" there. Could it be that there actually was some use of the higher tones, or was it just a a matter of calculation of snrm differently across the (assumed) mode change?
-
Lines 2 and 3, old modem vs new modem (different physical units)
Line 2
Old modem line 2 :
BT Test xDSL Status Check:Pass Standalone sub test passed successfully.Pass OK. Circuit In Sync
BRAS=2280Kb/s FTR=2272Kb/s MSR=2272Kb/s ServOpt=1 I/L=I
Up Sync=396Kb/s LoopLoss=43.6dB SNR=5.9dB ErrSec=0 HECErr=0 Cells=0
Down Sync=2656Kb/s LoopLoss=68.6dB SNR=6dB ErrSec=0 HECErr=N/A Cells=0
New modem line 2 :
BT Test xDSL Status Check:Pass Standalone sub test passed successfully.Pass OK. Circuit In Sync
BRAS=2280Kb/s FTR=2272Kb/s MSR=2272Kb/s ServOpt=1 I/L=I
Up Sync=399Kb/s LoopLoss=43.5dB SNR=6.1dB ErrSec=0 HECErr=0 Cells=0
Down Sync=2670Kb/s LoopLoss=66.2dB SNR=6dB ErrSec=0 HECErr=N/A Cells=0
--
Line 3
Old modem line 3 :
BT Test xDSL Status Check:Pass Standalone sub test passed successfully.Pass OK. Circuit In Sync
BRAS=2455Kb/s FTR=2272Kb/s MSR=2272Kb/s ServOpt=1 I/L=I
Up Sync=496Kb/s LoopLoss=42dB SNR=4.7dB ErrSec=0 HECErr=0 Cells=0
Down Sync=2791Kb/s LoopLoss=67.4dB SNR=6dB ErrSec=15 HECErr=N/A Cells=0
New modem line 3 :
BT Test xDSL Status Check:Pass Standalone sub test passed successfully.Pass OK. Circuit In Sync
BRAS=2455Kb/s FTR=2272Kb/s MSR=2272Kb/s ServOpt=1 I/L=I
Up Sync=433Kb/s LoopLoss=42.1dB SNR=5.8dB ErrSec=0 HECErr=0 Cells=0
Down Sync=2873Kb/s LoopLoss=65.1dB SNR=6dB ErrSec=6 HECErr=N/A Cells=0
-
Notice the d/s SNRM changes each time. Wouldn't that support the original premise - that they have changed operating mode?
It doesn't seem to be possible then for this to be just a DLM effect, or due to the difference between new vs old modems.
I have seven DLinks now. I'm going to be good for any lightning strikes. Or else I could start a DSL-320B-Z1 museum.
-
How did A & A save the changes after they programmed the modems for me? I'm aware that some devices need you to hit something to commit the changes.
(I can see that they did it as it's shown in the web UI setting pages.)
Perhaps there's nothing needed, could be it writes to flash straight away. I could of course test it by power-cycling and seeing if the choice of mode setting still shows up in the UI.
-
Btw, many many thanks for extremely helpful criticism of this original post. ;D
You've helped enormously.
-
> the difference between ADSL2+ and ADSL2 affects the upstream and the lowest downstream frequencies
Note that there is a difference in the u/s performance, not a consistent improvement.
-
I notice from ejs's bits-per-bin graph that there's no chance for me of getting near using my higher octave, as ejs doesn't even get to tone 255 in adsl2+ mode with his lower d/s attn.
-
Next day (ADSL2 mode)
2015-12-26 ::
Line 1 :
BRAS=2161Kb/s FTR=2272Kb/s MSR=2272Kb/s ServOpt=1 I/L=I
Up Sync=640Kb/s LoopLoss=41.6dB SNR=6dB ErrSec=0 HECErr=0 Cells=0
Down Sync=2784Kb/s LoopLoss=64.5dB SNR=6dB ErrSec=0 HECErr=N/A Cells=0
Line 2 :
BRAS=2280Kb/s FTR=2272Kb/s MSR=2272Kb/s ServOpt=1 I/L=I
Up Sync=419Kb/s LoopLoss=43dB SNR=4.3dB ErrSec=5 HECErr=0 Cells=0
Down Sync=2738Kb/s LoopLoss=65.2dB SNR=6dB ErrSec=46 HECErr=N/A Cells=0
Line 3 :
BRAS=2455Kb/s FTR=2272Kb/s MSR=2272Kb/s ServOpt=1 I/L=I
Up Sync=496Kb/s LoopLoss=41.9dB SNR=5.8dB ErrSec=0 HECErr=0 Cells=0
Down Sync=2902Kb/s LoopLoss=65.1dB SNR=6dB ErrSec=0 HECErr=N/A Cells=0
-
And when the lines came out of the training phase, everything changed. Bizarrely.
Speeds that had been seen around ~2700k sync rate d/s at 6dB SNRM came down to something like 2200k at the same speed or at 9dB SNRM.
I have absolutely no idea what's going on. I would have thought that the training phase would automatically ensure that you come out with the right speed to match a 6dB chosen target SNRM?
Anyway, I lost a huge amount of speed at the end of the training phase.