[vlc-devel] Re: Is there a way to force PES alignment to a TS boundary when Streaming UDP?

Måns Rullgård mru at inprovide.com
Thu Mar 16 03:14:28 CET 2006

"Robert Scheffler" <Bob at pricom.com> writes:

>>> Sorry about the email tread screw-up, my mistake, I won't do it again!

Could you also not top-post, please.

>>> The requirement is that every access unit in the video elementary
>>> stream should be aligned to a PES header. That is the first byte of
>>> each PES packet payload is the first byte of an access unit. The
>>> data_alignment_indicator bit should also be set in the PES
>>> header. This type of alignment corresponds to an alignment_type of
>>> two in the data_stream_alignment_descriptor.
>> OK, that's a different type of alignment.  Are you sure this is the
>> problem?  I've never seen any hardware or software requiring such
>> alignment.
> The text was copied from the guy that wrote the receiving code. I
> have to admit this is not my strong point, which is why I had
> formulated the question wrong in the first place. I asked that the
> developer pose the question so I could pass it to this fine group.

Making a device/program that is incompatible with 99.98% of all
streams seems like a bad idea to me.

> Ideally, I would like to use a capture card and send MPEG2-TS/UDP/IP
> to this device in question that conforms to the rule stated
> below. All works well except that at random intervals (5 secs to 30
> minutes) the alignment goes out enough to crash the receiver. The
> receiver doesn't give much more than a "PUSI_SET_PES_HEADER_OVERFLOW"
> message.

Perhaps you could ask the developer to explain under what conditions
it produces this message.  I assume PUSI is referring to the
payload_unit_start_indicator bit in the TS packet header.  This bit
has nothing to do with the data_alignment_indicator in the PES header.

> I don't have any way to look at the TS/UDP/IP packets so I really
> can't tell what is going on. I was hoping someone here would be
> better armed to help.  If there are any suggestions or tests I can
> perform, please let me know.

Getting a sample that triggers the problem would be helpful.  There
must be some way you can capture the stream that is going out.  If
nothing else works, dumping the outgoing IP packets (e.g. with
tcpdump) should be possible.

Måns Rullgård
mru at inprovide.com

This is the vlc-devel mailing-list, see http://www.videolan.org/vlc/
To unsubscribe, please read http://developers.videolan.org/lists.html

More information about the vlc-devel mailing list