kmac1
Off in new directions
Premium,VIP
join:2001-06-07
Port Neches, TX | Need some help from the Cayman gurus!!
I'm posting a link to a thread over in SBC land that has us scratching our heads.
»Decaying Noise Margin
Maybe someone here could give a hand. |
|
FatJack
Premium
join:2001-06-26
Barnwell, SC
1 edit | I've had a the same problem with my 3546 with Bellsouth. I just quit using my Cayman 3546 and started using the Westell 2200. It never misses a beat. |
|
MgA_ODEN
Excessive
Premium
join:2002-12-28
Spring, TX
clubs: | reply to kmac1
I'm on a RT and changed my settings from G.DMT to ANSI T1.413 . That worked me . 
--
Happiness is 6016/608 |
|
d_l
Barsoom
Premium,MVM
join:2002-12-08
Reno, NV
1 edit | reply to kmac1
Typical Cayman Noise Margin Decay | 
5-27 Noise Margins | 
6-03 Noise Margins | 
Typical bit loading |
A rehash. A typical Cayman noise margin decay is shown in the graph. The sync start point and loss of sync re-train are marked on the bottom. 2.25 - 2.5 days was typical for my decay period. Taking a surge suppressor off the line and "forcing" the Cayman to train in ANSI T1.413 mode instead of the G.DMT mode that it normally did, may have lengthened the decay period by a quarter or half a day to 2.5 - 2.75 days and it also may have made the noise margin swing amplitude smaller. All my Cayman measurements were made with the outside air temperatures between 25 to 45°F. I returned the Cayman because it wasn't stable on my line.
I replaced it with a P645M and noticed that my line had wide noise margin swings, but the modem never decayed or lost sync (re-trained)! Several typical graphs from recent days are shown. I've found that the swing cycle length is related to the outside air temperature (data taken from a nearby golf course weather monitoring station). The average temperatures for each hourly interval is plotted above the noise margin line. Cool temps (under 68°) have faster swing cycles and hot temps (greater than 78°) may not have swings. The phone lines in my neighborhood are buried.
I think that the Cayman just can't handle these noise margin swings and that only those with fast profiles (6016/608) on RT connections have this particular problem.
trader22  was told that he had multiple bridged taps and that was probably what caused his decay problem. I've extracted and plotted bits per tone and SNR data from my modem similar to what would be shown on a SunSet xDSL unit and I can see none of the cyclic tone drop-outs that are typical of a bridged tap.
If you need any other info please let me know. I've been collecting noise margin and CRC data since the start of March. The swings aren't a bother, but the occasional CRC bursts related to them can be a nuisance. |
|
timcuth
Braves Fan
Premium
join:2000-09-18
Pelham, AL
clubs:
 ·AT&T Southeast
|  This is very thought provoking. I have a couple of questions, please:
1) What (briefly) is the difference between ANSI T1.413 mode and G.DMT?
2) How do you "force" the Cayman to train on it? That term implies some difficulty. Is it simply a matter of changing the mode and restarting the Cayman, or is it more complicated than that?
Thanks, Tim
--
As Milton Berle said, "My wife and I have a perfect understanding. I don't try to run her life and I don't try to run mine." |
|
rameus
Bad Craziness
Premium
join:2001-01-28
Martinez, CA
clubs:
·Pacific Bell - SBC
| reply to kmac1
"I think that the Cayman just can't handle these noise margin swings and that only those with fast profiles (6016/608) on RT connections have this particular problem."
I am on a 6016/608 profile via copper CO line and started logging my noise margins beginning yesterday.
I run a 3220H-W11 w/ 6.3.0R7 COS.
06/07/04 9:35 PM Noise Margin : 4.0 dB 16.0 dB
06/07/04 9:38 PM Noise Margin : 14.0 dB 16.0 dB
(after reboot)
06/08/04 8:07 AM Noise Margin : 14.5 dB 16.0 dB
--
One ping only, Please. Proud to be a member of Team Discovery |
|
LBB
Premium,MVM
join:2001-06-14
Bama
·AT&T Southeast
| reply to timcuth
said by timcuth :
This is very thought provoking. I have a couple of questions, please:
1) What (briefly) is the difference between ANSI T1.413 mode and G.DMT?
