when should one perform fft based filtering?
in [DSP]
|
From: bharat pathak on 24 Jul 2010 10:45 what are the problems associated with FFT based filtering? and when should one decide to switchover from convolution based time domain filtering to FFT based filtering? bharat
From: Vladimir Vassilevsky on 24 Jul 2010 11:12 bharat pathak wrote: > what are the problems associated with FFT based filtering? FFT vs time domain filter == processing delay vs efficiency > and when should one decide to switchover from convolution > based time domain filtering to FFT based filtering? At the beginning of the project? VLV
From: robert bristow-johnson on 24 Jul 2010 11:39 On Jul 24, 10:47 am, "bharat pathak" <bharat(a)n_o_s_p_a_m.arithos.com> wrote: > what are the problems associated with FFT based filtering? > > and when should one decide to switchover from convolution > based time domain filtering to FFT based filtering? there was a thread about this a while back: " fast convolution crossover/break even point for DSPs, FPGAs " http://groups.google.com/group/comp.dsp/browse_frm/thread/985ad0ca9d10309/ you need to get a handle on the cost of doing the FFT and the overlap- save proceedure. from that you can get the coefficients A, B, C, and D of the cost function (see my post in that thread). and from that you can get the optimal N for doing it the FFT way, *then* you can compare the cost function to the time-domain cost which will be proportional to L. r b-j
From: Nasser M. Abbasi on 24 Jul 2010 15:10 On 7/24/2010 11:08 AM, Steve Pope wrote: > I so often here a question of the form, "How do I use an FFT > to do X", without the questioner having considered that > they can solve their problem in the time domain. > The teacher keep telling our class that the most common operation in digital world is convolution, and fft is a fast way to do this. So why would any one do convolution in time domain? isn't that much slower? so, for convolution, first choice is fft, right? I am just talking about convolution here. btw, the overlap save procedure (and there was another one like it?) was hard to understand for me, but luckly we did not get that in the exam, else I might have failed. --Nasser
From: Muzaffer Kal on 24 Jul 2010 15:10
On Sat, 24 Jul 2010 18:08:35 +0000 (UTC), spope33(a)speedymail.org (Steve Pope) wrote: >Vladimir Vassilevsky <nospam(a)nowhere.com> wrote: > >>bharat pathak wrote: > >>> and when should one decide to switchover from convolution >>> based time domain filtering to FFT based filtering? > >>At the beginning of the project? > >My answer was going to be sort-of the opposite: hardly ever. > >I so often here a question of the form, "How do I use an FFT >to do X", without the questioner having considered that >they can solve their problem in the time domain. That's true but some problems are just too costly to solve in time domain. As an example I think it would be quite challenging to do a 10GBT PHY (aminly the cancellers) in purely time domain with today's technology. With a 28nm process maybe but not just yet today. -- Muzaffer Kal DSPIA INC. ASIC/FPGA Design Services http://www.dspia.com |