Hi, is there a way to decimate a signal (reduce sampling rate) reducing aliasing without implementing a computationally intensive low-pass filter? Consider that this signal is supposed to have frequency components higher than the reduced sampling rate. thanks

# antialiasing and decimation

Started by ●February 27, 2006

Reply by ●February 27, 20062006-02-27

Sure... use an analog lowpass filter ahead of the sampler. You have just exchanged computational load for circuit complexity - not a bad tradeoff in some applications. BR, Dan

Reply by ●February 27, 20062006-02-27

If your signal is narrowband than you can use a Hilbert transform (if your signal is real, this will reduce redundancy in the spectrum) and sample it at twice it's bandwidth (***as opposed to twice the highest frequency***). The reconstruction is a little more tricky but still possible.

Reply by ●February 27, 20062006-02-27

"andrea" <andrea.zonca@gmail.com> wrote in message news:1141046701.724840.137370@i40g2000cwc.googlegroups.com...> Hi, > is there a way to decimate a signal (reduce sampling rate) reducing > aliasing without implementing a computationally intensive low-pass > filter? > Consider that this signal is supposed to have frequency components > higher than the reduced sampling rate. > thanks >I guess you mean, "this signal is known to have frequency components higher than 1/2 the reduced sampling rate" Note the "1/2". Can you decimate, with "reduced" aliasing without computational intensive operations? The question, and thus the answer, is a matter of degree. A) You can decimate with no filtering and get aliasing. This may be acceptable in some circumstances. B) You can prefilter simply and get less aliasing than in (A) above. This may be acceptable in some circumstances. C) You can prefilter better and get even less aliasing than in (B). D) You can prefilter as best one could and still get aliasing that's noticeble in some circumstances. So, the logical answer to your question is Yes. You can decimate with reduced aliasing without computational intensive operations by using a less-than "intensive" filter. Fred

Reply by ●February 27, 20062006-02-27

andrea wrote:> is there a way to decimate a signal (reduce sampling rate) reducing > aliasing without implementing a computationally intensive low-pass > filter? > Consider that this signal is supposed to have frequency components > higher than the reduced sampling rate.If the frequency components higher than half the reduced sampling are completely characterized, you might be able to subtract or otherwise account for them after decimation without low-pass filtering. e.g. you could implement a computationally intensive high-pass filter instead, decimate the original and the high band component, then subtract. This might actually be useful if the there is something known about the high band component that makes it easier to filter. Or perhaps the low band signal of interest does not overlap with the aliased image of the high band component. In that case you might not even need to subtract the interference after decimating. IMHO. YMMV. -- rhn A.T nicholson d.0.t C-o-M