how-to block ads
|
dslgr
join:2008-05-07
| reply to Bicephale
Re: SpeedTouch 516 Tweaking :)
@Bicephale first of all, I would like to thank you for the comprehensive post you've made above.
I can see your excitement for the USR's tweaking capabilities, as it matches mine, the first time I realized I could "play" with my line and try to get it to reach its limits.
Sadly though, my initial expectations weren't met and that's why I asked you in my previous post if there was something more besides tone selection, that could have slipped my mind.
I agree that tone selection is a nice feature but imo it can't compare with the tweaking options the GNet, for example, has to offer, which you are very familiar with, as you made the best out of them.
From what I gained from my experience, there's not much tweaking you can do and benefit from it. It's straight forward really, you just select which tones to use and I have "played" both with the cli --tones parameters and with the GUI too. Eventually I realized that besides the extra parameters, it gives you the exact same results.
Reading that the cli command lets you specify the starting tone used in the upstream and downstream direction, and the number of tones in each, my mind went at once to spectral shaping. And with spectral shaping I mean something like that. Unfortunately there's no such option, or at least I couldn't make it work. Regardless of the bitmap I sent to the router, it just masked some tones and didn't change the spectrum.
The only use I could find for the tone selection feature is if you have some kind of RFI which you can't avoid and which hits at random time and in specific frequencies. With tone selection you could exclude those problematic tones, so you won't have to worry if the RFI is on or not, leaving your line steady in all cases.
In your post, you mention another benefit from it, which I hadn't really considered and that's the reduction of internal crosstalk. I probably didn't think of it because I value the upstream rate too much (and it already has a hit, as an annex b line, and from my crappy NT1 device), to sacrifice it for a bit more downstream, but I would be more than glad to give it a try first chance and present you with some results.
In fact, if you like, you could give me specific parameters you would like me to test (like specific tones, ranges and such). The only problem is that I can't really give you longterm statistics (such as crc errors in days uptime) because unfortunately my line is not so stable. With the ST I will probably have a drop every one or two days, which doesn't really bother me, but with the USR, that number goes up to 4-5 a day, so you see I can't collect enough data.
Those are my thoughts for now. I will be waiting for a response while getting the USR out of the drawer. Hope there's more to do with it.
Greetings from Greece and a fellow tweaker.
Thank you. | |
Bicephale
join:2005-09-24
 ·TekSavvy Solutions..
·Velcom
| Hi,
Perhaps there's next to nothing to tweak on your side
simply because nothing is really broken as some of us
around here would consider 24 Mbps with pure envy!...
DesiGuy1981 would be lucky if could get 2496 Kbps; if
he were the one with a 3Com, i'd recommend him that he
tries an UpStream band starting with bin 6, 7 or 8 and
ending with bin 14. This should limit the UpStream
signal to 256 Kbps or so and then, maybe, he would be
allowed to dream of a DownStream speed of 2496 Kbps...
I expect the DownStream starting bin to be somewhere
between 32 and 36 and i find it even more improper to
guess what the ending bin should be but lets put it
in the 70 to 248 range. Most unfortunately, your own
DSL signal doesn't compare with that of DesiGuy1981 so
i'd rather not draw definitive conclusions from future
tests: i've learned to be carefull with predictions, not to
mention i found no gain/bit-loading CLI commands neither.
I suppose a Bell technician wouldn't raise his profile
unless there's a strong basis to request it anyway.
Fingers crossed...
 | |
dslgr
join:2008-05-07
edit:
May 8th, @09:54PM
| Well my line has its problems too. After all, that's what got me into this tweaking expendition I started some time ago. If everything worked as it should, then I probably wouldn't look any further.
First of all my line is up to 24Mbps/1Mbps, which for me that I live about 2,5-3km away from the CO, means that I get to enjoy about half of that. My line is ADSL2+ over ISDN (annex b), rate adaptive with a target SNRM of 9db for the downstream.
It usually synchronizes around 10-11Mbps DS and 650-700kbps US. I should have been able to have 1Mbps upstream but that's where my second problem arises. The upstream takes a big hit, mostly from my NT1 device (the device where the phones connect in a ISDN line). That's why I was looking for a spectral shaping solution, but unfortunately haven't been able to find one till now.
