Re: FFT Clutter Depending On LO Offset

On 11/04/2019 08:25 AM, Meelis Nõmm wrote:
>
> Hello everyone,
>
> We have been using the USRP N210 with the WBX 50-2200 (40 MHz)
> daughter board to generate FFT plots at various frequencies. Note that
> the experiments were done done without the antenna being connected.
> Our objective is to get as clean of an FFT view as possible.
>
> In order to get rid of the large DC spike in the middle of the FFT, we
> started offsetting the LO frequency with the use of tune requests.
> However, this produces unexpected results. In theory, we should be
> able to move the DC spike of the LO "out of view" by applying an
> offset that's greater than half the sample rate of the FFT. However,
> depending on the sample rate of the USRP, the center frequency, and
> the exact offset given to the LO, smaller secondary spikes will appear
> at unexpected locations on the FFT. See attachments for illustration:
> they're all taken at the same sample rate, with the same centre
> frequency, and only the LO offset is being modified.
>
> The research that we've done into the matter indicates that it might
> be impossible to wholly remove the DC spike in the middle of the FFT,
> however, with the LO moved off of it, the remaining spike isn't that
> large of an issue. What is an issue, however, is the fact that we have
> secondary spikes that appear depending on the LO offset frequency, and
> sometimes on the center frequency of the FFT. We are also aware of the
> fact that we should not be moving the LO out of the bandwidth of the
> daughter board (40 MHz total).
>
> What might be the reasons for such spikes, and what other factors
> should we look at when trying to get a clean FFT picture?
>
> Regards,
>
> Meelis
>
Try terminating the antenna input with a terminator--does that make a
difference?

At lower frequencies you can fully-expect there to be spurs picked up
from the environment, even with the antenna disconnected--terminating
the input will give a better picture.

But spurs are inevitable with any widely-tunable receiver--whether SDR
or a more-traditional architecture--particularly if there are digitial
circuits within the receiver architecture. One of the reasons that
high-end laboratory RF instrumentation is so costly (10s of $K) is that each
subsystem is EM-isolated from other subsystem with individual
sub-enclosures. Even THAT doesn't eliminate all spurs.