On 2007-02-24, Larry <noone@home.com> wrote:
> own echos from cellphone to cellphone! The call was clearer on Skype as
> we were only using one slow codec, his cellphone.
>
> I called a local number from the Skype laptop that always puts music-on-
> hold on the line while you wait and wait to talk to the lady at the
> electric company. We listened to the elevator music..yecch..on Skype and
> noted its fidelity. Then we called it from his Cingular cellphone and my
> Alltel cellphone, both on different digital schemes to compare. The
> Skype rendition of the same music-on-hold was MUCH better. That stands
> to reason because as anyone knows music on digital cellular just SUCKS
> because of the awful slow codec cellular uses to increase traffic.

If you look in the Advanced section in Skype Preferences (on my Mac at
least, I don't know the PC equivalent) and tick of "Display Call Technical
Info", then when you make a Skype call it will pop open a window which
tells you, among other things, the codec being used. For SkypeIn and
SkypeOut calls the codec is always listed as G.729, the same as all VoIP
providers, as it must be since this is the standard used by the operators
who deliver Skype calls to the PSTN. If you look at the overview here:

http://en.wikipedia.org/wiki/G.729

you'll find that this is an 8 kbps codec, compared to the 12 kbps codec
used for GSM and the up to 12 kbps codec used for CDMA. I'm not sure
what this codec does to music, but note that that DTMF tones can't be
sent through this codec with sufficient fidelity to be reliably decoded
and instead need special-case handling (a complexity which I suspect
is the source of the periodic griping on Skype forums about DTMF not
working reliably). There's nothing "fast" about this codec, the "fast"
codec (iSAC) is only used for Skype-Skype calls.

Despite this, I too have noticed that Skype-PSTN calls seem to me to
sound better at both ends than cell phone-cell phone calls. For what
it is worth my current theory is that this is because the quality of
PC audio hardware is sufficiently higher than the audio hardware in
a typical handset for you to hear the difference. It certainly has
nothing to do with the codec in use, even if one accepted the premise
that using a better codec for only part of the connection would improve
anything.

> Encrypting your audio takes quite a bit of computer power.

While encryption can consume lots of CPU cycles at high data rates, the
data rate for a voice call is tiny. I suspect most of the CPU is
consumed by the codec, since squeezing even phone-quality voice down
into 8 kbps with reasonable fidelity takes some work.

Dennis Ferguson