Impact of hardware with bad SNR & Attenuation

I'm struggling to find comprehensive (or at least convincingly accurate)
information on what I'm sure must be a FAQ:

- What are the "good" values for SNR/attenuation for
different line speeds, and more importantly:
- what are the symptoms of borderline low SNR?
- what are the symptoms of borderline high
attenuation?
- what (in each case) if the upstream figures
are good and downstream bad?
- Vice versa?

- Regarding CRC errors:
- what causes them?
- what is a "good reading? (Obviously 0 is good
but at what figure should I worry about them?)
- what impact do they have?

- What impact does hardware play in making up for
borderline figures? Eg:
- having a crap filter?
- having crap cable?
- having a crap modem?
- having too much or crap telephone equipment
hung on the same telephone line?

- What does environmental conditions (eg weather)
affect:
- the SNR and attenuation readings?
- the impact of bad values?

To put the whole question another way: I want to have a reasonable idea if I
visit a customer with ADSL issues, when I look at the information the router
can tell me, what I should do. In which cases will changing the hardware
help (and in those cases what should I be changing).

 

sounds like you need a training course ;-)

6 dB of SNR margin is a typical minimum for reasonable service, 10 dB
or higher should be robust.

disconnections, lots of errors.

none, unless it results in low SNR margin

BT's 512k products will slow down the data rate if the upstream margin
is too low, should show up as a low upstream speed (fairly rare). 1M
and above will fail to sync if inadequate upstream SNR margin.

errored seconds (seconds with one or more errors) can be useful, 5 in
an hour isn't a problem, is. CRC errors arise when the signal is
corrupted by interference or by being too weak for the modem.

all of these things can reduce the SNR, attenuation is more robust.

thunderstorms generate noise that reduces SNR. Rain can affect damaged
cables.

if you are going in as a diagnostic expert I would suggest you carry a
"golden modem" known to work well on a battery powered laptop. Take
along a known good filter and an RJ11-BT adaptor too. Diagnosis by
changing / removing things seems the most practical.

The most common problem area IMO is the domestic extension phone
wiring. Use the test socket (where available) to separate this out
http://www.adslguide.org.uk/newsarchive.asp?item=2182

Phil

 

Is this for your course assignment?

Increased [bit] error rates, longer start up synchronisation, hiccups in
transmission.

The same, plus possibly periods without connection. Eventually resets and
possibly data loss.

Unlikely to be asynchronous, however the use of lower bit rates for upstream
is a common way of reducing the impact of this problem (e.g. 75/1200 and
ADSL)

Bits dropped out, high noise, interference, loss of signal, poor receiver
etc. etc.

Frames / packets with bad CRC are discarded. Missing data frames will be
noticed from the incorrect sequence and retransmission requested. This may
take some time. If supervisory frames are missed they may or may not have a
major impact depending on the protcol being used.

All of these can reduce signal and / or increase noise.

Water in cables increases noise and often reduces signal. Lightening can
cause spikes of interference.

Take a spare and swap it in. If the fault goes then the equipment was
faulty, if not then look to the wiring.

 

While I don't have any concrete information for higher speeds the
general guidelines (from BT) are as follows

Attenuation:

512kb I've seen working up at around 75dB downstream attenuation - BT
will activate anything on an enabled exchange and attempt to get it
working. If it doesn't work on long lines initially an appointment may
be required to fit a filtered faceplate or swap pairs for one with more
favourable routing, etc.

1Mbit <=60dB is the limit set by BT, but again I've seen this working
higher. It depends on the SNR that is obtained, more on that later.

2Mbit <= 43dB is BT's limit for this. Again, some lines can be
absolutely fine at 2Mbit at significantly higher attenuation, even
though it's a logarithmic scale.

In regards to the higher speeds I gather the attenuation limits will be
about the same (for ~4Mbit anyway), it will depend on your SNR on what
actual line speed will be stable. Naturally, the lower your attenuation
the harder it is for noise to impact on the signal.

The upstream attenuation should be less than 80% of the downstream
attenuation on non-faulty lines, for downstream attenuation above 20dB.
Typically it's about 60% of the numeric value in dB.

