[Drily Lit Raga]

I~m working with a team developing streaming audio-video systems. We
need to make some competitive comparisons of our audio codec. I~m
thinking that pure frequency response won~t cut it, because supposing
all codecs are sampling at 32 kHz, there won~t be much difference
shown between say, 64 kbps and 96 kbps. I do know from personal
experimentation with MP3 that 192 kbps sounds ^pretty good^, higher is
of course better, and below 128 kbps it starts to sound bad, but
mostly getting effects like transient smearing, which you wouldn~t
expect to get from sine waves or sweeps.

I suppose one could normalize the encoded output of any complex bit of
sound, e.g. speech, music, typical TV noises (it is for sending
broadcast TV signals mostly), and then subtract it from the original
signal to get a residual, but this seems highly subject to correct
level matching, time alignment, blah blah blah.

Suggestions (question mark) my keyboard~s gone all screwy on me !

Thanks.

[Les Cargill]

Drily Lit Raga wrote:
I~m working with a team developing streaming audio-video systems. We
need to make some competitive comparisons of our audio codec. I~m
thinking that pure frequency response won~t cut it, because supposing
all codecs are sampling at 32 kHz, there won~t be much difference
shown between say, 64 kbps and 96 kbps. I do know from personal
experimentation with MP3 that 192 kbps sounds ^pretty good^, higher is
of course better, and below 128 kbps it starts to sound bad, but
mostly getting effects like transient smearing, which you wouldn~t
expect to get from sine waves or sweeps.

I suppose one could normalize the encoded output of any complex bit of
sound, e.g. speech, music, typical TV noises (it is for sending
broadcast TV signals mostly), and then subtract it from the original
signal to get a residual, but this seems highly subject to correct
level matching, time alignment, blah blah blah.

Suggestions (question mark) my keyboard~s gone all screwy on me !

Thanks.

Call Spirent or Agilent.

--
Les Cargill

[Arny Krueger]

"Drily Lit Raga" wrote in message


I~m working with a team developing streaming audio-video
systems.

As if we don't have enough of those already!

We need to make some competitive comparisons of
our audio codec.

How have the subjective comparisons worked out for you?

In some sense, you should know the answer to the question if you are
developers, because you should know what you did to your product to improve
it when it didn't sound good enough.

I~m thinking that pure frequency
response won~t cut it, because supposing all codecs are
sampling at 32 kHz, there won~t be much difference shown
between say, 64 kbps and 96 kbps.

As a rule, codecs have had good frequency response for tones and sweeps for
a long time.

The next stage of complexity after simple frequency response testing is
testing with complex multiones. As codecs have improved over the years,
this has also improved.

I do know from
personal experimentation with MP3 that 192 kbps sounds
^pretty good^, higher is of course better,

Not that much better, after 192.

and below 128 kbps it starts to sound bad,

Agreed.

ut mostly getting effects
like transient smearing, which you wouldn~t expect to get
from sine waves or sweeps.

Seems like bursts of carefully-tuned multitones might be a productive way
to go.

[Mark]

ut mostly getting effects
like transient smearing, which you wouldn~t expect to get
from sine waves or sweeps.

Seems like bursts of *carefully-tuned multitones might be a productive way
to go.

see:
http://www.avisoft.com/compression.htm

Mark

[Drily Lit Raga]

On Oct 15, 11:50*am, Mark wrote:
ut mostly getting effects
like transient smearing, which you wouldn~t expect to get
from sine waves or sweeps.

Seems like bursts of *carefully-tuned multitones might be a productive way
to go.

see:http://www.avisoft.com/compression.htm

Mark

Interesting! I wonder how my colleagues will respond to the
variations in intelligibility of a series of mouse, bat, cricket and
cockroach chirps? hmmmm 8^) some of them might get hungry....

Actually it was an intersting web site, thx!

DLR