[Asterisk-Dev] libiax2 OK for production?

[Steve Kann]

Steve Underwood wrote:

> Bruno Hertz wrote:
>
>> On Mon, 2005-01-31 at 18:41 -0500, Steve Kann wrote:
>>
>>  
>>
>>> What kind of latency are you talking about here?  There's about 60ms 
>>> of latency, on the output side (i.e. to the speakers) in iaxclient 
>>> proper that could be removed if we changed the architecture a bit 
>>> [removing the buffers in-between decoding and the audio layer], but 
>>> if you're seeing a _lot_ of latency, then there's probably some 
>>> particular issue that can be resolved.
>>>   
>>
>>
>> About one to two seconds on Fedora Core 3. On my system, it just can't
>> compete. Of all clients I tried, linphone, sjphone, cornfed,
>> gnomemeeting, ... iaxclient is worst latency wise. So it might be that
>> particular issue, but I myself see no easy way pinpointing it.
>>
>> Reinventing the wheel, not really I guess. Portability is not my
>> concern, and as said miniphone already does what I want at a basic
>> level. The rest, like codec selection and comfort stuff, is just
>> fun hacking I'd say.
>>
>> Thanks, Bruno.
>>  
>>
> This came up before. iaxcomm itself doesn't give such long delays, but 
> some sound card drivers do. My desktop machine give little latency. My 
> notebook gives a second or more. I think your sound card is most 
> likely the source of this latency. However, the fact you get better 
> results with miniphone means it is not a fundamental problem with the 
> card and driver (I assume you tried these tests on the same machine, 
> or they are meaningless). I believe miniphone and iaxcomm use 
> different ways to drive the sound card. Maybe iaxcomm is doing 
> something wrong. Sound cards normally work in a high latency mode, 
> which provides lots of buffering, and have to be switched to low 
> latency by the application. Maybe it isn't done correctly in iaxcomm 
> for all cards, or maybe there is some OSS/ALSA conflict. Do you know 
> what sound card hardware you have, and which driver?


I haven't examined all the code, but I'm pretty sure the major 
parameters of what they're doing should be similar:

They both use OSS
They both set the card to 8khz mono
They both set the card to full-duplex.

There are other things that might be different, though..


-SteveK