[linux-dvb] Re: DVB-C it got worse!

[Holger Waechtler <holger@convergence.de>]

Hi Micael,

Micael Beronius wrote:
> HW> Hmm, here were two explicit parentheses missing, I added them. Please
> HW> update your sources and tell me if it works now...
> 
> OK, I have found the main problem (apart from the missing
> parenthesis). The last reset command sequence does not set CLB bit to
> one after soft reset.
> 
> It should be
>   ves1820_writereg (fe, 0x00, reg0 | 0x01 );
>                                    ^^^^^^
> 
> But I do not really think that there is a need to reset the chip up to
> two times each call to ves1820_setup_reg0(). I would like to take, at
> least, the last reset calls out.

It is. When you want to see if your changed parameter set causes a 
frontend lock you have to apply this parameter set by toggling CLB.


> I also for now have made the function to use auto inversion only if so
> specified, since I feel it may work a bit better, but I'll test things

This is not possible. If the I/Q pins (and so the meaning of the 
inversion bit) are reversed or not depends on the used hardware. The 
ves1820 based cable version of the dbox2 has the PLL wired correctly, 
and so the meaning of the inversion bit is different from the one of the 
Hauppauge DVB-C cards.


> a bit more, primarily I will use the last cvs with the above patch to
> see if I feel it is good enough. My gut feeling says that it is not as
> good as the manual inversion setting when the zig-zag tuning starts,
> which by the way always happen to me if I stay at the "correct"
> frequency ( 322000000 for example always zigzags down to something
> like 321625000 ).

This phenomenon should not be caused by the autoinversion code. In 
Finland and here in Germany are some tronsponders offset'd by +-125kHz 
or +-250kHz too.


> By the way, I have also seen that I always get 2 outcommand error 1 when the
> Convergence copyright are scrolling down, it seams that the arm is busy with
> this little (unnecessary) scroll.
> 
> 
> 
> This is what I use now, for reference:
> 
> int ves1820_setup_reg0 (struct dvb_frontend *fe, u8 reg0, fe_spectral_inversion_t inversion)
> {
>     int autoinv = 0;
> 
>     reg0 |= GET_REG0(fe->data) & 0x42;   // Start without inversion
> 
>     switch (inversion) {
>     case INVERSION_OFF:      /* XXX FIXME: reversed?? p. 25  Yes, this is reversed! /MBe */
>         reg0 |= 0x20;
>         break;
>     case INVERSION_ON:
>         break;
>     case INVERSION_AUTO:
>         autoinv++;
>         break;
>     }

I won't apply this, see above.



>     ves1820_writereg (fe, 0x00, reg0 & 0xfe);  // enter reset
>     ves1820_writereg (fe, 0x00, reg0 | 0x01);  // leave reset
>
>         /**
>          *  check lock and toggle inversion bit if required...
>          */
>     if ( autoinv )  {
>         if (!(ves1820_readreg (fe, 0x11) & 0x08)) {  // Locked?
>             ddelay(1);
>             if (!(ves1820_readreg (fe, 0x11) & 0x08)) {  // Locked?
>                 reg0 ^= 0x20;
>                 dprintk ("ves1820_setup_reg0: Setting Inversion bit to %d\n", (1 == (reg0 & 0x20)) );
>                 ves1820_writereg (fe, 0x00, reg0 | 0x01 );

don't we have to clear the CLB bit here before we set it again? I'll add 
this to the driver...


>             }
>         }
>     }
>     SET_REG0(fe->data, reg0);
>     return 0;
> }

many thanks for your efforts, I'll apply the fixes to CVS, please test 
and report if everything is working as expected,

Holger



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