[vdr] Re: ANNOUNCE: Softdevice plugin. Use /dev/fb for output

[Holger Waechtler <holger@convergence.de>]

Hi,

Roland Praml wrote:
> Hello
>
> I've uploaded a new version. This one should also compile with gcc 3.2
> (I hope)
>
> http://www.k13zoo.de/vdr/vdr-softdevice-0.0.2.tar.bz2
>
> Comments are welcome.
I had only a short look over your code, but there came three things to 
my mind.

I'm not sure whether your GET_MPEG2_PTS() macro is correct, I'd write it 
like this:

#define GET_MPEG2_PTS(x)   ( ((uint64_t)x[4]&0xFE) >>1     | \
                             ((uint64_t)x[3])      <<7     | \
                             ((uint64_t)x[2]&0xFE) <<14    | \
-                            ((uint64_t)x[1])      <<22    | \
+                            ((uint64_t)x[1])      <<21    | \
                             ((uint64_t)x[0]&0x0E) <<29 )

(x[1] contains PTS[29:21], so you need to shift it 21 bits to the left, 
not 22 bits)

Then you have to synchronize your local clock to the Stream clock in the 
playout center using the Program Clock Reference announced in the TS 
packet header on the PCR PID. See ITU-T H.222.0 or ISO IEC 13818-1 for a 
detailed explanation and implementation suggestions.

Since you usually can't adjust the clock on your sound card you need to 
resample the audio stream before sending it to the sound card, if the 
stream from air is arriving faster or slower than you can decode it you 
get buffer overruns or underruns otherwise.

ffplay.c is just inserting or removing dummy samples, FusionSound (part 
of the DirectFB project) is doing an interpolating resampling based on 
the pitch floating point parameter, 1.0 means no resampling. Not too 
hard to implement, you need probably something similiar.

Holger



--
Info:
To unsubscribe send a mail to ecartis@linuxtv.org with "unsubscribe vdr" as subject.