Page 1 of 1

Is there an alternative to FFT prim?

PostPosted: Sat Nov 16, 2019 8:31 pm
by guyman
Is there an alternative method to, or different use of, the FFT prim that "captures" & displays more low frequency content? perhaps another octave or so? Even if accuracy decreases this far down, I want bins to display freq content below what our FFT currently spits out...

I've thought maybe it could be slid over, and combined with an envelope follower, on a filtered signal, and filing the bins with that.

Couldn't make it work.


Possible? workaround?

Re: Is there an alternative to FFT prim?

PostPosted: Sat Nov 16, 2019 8:48 pm
by MichaelBenjamin
the tradeoff/gain dilemma with FFT is, that the bigger the window is, the more resolution the frequency space gets, but the less resolution the amplitude over time for these frequency gets. so if you want more resolution in frequency, enlarge the window size.

Re: Is there an alternative to FFT prim?

PostPosted: Sat Nov 16, 2019 9:06 pm
by guyman
Thanks, but I'm not seeking more resolution in the lowend of the spectrum, as much as I am seeking to EXTEND the RANGE of the spectrum dsiplayed. I think the graph is only going down to 20hz, I want it to go down to about 10hz..
or am I misunderstanding what's taking place...

Re: Is there an alternative to FFT prim?

PostPosted: Sat Nov 16, 2019 9:17 pm
by MichaelBenjamin
when you got a samplerate of 48000 hz and want to detect freqs down to 10hz, then a 10hz sinewave would take 4800 samples for a full rotation, so the window size should be atleast this big, cant you set the window size to 8192?

Re: Is there an alternative to FFT prim?

PostPosted: Sat Nov 16, 2019 9:28 pm
by guyman
I do, but it isn't showing a proportionate change in distance in the display view... the octaves in a display window of the spectrum are equidistant. so if I conduct some operation that pulls a mono sig into fft, and run a sine wave, or filter bells in and observe where the peaks are, they are all equal is distance to one another, but if I have a wave form below say ~20-25hz, it goes off screen.. I haven't seen this in some other fft displays in vsts, I assume it's because the prim is calculating an array that only displays the content above ~20hz, but if I want to show say 8 - 9 - 15 hz, I'm getting no proportionate shift in the spectrum, you would think if i did something to do this, all the octaves would shift to the right by the distance equal to one octave in the display.

I understand what you are saying about grabbing a large enough window of samples, but I think this has less to do with the math not being done to calculate those freq when given the appropriate window, as so uch the prim has a cutoff of what populates the array...

I don't know how the prim does the math, but I was wondering after typing all of this if ruby were a feasible alternative.... could one run a mono2float @ 100ticks or more(custom) and then run that number into ruby and conduct the transform... with custom fft parameters, and spit out an array on the other side that gives us a spectrum of bins of say 0-30,000... (40,000)...

I am imaginative and bright, but ignorant so always feel free to correct my presumptions of possibilities.

Re: Is there an alternative to FFT prim?

PostPosted: Sat Nov 16, 2019 9:41 pm
by MichaelBenjamin
hm, without images it's hard to say what goes on there.

basically when you set window size to 8192, you get back an array of size 8192 which describes volume at a certain frequency, each calculated by 48000hz/(1...to...8192)hz, so practically from 24000hz down do some 5.8hz in 8192 steps

Re: Is there an alternative to FFT prim?

PostPosted: Sat Nov 16, 2019 9:45 pm
by guyman
I'll contribute more to this convo after I study more on the actual formula of the fast fourier transform..

without posting my whole toy, all I can say is the graph is proportional, but doesn't show 11 octaves, only 10...

but back to the ruby thing, can custom FFT be done in ruby? like custom ranges, etc...?

Re: Is there an alternative to FFT prim?

PostPosted: Sat Nov 16, 2019 9:49 pm
by MichaelBenjamin

Re: Is there an alternative to FFT prim?

PostPosted: Sat Nov 16, 2019 9:55 pm
by guyman
Lol thanks :lol: but that is a bit above where I am today. I'll have to really dig in..... :lol: