[Asterisk-Users] Analog FXO Woes Continue

[Jim Van Meggelen]

asterisk-users-bounces at lists.digium.com wrote:
>> I've been struggling with a test * install for a couple months now in
>> a small office and am just about ready to give up on it.  It's not
>> that the system itself is a problem.  I've got everything (attendant,
>> voicemail, FXS extensions, Cisco and Polycom hard-IP phones, and 2
>> VOIP carriers) working except for the frigging analog FXO interfaces.
>> These things are driving me completely mad.  Since this is obvioiusly
>> a deal breaker, I'm looking for any more suggestions on how I might
>> fet these things working. 
>> 
>> The hitch is pretty clearly the quality of the lines I have from
>> BellSouth but I can't get thim to identify anything wrong.  I have
>> tried a Digium 1-port FXO card (can't remember part number and it's
>> no longer on  their site, hmmm...) as well as a Sipura SPA3000.  With
>> both of these interfaces, I'm getting consistent mis-dials on
>> outbound calls, broken inbound fax-detection, broken DTMF detection
>> in the attendant menus. Hours of adjustments to the gains on the
>> Digium card only added echo and failed to reduce the offurenc of the
>> other issues. These same two interfaces worked fine on a line at my
>> office so I'm pretty sure the issue is with the lines at the test
>> site. 
>> 
>> So, what are my options here for interfacing with these lines?  Would
>> the channel-bank route affect this?
>> 
>> Thanks in advance for any suggestions,
> 
> Don't have any real answers, but might check the following...
> at least to rule them out.
> 
> Telco folks _always_ check lines from their demarc (which in
> some cases is the protector box on the outside of the
> building). Most will not come inside to measure anything from
> the customer equipment jack. If that's true in your case,
> then you have to question the cabling inside the building (to
> asterisk). That cabling is most often simple inside wire that
> can easily pick up noise (eg, induction from florescent lights,
> motors, wall-wart transformers, some desk lamps). If you
> don't know where the inside wire is run, might try to find
> out or bypass it with cabling laying on the floor for at
> least an elementary test.

Testing from the demarcation point is essential, and poor inside cabling
can contribute to the problem, but if the cable is Cat 3 or better, it
is unlikely that it will be succeptible to induced noise; that's why
twisted pair is twisted - to protect it from induced noise.

> If you did not _see_ a telco person on site doing the
> transmission checks, you have to assume that someone did them
> from the central office (most common approach). That's okay
> in many cases, but its not okay in other more serious cases.
> The majority of the telco people that would be dispatched for
> testing only know enough to follow printed procedures using
> whatever testset they've been given;
> they don't have the skills to actually interpret the readings
> for cases they've never seen or been trained to recognize.
> 
> Its not hard to plug an ordinary phone into the same rj11 jack
> used by asterisk. Do it and listen close. Given the problems
> that you've stated, it should not be difficult to hear noise, hum,
> low volume, etc, if it is in fact bad lines. Also, compare
> lines; it is not very often four of four lines go bad in
> exactly the same way. Can you hear any difference between lines?

This is not a bad idea, but is not always conclusive. I've done numerous
tests on circuits where it sounded great on a butt set, but was
nevertheless out of spec. Also, if the problem is due to loss, it is
quite reasonable to expect all the lines to have the exact same problem,
because they will all be exactly the same distance from the C.O.

> Bridge an ordinary phone on the same pstn line as asterisk.
> Place some calls from asterisk and listen to what's going on
> via the analog phone. (Example: some central offices don't
> like dtmf tones within xxx milliseconds after going off-hook. You'll
> get wrong numbers, etc. Insert the 'w' option in your Dial statement
> to delay those dtmf tones a little bit.) To be a little sneaky,
> unscrew and remove the mouthpiece from the analog phone and
> you can monitor calls all day long without impacting
> asterisk's ability to handle calls. 

Say WHAT?!?!

OK look, I'm sorry, but this is just plain wrong. Disconnecting the
transmitter in your handset will not alter the fact that you have
introduced a device in the loop that is in an off-hook condition.

To do what you are suggesting, one needs a butt set; which is equipped
to passively monitor the line without affecting it.

> If asterisk is having an
> echo issue (as an example) and you don't hear it with the
> bridged phone, you at least know where to look.

That isn't really true. Since the analogue phone will not have a
transcoding delay, the echo might still be there, just ocurring at the
same time as the sidetone.

> If you messed with the txgain/rxgain for your analog lines,
> go back to zero gain, use  echocancel=yes  echotraining=800
> rxgain=0.0  txgain=0.0 on each pstn line, reboot the server,
> and test using some of the above steps to verify problems.

> If you're still not sure what's going on, transmission test
> sets are sold by many different companies that you can use
> from the asterisk rj11 jack to prove line quality. New sets
> run about $400 to $600 for what you need; check ebay for used
> pricing. 

Wilcom T136Bs can be had for $20 on eBay - if you're patient.

> The telco's have a telephone number for a "quiet
> termination" and another one for their "milliwatt generator".
> Get those numbers and use the test set to measure noise
> (quiet termination) and loss (milliwatt generator). If those
> results are reaonable, then you've got an asterisk
> configuration problem (and/or digium card problem).

Yes. And test the lines from the demarcation point, with all Customer
Premise Equipment (CPE) REMOVED from the circuit (if you leave your
inside wiring connected, you are not isolating the telco circuit from
CPE). After testing from the demarc, test from the final termination
(i.e. the cord you've plugged into your FXO). Any difference is your
responsibility.

The spec for loss on a circuit is -8.5dB. Nominal is -3dB to -6db.

Cheers,

Jim.