Wish I could do some tests for DesiGuy1981, but as you have already said, our lines are disproportionate. | |
DesiGuy1981
join:2008-05-01
Scarborough, ON
| reply to Bicephale
than bicelphale..i wonder whats a 3com? is it a software like DMT? if it is how can i download it?
r u suggesting me to download U.S.ROBOTICS software? i can give it a try...
i luv readin wat u guys write but the only problem is u guys use so professional wrds that so go over my head..lol
im very new on this..so plz bare with me. thanx again both of u..i do appreciate ur help. | |
Bicephale
join:2005-09-24
·TekSavvy Solutions..
·Velcom
edit:
May 9th, @10:07PM
| Hi DesiGuy1981,
I don't know about dslgr but i was as professional
as any other user when i subscribed to DSL for the
1st time so i'd like to help you save the time and
frustration if i can. You don't have to embrace a
learning curve as steep as mine was, just consider
that the existing 3rd-party trouble-shooting tools
will help you to acquire a proper understanding of
the situation. 'DMT' is the most popular, use the
version corresponding to your MoDem, if supported.
My favourite trouble-shooting tool is the 24 h CRC
Error ("noise") curves generated by a GNet BB0060B
using 'DMT v2.20'; those from an Efficient SS4200
with generic FirmWare combined to 'DMT v9.01' seem
even better but i haven't tried that one just yet.
Typical SNR Margins are approximative as they vary
throughout the day, sometimes they're even totally
paradoxal. See for yourself by reading this post:
SpeedTouch 516 Tweaking :), Bicephale, 2007-Dec-11
According to the record shown above, it's possible
to register normal SNR Margins during a disruption
and then to read mediocre values once the noise is
gone! Of course, some people here will argue that
my GNet was faulty but my reply to them is that it
must be a self-healing GNet then! The same people
will brag around about the SpeedTouch which didn't
provide any "noise" curve at all until i asked the
author of 'STMT' to include this function; try it
if you have no alternative but be warned that i've
catched my ST resetting its counters unexpectedly.
Anyway, i suggest you keep in mind that trade-offs
are required while tweaking and that you shouldn't
rely on SNR Margins (static numbers) only: try to
gather stable feedback from a mix of RCO figures &
long term CRC Error rate curves as you toy around.
Now, here's what i mean by 3Com: "3Com USR-910x".
That's a MoDem brand/model and the D-Link DSL2320B
has an equivalent Spectral Shaping capability. My
underlying thought, i must confess, is that i'd be
more optimistic if you could use other HardWare!!!
Only three units are known to do Spectral Shaping,
only one of them got its tweaking abilities tested
(GNet's BB0060B) and this item was discontinued so
you probably won't find one of those in your local
electronics store. As i wrote before, practically
no interest could ever be raised relatively to the
Spectral Shaping topic and all we got to gather an
overall picture of this feature is my own numbers.
In other words, you may end being your own guru...
Congratulations, the average unsatisfied DSL users
would rather switch to cable than enter this quest
alone! Phone companies like to lower profiles, it
isn't profitable to send skilled technicians every
time a customer complains and 3rd-party ISPs don't
give orders to them: at best they open "tickets".
Normally, only you can pinpoint wiring problems or
evaluate devices in search for the ideal candidate
and there's no garantee you'll succeed in the end.
Phone companies like Bell would go bankrupt should
they compete with cable via pre-emptive techniques
such as systematic twisted-pair wiring, shielding,
RFI control by using ferrite toroids if necessary,
and, why not, on-site DSL unit tweaking (if it can
support it - which it doesn't, usually)! I may be
mistaking but it looks like Bell and others chosen
distributed filtering over existing phone cabling,
in hope that the subscribers would be satisfied to
finance the next fight (with help from ADSL2+ when
the time has come). I bet our forum didn't become
so popular because the phone guys guessed right...
Their plan left a few customers behind, welcome to
your new DSL neighbourhood! After a while it will
feel like amateur radio: the hobby makes us busy.
Well, i don't want to end my message on a negative
note so i added captures to boost your motivation:
Now, imagine if we knew the answer to this riddle:
 | |
Bicephale
join:2005-09-24
·TekSavvy Solutions..
