SUMMARY: [asterisk-biz] Does VoIP Really Work for Serious Business?

[Peter Wemm]

On Friday 10 March 2006 05:30 am, Matthew Crocker wrote:
> On Mar 10, 2006, at 8:24 AM, Hugh L. Johnson wrote:
> > It's all about latency.  The roundtrip ping times across my DSL
> > circuit
> > (Verizon ATM cloud) to my ISP's router (One hop away) are approx 25ms.
> > My ISP has a PRI which I use for all incoming and local outgoing
> > calls.
> > For these calls the quality is better than when I had an analog
> > phone at
> > home.
>
> No, it is all about jitter.  Latency is irrelevant unless it is above
> 150-200ms.  Rapidly changing latency (aka Jitter) is what kills VoIP
> quality because the echo cancelers  can't train up properly.

The only option for coping with significant jitter is to have echo cancelling 
at both ends.  If echo crosses the audio stream with jitter in it, it becomes 
a nightmare to fix.  The ideal way around it is to stop the echo in the first 
place.  Unfortunately this means you need control of all the endpoints before 
they cross the public internet.

eg: if you have Office 1 with a PRI and an asterisk box, and a remote Office 
2, with public internet in between, you need to echo cancel at both Offices.  
You've got to stop the echo coming in from the PSTN via the PRI before it 
travels over the internet to office 2.  If echo reaches that far with jitter, 
it can't be reliably cancelled.

Of course, there is a lot more to it than that, but in some circumstances you 
can mitigate the echo+jitter problem.

But that generally does not help you with Internet VoIP LD providers.  They 
seem to give you the works, including the echo.  By the time you get it, the 
jitter makes cancelling it a real challenge.

The real niche for enterprise VoIP is inside the enterprise network.  Smart 
desktop handsets (portable! no need to reconfigure for moves!), integration 
into your office functions (voicemail messages sent over email or accessed 
via http, for example), and so on.  When you control your enterprise network 
and have the proper infrastructure (end-to-end QoS and sufficient capacity), 
that should be unbeatable if it is done right.  

-Peter