[linux-dvb] Re: Scanning satellite for programs?

[Johannes Stezenbach <js@linuxtv.org>]

On Mon, Nov 22, 2004 at 04:46:48PM +0100, Josef Wolf wrote:
> This looks to be O(n^2) on the first glance.  But I think there is pretty
> much room for optimizations (please correct me if I write nonsense):
> 
> 1. You need not to probe _every_ frequency.  As soon as you find any
>    arbitrary transponder, the information about other transponders can
>    be read from there.

In theory yes, but in practise the NITs are often incomplete so it
isn't sufficient to find one TP to acquire the whole network :-(

> 2. There seem to be a lot of tolerance on the frequency.  I just tried to
>    check the tolerancy: With szap, I can tune for "superl rtl" on the range
>    12171..12204.  This is a frequency range of about 34MHz.  This might
>    vary with the signal quality (size of the dish, single/multi-feed),
>    but I would still expect that probing in 10 or 20MHz steps would be
>    sufficient.

The driver does zig-zag scan, but for frequency scanning this is
actually a bad thing. We'd need an option to disable this and
let the app control the scan process. The actual range depends
on the demodulator chip and the symbol rate. FE_GET_INFO
should give you the information about the demods capabilites.
(Although I wouldn't count on its correctness.)

> 3. The frequencies (at least on astra) seem to be assigned in about 41MHz
>    steps in the hi-band and in about 166MHz-steps in the lo-band.  So when
>    probing in 10MHz steps, there should be a pretty high  probability to
>    find a transponder after about 5..20 probes.

Not true. For satellites there is no such rule. For DVB-C and DVB-T
the frequencies are often the same as the old analog equipment used,
i.e. CCIR-something in Germany. Xawtv comes with many country
specific lists.

> 4. There seem to be some standard numbers for the symbol rate. 70% of the
>    entries in my channels.conf have 27500.  12% have 22000.  So probing only
>    for those two symrates would probably speed up scanning further.

There is no standard symbolrate. Astra 13.0E apparently uses the same
equipment for many TPs so they use mostly these two symbolrates.
You cannot count on this for other satellites. Some set-top-boxes
I've seen allow you to manually specify a list of symbolrates for
the scan, as this is apparently the only thing one can do.

> So the symrate is some sort of baudrate?  Does a list of all the valid
> symrates exist?  Would it be to autodetect the symrate?

Some demods (e.g. the Zarlink ones) have support for scanning
and can find out the symbolrates. Someone proposed an API for
that (search the list archives). Most demods will give you
no usable information when you guessed the symbolrate wrong,
so in general symbolrate scanning won't work unless the
demod has support for it.


Johannes