Installing Gnu Radio on the Parallella

[tjt7a]

Below are the steps I used to install GNU Radio on the Parallella from the github source. This install includes the Gnuradio-companion, which is a graphical editor for GNU Radio programs. It is a bit heavy, and may feel slow on the board. It is completely optional, and I recommend running/writing GNU Radio programs from the Terminal instead.

One part of the build process involves wrapping c++ code in Python (SWIG), which is very memory intensive; in fact it requires more memory than the board can sustain. To get around the memory shortage, you can choose to remove Python support, and only support c++ GNU Radio programs, but this is not ideal. In order to get everything running with python support, you need to add additional swap space. What I did was use a USB hub to add swap space in the form of a USB memory stick. I also did the build on an external HDD which was also connected via the hub to reduce the wear on the SD card, increase storage space, and to allow for future image updates without losing my source / built files.

Below are the steps:

Install Dependencies
sudo apt-get install libusb-dev python-qt4 python-numpy libtool g++ python-dev swig libboost-all-dev autoconf libcppunit-dev python-lxml python-qwt5-qt4 libqt4-opengl-dev python-cheetah cmake python-wxgtk2.8 libqwt5-qt4-dev fftw3-dev fftw3 libfftw3-dev libqwt5-qt4-dev python-gtk2 libasound2-dev gcc-multilib

Grab Source, Build, and Install
>cd to src directory>
git clone https://github.com/gnuradio/gnuradio.git
cd gnuradio
mkdir build

cmake -Dhave_mfpu_neon=1 -DCMAKE_CXX_FLAGS:STRING="-march=armv7 -mfpu=vfp -mfloat-abi=soft" -DCMAKE_C_FLAGS:STRING="-march=armv7 -mfpu=vfp -mfloat-abi=soft" ../

WARNING: This will take more than 24 hrs, and will slow down your board to an otherwise-unusable state. Both cores are being utilized at near 100% for the build
make -j 2
sudo make install

[Gerald]

Hi tjt7a,

do you have a radio connected to the parallella? which one? can you already send data?

Thanks,
Gerald

[tjt7a]

[aolofsson]

Wow, that's really cool! Can you post some screenshots?
Andreas

[Gerald]

@tjt7a good work! is ther enough memory to implement a USDR on the parallella?

[tnt]

Why would you disable NEON ?!?
Also, are you sure the ABI of the distributed SD card image is hard float ? I though it was softfp

Finally, an option is to disable python which makes the build _much_ faster and depending on your use case, it might be acceptable.

[tjt7a]

[tjt7a]

I ssh into the Parallella and forward X to get the gnuradio-companion on my laptop.

Some screenshots of what I have working right now:

Hardware Setup:

Hardware

Photo Apr 19, 12 15 25.jpg (100.26 KiB) Viewed 59023 times

Successfully Reading From the USB TV Tuner:

Reading From TV Tuner

Photo Apr 19, 12 15 53.jpg (70.42 KiB) Viewed 59023 times

GNU Radio Scope view of FM demodulated signal at 89.5MHz:

89.5

Photo Apr 19, 12 15 17.jpg (89.56 KiB) Viewed 58998 times

Tom

[Gerald]

@tjt7a is ther enough memory to implement a USDR on the parallella?

[theover]

That's some serious work, it almost makes us feel back the power of the Unix systems of old, successfully running a make on such small computer and some USB stuff.

Of course in case of the system at hand additional work is interesting on getting the FPGA fabric to take a role like in the software radio hardware, and to let the Adaptiva chip compute things... It always bugs me that the SR guys usually decimate and crude FFT at will, without much subtlety, this board can do better. Of course connecting up AD converters, maybe audio converters and such may be a lot of work, and I don't know if the webpack ISE tools are very fast to use on a big SR design (WRT the "design loop").

T.V.