On Off Keying

Hi,
I am trying to send a stream of bits using On Off keying and am
having
some issues.

At this stage, I just want to check if 1’s and 0’s are getting received
with
a high and low amplitude respectively. I have modified dbpsk.py setting
the
constellation to 0+0i and 1+0i in psk.py and invoke the tx/rx as in
benchmark_tx,benchmark_rx …

My flowgraph is
Bytes2symbols ->RRCFilter->USRP

USRP->filesink

I have a Bytes2symbols file which just writes the complex symbols for a
given set of bytes as in gr_chunks2symbols_bc.cc.

I have also checked that the complex symbols entering the USRP at the
transmitter are as expected.

However, at the receiver (USRP baseband samples without any processing)
when
I measure the power, I do not see the power going low for the 0 bits.
Specifically, when I send a 101010… bit stream of 128 samples (just
these
bits without any headers/trailers). The transmitted baseband complex
symbols are as expected with the real part going between 1 and 0
alternatively. At the receiver, the received power stays almost the
same
high value throughout the packet duration, whereas I would have expected
it
to alternatively go high and low
.

Adding or removing the RRC filter doesn’t affect the observation. The
following observations are true for the power and the real part of the
baseband samples.

  1. For a tx. stream of all 1’s, i can see the beat frequency or the
    frequency offset for the duration of the packet (as expected).
  2. For a tx. stream of all 0’s , i see a low received value. (almost
    close
    to the noise levels) as expected.
  3. However, for a tx. stream of 1’s and 0’s mixed, I still see the
    received
    amplitude (real part) showing the beat frequency continuously and not
    going
    to 0 for the 0 bits.
    I am using the latest stable version i.e gnuradio-3.1.3 on Ubuntu
    laptops.
    Could this be Inter Symbol Interference or a setting which makes the
    (carrier) power coming out of the USRP constant for the packet duration
    irrespective of the tx.data?

Any help is greatly appreciated.

Thanks,
Sri.

On Mon, Sep 29, 2008 at 11:52 AM, sri ram [email protected] wrote:

  1. However, for a tx. stream of 1’s and 0’s mixed, I still see the received
    amplitude (real part) showing the beat frequency continuously and not going
    to 0 for the 0 bits.

When you send alternating 1’s and 0’s, you are creating a baseband
square wave of constant power. The DC offset is half your baseband
transmit amplitude, and that energy at DC is upconverted to your
carrier frequency. On receive, since you have a frequency offset, you
will see a continuous beat frequency resulting from this constant
carrier. Superimposed on this will be the harmonics of your square
wave up to the Nyquist limit of your baseband sampling rate, or up to
the cutoff frequency of the RRC filter if it is in use.

You could of course change your baseband to be bipolar (-1, 1), but
then again, that’s just BPSK.


Johnathan C.
Corgan Enterprises LLC
http://corganenterprises.com/

On Mon, Sep 29, 2008 at 4:40 PM, Johnathan C.
[email protected] wrote:

will see a continuous beat frequency resulting from this constant
carrier. Superimposed on this will be the harmonics of your square
wave up to the Nyquist limit of your baseband sampling rate, or up to
the cutoff frequency of the RRC filter if it is in use.

Don’t some of the daughterboards also have some AGC built in? I can
see if the interpolation rate is not high enough, the signal power
will not go down enough (especially after the RRC filtering) to really
look like much of a difference if any due to the AGC circuitry and
other transients that may occur on signals quickly coming on then off.

Please correct me if I am wrong, but I think using a very large
interpolation rate might help clarify the situation.

Brian

On Mon, Sep 29, 2008 at 2:28 PM, Brian P. [email protected]
wrote:

Don’t some of the daughterboards also have some AGC built in? I can
see if the interpolation rate is not high enough, the signal power
will not go down enough (especially after the RRC filtering) to really
look like much of a difference if any due to the AGC circuitry and
other transients that may occur on signals quickly coming on then off.

None of the daughterboards have AGC as far as I’m aware (Matt,
correct me if I’m wrong on this.)

It’s not that the filtering is preventing the envelope from going to
zero (though it might be; RRC’s are intentionally designed to
introduce ISI in a very specific way). It’s just that with the
waveform he’s sending, there is a strong carrier component at passband
that shows up as a constant beat frequency in the receiver due to
frequency offset.

Please correct me if I am wrong, but I think using a very large
interpolation rate might help clarify the situation.

It would improve things. If the baseband “square wave” had it’s
fundamental frequency near the RRC filter limit, or near the Nyquist
limit of the baseband sampling rate if the RRC wasn’t in use, there
would be few to no harmonics that make it through the filtering and/or
interpolation. The transmitted waveform would be a carrier and two
sidebands, a classic AM waveform. I think that’s what he is seeing.


Johnathan C.
Corgan Enterprises LLC
http://corganenterprises.com/

Hi,

Johnathan C. [email protected] wrote:

None of the daughterboards have AGC as far as I’m aware (Matt,
correct me if I’m wrong on this.)
I’m not very sure, but I think TVRX daughter board has an AGC.

Regards,

Firas

Thanks for all the discussions. It has definitely given me clarity into
what is happening. .
I understand that the DC component of the transmitted signal, along with
the
frequency offset causes a strong tone at the receiver. I have linked two
plots of the real part of the received baseband samples with and without
RRC
pulse shaping.
http://www.ece.gatech.edu/~sriram/rrc.gif
http://www.ece.gatech.edu/~sriram/norrc.gif

It is clear that there is a strong tone corresponding to the frequency
offset and the data modulation can be seen although not clearly. Also,
using
the RRC pulse shaping seems to be better when I observe the real part of
the
received sample (although the power plots are similar with and without
RRC).

I am just wondering how I could demodulate this and how practical OOK
receivers operate.
Should this tone be filtered out using a high pass filter (difference of
consecutive samples?) ? Since the data sample variation is faster than
the
frequency offset ( about 100 samples for one cycle of the offset
frequency),
it looks likely that a filter which passes the high frequency (Data)
variations and filters out the low frequency (frequency offset) must be
helpful? Am I correct in thinking this way?

Feeding this to the costas loop does not help much. The costas loop
(with
default parameters that work for dBpsk) is unable to demodulate the bits
correctly. I am wondering if tweaking its parameters to values
different
from the default ones might help.

I currently use a -i 256 and -d 128, since my laptops don’t keep up for
lower values. I will experiment to see what happens if they are
increased.

Thanks,
Sri

Uhhhh. I thought the ADC’s or an external gain element was on the
USRP1.

ARRL SDR Working Group Chair

Member: ARRL, AMSAT, AMSAT-DL, TAPR, Packrats,

NJQRP, QRP ARCI, QCWA, FRC.

“Trample the slow … Hurdle the dead”

From: discuss-gnuradio-bounces+rwmcgwier=removed_email_address@domain.invalid
[mailto:discuss-gnuradio-bounces+rwmcgwier=removed_email_address@domain.invalid] On Behalf
Of
Firas A.
Sent: Tuesday, September 30, 2008 3:22 PM
To: Brian P.; Johnathan C.
Cc: sri ram; [email protected]
Subject: Re: [Discuss-gnuradio] On Off Keying

Hi,

Johnathan C. [email protected] wrote:

None of the daughterboards have AGC as far as I’m aware (Matt,

correct me if I’m wrong on this.)

I’m not very sure, but I think TVRX daughter board has an AGC.

Regards,

Firas

Johnathan C. wrote:

None of the daughterboards have AGC as far as I’m aware (Matt,
correct me if I’m wrong on this.)

It is not automatic, but there is gain control under software control.

Matt