·Velcom
edit:
May 9th, @10:08PM
| reply to dslgr
I don't want to minimize your problems, i just think
that the ADSL2+ protocol probably serves you right
under these circumstances. You mentioned being on
ADSL2+ Annex B, here's the illustration i could get:
You may have to use relatively expensive filtering
circuits to reject the ISDN band as it is closer to
the DSL band. Perhaps you should ask your provider
to do the tweaking for you: for example, request a
change to 9984 Kbps/512 Kbps if they don't mind...
Euh... But i read that you want to win on all bets
so that won't work for you i'm afraid! So, if no
trade-off is possible on the software side then you
are left with your phone wiring. Anything you can
improve? Still in the HardWare department, i have
this riddle for you. The answer might point at a
solution which many of us might also benefit from:
| |
dslgr
join:2008-05-07
| The illustration you presented is correct, with the addition that the downstream goes up to 2200kHz. Tone 32 (138kHz) to 64 (276kHz) for upstream and 64 to 512 (2200kHz) for downstream.
You are also correct about the fact that a trade-off (especially in the upstream rate) is not preferable and that one solution is expensive filtering circuits.
I've searched the internet extensively for a high quality filter that could cut the "noise" that my NT1 is injecting into my line in frequencies that it shouldn't, but unfortunately, although I tried a couple of splitters, I couldn't get one that does the job. My next move will be to try to construct one myself, when I find the time.
Changing that crappy device for a newer model would possibly make things better too, but that's not an option atm, so for now I've decided to live with it.
For your question now, because I am fairly new with the forum and I haven't been used to following lots of threads easily, could you provide me with some details regarding the graph?
From what I get, you are looking for a router's chipset that could provide a DS rate of around 2Mbps, for a 78db attenuated line? Where does that graph data come from? | |
Bicephale
join:2005-09-24
·TekSavvy Solutions..
·Velcom
| Ho DSLGr,
We're seeing more and more degradation cases where the
customer complains that his UpStream speed is lower than
it used to be. In my opinion, Cross-Talk would explain
this situation easily if we could accept that DSL has
gained popularity through time - which means Cross-Talk
levels must have increased a great deal during the last few
years. In my region, the phone company is using strudy
flat copper pairs to bring our DSL signal from the pole
to the house, that's where some progress can be made: i
showned it by turning a noisy & slightly unstable 4032/800
Kbps FastPath access into a quiet & stable 6016/512 Kbps
InterLeaved one using shielding and some RFI control...
The previous ADSL graph is for the 1st generation DSL
service, i believe ADSL2+ would have twice as many bins,
e.g. 512 bins instead of 256. By the way, did you try
to force your MoDem to negociate ADSL v1 only? That's
not going to solve your problem definitively but the
additional stability that it might bring would be one
more clue to guide you in the right direction... Sorry,
i don't see how to help you other than to mention that
TerTech is using a SpeedTouch with FirmWare v6.1.0.5.0
('ZZQOAA6.105') in combination with 'DMT' MoDemOptions.
The same person also shown me where his DSL device is
located on the graph by John Bullas (from Flickr site)
and i'm afraid that your speed might be even worst if
you were living around here, according to the curves;
DesiGuy1981, on the other hand, can expect up to 2M5
bps unless TerTech's unit is marginal. I wish i knew
what MoDem John Bullas tested to read 1M7 bps @ 78 dB
but i don't suppose he kept a record of that...
Have a nice day! | |
Bicephale
join:2005-09-24 | reply to dslgr
Euh...
Here's where i refered to John Bullas for the 1st time:
»ThoughtFood | |
dslgr
join:2008-05-07
edit:
May 11th, @02:30PM
| reply to Bicephale
It's true that increasing adsl popularity, increases crosstalk and therefore reduces rates.
Up to a point that is, since imo we can't blame crosstalk for everything (after all, a network should be optimized and should follow the standards, thus the rate hit by crosstalk should be under control and in the acceptable range).
But in my case, crosstalk is not an issue. Taking the NT1 out of the loop, gets my upstream high up to where it should have been in the first place.
Looking at the graph above, I guess my speed would be close to what I have here, if I lived in your area.
Now for that router and its settings that got 1,7Mbps with 78db attenuation, since we currently have no extra info, all I can say is that this is not a common thing to see.
As for DesiGuy1981, I agree with you, that he could get around 2Mbps, if all things worked at maximum efficiency. | |
Bicephale
join:2005-09-24
·TekSavvy Solutions..
