cwright's picture

While poking at kineme3d, I took a stroll down Accelerate.framework lane to see if I could use it to speed up some deformers. Unfortunately, it's not really handy for Kineme3D's data layout, but on the plus side, I got to experiment with it some, and figure out how to use the FFT functions as well. (FFT isn't at all useful in kineme3d, mind you.)

So, with the next build of AudioTools, we'll have frequency support (512-channel frequency support, no less :) [for the math nerds keeping score at home: yes, it's impossible to get 512 frequencies from a 512-sample set (Nyquist frequency and all that) -- however, FFT provides us with Complex Numbers (a+bi), and we export both real and imaginary parts side-by-side, for 512 "frequencies"]

Here's a sneak peak:

(until vimeo gets it converted etc)

FreqTest.png26.8 KB

Michael Neely's picture

Thanks Christopher. This will be a great addition to the Audio Tools. M

gtoledo3's picture
That's cool... doesn't

That's cool... doesn't really look quite as I would expect from the sound, given looking at various "studio" devices... that's probably because of the way that many audio gear makers would tends to give less weight to stuff above say 12k, and below 90hz, and probably tweaked in the 1k region to account for the fact that the ear is more sensitive- kind of "weighted" to account for the ear... I imagine this is more accurate, and hence, not what I'm used to seeing!

Isn't this a keyboard patch from the Roland XP series?

I'm really excited about this, because I've wanted to be to the values of various frequencies to do render based stuff for awhile now! (thought it kind of can be done with VH, sorta..._

smokris's picture
Linear vs. Logarithmic

gtoledo3 wrote:
doesn't really look quite as I would expect from the sound, given looking at various "studio" devices...

Right. Most "studio" audio analyzers display the frequency axis logarithmically, whereas the demo above displays the frequency axis linearly --- hence way more of the graph is focused on the (relatively uninteresting) higher frequencies.

I'm working on a modification to this to allow switching between raw linear data and logarithmic averaging.

gtoledo3's picture
Ah, exactly! There were

Ah, exactly! There were cobwebs in the noggin' on that one. The modification for that option would be rather cool.

The idea in general really opens up a whole new niche of applications for QC, and it's actually heartening that there IS a linear representation, which is interesting for seeing that detail up in the highs -which could allow one to see irritating high end artifacts of a piece of an audio chain... or aid in allowing one to place "dog whistle" tones in audio to effect imagery in QC in unique ways.

I am all of a sudden having this weird desire to make a QC "strobocon".