- This topic is empty.
27th March 2001 at 22:03 #20186WebproGuest
Looking for some results in sound sample form (best bet) or report/analysis form for voice as it is processed through multiple vocoders.
Would be real cool to have something processed through say g.726 then through g.729. Having source sample — g.726 vocoded sample — g.729 vodcoded sample would be slick. Having any vocoder — vocoder samples would be excelent.
As VoIP is more widely deployed, I expect there to be real issues with voice quality as it passes through multiple vocoders, especially for “on-net” calls. There will be similar issues for vocoder — PSTN — vocoder as well.
Anyone seen this kind of analysis? Know where it can be done?4th April 2001 at 01:54 #20187RGuest
This could be done in any lab with a few voip gateways, but why? If properly designed this should not happen even in a Service provider environment where various IXCs are involved. VoIP only came about because of toll arbitrage. If I have sent it the brunt of its distance for delivery I should be close enough to the user not to cause a tandem encoding scenario. I guess the issue could come in when a call is originated by an enterprise voip deployment then sent to traditional pstn which uses an ITSP(internet telephony service provider) to carry its traffic back to a VoIP enterprise environment. BUT the enterprise voip environment should be using G711 on campus, and the issue should be moot again.4th April 2001 at 08:16 #20188DonGuest
Consider a call originating in a ‘traditional’ PABX,across a private network (VoIP) to an extension which is on ‘call-forward’ to a mobile phone.
The call will originate in G711(originating PABX), convert to a VoIP standard(private network), back to G711(terminal PABX), possibly back to analogue(POTS line on terminal PABX), probably back into g711 again(Carriers digital exchange) before being forwarded to the mobile network(which encoding? which mobile—anyones guess!!)
This is a simple scenario which involves up to 6 encoding changes.
Yes, this type of modelling is long overdue.4th April 2001 at 19:37 #20189WebproGuest
Oh, it will definitly happen. I can give get some MOS (yea right) or PSQM scores in the above mentioned scenario. I have done g.711 to PSTN to G.711 to PSTN and PSTN to G.711 to PSTN and PSTN to G.711 to PSNT to G.711.
What I can not do is take some sort of standardized sample, go from G.711 to G.723.1M or G.729a and save the resulting sample.
Anyone know of a lab or testing site for this kind of modeling?