2) How do you "force" the Cayman to train on it? That term implies some difficulty. Is it simply a matter of changing the mode and restarting the Cayman, or is it more complicated than that?
Thanks, Tim
I'm not sure what the differences are between the 2 but it's really easy to change it in the Cayman if you would like to test. Here's how.
Telnet into the Cayman
At the Prompt type in configure and enter
type in set dmt type and enter
you will have options of [ lite |dmt |ansi | multi ]
type in ansi and enter
type in save and enter
type in quit
then restart
LB
|
|
d_l
Barsoom
Premium,MVM
join:2002-12-08
Reno, NV
2 edits | reply to timcuth
1) ANSI T1.413 mode was the original design spec defining all the DSL parameters and then G.DMT mode was defined later based on ANSI T1.413 but with European input. Basically they are almost identical, but G.DMT has some slight differences in the training sequence Edit: Different handshaking procedures was the word I couldn't think of earlier (there maybe some other differences too).
SBC users have been told by one ASI tech and others that RT DSLAM's used ANSI T1.413 and the CO DSLAM's used G.DMT. Supposedly the modem worked best when it was in the proper operating mode for the DSLAM it was connecting to. Netopia also told me that some RT's operated in either mode. The 3546 version that SBC sells is shipped in multi-mode (shown as Type: ALC MULTI-MODE CP on your stats page) and it is supposed to choose the correct mode for the type of DSLAM. The actual mode it uses is shown to the right of the Operating Mode line.
2) To "force" the Cayman to the ANSI T1.413 mode for my RT connection, I simply had to use the "set dmt type ansi" command from the CLI (read Appendix A - About Config Commands in the User Guide for more info). Remember to restart the Cayman after saving the new command to enable it to retrain.
In addition to what I mentioned before about the benefits of that mode, it bumped up the starting noise margin slightly (0.5 to 1.0 dB) and the Maximum Rate by a few hundred kbps. |
|
timcuth
Braves Fan
Premium
join:2000-09-18
Pelham, AL
clubs: |  reply to kmac1
Thanks, LBB and Barsoom. I am going to try it when I get home, this evening.
Tim |
|
DyerUser
Premium
join:2004-01-29
Dyer, IN
·AT&T Yahoo
| reply to d_l
I don't have any answers, but I did want to toss my limited experience into the mix here.
I posted a few items in the SBC Southwestern Bell forum yesterday in this thread: »Decaying Noise Margin
I have had a 3564 since DSL was first installed on my RT at the end of Jan. I upgraded to the 6016/608 package as soon as it was available. I have not experienced serious problems, with the exception of a few issues during extremely heavy rains. (My lines are underground and the house is over 30 years old). I have noticed the modem losing synch occasionally, but not often. (less than once a week.) Sometimes it will attempt to re-synch before the DSLAM noticed I dropped and so I get the bad username/password error until I reset the modem--but again, this has not happened often. I also have seen that upgrading my firmware may fix this, but haven't been concerned enough to upgrade yet.
When I saw a thread yesterday in the SBC Ameritech forum pointing to this problem, I checked my stats. I was showing 4.5 db Noise Up--but less than 20 CRC errors. I reset the modem and it came back with 16 db. That was yesterday at about 4pm local time. Finally got home and got a chance to look at it at about 9pm. At that point the noise margin was down to 10db. I have not paid attention to the numbers the few times I have looked at them in the past, but I am relatively certain that until yesterday I had not seen them that low or changing that frequently.
So, last night I wrote a perl script to grab the stats and dump it to a text file every 29 seconds. I have over 16 hours worth of data, and my noise margin has bounced between 16.5 and 16db during that entire time. It has not gone below 16. I realize from d_l's post above, that his typical decay did not start until after 24 hours, but I had certainly seen something in 6 hours yesterday.
There are only two things I can think that are different on my end between yesterday and today:
1) I'm querying the 3564 on a pretty regular basis.
2) Yesterday evening was the first time I turned the AC on in my house this year and I left it on today. My router is in a basement room, so it never got very hot, but the room does have a few running machines in it and the humidity had risen. We are maybe talking a difference between 80/85 F yesterday and 65/70 F now with a pretty large relative humidity difference. I can't imagine that would make such a huge difference, but in the post by d_l above with all the graphs, he mentions that he thought there was an inverse temperature relationship.
