I found out recently that if I power cycle my ADSL modem
that it takes hours to re-establish the connection. I
rarely power-cycle (the modem is on a UPS) so this
problem may have be present for a while. However once
the connection is made it stays up.

The phone side of things works perfectly.

I'd appeaciate advice on whether this is an ISP problem,
a line problem or if my modem is about to die.

The modems log file repeats the same log entries but I've
no idea what the numbers mean:
(newest log entry at the top)

01/01/1970 04:58:20> PPP1 Session is down.
01/01/1970 04:58:20> ADSL disconnected
01/01/1970 04:58:20> ATM Disconnected
01/01/1970 04:58:16> PPP1 PPPoA Connected
01/01/1970 04:58:16> ATM Tx Error Count: 0, Rx Error Count: 0
01/01/1970 04:58:16> u/s total bit capacity: 95bits, used: 95bits
01/01/1970 04:58:16> d/s total bit capacity: 398bits, used: 152bits
01/01/1970 04:58:16> d/s Latency: FAST, u/s Latency: FAST
01/01/1970 04:58:16> d/s Data Rate: 576kbps, u/s Data Rate: 288kbps
01/01/1970 04:58:16> d/s Attenuation: 56.4dB, u/s Attenuation: 31.5dB
01/01/1970 04:58:16> d/s SNR: 4.9dB, u/s SNR: 19.0dB
01/01/1970 04:58:16> CO Vendor: TEXAS_INSTRUMENTS
01/01/1970 04:58:16> Max Tx Power: -38dBm/Hz
01/01/1970 04:58:16> Annex Mode: ANNEX_A
01/01/1970 04:58:16> Modulation: G.dmt
01/01/1970 04:58:16> ADSL connected
01/01/1970 04:58:16> ATM layer is up, cell delineation achieved
01/01/1970 04:58:16> ATM Connected
01/01/1970 04:58:11> ADSL detected

This repeats on a regular basis but the numbers change each time.
These are some randomly chosen lines from the log that shown
the numbers that change:

01/01/1970 04:58:01> d/s total bit capacity: 403bits, used: 152bits
01/01/1970 00:06:37> d/s total bit capacity: 399bits, used: 152bits
01/01/1970 00:04:08> d/s total bit capacity: 389bits, used: 152bits
01/01/1970 00:02:37> d/s total bit capacity: 384bits, used: 152bits
01/01/1970 00:01:46> d/s total bit capacity: 407bits, used: 152bits

01/01/1970 04:58:01> d/s Attenuation: 56.7dB, u/s Attenuation: 31.5dB
01/01/1970 00:10:13> d/s Attenuation: 55.8dB, u/s Attenuation: 31.5dB
01/01/1970 00:09:47> d/s Attenuation: 57.3dB, u/s Attenuation: 31.5dB
01/01/1970 00:09:02> d/s Attenuation: 56.8dB, u/s Attenuation: 31.5dB
01/01/1970 00:06:07> d/s Attenuation: 57.1dB, u/s Attenuation: 31.5dB

01/01/1970 00:13:36> d/s SNR: 5.4dB, u/s SNR: 19.0dB
01/01/1970 00:12:35> d/s SNR: 4.9dB, u/s SNR: 20.0dB
01/01/1970 00:11:13> d/s SNR: 4.6dB, u/s SNR: 19.0dB
01/01/1970 00:09:17> d/s SNR: 5.8dB, u/s SNR: 20.0dB
01/01/1970 00:07:37> d/s SNR: 4.3dB, u/s SNR: 20.0dB

When it finally gets a connection the successful log entry is:

08/09/2007 07:46:09> Received time from Time Server 128.138.140.44
01/01/1970 04:33:51> PPP1 Session is up.
01/01/1970 04:33:51> DNS: Add IP address 205.188.146.145 (Auto
discovered)
01/01/1970 04:33:50> Firewall received interface change notice.
01/01/1970 04:33:50> No Static Session Information is defined.
01/01/1970 04:33:50> Initialized NAT Virtual Servers.
01/01/1970 04:33:50> NAT/NAPT Session Start: interface ppp1, WAN IP is
172.189.109.192
01/01/1970 04:33:50> PPP1: DNS Primary IP address is 205.188.146.145
01/01/1970 04:33:50> PPP1: PPP Gateway IP address is 10.64.96.7
01/01/1970 04:33:50> PPP1: PPP IP address is 172.189.109.192
01/01/1970 04:33:50> PPP1 CHAP Authentication success
01/01/1970 04:33:34> PPP1 PPPoA Connected
01/01/1970 04:33:34> ATM Tx Error Count: 0, Rx Error Count: 0
01/01/1970 04:33:34> u/s total bit capacity: 95bits, used: 95bits
01/01/1970 04:33:34> d/s total bit capacity: 560bits, used: 152bits
01/01/1970 04:33:34> d/s Latency: FAST, u/s Latency: FAST
01/01/1970 04:33:34> d/s Data Rate: 576kbps, u/s Data Rate: 288kbps
01/01/1970 04:33:34> d/s Attenuation: 57.4dB, u/s Attenuation: 31.5dB
01/01/1970 04:33:34> d/s SNR: 14.1dB, u/s SNR: 20.0dB
01/01/1970 04:33:34> CO Vendor: TEXAS_INSTRUMENTS
01/01/1970 04:33:34> Max Tx Power: -38dBm/Hz
01/01/1970 04:33:34> Annex Mode: ANNEX_A
01/01/1970 04:33:34> Modulation: G.dmt
01/01/1970 04:33:34> ADSL connected
01/01/1970 04:33:34> ATM layer is up, cell delineation achieved
01/01/1970 04:33:34> ATM Connected
01/01/1970 04:33:27> ADSL detected

Advice please.
Nick.

