Change of complex binary file format for Matlab

GNU Radio
1

Hello,
I am trying to use “rx_samples_to_file.cpp” in UHD to save samples
received.
UHD seems save samples in complex integer format in c++.
However, it is not possible to read that format from Matlab.
I have checked the Matlab utilities in Gnuradio, but they were not
useful.
Is that any simpler way to change the formats of the saving file so that
Matlab can easily read?

Another question is that why the saved samples are integer format? not
float?
Thanks for reading.

std::vector<std::complex<short> >

buff(dev->get_max_recv_samps_per_packet());
std::ofstream outfile(file.c_str(), std::ofstream::binary);

while(num_acc_samps < total_num_samps){
    size_t num_rx_samps = dev->recv(
        &buff.front(), buff.size(), md,
        uhd::io_type_t::COMPLEX_INT16,
        uhd::device::RECV_MODE_ONE_PACKET
    );

    outfile.write((const char*)&buff[0], num_rx_samps *

sizeof(std::complex));

– *
*

2

I am trying to use “rx_samples_to_file.cpp” in UHD to save samples received.
UHD seems save samples in complex integer format in c++.

yes, COMPLEX_INT16

However, it is not possible to read that format from Matlab.

I’m not a huge matlab fan, but I am sure that it can read integers from
a file. http://tinyurl.com/6kyxdk4

I have checked the Matlab utilities in Gnuradio, but they were not useful.

Its looks like you are not very familiar with matlab, so I suggest (out
of my person preference) that you take this time to learn python+scipy
http://www.scipy.org/ For one, its free. And I have always found myself
more productive with python, and I have found the plotting widgets and
mathematics that come with scipy give me everything I need.

Is that any simpler way to change the formats of the saving file so that
Matlab can easily read?

You can change the example easily to write binary floats. And with a
tiny bit more effort and a for loop you can write a text formatted file
as well.

Another question is that why the saved samples are integer format? not
float?

To save disk space and IO overhead for large captures or high data
rates. You will appreciate this if you change the example to output a
text format.

Good luck,
-Josh

3

Hi

I am not that coded up on UHD, but I would think that they have the same
type of file writing system as is used in the GNU Radio code. In the GNU
Radio code they use a binary file writing system. I have attached a
matlab M
file that reads these types of files, the file can be found in GNU Radio
file address ~/gnuradiio/gnuradio-core/src/utils. There are a large
number
of matlab M files there so just have a look for the type of stuff you
want
to do.

Hope it help

Regards
Alan Jones

4

Thanks a lot.
I found the problem. I should have used ‘int16’ instead of ‘int’ nor
‘float’.
I think I have better spectrum shape for 20Mhz signal using UHD compared
to
gnuradio
usrp2_rx_cfile.py.

But I found high frequency portion is somewhat suppressed than the
original
samples transmitted.

One question is that whether it is possible to demodulate 20Mhz signal
using
Gnuradio. I think many are trying to demodulate 802.11 packets, but
still
cannot found any one succeeded.

The maximum sampling that Gnuradio can provide rate is 25Mhz with
decimation 4. But I am curious about whether it is possible to
demodulate
the 20Mhz signal without 2x or 4x oversampling.