I will eventually back off on my logging frequency, but I will continue logging and will make them available to anyone who is interested. |
|
LBB
Premium,MVM
join:2001-06-14
Bama
·AT&T Southeast
| said by DyerUser :
So, last night I wrote a perl script to grab the stats and dump it to a text file every 29 seconds.
So, how hard is it to do that? Can you share the steps with us if it's not to involved?
BTW, I'm not trying to get this thread off topic and I hope we stick to the main topic but I would like (if you would) share how we can do monitoring for ourselves.
LB |
|
timcuth
Braves Fan
Premium
join:2000-09-18
Pelham, AL
clubs:
·AT&T Southeast
|  reply to kmac1
I am not sure if my problem is like the one described in this thread, at all. My modem has been up for 2 days, 3 hours. It just had a loss of sync failure a few minutes ago. But, here are the current stats (note the low number of CRC errors, which are cumulative over the 2 days, 3 hours):
DSL Statistics:
Type : ALC MULTI-MODE CP
Datapump HW Rev : ff
Datapump FW Rev : 3.9.16
Datapump Vendor ID : ffff
Current Status : LINK UP
Operating Mode : G.DMT
Data Path : Fast
Downstream Upstream
Current Rate : 3008 Kbps 384 Kbps
Maximum Rate : 3234 Kbps 711 Kbps
Noise Margin : 8.5 dB 12.0 dB
Attenuation : 55.5 dB 31.5 dB
Out Power : 19.5 dB 6.0 dB
Near Far
FEC Counts Fast : 0 4
CRC Counts Fast : 188 1
HEC Counts Fast : 113 3
Tim
--
As Milton Berle said, "My wife and I have a perfect understanding. I don't try to run her life and I don't try to run mine." |
|
timcuth
Braves Fan
Premium
join:2000-09-18
Pelham, AL
clubs:
·AT&T Southeast
|  And, here are my stats after switching it to ANSI mode. Little or no difference, especially in the SNR. I'm going to go with it for a few days, though, to see if the LOS failures stop or slow down.
DSL Statistics:
Type : ALC ANSI DMT CP
Datapump HW Rev : ff
Datapump FW Rev : 3.9.16
Datapump Vendor ID : ffff
Current Status : LINK UP
Operating Mode : Ansi T1.413
Data Path : Fast
Downstream Upstream
Current Rate : 3008 Kbps 384 Kbps
Maximum Rate : 3379 Kbps 711 Kbps
Noise Margin : 8.5 dB 12.0 dB
Attenuation : 55.5 dB 31.5 dB
Out Power : 19.5 dB 5.5 dB
Near Far
FEC Counts Fast : 0 0
CRC Counts Fast : 14 0
HEC Counts Fast : 7 0
Tim
--
As Milton Berle said, "My wife and I have a perfect understanding. I don't try to run her life and I don't try to run mine." |
|
FatJack
Premium
join:2001-06-26
Barnwell, SC | Tim, be sure to post back if it makes any difference. |
|
DyerUser
Premium
join:2004-01-29
Dyer, IN
·AT&T Yahoo
| reply to LBB
said by LBB :
So, how hard is it to do that? Can you share the steps with us if it's not to involved?
I'll share the script. It's rough, sloppy and not something I wrote for public consumption--and I'm too lazy to re-write it to accept variables passed in.
However, here it is. I have a Linux box, so I run it from there. It needs to have Net::Telnet installed. I suspect it would run fine on a Windows box that had a perl installation, but I am not sure. I suspect that others have something better.
All of that said, I've attached it as a text document. Comments inline for places where I've hard-coded things that you will need to update for your installation.
Simply save this file as something like check_dsl_stats.pl and launch it from a command line with
perl check_dsl_stats.pl
btw, my Noise Margin has dropped down to 14 in 1.5 hours after bouncing between 16 and 16.5 for nearly 17 hours. Still watching it. |
|
timcuth
Braves Fan
Premium
join:2000-09-18
Pelham, AL
clubs:
·AT&T Southeast
| reply to kmac1
No apparent help from the change:
Wed Jun 09 01:55:33 2004(UTC) L3 DSLU: Link LOS Failure
Wed Jun 09 01:55:33 2004(UTC) L3 DSLU: Resetting modem hardware
Wed Jun 09 01:55:33 2004(UTC) L4 ATM: PHY layer has gone down
Wed Jun 09 01:55:33 2004(UTC) L3 IPCP: (ppp/vcc1) PPP protocol down
This was just about 3 hours since I re-sync'ed in ANSI mode.
