OFDM benchmark transmit signal bandwidth

GNU Radio
1

Hi all

I am currently using OFDM benchmark to generate OFDM signal under the
setting of FFT len, CP length, occupied-tones and something.
But I can not find out what is the real bandwidth of signal it
generated.
Because when I changed the Interpolation rate (sampling rate), the
bandwidth
at RX changed as well.
Ideally we know that setting enough large sampling rate ( In USRP2, the
max
fs = 25MHz), I should observe the constant signal with fixed BW.
It seems to me that BW of the generated signal is too large.

My question is: how to determine the BW of transmit signal in the codes?
where I can change it.
All I found is actual bit rate = (converter_) / xrate /
samples_per_symbol
= 100MHz/4/2. But this one seems not related to the BW of signal itself.

Thanks for any suggestions!

2

On Thu, Sep 8, 2011 at 6:47 PM, Guanbo Z. [email protected] wrote:

My question is: how to determine the BW of transmit signal in the codes?
where I can change it.
All I found is actual bit rate = (converter_) / xrate / samples_per_symbol
= 100MHz/4/2. But this one seems not related to the BW of signal itself.

Thanks for any suggestions!

Regards,
Guanbo

Guanbo,
The bandwidth of the signal changes with the interpolation rate. If you
set
the interpolation rate such that you get 25 MHz of bandwidth out, then
the
OFDM signal will also have a 25 MHz bandwidth. What you will see over
the
air is 25e6 * (occupided_tones/fft_length), since the ratio of the used
tones to the number of subcarriers is the amount of occupied bandwidth.

You can also think of it this way. The bandwidth of a subcarrier is
BW/fft_length, where BW is the sample rate out of the USRP.

Tom

3

On Fri, Sep 9, 2011 at 8:18 AM, Tom R. [email protected]
wrote:

max fs = 25MHz), I should observe the constant signal with fixed BW.
Regards,

You can also think of it this way. The bandwidth of a subcarrier is
BW/fft_length, where BW is the sample rate out of the USRP.

Tom

Hi Tom

What you means that, the bandwidth of OFDM signal is actually equal to
the
sampling rate*occupided_tones/fft_length.

Then how to understand the sampling theory, in which sampling rate is
twice
of bandwidth?

Thanks,
Guanbo

4

On Fri, Sep 9, 2011 at 6:36 PM, Guanbo Z. [email protected] wrote:

But I can not find out what is the real bandwidth of signal it generated.
samples_per_symbol = 100MHz/4/2. But this one seems not related to the BW of
The bandwidth of the signal changes with the interpolation rate. If you

Hi Tom

What you means that, the bandwidth of OFDM signal is actually equal to the
sampling rate*occupided_tones/fft_length.

I mean exactly that :slight_smile:

Then how to understand the sampling theory, in which sampling rate is twice
of bandwidth?

Complex signals. Sample rate is the bandwidth. Have a sample for I and
Q, so
we still have enough information so as not to violate Nyquist.

Thanks,
Guanbo

–
Regards,
Guanbo

Tom

5

Hi again,

I am facing a similar problem and I would greatly appreciate your
feedback.

I am transmitting an OFDM signal with 512 subcarriers, 200 occupied
tones, BPSK, with a sampling rate set to 2MHz. Accordingly, I would
expect to see a bandwidth of 200*(2*10^6)/512 = 781.25KHz of occupied
bandwidth in the corresponding FFT plot. Am I correct?

However, what I observe at the RX is a bandwidth of 200KHz. Can someone
explain this or suggest any changes? Since the number of occupied tones
is 200, this means that the bandwidth of each occupied tone is 1KHz.

I also noticed that changing the sample rate at the Rx or Tx does not
affect the bandwidth of the OFDM signal, which now makes me wonder what
is the exact relationship between these two?

Thanks for your time.

Regards,

Konstantinos

6

Great!! Thanks for your explanation!

Guanbo