Hi,
On 01/16/2017 05:37 AM, Nikita Airee wrote:
I haven't modified the flowgraphs from the examples folder at all,
except changing the variables/values for the experiment. It is for those
original flowgraphs of wifi_tx and wifi_rx that I get the deterministic
error.
Besides checking the gnuradio threads, is there anything else I could do
to find the cause for such errors?
My current understanding is that simulations work as expected, but
experiments show deterministic errors.
I would:
- Make sure that there are no overruns or underruns. Disable all logging
to the console and ensure that there are really no 'U's and 'O's.
- Try on a different frequency (just to make sure that there is no weird
device).
- Make sure that the xy plot shows clear constellations.
tx_gain is set to 30dB and rx_gain to 0.
If constellations don't look nice, you could try lowering the tx and
increasing the rx gain. (also make sure that the antennas are connected
to the correct ports).
Best,
Bastian
Bests,
Nikita
On Sun, Jan 15, 2017 at 2:26 PM, Bastian Bloessl <bloe...@ccs-labs.org
<mailto:bloe...@ccs-labs.org>> wrote:
Hi,
On 01/15/2017 07:03 AM, Nikita Airee wrote:
The packet padding was enabled when I ran the flowgraph earlier
with the
pad front/back set to the default values of 1000/100. So I
increased the
values of the padding which lead to some improvement. However those
warnings wouldn't go away compeletely for values as high as
10000 zeros.
I compared my flowgraph (PFA: wifi_tx_rx_jan) to the loopback in the
examples folder and couldn't find any difference besides the
fact the 3
blocks for channel effects(multiply const, channel model and
polyphase
resampler) were not present between the tx and the rx part
initially.
With these blocks in between, the frame rate of 10/s doesn't lead to
those warnings. Is that what going wrong or the presence of a
channel
shouldn't matter?
The receiver calculates the autocorrelation coefficient to detect
frames. If you pad 0's, it will constantly divide by zero, leading
to nondeterminism. With a bit noise (added by the channel model),
this is avoided.
I have yet to check for the threads, which I will only be able to do
tomorrow as I dont have access the hardware today.
Should I upload more of the project?
Depends. If simulations work as expected that will not help much.
I would recommend to start the hardware experiments with the
(unmodified) flow graphs from the examples folder and test whether
they also cause this deterministic errors.
Best,
Bastian
Thanks again for your help Bastian,
Nikita
On Fri, Jan 13, 2017 at 10:39 PM, Bastian Bloessl
<bloe...@ccs-labs.org <mailto:bloe...@ccs-labs.org>
<javascript:_e(%7B%7D,'cvml','bloe...@ccs-labs.org
<mailto:bloe...@ccs-labs.org>');>> wrote:
Hi
On 01/13/2017 05:50 PM, Nikita Airee wrote:
> Thanks for your prompt response!
>
> I have tried connecting the wifi_tx and wifi_rx in loopback
> configuration. At 10Packets/s, it would give me a lot of
"Warning:
> starting to receive new frame before old frame was
complete" messages.
You should pad some 0's between frames (as in the example
loopback flow
graph). Otherwise, frames are really back to back in
simulations.
> So quite a lot of packets didn't reach the MAC decoding stage.
> To eliminate the above warning completely, I had to
increase the
> interval to 20s. But then these sequence numbers have
correct checksum
> values.
This is very strange. Something seems to go horribly wrong
in software,
as these simulations shouldn't depend at all on the frame
rate. Maybe
you could check for threads that are in a live-lock (are
always at 100%
even if nothing is sent).
I would recommend to try the (unmodified) loopback flow
graph in the
examples directory and, then, make your modifications step
by step to
see where it goes wrong.
> There is no under/overflow at the tx/rx in this
configuration.I believe
> this channel to be quite clean, I checked with the IT
department of my
> institute and also listened to the channel using a
spectrum analyzer.
> Even if it weren't, you are right that it shouldn't mess
up the same
> sequence numbers every time.
Could you maybe upload your project? Or produce a minimal
example for
the problem.
Best,
Bastian
> Is there something else I could look into?
>
> Bests,
> Nikita
>
> On Fri, Jan 13, 2017 at 8:30 PM, Bastian Bloessl
<bloe...@ccs-labs.org <mailto:bloe...@ccs-labs.org>
<javascript:_e(%7B%7D,'cvml','bloe...@ccs-labs.org
<mailto:bloe...@ccs-labs.org>');>
> <mailto:bloe...@ccs-labs.org <mailto:bloe...@ccs-labs.org>
<javascript:_e(%7B%7D,'cvml','bloe...@ccs-labs.org
<mailto:bloe...@ccs-labs.org>');>>> wrote:
>
> Hi,
>
> On 01/13/2017 11:17 AM, Nikita Airee wrote:
> > Hi everyone!
> >
> > * Ubuntu 14.04
> > * Gnuradio version : 3.7.10.1,
UHD_3.11.0.git-28-gc66cb1ba
> > * 2 USRP 2953R(x310 + cbx40) connected to host laptops
using pcie
> > cable, antennae=vert2450
> > * center frequency=2.484GHz, samp_rate=10MHz
> >
> > I have been transmitting at the rate of 10
Packets/s over
wireless link
> > with distance between the tx and rx 3m and 6m.
tx_gain is
set to 30dB
> > and rx_gain to 0. (I have varied these gains to find no
improvement).
> > I get a constant total frame error rate at the
receiver ~10%
for a
> > payload of size 50 bytes. The problem is that the
frames for
which the
> > checksum fails are always the same sequence
> > numbers(10,12,31,39,49,58,89,93,94,95...).
> > I have tried the different channel estimation
algorithms and
tried
> > sending the same packet(seq number 0) over and over but
still the same
> > numbers are are dropped. I have also increased the
interval
between
> > transmission to no affect.
>
> really strange that always the same frames are lost.
Can you
run the
> flow graph in loopback mode without any hardware (i.e.,
connect RX to TX
> in GRC), just to rule out any software issues.
>
> Also assert that there are no overruns (in the
receiver) and no
> underruns (in the transmitter).
>
> There could also be interference on 2.4GHz, but very
unlikely
that it
> always hits the same frames. I assume you are sending
frames
that don't
> trigger any response from APs (or other nodes that
might be in
radio
> range).
>
> Best,
> Bastian
>
>
--
Dipl.-Inform. Bastian Bloessl
Distributed Embedded Systems Group
University of Paderborn, Germany
http://www.ccs-labs.org/~bloessl/
<http://www.ccs-labs.org/~bloessl/>
<http://www.ccs-labs.org/~bloessl/
<http://www.ccs-labs.org/~bloessl/>>
--
Dipl.-Inform. Bastian Bloessl
Distributed Embedded Systems Group
University of Paderborn, Germany
http://www.ccs-labs.org/~bloessl/ <http://www.ccs-labs.org/~bloessl/>
--
Dipl.-Inform. Bastian Bloessl
Distributed Embedded Systems Group
University of Paderborn, Germany
http://www.ccs-labs.org/~bloessl/
_______________________________________________
Discuss-gnuradio mailing list
Discuss-gnuradio@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnuradio
