OFDM Transmitter and Receiver - poor synchronization with USRP1

GNU Radio
1

Hi,
I am trying to implement OFDM transmitter and receiver (both in one
flowgraph) using 2 USRP1s (both equipped with WBX). I have adjusted the
signal so there’s no peak above 1 and -1, but I am still failing at
syncrhonization. I used exmaples mentioned in Martin B.'s talk at
FOSDEM
14 which are now available in digital/ofdm/ folder.

Running the flowgraph without USRPs seems to be fine at lower sample
rates
(~125k), but I am struggling at adjusting sample rates for USRP1 which
cannot go lower than 250k. I have noticed syncrhonization being much
better
at high sample rates (2-4M).

Are there any guides on how to achieve better synchronization with two
USRP1s?

Thanks in advance.
Michal

2

Is the detection working reliably?

M

3

Doesn’t seem so - by connecting scope to DETECT port on S&C block I can
see
only randomly timed peeks as opposed to running with channel model where
the peeks happen at fixed time periods.

Detection seems to get better with higher sample rates.

2015-05-09 23:25 GMT+02:00 Martin B. [email protected]:

4

On 20.05.2015 20:19, Michal Vaclk wrote:

Was OFDM improved in some kind of way in v3.7.7.1 (I don’t see anything
related to OFDM in changelog)? I simply launched the same flowgraph on
newer version and it works much better.

Hm, not that I recall… but good for you :slight_smile:

I forget, were you using the example codes with default settings? If so,
does doubling the sampling rate without changing the total bandwidth
(i.e. doubling the FFT size) help with sync at lower bandwidths?

M

5

Hi, I am sorry for the delayed response.

I checked the centering of the OFDM signal and it looks pretty good.
Even
looking at the signal’s spectrum using SDR# on other computer was very
nice

  • book-grade example of nicely shaped OFDM spectrum. Today I upgraded
    from
    GNU Radio v 3.7.6.1 to 3.7.7.1 and reception (syncrhonization and
    detection) has improved dramatically. Even at relatively low sample rate
    (1M, anything lower simply doesn’t synchronizie). Unfortunately the
    received packets still weren’t enough for any kind of real world
    transmission.

Was OFDM improved in some kind of way in v3.7.7.1 (I don’t see anything
related to OFDM in changelog)? I simply launched the same flowgraph on
newer version and it works much better.

2015-05-11 21:02 GMT+02:00 Martin B. [email protected]:

6

On 09.05.2015 15:47, Michal Vaclk wrote:

Doesn’t seem so - by connecting scope to DETECT port on S&C block I can
see only randomly timed peeks as opposed to running with channel model
where the peeks happen at fixed time periods.

Detection seems to get better with higher sample rates.

This might be an issue of high relative frequency offsets (which get
less severe for higher sub-carrier bandwidths). You could check how well
your OFDM signal is centered.

M

7

Yes, I am using default flowgraphs (examples) just adding Multiply Const
block to avoid clipping due to high PAPR and obviously UHD blocks. Using
FFT of 64 and default carrier configuration, sync does get better if I
stretch this to wider bandwidths by setting higher sample rates at USRP.

2015-05-21 18:13 GMT+02:00 Martin B. [email protected]:

8

On 22.05.2015 10:54, Michal Vaclk wrote:

Yes, I am using default flowgraphs (examples) just adding Multiply Const
block to avoid clipping due to high PAPR and obviously UHD blocks. Using
FFT of 64 and default carrier configuration, sync does get better if I
stretch this to wider bandwidths by setting higher sample rates at USRP.

Hey Michal,

I meant using the same signal bandwidth at higher sampling rates (i.e.
bigger FFTs, same number of occupied bins).

M