SNR Margin:

BT will classify a line to be possibly faulty when the upstream or
downstream signal-to-noise ratio margin is 5dB or less. However, many
modems won't cope very well when the SNR drops below around 8dB. This
borderline nature may cause intermittent loss of synch, loss of
connectivity (PPP drops), slow speeds, or all three. Essentially what
you need to know is even if a line is out of attenuation limits for a
service, doesn't mean it's the cause of the problem you're
investigating. If the upstream/downstream SNR is fine (say, above 10dB)
then there may well be another problem.

SNR will roughly drop by around 6dB per doubling of the line speed on
in-limits lines. So if you have 12dB on a 2Mbit line, you'd be very
lucky to have a stable 4Mbit circuit.

Even if the line doesn't have low SNR or high attenuation, it can still
be faulty so you can't always blame the modem. At PlusNet we find around
25% of faults to be caused by Customer Premesis Equipment (CPE), so
modems, wiring, filters, or sometimes even dodgy washing machines or
flatbed scanners generating radio frequency interference.

CRC (sometimes known as HEC) errors on ADSL are caused by an ATM packet
getting corrupted enroute - possibly due to the modem not being able to
correctly decypher the information from the noise.

You can essentially ignore these unless you're having problems of the
nature described in the section above. They'll be higher if the
connection is in constant use than if it's been idle, and you can't
really easily guage this. Extremely high counts (eg thousands per hour)
can cause problems of slow speeds and packet loss, due to
retransmissions having to be requested.

This is the wrong way of asking this question - the hardware you're
using can be *causing* the borderline figures reported by the modem. Or
the modem may be misreporting the statistics - they can vary
significantly between modems. The type of filter and wiring you use can
mean a difference in the SNR you get, the attenuation is generally not
something you can change from within the premesis because it largely
depends on the line's length from the exchange.

Faulty or unfiltered equipment sharing the same line can have an impact
on the SNR or cause the ADSL to not work at all. The easiest way of
eliminating internal wiring and devices entirely is to just unscrew the
master phone socket's faceplate and plug your ADSL modem into the test
socket behind that to see how the figures vary. In some cases this will
cause your SNR to be quite a lot higher, as internal wiring can pick up
all sorts of electrical noise and feed it back to the master socket,
affecting the ADSL signal.

It's then a case of either finding out what bit of wiring is causing the
problem (might be a loop, faulty bell wiring, polarity, second master as
an extension, DECT phone on an extension...), or fitting a filtered NTE5
faceplate with dedicated CAT5 cabling laid to the place you want the
ADSL modem to be fitted.

As mentioned above, the attenuation shouldn't really change much with
the weather conditions. SNR will vary according to weather, time of day,
phases of the moon, you name it. Thunderstorms are the thing that affect
it the most significantly, as you might expect from the electrical noise
generated. This will be all the more significant on longer lines - some
people find they can't obtain a reliable ADSL service during a storm.

A leaky junction box along the line's route could mean the ADSL drops
out in wet weather and miraculously recovers when it dries out.
Similarly, SNR can be affected by the time of day - the neighbours may
have an outside light that only comes on at certain times, halving your
SNR for instance. A TV transformer or central heating pump may be
faulty, giving the same effect. When we get customers with time-related
problems we try to identify anything in their premesis or nearby (e.g
street lighting) that may come on at the time the problem is
experienced.

As eluded to above, you can't always tell if it's the modem's fault,
however you can try to eliminate that possibility. Trying the connection
in the master socket (ideally the test socket), and changing the filters
are the first steps you'd be required to do before you could report this
to your ISP (or your customer's ISP rather), who can run some further
tests and report the fault to BT.

With sporadic connection drops on good quality lines, if no fault is
found initially, the first BT engineer will most likely turn up, confirm
the attenuation and SNR with their own hardware, make sure everything at
the exchange still checks out, and then claim the line is ok and ask you
to test with a different modem, so it's probably best to start with this
after doing the wiring checks.

Hope that helps.

Regards,

 

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

<snip>

Interesting read - thanks!


