Re: [Discuss-gnuradio] Error in testing benchmark_tx/rx.py

[Tom Rondeau]

On Sun, Nov 20, 2011 at 1:57 AM, Songsong Gee <address@hidden> wrote:

Thank you for all of you.

I've got some progress on it.

From Tom's suggestion I give options to benchmark_tx.py

$ ./benchmark_tx.py -f 450e6 --tx-amplitude=0.02 --tx-gain 30 -S 10

And looking its spectrum,

<snip>

 

I could get a distinct frequency component around 450 MHz.

After that, I ran benchmark_rx.py

$ benchmark_rx.py -f 450e6 --rx-gain=50 -S 10

(I modified receive_path.py to have threshold with -80 dB to filter out others)

And I got a following result.

<snip>

 

(Sorry that this is an image that captured text because (pktno, ) = unpack('!H', payload[0:2]) throws an error and I could not type such ugly characters with my keyboard)

error: unpack requires a string argument of length 2

As you can see, benchmark_rx.py receives something. but it seems that it fails to check CRC.

And I'm curious that the reason unpack() function throwing errors is because payload contains corrupted data and this also causes CRC check failure

This means that I did wrong in either trasmitting or receiving.

1. Still signal is clipped (but, I guess not, --tx-amplitude=0.02)

2. Threshold value is mistaken (threshold = -80 [dB])

3. sample per symbol (-S 10)

4. --tx-gain, --rx-gain might be wrong.

5. or something I am missing.

Please give your suggestions. I am almost close to the goal.

And thank you for all of you in advance

Yes, you are close. It's probably a matter of tweaking the receiver gain and the transmit power to the right levels. We have a digital AGC, but that doesn't do much for the important analog gain stages you need to get the signal right.

It could also be a frequency offset, although your graph showed that the signal looks pretty well centered. It be good to zoom in, though, or use the uhd_siggen to transmit just a tone and find out how far off frequency you are.

Tom

 

2011/11/20 Tom Rondeau <address@hidden>

On Sat, Nov 19, 2011 at 8:13 AM, Songsong Gee <address@hidden> wrote:

FYI,

this is a fft scope WIHTOU TX (uhd_fft.py -f 450e6 -A TX/RX)

and this is the fft scope WITH TX (benchmark_tx.py -f 450e6 -A TX/RX)

as you can see, gain increased when TX is turned on.

Songsong,

That signal looks horrible. I'm pretty sure you're clipping on either the transmitter or receiver. Notice that the signal level is -10 dBm! The skirts only go down to -60 dBm, too, so you have some serious nonlinearities occurring there. Try backing off the power on the transmitted signal (--tx-amp=) and see how that does.

Tom

--

Seokseong Jeon, PhD Candidate

Communication & Networks Lab

IT Convergence Engineering (ITCE), POSTECH, Korea

+82 10 8338 1229, gee.songsong at gmail . com