# Confused about anti-aliasing in analog filters

I'm a newbeginner when it comes to signal processing. I'm reading up about anti-aliasing, Nyquist and band-pass filters. There's something I don't get: Anti-aliasing in analog filters.

Lemme try to explain. I use a 8th order analog butterworth band-pass filter (centered at 17.5 Hz with a 25 Hz 3dB bandwidth) on the analog signal. So it cuts at 5 and 30 Hertz and ofcourse there's the transition-band before the stop-band. It is supposed to be an anti-aliasing filter. Where is this anti-aliasing happening?

I asked my teacher before he left why the butterworth was called anti-aliasing filter when it didn't really do it. He answered that since I was working on the analog signal, it was just a band-pass with no anti-aliasing because this was before ADC. The aliasing only occur after the ADC when the analog signal is made to digital and digital filters are used. WUT?

I mean, the anti-aliasing still happens when the signal goes through ADC, right? (well, I sample at 512Hz so I oversample the signal but stil...) So why is the filter called anti-aliasing filter?

Could someone explain this to me? I totally don't get it. Why is it called anti-aliasing filter and where does this anti-aliasing happen?

Aliasing does not happen in the analog world. It only happens when you sample a signal.

If you have an ADC with sampling frequency $f_s$, then you'll have aliasing if the input signal has frequency components larger than $f_N=f_s/2$. To avoid aliasing, it's common to low-pass filter the signal before it is input into the ADC. That way, you make sure that the digital signal out of the ADC has no aliasing.

Such an analog filter is called an "anti-alias filter", not because it will experience anything related to aliasing itself, but because its purpose is to avoid aliasing in the ADC.

• Thanks for the explanation. It really put some pieces down in the puzzle. I'm all clear on this now.^^ – spginner Mar 3 '15 at 19:00

Another way to think about it is that aliasing always occurs during sampling. The entire spectrum always gets folded down and aliased with the signal of interest. Even if that spectrum content is zero.

The closer you can get the unwanted spectrum to zero, the less aliasing noise you will have contaminating your signal band of interest. The best case would be to alias zero noise. (In the real world, that isn't possible, but if you can reduce any alias noise below your sampling quantization noise and numerical processing rounding "errors", it likely becomes insignificant).

So (part of) your bandpass filter's purpose, before the sampler, is to reduce the unwanted (out-of-band) spectrum content so that when it does alias (it will), the alias noise is below your desired noise floor.

Thus, calling it an anti-alias filter as at least part of its purpose. Other purposes might include to keep it from overloading a high-bandwidth input amplifier, or from coupling parasitic noise from or to the power supply, etc.