- --
Adam Piggott, Proprietor, Proactive Services (Computing).
http://www.proactiveservices.co.uk/

Please replace dot invalid with dot uk to email me.
Apply personally for PGP public key.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.0 (MingW32)

iD8DBQFDFbHo7uRVdtPsXDkRAqrfAJ9x7B/kVC8UECwVsUUJVJOWLb0InQCeJ1b4
aQ1+4WKYian4blNbZGxPcPY=
=u+m4
-----END PGP SIGNATURE-----

 

Crap filters can be very bad.
For example, changing from a 99p ebuyer one to a faceplate filter from
http://www.solwise.co.uk/ added some 10dB to my SNR margin, when
other devices (cordless phones) were plugged in.

 

Interesting what you say Alex about BT and their processes. Friend of mine
living locally was activated but he could not receive any adsl signal. He
reported the fault to his ISP. BT then apparently did the check without
going to the house and according to the BT W checker this number is now
showing as not capable of receiving broadband at any speed following an
engineer visit!.He is about 5km from exchange and others further down the
cable seem to be getting 512 ok. BT dont seem to have made any attempt to
switch pairs. They also dont seem to have told his ISP yet!
Alan

 

Hi there,

While I couldn't comment on this specific case it does really depend on
the state of play in regards to the line itself. BT's data may suggest
the line is routed in a specific way meaning there's no point sending an
engineer out, or it's TPON, or various other possibilities. I'd suggest
you get your friend to find out from his ISP why they can't supply ADSL
- the ISP if necessary will be able to call BTw and obtain further
information. It's entirely possible that BT have made a mistake and
misclassified the line.

Regards,

 

You forgot the infamous Xmas tree lights/rope lights, I've also known
electrical relays to cause dropouts, one case in particular all the ADSL
circuits in the building dropped when the relay controlling the carpark
lighting switched over. In the end they had to live without the lights
as we couldn't pin it down to a voltage spike or RF spike causing the
problems & the end user wasn't incline to investigate the car park
wiring & the customer was more worried about the broad band circuits
than having some halogen spot lamps on ....

 

Hi Alex,
Thanks for the response. Would you mind if I asked my friend to contact you
direct as he is with your ISP and this matter has been dragging on since the
6 Aug when our lines locally were activated. He is getting increasingly
frustrated with the lack of action from BT.

Alan

 

Hi Alan,

He can contact me, but I'll be brutally honest, I'm only working for
PlusNet for one more day so I'm not sure I'll get a chance to chase this
up, especially if the fault is with BT High Level Complaints. He may be
best off calling our support guys or using the contact us system on the
website if he doesn't hear anything from me directly.

Regards,

 

Thanks Alex. I will suggest that he tries Support again though they do not
seem to getting very far with BT.
Regards,

Alan

 

When I had 9-14db downstream noise margin I found that the line kept dropping
particulary about 9-10pm - most popular time for internet use!

with 22-23db noise margin my line is completely stable. (512k line about 60db
attentuation, only connected when BT disontinued the line loss limit last
Sept).)

I obtained a stable connection my connecting my router to the master socket and
connecting all the household wiring and extensions via a filter.

Trees falling on local overhead phone lines have spoilt my connection! I have
lost a 56K modem to lightning.

If you can not get a stable connection by the above self help measure (i.e. if
the connection fails more than 1 per week.) I would suggest complaining to
your ISP, and asking them to get BT to investigate.

 

Phil Thompson wrote...

It wouldn't be a bad idea! Any such things online? Doubt I'll have the
budget for a professional course, since it's mostly for my own interest.

So the ~3dB I was getting isn't much cop then

[I already knew that, of-course, but it was the investiagtions surrounding
that which made me realise how little I knew and prompted the questions
here.]

I'm certainly getting the impression that attenuation is pretty unimportant
on its own.

This will show in the data rates the modem presents to me, I assume?

So 1M won't downgrade - that's useful to know. Borderline 512 = slow,
borderline 1M = nowt at all.

Can you expand on that? (I can't make sense of the sentence.)

We've consistently had problems during bad weather (ie heavy rain). The line
has not really degraded for calls as far as I can tell, though. The problems
started when we moved offices (in the same building) and went from one cable
to another (the current one being an old weathered cable running around the
outside of the buildig which tests fine when BT test it).

