Re: Adaptation of Federico La Roccas ISDB-T blocks for DVB-T

[Ralf Gorholt]

    Hello Federico,
    
    I have a lot of difficulties to generate the output and the tags for
    the blocks that follow my TPS Decoder in the flowgraph. That's why I
    had the idea to take a different approach.
    
    The OFDM Synchronization block outputs the payload carriers (1705 in
    2K mode) whereas the Demod Reference Symbols block that normally
    follows the FFT expects 2048 carriers. My question sounds certainly
    silly but would it be possible to fill in the missing carriers with
    zero carriers (amplitude and phase == 0) to get 2048 carriers again
    and thus be able to use the original DVB-T blocks? As far as I know,
    in DVB-T the carriers at the borders are all set to zero (and
    perhaps the one in the middle too). Perhaps this would be too simple
    to be true :-)
    
    Regards,
    
    Ralf

Am 10.12.2021 um 14:41 schrieb Ralf Gorholt:

Hi Federico,

indeed, the "symbol_index" tag that is normally sent for each OFDM symbol is missing. This might cause an unexpected situation for the following deinterleaver block, "Access not within mapped region" according to valgrind.

I will see how I can generate the symbol index for each symbol. The "Demod Reference Signals" block uses the dvbt_pilot_gen object to parse the input data and to generate symbol and frame indices.

As far as I have understood, in your block OFDM Synchronization you have combined the symbol acquisition and the FFT. Then you connect a TMCC decoder that eliminates the TMCC and auxiliary carriers, generates tags and outputs the data carriers. My idea was to do the same for the TPS signals in DVB-T.

To my surprise, in principle what I have done seems to work. If not, my TPS decoder would not be able to correctly decode the TPS information that changes accordingly when I change settings in the transmitter, for instance the modulation scheme.

I will focus on the tags and see what is missing.

Regards,

Ralf

Am 10.12.2021 um 14:25 schrieb Federico 'Larroca' La Rocca:

Hi, 

I'd be more than happy to help. A couple of things that come into my mind. 

The OFDM Synchronization block is a combination of our "old" OFDM Symbol acquisition (for a while now it's been part of GNU Radio) and Sync and Channel estimation blocks (which performed equalization and integer frequency correction) . The most important difference is that OFDM Synchronization includes a loop with the estimated channel gains, which in turn is used to estimate the sampling error (plus fine frequency errors). It also indicates some events downstream via tags, just like the older blocks. This new "DVB-T OFDM Synchronization" block should then be a combination, if I'm not mistaken, of OFDM Symbol Acquisition plus Demod Reference Signals (I'm sure Ron will know more on this). 

Anyhow, my point is that you should take a look at the OFDM Symbol Acquisition and Demod Reference Signals blocks in GNU Radio, and check which tags are used and when. Maybe this lack of tags is generating an unforeseen situation on the downstream blocks which generate the segfault? Furthermore, if I'm not mistaken, the pilots in DVB-T (in particular continuous pilots) are not exactly the same as in ISDB-T. Another possibility is that the Demod Reference Signals block is not equivalent to our Sync and Channel estimation block, and further processing is needed for it to be ready for the DVB-T Demap...

best

Federico

El vie, 10 dic 2021 a las 9:55, Ralf Gorholt (<ralf.gorholt@gmx.de>) escribió:

Hi Vasil,

thank you for your message. As I have no experience with GNU Radio and
command line debugging, your hints may be really helpful. I have
attached the gdb and valgrind output to this email.

In the gdb output thread 27 that receives the SIGSEGV is the DVB-T
"Symbol Inner Interleaver" that comes with GNU Radio, not one of my blocks.

As far as valgrind is concerned, it tells me for my block OFDM
Synchronization: "Conditional jump or move depends on uninitialised
value(s)". I will see if I can find out which variable is uninitialized
and how I can get rid of this problem.

Kind regards,

Ralf

Am 10.12.2021 um 12:35 schrieb Vasil Velichkov:
> Hi Ralf,
>
> On 10/12/2021 11.52, Ralf Gorholt wrote:
>> Unfortunately, when I deactivate the original flowgraph, it does no
>> longer work and I get a -11 return code.
> The "-11" value means that you got a segmentation fault and the process was kill with signal 11 (SIGSEGV)
>
> https://docs.python.org/3.8/library/subprocess.html#subprocess.CalledProcessError.returncode
> https://www.man7.org/linux/man-pages/man7/signal.7.html
>
> In my opinion the easiest way to debug segfaults is to run the flowgraph under gdb and valgrind. Open your flowgraph in the gnuradio-companion and then Generate (F5) but do not Execute (F6). The open a terminal, go where the flowgraph python (.py) file was generated and execute
>
>   gdb -ex run --args /usr/bin/python3 test.py
>
> and then when it stops execute `bt` command in the gdb's shell and provide the full output. To run it under valgrind execute
>
>   valgrind --tool=memcheck /usr/bin/python3 test.py
>
> Adjust the path to your python interpreter and its version if needed.
>
> Regards,
> Vasil