[e2e] Re: crippled Internet

Vernon Schryver vjs at calcite.rhyolite.com
Wed Apr 25 22:27:09 PDT 2001

> From: Randy Bush <randy at psg.com>

> ...
> so the jitter characterization, "50ms or so of frequent variation," would
> seem to be the interesting issue.  what is "frequent variation," and how
> should it be measured?  i.e. if 3% of the samples are outliers, how they
> are distributed in time would seem to be critical.

Beware of the old IBM token ring and HP 100VG-AnyLAN calumny that 50
or even 100 ms of jitter is fatal for voice traffic and therefore it
is absolutely impossible to use 10 MHz or even 100 MHz Ethernet for
"multimedia."  If network stuff consisted of coils and other passive
components, that stuff might make sense, but we've got computers that
can create as much de-jitter buffering as users will tolerate delay.

More of the token ring salescritters' standard nonsense is that total
delay including de-jitter buffering must not be more than 20 or 30 ms.
The speed of sound is about 1 foot per millisecond, so that if 50 ms were
a problem, people wouldn't be able to talk across a large room.

I've found an old note from Al Varney who wrote a previous cycle of
the many of this old discussion in 1994:

] need echo control.  On the other hand, 2 satellite hops is >500 ms, which
] exceeds the ITU-T G.114 limit of 400 ms one-way delay.)
] ...

]    According to G.114, one-way delays up to 150 ms should cause no problems
] with typical voice communications.  From 150 to 400, loss/noise/etc. become
] a bigger issue and "long-delay" echo cancelers are needed.  (But some form
] of echo canceler/suppression is needed on ALL delays beyond 25 ms.)
] ...

]   So I'd summarize by saying that human factors in dealing with one-way
] delay sets in at about 150-250 ms.  The echo cancellation needs start at
] around 25 ms, but ECs can handle one-way delays beyond 400 ms, assuming
] they are within 30 ms or so of the end they are canceling.

Given the sort of digital processing against echos that is now routinely
done for good speakerphones and the CPU cycles available in modern personal
computers, I wonder if the old worries about echo cancelers are irrelevant.
There's also the fact that everything is digital in VoIP.

Vernon Schryver    vjs at rhyolite.com

More information about the end2end-interest mailing list