Hello.
I am using the LabVIEW CSDS 2.0 and the USRP Driver 15.1 with both the USRP-2950R and the USRP-2953R Device. On each device, we have simply connected both ports from RF 0 together such that we can transmit a signal directly from port TX0->RX0 or TX1->RX1 (see attached picture).
We have implemented a basic OFDM transmitter/receiver on the FPGA with 512 subcarriers and adaptive modulation. While implementing and testing automatic classification algorithm to estimate the bit loading table applied on the transmitter side I found an interesting behavior regarding the signal-to-noise ratio.
I have attached two Matlab figures with measurements and analysis from the USRP 2950R and 2950R. We have analyzed the channel transfer function and the signal-to-noise ratio on each subcarrier. Each graph shows the accumulation of 2000 subsequent measurements. You can see that although the CTF is flat in the frequency range of interest (Subcarrier 48-465) the SNR shows a degrading behavior around the carrier frequency (Subcarrier 257). The amount of degradation is also effected by the transmit signal power (see SNR_CTF.png).
In SNR_DifferentMeas.png I have restarted the measurement three times. On the left side we can see that although, SNR behavior is the same the signal values differ. In addition, each time the USRP is initialized there is a different I-Q offset. Subtracting the I-Q offset has no effect on the demodulation process.
My question are:
- Is this behavior expected from the build in hardware components? Both the USRP 2950R and the 2953R exhibit it.
- Is there any way to measure and compensate this effect?
Thank your time and I am happy to answer any questions you might have.
