Re: [codec] this weeks summary

Hi Cullen,

In my original email discussion of VoIP gateways, I did not use the 
phrase "ultra-low delay".  What I did say was that due to the large 
number of voice channels that the DSPs and the (possibly two layers of) 
micro-controllers have to handle simultaneously, the timing/scheduling 
can be quite complex, and the necessary buffering at the (possibly 
three) hierarchical layers of processors can add substantial delay.  
Therefore, the worst-case codec-dependent one-way delay may be up to 5X 
to 6X the codec frame size before delay optimization, and even after 
the engineers "optimize the delay to death", they could only manage to 
get the worst-case codec-dependent delay down to around 3X codec frame 
size.  

Again, I didn't come up with that myself.  A senior technical guy who 
was deeply involved in high-density VoIP gateway designs told me that, 
and another technical guy who worked at a different company than the 
first guy and who was involved in the design of a different VoIP 
gateway based on a different DSP chip and a different system 
architecture independently gave me a range of codec-dependent one-way 
delay around 3X codec frame size, and he didn't know that I talked to 
the first guy.

Due to this 3X multiplier, an increase in the codec frame size will 
have a larger delay impact in VoIP gateways than, say, a softphone 
that has a lower multiplier.  Perhaps it was due to this larger 
multiplier that Christian used the phrase "ultra-low delay", but I 
myself didn't use that phrase.

I hope this answered your question.

Best Regards,

Raymond

-----Original Message-----
From: codec-bounces@ietf.org [mailto:codec-bounces@ietf.org] On Behalf Of Cullen Jennings
Sent: Wednesday, May 12, 2010 8:21 AM
To: Christian Hoene
Cc: codec@ietf.org
Subject: Re: [codec] this weeks summary

On May 9, 2010, at 12:10 PM, Christian Hoene wrote:

> Raymond did a great job explaining the special requirements of high-density VoIP gateway. They are differ substantially to those of
> soft-phones (but are somewhat similar to VoIP phones with embedded CPUs). The requirements include low memory footprint, ultra-low
> delay,

Uh, can someone walk me through the low delay part? I either missed of failed to understand that 

> low complexity and mostly narrow band. But aren't those requirements already covered by existing codecs? Shall this working
> group provide a codec profile for end-to-gateway beside the profile for an end-to-end codec?

Cullen Jennings
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/index.html

_______________________________________________
codec mailing list
codec@ietf.org
https://www.ietf.org/mailman/listinfo/codec