On 2022-04-08 09:56, Tobias Kronauer wrote:
Hi all,

we use a X410 and send 60k samples via localhost. We increase the frequency within the frame every 6k samples. We observe a frequency selectivity of the USRP. Can you help us on that? In the attachment [0], you can find the MWE (we tried to keep it as small as possible). Let us explain the problem in the following plot:


In the left column, you can find the transmitted signal, in the right column the received signal. In the lower row, you see the spectrogram. One can clearly see that the signal gets attenuated, depending on the frequency. One would expect the amplitude to remain constant.

The transmitted signal changes every 6k signal its frequency. Within these 6k samples, we simply send complex exponentials. The frequencies are: [-25, -19.44 -13.9,  -8.3, - 2.78   2.78   8.3  13.9, 19.4  25].
Rf config:
```cpp
     const double FS = 50e6;
    Config conf;
    conf.txAnalogFilterBw = 400e6;
    conf.rxAnalogFilterBw = 400e6;
    conf.txGain = 35;
    conf.rxGain = 35;
    conf.txCarrierFrequency = 2e9;
    conf.rxCarrierFrequency = 2e9;
    conf.txSamplingRate = FS;
    conf.rxSamplingRate = FS;
```
For different RfConfigs, the results are less distinct.

Kind regards,

Tobias

Attachments:

[0]: usrp_mwe.cpp


Your plots aren't particularly clear in what they're showing, neither is your problem description.  Is your list of "frequencies" in terms of the baseband?  In Hz?  Khz?  MHz?

ANY radio receiver (or transmitter) will have a passband response that is dependent on the analog passband response of the various components in the system.  With   a DSP radio the passband response is conceptually the convolution of the  analog and digital responses.  The passband for any SDR will exhibit roll-off at the edges   of your passband--that roll-off CANNOT be infinitely steep without the filters also being infinitely long.

I don't have an X410, so I don't know what the passband *should* look like, but I imagine that there are both edge roll-off effects and the passband is not
  perfectly "flat".

If you inject a laboratory noise source into the receiver, and just look at the spectrum integrated over a few seconds, you will see the pass-band response of the
  radio at the desired frequency, at the desired sample-rate.

_______________________________________________
USRP-users mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to