I thought as much, but I'm not sure I really know how to determine which
devices I have (or get) are "good" other than by spending a bit more on them
than usual - eBuyer microfilters are out If I plug different hardware
into the same line and look at the SNR or other figures, wil that help me
pick the good hardware, or will the readings I get from the hardware not be
good enough for that? I guess this should work for the filter at least, but
if I compare two modems which give different readings does that mean the one
with the better readings is a better modem?

For test purposes, are ethernet or USB modems best?

Thanks for the help information - all very useful stuff.

 

R. Mark Clayton wrote...

No, but looking at it now I can see why you'd think so. Been around a bit
too long for that to be the case, though.

Lots of useful stuff snipped - but all read and digested and I'm very
grateful for it.

I've been doing that a lot recently, but without really knowing which
hardware is "good" (ie it all worked in the past) I feel like I'm just
guessing. The information in this thread is therefore invaluable.

 

BT have just fitted a couple of faceplates here - we'll have to see how much
difference that makes. To be fair the filters were cheap ones (part of the
reason for asking my original question was to find out which things the
filter will mess up).

On a line with no phone, where the filter is being used only as a way to
convert from RJ11 to BT plug will there be any difference between a cheap
(crap) filter and a good one? Ie does unplugging all other telephones act as
a good test as to whether the filter is the problem?

Reading between the lines here, a good modem is better at dealing with poor
SNR than a cheap one. What's a good modem to go for in these cases? How much
difference do they make - eg do they typically cope with 1-2dB less SNR than
a poor modem, or more/less?

If the problems are with CPE (other than the mode/wiring/etc), what are the
best solutions? Eg if the problem is a washing machine or scanner, is it
proximity to the device that matters, or is it electrical noise? Some
hardware better than others at coping with the noise? Do things like surge
protectors or UPSs help?

I had a situation recetly where we could get sync but no IP address
(including from [email protected]_domain). The error count was increasing by
about 1 or 2 per second (which given the lack of traffic without even having
an IP I guess is pretty high). The SNR was very bad though (<3dB).

If I have sync but no IP, what does this mean? For that matter, what does
sync actually mean? What do all the "ACTIVATION" to "SHOWTIME" actually
statuses mean?

I tend to assume that there's crap hardware which causes (or at least
exacerbates) the problem, standard hardware that does works fine in normal
situations and really good hardware that may work when bad line quality
means that normal kit can't cope. Is that fair? If so, which hardware fits
into which bracket - is it simply a case of cheap is crap, expensive is
good?

I almost always locate the modem (ie modem/router) next to the phone socket
and CAT5 from there to where the network is needed - I assume that this is
preferable to extending the cable on the ADSL side?

Good advice. Are there any other good tests to perform before calling on BT
(even before calling on the ISP) that help to say "I have tested this and
it's almost certainly a BT/ISP problem"? So far I normally:
- Swap routers and filters, and test each without anything in the phone
socket
- Make notes of SNR/attenuation in each case
- Try [email protected]_domain login

Again - thanks for all the help.

I'm going to collect all the information in this thread into a single
document for internal use - does anyone have any objection if I put the
results online somewhere for everyone-else's benefit (properly attributed,
of-course)?

Mark Rogers,
More Solutions Ltd

 

I'm getting a good feel for the impact of SNR/attenuation/etc.

What would be really interesting (to me anyway) would be to hear what really
bad ADSL conditions people are managing to make work for them, and with what
hardware?

What I'm thinking of is low SNR and/or high attenuation, or long distance
from exchange, etc, and ADSL speeds you get with what hardware (modem
make/model, filter make/model).

Does a similar survey exist somewhere already?

For kickoff, my two lines (which come into my building on the same BT cable)
are:

Line 1:
55.1/31.5dB Attenuation
25.9/29.0dB SNR
1152/288 kbps
Safecom (eBuyer) router with BT faceplate filter

Line 2:
61.4/31.5dB Attenuation
22.1/25.0dB SNR
576/288 kbps
Billion (Solwise) router with BT faceplate filter


Line 2 was showing around 3dB SNR until the cheap filter was swapped for the
BT faceplate. (It wasn't working at all well, of-course, hence the BT
visit.)

Even now it's showing a lot of errors (7277 error seconds in around 12 hrs
since BT man left), and since its really just a spare line it hardly sees
any data through it. So it probably doesn't qualify as reliable.

