fftshift implementation in Simulink

I need to generate IPs for HDL implementation of an algorithm. I need to do fftshift in simulink but there is no fftshift block avaiable. Is there any way we can do fftshift of the output data from FFT block from DSP HDL toolbox? I am attaching FFT model for reference.

[Hamza]

@Walter Roberson can you help me with this query?

Answer 1

[Walter Roberson]

Fork the fft output. Use https://www.mathworks.com/help/dsp/ref/variableselector.html variable selector block on each of the branches, with Fixed selector. In one of the branches select 1:floor(signal_length)/2 and in the other branch select floor(signal_length)/2+1:signal_length . Now concatenate those two together in reverse order.

This potentially requires hard-coding the signal length.

[Paul]

This approach doesn't work for odd signal_length assuming the intent is floor(signal_length/2) (not floor(signal_length)/2)

x = 1:9;
signal_length = numel(x);
fftshift(x)

ans = 1×9

6 7 8 9 1 2 3 4 5

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

[x(floor(signal_length/2)+1:signal_length), x(1:floor(signal_length/2))]

ans = 1×9

5 6 7 8 9 1 2 3 4

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

[Walter Roberson]

@Paul

good point

x = 1:9;
signal_length = numel(x);
fftshift(x)

ans = 1×9

6 7 8 9 1 2 3 4 5

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

[x(ceil(signal_length/2)+1:signal_length), x(1:ceil(signal_length/2))]

ans = 1×9

6 7 8 9 1 2 3 4 5

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

[Paul]

I just realized that this block requires the FFT length to be a power of 2, so the ceil() isn't necessary for this particular use case.

Answer 2

[Paul]

fftshift is extended to code generation. Assuming the data output from the FFT block is in linear order, you can call fftshift from inside a MatlabFunction block.