Re: [Discuss-gnuradio] Please help me understand wave.jpg [attached] representing the plot of the usrp_rx_cfile.py output

[Tom Rondeau]

On Fri, Mar 6, 2009 at 5:07 AM, Bishal Thapa <address@hidden> wrote:
> Dear all,
>   Please help me understand this output from usrp_rx_cfile.py. I have
> attached the plot as the jpeg at the end:
>
> 1. I send a file with 1500 bytes of spaces (Space = 0x20 = 0010 0000 in
> binary)
> 2. I use usrp_rx_cfile.py to capture the samples, and focus in on roughly
> 24600 samples (because 1500 bdpsk modulated bytes = 1500*8*2 + preamble +
> CRC..)
> 3. multiplied by 2 because its 2 samples per symbol (one bit per symbol).
> Therefore, I ignore every other sample. So down to 12300 samples roughly.
> 4. I turned off whitening and dewhitening. Hence, I should now see dbpsk
> samples that represent 0010 0000 0010 0000.....
> 5. I do see the pattern when I look at the angle formed by those complex
> symbols in MATLAB [using  plot(angle(read_complex_binary('s.dat')))]
> 6. But I also see some kind of unexplainable shifting once I take the
> absolute value of the angles of the complex symbols [shown in the jpg].
>
> Can somebody please tell me if there is some kind of shifting implemented in
> the dbpsk modulation which causes the plot of the angle of these complex
> samples
> captured by usrp_rx_cfile.py to look like this. Please note that I turn off
> whitening/dewhitening so the pattern do resemble 0010 0000 0010 0000
> somewhat.
>
> -Thank you for your time,
> -B.

Bishal,

In the future, please label your axes.

If I'm looking at what I think I'm looking at, then the answer is very
simple. You have a frequency mismatch between your transmitter and
receiver. And remember, you ALWAYS have a frequency mismatch unless
you are driven by some universal standard, and even then, well, we
won't get into that.

The digital demodulators we have implemented go through a lot of
trouble to track the frequency, phase, and timing at the receivers. If
you hare just using usrp_rx_cfile, there is no frequency correction
going on. The different frequencies used in both USRPs will cause an
envelop like you are seeing of fdel = |f_tx - f_rx|.

Tom