Megaconference

[Erik Dobbelsteijn <>]

Actually, to perform decent echo cancellation you need three things:
- an adaptive though robust algorithm
- input signals and output signals with decent Signal to Noise Ration that 
are as linear as possible.
- the ability to subtract the environmental noise on your (!) side from your 
(!) input signal before sending it to the far side

(bear in mind that it is always the one on the other side who has to take the 
effort/cost to cancel echo for you. 

On the side of the algorithms: the most popular is the LMS algorithm, since 
it is simple and can easily made more robust. It has the least damping of the 
environmental audio, though. It really doesn't matter whether it is 
implemented in hardware or software (and anyways: a DSP runs software, too 
:-) ) as long as the audio processing path is as real time as possible (short 
and consistently short delays).

That brings us to the ability to get the right signals into the algorithm 
processor. This has been a problem in pc's for a long time. As stated by 
Simon, dedicated echo cancelling equipment like the long time used line echo 
cancellers in long distance telephony infrastructures but also hardware video 
conferencing stations have had hardware DSP's on board for a long time and 
the inputs and outputs were wired in such a way that they were available to 
the algorithm. Most of them, Apple in particular, do not provide any info on 
how they implement it (at least not that I have found). Telco grade echo 
cancellers come with detailed info on how to adjust parameters like the tail 
length of the adaptive filter in the algorithm and such.
The quality of microphones, use of array microphones and wideband loudspeaker 
drivers make sure that the signals are of such quality that the algorithm can 
reduce the environmental noise as much as possible. Quality of (laptop) 
loudspeakers and microphones have drastically increased in the last 10 years 
or so, fortunately.

Since Windows Vista (not my favorite OS), Microsoft has enabled programmers 
to use both inputs and outputs in one algorithm, and is actually providing 
basic echo cancellation that any application can use. Windows 7 has 
significantly improved in this area. Still, though, video conferencing 
applications are far ahead when it comes to the quality of the algorithm, 
Skype being one of the best with Mirial, Polycom and Vidyo as very good 
runners up (in my personal experience, have not taken measurements. that 
would not be very simple since it cannot be turned of most of the times).

...and yes, door slams can be canceled if the algorithm looks a little ahead 
of time. That only has to be one sample. As soon as the adaptive algorithm 
has 'learned' the characteristics of the room, it can pretty well block any 
noise travelling through it.


Erik Dobbelsteijn

On 1 Dec 2012, at 8:08 PM, Simon Horne 
<>
 wrote:

> Christian
> 
>> Beyond your point about Windows being so difficult to master regarding
> sound processing, it's a little bit scary to see that every *application*
> needs a different hardware!
> 
> The inherent problem is that unlike MAC,  Windows does not control the
> hardware so by default it has no hardware echo cancellation and every vendor
> has their own software implementation which of course requires different
> configured devices to work around. Oh it’s truly a mess. 
> 
> 
>> Regarding Spranto, I gave it a (very short and incomplete) trial a couple
> a weeks ago. I just called the Esnet MCU from a new iPAD (without remote
> participants). The video was often 
>> quite pixellated, under conditions where other systems (Cisco Jabber,
> zoom.us, Lifesize softphone) are performing quite well.
> 
> And I certainly would expect that to be the case. Let me explain.  Firstly
> zoom.us is a closed island so let's just forget about that.
> Both these softphones are SIP based and almost certainly would require media
> proxy if behind a NAT to reach a standalone endpoint or MCU. I think (I
> could be mistaken) that video size received on the IPad to be around 352x288
> (CIF) or 320x240 (QVGA). My testing maybe out of date so please someone
> correct me if I am wrong?!?
> 
> Spranto utilizes direct media meaning there is absolutely no media proxying
> in most cases even to existing MCU's such as Esnet. This allows the new IPad
> to be tuned by default to receive a higher 640x480 (VGA)  and with the HD
> switch turned on up to 1280x720 (720p) which is between 3 to 12 times the
> bandwidth. The picture quality would be superior on the retina display but
> it does comes at a heavy price and that is wifi range. It is very difficult
> to maintain the higher bitrate over the wifi with the same distance from the
> access point. We currently opt for quality over range however maybe we need
> to add a 3 way switch in the settings  LD > SD > HD to compensate for
> greater range. Thanks for the feedback.
> 
> For more information on direct media. 
> http://hive.packetizer.com/users/h323forum/specifications/h323_nat_fw_soluti
> on.pdf
> 
> 
> Simon
> 
> -----Original Message-----
> From: 
> 
> [mailto:]
>  On Behalf Of Christian Helft
> Sent: 02 December 2012 00:51
> To: Simon Horne
> Cc: 
> 
> Subject: Hardware vs software sound processing [was: [Megacon] Message from
> Platinum sponsor (instructions and notes)]
> 
> Thanks a lot Simon. You make very good points and answer some of my long
> lasting interrogations about software vs hardware echo cancellation.
> 
> I doubt that people not bearing hearing door slams appreciate too much
> watching you with a headphone hiding your ears and mouth. And unfortunately
> echo cancellers DON'T cancel door slams (maybe I'm wrong, do they have a
> short range sensitivity like headphone microphones?). For me, pervasive
> videoconferencing requires hands-(or more exactly ears- and mouth-)free user
> experience, period (exit headphone). And my experience is that I'm using
> less and less often an external echo canceller (sometimes I forget to hook
> one, and ... no one complains!).
> 
> Granted, I'm using more Macs than Windows machines, and you're shedding a
> new light on my understanding of the mysteries of software vs hardware
> implementation of echo cancellation. Now I understand why my iPad will
> probably never see an echo canceller during its whole life.
> 
> Beyond your point about Windows being so difficult to master regarding sound
> processing, it's a little bit scary to see that every *application* needs a
> different hardware!
> 
> Regarding Spranto, I gave it a (very short and incomplete) trial a couple a
> weeks ago. I just called the Esnet MCU from a new iPAD (without remote
> participants). The video was often quite pixellated, under conditions where
> other systems (Cisco Jabber, zoom.us, Lifesize softphone) are performing
> quite well.
> 
> Regards,
> 
> Christian.
> 
> Le 1 déc. 2012 à 14:03, "Simon Horne" 
> <>
>  a écrit :
> 
>> Christian
>> 
>> I think you might be a little misdirected. The issue is not the software
> but
>> the curse of the OS, location or just simply the hardware itself.
>> 
>> People in Professional video conferences do not want to get blasted in
> their
>> ear when you receive an email or IM or hear the door slam down the hall or
>> people talking noisily in the other room. That can get really taxing on
> the
>> other participants in the meeting. This is an inherent problem with ANY PC
>> application.  People using Skype might think it is ok to occasionally get
>> loopback,  beep of your IM or external noise but people who are using
>> profession video equipment like Polycom or Tandberg room systems are not.
>> Windows PC's are terrible to get software echo cancellation/noise
> reduction
>> consistently right. Mac and iOS are much much better. For instance Spranto
>> iOS client don't need any such devices it works really well.
>> 
>> For professional video conferencing on PCs, companies like ClearOne offer
> a
>> range of device tailored to PC applications to provide professional
> crystal
>> clear audio like a professional room system.
>> Here are some of their range.
>> For General Softphones
>> http://www.clearone.com/products_chat50
>> For Skype
>> http://shop.skype.com/phones/speakerphones/clear-chat-60/
>> For Lync
>> http://www.clearone.com/products_chat70u
>> 
>> I highly recommend the Logitech BCC950. It comes with a 1080p camera and
>> high quality speakerphone and a PTZ that can be controlled remotely during
> a
>> call.
>> http://www.logitech.com/en-us/product/Conferencecam?crid=1252
>> We will be giving one of them away as part of our package at megacon.
>> 
>> Simon
>> 
>> -----Original Message-----
>> From: 
>> 
>> [mailto:]
>>  On Behalf Of Christian Helft
>> Sent: 01 December 2012 16:33
>> To: Simon Horne
>> Cc: 
>> 
>> Subject: Re: [Megacon] Message from Platinum sponsor (instructions and
>> notes)
>> 
>> Hi,
>> 
>> Le 30 nov. 2012 à 23:15, "Simon Horne" 
>> <>
>>  a écrit :
>> 
>>> 2.       Recommend Headphones or echo cancelling speakerphone. (for
>> everyone’s sanity)
>>> 
>> 
>> Can you comment on the fact that any reasonably recent videoconferencing
>> "softphone" (you name it: VSee, zoom.us, Jabber video, Messages (for Mac),
>> Vid, etc. (even Skype)) doesn't require external echo cancellation any
> more,
>> while there seems to be a malediction with truly H323 compliant systems?
>> 
>> Regards,
>> 
>> Christian. 
>> --
>> Christian Helft LAL - IN2P3 - CNRS
>> LAL Bat 200 BP 34 91898 Orsay CEDEX France
>> +33 1 64 46 84 12
>> Mobile +33 6 32 90 97 09
>> http://www.lal.in2p3.fr
>> 
> 
> --
> Christian Helft LAL - IN2P3 - CNRS
> LAL Bat 200 BP 34 91898 Orsay CEDEX France
> +33 1 64 46 84 12
> Mobile +33 6 32 90 97 09
> http://www.lal.in2p3.fr
>