Redpitaya Forum

Hello everybody,

Is there a way to make an anti-aliasing filter inside the FPGA?
thanks for any comments/suggestions

TOM

Digital anti-aliazing filters work with an oversampled signal. If the (fixed) sampling rate of 125MSps is sufficiently oversampled for your desired target bandwidth, then the answer is yes.

You can implement a digital filter in the FPGA that reduces the bandwidth of the sampled signal (62.5MHz) to your target bandwidth (x), and then apply decimation to reduce the oversampled rate to the rate you actually need. This avoids aliazing effects of signal energy in the band from x to 62.5MHz that you would get by just decimating.

Of course, this is constrained by the available computing resources of the FPGA.

thank you Nils for your answer, this is exactly what i would like to do.
But unfortunately i don't know how to make such filter inside a FPGA( for now i only know how to work on uP side ).

Could you point me to some examples that show how to modify the actual FPGA bit file?

As you said, this is constrained by the available computing resources of the FPGA,
So the question is does the FPGA as enough resources to do this?
Or should i remove many functionalities to have enough resources?

thanks for any comments/suggestions

TOM

Could you point me to some examples that show how to modify the actual FPGA bit file?

You modify the bitfile by changing things in the Verilog code or block design and then rebuilding it.
Here is an example for adding a filter module in the Verilog code. For block designs, I believe there is a filter wizard in Vivado.

So the question is does the FPGA as enough resources to do this?

That depends on the complexity of your filter. I have no practical experience in designing digital (or analog) filters, so I can't give an estimate.

Are you sure it doesn't have this feature already?
I fed a 16kHz 1Vpp signal into the Red Pitaya. At sampling rate of 122 kHz it showed up fine.
At a sampling rate of 15.258kHz I would expect a 1Vpp aliased signal at .742 kHz, but instead I got a .04Vpp aliased signal at .742kHz.
This also happens if you use the oscilloscope app and change the time scale.