Tim
--
As Milton Berle said, "My wife and I have a perfect understanding. I don't try to run her life and I don't try to run mine." |
|
DyerUser
Premium
join:2004-01-29
Dyer, IN
·AT&T Yahoo
1 edit | reply to d_l
Well, my chart is neither as pretty nor as detailed as d_l's above, but here it is. My charts also do not look much like his. I am not seeing the wide noise margin swings during the drop that he did, but my drop is more precipitous than his was. I should also note that my CRC errors are almost non-existent compared to d_l's. I'm seeing a grand total of 22 CRC errors for the 34+ hours the graph covers. I'm still collecting data.
I have not lost synch, and in fact my downloads seem fine. The plot is noise margin against total elapsed time. I went for about 19 hours with noise margins bouncing between 16.5 and 16 and then starting at 5:45pm last night, it dropped quickly to a low of 5.5 and then started a climb back up to 9.5 before beginning another drop.
Again, I have not noticed this problem much before, which could mean that it has been going on but since I'm not getting CRC errors I just never noticed. It could also mean that it is only a relatively recent issue for me. Again, Chicago has only just this week begun getting typical hot, humid summer days. My drop-off began late in the day after the temp had climbed above 90. It began falling and then late at night it began to climb again, until the temps started to warm back up. I don't have graphs from Monday, but on Monday in the late afternoon, I restarted my 3564 and by 10pm that night it had a noise margin of 10, so the fall-off for me on Monday happened much sooner than the Tuesday decay, but would have been at about the same time of day. Is it possible that the 3564's are more sensitive to some effect on the RT equipment that is related to ambient temperatute?
[edited to mention CRC errors and provide speculation in final paragraph.] |
|
d_l
Barsoom
Premium,MVM
join:2002-12-08
Reno, NV
| reply to kmac1
Trader22's Cayman Data |
So I think when I had my Cayman, I had:
• the Cayman noise margin decay problem,
• compounded by noise margin swings possibly from the RT,
• further compounded by CRC errors from impulse noise sources within my house (poor IW) and unknown external sources (perhaps along the phone lines or around the RT).
LOL Whew! Can you say bedeviled DSL?  Got rid of the Cayman and have a temporary home run. Maybe my RT needs to be looked at.
My Cayman noise margin decay results may not be typical of everyone elses decay curves. |
|
DyerUser
Premium
join:2004-01-29
Dyer, IN
·AT&T Yahoo
1 edit | said by d_l :
Well, doh! More decay data was staring me in the face for the last three months right here: »Decaying Noise Margin So I plotted trader22 's data that he collected every 10 minutes for about five days. His data may be more typical of the Cayman noise margin decay process. Note he didn't have any noise margin oscillations other than the expected 0.5 dB variations between readings!
Yeah, I had looked at his raw data as well.
I think my initial decay looks more like his. However, I am now seeing the oscillations that apper in your decay from the very beginning. haven't hit a sync loss yet--and if this is what has been causing my occasional sync loss, it will likely be more than a week before I do. I won't give hourly reports, but will contribute when it makes sense for comparison purposes.
I've not plotted my CRC errors, because again, I'm not seeing much. My data collection has been going on for 39+ hours and I've got 25 CRC errors total. |
|
Ogy1012
join:2004-05-09
Austin, TX
clubs:
| reply to kmac1
I don't have an easy way to graph my logged data yet. So for now I'll post in raw form.
This set of data shows date,time,downlink noise margin:
NM1.zip
contains a decay cycle in progress along with the
stats after the router re-trained.
This is a running set of current stats covering 4+ days:
NM.zip
Contains one complete decay cycle and one decay cycle currently in progress. The in progress cycle is different in that it drop down to 10 dB and then climbed back to 13.5 dB and has somewhat leveled off. |
|
