Forums

Polyphase decimation filter, is it only a mathematical gimmick?

Started by Mr. Ken May 16, 2006
My decimation filter is 50 taps, and need to decimate by 10 times.
First approach is by normal FIR, but latch outputs every 10th cycle.
2nd approach is by dividing them into 10 polyphase filters each with 5 taps.

Is it correct to say, each decimated output I get from the two approaches
comprises of multiplication and summation of the same data and coefficients?



Yes.  The polyphase technique is a way of achieving the same output
with a reduced number of arithmetic operations.

You are on the right track, I think. However, I prefer to think of
decimation by M like this: you're computing a convolution of the input
sequence with a filter, then throwing away M-1 of every M
outputs. Well, why compute those outputs in the first place if you're
just going to throw them away?!?!!!

Splitting the one-of-M output's computation into "10 polyphase filters
of 5 taps each" sounds like a much more complex way of saying you just
run the filter on the input sequence *at that point*.

"Polyphase filtering' seems to be fancy words for essentially a simple
concept.

--Randy

Randy Yates wrote:

  ...

> "Polyphase filtering' seems to be fancy words for essentially a simple > concept.
Well, one name is nearly as good as another, as long as it's widely understood. Occasionally, it can lift one's spirits to discover that he knows how to talk in prose! :+) Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
"Jerry Avins" <jya@ieee.org> wrote in message 
news:W7ydnWbs1Pr-IvHZnZ2dneKdnZydnZ2d@rcn.net...
> Randy Yates wrote: > > ... > >> "Polyphase filtering' seems to be fancy words for essentially a simple >> concept. > > Well, one name is nearly as good as another, as long as it's widely > understood. Occasionally, it can lift one's spirits to discover that he > knows how to talk in prose! :+)
It's a pretty obvious optimization of the brute force zero-pad/up-sample, filter, down-sample procedure found in the texts. (At least to those with an optimization mindset.) A lot of people probably do it without knowing what it is called.
Jerry Avins wrote:
> Randy Yates wrote: > > ... > > > "Polyphase filtering' seems to be fancy words for essentially a simple > > concept. > > Well, one name is nearly as good as another, as long as it's widely > understood. Occasionally, it can lift one's spirits to discover that he > knows how to talk in prose! :+)
Yes... that's the difference between voodoo and science. A scientific concept, albeit with a fancy name, can often be described in rather simple terms. In fact, I used to learn new stuff by trying to concoct as simple explanations as possible for various "advanced" concepts. What effects that had on my spirits, is a completely different matter. Rune