Re: [GIT PULL for v3.11] media patches for v3.11

[Manu Abraham]

On Mon, Jul 1, 2013 at 9:05 PM, Mauro Carvalho Chehab
<mchehab@xxxxxxxxxx> wrote:
> Em Mon, 1 Jul 2013 16:37:58 +0530
> Manu Abraham <abraham.manu@xxxxxxxxx> escreveu:
>
>> Mauro,
>>
>> On Mon, Jul 1, 2013 at 4:28 PM, Mauro Carvalho Chehab
>> <mchehab@xxxxxxxxxx> wrote:
>> > Hi Linus,
>> >
>> > Please pull from:
>> >   git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media v4l_for_linus
>> >
>> > For the media patches for Kernel v3.11.
>> >
>>
>> >
>> > Zoran Turalija (2):
>> >       [media] stb0899: allow minimum symbol rate of 1000000
>> >       [media] stb0899: allow minimum symbol rate of 2000000
>>
>>
>> Somehow, I missed these patches; These are incorrect. Please revert
>> these changes.
>> Simply changing the advertized minima values don't change the search algorithm
>> behaviour, it simply leads to broken behaviour.
>>
>> NACK for these changes.
>
> While this patch came from a sub-maintainer's tree, looking at its
> history, the patch was proposed here:
>         https://linuxtv.org/patch/18341/
>


Wherever it came from, the patch is incorrect. Anyone can throw in
any patch as they want.


> From what it is said there, with this patch, 6 additional channels
> were discovered when using with Eutelsat 16A, that uses a symbol
> rate between 2MS/s to 5 MS/s. Without this patch, those channels won't
> be discovered, as the core won't try to use a symbol rate outside
> the range.

What you are stating is a hit and miss scenario, sometimes it might lock
and sometimes it wouldn't.

The scanning algorithm that I implemented for the demodulator works
with a symbol rate as low as 5 MSPS alone. Anything lower than that
is hit and miss.


> Of course, transponders with a symbol rate equal or upper than 5MS/s
> won't be affected by this patch.
>


How can you be sure ? I myself am not very sure. While we worked on
the demodulator in the early days, we had different situations where a
previous failed state could cause lockup of the demodulator, eventually
resulting tuning failures.


> Even if this is not a perfect patch and some changes would be
> needed to improve tuning for those low symbol rate transponders,
> it seems better than before, as at least now some channels are tuned.
>
> The only reason I can see to reverse this patch is that if setting
> the frontend to low bit ranges could damage the frontend or could
> hit some bug on the hardware (or internal firmware).
>
> Yet, from the datasheet pointed by the patch author, it seems that
> this frontend allows such low symbol rates:
>         http://comtech.sg1002.myweb.hinet.net/pdf/dvbs2-6899.pdf


The frontend allows a different lower symbol rate with a different
scanning algorithm, not with this existing current one.

I am pretty sure, that author saw some specifications written some
place and simply copied those numbers in here. Also sure that he
has no idea about the algorithm in use.

According to ST itself, a 2MSPS algorithm was created for a very
specific customer requirement, which is not applicable to the existing
algorithm in use with the Linux STB0899 demodulator driver.


Regards,
Manu
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/