[asterisk-users] Best Practices: Empirical measure of call latency

[Steve Edwards]

> On Tue, 01 Jul 2008 22:40:20 -0500, "Michael Graves" <mgraves at mstvp.com>
> said:
>> On Tue, 01 Jul 2008 17:57:31 -0500, asterisk-users at kfife.mailworks.org
>> wrote:
>>
>>> I would like to hear your favored method to obtain an empirical measure
>>> of latency in the media path.
>>
>> I had a project not long ago where I thought I was going to have to
>> make  a comparison between the latency presented by two different call
>> paths. In the end it wasn't necessary, but it did get me thinking about
>> what I could do, lacking for any special equipment.
>>
>> I had thought that I'd locate an echo test on a remote server. Free
>> World Dialup still runs one that's accessible by both SIP and IAX2. My
>> hosted PBX provider has one accessible via PSTN or SIP.
>>
>> Then I'd use a mechanical click generator (impulse) at the handset
>> while recording the call. Then take the recording into a waveform
>> editor software and measure the timing differences between the various
>> paths.

On Wed, 2 Jul 2008, Karl Fife wrote:

> I like your idea Michael.  Is the increment of delay of the echo service
> known?  I suppose you'd have to back that out of the measurement.
>
> I was thinking of something similar (using audio editing software to
> measure time between 'clicks') but more kludgy than your idea -- my idea
> was to test the services in the form of LOOPS so I could HEAR the delay
> myself.  Then the idea was to mesure the time between the first click
> and the return click.
>
> I imagine that someone out ther must have created a more automated way
> to do this.
> Maybe the best reasons to have it automated would be to test for
> variance over time.

Several years ago, I replaced an aging Dialogic based adult chat system 
with Asterisk.

One day I made the mistake of letting the client listen to the system with 
a handset on each ear. The delay was noticeable and the client was a 
stickler. (Note to self - NEVER let a client do that :))

I used 2 RadioShack 43-228A telephone recording controls ("left" and 
"right") feeding into a "Y" and then into my laptop. Using Audacity, I 
would tap the "left" handset and you could measure the delay until the tap 
registered on the "right."

Only by demonstrating that the measured delay was below what several 
studies showed the threshold for interfering with conversation saved the 
project.

That and demonstrating the delay in a cell to cell based call with a 
handset on each ear.

Thanks in advance,
------------------------------------------------------------------------
Steve Edwards      sedwards at sedwards.com      Voice: +1-760-468-3867 PST
Newline                                             Fax: +1-760-731-3000