I am trying to transmit a multi-tone signal from one USRP (N210) to another and have written an out of tree block to generate the multi tone signals using an ifft. The two USRPs are connected directly via a cable and a $50\ \Omega, 20\textrm{ dB}$-attenuator.

I am finding that the signal is received clearly for 1 and 2 tones but not for 3+ tones. I'm not sure what I can do digitally to make the frequency components are lot more distinct. The received signal is amplified by $10\textrm{ dB}$ and passed through a LPF and I've attached some FFT plots showing the results.

2 tone receiver

4 tone receiver

I don't have enough reputation to show the transmitted signals but they are just direct spikes evenly distributed through the bandwidth and very distinct.

The received signals at the USRP are complex and so the FFT plots shown are only the real components.

Flow graph

Flow graph

Relevant graphs for a series of different tones

  • Single tone


    single tone received


    single tone transmitted

  • Two tones


    two tone received


    two tone transmit

  • Three tones


    3-tone RX


    3-tone TX

The number sweep block is an out of tree module that produces the frequency array that is passed into the FFT block.

The two USRPs are connected via a MIMO cable to ensure synchronisation of their clocks.

  • $\begingroup$ that could be due to the spectral characteristics of the signals. Are you sending pure tones or pulse modulated tones? $\endgroup$
    – vaz
    Jul 28, 2016 at 15:18
  • $\begingroup$ From what I know of how the USRP works, I assume the transmitted wave will be pulse modulated (I'm not 100% sure). Its worth noting that I am using an oversampling of 5 (1MHz) $\endgroup$
    – Roy
    Jul 28, 2016 at 15:34
  • $\begingroup$ How many tones are you transmitting, over what bandwidth? What is the duration of each tone? $\endgroup$
    – MBaz
    Jul 28, 2016 at 15:58
  • $\begingroup$ why is your spectrum single sided? The USRP gives you complex baseband, so your spectrum should basically span $[-\frac{f_s}2;\frac{f_s}2]$ No, the USRP doesn't do any baseband modulation itself, it just mixes up the baseband signal you supply it with to the carrier frequency you specify. Which is that? Which model is your USRP? If applicable: which daughterboard? can you include the flow graph? This is all very much stumbling in the dark. $\endgroup$ Jul 28, 2016 at 17:28
  • $\begingroup$ @MarcusMüller I'm using the USRP N210 with a carrier frequency of 2.8GHz. Yes, the plots do go into the negative frequencies and the fft plots in the question are the real parts of the received complex signal. I've edited the question to include the relevant information that I missed out. I've tried experimenting with different oversampling rates and windows for the fft block and LPF. $\endgroup$
    – Roy
    Jul 29, 2016 at 8:52

1 Answer 1


The problem was from the fact that I was using a rectangular window for the ifft at the transmitter.

Using any other window at the ifft ensures there are no harmonics outside the 200kHz (which are interrupting the signal).

Here is an example of the signal using the rectangular window enter image description here

And using a hamming window enter image description here

Four tones

Received 4-tone RX

Transmitted 4-tone TX

  • $\begingroup$ you might want to say something why that solves things, mathematically, for the afterworld :) $\endgroup$ Jul 29, 2016 at 10:37

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.