Reply by Eric Jacobsen August 13, 20122012-08-13
On Mon, 13 Aug 2012 11:31:21 -0500, "colin22" <61148@dsprelated>
wrote:

>I know what a farrow filter is and I'm familiar with the efficient >polyphase implementation, but the structures I've seen in articles about >Farrow filters and the structures in the articles about "polyphase" Farrow >filters look pretty much the same. I'm confused why the author would >specify that a Farrow filter was a polyphase Farrow filter if all Farrow >filters are polyphase. Could I have an example of a Farrow filter that >isn't polyphase? > >Maybe the source of my confusion is similar to the end of this thread >http://www.dsprelated.com/showmessage/173840/1.php > >I posted: > >"Powerpoint about polyphase filters >http://www.ee.ic.ac.uk/hp/staff/dmb/courses/DSPDF/01200_Polyphase.pdf" > >and someone replied: > >"note that the above course notes describe a filter that has equal input- >and output rate. This is quite a bit more complicated than a plain >polyphase interpolator.
I think polyphase filters may or may not do rate changes, since they can be used for interpolation or decimation or just timing adjustment without rate change.
>The term 'phase' refers to a time-delayed replica of the signal. A >'poly'phase filter simply means that several 'phases' are computed in >parallel, > >What I'm calling a 'polyphase' interpolator is referred to as 'commutator' >in the slides ('upsampler implementation')." > >and by that definition, aren't all Farrow filters polyphase filters?
I think one could call them that, but there may be a distinction as well. Generally a Farrow filter computes coefficients (usually on the fly) from an approximated curve fit of the desired impulse response sections. A polyphase filter generally has all the possible phases of the desired (not approximated) impulse response stored and accesses them as needed. It is possible to pre-compute all the expected possible phases of a Farrow filter's approximated coefficients and store them like in a polyphase filter, but if you do that I don't know why you wouldn't just use the exact rather than approximated coefficients. One twist is that, depending on the implementation, a Farrow filter may be able to approximate much finer resolution in the "phases" by interpolating the approximate curve-fits on the fly than a polyphase could practically do by storing a large quantity of coefficients. There is a tradeoff between complexity and accuracy in the impulse response. So, a Farrow filter is a "polyphase" in the sense that it can filter different phases of an impulse response, but it may not store the coefficient "phases" the way that most polyphase filters do. Don't get hung up in the definitions or semantics. There's usually a lot of overlap in these sorts of definitions. The main thing is to understand the filters and their distinctions. Eric Jacobsen Anchor Hill Communications www.anchorhill.com
Reply by colin22 August 13, 20122012-08-13
I know what a farrow filter is and I'm familiar with the efficient
polyphase implementation, but the structures I've seen in articles about
Farrow filters and the structures in the articles about "polyphase" Farrow
filters look pretty much the same. I'm confused why the author would
specify that a Farrow filter was a polyphase Farrow filter if all Farrow
filters are polyphase. Could I have an example of a Farrow filter that
isn't polyphase?

Maybe the source of my confusion is similar to the end of this thread
http://www.dsprelated.com/showmessage/173840/1.php

I posted:

"Powerpoint about polyphase filters
http://www.ee.ic.ac.uk/hp/staff/dmb/courses/DSPDF/01200_Polyphase.pdf"

and someone replied:

"note that the above course notes describe a filter that has equal input-
and output rate. This is quite a bit more complicated than a plain
polyphase interpolator. 

The term 'phase' refers to a time-delayed replica of the signal. A
'poly'phase filter simply means that several 'phases' are computed in
parallel, 

What I'm calling a 'polyphase' interpolator is referred to as 'commutator'
in the slides ('upsampler implementation')."

and by that definition, aren't all Farrow filters polyphase filters?