Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[aesmith]

Hi,

Just a little funny.  The router had been up for around 24 hours prior to this, and was synched at 4416K with noise margin varying between 3.5 and 4.5dB, error rate was quite high.  I did a graceful disconnect followed by reboot, expecting to reconnect at a lower speed.  Instead what happened was it connected at 6dB noise margin as you'd expect, but exactly the same speed of 4416K which was definitely not expected.  Error rate dropped right off.   What do you think happened there?  Was the router screwed up in some way?    A Thomson 582N by the way, on 20CN Max.

Tony S

[Weaver]

This has happened to me lots of times, although as far as I've noticed there's always a loss of 1 notch=32kbps of downstream sync speed. With me error rate was possibly high reflected in the fact that that there was enormous packet loss, that is, IP packet loss and loss of the error checking PPP LCP echo request probe packets that my ISP, Andrews and Arnold, sends continuously to monitor link quality. (See
    http://aa.net.uk/kb-broadband-cqm.html ).

The huge packet loss means that streaming videos won't play at all, Internet speed tester tools give horrific figures (down by ~90-95%). I've ruled out a DDoS attack as a reason for slowness by getting a packet capture from the link. Also the SNRM doesn't drop through the floor.

Whatever it is, I have just assumed it is the modem, a DLink DSL-320B-Z1, getting its knickers in a twist in some way, or the line going really bad, or else the DSLAM being weird. But whatever with me there is always some very serious wrongness that is real.

I don't know if this is similar enough to your scenario, the packet thing (do a speed test) would be one way of looking into it.

[Chrysalis]

bit swapping will disable weak tones which decreases overall snrm but a resync will recover those tones

[ejs]

On ADSL2/2+ at least, the 582n with 10.2 firmware is capable of re-allocating bits to tones that get reduced to having zero bits allocated to them. Some older modems can't do that, but I don't know if it's possible with ADSL1.

[Weaver]

That would seem like a very poor design decision, not allowing bitswap to allocate zero bits to a bin, unless the standards themselves forbid it, but in that case it would seem like a design weakness in the standard. I can't understand why you would want to remain stuck with a useless frequency.

[ejs]

Sorry, I meant that all modems can bitswap a tone down to zero bits. Some modems will then have to leave that tone at zero bits, even if conditions later improve.

[Chrysalis]

Thats what used to happen to me on adsl, the higher tones would be all but wiped out during the first night, but the next day they stayed turned off and for the rest of the lifetime of the sync.  Thats the basis of my above reply, I seen the same behaviour on different dslam's with different modem's.

[ejs]

11.2.4 Extended bit swap request
Any on-line adaptation may be encoded in an extended bit swap request. However, because a single-
bit subcarrier is not allowed, an extended bit swap request containing 6 fields shall be used when
decreasing the number of bits on a subcarrier from 2 to 0, or when increasing the number of bits on a
subcarrier from 0 to 2.

The ADSL1 document seems to allow for bit swapping from 0 to 2. aesmith should be able to get the bits per tone data from the 582n and see if it ever actually happens.
The on-line reconfiguration messages can also adjust the power transmitted on a particular tone, although of course it would have to remain within the limit for the maximum power per tone and the maximum overall power.

[aesmith]

Looking at the saved graphs I can't see an example where a tone that's been zeroed coming back into use.