2007/10/9, Manu Abraham <[EMAIL PROTECTED]>:
> jlacvdr wrote:
> > 2007/10/9, Manu Abraham <[EMAIL PROTECTED]>:
> >> jlacvdr wrote:
> >>> I rewrote a part of stb0899_dvbs2_algo( ) to remove duplicate source
> >>> code and to get a source code more readable.
> >>> I added also re-initialization of some registers.
> >>> The result is a lower probability to don't obtain the lock.
> >>>
> >>> I tested on TT-s2 3200 and on Knc1 dvb-s2 plus.
> >>> It must work on other cards, the patch touch only stb0899 part.
> >>>
> >>> I don't remember the time duration to lock, but if you want I can measure
> >>> it.
> >>>
> >>> The aim of this patch, it's obtain the lock all the time.
> >>>
> >>> Regards,
> >>>
> >> I doubt whether there is a problem in acquiring LOCK on the KNC1, Satelco
> >> and the TT cards.
> >
> > without the patch, I can't lock all the time, and it's seems that
> > other persons have the same problem.
> >
>
> Some cards, yes. What transponder where you trying to tune with KNC1 card and
> having issues ?
on Astra 19.2E: 12581,12721,12522
and HotBird 13.0E
>
> >
> >> Julian, Marco can you guys please confirm ?
> >>
> >>
> >>
> >> Let me walk through your patch ....
> >> (I feel something wrong in the patch)
> >>
> >>
> >> --- linux/drivers/media/dvb/frontends/stb0899_algo.c.org 2007-10-08
> >> 21:25:09.000000000 +0200
> >> +++ linux/drivers/media/dvb/frontends/stb0899_algo.c 2007-10-08
> >> 21:25:06.000000000 +0200
> >> @@ -1351,7 +1351,17 @@
> >>
> >> s32 offsetfreq, searchTime, FecLockTime, pilots, iqSpectrum;
> >> int i = 0;
> >> + int cpt = 0;
> >> u32 reg, csm1;
> >> + s32 retry = 0;
> >> +
> >> +
> >> + reg = STB0899_READ_S2REG( STB0899_S2DEMOD, DMD_CNTRL2 );
> >> + STB0899_SETFIELD_VAL( SPECTRUM_INVERT, reg, 1 );
> >> + stb0899_write_s2reg( state, STB0899_S2DEMOD, STB0899_BASE_DMD_CNTRL2,
> >> STB0899_OFF0_DMD_CNTRL2, reg );
> >>
> >>
> >>
> >> Why do you forcefully invert the I/Q inputs ? Different hardware have
> >> different configurations.
> >
> > yes, there are different hardware configurations, so with this patch
> > the algo find automatically the good configuration.
> > This is an initialiation, I begin with a I/Q configuration, if failure
> > the configuration is inverted.
>
>
> Q) Does it work for you, if you started with SWAP_OFF, rather than SWAP_ON,
> retaining your logic. Or do you see any other issues/degradation when you
> change
> the initial INVERSION state ?
>
> Currently there will be a case of inversion in your logic, but for the
> hardware that we
> have and given the networks, the chances of finding an inversion is much
> less. Can
> you please check whether INVERSION_OFF in the initial case does work better
> still, if
> not worser ?
I think the resultat is the same, perhaps more slow to lock (when 2 try needed).
I test it soon.
>
>
> >>
> >> +
> >> + do
> >> + {
> >>
> >> if (internal->srate <= 2000000) {
> >> searchTime = 5000; /* 5000 ms max time to lock UWP
> >> and CSM, SYMB <= 2Mbs */
> >> @@ -1381,9 +1391,16 @@
> >> STB0899_SETFIELD_VAL(FRESRS, reg, 1);
> >> stb0899_write_reg(state, STB0899_TSTRES, reg);
> >>
> >> + reg = STB0899_READ_S2REG( STB0899_S2DEMOD, CRL_NOM_FREQ );
> >> + STB0899_SETFIELD_VAL( CRL_NOM_FREQ, reg, 0xd0 );
> >> + stb0899_write_s2reg( state, STB0899_S2DEMOD,
> >> STB0899_BASE_CRL_NOM_FREQ, STB0899_OFF0_CRL_NOM_FREQ, 0xd0 );
> >> +
> >>
> >>
> >>
> >> You read Nominal Frequency into "reg", set 208 Hz into "reg", ignore reg ..
> >> After which you forcefully enter 208 Hz into the Complex mixer straight
> >> away.
> >> Even if you required to reprogram the mixer to 208 Hz, i don't see then why
> >> do you need the Complex mixer current setting as a part of the Carrier
> >> Recovery ?
> >>
> >> But, i don't see this "magic frequency" in any of the specs from STM.
> >>
> >> Although one question i have, irrespective to your patch, does setting the
> >> Complex Mixer to "0" Hz (30 bit unsigned) help in your case, as applicable
> >> to you ?
> >>
> >>
> >
> > Before starting acquisition, I restore the initial conditions.
> > The 208 is the initial value of nominal frequency in CRL_NOM_FREQ
> > register. (can be found is the datasheet)
>
> Ok. reasonable, missed that one. Thanks for pointing it out. Will test it
> out, but
> please do see the first question that i raised.
>
> Manu
>
>
_______________________________________________
linux-dvb mailing list
[email protected]
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb
