What is an FFT filter?

I am a guitarist/composer attempting to create my own demo recordings. It’s a classic case of old dog/new tricks. I know what most of the features are in the Effects drop list, but I have not been able to find much about FFT filters.

First, what does FFT stand for; and second, what does it filter?

I am doing mostly song demos: 7 - 12 tracks at the most. Would this function do me any good, and at what stage would I use it. Is it more effective on an individual track, or on a “global” application on a final mix?

Thanks for any help on this.

Tom Rasely
guitarist
(http://www.rasely.com)

Fast Fourier Transform.

Note that the FFT filter is a cousin to Equalization. Both have a “rubber band” line that you can drag around to get different effects.

http://kozco.com/tech/audacity/voiceEq.jpg

Pull the work window bigger and bigger to get more and more accuracy.

I’m not totally sure what the difference between those two tools is, but I know FFT made most modern audio tools possible.

Quick Math lesson. If you have a flute and an oboe playing the exact same note, how can you tell them apart? Overtones, right? Fourier is the science of predicting with mind-bending accuracy exactly why the oboe sounds like that and writing down, with a tiny pencil on a really big legal pad, exactly what and where each overtone is. Normally, this would take, like, a bazillion years to do on a musical composition, but FFT is a tool that makes certain assumptions and restrictions and greatly reduces the amount of work that the computer has to do. This can give you musical tone controls on steroids and caffeine.

Koz

Thank you. That was a good explanation, and a good demonstration with the link. So do I assume that this is something I would do individually to selected tracks? Or can this kind of EQ alteration be applied to a final mix?
Rasely

You can apply all these tools to any audio you select, the entire performance or even just the right or just the left of a small segment if you want. There are people applying effects to each click and pop on a photograph record capture.

Koz

FFT is explained in detail here,

this is another approach. more general,

You have in signal processing, the laplace, the fourier and the z transform.

the FFT is a special case of the Fourier transform.

in most cases it is only used as a marketing gimmick.

the more interesting question is what it does,
then I need to know more about the DSP digital signal processor used in the equipment,

Applications

The main applications of DSP are audio signal processing, audio compression,

and many more, tracking filter for preventing feedback, acoustic feedback,

another application is to sing and then in real time get your voice translated to sound like Elvis Presleys voice,
this is done in some films. An example of an analog signal processor is the moog synthesizer.

anything a moog can do your guitar plus a DSP also can do, you don’t really need FFT for that,
some engineers and programmers are a little bit old fashioned and still only use FFT but it is most likely possible to do a lot more,
with the right software. Which you can even program yourself, if the manufacturer gives you the specs.

What audacity does in a sound engineers studio could even be implemented and done on stage in real time,
with very low latency. You could use your guitar with a wlan antenna and transfer the guitar audio acoustic output to
tcp/ip and use a massive beowolf cluster of say playstations to calculate effects in near real time, say 200 ms,
which could have taken hours to weeks in a traditional studio.

With the right software.

This is all very helpful. Although I’ve been involved with recording music since 1980, I have only begun making my own recordings on my computer this year. There are times when I feel that the learning curve is a little overwhelming. So thank you very much for your assistance.
Rasely

I have to take serious issue with this statement:

the FFT is a special case of the Fourier transform.
in most cases it is only used as a marketing gimmick.

What? Marketing gimmick?

FFT isn’t an effect, it’s just a description of how the effect is being applied. The FFT function makes use of some of the properties of a digital signal in order to compute a very difficult equation is a much shorter period of time. Just about any digital audio filter is going to make use of a reverse-FFT function to apply the filter curve to the output. I would hardly call it a marketing gimmick.

The only difference that I can see between Audacity’s FFT Filter and the Equalization effect is that the FFT filter is displayed linearly with respect to frequency while the EQ is displayed Logarithmically. The upshot of this is that the EQ is much more useful for musical signals, while the FFT is sometimes more useful for non-musical audio (like test signals or audio forensics).

Audacity’s FFT Filter and the Equalization effect

I will take a close look on Audacitys FFT Filter.
I’m still learning how to use all features.

