[Asterisk-Users] SpanDSP issues (oh fun!)

[Tom Christensen]

I am having classic "frame slip" symptoms on 4 different systems with 4
different providers (all full PRIs, Qwest, XO, Xspedius, and First Digital
are the providers).  By classic I mean pages cut short.  I do not hear
clicks on calls however, and faxing from a fax machine plugged into the
asterisk box via an fxs port (no VoIP) faxes just fine through the PRI.
Also, faxing directly from said fxs port to spandsp (IE not traversing the
PRI) works wonderfully.

I do not see any missed interupts on any of these systems, the digium cards
(tdm400's for fxs ports, and te110p's for PRI termination) are not sharing
IRQs with anything else, and they are not on the same IRQ themselves IE,
they are totally isolated in the IRQ space.

Generally shorter faxes (IE < 3pages) get through just fine, however
anything longer than that will have a failed page about 50% of the time, and
anything over 10 pages always has a failed page (except on Xspedius, which
has about a 50% fail rate above 10 pages).

I have searched everywhere, and the "answer" to this problem seems to be
frame slips on the PRI however, all 4 systems are set to sync with the PSTN
connection, I'm not running X on any of these systems, (although turning X
on doesn't negatively effect them either... IE I get the same number of
errors with X running or not).  Also, there is no mention anywhere of a
"fix" or even anything to try to reduce or eliminate frame slips.

I recompiled spandsp to log the audio files, does anyone know what type of
file that is? can I listen to it? Or is Steve the only one who knows how?

 how come a standard fax machine can fax over this PRI connection just fine,
but spandsp cannot use it?  If frame slips are occurring, shouldn't they
break faxing on the machine as well?  If not, the machines must be doing
some error checking or something that spandsp does not currently support so
that they can deal with frame slips.  Does spandsp do any sort of error
checking/page resend requesting?  What would it take to make spandsp more
robust in this regard?  Or is it just something that can't be overcome in
software, but that hardware modems in fax machines can handle?  I am more
than willing to help out any way I can (including code, although I'd need a
little help getting up to speed on the code) to get spandsp working on these
boxes.

Is there anything else to try?

Basically all of the posts I've been able to find say "Don't share IRQs,
turn off X, and make sure the PRI is configured properly", well I've done
all of that.  Do I just have 4 bad PRIs that can't keep time?  These systems
are deployed as a test solution for fax-email, they are the only 4 I've
setup to use spandsp so 100% of the systems I have installed show this
problem, and they show it across telecom vendors all other things being
equal (exact same hardware, exact same OS, exact same asterisk, spandsp, etc
versions asterisk 1.2.4, zaptel 1.2.4, libpri 1.2.2, spandsp 0.0.2pre25).  I
have tried later versions of asterisk, zaptel, and libpri, but not a
0.0.3version of spandsp.  I have sent faxes from many different fax
machines, all
show the problem.

I have attempted to run a utility I found called ztclock... suprisingly of
the 4 pri's the "best" as far as spandsp results are concerned is Xspedius
(of the 4 it is the only one that ever receives 10+page faxes without error,
still only about 50% of the time, but the other 3 never receive more than 10
pages without failing a page).  However, according to ztclock it is the
worst as far as frame slips are concerned.  Here are the results of Qwest vs
Xspedius.

Qwest PRI:
ztclock - clock source accuracy test (3 passes)

Flushing input buffer...
Flush Complete.

Test is approximately 3 minutes.  Please wait...

483328 samples in 60.416021 sec. (483329 sample intervals) 99.999794%
483328 samples in 60.416041 sec. (483329 sample intervals) 99.999794%
483328 samples in 60.416097 sec. (483329 sample intervals) 99.999794%

Estimate 8 frame slips every 483.328003 seconds.


Xspedius PRI:


ztclock - clock source accuracy test (3 passes)

Flushing input buffer...
Flush Complete.

Test is approximately 3 minutes.  Please wait...

483328 samples in 60.416018 sec. (483329 sample intervals) 99.999794%
483328 samples in 60.416037 sec. (483329 sample intervals) 99.999794%
483328 samples in 60.514997 sec. (484120 sample intervals) 99.836136%

Estimate 8 frame slips every 0.610263 seconds.

The other 2 PRIs show ztclock results similar to Qwest, 8 frame slips every
400-500 seconds.
Any help or suggestions would be greatly appreciated.

Thanks for your time,
Tom Christensen
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.digium.com/pipermail/asterisk-users/attachments/20060519/c29926e4/attachment.htm