On Thu, 09 Aug 2007 19:00:39 +0100, Nick Austin
<see_reply_to_in_headers@nospam.com> wrote:

You have a poor line. 4.9 for a SNR margin is close to dropping and
that is only giving you a mere 576k downstream speed.


Jim.

On 9-Aug-2007, James Egan <jegan@jegan.com> wrote:

For some routers you can get a poor line software flash.
In the old days modems were equalised with knobs for
phase and gain, while watching the eye pattern. Now the
knobs are twiddled electronically, but some modems have
a wider range of adjustment, and cope better with bad lines than
others.

On Thu, 09 Aug 2007 21:04:22 +0100, James Egan <jegan@jegan.com>
wrote:

OK, so why does the SNR jump from 5 when trying to connect
to around 14 when it is connected. It's done this to me
twice during the last 24 hours and it would be extremely
co-indental that I happened to power cycle just as the
quality of the line dropped. I've not yet been kicked
off, the problem always appears just after a power
cycle.

The status web page is currently showing between 13.9 and
14.1 and has been every time I looked at it recently.

The reason I only get 512/256 is because that was the deal I
originally signed up to. I'm not due to be forced upgraded
until October.

Nick.

On 9-Aug-2007, Nick Austin <see_reply_to_in_headers@nospam.com> wrote:

I'd speculate that being a long line the signal is weak and
that during the training period the gain is turned full up, adding
some additional noise and lowering the S/N ratio. Then it
starts adjusting gain and phase to get the best eye pattern.
The eye pattern being a measure of the errors, and correlates
with the error rate. Equalisation takes time, but eventually
the line is compensated for, the gain reduced in stages, and
the resultant error rate the best that can be achieved. As the
gain is reduced and the line characteristics compensated
for the S/N ratio improves.
Google for "eye pattern"
In the old days equalisation was done manually and
you could see what the line was like, and the effect of
adjusting the equalisation gain and phase knobs.
If you power cycle then the whole equalisation process
starts from square one, I don't thing designers have
included a start from the last settings that worked.
My own router shows a spectral analysis histogram, and
it takes a while to stabilise to it's normal pattern
after a power cycle.
Some routers can be flashed with long and difficult line
software.

On Thu, 9 Aug 2007 23:17:27 GMT, ato_zee@hotmail.com wrote:

I think I see.

It's a Safecom SAMR 4114 that was flashed two years ago with the
firmware from here:
http://safecom.cn/code/product/adsl/samr-4114/fm.htm

As far as I can tell there isn't any alternative firmware so I
guess I need a new ADSL router. Care to recommend one? I only
use one ethernet port (which connects to a hub) and require no
special features except NAT and a small number of port forwarding
rules.

Nick.

On 10-Aug-2007, Nick Austin <see_reply_to_in_headers@nospam.com> wrote:

Some of the Draytek's come with long and difficult line UK2
software.
------I quote
The special 'UK2' version of the firmware is only for use where the
standard firmware cannot sync. reliably on your ADSL line, for example
on very long line with high attenuation. Use the standard firmware
otherwise.
------Unquote
Draytek has a users forum, and you don't need a Draytek to register, plus
they only ask for minimal information to register.
Users experiences of the UK2 long line software is discussed in the
forums, and software flashes are available for the older routers.
As you won't get ADSLmax on a poor line you might find one
of the early Vigor 2600 routers cheap on eBay, a good source
for stuff people are upgrading from. Recently saw one go for
a fiver plus postage.
The 2600 has a 4-port RJ45 capability, you could add wireless
if needed with say a Netgear ME102 or a more recent access point.
The 2600's seem to uTorrent port forward ok with factory defaults,
also connect VPN outgoing out of the box. Haven't tried setting up
VPN inbound on these. Shields Up says attached PC's are
(unusually) completely invisible, no response to ICMP echo
requests, and invisible to port scans.

There are other quality routers out there but I'm more familiar
with the Draytek's.

One of the problems with long lines is crosstalk between
other pairs on the cable, so on a long line your signal is
weak by the time it reaches the exchange, whilst picking
up competeing stronger signals in adjacent pairs being
generated at the exchange end. And vice versa.
All that modem designers can do is make their front ends as
low noise as possible, and to provide a wide a range of
compensation for line characteristics. Which is
what I guess the UK2 flash tries to achieve. Then all
that can be done is drop down in speed until the error
rate is acceptable.

On Fri, 10 Aug 2007 18:51:10 GMT, ato_zee@hotmail.com wrote:

Thanks for the suggestion, both to you and to Kris.

Nick.