I did FFT with Hypersignal
http://www.signalogic.com/

meanwhile I have discovered some nice pages I have previously
only seen in music magazines.

http://freesoftware.ircam.fr/

http://freesoftware.ircam.fr/rubrique.php3?id_rubrique=14

http://www.ircam.fr/

http://mediatheque.ircam.fr/articles/textes/Boulez88c/

I was there many many years ago and I have seen an introduction
to what at that time was possible.


some nice introductory topics
http://www.dsptutor.freeuk.com/


the standard DSP textbooks

http://www.dspguru.com/info/books/favor2.htm

http://www.dspguru.com/info/books/ref.htm

My personal preferred white paper

A Mathematical Theory of Communication
http://plan9.bell-labs.com/cm/ms/what/shannonday/shannon1948.pdf



It is much better we agree about what DSPs can do,
most CPUs has the FLOPS required to do what a special purpose DSP
could at most ideally do few years ago.

Basically latency is for most practical applications like gaming or
video conferencing always a limiting factor.

By splitting the signal in small short time windows,
and applying some signal processing on each of these frames you get what you want.

The duration of this time windows will
always result in some kind of latency,

on a normal CPU this latency will always increase by threads and CPU load.

so striving for zero latency and claiming to be some application of FFT
is to me a contradiction.
After one time window sample has elapsed the processing can at most start.

constant latency and FFT is possible.
Zero latency and FFT plus inverse FFT is impossible.

What is neglectible latency is a personal matter.
Young persons can hear better than their parents.

only a DSP with application software and operating system
designed for real time can offer that.


If you google for fft filter marketing gimmick
you will find this


Jitter spectral extraction for multi-gigahertz signal - Design …
With marketing gimmicks pushing for higher processor … FFT. to extract the peaks and frequencies of the. sinusoidal jitters. …
ieeexplore.ieee.org/iel5/9284/29505/01337584.pdf - Similar pages

I don’t have access to the paper.

Esp in consumer oriented music for multimedia
they are often free to write anything.
inclusive FFT and IFFT.

Jitter congestion latency and round trip time is hot topics in real time gaming and
next generation network,
which could be a concert conducted over the internet.

Here is a nice podcast about post production.

http://www.oreillynet.com/pub/a/oreilly/digitalmedia/2008/03/28/dmi-23-inside-intel-inside.html


some nice hardware.
http://www.pcisource.com/pciprod/pciprocessors/dsp.shtml

http://focus.ti.com/dsp/docs/dsphome.tsp?sectionId=46

http://focus.ti.com/dsp/docs/dspfindtoolswbytooltype.tsp?sectionId=3&tabId=2095&toolTypeId=5&familyId=44

discusses fft pro and contra.
http://homerecording.com/bbs/showthread.php?t=245320

Offers an alternative to FFT which I have never heard about.

Real-time FFT filtering
http://www.stanford.edu/~sols/course_projects/ee595k_report.pdf

As an example red book specified 44.1 KHz 16 bit stereo
K at that time was 1000

In marketing it can be 2^10 = 1024

mp3 is a file format or streaming format which reduces bandwidth.

and saves file space. the same filter applied on all three signals
I feel will not give the same effect. analog (through ADC + DSP + DAC ), PCM, or mp3

The best evidence for that is aliasing.
for that we have anti aliasing filter.

Audio Critic’s Ten Biggest Lies
http://forums.audioholics.com/forums/showthread.php?t=12599

Shows that it exists smoke without fire.

Ideally an FFT filter is a piece of software which runs well on a DSP and uses a RTOS.
if it is not a DSP and not an RTOS the same software behaves different and does
not do what normally an FFT filter was designed to do.

This is independent of audacity, the OP had a Q about FFT in general,
FFT does not need to filter anything.

as in mp3 it does not filter anything, it simply reduces the bandwidth.
and shortens the download.

best regards

morten44,

I have no idea what you’re trying to say. Whenever I see the words “zero-latency” on an audio product, they’re usually talking about the monitoring signal. In that case, it’s an analog circuit with no latency (unless you consider a few nano-seconds to be latency).

It’s perfectly reasonable to have a piece of equipment that does zero-latency monitoring and also does DSP effects. They’re simply talking about two different outputs. Personally I wouldn’t call that a marketing gimmick, especially if it’s the simple truth.