 

Could be a good place to start, old cracked cables, allowing the ingress
of water, may test ok but doens't mean to say that it's any good for
ADSL, especially if they are low resistance wire to wire
I found that despite all the problems I had with my Netgear DG834G the
diagnostics page tallied quite closely with an APTS test result. So if
you're not a BT person with the correct software & modem (Voyager 105) I
would suggest you kit yourself out with a Netgear & a supply of known
good filters, including a few of the SSFP type filters BT fit, see if
you can get some RF3 filters as well as they can occaisionally do
wonders, but only occaisionally.

I've experimented with various routers (had a hardware problem brought
on by my usage, shhhhhh P2P) & found that each make gave wildly
different test results, some being that far out if the reading were true
the line wasn't working, which leads back to my recommendation above...

Unless you've got the correct software re above, use a router it will
contain it's own diagnostic software

Once you've made your decision stick with the one modem/router, get to
know how it feels/works. I can often get a feel that the APT test will
give good or bad result just by watching how my Voyager 105 synch's up.

Hope you don't mind me jumping in with my ha'ppence worth. whatever
router you do go with (the USB route as far as I'm concerned is a no no)
stick with the one, learn it's foibles & you won't go far wrong...

Best of luck (trying to put me out of a job, huh)

 

The faceplate filters help stop the internal electronic mush, picked up
by the bell wire of the extensions feeding back & mixing with the ADSL
signal, via the the master socket

Depends on how many/much extension wiring is attached, ref the above
bullet point.
In the field the engineers try to get a +10dB as in a lot of situations
the SNR will vary & normally will drop in the evening, so 10dB does give
you a bit of lee way..

Not neccesarily, depends a lot on how much you want to spend (a really
good one can be £300 or there abouts) & after that you then have to get
the correct long line firmware installed as well.

Nope none of the above will help, what may (repeat may) help is fitting
a RF3 filter in line before the master socket, but even that isn't
guaranteed, depending on high big the RF spike is...
Speaking from personal experience you can cut down on excessive CRC (or
even HEC) errors by using the ADSL circuit in the master socket via a
SSFP, or even just disconnecting all the extensions connected to that
line..

Showtime is when the circuit is synch'd up, Activation I must be honest
I cannot recall ever seeing (but it could mean that the monem/router is
attempting to negotiate a connection with the exchange)..

All of the above can & will cause problems, which is why it's always
best to start from the test socket on the NTE5 (if you haven't got one
then disconnect all your extensions) with a known, good, working filter

Not neccesarily as different users in different situations can get
different performance from the same piece of kit, even the expensive kit
can be a lemon...

You can run anything upto 30 metres of cat3/5 cable from the NTE
faceplate filter to where ever you place your router/modem, but doing it
your way can be simpler, easier & it does rule out one possible
problem...

Hopefully they would also test the line (from the exchange), for
capacitance balance as well as the normal battery, earth, loop & low
resistance test. Also they would check the exchange equipment under
load, using artifical line extenders (plug in dongles), personally I
always check with 80dB but some only use 60dB, which is no matter as
either way it is putting a loading onto the DSLAM which could prove a
problem into the DSLAM card you are connected to.

DO NOT call BT for any ADSL problems. All ADSL problems have to be fed
via your ISP (yes even if it's BT Yahoo/Broadband/Openworld or
whatever). You run the very real risk or getting visit charges raised
against you & little else if you do. Raise it with the ISP, that way
you will get an ADSL engineer who will be able to test your ADSL service
& hopefully either fix the problem if it's on the BT network side or
give you some advice if it's on your private wiring

 

Just because the 2 lines enter your building on the same cable doesn't
mean that they follow the same route back to the exchange. They almost
certainly do to the PCP, but from there they could be on completely
different cables, with different routes & compositions back to the
exchange. This could account for the different test results you have
given.

I would almost certainly say that the first circuit has a far better
path back to the exchange than the second, but having noticed that you
are using 2 different routers you can't really compare the results, what
you need to do is get the results using the same router on both lines in
turn & then see how they compare (the other way is to mug a BT ADSL
engineer & get him to test your lines with his test equipment, but of
course I can't condome this, for a start you've got to find one first,
then you've got to catch him as he's driving hell for leather to his
next job 50 miles away, yes it is like that around here)