·Velcom
edit:
May 11th, @04:07PM
| Hi,
Based on TerTech's own post in the other thread
i considered that the curve on our side of the
Atlantic may happen to be slighty lower and this
is why i devaluated your maximum attainable speed
should you live around here. There has to be room
for variance, in any case, so i guess that you'd
still be justified to hope for 10 Mbps. Perhaps
i got something to help you filter the ISDN band:
it's from a European guy, probably a neighbour...
Do you think 60 dB rejection @ 40 KHz would be OK?
If so, see the filter circuit shown in this post:
»Re: Is there life after the SpeedTouch?
The original .PDF document lies around, somewhere
in a post from Old Computer. His profile follows:
»/profile/615397 | |
Bicephale
join:2005-09-24 | reply to dslgr
Oh, i found it:
»Re: A wire is a wire is a wire
 | |
Bicephale
join:2005-09-24
·TekSavvy Solutions..
·Velcom
| reply to dslgr
Oups!
The graphic at the bottom of the .PDF document
shows that this is a low-pass filter capable of
rejecting most of the ISDN band and also beyond,
including your DSL signal!... Perhaps, if you're
very lucky, some manufacturer may have a band-
pass filter on his shelves that rejects ISDN only...
| |
Bicephale
join:2005-09-24 | reply to dslgr
Oups again!
It's a high-pass filter you'd need to cut ISDN noise... | |
dslgr
join:2008-05-07 | Thank you for the diagram. I will keep it for future reference when I find the time to start my filter project. | |
Bicephale
join:2005-09-24
·TekSavvy Solutions..
·Velcom
| I'm afraid a low-pass filter isn't the fix you had
in mind unless your house wiring is acting like an
A.M. radio antenna. The ISDN and DSL bands are so
close i doubt a filter made of analog components is
even possible, actually. The MoDem is responsible
for ISDN rejection: because of ISDN you lose all bins
below 37, i believe, and digital filtering does the
rest. If ISDN signals are a problem then the most
practical solution would be to try another device...
Godd luck in your quest towards a better service!
 | |
dslgr
join:2008-05-07
| Actually that's what I have in mind, a low pass filter that would allow ISDN signals to pass and cut off frequencies above, that could interfere with the adsl upstream.
I am using one (with analog components) at the moment, like everyone else with an adsl over isdn line, in order for the NT1 device (and with it the telephone) to work. All I need is one of better quality. Is it possible? I'll find out when I try it.
It's true that ISDN and DSL bands are close, although after looking once again at the illustration you posted above, I saw a difference that I hadn't spotted in the first place, which is the range of the ISDN signal.
Here in Greece we have 2B1Q ISDN and not 4B3T, which means the ISDN signal goes up to 80kHz and not 120kHz as the illustration shows. | |
Bicephale
join:2005-09-24
·TekSavvy Solutions..
·Velcom
| A low-pass-filter would be great for the ISDN device but
catastrophic for the DSL one, don't you think? If you
filter-out the DSL signal, the MoDem just won't sync...
Usually, the filter isn't installed ahead of the DSL unit:
it filters non-voice frequencies, ahead of ordinary phone
equipment. A filter meant to be installed in front of
your ISDN-compatible MoDem that filters out the ISDN part
without affecting the DSL part would be unselective and/or
introduce excessive distortion using analog components.
This situation calls for digital techniques and that's
exactly how the DSL MoDem works, hence my suggestion that
you simply need to find a more selective DSL device.
| |
dslgr
join:2008-05-07
edit:
May 14th, @08:24PM
| OK, I think we got confused here, so I hope the above diagram will clear things out.
As you can see at the setup, there is a different device for the adsl and a different device for the telephones.
Now, the ISDN splitter either consists of a high-pass filter (adsl output) and a low-pass filter (phone output) or just a low-pass filter (phone output) with the signal at the adsl output unfiltered, without having any problems at all (the router handles the situation as it should).
When I have my NT1 device connected into the loop, I get instantly a decrease of about 5db in my upstream SNRM, which means there is "noise" coming from that NT1 in frequencies that it shouldn't (in the adsl upstream range) and the low pass filter is not able to cut them off.
That's why I need to change my NT1 device, or get a better filter.
PS. My adsl annex b devices are a USR 9107 and a speedtouch 536i. | |
|
