[Asterisk-Users] Any 24 (or 30) way FXS PCI cards?

[Steven Critchfield]

On Sun, 2005-03-20 at 09:54 -0600, Rich Adamson wrote:
> > > It seems to me silly to have a T1/E1 card to connect to a channel bank
> > > when you could just have a 24/30 way FXS card in the slot in the first
> > > place.
> > > 
> > > Does such a thing exist?
> > > 
> > > Wouldn't Digium have a lot of customers if they could produce one for
> > > say < $1000 retail?
> > 
> > Trouble is power. Unless there is more power made available, you may not
> > be able to drive the ring voltage of several consecutive lines at once. 
> > 
> > Take for instance the Adit 600, it has a 130w power supply for just 25
> > ren capability. Think of the troubles that would cause trying to be
> > regulated through your standard PC PSU of 300w. Won't you just love
> > trying to diagnose random reboots right after a phone call comes in and
> > over draws your PSU capacity and it goes into short protection where it
> > begins pulsing power.
> 
> The entire telephony system relies heavily on statistical probabilities,
> including central office switches, inter-office trunking, and ringing.
> Applying the above thought process to a central office, there is no way
> a single central office could possibly ring all telephones associated
> with that office simultanously. The ring generators could never handle
> it.
> 
> Likewise for a pc card supporting 24 fxs lines. The probability of three
> or more lines ringing at exactly the same time are very small. With at
> least a little engineering forethought, its not that difficult to
> create ring cycles where ports 1 through 6 ring during some period,
> followed by 7 through 12, etc. (That's actually how most central 
> offices handle ringing now with a couple of exceptions.)

Maybe, but is that something you would expect in a PBX? And do you do
that in drivers or in circuits?

> A bigger issue is really oriented around how many ringers exist on a
> single fxs port, and that _could_ be limited by specifically limiting
> user implementation (via specs) to one phone per port (or whatever), etc.
> 
> Couple that with the fact that current ringer designs require 
> substantially less power then the old electomechanical ringers, it
> certainly isn't that difficult to design a PC board to support
> 24 fxs ports.

You are right, but you still hit the same problem just at a lower
probability of major problems. I still contend that any design is going
to need it's own external powersupply so as to not over draw the PC's
PSU and cause it to either fail or pulse causing system stability
problems. 
-- 
Steven Critchfield <critch at basesys.com>