[linux-dvb] Re: [vdr] EPG scan cause high load and AC3 drop outs - maybe not solved yet ?

["Christian Jacobsen" <christian.jacobsen@stageholding.de>]

Hallo,

> On Thu, Mar 27, 2003 at 08:58:52AM +0100, Christian Jacobsen wrote:
> >>
> >> While using VDR 1.1.26 I Yesterday saw a kdvb-fe process using 29% CPU
> >Time
> >> again (Pentium3 933).
> >>
> >> I used "top" so it is only snapshots of what really happens. Did not
> >watch
> >> it that long - but Sometimes it uses less that 1% and sometimes it uses
> >29%.
> >> Is there a better tool for watching such things ?
> >> As far as i remember the exact name of the process was kdvb-fe1:0 (or
> >> kdvb-fe-1:0)
> >> IIRC this is the process that switches channels ? Does it do other
> >things to
> >> ?
> >>
> >> Could it be involved in the "AC3-processing" - so that the
> high IRQ load
> >> also causes this high load on kdvb-fe ?
> >>
> >> Is 1:0 somehow related to the card it is running on ?
>
> see driver/dvb_frontend.c at dvb_frontend_thread() ...
>
> [...]
>         snprintf (current->comm, sizeof (current->comm), "kdvb-fe-%i:%i",
>                   fe->frontend.i2c->adapter->num, fe->frontend.i2c->id);
> [...]
>
> ... IMHO the first is the card used on the i2c bus (1 == the second), the
> second number is the i2c bus its self. What I2C is for see
> /usr/src/linux/Documentation/Configure.help , search for CONFIG_I2C, and
> /usr/src/linux/Documentation/i2c/summary .
>
> I2C is a slow dog but is used for programming/reading registers of the
> DVB cards. IMHO there are also missed some
> request_mem_region()/release_mem_region() within av7110/saa7146_core.c
> also a better wait mechanism as mdelay().  Maybe the patch attached
> will help to for this.
>
> The thread dvb_frontend_thread() calls dvb_frontend_recover()
> which is used
> for compensating LNB drift.  IMHO this should not happen (nevertheless it
> could be the first reason of the high load!!).
> Beside this, this is what I've found by searching for lost_sync_count:
>
> /**
>  *  if 2 tuners are located side by side you can get interferences when
>  *  they try to tune to the same frequency, so both lose sync.
>  *  We will slightly mistune in this case. The AFC of the demodulator
>  *  should make it still possible to receive the requested transponder
>  *  on both tuners...
>  */
>
> ... and this could also explain why users of two or more DVB cards (full
> featured or not) are `able' to see the phenomena you've described during
> EPG scans.
>
> IMHO the ml linux-dvb@linuxtv.org should notice that (therefore a cross
> posting is needed IMHO:). For readers of linux-dvb@linuxtv.org: the
> drop outs are happen during live view of Pro7 which means that the TS
> stream has longer timeouts beginning with some bit failures which are
> found by a CRC check of a VDR plugin forwarding the AC3 data stream
> to the S/P-DIF of a sound card.
>
>
>         Werner
>
>
> -- Attached file included as plaintext by Listar --
> -- Desc: dvb-20030326-driver.dif

I tried the patch but unfortunatly it did not change the AC3/DD dropout
situation :(

Configuration :
VDR 1.1.26
ALSA 0.9.2
Bitstreamout 0.44d
DVB 1.0.0-pre2
DVB CVS 200326 with included patch from this mail.

Hardware :
CMI8738 Soundcard (Nightingale 4.1 with external SPDIF out)
1 DVB-s rev. 1.3
1 WinTV Nova-s model 541
Intel i815 Maniboard

Greetings
Christian




-- 
Info:
To unsubscribe send a mail to listar@linuxtv.org with "unsubscribe linux-dvb" as subject.