Logitech C920's H264 over RTP

Jesper Larsen jesper.larsen at ixonos.com
Tue Jul 23 23:09:21 PDT 2013


On Wed, Jul 24, 2013 at 5:36 AM, Adam Goodwin <adam.goodwin.nz at gmail.com>wrote:
>
>
> First of all, the sending pipeline doesn't work without the h264 parser,
> but I don't really understand why. If the camera is outputting H264 format,
> isn't that all that the payloader should need?
>

> Second, I've set the config-interval of the payloader to 4. As I
> understand it, this should mean that if I run the receiver at any time
> after the sender has already started, it should take at most 4 seconds for
> the stream to be recognised. However, it ends up taking more like 8-10
> seconds before I start to see the video at the receiver. What am I missing?
> In practice is there a lower limit at around 8-10 seconds?
>

You need the h264parser because it will take care of properly extracting
the NAL units from the h264 stream. With config-interval you set the
interval of when the payloader sends SPS/PPS information about the stream.
SPS/PPS are embedded in the h264stream, so the payloader is a bit more
intelligent then just putting a header on some data.

I'm not actually sure what the payloader does in the case, that the SPS/PPS
information in the h264 stream has a greater interval than the
config-interval. My guess is that it will just re transmit the last one it
got since SPS/PPS would not change unless the source/encoder changes.

However, the decoder still needs a key frame (I frame) before it can start
decoding. So if the camera doesn't output key frames at at least 4 seconds
interval that will cause further delay.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/gstreamer-devel/attachments/20130724/68f20989/attachment.html>


More information about the gstreamer-devel mailing list