new TV Tuner Chip, the Si2177

Leif Asbrink leif at sm5bsz.com
Tue Sep 24 13:57:24 UTC 2013


Hello Jürgen,


> That's funny. Charged raindrops causing pulses.
> But it makes sense that a higher sampling rate can simply
> removal of such disturbances. A sampling rate of 5 to 10 MHz
> is easily implemented with the PCIe-9842. My first trials with my
> own software used 5 and later 20 MHz. At the moment, my
> recordings use 200 MHz by default. Thanks to the PCI express
> interface and a good driver software, the transfer of 400 MB/s
> puts no load onto the CPU. It is only the signal processing that
> causes a high computational burden onto the hardware.
Very interesting:-)

Linrad can process 8 MHz bandwidth when run without the noise 
blanker on my now old super-computer, an Intel D5400XS board
with two 2 Xeon E5410 CPUs and 8 cpu cores while producing decent
filtering.

With the noise blanker enabled it can do 4 MHz. 
If I apply less good filters Linrad can do at least twice 
the bandwidths.

The limitation is the time to do FFT. I guess that could be 
moved to a GPU but I have no idea about how difficult it might be.

Linrad goes back and forth between the time and the frequency
domain. Four FFT computations are done after one another
for the noise blanker and they are placed in separate threads.
The blanker FFTs are double but placed in the same thread
and for that reason the blanker fills up a core at half the bandwidth.
That should not be too difficult to fix but I never had any reason.
I have not yet had access to any high speed hardware... 

> ADLINK is a taiwanese company, as far as I know.
> The Swedish company you mean is probably Strategic Test in Stockholm.
> They offer excellent cards with Linux support, but too expensive for me.
No, it was another Swedish company. I called them today but
the price they asked was pretty expensive so I did not order.

> I have downloaded the source code of Linrad and had a short look at it.
> Why don't you put the source code into a Subversion repository at SourceForge ?
> Storing the source code there costs nothing and simplifies access.
That is because I have no idea how to do it.
I do not want to spend time on learning those kinds of things.
I have several far more exciting things on my agenda:-)

> It looks like there is a systematic way of integrating new devices into
> the Linrad source code. If I tried to integrate the PCIe-9842, would
> you consider my source code for acceptance to  into Linrad ?
Absolutely. I would be delighted:-)
If you report sucess I would also want to buy such a card even
if I would have to pay more than I really want for it. On the
Internet I find the price to be something like 2000 Euro. 

Regards

Leif





More information about the osmocom-sdr mailing list