Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: [RE] Time sensitivities in ResE. Part 2: Latency



Your 100ms control response figure is in line with my experience. As you've
pointed out, there are many systems that stretch this. Sluggishness does
bother people but apparently not enough to get them to open their wallets
and address it.

For case 2 we need to decide if we're addressing the special case of a
singer or DJ wearing headphones or in-ear monitors. For that case, anything
above 0.5ms is noticeable. Since A/D and D/A conversion itself consumes over
1ms, we'll not be able to make anyone happy here.

Musicians are most comfortable playing in reverberant spaces so latencies on
the order of 5-50ms are what we're shooting for. A paper presented at the
117th AES convention, "Network Time Delay and Ensemble Accuracy" (#6808)
corroborates this. The finding here is that a constant latency of around
20ms is most conducive to ensemble performance. Shorter latency tended to
induce tempo acceleration.

-----Original Message-----
From: owner-stds-802-3-re@IEEE.ORG [mailto:owner-stds-802-3-re@IEEE.ORG] On
Behalf Of Michael Johas Teener
Sent: Tuesday, April 12, 2005 6:22 PM
To: STDS-802-3-RE@LISTSERV.IEEE.ORG
Subject: [RE] Time sensitivities in ResE. Part 2: Latency

Once again, there are several reasons why latency matters in a
media-streaming network.

1) Control response ... in a normal CE system, the latency of most
importance is that between the time a consumer presses a button the remote
control and the time the appropriate response is perceived by that same
consumer. Response times of current consumer devices is on the order of 0.2
to 0.5 sec. Some devices take longer, but are disliked (controls on an
airplane entertainment system are typically pretty awful ... try them
sometime). Unfortunately, much of the response time is consumed by the CE
devices themselves, and any significant addition caused by the network would
be unacceptable ... so let's say that the network should hold control
latency to less than 100ms. Not too bad.

2) Perceptual feedback ... in system where the network is part of a musical
instrument or gaming feedback, the requirements are much more severe.
Average people start noticing problems at 8-10ms, and some professionals
much sooner. As a network latency, this is not much of a problem, but the
network isn't the only consumer of latency: in a typical system, the
keyboard or guitar uses some, the audio processor (mixer/effects) uses some,
and the amplifier uses some ... Alexei described this process in SG
presentations some time ago, and came up with the 2ms figure for maximum
network transit delay (did I get that right, Alexei?).

This last one is particularly significant, since this latency figure is
absolute, not statistical (any late data will cause glitches in the sound
since we can't do much buffering here).

-- 
---------------------------------------------------------------
         Michael D. Johas Teener < mikejt@broadcom.com
office +1-408-922-7542 cell +1-831-247-9666 fax +1-831-480-5845
http://public.xdi.org/=Michael.Johas.Teener - PGP ID 0x3179D202
---------------------------------------------------------------