# Weird noise floor slope when capturing with USRP E310

I'm experiencing a weird effect on my captures. It may just be due to my bad dsp understanding.

I'm using an E310 SDR platform to capture a multitone (4 carriers) signal:

• Fc = 150 MHz
• Bw = 286 KHz
• fs = 3 Msps (I tried several values ranging from 300 ksps to 3 Msps)
• format = short (IQ components, 16 bit each)
• Time of capture: 10 s

And then using Matlab to plot it:

% Read signal from file
fid = open(filename, 'r');
rx_signal = complex(IQ(1:2:end), IQ(2:2:end));

% Plot PSD
pwelch(rx_signal, [], [], [], 3e6, 'centered')


And the result is shown in the figure below:

That ~10dB noise floor slope does not look good. Shouldn't the noise floor be flatter? What may I be doing wrong?

Cheers,

~Leroy

ps- I'm new to this community but I hope I'll be around for a good while :)

• Since this is specific to the USRP hardware, you might have better luck asking in their mailing list. – MBaz Feb 5 '18 at 13:46
• @MBaz, yes, you are right. I was just suspicious that I could be doing something wrong in Matlab, or that maybe this noise floor slope was actually normal. If nobody else has any idea what this might be I'll close the question and then ask in Ettus mailing list. Cheers – Leroy Feb 5 '18 at 18:38
• One way to test the frequency response of your system is to transmit white noise covering the entire band. In the receiver, set the FFT visual to average, and let it run for a few seconds. Any clear deviation from flatness is due to the frequency response of some part of your tx-rx chain. – MBaz Feb 5 '18 at 20:44