[USRP-users] X300/CBX-120 at high sample rates: Filter issues?

Marcus D. Leech mleech at ripnet.com
Fri Mar 11 16:53:02 EST 2016


On 02/19/2016 02:49 PM, Merlin Chlosta via USRP-users wrote:
> Hi all,
>
> to measure channel transfer functions we'd like to operate the X300/CBX-120 at high sample rates to use 100 MHz bandwidth, i.e. at least at 100 MSamples/s. Because the resulting transfer function looks odd, we tried with different center frequencies. An image of overlapping transfer functions is attached.
>
> [200MS_overlapping.svg]
> We see each transfer function filters 120 MHz, but also attenuation rises around center frequency of each measurement. Center frequencies are 50 MHz apart, so for example two measurements fc=3.00 GHz and fc=3.05 GHz should agree between 3.00 and 3.05 GHz. Instead they differ notably. With a lower sample rate this does not happen. What shapes the transfer function like that, besides the lowpass cutoff?
>
> [100MS_overlapping.svg]
> This is 100 MS/s with center frequencies 25 MHz apart. It's more like what I expected, but there seem to be some filter effects towards the edges. I only send a 80 MHz wide signal -- the 120 MHz analog filter should cover that. Could this be caused by the digital filters? Are transfer functions for the digital filters available? My intention was to avoid them with 200MS/s mode.
>
> What am I missing here? I appreciate any help.
>
> --Merlin Chlosta
There will be a narrow divot in the spectral response around 0Hz (DC) 
due to DC-offset removal.    What resolution are you using in your FFTs--
   lower FFT resolution will make the DC offset correction seem larger 
than it actually is.

I don't see anything unusual in the band-edge effects in your plots.  
The FPGA code is freely available, where you can look at the filter
   implementations. 
http://files.ettus.com/manual/page_build_guide.html#build_get_source







More information about the USRP-users mailing list