 1 edit | reply to gmsron
Re: Need help with gnet bb0060a Hi,
I wish i had nice graphics to explain what i guess it does, instead i only got this late test sample:
DMT - US/DS frequency bands/Tx Bin Adjust: [ ] TelNet - modify dsl config txbinadj enable DMT - Upstream (TX) Bins: 6 - 13 TelNet - modify dsl config rxstart 6 TelNet - modify dsl config rxend 13 -------------------------------------------------- DMT - US/DS frequency bands/Rx Bin Adjust: [ ] TelNet - modify dsl config rxbinadj enable DMT - Downstream (RX) Bins: 14 - 248 TelNet - modify dsl config rxstart 14 TelNet - modify dsl config rxend 248 -------------------------------------------------- DMT - ec fdm mode: FDM TelNet - modify dsl config ecfdmmode fdm -------------------------------------------------- DMT - Tx Power Atten.: -1 dB TelNet - modify dsl config txatten 1 WEB - Local Tx Power: 6.36 dB (11.95) WEB - Remote Line Atten.: 27.0 dB (30) WEB - LS0(kbps)-Up/Fast: 256 (800) WEB - Remote SNR Margin: 8.0 dB -------------------------------------------------- DMT - Coding Gain: 7 dB TelNet - modify dsl config gain 7 WEB - Local Line Atten.: 30.0 dB WEB - AS0(kbps)-Down/Fast: 4032 WEB - Local SNR Margin: 23.5 dB (18-21.5)
If you try tweaks based on the CLI TelNet commands instead, type "commit" then "reboot" in the end...
The DMT part was tested before i checked my TelNet one, today's numbers varry somewhat (19.5 -> 22 dB Local SNR Margin improvement, starting with a non- optimized 'commitedcfg.cfg' file) then my CLI fix.

My understanding is that i reshaped the juxtaposed UpStream (1-70) and DownStream (10-255) bandwidths of the DSL signal so that it will suit my specific phone-line conditions better. It really helped me a lot to use the 'DMT' utility while searching for my own sweet spot; i'm sure there could have been a few more variations to try but it already raises my Local SNR Margin by 2.5 dB and lowers my Remote Line attenuation level by 3 dB, with no noticeable effect on the other parameters. Well, i'm puzzled quite a bit, actually, because my TelNet test gets a nice 4032/800 Kbps profile when it was 4032/256!

Hummm... Perhaps this is where the 1.5 dB SNR gap between 23.5 dB (DMT) and 22 dB (CLI) has gone!...
I think i'll need to investigate this, eventually!

Addendum
Here's a nice graphic:

 |
|
 1 edit | Oh... I can see it now, typos around "tx" and "rx"!...
'DMT' can read back the resulting values and i produced an overlapped configuration! I didn't think i could...
 |
|
 | reply to Bicephale Hmm...

»img259.imageshack.us/img259/1153/dmt8bm.jpg
I'll test those settings out in a bit |
|
 1 edit | reply to Bicephale Hi,
According to the graphic above you need to click on the [On] button of the "automatic DSL-monitoring" field and to check "activate Auto-DSL-Monitoring on startup" too.
Keep in mind that numbers which are right over here may not be optimal for the kind of problem you're facing...
I'd say that Tx = 1 to 32 and Rx = 40 to 255 happens to be the default configuration. You can try to sacrifice some UpStream bandwidth to improve the DownStream side.
 |
|
 | Tried out the settings you had in the test. I'm not disconnecting, but the download speed is very very slow. The upload is also really slow. I'm not sure what settings affect what. So, I'm not sure how I would even sacrafice upstream for downstream.
»img407.imageshack.us/img407/7277/dmt1jz.jpg |
|
 | Hi,
I guess the UpStream signal still has some room at -1 dB, i'd try -2 or even -3 dB but the SNR figure associated to it must remain above 6 dB. UpStream speed can be made lower (i use 288 Kbps right now) and, euh... Oh my God! 7808 KBps! I'm sure that isn't your true DownLoad speed!... Well, it would still be 4032 Kbps if you use bins #36 to #248 but this device reacts slightly differently than mine.

If only 7808 KBps was real! I tried other numbers for the "max. Bits/Bin" field in the "Limitations" section but 14 was about right for me, i'd give it a try once you found a sweeter spot. I'd say that a practical limit for the DownStream bins would be around #70 and #192 for the lower and upper edges, respectively. I'll explain. Here's what i gather from my tweak sessions so far: your "start" field (the one associated to the DownStream line) can go as low as 14 but as high as 70, if i'm correct. A lower Start Bin brings your two bandwidths closer, which exposes the DownStream signal to distortions associated to the UpStream one; hence causing the MoDem to disconnect more frequently - should noise be your actual problem. Raising the Start Bin can create enough void between the bandwidths to bring the error rate down but you have to know that this also affects the DownStream SNR figure because the function between the number of bins and SNR is not linear (it's probably closer to an asymetric bell- shaped curve instead)... In short, the DownStream bandwidth should have some clearance from UpStream noise but not too much because the SNR will suffer past a given point - SNR seems to depend on width, the larger the better but too much of it and noise from the UpStream band would be too close. As for the DownStream End Bin, this one's behaviour seems much easier to grasp: attenuation goes up rapidly as a function of frequency and a point comes where noise is predominant; the lower this edge is kept the less attenuation (hence noise) you get - but a lower End Bin would mean less bandwidth and past a sweet spot the SNR goes down again. As i wrote, i would have needed a graphic illustration for this, i hope my long explanations didn't discourage you!

Finally, if you're goal is to obtain a stable link and peer-to-peer performance, euh... i'm afraid i could only buy you some time until a better fix to your specific situation can be found. These tweak trics only have marginal effects on signal quality in the end, we may be talking about 2-3 dB here...
 |
|
 | reply to Bicephale I give up... I tried a bunch of different settings, but the download speed would always be very low and if it were at even half my regular speeds...then it disconnects. Tweaking modem settings isn't for me. Thanks for all the help though. |
|