# How do I filter out the specific components at 20MHz, in this figure?

I used the filter builder (lowpass) , to cut off the frequency at 0.5 MHz, and thus used a stopband of 0.1pi, and a passband of 0.0125pi. (Sampling Frequency= 80MHz). However, the output was still unsatisfactory, as the spikes at +20 and -20 MHz are still visible. Hence, I reduced the stopband to 0.08pi, from 0.1pi, but then the spikes at 20 and -20 became larger, but in the negative direction.

What could be wrong? And what is the reason for the spikes becoming more negative as the stopband frequency is decreased?

Input Spectrum- Figure 2

Output of filter- Figure 3

Output of filter, with reduced stopband- Figure 4

• "still visible" is not a relevant criterion, typically. What's the suppression you need to achieve? And those aren't "negative spikes"; this is the effect of filtering away exactly these frequencies. I think you might be confusing the meaning of what you're looking at with a signal or something. – Marcus Müller May 23 '19 at 7:58
• I want to be left with only the 0MHz component, that is, I don't want my 20MHz components to be present in the spectrum. I don't really understand what you mean by "the effect of filtering away exactly these frequencies". Shouldn't the magnitude at ± 20 MHz be decreased to a complete 0, if they are filtered away? – Toonlooney 97 May 23 '19 at 8:57
• you can't mathematically ever achieve a perfect suppression to zero, with any signal that is finite in length. So, no! Also, nobody needs that – if you think about it, do you really care if you have these other two frequency components in there if they are 100000 times weaker than the main component? Is whatever your system is doing with that afterwards that sensitive? The answer might be yes, but then you'd really need to model what your overall system needs. "I still can see them" really is no metric for "goodness" of any filtering. – Marcus Müller May 23 '19 at 9:21
• Yes, that makes sense, thank you! – Toonlooney 97 May 23 '19 at 9:49