[srslte-users] "Digital scaling" in srsUE

Mathias Rønholt Kielgast mkielg11 at student.aau.dk
Sun May 7 18:46:23 UTC 2017

Thanks for your response! I generally also use attenuators, however, I was trying to make a "software defined" way to change the SNR using without changing the physical setup.

Where in the srsLTE library can I find the functionality to compute the scaling using the RSSI sensor on the USRP board? I haven't been able to find it.

Yours sincerely,
Mathias Rønholt Kielgast
Master's Student, Wireless Communication Systems
Aalborg University
From: Ismael Gomez [ismael.gomez at softwareradiosystems.com]
Sent: 05 May 2017 17:21
To: Mathias Rønholt Kielgast; srslte-users at lists.softwareradiosystems.com
Subject: Re: [srslte-users] "Digital scaling" in srsUE

Hi Mathias,

On Fri, 5 May 2017 at 11:26 Mathias Rønholt Kielgast <mkielg11 at student.aau.dk<mailto:mkielg11 at student.aau.dk>> wrote:

I am trying to artificially add additional noise in srsUE to decrease the SNR.

This is not really a good idea. The best approach to simulate low SNR is to use a cable and a variable attenuator. This is what we've used so far with excellent results.

I have difficulties however, as I struggle to find the "correlation" between the "digital power" (i.e. the values of the complex floats in the baseband samples handled before decoding) and the actual input power.

That's difficult yes. You need to calibrate it for your setup.

I want to add power at a level compared to the current noise level (i.e decreasing the SNR with a factor of 3 dB) using the noise estimation performed by the srsUE functions.

An example: There is estimated a noise power of 12 dBm on a channel (and a 27 dBm RSRP, so a good channel).
I want to add noise to decrease the SNR with 3 dB, i.e. I (on a linear scale) must double the noise. If I generate samples with N(tau) = sqrt(0.5*P_N) * (N_I + iN_Q), where P_N is 0.15 (12 dBm in linear scale) and N_I/N_Q are Gassian random numbers, I then should have generated an array with noise power equal to what is on the incoming signal.
However, this array of noise samples will have "digital power" of 0.15, which is very large compared to the "digital power" of the samples given (which are about 0.0003). So adding the generated noise to the signal wil ruin the SNR.

How can I relate the estimated powers to the "digital power"? What kind of scaling is being used? I have tried to figure it our from the noise estimation functions themself, however to no avail.

What we've used so far is to use the RSSI sensor on the ADI chipset in USRP devices. With that RSSI power and the RSSI power we compute digitally we estimate the scaling factor.

This is already provided in srsLTE.


Hope for some guidance, thanks in advance!

Yours sincerely,
Mathias Rønholt Kielgast
Master's Student, Wireless Communication Systems
Aalborg University
srslte-users mailing list
srslte-users at lists.softwareradiosystems.com<mailto:srslte-users at lists.softwareradiosystems.com>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.softwareradiosystems.com/pipermail/srslte-users/attachments/20170507/c037db20/attachment-0001.html>

More information about the srslte-users mailing list