Improving audio in BigBlueButton 0.8 -- test and give us feedback

[Fred Dixon]

We’ve been working on BigBlueButton 0.8: adding pre-flight check-list,
record and playback, and exploring ways to improve the audio.

Currently, the audio delay in BigBlueButton 0.71a is caused by the
Java-based transcoding of audio packets from nellymoser <-> mlaw
within red5phone. See FAQ.

http://code.google.com/p/bigbluebutton/wiki/FAQ#Why_is_there_a_delay_in_the_audio_when_I_use_VoIP?

This means, even when a BigBlueButton server running is running within
a LAN environment (no network latency, no packet loss), with only two
people on the server, the audio delay is still about 1.5 seconds.

We’ve been thinking about our approach to audio and realize we’re
trying to solve two problems at once: (1) quality and (2) scalability.
But the feedback from many is “what good is scalability if the audio
lag is still 2 seconds for small groups?” For many -- especially
when doing small collaboration or one-to-one sessions -- low latency
is the key.

Flash supports both nellymoser and speex. Recently, we’ve been
working on optimizing the use of 16 kHz wide-band speex within
BigBlueButton. Turns out, this significantly reduces the lag in
audio. Why? When using speex, red5phone does *no* transcoding.
None. It passes the packets straight through to FreeSWITCH, which
encodes/decodes the speex packets in native code.

We have setup a test server for you to try out the audio in
BigBlueButton using speex. This server is running an internal build
of BigBlueButton.

http://test.blindsidenetworks.com/

So, grab a set of headphone, login to the server, and give us your feedback.

A good way to determine the latency is to login *twice* (use two
browser tabs), mute yourself in one session, then speak. Your audio
packets will go to the server in one session and come back to you in
the other. The difference between your speaking and hearing is the
latency.

There are still more problems to solve, and there is no free lunch.
Encoding speex is CPU intensive, so you’ll find FreeSWITCH taking more
CPU as the number of users increase. The Flash client is still using
restricted to transmit using TCP, which means if there are corrupted
packets, the user’s client will resend corrupted packets, and these
resent packets may arrive too late and be discarded as FreeSWITCH
attempts to keep the user in sync.

We look forward to your feedback as we keep exploring ways to improve the audio.

Regards,... Fred

[Joe Miller]

Fred,

After my initial test run, all I can say is "Wow!"  This makes a huge difference in the quality of experience for small workgroups and the speex 16Khz codec compares quite well with the previous Nellymoser quality.  Congrats on this big design win!  Letting FreeSWITCH do this work is the right call, and may enable other (less cpu-intensive) codecs, such as G722.1C to be tried for "HD" quality audio in the future.

Regards,
-- Joe Miller

--
You received this message because you are subscribed to the Google Groups "BigBlueButton-dev" group.
To post to this group, send email to bigblueb...@googlegroups.com.
To unsubscribe from this group, send email to bigbluebutton-...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/bigbluebutton-dev?hl=en.

[hamdy fadhloun]

really good work.
is it the same when we use asterisk ?

2011/3/17 Joe Miller <jmi...@perilux.com>

[Felipe Cecagno]

Hi Fred, great work, congrats for all the team!

We've tested here and tried to find a way to measure the delay. So, we've thought about to use a metronome and two computers and try to synchronize the beats, source and destination, with 1 beat delay.

Applying that method using http://www.metronomeonline.com/, we've found a delay of 0,91s - exactly 66 beats-per-minute.

Regards,

Felipe

[Fred Dixon]

Hi Hamdy,

Not at the moment.

The good folks at Generic Conferencing paid an external consultant to
add speex support to app_konference, the conference module used by
BigBlueButton in Asterisk, but we've not tested our changes with
Asterisk yet.

Regards,... Fred

[hamdy fadhloun]

thank you for your response Fred.

never mind we'll wait to test it.

thank you for all the good work you've done.

Best regards 
hamdy

2011/3/17 Fred Dixon <ffd...@gmail.com>

[Chris]

I just tried it out and it is MUCH improved. About how much of a load
increase does this place on a server?

Can't wait for the .8 release! The audio is a major improvement.

-Chris

On Mar 17, 11:11 am, Fred Dixon <ffdi...@gmail.com> wrote:
> We’ve been working on BigBlueButton 0.8: adding pre-flight check-list,
> record and playback, and exploring ways to improve the audio.
>
> Currently, the audio delay in BigBlueButton 0.71a is caused by the
> Java-based transcoding of audio packets from nellymoser <-> mlaw
> within red5phone.  See FAQ.
>

>  http://code.google.com/p/bigbluebutton/wiki/FAQ#Why_is_there_a_delay_...

[Fred Dixon]

> About how much of a load increase does this place on a server?

Our plan is to schedule a specific time in the near future, invite
everyone on the mailing list to join at that time, and watch the
server load.

We're thinking Tuesday, March 22, at 9-10am EST.

Regards,.. Fred

[Andrew E]

Glad to hear about this progress! Are you considering enabling 16 kHz
speex by default in 0.8?

Also, when you do the scheduled test, will you be able to tell us the
server specs so we can have a baseline to compare to?

Thanks,

Andrew

[HostBBB.com]

Great Job, watch out Skype and Web-x you guys have a contender in
the making.

This fix, will negate the need for a lot of people trying to configure
sip clients simply to gain a little less audio delay.

regards,
Stephen

> > > For more options, visit this group athttp://groups.google.com/group/bigbluebutton-dev?hl=en.- Hide quoted text -
>
> - Show quoted text -

[IBRAHIM]

Hi Fred,
My major problems with 0.71 are about the audio that I couldn't use
BBB on language teaching (which needs quality sound). That's why I'm
waiting for BBB 0.8...
(1) Audio quality (8 Khz),
(2) Audio delay: Synchronization between video and audio.

I tried the test server. Is is definately WONDERFUL. Increased audio
quality (16 Khz). Delay between video and audio is acceptable, you
can't event notice it if you are not so careful. Thank you all for
this great work.

Ibrahim

[paulik]

Excellent work!
We trying to speak on test server with 6 people and it's work fine!
Audio quality and delay like a Skype, even better!
Synchronization of our video and audio on the excellent level.

Thank you for your great work!
Can you publish this release? so i can test server load with many
users and public results here.

[hamdy fadhloun]

yeah can't wait for the new release :)

2011/3/19 paulik <bypa...@gmail.com>

[Vittorio Scarano]

Hi everybody. I confirm a very good users' experience.. tried with 4
webcams and the audio worked fine, no significative delay and the
synch with video was almost perfect... Some annoying echo, but only
because one of the team had no headset. In conclusion.. I can't wait
for the 0.8 to come out.. :-)

Vittorio Scarano

--
_____________________________________________________________________________
Vittorio Scarano
Dipartimento di Informatica,
Università di Salerno, Fisciano (SA) -- 84084 ITALY

E-mail:          vit...@dia.unisa.it
Homepage:    http://www.dia.unisa.it/professori/vitsca
ISISLab:        http://www.isislab.it
Phone :        +39 089 96 9733 (office), 9308 (ISISLab 1), 9307 (ISISLab 2)
Fax:             +39 089 96 9600

[Denis Zgonjanin]

The echo problem should be solved now with Flash Player 10.3 coming out.

According to the docs however, enabling echo cancellation adds 100-200ms of additional delay. So we'll need to make it optional for users, since if you use a headset you don't need echo cancelation.

- Denis

[hamdy fadhloun]

Hi
is there any estimated date to launch the 0.8 version of bigbluebutton ?
thanks

2011/3/21 Denis Zgonjanin <deniszg...@gmail.com>

[Fred Dixon]

Hi Hamdy,

We don't give out estimates for delivery dates. We release on
quality, not dates.

However, you'll know when we're getting closer to a delivery for 0.8
when we start our testing cycle. During the testing cycle, we shift
our focus to stability and bug fixing. For an overview of our
development and testing process, see

http://groups.google.com/group/bigbluebutton-dev/browse_thread/thread/8fc2360646bc6a09

We're not there yet for the testing cycle for BigBlueButton 0.8, but
working hard towards it.

Regards,... Fred

[Marc Laporte]

Hi!

I confirm a very positive test in the last few minutes, with 4 people.
No noticeable audio lag.

Congratulations to all involved!

Best regards,

M ;-)