[opendtv] Re: Lip sync problem resurfaces
- From: Eory Frank-p22212 <Frank.Eory@xxxxxxxxxxxx>
- To: "'opendtv@xxxxxxxxxxxxx'" <opendtv@xxxxxxxxxxxxx>
- Date: Tue, 20 Apr 2004 00:59:43 -0700
Comments interleaved.
>
> From: jeroen.stessen@xxxxxxxxxxx
> Subject: [opendtv] Re: Lip sync problem resurfaces
> Date: Tue, 20 Apr 2004 08:38:55 +0200
>
>
>
>
>
> Hello all,
>
> Frank Eory replied:
> > There is no excuse for this. The system that decodes and processes
> > the audio & video has all the tools it needs to maintain A/V sync.
>
> It is not so simple !
>
> It is simple enough to delay the
> audio IF you process it, but that is not always the case anymore.
This was sort of my point -- the audio & video should be processed by the same
device, so that latencies can be matched. If you are building a display-only,
perhaps the audio should be required to loop through the display, so that the
correct delay can be applied to it to match the video delay.
>
> I have once read a document in which it was stated that (IIRC) the
> human tolerance for audio delay is between -30 and +80 ms relative
> to the video. In other words: we tolerate better that the audio is
> late than that the audio is early. Unfortunately, the typical
> latency of even a minimal amount of video processing is already in
> the order of 30 ms.
I think the negative delay (audio leading video) is even tighter than 30 ms.
Due to the slow speed of sound relative to the speed of light, it seems that
our brains are comfortable with a small amount of audio lag, since this occurs
in the natural environment. But when audio leads video, even by a small amount,
it is very unnatural and disturbing.
> We would be greatly helped if players and set-
> top boxes would apply a nominal delay of e.g. 30 ms to the audio.
> This would be well within the tolerable limits for a zero-delay
> (i.e. plain CRT) display, and it would give other displays an extra
> margin of 30 ms for high-latency video processing. Unfortunately,
> everybody do their best to achieve zero delay difference, and then
> it is the display processing that makes the delays unequal again.
So the audio should be deliberately delayed at the STB outputs, thus skewing
A/V sync for CRTs? I think that's a recipe for trouble. Everyone knows what to
do if audio & video are delay-matched at every component interface.
>
> > If that system is embedded in the display and there is an A/V
> > sync issue, I call that a major design flaw.
>
> It is a flaw, but you tell me how to avoid it ?!
>
Require audio loop-through, so you can delay audio appropriately.
> > If that system is in an external STB, and the display does not
> > provide a mechanism for bypassing its internal "value-added"
> > video processing, I call that a major marketing flaw.
>
> We have spent countless manyears in designing low-latency video
> processing. Life would have been a lot easier if a higher latency
> were allowed. But for the reasons stated above, it is not possible.
> Also, some of this video processing can not be bypassed. For a
> progressive matrix display, it is not an option to bypass the de-
> interlacer.
An HDTV or EDTV set-top box should include a good deinterlacer, frame-rate
conversion, any-format-in-any-format-out decoding & scaling, and again, should
have audio & video delay-matched at its outputs.
> The plasma and micromirror displays
> have essential frame memories for creating the sub-fields for their
> digital pulse width modulation. The liquid crystal display has an
> inevitable latency because the molecules need time to rotate to a
> new state. And with a temporal aperture of a whole frame period,
> even the fastest LCD would have an average latency of half that
> period. This will also be true for continuously lit OLED displays.
> (Some latency necessarily comes with reduction of field flicker.)
>
> As I said: life would be a lot easier if audio came with a nominal
> delay of e.g. 30 ms. Almost no modern display has zero latency
> anymore. Only the classical (50-60 Hz) CRT can be "accused" of that.
Due to the variability of video latency through different modern displays (and
video processing circuits within), even a nominal 30 ms audio delay from the
STB does not solve the problem. The only solution is audio loop-through, so the
display can clean up its own lip sync mess, so to speak.
-- Frank
----------------------------------------------------------------------
You can UNSUBSCRIBE from the OpenDTV list in two ways:
- Using the UNSUBSCRIBE command in your user configuration settings at
FreeLists.org
- By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word
unsubscribe in the subject line.
Other related posts:
