[Asterisk-Users] An anniversary and a lament for FXOs

[Rich Adamson]

> So then the system I'm building (P4 2.66GHz) should do fine in the
> encoding/decoding (which I assume accounts for most of the echos, and poor
> quality).
<snip>
> > use Sipura SPA-3000s but they're only marginally better. How is it that
> > my Panasonic 4 line SOHO phone system (KX-TG4000B) can have four
> > stable, reliable FXOs with no echo at all in a device with a total cost
> > of <$500? It seems to me that there ought to be hardware available that
> > behaves just as well, but bridges the PSTN to the SIP/IAX domain?
> 
> VOIP introduces latencies that Just Do Not Exist in non-VOIP scenarios.  I 
> would have thought that after a year with Asterisk you would have come to 
> that conclusion.
> 
> > I've read a lot on the list about how difficult designing FXOs can be,
> > but that flies in the face of the fact that every small multi-line
> > phone system has them...and without expection those behave better than
> > the devices I've been able to try with Asterisk. The Sipura SPA-3000
> > has several settings to adjust for line impedance and
> > inductive/capacitive line loading....lots of settings, but it provides
> > nowhere near the basic performance of one of the lines on the Panasonic
> > KSU. It's simply mind boggling.
> 
> You're tuning the hybrid; the problem lies in the additional latency
> incurred 
> with VOIP -- If you took the Panasonic system hardware and VOIP'd it you 
> would likely see similar issues that you see with the X100P, TDM400P and
> even 
> channel banks.  You need a good stable and performance system to get good 
> VOIP quality.  There is no substitute.
> 
> Case in point:  I was running an X100P and a TDM410P in a P90MMX -- VOIP
> calls 
> were pretty good but anything over the POTS was kind of shitty... put the 
> exact same hardware into a P3-700 and the POTS problems went away for the 
> most part.

The majority of the above statements are either not true, or inaccurate
conclusions were drawn based on observations as opposed to facts.

We've been through this topic at least several hundred times over the
past year, and the conclusions that seem to hold the most truth are:
 - systems that have had large amounts of echo with the tdm card have
   none (or near zero) by simply swapping motherboards,
 - the amount of echo associated with x100p verses tdm04b in the US
   is the same (not true outside the US),
 - the brand name of the system (eg, Compaq, Dell, HP) has nothing to 
   do with the echo problems,
 - swapping motherboards from a dual-something-or-another to a slower
   P3 has fixed some echo issues (suggesting motherboard speed and/or
   number of processors has nothing to do with the issue),
 - near-end vs far-end echo (and assoicated hybrids) has nothing to do
   with the issues in the majority of US cases,
 - those non-asterisk users that are heavy into audio projects have
   known for some time that PC motherboards (as opposed to Apple boards)
   have serious throughput issues with PCI controllers and/or interrupt
   latency,
 - multiple individuals have suggested writing a routine to eval the
   latency issues, but no one has completed such an app,
 - at least in my case (with over 20 years of hands-on transmission
   engineering experience), extensive testing of four pstn lines and
   associated hybrid settings yield absolutely no improvements,
 - the majority of tdm echo issues seem to be related to the latency
   from sending data to the tdm card and receiving back the "echo",
   and the limited window in which the existing * canceller routines
   actually are able to operate (per multiple developers with extensive
   knowledge on the subject),
 - external pstn gateways (such as the Mediatrix 1204) have absolutely
   no echo issues on the same pstn lines used with the tdm04b (however
   the mediatrix has other issues that preclude long-term production
   use).

There is no doublt that mismatches associated with the x100p and/or tdm
fxo cards _can_ be great enough to cause echo issues, but there are
pletty of real world examples where those mismatches have been corrected
yet the echo issue is far greater then what non-technical users would
put up with.

It would _really_ be nice if we could assemble a list of motherboards
with specific part numbers that do/don't have the issues. Anyone care
to contribute?

Rich