Canadian Broadband → Re: Tweaks

Now i think i'm capturing my neighbour's cooking
hour:  the peak at 11h30 didn't come from here!

wow, your stats changed quite a bit... I guess the most notable one would be the downstream attenuation.

What did you use for your shielding?

I think I may be doing the same thing in a bout a weeks time when a newer modem arrives.
I can't do much about the outside (underground)
But I should be able to figure out if internal appliances are effecting me(it feels like it).

(my current adsl modems only show upstream numbers [3com duallink, ss5200 240]) soon to have 6520 and usrobotics 9108

Our upstream profiles match (your older profile), but I don't think my downstream is anywhere near yours

Hi,

At first it was a shock because my latency went up
from an average of 47 ms (+16 %/-31 %) at 5056/800
Kbps FastPath (with VelCom) to 75 ms using the new
6016/512 Kbps InterLeave profile...  That's a 60 %
increase but the higher DownLoad speed compensates
for the initial delay to some point, it seems;  my
browsing feels more responsive, actually, but it's
still too early to draw any definitive conclusion.

What i know is that Bell Sympatico would bring the
latency around 35 ms since the signal doesn't have
to travel between me, Québec city and Toronto just
to reach the ISP (before it really does anything):

»/r0/do ··· 0%20.GIF
Local vs Québec-Toronto transit



I'm afraid i don't agree about the attenuation but
you may have mistaken my GNet statistics for those
i obtain from the SpeedTouch.  Here are more 'DMT'
captures generated while having the ST546 on line:

Thomson SpeedTouch 5x6v6 FirmWare v6.2.H (v6.2.17.5), Bicephale, 2007-Sep-28
Thomson SpeedTouch 5x6v6 FirmWare v6.2.H (v6.2.17.5), Bicephale, 2007-Oct-16
Thomson SpeedTouch 5x6v6 FirmWare v6.2.H (v6.2.17.5), Bicephale, 2007-Oct-24
Thomson SpeedTouch 5x6v6 FirmWare v6.2.H (v6.2.17.5), Bicephale, 2007-Nov-3

I've tried eleven FirmWares with the SpeedTouch so
far, it's unclear which is best for me but i shown
that shielding does have a positive effect, in any
case.  All that matters is to use galvanized iron:
it's para-magnetic and the coating eases soldering
a lot.  I use massive iron tubing too but i hardly
recommend it unless the craftwork is no problem...

Slotted sleeves as these can do the job just fine:

»/r0/do ··· s%20.JPG
Ferro-Magnetic Metal Sleeves

The shields will work if they're grounded properly
(hence the need to solder).  Of course, i guess it
is evident for most of us that the close proximity
to AC power lines means caution must be mandatory!



A couple 24 error rate records will be required if
you need to evaluate the noise sources relative to
domestic activity, instantaneous spectral response
curves are practically useless for this task.  I'd
suggest you focus on your SpeedStream 5200 and try
to make 'DMT v9.x' work with it, should the SS4200
happen to be close enough that means 'DMT can work
with it but i'd appreciate too see some proof over
this...  The SS6520 and the 3Com USR-9108 are even
less documented relatively to 'DMT', i'm afraid...

If there's room for tweaking in your own situation
then my advice is to play with the 3Com since that
device does support BandWidth Shaping extensively:
on a tone-per-tone basis, on top of that.  Maybe a
genuine SpeedStream can do as much using 'DMT' but
it remains to be seen.  Feel free to be a pioneer!

I tried dmt on the 5200. It only gives me some upstream numbers and i couldn't figure out how to have it run as a bridge. The 3com has virtually no play, but it is an old alcatel chipset that may work with dmt software but Ive searched for the telnet password with no success (found some that didn't work).

The 6520, I couldn't find much information on it for adsl stats for when it comes.

the 9108 uses dmt v8. uses the broadcom 9348 chip which a lot of modems use that supposedly can withstand iffy connections.

Your gnet numbers are similar to what I get currently, so when I get a modem that shows me stats both ways, I would like to see what I get, cause I don't feel my potential will be the same as yours currently.

*my connection is fine for a long period of time (upto about 24hrs) and then it just crc errors about 300 times faster then I can refresh the stats. --happens for a duration of a few seconds, then its fine for many many hours again. (I would like to use that dmt graph you have and use a process of elimination, etc...)

Was the cut/worn area done by you or by contact with the aluminum soffit?

Hi DKS,

I didn't climb up there, most unfortunately!  If i
had, no one would wonder about the tape nor what's
hidden behind because it would be long gone as i'd
have been unable to ignore such a nice opportunity
to get this potential problem fixed.  I suppose it
was damaged consequently to repeated friction with
an edge of the alluminium cladding and yet it does
not mean i can totally exclude another cause.  For
example, we had the roof worked on this summer and
hence it's also possible that a worker severed the
old tape somehow, although that's not as likely...

Anyway, the detail which really got me worried was
the tape itself and more specifically what it does
not allow us to see.  My pictures provide evidence
that this tape may now behave more like a compound
of tissue with tar spots than a real insulator, if
there happens to be a splice underneath then we're
talking about a cell where the growth of dendrites
and whiskers could end up being actually promoted!

The major disruption episodes which "hammer pulse-
dialing" can cure temporarily are far from being a
frequent event, that part of my triple-personality
line resurfaces only a few times a year and i have
little hope that Bell would get here right in time
to catch it in action since they don't even return
calls...  How can one explain this to them anyway?

If they make such a big difference, somebody should make a modem with this sort of filtering built-in. It's my firm belief that anything you can do to the line beyond shielding (such as filters, or coils) could also be built into the device using the line. After all, you're not doing anything that the device couldn't do if the manufacturer decided to do it.

One day, you should produce your own line of modems, with all sorts of line filtering equipment built-in. You can market it to all those pour souls stuck on noisy lines

Not that contradictory when you think about it: the choke filters out some frequencies, so while the modem is training, it does not allocate bits to these buckets. The result is: less buckets, more bits per bucket, less SN ratio:

Without choke: 193 buckets are used (tones)
With choke: 214

Also, look at the number of bits per tone with and without the choke.

The logic is:

1- Find how many bits I can put in each bucket with SNR>=6
2- Use the better buckets (highest SNR), with the least amount of bits per bucket to support the required BW.

So, you most likely have less errors with the choke because you are removing the tones that are more susceptible to errors (and this is specific to your line). I bet you would get the same result if you you were able to «disable» these tones.