2

I know that to downsample by a factor of 2.5x one can upsample a signal by a factor of 2 and downsample by a factor of 5. I am wondering what is the best and more efficient way of downsampling by a factor of 2.5x?

TJ1
  • 177
  • 5
  • You may have to benchmark your actual hardware to accurately determine efficiency. If you require a FIR anti-alias filter for your desired result quality, you only need a 2 phase polyphase kernel with which you can avoid upsampling. – hotpaw2 Aug 05 '13 at 19:37
  • @hotpaw2 thanks for the response. The 2 phase polyphase that you referring to: how does it do the downsampling by a factor of 2.5? The downsample by a factor of 2 using two phase polyphase is straightforward, how about a factor of 2.5? – TJ1 Aug 05 '13 at 20:33
  • @TJ1: Without providing enough detail to make this an answer, think of a polyphase structure as a way of efficiently changing sample rate. As you noted, one way to solve your problem would be to interpolate by 2 and then downsample by 5. However, you're wasting work by first increasing your sample rate. Instead, a polyphase structure is cleverly structured so you just don't waste time calculating outputs that you won't use. – Jason R Aug 06 '13 at 13:04
  • @JasonR thanks for the response. What details one would need to answer this question? Can you show how a downsample by 2.5 can be done with polyphase structure? A drawing or reference to an article would be great. – TJ1 Aug 06 '13 at 16:50

0 Answers0