[VoIP] Zaptel.c patch?

[ikjtel]

Hi Don

I've been following with great interest your MF and
2600 project, delighted to hear from you.  If there's
anything I might do to help, let me know.

There are actually two separate things going on in the
patch, perhaps should have been separated, but...

The 2600 additions were to enable SF-to-DC conversion
for incoming VOIP audio (assuming an FXO leg) for
application to an incoming selector in a SxS office. 
The intent was to allow distant callers (who must be
equipped for SF outpulsing) directly to dial through
the switches in remote EM offices *and* to hear the
switches operating in real time.

The X100P (junk) cards won't do it, but most channel
bank FXO's will also pass audio back to the remote
caller EVEN WHEN ONHOOK.  On John Novack's W.E. SxS
machine you used to be able to hear the dialtone
motor/generator set winding down like an old record
player turntable by blasting 2600 at the 1st selector
and listening as the FXO line goes onhook....

I believe Mr. Ratguy (Jayson Smith) also has
successfully SF'ed into offices equipped with some
versions of this patch.

As for the code itself, the bulk of the zaptel changes
were simply to move the existing sf_detect() function
upwards in the file so its fn definition is defined at
an earlier point.  There is code added to invoke
sf_detect for audio received via the WAN.  This is the
actual SF-to-DC conversion.  It implements a digital
LPF to try to cover up IP packet loss/latency; the
resulting filtered DC signal is used to toggle the
FXO's transmitted hook state.  If I recall correctly,
there's code to notch out any 2600 that might appear
in frames of audio that are flying by (in that same
direction of flow), although in this particular case
it's sort of moot because by definition the very next
link in the chain is in the onhook state (where no
audio is presumed to be proceeding to whatever follows
next ;)

Also, note that it's NOT true that you only need to
defeat the built-in audio muting just in zaptel
(unless you're using zaptel by itself without
asterisk, which most people don't do, although perhaps
you are)...  In addition to commenting out the section
of zaptel code you mentioned (below), there's also
some "undesirable" muting code in asterisk
(chan_zap.c) that you need to defeat.  Also, for my
application it seemed that native bridging needed to
go (an interesting feature, an optimization, not
strictly essential).

Here are the current links if I recall correctly
http://www.lightlink.com/mhp/sf/sf2.patch
http://www.lightlink.com/mhp/sf/sf-0.2.tar.gz

Best

Max

--- Donald Froula <dfroula at sbcglobal.net> wrote:

> Max, I'm a CNET member on extension 1-762.
> 
> On my Asterisk box, I'm running some Zaptel patches
> that correctly enable 2600 SF/MF signalling,like
> that
> of the old blue box days. I have this working over a
> TDMOE 24 member trunk-group using two extra looped
> Ethernet cards.
> 
> All works very well, except the MF outpulsing over
> the
> trunks can not be heard when a call is set up. I was
> hoping your patch on the CNET web site might help
> here.
> 
> However, the patch seems to fool with the SF notch
> filter, which I am using for the trunks.
> 
> I seems the only thing needed to unmute Zaptel on
> call
> setup is to comment out the following block of code,
> which your patch does:
> 
> if 0
> 
>      if (ms->dialing) ms->afterdialingtimer = 50;
>         else if (ms->afterdialingtimer)
> ms->afterdialingtimer--;
>         if (ms->afterdialingtimer && (!(ms->flags &
> ZT_FLAG_PSEUDO))) {
>                 /* Be careful since memset is likely
> a
> macro */
>                 rxb[0] = ZT_LIN2X(0, ms);
>                 memset(&rxb[1], rxb[0], ZT_CHUNKSIZE
> -
> 1);  /* receive as silence if dialing */
>         }
> 
> #endif
> 
> Could you explain a bit how the patch works,
> especially th part that affects the notch filter?
> 
> Best,
> 
> Don Froula
> 
> 


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com