in the main window of the AD9361 Filter Wizard the Magnitude Response of the programmable FIR is shown. What I don't understand is the frequency axis. If I want the FIR to interpolate by 2 or 4, shouldn't the frequency axis go from 0 to 2*Fs/2 or 4*Fs/2. With Fs I mean the Data rate, that is entered by the user in the UIControl on the left.
An interpolating filter first upsamples the signal and then filters. So at the input of the filter the signal has a higher sampling rate, in our case double of quadruple.
Shouldn't the frequency axis, and thus the bandwidth of the filter then conform to the higher sampling rate?
I attached an image indicating what I mean.
With best regards,
The FIR filters in the part are polyphase therefore they filter then interpolate on the TX, and decimate then filter on the RX. Therefore, the bandwidth will always be with respect to the data rate for the wizard. The filters are also component independent (no cross multiplies), so their frequency response will always be symmetrical.
I like your comment here and hope you are correct, because my understanding of polyphase filters is the same as yours.
For Tx Interpolation, Filter first, then interpolate, this is minimal computationally
For Rx Decimation, decimate first, then filter. Same reasons - less samples to work on, faster etc
I believe that the FIR structure is a (16tap - 8 rows) to get the 128 tap ?
Meanwhile, the in the user guide UG570 rev a. ( I can't find an updated manual)
it says the opposite to this, and that has confused me.
UG570 rev A, on page 31 (Tx DIGITAL FILTER BLOCKS section) it says.
"In each filter, interpolation is performed first followed by the filter transfer function"
On page 33 (Rx DIGITAL FILTER BLOCKS section) it says.
"In each filter, decimation is performed after filtering has taken place"
….So I'm hoping that you can clarify that you are indeed right and the manual is wrong...…